1 | The following changes since commit 411ad78115ebeb3411cf4b7622784b93dfabe259: | 1 | The following changes since commit db754f8ccaf2f073c9aed46a4389e9c0c2080399: |
---|---|---|---|
2 | 2 | ||
3 | Merge remote-tracking branch 'remotes/stefanberger/tags/pull-tpm-2017-12-15-1' into staging (2017-12-17 15:27:41 +0000) | 3 | Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210202' into staging (2021-02-03 19:35:57 +0000) |
4 | 4 | ||
5 | are available in the Git repository at: | 5 | are available in the Git repository at: |
6 | 6 | ||
7 | git://github.com/stefanha/qemu.git tags/block-pull-request | 7 | https://gitlab.com/stefanha/qemu.git tags/block-pull-request |
8 | 8 | ||
9 | for you to fetch changes up to 585426c518958aa768564596091474be786aae51: | 9 | for you to fetch changes up to abe42229db7b87caa11b3c02835ebf9d384e0bd4: |
10 | 10 | ||
11 | qemu-iotests: add 203 savevm with IOThreads test (2017-12-18 13:12:53 +0000) | 11 | docs: fix Parallels Image "dirty bitmap" section (2021-02-04 15:17:10 +0000) |
12 | |||
13 | ---------------------------------------------------------------- | ||
14 | Pull request | ||
15 | |||
16 | v2: | ||
17 | * Rebase to resolve memory_region_init_ram_from_file() conflict due to the new | ||
18 | offset argument that was added in qemu.git/master in the meantime [Peter] | ||
12 | 19 | ||
13 | ---------------------------------------------------------------- | 20 | ---------------------------------------------------------------- |
14 | 21 | ||
15 | ---------------------------------------------------------------- | 22 | Denis V. Lunev (1): |
23 | docs: fix Parallels Image "dirty bitmap" section | ||
16 | 24 | ||
17 | Mao Zhongyi (4): | 25 | Elena Ufimtseva (8): |
18 | hw/block/nvme: Convert to realize | 26 | multi-process: add configure and usage information |
19 | hw/block: Fix the return type | 27 | io: add qio_channel_writev_full_all helper |
20 | hw/block: Use errp directly rather than local_err | 28 | io: add qio_channel_readv_full_all_eof & qio_channel_readv_full_all |
21 | dev-storage: Fix the unusual function name | 29 | helpers |
30 | multi-process: define MPQemuMsg format and transmission functions | ||
31 | multi-process: introduce proxy object | ||
32 | multi-process: add proxy communication functions | ||
33 | multi-process: Forward PCI config space acceses to the remote process | ||
34 | multi-process: perform device reset in the remote process | ||
22 | 35 | ||
23 | Mark Kanda (2): | 36 | Jagannathan Raman (11): |
24 | virtio-blk: make queue size configurable | 37 | memory: alloc RAM from file at offset |
25 | virtio-blk: reject configs with logical block size > physical block | 38 | multi-process: Add config option for multi-process QEMU |
26 | size | 39 | multi-process: setup PCI host bridge for remote device |
40 | multi-process: setup a machine object for remote device process | ||
41 | multi-process: Initialize message handler in remote device | ||
42 | multi-process: Associate fd of a PCIDevice with its object | ||
43 | multi-process: setup memory manager for remote device | ||
44 | multi-process: PCI BAR read/write handling for proxy & remote | ||
45 | endpoints | ||
46 | multi-process: Synchronize remote memory | ||
47 | multi-process: create IOHUB object to handle irq | ||
48 | multi-process: Retrieve PCI info from remote process | ||
27 | 49 | ||
28 | Paolo Bonzini (1): | 50 | John G Johnson (1): |
29 | block: avoid recursive AioContext acquire in bdrv_inactivate_all() | 51 | multi-process: add the concept description to |
52 | docs/devel/qemu-multiprocess | ||
30 | 53 | ||
31 | Stefan Hajnoczi (16): | 54 | Stefan Hajnoczi (6): |
32 | coroutine: simplify co_aio_sleep_ns() prototype | 55 | .github: point Repo Lockdown bot to GitLab repo |
33 | qdev: drop unused #include "sysemu/iothread.h" | 56 | gitmodules: use GitLab repos instead of qemu.org |
34 | blockdev: hold AioContext for bdrv_unref() in | 57 | gitlab-ci: remove redundant GitLab repo URL command |
35 | external_snapshot_clean() | 58 | docs: update README to use GitLab repo URLs |
36 | block: don't keep AioContext acquired after | 59 | pc-bios: update mirror URLs to GitLab |
37 | external_snapshot_prepare() | 60 | get_maintainer: update repo URL to GitLab |
38 | block: don't keep AioContext acquired after drive_backup_prepare() | ||
39 | block: don't keep AioContext acquired after blockdev_backup_prepare() | ||
40 | block: don't keep AioContext acquired after | ||
41 | internal_snapshot_prepare() | ||
42 | block: drop unused BlockDirtyBitmapState->aio_context field | ||
43 | iothread: add iothread_by_id() API | ||
44 | blockdev: add x-blockdev-set-iothread testing command | ||
45 | qemu-iotests: add 202 external snapshots IOThread test | ||
46 | docs: mark nested AioContext locking as a legacy API | ||
47 | blockdev: add x-blockdev-set-iothread force boolean | ||
48 | iotests: add VM.add_object() | ||
49 | iothread: fix iothread_stop() race condition | ||
50 | qemu-iotests: add 203 savevm with IOThreads test | ||
51 | 61 | ||
52 | docs/devel/multiple-iothreads.txt | 7 +- | 62 | MAINTAINERS | 24 + |
53 | qapi/block-core.json | 40 ++++++ | 63 | README.rst | 4 +- |
54 | hw/block/dataplane/virtio-blk.h | 2 +- | 64 | docs/devel/index.rst | 1 + |
55 | include/hw/block/block.h | 4 +- | 65 | docs/devel/multi-process.rst | 966 ++++++++++++++++++++++ |
56 | include/hw/virtio/virtio-blk.h | 1 + | 66 | docs/system/index.rst | 1 + |
57 | include/qemu/coroutine.h | 6 +- | 67 | docs/system/multi-process.rst | 64 ++ |
58 | include/sysemu/iothread.h | 4 +- | 68 | docs/interop/parallels.txt | 2 +- |
59 | block.c | 14 ++- | 69 | configure | 10 + |
60 | block/null.c | 3 +- | 70 | meson.build | 5 +- |
61 | block/sheepdog.c | 3 +- | 71 | hw/remote/trace.h | 1 + |
62 | blockdev.c | 259 +++++++++++++++++++++++++++----------- | 72 | include/exec/memory.h | 2 + |
63 | hw/block/block.c | 15 ++- | 73 | include/exec/ram_addr.h | 4 +- |
64 | hw/block/dataplane/virtio-blk.c | 12 +- | 74 | include/hw/pci-host/remote.h | 30 + |
65 | hw/block/fdc.c | 17 +-- | 75 | include/hw/pci/pci_ids.h | 3 + |
66 | hw/block/nvme.c | 23 ++-- | 76 | include/hw/remote/iohub.h | 42 + |
67 | hw/block/virtio-blk.c | 35 ++++-- | 77 | include/hw/remote/machine.h | 38 + |
68 | hw/core/qdev-properties-system.c | 1 - | 78 | include/hw/remote/memory.h | 19 + |
69 | hw/ide/qdev.c | 12 +- | 79 | include/hw/remote/mpqemu-link.h | 99 +++ |
70 | hw/scsi/scsi-disk.c | 13 +- | 80 | include/hw/remote/proxy-memory-listener.h | 28 + |
71 | hw/usb/dev-storage.c | 29 ++--- | 81 | include/hw/remote/proxy.h | 48 ++ |
72 | iothread.c | 27 +++- | 82 | include/io/channel.h | 78 ++ |
73 | util/qemu-coroutine-sleep.c | 4 +- | 83 | include/qemu/mmap-alloc.h | 4 +- |
74 | tests/qemu-iotests/202 | 95 ++++++++++++++ | 84 | include/sysemu/iothread.h | 6 + |
75 | tests/qemu-iotests/202.out | 11 ++ | 85 | backends/hostmem-memfd.c | 2 +- |
76 | tests/qemu-iotests/203 | 59 +++++++++ | 86 | hw/misc/ivshmem.c | 3 +- |
77 | tests/qemu-iotests/203.out | 6 + | 87 | hw/pci-host/remote.c | 75 ++ |
78 | tests/qemu-iotests/group | 2 + | 88 | hw/remote/iohub.c | 119 +++ |
79 | tests/qemu-iotests/iotests.py | 5 + | 89 | hw/remote/machine.c | 80 ++ |
80 | 28 files changed, 532 insertions(+), 177 deletions(-) | 90 | hw/remote/memory.c | 65 ++ |
81 | create mode 100755 tests/qemu-iotests/202 | 91 | hw/remote/message.c | 230 ++++++ |
82 | create mode 100644 tests/qemu-iotests/202.out | 92 | hw/remote/mpqemu-link.c | 267 ++++++ |
83 | create mode 100755 tests/qemu-iotests/203 | 93 | hw/remote/proxy-memory-listener.c | 227 +++++ |
84 | create mode 100644 tests/qemu-iotests/203.out | 94 | hw/remote/proxy.c | 379 +++++++++ |
95 | hw/remote/remote-obj.c | 203 +++++ | ||
96 | io/channel.c | 116 ++- | ||
97 | iothread.c | 6 + | ||
98 | softmmu/memory.c | 3 +- | ||
99 | softmmu/physmem.c | 12 +- | ||
100 | util/mmap-alloc.c | 8 +- | ||
101 | util/oslib-posix.c | 2 +- | ||
102 | .github/lockdown.yml | 8 +- | ||
103 | .gitlab-ci.yml | 1 - | ||
104 | .gitmodules | 44 +- | ||
105 | Kconfig.host | 4 + | ||
106 | hw/Kconfig | 1 + | ||
107 | hw/meson.build | 1 + | ||
108 | hw/pci-host/Kconfig | 3 + | ||
109 | hw/pci-host/meson.build | 1 + | ||
110 | hw/remote/Kconfig | 4 + | ||
111 | hw/remote/meson.build | 13 + | ||
112 | hw/remote/trace-events | 4 + | ||
113 | pc-bios/README | 4 +- | ||
114 | scripts/get_maintainer.pl | 2 +- | ||
115 | 53 files changed, 3296 insertions(+), 70 deletions(-) | ||
116 | create mode 100644 docs/devel/multi-process.rst | ||
117 | create mode 100644 docs/system/multi-process.rst | ||
118 | create mode 100644 hw/remote/trace.h | ||
119 | create mode 100644 include/hw/pci-host/remote.h | ||
120 | create mode 100644 include/hw/remote/iohub.h | ||
121 | create mode 100644 include/hw/remote/machine.h | ||
122 | create mode 100644 include/hw/remote/memory.h | ||
123 | create mode 100644 include/hw/remote/mpqemu-link.h | ||
124 | create mode 100644 include/hw/remote/proxy-memory-listener.h | ||
125 | create mode 100644 include/hw/remote/proxy.h | ||
126 | create mode 100644 hw/pci-host/remote.c | ||
127 | create mode 100644 hw/remote/iohub.c | ||
128 | create mode 100644 hw/remote/machine.c | ||
129 | create mode 100644 hw/remote/memory.c | ||
130 | create mode 100644 hw/remote/message.c | ||
131 | create mode 100644 hw/remote/mpqemu-link.c | ||
132 | create mode 100644 hw/remote/proxy-memory-listener.c | ||
133 | create mode 100644 hw/remote/proxy.c | ||
134 | create mode 100644 hw/remote/remote-obj.c | ||
135 | create mode 100644 hw/remote/Kconfig | ||
136 | create mode 100644 hw/remote/meson.build | ||
137 | create mode 100644 hw/remote/trace-events | ||
85 | 138 | ||
86 | -- | 139 | -- |
87 | 2.14.3 | 140 | 2.29.2 |
88 | 141 | ||
89 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | Use the GitLab repo URL as the main repo location in order to reduce | ||
2 | load on qemu.org. | ||
1 | 3 | ||
4 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
5 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> | ||
6 | Reviewed-by: Thomas Huth <thuth@redhat.com> | ||
7 | Message-id: 20210111115017.156802-2-stefanha@redhat.com | ||
8 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
9 | --- | ||
10 | .github/lockdown.yml | 8 ++++---- | ||
11 | 1 file changed, 4 insertions(+), 4 deletions(-) | ||
12 | |||
13 | diff --git a/.github/lockdown.yml b/.github/lockdown.yml | ||
14 | index XXXXXXX..XXXXXXX 100644 | ||
15 | --- a/.github/lockdown.yml | ||
16 | +++ b/.github/lockdown.yml | ||
17 | @@ -XXX,XX +XXX,XX @@ issues: | ||
18 | comment: | | ||
19 | Thank you for your interest in the QEMU project. | ||
20 | |||
21 | - This repository is a read-only mirror of the project's master | ||
22 | - repostories hosted on https://git.qemu.org/git/qemu.git. | ||
23 | + This repository is a read-only mirror of the project's repostories hosted | ||
24 | + at https://gitlab.com/qemu-project/qemu.git. | ||
25 | The project does not process issues filed on GitHub. | ||
26 | |||
27 | The project issues are tracked on Launchpad: | ||
28 | @@ -XXX,XX +XXX,XX @@ pulls: | ||
29 | comment: | | ||
30 | Thank you for your interest in the QEMU project. | ||
31 | |||
32 | - This repository is a read-only mirror of the project's master | ||
33 | - repostories hosted on https://git.qemu.org/git/qemu.git. | ||
34 | + This repository is a read-only mirror of the project's repostories hosted | ||
35 | + on https://gitlab.com/qemu-project/qemu.git. | ||
36 | The project does not process merge requests filed on GitHub. | ||
37 | |||
38 | QEMU welcomes contributions of code (either fixing bugs or adding new | ||
39 | -- | ||
40 | 2.29.2 | ||
41 | diff view generated by jsdifflib |
1 | The VM.add_object() method can be used to add IOThreads or memory | 1 | qemu.org is running out of bandwidth and the QEMU project is moving |
---|---|---|---|
2 | backend objects. | 2 | towards a gating CI on GitLab. Use the GitLab repos instead of qemu.org |
3 | (they will become mirrors). | ||
3 | 4 | ||
4 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
5 | Reviewed-by: Eric Blake <eblake@redhat.com> | 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> |
6 | Message-id: 20171207201320.19284-5-stefanha@redhat.com | 7 | Reviewed-by: Thomas Huth <thuth@redhat.com> |
8 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
9 | Message-id: 20210111115017.156802-3-stefanha@redhat.com | ||
7 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
8 | --- | 11 | --- |
9 | tests/qemu-iotests/iotests.py | 5 +++++ | 12 | .gitmodules | 44 ++++++++++++++++++++++---------------------- |
10 | 1 file changed, 5 insertions(+) | 13 | 1 file changed, 22 insertions(+), 22 deletions(-) |
11 | 14 | ||
12 | diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py | 15 | diff --git a/.gitmodules b/.gitmodules |
13 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
14 | --- a/tests/qemu-iotests/iotests.py | 17 | --- a/.gitmodules |
15 | +++ b/tests/qemu-iotests/iotests.py | 18 | +++ b/.gitmodules |
16 | @@ -XXX,XX +XXX,XX @@ class VM(qtest.QEMUQtestMachine): | 19 | @@ -XXX,XX +XXX,XX @@ |
17 | socket_scm_helper=socket_scm_helper) | 20 | [submodule "roms/seabios"] |
18 | self._num_drives = 0 | 21 | path = roms/seabios |
19 | 22 | - url = https://git.qemu.org/git/seabios.git/ | |
20 | + def add_object(self, opts): | 23 | + url = https://gitlab.com/qemu-project/seabios.git/ |
21 | + self._args.append('-object') | 24 | [submodule "roms/SLOF"] |
22 | + self._args.append(opts) | 25 | path = roms/SLOF |
23 | + return self | 26 | - url = https://git.qemu.org/git/SLOF.git |
24 | + | 27 | + url = https://gitlab.com/qemu-project/SLOF.git |
25 | def add_device(self, opts): | 28 | [submodule "roms/ipxe"] |
26 | self._args.append('-device') | 29 | path = roms/ipxe |
27 | self._args.append(opts) | 30 | - url = https://git.qemu.org/git/ipxe.git |
31 | + url = https://gitlab.com/qemu-project/ipxe.git | ||
32 | [submodule "roms/openbios"] | ||
33 | path = roms/openbios | ||
34 | - url = https://git.qemu.org/git/openbios.git | ||
35 | + url = https://gitlab.com/qemu-project/openbios.git | ||
36 | [submodule "roms/qemu-palcode"] | ||
37 | path = roms/qemu-palcode | ||
38 | - url = https://git.qemu.org/git/qemu-palcode.git | ||
39 | + url = https://gitlab.com/qemu-project/qemu-palcode.git | ||
40 | [submodule "roms/sgabios"] | ||
41 | path = roms/sgabios | ||
42 | - url = https://git.qemu.org/git/sgabios.git | ||
43 | + url = https://gitlab.com/qemu-project/sgabios.git | ||
44 | [submodule "dtc"] | ||
45 | path = dtc | ||
46 | - url = https://git.qemu.org/git/dtc.git | ||
47 | + url = https://gitlab.com/qemu-project/dtc.git | ||
48 | [submodule "roms/u-boot"] | ||
49 | path = roms/u-boot | ||
50 | - url = https://git.qemu.org/git/u-boot.git | ||
51 | + url = https://gitlab.com/qemu-project/u-boot.git | ||
52 | [submodule "roms/skiboot"] | ||
53 | path = roms/skiboot | ||
54 | - url = https://git.qemu.org/git/skiboot.git | ||
55 | + url = https://gitlab.com/qemu-project/skiboot.git | ||
56 | [submodule "roms/QemuMacDrivers"] | ||
57 | path = roms/QemuMacDrivers | ||
58 | - url = https://git.qemu.org/git/QemuMacDrivers.git | ||
59 | + url = https://gitlab.com/qemu-project/QemuMacDrivers.git | ||
60 | [submodule "ui/keycodemapdb"] | ||
61 | path = ui/keycodemapdb | ||
62 | - url = https://git.qemu.org/git/keycodemapdb.git | ||
63 | + url = https://gitlab.com/qemu-project/keycodemapdb.git | ||
64 | [submodule "capstone"] | ||
65 | path = capstone | ||
66 | - url = https://git.qemu.org/git/capstone.git | ||
67 | + url = https://gitlab.com/qemu-project/capstone.git | ||
68 | [submodule "roms/seabios-hppa"] | ||
69 | path = roms/seabios-hppa | ||
70 | - url = https://git.qemu.org/git/seabios-hppa.git | ||
71 | + url = https://gitlab.com/qemu-project/seabios-hppa.git | ||
72 | [submodule "roms/u-boot-sam460ex"] | ||
73 | path = roms/u-boot-sam460ex | ||
74 | - url = https://git.qemu.org/git/u-boot-sam460ex.git | ||
75 | + url = https://gitlab.com/qemu-project/u-boot-sam460ex.git | ||
76 | [submodule "tests/fp/berkeley-testfloat-3"] | ||
77 | path = tests/fp/berkeley-testfloat-3 | ||
78 | - url = https://git.qemu.org/git/berkeley-testfloat-3.git | ||
79 | + url = https://gitlab.com/qemu-project/berkeley-testfloat-3.git | ||
80 | [submodule "tests/fp/berkeley-softfloat-3"] | ||
81 | path = tests/fp/berkeley-softfloat-3 | ||
82 | - url = https://git.qemu.org/git/berkeley-softfloat-3.git | ||
83 | + url = https://gitlab.com/qemu-project/berkeley-softfloat-3.git | ||
84 | [submodule "roms/edk2"] | ||
85 | path = roms/edk2 | ||
86 | - url = https://git.qemu.org/git/edk2.git | ||
87 | + url = https://gitlab.com/qemu-project/edk2.git | ||
88 | [submodule "slirp"] | ||
89 | path = slirp | ||
90 | - url = https://git.qemu.org/git/libslirp.git | ||
91 | + url = https://gitlab.com/qemu-project/libslirp.git | ||
92 | [submodule "roms/opensbi"] | ||
93 | path = roms/opensbi | ||
94 | - url = https://git.qemu.org/git/opensbi.git | ||
95 | + url = https://gitlab.com/qemu-project/opensbi.git | ||
96 | [submodule "roms/qboot"] | ||
97 | path = roms/qboot | ||
98 | - url = https://git.qemu.org/git/qboot.git | ||
99 | + url = https://gitlab.com/qemu-project/qboot.git | ||
100 | [submodule "meson"] | ||
101 | path = meson | ||
102 | - url = https://git.qemu.org/git/meson.git | ||
103 | + url = https://gitlab.com/qemu-project/meson.git | ||
104 | [submodule "roms/vbootrom"] | ||
105 | path = roms/vbootrom | ||
106 | - url = https://git.qemu.org/git/vbootrom.git | ||
107 | + url = https://gitlab.com/qemu-project/vbootrom.git | ||
28 | -- | 108 | -- |
29 | 2.14.3 | 109 | 2.29.2 |
30 | 110 | ||
31 | diff view generated by jsdifflib |
1 | Commit 1351d1ec89eabebc9fdff20451a62c413d7accc1 ("qdev: drop iothread | 1 | It is no longer necessary to point .gitmodules at GitLab repos when |
---|---|---|---|
2 | property type") forgot to remove this include. | 2 | running in GitLab CI since they are now used all the time. |
3 | 3 | ||
4 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 4 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
5 | Message-id: 20171205133954.31006-1-stefanha@redhat.com | 5 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> |
6 | Reviewed-by: Thomas Huth <thuth@redhat.com> | ||
7 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
8 | Message-id: 20210111115017.156802-4-stefanha@redhat.com | ||
6 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 9 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
7 | --- | 10 | --- |
8 | hw/core/qdev-properties-system.c | 1 - | 11 | .gitlab-ci.yml | 1 - |
9 | 1 file changed, 1 deletion(-) | 12 | 1 file changed, 1 deletion(-) |
10 | 13 | ||
11 | diff --git a/hw/core/qdev-properties-system.c b/hw/core/qdev-properties-system.c | 14 | diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml |
12 | index XXXXXXX..XXXXXXX 100644 | 15 | index XXXXXXX..XXXXXXX 100644 |
13 | --- a/hw/core/qdev-properties-system.c | 16 | --- a/.gitlab-ci.yml |
14 | +++ b/hw/core/qdev-properties-system.c | 17 | +++ b/.gitlab-ci.yml |
15 | @@ -XXX,XX +XXX,XX @@ | 18 | @@ -XXX,XX +XXX,XX @@ include: |
16 | #include "qapi/visitor.h" | 19 | image: $CI_REGISTRY_IMAGE/qemu/$IMAGE:latest |
17 | #include "chardev/char-fe.h" | 20 | before_script: |
18 | #include "sysemu/tpm_backend.h" | 21 | - JOBS=$(expr $(nproc) + 1) |
19 | -#include "sysemu/iothread.h" | 22 | - - sed -i s,git.qemu.org/git,gitlab.com/qemu-project, .gitmodules |
20 | 23 | script: | |
21 | static void get_pointer(Object *obj, Visitor *v, Property *prop, | 24 | - mkdir build |
22 | char *(*print)(void *ptr), | 25 | - cd build |
23 | -- | 26 | -- |
24 | 2.14.3 | 27 | 2.29.2 |
25 | 28 | ||
26 | diff view generated by jsdifflib |
1 | When a node is already associated with a BlockBackend the | 1 | qemu.org is running out of bandwidth and the QEMU project is moving |
---|---|---|---|
2 | x-blockdev-set-iothread command refuses to set the IOThread. This is to | 2 | towards a gating CI on GitLab. Use the GitLab repos instead of qemu.org |
3 | prevent accidentally changing the IOThread when the nodes are in use. | 3 | (they will become mirrors). |
4 | |||
5 | When the nodes are created with -drive they automatically get a | ||
6 | BlockBackend. In that case we know nothing is using them yet and it's | ||
7 | safe to set the IOThread. Add a force boolean to override the check. | ||
8 | 4 | ||
9 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
10 | Reviewed-by: Eric Blake <eblake@redhat.com> | 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> |
11 | Message-id: 20171207201320.19284-4-stefanha@redhat.com | 7 | Reviewed-by: Thomas Huth <thuth@redhat.com> |
8 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
9 | Message-id: 20210111115017.156802-5-stefanha@redhat.com | ||
12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
13 | --- | 11 | --- |
14 | qapi/block-core.json | 6 +++++- | 12 | README.rst | 4 ++-- |
15 | blockdev.c | 11 ++++++----- | 13 | 1 file changed, 2 insertions(+), 2 deletions(-) |
16 | 2 files changed, 11 insertions(+), 6 deletions(-) | ||
17 | 14 | ||
18 | diff --git a/qapi/block-core.json b/qapi/block-core.json | 15 | diff --git a/README.rst b/README.rst |
19 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
20 | --- a/qapi/block-core.json | 17 | --- a/README.rst |
21 | +++ b/qapi/block-core.json | 18 | +++ b/README.rst |
22 | @@ -XXX,XX +XXX,XX @@ | 19 | @@ -XXX,XX +XXX,XX @@ The QEMU source code is maintained under the GIT version control system. |
23 | # | 20 | |
24 | # @iothread: the name of the IOThread object or null for the main loop | 21 | .. code-block:: shell |
25 | # | 22 | |
26 | +# @force: true if the node and its children should be moved when a BlockBackend | 23 | - git clone https://git.qemu.org/git/qemu.git |
27 | +# is already attached | 24 | + git clone https://gitlab.com/qemu-project/qemu.git |
28 | +# | 25 | |
29 | # Note: this command is experimental and intended for test cases that need | 26 | When submitting patches, one common approach is to use 'git |
30 | # control over IOThreads only. | 27 | format-patch' and/or 'git send-email' to format & send the mail to the |
31 | # | 28 | @@ -XXX,XX +XXX,XX @@ The QEMU website is also maintained under source control. |
32 | @@ -XXX,XX +XXX,XX @@ | 29 | |
33 | ## | 30 | .. code-block:: shell |
34 | { 'command': 'x-blockdev-set-iothread', | 31 | |
35 | 'data' : { 'node-name': 'str', | 32 | - git clone https://git.qemu.org/git/qemu-web.git |
36 | - 'iothread': 'StrOrNull' } } | 33 | + git clone https://gitlab.com/qemu-project/qemu-web.git |
37 | + 'iothread': 'StrOrNull', | 34 | |
38 | + '*force': 'bool' } } | 35 | * `<https://www.qemu.org/2017/02/04/the-new-qemu-website-is-up/>`_ |
39 | diff --git a/blockdev.c b/blockdev.c | ||
40 | index XXXXXXX..XXXXXXX 100644 | ||
41 | --- a/blockdev.c | ||
42 | +++ b/blockdev.c | ||
43 | @@ -XXX,XX +XXX,XX @@ BlockJobInfoList *qmp_query_block_jobs(Error **errp) | ||
44 | } | ||
45 | |||
46 | void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread, | ||
47 | - Error **errp) | ||
48 | + bool has_force, bool force, Error **errp) | ||
49 | { | ||
50 | AioContext *old_context; | ||
51 | AioContext *new_context; | ||
52 | @@ -XXX,XX +XXX,XX @@ void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread, | ||
53 | return; | ||
54 | } | ||
55 | |||
56 | - /* If we want to allow more extreme test scenarios this guard could be | ||
57 | - * removed. For now it protects against accidents. */ | ||
58 | - if (bdrv_has_blk(bs)) { | ||
59 | - error_setg(errp, "Node %s is in use", node_name); | ||
60 | + /* Protects against accidents. */ | ||
61 | + if (!(has_force && force) && bdrv_has_blk(bs)) { | ||
62 | + error_setg(errp, "Node %s is associated with a BlockBackend and could " | ||
63 | + "be in use (use force=true to override this check)", | ||
64 | + node_name); | ||
65 | return; | ||
66 | } | ||
67 | 36 | ||
68 | -- | 37 | -- |
69 | 2.14.3 | 38 | 2.29.2 |
70 | 39 | ||
71 | diff view generated by jsdifflib |
1 | See the patch for why nested AioContext locking is no longer allowed. | 1 | qemu.org is running out of bandwidth and the QEMU project is moving |
---|---|---|---|
2 | towards a gating CI on GitLab. Use the GitLab repos instead of qemu.org | ||
3 | (they will become mirrors). | ||
2 | 4 | ||
3 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
4 | Reviewed-by: Eric Blake <eblake@redhat.com> | 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> |
5 | Message-id: 20171207201320.19284-3-stefanha@redhat.com | 7 | Reviewed-by: Thomas Huth <thuth@redhat.com> |
8 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
9 | Message-id: 20210111115017.156802-6-stefanha@redhat.com | ||
6 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
7 | --- | 11 | --- |
8 | docs/devel/multiple-iothreads.txt | 7 ++++--- | 12 | pc-bios/README | 4 ++-- |
9 | 1 file changed, 4 insertions(+), 3 deletions(-) | 13 | 1 file changed, 2 insertions(+), 2 deletions(-) |
10 | 14 | ||
11 | diff --git a/docs/devel/multiple-iothreads.txt b/docs/devel/multiple-iothreads.txt | 15 | diff --git a/pc-bios/README b/pc-bios/README |
12 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
13 | --- a/docs/devel/multiple-iothreads.txt | 17 | --- a/pc-bios/README |
14 | +++ b/docs/devel/multiple-iothreads.txt | 18 | +++ b/pc-bios/README |
15 | @@ -XXX,XX +XXX,XX @@ | 19 | @@ -XXX,XX +XXX,XX @@ |
16 | -Copyright (c) 2014 Red Hat Inc. | 20 | legacy x86 software to communicate with an attached serial console as |
17 | +Copyright (c) 2014-2017 Red Hat Inc. | 21 | if a video card were attached. The master sources reside in a subversion |
18 | 22 | repository at http://sgabios.googlecode.com/svn/trunk. A git mirror is | |
19 | This work is licensed under the terms of the GNU GPL, version 2 or later. See | 23 | - available at https://git.qemu.org/git/sgabios.git. |
20 | the COPYING file in the top-level directory. | 24 | + available at https://gitlab.com/qemu-project/sgabios.git. |
21 | @@ -XXX,XX +XXX,XX @@ aio_context_acquire()/aio_context_release() for mutual exclusion. Once the | 25 | |
22 | context is acquired no other thread can access it or run event loop iterations | 26 | - The PXE roms come from the iPXE project. Built with BANNER_TIME 0. |
23 | in this AioContext. | 27 | Sources available at http://ipxe.org. Vendor:Device ID -> ROM mapping: |
24 | 28 | @@ -XXX,XX +XXX,XX @@ | |
25 | -aio_context_acquire()/aio_context_release() calls may be nested. This | 29 | |
26 | -means you can call them if you're not sure whether #2 applies. | 30 | - The u-boot binary for e500 comes from the upstream denx u-boot project where |
27 | +Legacy code sometimes nests aio_context_acquire()/aio_context_release() calls. | 31 | it was compiled using the qemu-ppce500 target. |
28 | +Do not use nesting anymore, it is incompatible with the BDRV_POLL_WHILE() macro | 32 | - A git mirror is available at: https://git.qemu.org/git/u-boot.git |
29 | +used in the block layer and can lead to hangs. | 33 | + A git mirror is available at: https://gitlab.com/qemu-project/u-boot.git |
30 | 34 | The hash used to compile the current version is: 2072e72 | |
31 | There is currently no lock ordering rule if a thread needs to acquire multiple | 35 | |
32 | AioContexts simultaneously. Therefore, it is only safe for code holding the | 36 | - Skiboot (https://github.com/open-power/skiboot/) is an OPAL |
33 | -- | 37 | -- |
34 | 2.14.3 | 38 | 2.29.2 |
35 | 39 | ||
36 | diff view generated by jsdifflib |
1 | bdrv_unref() requires the AioContext lock because bdrv_flush() uses | 1 | qemu.org is running out of bandwidth and the QEMU project is moving |
---|---|---|---|
2 | BDRV_POLL_WHILE(), which assumes the AioContext is currently held. If | 2 | towards a gating CI on GitLab. Use the GitLab repos instead of qemu.org |
3 | BDRV_POLL_WHILE() runs without AioContext held the | 3 | (they will become mirrors). |
4 | pthread_mutex_unlock() call in aio_context_release() fails. | ||
5 | |||
6 | This patch moves bdrv_unref() into the AioContext locked region to solve | ||
7 | the following pthread_mutex_unlock() failure: | ||
8 | |||
9 | #0 0x00007f566181969b in raise () at /lib64/libc.so.6 | ||
10 | #1 0x00007f566181b3b1 in abort () at /lib64/libc.so.6 | ||
11 | #2 0x00005592cd590458 in error_exit (err=<optimized out>, msg=msg@entry=0x5592cdaf6d60 <__func__.23977> "qemu_mutex_unlock") at util/qemu-thread-posix.c:36 | ||
12 | #3 0x00005592cd96e738 in qemu_mutex_unlock (mutex=mutex@entry=0x5592ce9505e0) at util/qemu-thread-posix.c:96 | ||
13 | #4 0x00005592cd969b69 in aio_context_release (ctx=ctx@entry=0x5592ce950580) at util/async.c:507 | ||
14 | #5 0x00005592cd8ead78 in bdrv_flush (bs=bs@entry=0x5592cfa87210) at block/io.c:2478 | ||
15 | #6 0x00005592cd89df30 in bdrv_close (bs=0x5592cfa87210) at block.c:3207 | ||
16 | #7 0x00005592cd89df30 in bdrv_delete (bs=0x5592cfa87210) at block.c:3395 | ||
17 | #8 0x00005592cd89df30 in bdrv_unref (bs=0x5592cfa87210) at block.c:4418 | ||
18 | #9 0x00005592cd6b7f86 in qmp_transaction (dev_list=<optimized out>, has_props=<optimized out>, props=<optimized out>, errp=errp@entry=0x7ffe4a1fc9d8) at blockdev.c:2308 | ||
19 | 4 | ||
20 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
21 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> |
22 | Reviewed-by: Eric Blake <eblake@redhat.com> | 7 | Reviewed-by: Thomas Huth <thuth@redhat.com> |
23 | Message-id: 20171206144550.22295-2-stefanha@redhat.com | 8 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> |
9 | Message-id: 20210111115017.156802-7-stefanha@redhat.com | ||
24 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
25 | --- | 11 | --- |
26 | blockdev.c | 2 +- | 12 | scripts/get_maintainer.pl | 2 +- |
27 | 1 file changed, 1 insertion(+), 1 deletion(-) | 13 | 1 file changed, 1 insertion(+), 1 deletion(-) |
28 | 14 | ||
29 | diff --git a/blockdev.c b/blockdev.c | 15 | diff --git a/scripts/get_maintainer.pl b/scripts/get_maintainer.pl |
30 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100755 |
31 | --- a/blockdev.c | 17 | --- a/scripts/get_maintainer.pl |
32 | +++ b/blockdev.c | 18 | +++ b/scripts/get_maintainer.pl |
33 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_clean(BlkActionState *common) | 19 | @@ -XXX,XX +XXX,XX @@ sub vcs_exists { |
34 | DO_UPCAST(ExternalSnapshotState, common, common); | 20 | warn("$P: No supported VCS found. Add --nogit to options?\n"); |
35 | if (state->aio_context) { | 21 | warn("Using a git repository produces better results.\n"); |
36 | bdrv_drained_end(state->old_bs); | 22 | warn("Try latest git repository using:\n"); |
37 | - aio_context_release(state->aio_context); | 23 | - warn("git clone https://git.qemu.org/git/qemu.git\n"); |
38 | bdrv_unref(state->new_bs); | 24 | + warn("git clone https://gitlab.com/qemu-project/qemu.git\n"); |
39 | + aio_context_release(state->aio_context); | 25 | $printed_novcs = 1; |
40 | } | 26 | } |
41 | } | 27 | return 0; |
42 | |||
43 | -- | 28 | -- |
44 | 2.14.3 | 29 | 2.29.2 |
45 | 30 | ||
46 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: John G Johnson <john.g.johnson@oracle.com> | ||
1 | 2 | ||
3 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
4 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
5 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
6 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
7 | Message-id: 02a68adef99f5df6a380bf8fd7b90948777e411c.1611938319.git.jag.raman@oracle.com | ||
8 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
9 | --- | ||
10 | MAINTAINERS | 7 + | ||
11 | docs/devel/index.rst | 1 + | ||
12 | docs/devel/multi-process.rst | 966 +++++++++++++++++++++++++++++++++++ | ||
13 | 3 files changed, 974 insertions(+) | ||
14 | create mode 100644 docs/devel/multi-process.rst | ||
15 | |||
16 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
17 | index XXXXXXX..XXXXXXX 100644 | ||
18 | --- a/MAINTAINERS | ||
19 | +++ b/MAINTAINERS | ||
20 | @@ -XXX,XX +XXX,XX @@ S: Maintained | ||
21 | F: hw/semihosting/ | ||
22 | F: include/hw/semihosting/ | ||
23 | |||
24 | +Multi-process QEMU | ||
25 | +M: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
26 | +M: Jagannathan Raman <jag.raman@oracle.com> | ||
27 | +M: John G Johnson <john.g.johnson@oracle.com> | ||
28 | +S: Maintained | ||
29 | +F: docs/devel/multi-process.rst | ||
30 | + | ||
31 | Build and test automation | ||
32 | ------------------------- | ||
33 | Build and test automation | ||
34 | diff --git a/docs/devel/index.rst b/docs/devel/index.rst | ||
35 | index XXXXXXX..XXXXXXX 100644 | ||
36 | --- a/docs/devel/index.rst | ||
37 | +++ b/docs/devel/index.rst | ||
38 | @@ -XXX,XX +XXX,XX @@ Contents: | ||
39 | clocks | ||
40 | qom | ||
41 | block-coroutine-wrapper | ||
42 | + multi-process | ||
43 | diff --git a/docs/devel/multi-process.rst b/docs/devel/multi-process.rst | ||
44 | new file mode 100644 | ||
45 | index XXXXXXX..XXXXXXX | ||
46 | --- /dev/null | ||
47 | +++ b/docs/devel/multi-process.rst | ||
48 | @@ -XXX,XX +XXX,XX @@ | ||
49 | +This is the design document for multi-process QEMU. It does not | ||
50 | +necessarily reflect the status of the current implementation, which | ||
51 | +may lack features or be considerably different from what is described | ||
52 | +in this document. This document is still useful as a description of | ||
53 | +the goals and general direction of this feature. | ||
54 | + | ||
55 | +Please refer to the following wiki for latest details: | ||
56 | +https://wiki.qemu.org/Features/MultiProcessQEMU | ||
57 | + | ||
58 | +Multi-process QEMU | ||
59 | +=================== | ||
60 | + | ||
61 | +QEMU is often used as the hypervisor for virtual machines running in the | ||
62 | +Oracle cloud. Since one of the advantages of cloud computing is the | ||
63 | +ability to run many VMs from different tenants in the same cloud | ||
64 | +infrastructure, a guest that compromised its hypervisor could | ||
65 | +potentially use the hypervisor's access privileges to access data it is | ||
66 | +not authorized for. | ||
67 | + | ||
68 | +QEMU can be susceptible to security attacks because it is a large, | ||
69 | +monolithic program that provides many features to the VMs it services. | ||
70 | +Many of these features can be configured out of QEMU, but even a reduced | ||
71 | +configuration QEMU has a large amount of code a guest can potentially | ||
72 | +attack. Separating QEMU reduces the attack surface by aiding to | ||
73 | +limit each component in the system to only access the resources that | ||
74 | +it needs to perform its job. | ||
75 | + | ||
76 | +QEMU services | ||
77 | +------------- | ||
78 | + | ||
79 | +QEMU can be broadly described as providing three main services. One is a | ||
80 | +VM control point, where VMs can be created, migrated, re-configured, and | ||
81 | +destroyed. A second is to emulate the CPU instructions within the VM, | ||
82 | +often accelerated by HW virtualization features such as Intel's VT | ||
83 | +extensions. Finally, it provides IO services to the VM by emulating HW | ||
84 | +IO devices, such as disk and network devices. | ||
85 | + | ||
86 | +A multi-process QEMU | ||
87 | +~~~~~~~~~~~~~~~~~~~~ | ||
88 | + | ||
89 | +A multi-process QEMU involves separating QEMU services into separate | ||
90 | +host processes. Each of these processes can be given only the privileges | ||
91 | +it needs to provide its service, e.g., a disk service could be given | ||
92 | +access only to the disk images it provides, and not be allowed to | ||
93 | +access other files, or any network devices. An attacker who compromised | ||
94 | +this service would not be able to use this exploit to access files or | ||
95 | +devices beyond what the disk service was given access to. | ||
96 | + | ||
97 | +A QEMU control process would remain, but in multi-process mode, will | ||
98 | +have no direct interfaces to the VM. During VM execution, it would still | ||
99 | +provide the user interface to hot-plug devices or live migrate the VM. | ||
100 | + | ||
101 | +A first step in creating a multi-process QEMU is to separate IO services | ||
102 | +from the main QEMU program, which would continue to provide CPU | ||
103 | +emulation. i.e., the control process would also be the CPU emulation | ||
104 | +process. In a later phase, CPU emulation could be separated from the | ||
105 | +control process. | ||
106 | + | ||
107 | +Separating IO services | ||
108 | +---------------------- | ||
109 | + | ||
110 | +Separating IO services into individual host processes is a good place to | ||
111 | +begin for a couple of reasons. One is the sheer number of IO devices QEMU | ||
112 | +can emulate provides a large surface of interfaces which could potentially | ||
113 | +be exploited, and, indeed, have been a source of exploits in the past. | ||
114 | +Another is the modular nature of QEMU device emulation code provides | ||
115 | +interface points where the QEMU functions that perform device emulation | ||
116 | +can be separated from the QEMU functions that manage the emulation of | ||
117 | +guest CPU instructions. The devices emulated in the separate process are | ||
118 | +referred to as remote devices. | ||
119 | + | ||
120 | +QEMU device emulation | ||
121 | +~~~~~~~~~~~~~~~~~~~~~ | ||
122 | + | ||
123 | +QEMU uses an object oriented SW architecture for device emulation code. | ||
124 | +Configured objects are all compiled into the QEMU binary, then objects | ||
125 | +are instantiated by name when used by the guest VM. For example, the | ||
126 | +code to emulate a device named "foo" is always present in QEMU, but its | ||
127 | +instantiation code is only run when the device is included in the target | ||
128 | +VM. (e.g., via the QEMU command line as *-device foo*) | ||
129 | + | ||
130 | +The object model is hierarchical, so device emulation code names its | ||
131 | +parent object (such as "pci-device" for a PCI device) and QEMU will | ||
132 | +instantiate a parent object before calling the device's instantiation | ||
133 | +code. | ||
134 | + | ||
135 | +Current separation models | ||
136 | +~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
137 | + | ||
138 | +In order to separate the device emulation code from the CPU emulation | ||
139 | +code, the device object code must run in a different process. There are | ||
140 | +a couple of existing QEMU features that can run emulation code | ||
141 | +separately from the main QEMU process. These are examined below. | ||
142 | + | ||
143 | +vhost user model | ||
144 | +^^^^^^^^^^^^^^^^ | ||
145 | + | ||
146 | +Virtio guest device drivers can be connected to vhost user applications | ||
147 | +in order to perform their IO operations. This model uses special virtio | ||
148 | +device drivers in the guest and vhost user device objects in QEMU, but | ||
149 | +once the QEMU vhost user code has configured the vhost user application, | ||
150 | +mission-mode IO is performed by the application. The vhost user | ||
151 | +application is a daemon process that can be contacted via a known UNIX | ||
152 | +domain socket. | ||
153 | + | ||
154 | +vhost socket | ||
155 | +'''''''''''' | ||
156 | + | ||
157 | +As mentioned above, one of the tasks of the vhost device object within | ||
158 | +QEMU is to contact the vhost application and send it configuration | ||
159 | +information about this device instance. As part of the configuration | ||
160 | +process, the application can also be sent other file descriptors over | ||
161 | +the socket, which then can be used by the vhost user application in | ||
162 | +various ways, some of which are described below. | ||
163 | + | ||
164 | +vhost MMIO store acceleration | ||
165 | +''''''''''''''''''''''''''''' | ||
166 | + | ||
167 | +VMs are often run using HW virtualization features via the KVM kernel | ||
168 | +driver. This driver allows QEMU to accelerate the emulation of guest CPU | ||
169 | +instructions by running the guest in a virtual HW mode. When the guest | ||
170 | +executes instructions that cannot be executed by virtual HW mode, | ||
171 | +execution returns to the KVM driver so it can inform QEMU to emulate the | ||
172 | +instructions in SW. | ||
173 | + | ||
174 | +One of the events that can cause a return to QEMU is when a guest device | ||
175 | +driver accesses an IO location. QEMU then dispatches the memory | ||
176 | +operation to the corresponding QEMU device object. In the case of a | ||
177 | +vhost user device, the memory operation would need to be sent over a | ||
178 | +socket to the vhost application. This path is accelerated by the QEMU | ||
179 | +virtio code by setting up an eventfd file descriptor that the vhost | ||
180 | +application can directly receive MMIO store notifications from the KVM | ||
181 | +driver, instead of needing them to be sent to the QEMU process first. | ||
182 | + | ||
183 | +vhost interrupt acceleration | ||
184 | +'''''''''''''''''''''''''''' | ||
185 | + | ||
186 | +Another optimization used by the vhost application is the ability to | ||
187 | +directly inject interrupts into the VM via the KVM driver, again, | ||
188 | +bypassing the need to send the interrupt back to the QEMU process first. | ||
189 | +The QEMU virtio setup code configures the KVM driver with an eventfd | ||
190 | +that triggers the device interrupt in the guest when the eventfd is | ||
191 | +written. This irqfd file descriptor is then passed to the vhost user | ||
192 | +application program. | ||
193 | + | ||
194 | +vhost access to guest memory | ||
195 | +'''''''''''''''''''''''''''' | ||
196 | + | ||
197 | +The vhost application is also allowed to directly access guest memory, | ||
198 | +instead of needing to send the data as messages to QEMU. This is also | ||
199 | +done with file descriptors sent to the vhost user application by QEMU. | ||
200 | +These descriptors can be passed to ``mmap()`` by the vhost application | ||
201 | +to map the guest address space into the vhost application. | ||
202 | + | ||
203 | +IOMMUs introduce another level of complexity, since the address given to | ||
204 | +the guest virtio device to DMA to or from is not a guest physical | ||
205 | +address. This case is handled by having vhost code within QEMU register | ||
206 | +as a listener for IOMMU mapping changes. The vhost application maintains | ||
207 | +a cache of IOMMMU translations: sending translation requests back to | ||
208 | +QEMU on cache misses, and in turn receiving flush requests from QEMU | ||
209 | +when mappings are purged. | ||
210 | + | ||
211 | +applicability to device separation | ||
212 | +'''''''''''''''''''''''''''''''''' | ||
213 | + | ||
214 | +Much of the vhost model can be re-used by separated device emulation. In | ||
215 | +particular, the ideas of using a socket between QEMU and the device | ||
216 | +emulation application, using a file descriptor to inject interrupts into | ||
217 | +the VM via KVM, and allowing the application to ``mmap()`` the guest | ||
218 | +should be re used. | ||
219 | + | ||
220 | +There are, however, some notable differences between how a vhost | ||
221 | +application works and the needs of separated device emulation. The most | ||
222 | +basic is that vhost uses custom virtio device drivers which always | ||
223 | +trigger IO with MMIO stores. A separated device emulation model must | ||
224 | +work with existing IO device models and guest device drivers. MMIO loads | ||
225 | +break vhost store acceleration since they are synchronous - guest | ||
226 | +progress cannot continue until the load has been emulated. By contrast, | ||
227 | +stores are asynchronous, the guest can continue after the store event | ||
228 | +has been sent to the vhost application. | ||
229 | + | ||
230 | +Another difference is that in the vhost user model, a single daemon can | ||
231 | +support multiple QEMU instances. This is contrary to the security regime | ||
232 | +desired, in which the emulation application should only be allowed to | ||
233 | +access the files or devices the VM it's running on behalf of can access. | ||
234 | +#### qemu-io model | ||
235 | + | ||
236 | +Qemu-io is a test harness used to test changes to the QEMU block backend | ||
237 | +object code. (e.g., the code that implements disk images for disk driver | ||
238 | +emulation) Qemu-io is not a device emulation application per se, but it | ||
239 | +does compile the QEMU block objects into a separate binary from the main | ||
240 | +QEMU one. This could be useful for disk device emulation, since its | ||
241 | +emulation applications will need to include the QEMU block objects. | ||
242 | + | ||
243 | +New separation model based on proxy objects | ||
244 | +------------------------------------------- | ||
245 | + | ||
246 | +A different model based on proxy objects in the QEMU program | ||
247 | +communicating with remote emulation programs could provide separation | ||
248 | +while minimizing the changes needed to the device emulation code. The | ||
249 | +rest of this section is a discussion of how a proxy object model would | ||
250 | +work. | ||
251 | + | ||
252 | +Remote emulation processes | ||
253 | +~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
254 | + | ||
255 | +The remote emulation process will run the QEMU object hierarchy without | ||
256 | +modification. The device emulation objects will be also be based on the | ||
257 | +QEMU code, because for anything but the simplest device, it would not be | ||
258 | +a tractable to re-implement both the object model and the many device | ||
259 | +backends that QEMU has. | ||
260 | + | ||
261 | +The processes will communicate with the QEMU process over UNIX domain | ||
262 | +sockets. The processes can be executed either as standalone processes, | ||
263 | +or be executed by QEMU. In both cases, the host backends the emulation | ||
264 | +processes will provide are specified on its command line, as they would | ||
265 | +be for QEMU. For example: | ||
266 | + | ||
267 | +:: | ||
268 | + | ||
269 | + disk-proc -blockdev driver=file,node-name=file0,filename=disk-file0 \ | ||
270 | + -blockdev driver=qcow2,node-name=drive0,file=file0 | ||
271 | + | ||
272 | +would indicate process *disk-proc* uses a qcow2 emulated disk named | ||
273 | +*file0* as its backend. | ||
274 | + | ||
275 | +Emulation processes may emulate more than one guest controller. A common | ||
276 | +configuration might be to put all controllers of the same device class | ||
277 | +(e.g., disk, network, etc.) in a single process, so that all backends of | ||
278 | +the same type can be managed by a single QMP monitor. | ||
279 | + | ||
280 | +communication with QEMU | ||
281 | +^^^^^^^^^^^^^^^^^^^^^^^ | ||
282 | + | ||
283 | +The first argument to the remote emulation process will be a Unix domain | ||
284 | +socket that connects with the Proxy object. This is a required argument. | ||
285 | + | ||
286 | +:: | ||
287 | + | ||
288 | + disk-proc <socket number> <backend list> | ||
289 | + | ||
290 | +remote process QMP monitor | ||
291 | +^^^^^^^^^^^^^^^^^^^^^^^^^^ | ||
292 | + | ||
293 | +Remote emulation processes can be monitored via QMP, similar to QEMU | ||
294 | +itself. The QMP monitor socket is specified the same as for a QEMU | ||
295 | +process: | ||
296 | + | ||
297 | +:: | ||
298 | + | ||
299 | + disk-proc -qmp unix:/tmp/disk-mon,server | ||
300 | + | ||
301 | +can be monitored over the UNIX socket path */tmp/disk-mon*. | ||
302 | + | ||
303 | +QEMU command line | ||
304 | +~~~~~~~~~~~~~~~~~ | ||
305 | + | ||
306 | +Each remote device emulated in a remote process on the host is | ||
307 | +represented as a *-device* of type *pci-proxy-dev*. A socket | ||
308 | +sub-option to this option specifies the Unix socket that connects | ||
309 | +to the remote process. An *id* sub-option is required, and it should | ||
310 | +be the same id as used in the remote process. | ||
311 | + | ||
312 | +:: | ||
313 | + | ||
314 | + qemu-system-x86_64 ... -device pci-proxy-dev,id=lsi0,socket=3 | ||
315 | + | ||
316 | +can be used to add a device emulated in a remote process | ||
317 | + | ||
318 | + | ||
319 | +QEMU management of remote processes | ||
320 | +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
321 | + | ||
322 | +QEMU is not aware of the type of type of the remote PCI device. It is | ||
323 | +a pass through device as far as QEMU is concerned. | ||
324 | + | ||
325 | +communication with emulation process | ||
326 | +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ | ||
327 | + | ||
328 | +primary channel | ||
329 | +''''''''''''''' | ||
330 | + | ||
331 | +The primary channel (referred to as com in the code) is used to bootstrap | ||
332 | +the remote process. It is also used to pass on device-agnostic commands | ||
333 | +like reset. | ||
334 | + | ||
335 | +per-device channels | ||
336 | +''''''''''''''''''' | ||
337 | + | ||
338 | +Each remote device communicates with QEMU using a dedicated communication | ||
339 | +channel. The proxy object sets up this channel using the primary | ||
340 | +channel during its initialization. | ||
341 | + | ||
342 | +QEMU device proxy objects | ||
343 | +~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
344 | + | ||
345 | +QEMU has an object model based on sub-classes inherited from the | ||
346 | +"object" super-class. The sub-classes that are of interest here are the | ||
347 | +"device" and "bus" sub-classes whose child sub-classes make up the | ||
348 | +device tree of a QEMU emulated system. | ||
349 | + | ||
350 | +The proxy object model will use device proxy objects to replace the | ||
351 | +device emulation code within the QEMU process. These objects will live | ||
352 | +in the same place in the object and bus hierarchies as the objects they | ||
353 | +replace. i.e., the proxy object for an LSI SCSI controller will be a | ||
354 | +sub-class of the "pci-device" class, and will have the same PCI bus | ||
355 | +parent and the same SCSI bus child objects as the LSI controller object | ||
356 | +it replaces. | ||
357 | + | ||
358 | +It is worth noting that the same proxy object is used to mediate with | ||
359 | +all types of remote PCI devices. | ||
360 | + | ||
361 | +object initialization | ||
362 | +^^^^^^^^^^^^^^^^^^^^^ | ||
363 | + | ||
364 | +The Proxy device objects are initialized in the exact same manner in | ||
365 | +which any other QEMU device would be initialized. | ||
366 | + | ||
367 | +In addition, the Proxy objects perform the following two tasks: | ||
368 | +- Parses the "socket" sub option and connects to the remote process | ||
369 | +using this channel | ||
370 | +- Uses the "id" sub-option to connect to the emulated device on the | ||
371 | +separate process | ||
372 | + | ||
373 | +class\_init | ||
374 | +''''''''''' | ||
375 | + | ||
376 | +The ``class_init()`` method of a proxy object will, in general behave | ||
377 | +similarly to the object it replaces, including setting any static | ||
378 | +properties and methods needed by the proxy. | ||
379 | + | ||
380 | +instance\_init / realize | ||
381 | +'''''''''''''''''''''''' | ||
382 | + | ||
383 | +The ``instance_init()`` and ``realize()`` functions would only need to | ||
384 | +perform tasks related to being a proxy, such are registering its own | ||
385 | +MMIO handlers, or creating a child bus that other proxy devices can be | ||
386 | +attached to later. | ||
387 | + | ||
388 | +Other tasks will be device-specific. For example, PCI device objects | ||
389 | +will initialize the PCI config space in order to make a valid PCI device | ||
390 | +tree within the QEMU process. | ||
391 | + | ||
392 | +address space registration | ||
393 | +^^^^^^^^^^^^^^^^^^^^^^^^^^ | ||
394 | + | ||
395 | +Most devices are driven by guest device driver accesses to IO addresses | ||
396 | +or ports. The QEMU device emulation code uses QEMU's memory region | ||
397 | +function calls (such as ``memory_region_init_io()``) to add callback | ||
398 | +functions that QEMU will invoke when the guest accesses the device's | ||
399 | +areas of the IO address space. When a guest driver does access the | ||
400 | +device, the VM will exit HW virtualization mode and return to QEMU, | ||
401 | +which will then lookup and execute the corresponding callback function. | ||
402 | + | ||
403 | +A proxy object would need to mirror the memory region calls the actual | ||
404 | +device emulator would perform in its initialization code, but with its | ||
405 | +own callbacks. When invoked by QEMU as a result of a guest IO operation, | ||
406 | +they will forward the operation to the device emulation process. | ||
407 | + | ||
408 | +PCI config space | ||
409 | +^^^^^^^^^^^^^^^^ | ||
410 | + | ||
411 | +PCI devices also have a configuration space that can be accessed by the | ||
412 | +guest driver. Guest accesses to this space is not handled by the device | ||
413 | +emulation object, but by its PCI parent object. Much of this space is | ||
414 | +read-only, but certain registers (especially BAR and MSI-related ones) | ||
415 | +need to be propagated to the emulation process. | ||
416 | + | ||
417 | +PCI parent proxy | ||
418 | +'''''''''''''''' | ||
419 | + | ||
420 | +One way to propagate guest PCI config accesses is to create a | ||
421 | +"pci-device-proxy" class that can serve as the parent of a PCI device | ||
422 | +proxy object. This class's parent would be "pci-device" and it would | ||
423 | +override the PCI parent's ``config_read()`` and ``config_write()`` | ||
424 | +methods with ones that forward these operations to the emulation | ||
425 | +program. | ||
426 | + | ||
427 | +interrupt receipt | ||
428 | +^^^^^^^^^^^^^^^^^ | ||
429 | + | ||
430 | +A proxy for a device that generates interrupts will need to create a | ||
431 | +socket to receive interrupt indications from the emulation process. An | ||
432 | +incoming interrupt indication would then be sent up to its bus parent to | ||
433 | +be injected into the guest. For example, a PCI device object may use | ||
434 | +``pci_set_irq()``. | ||
435 | + | ||
436 | +live migration | ||
437 | +^^^^^^^^^^^^^^ | ||
438 | + | ||
439 | +The proxy will register to save and restore any *vmstate* it needs over | ||
440 | +a live migration event. The device proxy does not need to manage the | ||
441 | +remote device's *vmstate*; that will be handled by the remote process | ||
442 | +proxy (see below). | ||
443 | + | ||
444 | +QEMU remote device operation | ||
445 | +~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
446 | + | ||
447 | +Generic device operations, such as DMA, will be performed by the remote | ||
448 | +process proxy by sending messages to the remote process. | ||
449 | + | ||
450 | +DMA operations | ||
451 | +^^^^^^^^^^^^^^ | ||
452 | + | ||
453 | +DMA operations would be handled much like vhost applications do. One of | ||
454 | +the initial messages sent to the emulation process is a guest memory | ||
455 | +table. Each entry in this table consists of a file descriptor and size | ||
456 | +that the emulation process can ``mmap()`` to directly access guest | ||
457 | +memory, similar to ``vhost_user_set_mem_table()``. Note guest memory | ||
458 | +must be backed by file descriptors, such as when QEMU is given the | ||
459 | +*-mem-path* command line option. | ||
460 | + | ||
461 | +IOMMU operations | ||
462 | +^^^^^^^^^^^^^^^^ | ||
463 | + | ||
464 | +When the emulated system includes an IOMMU, the remote process proxy in | ||
465 | +QEMU will need to create a socket for IOMMU requests from the emulation | ||
466 | +process. It will handle those requests with an | ||
467 | +``address_space_get_iotlb_entry()`` call. In order to handle IOMMU | ||
468 | +unmaps, the remote process proxy will also register as a listener on the | ||
469 | +device's DMA address space. When an IOMMU memory region is created | ||
470 | +within the DMA address space, an IOMMU notifier for unmaps will be added | ||
471 | +to the memory region that will forward unmaps to the emulation process | ||
472 | +over the IOMMU socket. | ||
473 | + | ||
474 | +device hot-plug via QMP | ||
475 | +^^^^^^^^^^^^^^^^^^^^^^^ | ||
476 | + | ||
477 | +An QMP "device\_add" command can add a device emulated by a remote | ||
478 | +process. It will also have "rid" option to the command, just as the | ||
479 | +*-device* command line option does. The remote process may either be one | ||
480 | +started at QEMU startup, or be one added by the "add-process" QMP | ||
481 | +command described above. In either case, the remote process proxy will | ||
482 | +forward the new device's JSON description to the corresponding emulation | ||
483 | +process. | ||
484 | + | ||
485 | +live migration | ||
486 | +^^^^^^^^^^^^^^ | ||
487 | + | ||
488 | +The remote process proxy will also register for live migration | ||
489 | +notifications with ``vmstate_register()``. When called to save state, | ||
490 | +the proxy will send the remote process a secondary socket file | ||
491 | +descriptor to save the remote process's device *vmstate* over. The | ||
492 | +incoming byte stream length and data will be saved as the proxy's | ||
493 | +*vmstate*. When the proxy is resumed on its new host, this *vmstate* | ||
494 | +will be extracted, and a secondary socket file descriptor will be sent | ||
495 | +to the new remote process through which it receives the *vmstate* in | ||
496 | +order to restore the devices there. | ||
497 | + | ||
498 | +device emulation in remote process | ||
499 | +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
500 | + | ||
501 | +The parts of QEMU that the emulation program will need include the | ||
502 | +object model; the memory emulation objects; the device emulation objects | ||
503 | +of the targeted device, and any dependent devices; and, the device's | ||
504 | +backends. It will also need code to setup the machine environment, | ||
505 | +handle requests from the QEMU process, and route machine-level requests | ||
506 | +(such as interrupts or IOMMU mappings) back to the QEMU process. | ||
507 | + | ||
508 | +initialization | ||
509 | +^^^^^^^^^^^^^^ | ||
510 | + | ||
511 | +The process initialization sequence will follow the same sequence | ||
512 | +followed by QEMU. It will first initialize the backend objects, then | ||
513 | +device emulation objects. The JSON descriptions sent by the QEMU process | ||
514 | +will drive which objects need to be created. | ||
515 | + | ||
516 | +- address spaces | ||
517 | + | ||
518 | +Before the device objects are created, the initial address spaces and | ||
519 | +memory regions must be configured with ``memory_map_init()``. This | ||
520 | +creates a RAM memory region object (*system\_memory*) and an IO memory | ||
521 | +region object (*system\_io*). | ||
522 | + | ||
523 | +- RAM | ||
524 | + | ||
525 | +RAM memory region creation will follow how ``pc_memory_init()`` creates | ||
526 | +them, but must use ``memory_region_init_ram_from_fd()`` instead of | ||
527 | +``memory_region_allocate_system_memory()``. The file descriptors needed | ||
528 | +will be supplied by the guest memory table from above. Those RAM regions | ||
529 | +would then be added to the *system\_memory* memory region with | ||
530 | +``memory_region_add_subregion()``. | ||
531 | + | ||
532 | +- PCI | ||
533 | + | ||
534 | +IO initialization will be driven by the JSON descriptions sent from the | ||
535 | +QEMU process. For a PCI device, a PCI bus will need to be created with | ||
536 | +``pci_root_bus_new()``, and a PCI memory region will need to be created | ||
537 | +and added to the *system\_memory* memory region with | ||
538 | +``memory_region_add_subregion_overlap()``. The overlap version is | ||
539 | +required for architectures where PCI memory overlaps with RAM memory. | ||
540 | + | ||
541 | +MMIO handling | ||
542 | +^^^^^^^^^^^^^ | ||
543 | + | ||
544 | +The device emulation objects will use ``memory_region_init_io()`` to | ||
545 | +install their MMIO handlers, and ``pci_register_bar()`` to associate | ||
546 | +those handlers with a PCI BAR, as they do within QEMU currently. | ||
547 | + | ||
548 | +In order to use ``address_space_rw()`` in the emulation process to | ||
549 | +handle MMIO requests from QEMU, the PCI physical addresses must be the | ||
550 | +same in the QEMU process and the device emulation process. In order to | ||
551 | +accomplish that, guest BAR programming must also be forwarded from QEMU | ||
552 | +to the emulation process. | ||
553 | + | ||
554 | +interrupt injection | ||
555 | +^^^^^^^^^^^^^^^^^^^ | ||
556 | + | ||
557 | +When device emulation wants to inject an interrupt into the VM, the | ||
558 | +request climbs the device's bus object hierarchy until the point where a | ||
559 | +bus object knows how to signal the interrupt to the guest. The details | ||
560 | +depend on the type of interrupt being raised. | ||
561 | + | ||
562 | +- PCI pin interrupts | ||
563 | + | ||
564 | +On x86 systems, there is an emulated IOAPIC object attached to the root | ||
565 | +PCI bus object, and the root PCI object forwards interrupt requests to | ||
566 | +it. The IOAPIC object, in turn, calls the KVM driver to inject the | ||
567 | +corresponding interrupt into the VM. The simplest way to handle this in | ||
568 | +an emulation process would be to setup the root PCI bus driver (via | ||
569 | +``pci_bus_irqs()``) to send a interrupt request back to the QEMU | ||
570 | +process, and have the device proxy object reflect it up the PCI tree | ||
571 | +there. | ||
572 | + | ||
573 | +- PCI MSI/X interrupts | ||
574 | + | ||
575 | +PCI MSI/X interrupts are implemented in HW as DMA writes to a | ||
576 | +CPU-specific PCI address. In QEMU on x86, a KVM APIC object receives | ||
577 | +these DMA writes, then calls into the KVM driver to inject the interrupt | ||
578 | +into the VM. A simple emulation process implementation would be to send | ||
579 | +the MSI DMA address from QEMU as a message at initialization, then | ||
580 | +install an address space handler at that address which forwards the MSI | ||
581 | +message back to QEMU. | ||
582 | + | ||
583 | +DMA operations | ||
584 | +^^^^^^^^^^^^^^ | ||
585 | + | ||
586 | +When a emulation object wants to DMA into or out of guest memory, it | ||
587 | +first must use dma\_memory\_map() to convert the DMA address to a local | ||
588 | +virtual address. The emulation process memory region objects setup above | ||
589 | +will be used to translate the DMA address to a local virtual address the | ||
590 | +device emulation code can access. | ||
591 | + | ||
592 | +IOMMU | ||
593 | +^^^^^ | ||
594 | + | ||
595 | +When an IOMMU is in use in QEMU, DMA translation uses IOMMU memory | ||
596 | +regions to translate the DMA address to a guest physical address before | ||
597 | +that physical address can be translated to a local virtual address. The | ||
598 | +emulation process will need similar functionality. | ||
599 | + | ||
600 | +- IOTLB cache | ||
601 | + | ||
602 | +The emulation process will maintain a cache of recent IOMMU translations | ||
603 | +(the IOTLB). When the translate() callback of an IOMMU memory region is | ||
604 | +invoked, the IOTLB cache will be searched for an entry that will map the | ||
605 | +DMA address to a guest PA. On a cache miss, a message will be sent back | ||
606 | +to QEMU requesting the corresponding translation entry, which be both be | ||
607 | +used to return a guest address and be added to the cache. | ||
608 | + | ||
609 | +- IOTLB purge | ||
610 | + | ||
611 | +The IOMMU emulation will also need to act on unmap requests from QEMU. | ||
612 | +These happen when the guest IOMMU driver purges an entry from the | ||
613 | +guest's translation table. | ||
614 | + | ||
615 | +live migration | ||
616 | +^^^^^^^^^^^^^^ | ||
617 | + | ||
618 | +When a remote process receives a live migration indication from QEMU, it | ||
619 | +will set up a channel using the received file descriptor with | ||
620 | +``qio_channel_socket_new_fd()``. This channel will be used to create a | ||
621 | +*QEMUfile* that can be passed to ``qemu_save_device_state()`` to send | ||
622 | +the process's device state back to QEMU. This method will be reversed on | ||
623 | +restore - the channel will be passed to ``qemu_loadvm_state()`` to | ||
624 | +restore the device state. | ||
625 | + | ||
626 | +Accelerating device emulation | ||
627 | +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
628 | + | ||
629 | +The messages that are required to be sent between QEMU and the emulation | ||
630 | +process can add considerable latency to IO operations. The optimizations | ||
631 | +described below attempt to ameliorate this effect by allowing the | ||
632 | +emulation process to communicate directly with the kernel KVM driver. | ||
633 | +The KVM file descriptors created would be passed to the emulation process | ||
634 | +via initialization messages, much like the guest memory table is done. | ||
635 | +#### MMIO acceleration | ||
636 | + | ||
637 | +Vhost user applications can receive guest virtio driver stores directly | ||
638 | +from KVM. The issue with the eventfd mechanism used by vhost user is | ||
639 | +that it does not pass any data with the event indication, so it cannot | ||
640 | +handle guest loads or guest stores that carry store data. This concept | ||
641 | +could, however, be expanded to cover more cases. | ||
642 | + | ||
643 | +The expanded idea would require a new type of KVM device: | ||
644 | +*KVM\_DEV\_TYPE\_USER*. This device has two file descriptors: a master | ||
645 | +descriptor that QEMU can use for configuration, and a slave descriptor | ||
646 | +that the emulation process can use to receive MMIO notifications. QEMU | ||
647 | +would create both descriptors using the KVM driver, and pass the slave | ||
648 | +descriptor to the emulation process via an initialization message. | ||
649 | + | ||
650 | +data structures | ||
651 | +^^^^^^^^^^^^^^^ | ||
652 | + | ||
653 | +- guest physical range | ||
654 | + | ||
655 | +The guest physical range structure describes the address range that a | ||
656 | +device will respond to. It includes the base and length of the range, as | ||
657 | +well as which bus the range resides on (e.g., on an x86machine, it can | ||
658 | +specify whether the range refers to memory or IO addresses). | ||
659 | + | ||
660 | +A device can have multiple physical address ranges it responds to (e.g., | ||
661 | +a PCI device can have multiple BARs), so the structure will also include | ||
662 | +an enumerated identifier to specify which of the device's ranges is | ||
663 | +being referred to. | ||
664 | + | ||
665 | ++--------+----------------------------+ | ||
666 | +| Name | Description | | ||
667 | ++========+============================+ | ||
668 | +| addr | range base address | | ||
669 | ++--------+----------------------------+ | ||
670 | +| len | range length | | ||
671 | ++--------+----------------------------+ | ||
672 | +| bus | addr type (memory or IO) | | ||
673 | ++--------+----------------------------+ | ||
674 | +| id | range ID (e.g., PCI BAR) | | ||
675 | ++--------+----------------------------+ | ||
676 | + | ||
677 | +- MMIO request structure | ||
678 | + | ||
679 | +This structure describes an MMIO operation. It includes which guest | ||
680 | +physical range the MMIO was within, the offset within that range, the | ||
681 | +MMIO type (e.g., load or store), and its length and data. It also | ||
682 | +includes a sequence number that can be used to reply to the MMIO, and | ||
683 | +the CPU that issued the MMIO. | ||
684 | + | ||
685 | ++----------+------------------------+ | ||
686 | +| Name | Description | | ||
687 | ++==========+========================+ | ||
688 | +| rid | range MMIO is within | | ||
689 | ++----------+------------------------+ | ||
690 | +| offset | offset withing *rid* | | ||
691 | ++----------+------------------------+ | ||
692 | +| type | e.g., load or store | | ||
693 | ++----------+------------------------+ | ||
694 | +| len | MMIO length | | ||
695 | ++----------+------------------------+ | ||
696 | +| data | store data | | ||
697 | ++----------+------------------------+ | ||
698 | +| seq | sequence ID | | ||
699 | ++----------+------------------------+ | ||
700 | + | ||
701 | +- MMIO request queues | ||
702 | + | ||
703 | +MMIO request queues are FIFO arrays of MMIO request structures. There | ||
704 | +are two queues: pending queue is for MMIOs that haven't been read by the | ||
705 | +emulation program, and the sent queue is for MMIOs that haven't been | ||
706 | +acknowledged. The main use of the second queue is to validate MMIO | ||
707 | +replies from the emulation program. | ||
708 | + | ||
709 | +- scoreboard | ||
710 | + | ||
711 | +Each CPU in the VM is emulated in QEMU by a separate thread, so multiple | ||
712 | +MMIOs may be waiting to be consumed by an emulation program and multiple | ||
713 | +threads may be waiting for MMIO replies. The scoreboard would contain a | ||
714 | +wait queue and sequence number for the per-CPU threads, allowing them to | ||
715 | +be individually woken when the MMIO reply is received from the emulation | ||
716 | +program. It also tracks the number of posted MMIO stores to the device | ||
717 | +that haven't been replied to, in order to satisfy the PCI constraint | ||
718 | +that a load to a device will not complete until all previous stores to | ||
719 | +that device have been completed. | ||
720 | + | ||
721 | +- device shadow memory | ||
722 | + | ||
723 | +Some MMIO loads do not have device side-effects. These MMIOs can be | ||
724 | +completed without sending a MMIO request to the emulation program if the | ||
725 | +emulation program shares a shadow image of the device's memory image | ||
726 | +with the KVM driver. | ||
727 | + | ||
728 | +The emulation program will ask the KVM driver to allocate memory for the | ||
729 | +shadow image, and will then use ``mmap()`` to directly access it. The | ||
730 | +emulation program can control KVM access to the shadow image by sending | ||
731 | +KVM an access map telling it which areas of the image have no | ||
732 | +side-effects (and can be completed immediately), and which require a | ||
733 | +MMIO request to the emulation program. The access map can also inform | ||
734 | +the KVM drive which size accesses are allowed to the image. | ||
735 | + | ||
736 | +master descriptor | ||
737 | +^^^^^^^^^^^^^^^^^ | ||
738 | + | ||
739 | +The master descriptor is used by QEMU to configure the new KVM device. | ||
740 | +The descriptor would be returned by the KVM driver when QEMU issues a | ||
741 | +*KVM\_CREATE\_DEVICE* ``ioctl()`` with a *KVM\_DEV\_TYPE\_USER* type. | ||
742 | + | ||
743 | +KVM\_DEV\_TYPE\_USER device ops | ||
744 | + | ||
745 | + | ||
746 | +The *KVM\_DEV\_TYPE\_USER* operations vector will be registered by a | ||
747 | +``kvm_register_device_ops()`` call when the KVM system in initialized by | ||
748 | +``kvm_init()``. These device ops are called by the KVM driver when QEMU | ||
749 | +executes certain ``ioctl()`` operations on its KVM file descriptor. They | ||
750 | +include: | ||
751 | + | ||
752 | +- create | ||
753 | + | ||
754 | +This routine is called when QEMU issues a *KVM\_CREATE\_DEVICE* | ||
755 | +``ioctl()`` on its per-VM file descriptor. It will allocate and | ||
756 | +initialize a KVM user device specific data structure, and assign the | ||
757 | +*kvm\_device* private field to it. | ||
758 | + | ||
759 | +- ioctl | ||
760 | + | ||
761 | +This routine is invoked when QEMU issues an ``ioctl()`` on the master | ||
762 | +descriptor. The ``ioctl()`` commands supported are defined by the KVM | ||
763 | +device type. *KVM\_DEV\_TYPE\_USER* ones will need several commands: | ||
764 | + | ||
765 | +*KVM\_DEV\_USER\_SLAVE\_FD* creates the slave file descriptor that will | ||
766 | +be passed to the device emulation program. Only one slave can be created | ||
767 | +by each master descriptor. The file operations performed by this | ||
768 | +descriptor are described below. | ||
769 | + | ||
770 | +The *KVM\_DEV\_USER\_PA\_RANGE* command configures a guest physical | ||
771 | +address range that the slave descriptor will receive MMIO notifications | ||
772 | +for. The range is specified by a guest physical range structure | ||
773 | +argument. For buses that assign addresses to devices dynamically, this | ||
774 | +command can be executed while the guest is running, such as the case | ||
775 | +when a guest changes a device's PCI BAR registers. | ||
776 | + | ||
777 | +*KVM\_DEV\_USER\_PA\_RANGE* will use ``kvm_io_bus_register_dev()`` to | ||
778 | +register *kvm\_io\_device\_ops* callbacks to be invoked when the guest | ||
779 | +performs a MMIO operation within the range. When a range is changed, | ||
780 | +``kvm_io_bus_unregister_dev()`` is used to remove the previous | ||
781 | +instantiation. | ||
782 | + | ||
783 | +*KVM\_DEV\_USER\_TIMEOUT* will configure a timeout value that specifies | ||
784 | +how long KVM will wait for the emulation process to respond to a MMIO | ||
785 | +indication. | ||
786 | + | ||
787 | +- destroy | ||
788 | + | ||
789 | +This routine is called when the VM instance is destroyed. It will need | ||
790 | +to destroy the slave descriptor; and free any memory allocated by the | ||
791 | +driver, as well as the *kvm\_device* structure itself. | ||
792 | + | ||
793 | +slave descriptor | ||
794 | +^^^^^^^^^^^^^^^^ | ||
795 | + | ||
796 | +The slave descriptor will have its own file operations vector, which | ||
797 | +responds to system calls on the descriptor performed by the device | ||
798 | +emulation program. | ||
799 | + | ||
800 | +- read | ||
801 | + | ||
802 | +A read returns any pending MMIO requests from the KVM driver as MMIO | ||
803 | +request structures. Multiple structures can be returned if there are | ||
804 | +multiple MMIO operations pending. The MMIO requests are moved from the | ||
805 | +pending queue to the sent queue, and if there are threads waiting for | ||
806 | +space in the pending to add new MMIO operations, they will be woken | ||
807 | +here. | ||
808 | + | ||
809 | +- write | ||
810 | + | ||
811 | +A write also consists of a set of MMIO requests. They are compared to | ||
812 | +the MMIO requests in the sent queue. Matches are removed from the sent | ||
813 | +queue, and any threads waiting for the reply are woken. If a store is | ||
814 | +removed, then the number of posted stores in the per-CPU scoreboard is | ||
815 | +decremented. When the number is zero, and a non side-effect load was | ||
816 | +waiting for posted stores to complete, the load is continued. | ||
817 | + | ||
818 | +- ioctl | ||
819 | + | ||
820 | +There are several ioctl()s that can be performed on the slave | ||
821 | +descriptor. | ||
822 | + | ||
823 | +A *KVM\_DEV\_USER\_SHADOW\_SIZE* ``ioctl()`` causes the KVM driver to | ||
824 | +allocate memory for the shadow image. This memory can later be | ||
825 | +``mmap()``\ ed by the emulation process to share the emulation's view of | ||
826 | +device memory with the KVM driver. | ||
827 | + | ||
828 | +A *KVM\_DEV\_USER\_SHADOW\_CTRL* ``ioctl()`` controls access to the | ||
829 | +shadow image. It will send the KVM driver a shadow control map, which | ||
830 | +specifies which areas of the image can complete guest loads without | ||
831 | +sending the load request to the emulation program. It will also specify | ||
832 | +the size of load operations that are allowed. | ||
833 | + | ||
834 | +- poll | ||
835 | + | ||
836 | +An emulation program will use the ``poll()`` call with a *POLLIN* flag | ||
837 | +to determine if there are MMIO requests waiting to be read. It will | ||
838 | +return if the pending MMIO request queue is not empty. | ||
839 | + | ||
840 | +- mmap | ||
841 | + | ||
842 | +This call allows the emulation program to directly access the shadow | ||
843 | +image allocated by the KVM driver. As device emulation updates device | ||
844 | +memory, changes with no side-effects will be reflected in the shadow, | ||
845 | +and the KVM driver can satisfy guest loads from the shadow image without | ||
846 | +needing to wait for the emulation program. | ||
847 | + | ||
848 | +kvm\_io\_device ops | ||
849 | +^^^^^^^^^^^^^^^^^^^ | ||
850 | + | ||
851 | +Each KVM per-CPU thread can handle MMIO operation on behalf of the guest | ||
852 | +VM. KVM will use the MMIO's guest physical address to search for a | ||
853 | +matching *kvm\_io\_device* to see if the MMIO can be handled by the KVM | ||
854 | +driver instead of exiting back to QEMU. If a match is found, the | ||
855 | +corresponding callback will be invoked. | ||
856 | + | ||
857 | +- read | ||
858 | + | ||
859 | +This callback is invoked when the guest performs a load to the device. | ||
860 | +Loads with side-effects must be handled synchronously, with the KVM | ||
861 | +driver putting the QEMU thread to sleep waiting for the emulation | ||
862 | +process reply before re-starting the guest. Loads that do not have | ||
863 | +side-effects may be optimized by satisfying them from the shadow image, | ||
864 | +if there are no outstanding stores to the device by this CPU. PCI memory | ||
865 | +ordering demands that a load cannot complete before all older stores to | ||
866 | +the same device have been completed. | ||
867 | + | ||
868 | +- write | ||
869 | + | ||
870 | +Stores can be handled asynchronously unless the pending MMIO request | ||
871 | +queue is full. In this case, the QEMU thread must sleep waiting for | ||
872 | +space in the queue. Stores will increment the number of posted stores in | ||
873 | +the per-CPU scoreboard, in order to implement the PCI ordering | ||
874 | +constraint above. | ||
875 | + | ||
876 | +interrupt acceleration | ||
877 | +^^^^^^^^^^^^^^^^^^^^^^ | ||
878 | + | ||
879 | +This performance optimization would work much like a vhost user | ||
880 | +application does, where the QEMU process sets up *eventfds* that cause | ||
881 | +the device's corresponding interrupt to be triggered by the KVM driver. | ||
882 | +These irq file descriptors are sent to the emulation process at | ||
883 | +initialization, and are used when the emulation code raises a device | ||
884 | +interrupt. | ||
885 | + | ||
886 | +intx acceleration | ||
887 | +''''''''''''''''' | ||
888 | + | ||
889 | +Traditional PCI pin interrupts are level based, so, in addition to an | ||
890 | +irq file descriptor, a re-sampling file descriptor needs to be sent to | ||
891 | +the emulation program. This second file descriptor allows multiple | ||
892 | +devices sharing an irq to be notified when the interrupt has been | ||
893 | +acknowledged by the guest, so they can re-trigger the interrupt if their | ||
894 | +device has not de-asserted its interrupt. | ||
895 | + | ||
896 | +intx irq descriptor | ||
897 | + | ||
898 | + | ||
899 | +The irq descriptors are created by the proxy object | ||
900 | +``using event_notifier_init()`` to create the irq and re-sampling | ||
901 | +*eventds*, and ``kvm_vm_ioctl(KVM_IRQFD)`` to bind them to an interrupt. | ||
902 | +The interrupt route can be found with | ||
903 | +``pci_device_route_intx_to_irq()``. | ||
904 | + | ||
905 | +intx routing changes | ||
906 | + | ||
907 | + | ||
908 | +Intx routing can be changed when the guest programs the APIC the device | ||
909 | +pin is connected to. The proxy object in QEMU will use | ||
910 | +``pci_device_set_intx_routing_notifier()`` to be informed of any guest | ||
911 | +changes to the route. This handler will broadly follow the VFIO | ||
912 | +interrupt logic to change the route: de-assigning the existing irq | ||
913 | +descriptor from its route, then assigning it the new route. (see | ||
914 | +``vfio_intx_update()``) | ||
915 | + | ||
916 | +MSI/X acceleration | ||
917 | +'''''''''''''''''' | ||
918 | + | ||
919 | +MSI/X interrupts are sent as DMA transactions to the host. The interrupt | ||
920 | +data contains a vector that is programmed by the guest, A device may have | ||
921 | +multiple MSI interrupts associated with it, so multiple irq descriptors | ||
922 | +may need to be sent to the emulation program. | ||
923 | + | ||
924 | +MSI/X irq descriptor | ||
925 | + | ||
926 | + | ||
927 | +This case will also follow the VFIO example. For each MSI/X interrupt, | ||
928 | +an *eventfd* is created, a virtual interrupt is allocated by | ||
929 | +``kvm_irqchip_add_msi_route()``, and the virtual interrupt is bound to | ||
930 | +the eventfd with ``kvm_irqchip_add_irqfd_notifier()``. | ||
931 | + | ||
932 | +MSI/X config space changes | ||
933 | + | ||
934 | + | ||
935 | +The guest may dynamically update several MSI-related tables in the | ||
936 | +device's PCI config space. These include per-MSI interrupt enables and | ||
937 | +vector data. Additionally, MSIX tables exist in device memory space, not | ||
938 | +config space. Much like the BAR case above, the proxy object must look | ||
939 | +at guest config space programming to keep the MSI interrupt state | ||
940 | +consistent between QEMU and the emulation program. | ||
941 | + | ||
942 | +-------------- | ||
943 | + | ||
944 | +Disaggregated CPU emulation | ||
945 | +--------------------------- | ||
946 | + | ||
947 | +After IO services have been disaggregated, a second phase would be to | ||
948 | +separate a process to handle CPU instruction emulation from the main | ||
949 | +QEMU control function. There are no object separation points for this | ||
950 | +code, so the first task would be to create one. | ||
951 | + | ||
952 | +Host access controls | ||
953 | +-------------------- | ||
954 | + | ||
955 | +Separating QEMU relies on the host OS's access restriction mechanisms to | ||
956 | +enforce that the differing processes can only access the objects they | ||
957 | +are entitled to. There are a couple types of mechanisms usually provided | ||
958 | +by general purpose OSs. | ||
959 | + | ||
960 | +Discretionary access control | ||
961 | +~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | ||
962 | + | ||
963 | +Discretionary access control allows each user to control who can access | ||
964 | +their files. In Linux, this type of control is usually too coarse for | ||
965 | +QEMU separation, since it only provides three separate access controls: | ||
966 | +one for the same user ID, the second for users IDs with the same group | ||
967 | +ID, and the third for all other user IDs. Each device instance would | ||
968 | +need a separate user ID to provide access control, which is likely to be | ||
969 | +unwieldy for dynamically created VMs. | ||
970 | + | ||
971 | +Mandatory access control | ||
972 | +~~~~~~~~~~~~~~~~~~~~~~~~ | ||
973 | + | ||
974 | +Mandatory access control allows the OS to add an additional set of | ||
975 | +controls on top of discretionary access for the OS to control. It also | ||
976 | +adds other attributes to processes and files such as types, roles, and | ||
977 | +categories, and can establish rules for how processes and files can | ||
978 | +interact. | ||
979 | + | ||
980 | +Type enforcement | ||
981 | +^^^^^^^^^^^^^^^^ | ||
982 | + | ||
983 | +Type enforcement assigns a *type* attribute to processes and files, and | ||
984 | +allows rules to be written on what operations a process with a given | ||
985 | +type can perform on a file with a given type. QEMU separation could take | ||
986 | +advantage of type enforcement by running the emulation processes with | ||
987 | +different types, both from the main QEMU process, and from the emulation | ||
988 | +processes of different classes of devices. | ||
989 | + | ||
990 | +For example, guest disk images and disk emulation processes could have | ||
991 | +types separate from the main QEMU process and non-disk emulation | ||
992 | +processes, and the type rules could prevent processes other than disk | ||
993 | +emulation ones from accessing guest disk images. Similarly, network | ||
994 | +emulation processes can have a type separate from the main QEMU process | ||
995 | +and non-network emulation process, and only that type can access the | ||
996 | +host tun/tap device used to provide guest networking. | ||
997 | + | ||
998 | +Category enforcement | ||
999 | +^^^^^^^^^^^^^^^^^^^^ | ||
1000 | + | ||
1001 | +Category enforcement assigns a set of numbers within a given range to | ||
1002 | +the process or file. The process is granted access to the file if the | ||
1003 | +process's set is a superset of the file's set. This enforcement can be | ||
1004 | +used to separate multiple instances of devices in the same class. | ||
1005 | + | ||
1006 | +For example, if there are multiple disk devices provides to a guest, | ||
1007 | +each device emulation process could be provisioned with a separate | ||
1008 | +category. The different device emulation processes would not be able to | ||
1009 | +access each other's backing disk images. | ||
1010 | + | ||
1011 | +Alternatively, categories could be used in lieu of the type enforcement | ||
1012 | +scheme described above. In this scenario, different categories would be | ||
1013 | +used to prevent device emulation processes in different classes from | ||
1014 | +accessing resources assigned to other classes. | ||
1015 | -- | ||
1016 | 2.29.2 | ||
1017 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
1 | 2 | ||
3 | Adds documentation explaining the command-line arguments needed | ||
4 | to use multi-process. | ||
5 | |||
6 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
7 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
10 | Message-id: 49f757a84e5dd6fae14b22544897d1124c5fdbad.1611938319.git.jag.raman@oracle.com | ||
11 | |||
12 | [Move orphan docs/multi-process.rst document into docs/system/ and add | ||
13 | it to index.rst to prevent Sphinx "document isn't included in any | ||
14 | toctree" error. | ||
15 | --Stefan] | ||
16 | |||
17 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
18 | --- | ||
19 | MAINTAINERS | 1 + | ||
20 | docs/system/index.rst | 1 + | ||
21 | docs/system/multi-process.rst | 64 +++++++++++++++++++++++++++++++++++ | ||
22 | 3 files changed, 66 insertions(+) | ||
23 | create mode 100644 docs/system/multi-process.rst | ||
24 | |||
25 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
26 | index XXXXXXX..XXXXXXX 100644 | ||
27 | --- a/MAINTAINERS | ||
28 | +++ b/MAINTAINERS | ||
29 | @@ -XXX,XX +XXX,XX @@ M: Jagannathan Raman <jag.raman@oracle.com> | ||
30 | M: John G Johnson <john.g.johnson@oracle.com> | ||
31 | S: Maintained | ||
32 | F: docs/devel/multi-process.rst | ||
33 | +F: docs/system/multi-process.rst | ||
34 | |||
35 | Build and test automation | ||
36 | ------------------------- | ||
37 | diff --git a/docs/system/index.rst b/docs/system/index.rst | ||
38 | index XXXXXXX..XXXXXXX 100644 | ||
39 | --- a/docs/system/index.rst | ||
40 | +++ b/docs/system/index.rst | ||
41 | @@ -XXX,XX +XXX,XX @@ Contents: | ||
42 | pr-manager | ||
43 | targets | ||
44 | security | ||
45 | + multi-process | ||
46 | deprecated | ||
47 | removed-features | ||
48 | build-platforms | ||
49 | diff --git a/docs/system/multi-process.rst b/docs/system/multi-process.rst | ||
50 | new file mode 100644 | ||
51 | index XXXXXXX..XXXXXXX | ||
52 | --- /dev/null | ||
53 | +++ b/docs/system/multi-process.rst | ||
54 | @@ -XXX,XX +XXX,XX @@ | ||
55 | +Multi-process QEMU | ||
56 | +================== | ||
57 | + | ||
58 | +This document describes how to configure and use multi-process qemu. | ||
59 | +For the design document refer to docs/devel/qemu-multiprocess. | ||
60 | + | ||
61 | +1) Configuration | ||
62 | +---------------- | ||
63 | + | ||
64 | +multi-process is enabled by default for targets that enable KVM | ||
65 | + | ||
66 | + | ||
67 | +2) Usage | ||
68 | +-------- | ||
69 | + | ||
70 | +Multi-process QEMU requires an orchestrator to launch. | ||
71 | + | ||
72 | +Following is a description of command-line used to launch mpqemu. | ||
73 | + | ||
74 | +* Orchestrator: | ||
75 | + | ||
76 | + - The Orchestrator creates a unix socketpair | ||
77 | + | ||
78 | + - It launches the remote process and passes one of the | ||
79 | + sockets to it via command-line. | ||
80 | + | ||
81 | + - It then launches QEMU and specifies the other socket as an option | ||
82 | + to the Proxy device object | ||
83 | + | ||
84 | +* Remote Process: | ||
85 | + | ||
86 | + - QEMU can enter remote process mode by using the "remote" machine | ||
87 | + option. | ||
88 | + | ||
89 | + - The orchestrator creates a "remote-object" with details about | ||
90 | + the device and the file descriptor for the device | ||
91 | + | ||
92 | + - The remaining options are no different from how one launches QEMU with | ||
93 | + devices. | ||
94 | + | ||
95 | + - Example command-line for the remote process is as follows: | ||
96 | + | ||
97 | + /usr/bin/qemu-system-x86_64 \ | ||
98 | + -machine x-remote \ | ||
99 | + -device lsi53c895a,id=lsi0 \ | ||
100 | + -drive id=drive_image2,file=/build/ol7-nvme-test-1.qcow2 \ | ||
101 | + -device scsi-hd,id=drive2,drive=drive_image2,bus=lsi0.0,scsi-id=0 \ | ||
102 | + -object x-remote-object,id=robj1,devid=lsi1,fd=4, | ||
103 | + | ||
104 | +* QEMU: | ||
105 | + | ||
106 | + - Since parts of the RAM are shared between QEMU & remote process, a | ||
107 | + memory-backend-memfd is required to facilitate this, as follows: | ||
108 | + | ||
109 | + -object memory-backend-memfd,id=mem,size=2G | ||
110 | + | ||
111 | + - A "x-pci-proxy-dev" device is created for each of the PCI devices emulated | ||
112 | + in the remote process. A "socket" sub-option specifies the other end of | ||
113 | + unix channel created by orchestrator. The "id" sub-option must be specified | ||
114 | + and should be the same as the "id" specified for the remote PCI device | ||
115 | + | ||
116 | + - Example commandline for QEMU is as follows: | ||
117 | + | ||
118 | + -device x-pci-proxy-dev,id=lsi0,socket=3 | ||
119 | -- | ||
120 | 2.29.2 | ||
121 | diff view generated by jsdifflib |
1 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 2 | |
3 | Reviewed-by: Eric Blake <eblake@redhat.com> | 3 | Allow RAM MemoryRegion to be created from an offset in a file, instead |
4 | Message-id: 20171206144550.22295-6-stefanha@redhat.com | 4 | of allocating at offset of 0 by default. This is needed to synchronize |
5 | RAM between QEMU & remote process. | ||
6 | |||
7 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
9 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
11 | Message-id: 609996697ad8617e3b01df38accc5c208c24d74e.1611938319.git.jag.raman@oracle.com | ||
5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
6 | --- | 13 | --- |
7 | blockdev.c | 47 +++++++++++++++++++++++++++++++---------------- | 14 | include/exec/memory.h | 2 ++ |
8 | 1 file changed, 31 insertions(+), 16 deletions(-) | 15 | include/exec/ram_addr.h | 4 ++-- |
9 | 16 | include/qemu/mmap-alloc.h | 4 +++- | |
10 | diff --git a/blockdev.c b/blockdev.c | 17 | backends/hostmem-memfd.c | 2 +- |
11 | index XXXXXXX..XXXXXXX 100644 | 18 | hw/misc/ivshmem.c | 3 ++- |
12 | --- a/blockdev.c | 19 | softmmu/memory.c | 3 ++- |
13 | +++ b/blockdev.c | 20 | softmmu/physmem.c | 12 +++++++----- |
14 | @@ -XXX,XX +XXX,XX @@ struct BlkActionState { | 21 | util/mmap-alloc.c | 8 +++++--- |
15 | typedef struct InternalSnapshotState { | 22 | util/oslib-posix.c | 2 +- |
16 | BlkActionState common; | 23 | 9 files changed, 25 insertions(+), 15 deletions(-) |
17 | BlockDriverState *bs; | 24 | |
18 | - AioContext *aio_context; | 25 | diff --git a/include/exec/memory.h b/include/exec/memory.h |
19 | QEMUSnapshotInfo sn; | 26 | index XXXXXXX..XXXXXXX 100644 |
20 | bool created; | 27 | --- a/include/exec/memory.h |
21 | } InternalSnapshotState; | 28 | +++ b/include/exec/memory.h |
22 | @@ -XXX,XX +XXX,XX @@ static void internal_snapshot_prepare(BlkActionState *common, | 29 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_file(MemoryRegion *mr, |
23 | qemu_timeval tv; | 30 | * @size: size of the region. |
24 | BlockdevSnapshotInternal *internal; | 31 | * @share: %true if memory must be mmaped with the MAP_SHARED flag |
25 | InternalSnapshotState *state; | 32 | * @fd: the fd to mmap. |
26 | + AioContext *aio_context; | 33 | + * @offset: offset within the file referenced by fd |
27 | int ret1; | 34 | * @errp: pointer to Error*, to store an error if it happens. |
28 | 35 | * | |
29 | g_assert(common->action->type == | 36 | * Note that this function does not do anything to cause the data in the |
30 | @@ -XXX,XX +XXX,XX @@ static void internal_snapshot_prepare(BlkActionState *common, | 37 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_fd(MemoryRegion *mr, |
31 | return; | 38 | uint64_t size, |
32 | } | 39 | bool share, |
33 | 40 | int fd, | |
34 | - /* AioContext is released in .clean() */ | 41 | + ram_addr_t offset, |
35 | - state->aio_context = bdrv_get_aio_context(bs); | 42 | Error **errp); |
36 | - aio_context_acquire(state->aio_context); | 43 | #endif |
37 | + aio_context = bdrv_get_aio_context(bs); | 44 | |
38 | + aio_context_acquire(aio_context); | 45 | diff --git a/include/exec/ram_addr.h b/include/exec/ram_addr.h |
39 | 46 | index XXXXXXX..XXXXXXX 100644 | |
40 | state->bs = bs; | 47 | --- a/include/exec/ram_addr.h |
41 | + | 48 | +++ b/include/exec/ram_addr.h |
42 | + /* Paired with .clean() */ | 49 | @@ -XXX,XX +XXX,XX @@ RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr, |
43 | bdrv_drained_begin(bs); | 50 | uint32_t ram_flags, const char *mem_path, |
44 | 51 | bool readonly, Error **errp); | |
45 | if (bdrv_op_is_blocked(bs, BLOCK_OP_TYPE_INTERNAL_SNAPSHOT, errp)) { | 52 | RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr, |
46 | - return; | 53 | - uint32_t ram_flags, int fd, bool readonly, |
47 | + goto out; | 54 | - Error **errp); |
48 | } | 55 | + uint32_t ram_flags, int fd, off_t offset, |
49 | 56 | + bool readonly, Error **errp); | |
50 | if (bdrv_is_read_only(bs)) { | 57 | |
51 | error_setg(errp, "Device '%s' is read only", device); | 58 | RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host, |
52 | - return; | 59 | MemoryRegion *mr, Error **errp); |
53 | + goto out; | 60 | diff --git a/include/qemu/mmap-alloc.h b/include/qemu/mmap-alloc.h |
54 | } | 61 | index XXXXXXX..XXXXXXX 100644 |
55 | 62 | --- a/include/qemu/mmap-alloc.h | |
56 | if (!bdrv_can_snapshot(bs)) { | 63 | +++ b/include/qemu/mmap-alloc.h |
57 | error_setg(errp, "Block format '%s' used by device '%s' " | 64 | @@ -XXX,XX +XXX,XX @@ size_t qemu_mempath_getpagesize(const char *mem_path); |
58 | "does not support internal snapshots", | 65 | * @readonly: true for a read-only mapping, false for read/write. |
59 | bs->drv->format_name, device); | 66 | * @shared: map has RAM_SHARED flag. |
60 | - return; | 67 | * @is_pmem: map has RAM_PMEM flag. |
61 | + goto out; | 68 | + * @map_offset: map starts at offset of map_offset from the start of fd |
62 | } | 69 | * |
63 | 70 | * Return: | |
64 | if (!strlen(name)) { | 71 | * On success, return a pointer to the mapped area. |
65 | error_setg(errp, "Name is empty"); | 72 | @@ -XXX,XX +XXX,XX @@ void *qemu_ram_mmap(int fd, |
66 | - return; | 73 | size_t align, |
67 | + goto out; | 74 | bool readonly, |
68 | } | 75 | bool shared, |
69 | 76 | - bool is_pmem); | |
70 | /* check whether a snapshot with name exist */ | 77 | + bool is_pmem, |
71 | @@ -XXX,XX +XXX,XX @@ static void internal_snapshot_prepare(BlkActionState *common, | 78 | + off_t map_offset); |
72 | &local_err); | 79 | |
80 | void qemu_ram_munmap(int fd, void *ptr, size_t size); | ||
81 | |||
82 | diff --git a/backends/hostmem-memfd.c b/backends/hostmem-memfd.c | ||
83 | index XXXXXXX..XXXXXXX 100644 | ||
84 | --- a/backends/hostmem-memfd.c | ||
85 | +++ b/backends/hostmem-memfd.c | ||
86 | @@ -XXX,XX +XXX,XX @@ memfd_backend_memory_alloc(HostMemoryBackend *backend, Error **errp) | ||
87 | name = host_memory_backend_get_name(backend); | ||
88 | memory_region_init_ram_from_fd(&backend->mr, OBJECT(backend), | ||
89 | name, backend->size, | ||
90 | - backend->share, fd, errp); | ||
91 | + backend->share, fd, 0, errp); | ||
92 | g_free(name); | ||
93 | } | ||
94 | |||
95 | diff --git a/hw/misc/ivshmem.c b/hw/misc/ivshmem.c | ||
96 | index XXXXXXX..XXXXXXX 100644 | ||
97 | --- a/hw/misc/ivshmem.c | ||
98 | +++ b/hw/misc/ivshmem.c | ||
99 | @@ -XXX,XX +XXX,XX @@ static void process_msg_shmem(IVShmemState *s, int fd, Error **errp) | ||
100 | |||
101 | /* mmap the region and map into the BAR2 */ | ||
102 | memory_region_init_ram_from_fd(&s->server_bar2, OBJECT(s), | ||
103 | - "ivshmem.bar2", size, true, fd, &local_err); | ||
104 | + "ivshmem.bar2", size, true, fd, 0, | ||
105 | + &local_err); | ||
73 | if (local_err) { | 106 | if (local_err) { |
74 | error_propagate(errp, local_err); | 107 | error_propagate(errp, local_err); |
75 | - return; | 108 | return; |
76 | + goto out; | 109 | diff --git a/softmmu/memory.c b/softmmu/memory.c |
77 | } else if (ret) { | 110 | index XXXXXXX..XXXXXXX 100644 |
78 | error_setg(errp, | 111 | --- a/softmmu/memory.c |
79 | "Snapshot with name '%s' already exists on device '%s'", | 112 | +++ b/softmmu/memory.c |
80 | name, device); | 113 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_fd(MemoryRegion *mr, |
81 | - return; | 114 | uint64_t size, |
82 | + goto out; | 115 | bool share, |
116 | int fd, | ||
117 | + ram_addr_t offset, | ||
118 | Error **errp) | ||
119 | { | ||
120 | Error *err = NULL; | ||
121 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_fd(MemoryRegion *mr, | ||
122 | mr->destructor = memory_region_destructor_ram; | ||
123 | mr->ram_block = qemu_ram_alloc_from_fd(size, mr, | ||
124 | share ? RAM_SHARED : 0, | ||
125 | - fd, false, &err); | ||
126 | + fd, offset, false, &err); | ||
127 | if (err) { | ||
128 | mr->size = int128_zero(); | ||
129 | object_unparent(OBJECT(mr)); | ||
130 | diff --git a/softmmu/physmem.c b/softmmu/physmem.c | ||
131 | index XXXXXXX..XXXXXXX 100644 | ||
132 | --- a/softmmu/physmem.c | ||
133 | +++ b/softmmu/physmem.c | ||
134 | @@ -XXX,XX +XXX,XX @@ static void *file_ram_alloc(RAMBlock *block, | ||
135 | int fd, | ||
136 | bool readonly, | ||
137 | bool truncate, | ||
138 | + off_t offset, | ||
139 | Error **errp) | ||
140 | { | ||
141 | void *area; | ||
142 | @@ -XXX,XX +XXX,XX @@ static void *file_ram_alloc(RAMBlock *block, | ||
83 | } | 143 | } |
84 | 144 | ||
85 | /* 3. take the snapshot */ | 145 | area = qemu_ram_mmap(fd, memory, block->mr->align, readonly, |
86 | @@ -XXX,XX +XXX,XX @@ static void internal_snapshot_prepare(BlkActionState *common, | 146 | - block->flags & RAM_SHARED, block->flags & RAM_PMEM); |
87 | error_setg_errno(errp, -ret1, | 147 | + block->flags & RAM_SHARED, block->flags & RAM_PMEM, |
88 | "Failed to create snapshot '%s' on device '%s'", | 148 | + offset); |
89 | name, device); | 149 | if (area == MAP_FAILED) { |
90 | - return; | 150 | error_setg_errno(errp, errno, |
91 | + goto out; | 151 | "unable to map backing store for guest RAM"); |
152 | @@ -XXX,XX +XXX,XX @@ static void ram_block_add(RAMBlock *new_block, Error **errp, bool shared) | ||
153 | |||
154 | #ifdef CONFIG_POSIX | ||
155 | RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr, | ||
156 | - uint32_t ram_flags, int fd, bool readonly, | ||
157 | - Error **errp) | ||
158 | + uint32_t ram_flags, int fd, off_t offset, | ||
159 | + bool readonly, Error **errp) | ||
160 | { | ||
161 | RAMBlock *new_block; | ||
162 | Error *local_err = NULL; | ||
163 | @@ -XXX,XX +XXX,XX @@ RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr, | ||
164 | new_block->max_length = size; | ||
165 | new_block->flags = ram_flags; | ||
166 | new_block->host = file_ram_alloc(new_block, size, fd, readonly, | ||
167 | - !file_size, errp); | ||
168 | + !file_size, offset, errp); | ||
169 | if (!new_block->host) { | ||
170 | g_free(new_block); | ||
171 | return NULL; | ||
172 | @@ -XXX,XX +XXX,XX @@ RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr, | ||
173 | return NULL; | ||
92 | } | 174 | } |
93 | 175 | ||
94 | /* 4. succeed, mark a snapshot is created */ | 176 | - block = qemu_ram_alloc_from_fd(size, mr, ram_flags, fd, readonly, errp); |
95 | state->created = true; | 177 | + block = qemu_ram_alloc_from_fd(size, mr, ram_flags, fd, 0, readonly, errp); |
96 | + | 178 | if (!block) { |
97 | +out: | 179 | if (created) { |
98 | + aio_context_release(aio_context); | 180 | unlink(mem_path); |
99 | } | 181 | diff --git a/util/mmap-alloc.c b/util/mmap-alloc.c |
100 | 182 | index XXXXXXX..XXXXXXX 100644 | |
101 | static void internal_snapshot_abort(BlkActionState *common) | 183 | --- a/util/mmap-alloc.c |
102 | @@ -XXX,XX +XXX,XX @@ static void internal_snapshot_abort(BlkActionState *common) | 184 | +++ b/util/mmap-alloc.c |
103 | DO_UPCAST(InternalSnapshotState, common, common); | 185 | @@ -XXX,XX +XXX,XX @@ void *qemu_ram_mmap(int fd, |
104 | BlockDriverState *bs = state->bs; | 186 | size_t align, |
105 | QEMUSnapshotInfo *sn = &state->sn; | 187 | bool readonly, |
106 | + AioContext *aio_context; | 188 | bool shared, |
107 | Error *local_error = NULL; | 189 | - bool is_pmem) |
108 | 190 | + bool is_pmem, | |
109 | if (!state->created) { | 191 | + off_t map_offset) |
110 | return; | 192 | { |
193 | int prot; | ||
194 | int flags; | ||
195 | @@ -XXX,XX +XXX,XX @@ void *qemu_ram_mmap(int fd, | ||
196 | |||
197 | prot = PROT_READ | (readonly ? 0 : PROT_WRITE); | ||
198 | |||
199 | - ptr = mmap(guardptr + offset, size, prot, flags | map_sync_flags, fd, 0); | ||
200 | + ptr = mmap(guardptr + offset, size, prot, | ||
201 | + flags | map_sync_flags, fd, map_offset); | ||
202 | |||
203 | if (ptr == MAP_FAILED && map_sync_flags) { | ||
204 | if (errno == ENOTSUP) { | ||
205 | @@ -XXX,XX +XXX,XX @@ void *qemu_ram_mmap(int fd, | ||
206 | * if map failed with MAP_SHARED_VALIDATE | MAP_SYNC, | ||
207 | * we will remove these flags to handle compatibility. | ||
208 | */ | ||
209 | - ptr = mmap(guardptr + offset, size, prot, flags, fd, 0); | ||
210 | + ptr = mmap(guardptr + offset, size, prot, flags, fd, map_offset); | ||
111 | } | 211 | } |
112 | 212 | ||
113 | + aio_context = bdrv_get_aio_context(state->bs); | 213 | if (ptr == MAP_FAILED) { |
114 | + aio_context_acquire(aio_context); | 214 | diff --git a/util/oslib-posix.c b/util/oslib-posix.c |
115 | + | 215 | index XXXXXXX..XXXXXXX 100644 |
116 | if (bdrv_snapshot_delete(bs, sn->id_str, sn->name, &local_error) < 0) { | 216 | --- a/util/oslib-posix.c |
117 | error_reportf_err(local_error, | 217 | +++ b/util/oslib-posix.c |
118 | "Failed to delete snapshot with id '%s' and " | 218 | @@ -XXX,XX +XXX,XX @@ void *qemu_memalign(size_t alignment, size_t size) |
119 | @@ -XXX,XX +XXX,XX @@ static void internal_snapshot_abort(BlkActionState *common) | 219 | void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment, bool shared) |
120 | sn->id_str, sn->name, | 220 | { |
121 | bdrv_get_device_name(bs)); | 221 | size_t align = QEMU_VMALLOC_ALIGN; |
122 | } | 222 | - void *ptr = qemu_ram_mmap(-1, size, align, false, shared, false); |
123 | + | 223 | + void *ptr = qemu_ram_mmap(-1, size, align, false, shared, false, 0); |
124 | + aio_context_release(aio_context); | 224 | |
125 | } | 225 | if (ptr == MAP_FAILED) { |
126 | 226 | return NULL; | |
127 | static void internal_snapshot_clean(BlkActionState *common) | ||
128 | { | ||
129 | InternalSnapshotState *state = DO_UPCAST(InternalSnapshotState, | ||
130 | common, common); | ||
131 | + AioContext *aio_context; | ||
132 | |||
133 | - if (state->aio_context) { | ||
134 | - if (state->bs) { | ||
135 | - bdrv_drained_end(state->bs); | ||
136 | - } | ||
137 | - aio_context_release(state->aio_context); | ||
138 | + if (!state->bs) { | ||
139 | + return; | ||
140 | } | ||
141 | + | ||
142 | + aio_context = bdrv_get_aio_context(state->bs); | ||
143 | + aio_context_acquire(aio_context); | ||
144 | + | ||
145 | + bdrv_drained_end(state->bs); | ||
146 | + | ||
147 | + aio_context_release(aio_context); | ||
148 | } | ||
149 | |||
150 | /* external snapshot private data */ | ||
151 | -- | 227 | -- |
152 | 2.14.3 | 228 | 2.29.2 |
153 | 229 | ||
154 | diff view generated by jsdifflib |
New patch | |||
---|---|---|---|
1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
1 | 2 | ||
3 | Add configuration options to enable or disable multiprocess QEMU code | ||
4 | |||
5 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
6 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
7 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
8 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
9 | Message-id: 6cc37253e35418ebd7b675a31a3df6e3c7a12dc1.1611938319.git.jag.raman@oracle.com | ||
10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
11 | --- | ||
12 | configure | 10 ++++++++++ | ||
13 | meson.build | 4 +++- | ||
14 | Kconfig.host | 4 ++++ | ||
15 | hw/Kconfig | 1 + | ||
16 | hw/remote/Kconfig | 3 +++ | ||
17 | 5 files changed, 21 insertions(+), 1 deletion(-) | ||
18 | create mode 100644 hw/remote/Kconfig | ||
19 | |||
20 | diff --git a/configure b/configure | ||
21 | index XXXXXXX..XXXXXXX 100755 | ||
22 | --- a/configure | ||
23 | +++ b/configure | ||
24 | @@ -XXX,XX +XXX,XX @@ skip_meson=no | ||
25 | gettext="auto" | ||
26 | fuse="auto" | ||
27 | fuse_lseek="auto" | ||
28 | +multiprocess="no" | ||
29 | |||
30 | malloc_trim="auto" | ||
31 | |||
32 | @@ -XXX,XX +XXX,XX @@ Linux) | ||
33 | linux="yes" | ||
34 | linux_user="yes" | ||
35 | vhost_user=${default_feature:-yes} | ||
36 | + multiprocess=${default_feature:-yes} | ||
37 | ;; | ||
38 | esac | ||
39 | |||
40 | @@ -XXX,XX +XXX,XX @@ for opt do | ||
41 | ;; | ||
42 | --disable-fuse-lseek) fuse_lseek="disabled" | ||
43 | ;; | ||
44 | + --enable-multiprocess) multiprocess="yes" | ||
45 | + ;; | ||
46 | + --disable-multiprocess) multiprocess="no" | ||
47 | + ;; | ||
48 | *) | ||
49 | echo "ERROR: unknown option $opt" | ||
50 | echo "Try '$0 --help' for more information" | ||
51 | @@ -XXX,XX +XXX,XX @@ disabled with --disable-FEATURE, default is enabled if available | ||
52 | libdaxctl libdaxctl support | ||
53 | fuse FUSE block device export | ||
54 | fuse-lseek SEEK_HOLE/SEEK_DATA support for FUSE exports | ||
55 | + multiprocess Multiprocess QEMU support | ||
56 | |||
57 | NOTE: The object files are built at the place where configure is launched | ||
58 | EOF | ||
59 | @@ -XXX,XX +XXX,XX @@ fi | ||
60 | if test "$have_mlockall" = "yes" ; then | ||
61 | echo "HAVE_MLOCKALL=y" >> $config_host_mak | ||
62 | fi | ||
63 | +if test "$multiprocess" = "yes" ; then | ||
64 | + echo "CONFIG_MULTIPROCESS_ALLOWED=y" >> $config_host_mak | ||
65 | +fi | ||
66 | if test "$fuzzing" = "yes" ; then | ||
67 | # If LIB_FUZZING_ENGINE is set, assume we are running on OSS-Fuzz, and the | ||
68 | # needed CFLAGS have already been provided | ||
69 | diff --git a/meson.build b/meson.build | ||
70 | index XXXXXXX..XXXXXXX 100644 | ||
71 | --- a/meson.build | ||
72 | +++ b/meson.build | ||
73 | @@ -XXX,XX +XXX,XX @@ host_kconfig = \ | ||
74 | ('CONFIG_VHOST_KERNEL' in config_host ? ['CONFIG_VHOST_KERNEL=y'] : []) + \ | ||
75 | (have_virtfs ? ['CONFIG_VIRTFS=y'] : []) + \ | ||
76 | ('CONFIG_LINUX' in config_host ? ['CONFIG_LINUX=y'] : []) + \ | ||
77 | - ('CONFIG_PVRDMA' in config_host ? ['CONFIG_PVRDMA=y'] : []) | ||
78 | + ('CONFIG_PVRDMA' in config_host ? ['CONFIG_PVRDMA=y'] : []) + \ | ||
79 | + ('CONFIG_MULTIPROCESS_ALLOWED' in config_host ? ['CONFIG_MULTIPROCESS_ALLOWED=y'] : []) | ||
80 | |||
81 | ignored = [ 'TARGET_XML_FILES', 'TARGET_ABI_DIR', 'TARGET_ARCH' ] | ||
82 | |||
83 | @@ -XXX,XX +XXX,XX @@ summary_info += {'libpmem support': config_host.has_key('CONFIG_LIBPMEM')} | ||
84 | summary_info += {'libdaxctl support': config_host.has_key('CONFIG_LIBDAXCTL')} | ||
85 | summary_info += {'libudev': libudev.found()} | ||
86 | summary_info += {'FUSE lseek': fuse_lseek.found()} | ||
87 | +summary_info += {'Multiprocess QEMU': config_host.has_key('CONFIG_MULTIPROCESS_ALLOWED')} | ||
88 | summary(summary_info, bool_yn: true, section: 'Dependencies') | ||
89 | |||
90 | if not supported_cpus.contains(cpu) | ||
91 | diff --git a/Kconfig.host b/Kconfig.host | ||
92 | index XXXXXXX..XXXXXXX 100644 | ||
93 | --- a/Kconfig.host | ||
94 | +++ b/Kconfig.host | ||
95 | @@ -XXX,XX +XXX,XX @@ config VIRTFS | ||
96 | |||
97 | config PVRDMA | ||
98 | bool | ||
99 | + | ||
100 | +config MULTIPROCESS_ALLOWED | ||
101 | + bool | ||
102 | + imply MULTIPROCESS | ||
103 | diff --git a/hw/Kconfig b/hw/Kconfig | ||
104 | index XXXXXXX..XXXXXXX 100644 | ||
105 | --- a/hw/Kconfig | ||
106 | +++ b/hw/Kconfig | ||
107 | @@ -XXX,XX +XXX,XX @@ source pci-host/Kconfig | ||
108 | source pcmcia/Kconfig | ||
109 | source pci/Kconfig | ||
110 | source rdma/Kconfig | ||
111 | +source remote/Kconfig | ||
112 | source rtc/Kconfig | ||
113 | source scsi/Kconfig | ||
114 | source sd/Kconfig | ||
115 | diff --git a/hw/remote/Kconfig b/hw/remote/Kconfig | ||
116 | new file mode 100644 | ||
117 | index XXXXXXX..XXXXXXX | ||
118 | --- /dev/null | ||
119 | +++ b/hw/remote/Kconfig | ||
120 | @@ -XXX,XX +XXX,XX @@ | ||
121 | +config MULTIPROCESS | ||
122 | + bool | ||
123 | + depends on PCI && KVM | ||
124 | -- | ||
125 | 2.29.2 | ||
126 | diff view generated by jsdifflib |
1 | This test case will prevent future regressions with savevm and | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | IOThreads. | ||
3 | 2 | ||
4 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 3 | PCI host bridge is setup for the remote device process. It is |
5 | Reviewed-by: Eric Blake <eblake@redhat.com> | 4 | implemented using remote-pcihost object. It is an extension of the PCI |
6 | Message-id: 20171207201320.19284-7-stefanha@redhat.com | 5 | host bridge setup by QEMU. |
6 | Remote-pcihost configures a PCI bus which could be used by the remote | ||
7 | PCI device to latch on to. | ||
8 | |||
9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
10 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
11 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
12 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
13 | Message-id: 0871ba857abb2eafacde07e7fe66a3f12415bfb2.1611938319.git.jag.raman@oracle.com | ||
7 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
8 | --- | 15 | --- |
9 | tests/qemu-iotests/203 | 59 ++++++++++++++++++++++++++++++++++++++++++++++ | 16 | MAINTAINERS | 2 + |
10 | tests/qemu-iotests/203.out | 6 +++++ | 17 | include/hw/pci-host/remote.h | 29 ++++++++++++++ |
11 | tests/qemu-iotests/group | 1 + | 18 | hw/pci-host/remote.c | 75 ++++++++++++++++++++++++++++++++++++ |
12 | 3 files changed, 66 insertions(+) | 19 | hw/pci-host/Kconfig | 3 ++ |
13 | create mode 100755 tests/qemu-iotests/203 | 20 | hw/pci-host/meson.build | 1 + |
14 | create mode 100644 tests/qemu-iotests/203.out | 21 | hw/remote/Kconfig | 1 + |
22 | 6 files changed, 111 insertions(+) | ||
23 | create mode 100644 include/hw/pci-host/remote.h | ||
24 | create mode 100644 hw/pci-host/remote.c | ||
15 | 25 | ||
16 | diff --git a/tests/qemu-iotests/203 b/tests/qemu-iotests/203 | 26 | diff --git a/MAINTAINERS b/MAINTAINERS |
17 | new file mode 100755 | 27 | index XXXXXXX..XXXXXXX 100644 |
18 | index XXXXXXX..XXXXXXX | 28 | --- a/MAINTAINERS |
19 | --- /dev/null | 29 | +++ b/MAINTAINERS |
20 | +++ b/tests/qemu-iotests/203 | 30 | @@ -XXX,XX +XXX,XX @@ M: John G Johnson <john.g.johnson@oracle.com> |
21 | @@ -XXX,XX +XXX,XX @@ | 31 | S: Maintained |
22 | +#!/usr/bin/env python | 32 | F: docs/devel/multi-process.rst |
23 | +# | 33 | F: docs/system/multi-process.rst |
24 | +# Copyright (C) 2017 Red Hat, Inc. | 34 | +F: hw/pci-host/remote.c |
25 | +# | 35 | +F: include/hw/pci-host/remote.h |
26 | +# This program is free software; you can redistribute it and/or modify | 36 | |
27 | +# it under the terms of the GNU General Public License as published by | 37 | Build and test automation |
28 | +# the Free Software Foundation; either version 2 of the License, or | 38 | ------------------------- |
29 | +# (at your option) any later version. | 39 | diff --git a/include/hw/pci-host/remote.h b/include/hw/pci-host/remote.h |
30 | +# | ||
31 | +# This program is distributed in the hope that it will be useful, | ||
32 | +# but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
33 | +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
34 | +# GNU General Public License for more details. | ||
35 | +# | ||
36 | +# You should have received a copy of the GNU General Public License | ||
37 | +# along with this program. If not, see <http://www.gnu.org/licenses/>. | ||
38 | +# | ||
39 | +# Creator/Owner: Stefan Hajnoczi <stefanha@redhat.com> | ||
40 | +# | ||
41 | +# Check that QMP 'migrate' with multiple drives on a single IOThread completes | ||
42 | +# successfully. This particular command triggered a hang in the source QEMU | ||
43 | +# process due to recursive AioContext locking in bdrv_invalidate_all() and | ||
44 | +# BDRV_POLL_WHILE(). | ||
45 | + | ||
46 | +import iotests | ||
47 | + | ||
48 | +iotests.verify_image_format(supported_fmts=['qcow2']) | ||
49 | +iotests.verify_platform(['linux']) | ||
50 | + | ||
51 | +with iotests.FilePath('disk0.img') as disk0_img_path, \ | ||
52 | + iotests.FilePath('disk1.img') as disk1_img_path, \ | ||
53 | + iotests.VM() as vm: | ||
54 | + | ||
55 | + img_size = '10M' | ||
56 | + iotests.qemu_img_pipe('create', '-f', iotests.imgfmt, disk0_img_path, img_size) | ||
57 | + iotests.qemu_img_pipe('create', '-f', iotests.imgfmt, disk1_img_path, img_size) | ||
58 | + | ||
59 | + iotests.log('Launching VM...') | ||
60 | + (vm.add_object('iothread,id=iothread0') | ||
61 | + .add_drive(disk0_img_path, 'node-name=drive0-node', interface='none') | ||
62 | + .add_drive(disk1_img_path, 'node-name=drive1-node', interface='none') | ||
63 | + .launch()) | ||
64 | + | ||
65 | + iotests.log('Setting IOThreads...') | ||
66 | + iotests.log(vm.qmp('x-blockdev-set-iothread', | ||
67 | + node_name='drive0-node', iothread='iothread0', | ||
68 | + force=True)) | ||
69 | + iotests.log(vm.qmp('x-blockdev-set-iothread', | ||
70 | + node_name='drive1-node', iothread='iothread0', | ||
71 | + force=True)) | ||
72 | + | ||
73 | + iotests.log('Starting migration...') | ||
74 | + iotests.log(vm.qmp('migrate', uri='exec:cat >/dev/null')) | ||
75 | + while True: | ||
76 | + vm.get_qmp_event(wait=60.0) | ||
77 | + result = vm.qmp('query-migrate') | ||
78 | + status = result.get('return', {}).get('status', None) | ||
79 | + if status == 'completed': | ||
80 | + break | ||
81 | diff --git a/tests/qemu-iotests/203.out b/tests/qemu-iotests/203.out | ||
82 | new file mode 100644 | 40 | new file mode 100644 |
83 | index XXXXXXX..XXXXXXX | 41 | index XXXXXXX..XXXXXXX |
84 | --- /dev/null | 42 | --- /dev/null |
85 | +++ b/tests/qemu-iotests/203.out | 43 | +++ b/include/hw/pci-host/remote.h |
86 | @@ -XXX,XX +XXX,XX @@ | 44 | @@ -XXX,XX +XXX,XX @@ |
87 | +Launching VM... | 45 | +/* |
88 | +Setting IOThreads... | 46 | + * PCI Host for remote device |
89 | +{u'return': {}} | 47 | + * |
90 | +{u'return': {}} | 48 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. |
91 | +Starting migration... | 49 | + * |
92 | +{u'return': {}} | 50 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. |
93 | diff --git a/tests/qemu-iotests/group b/tests/qemu-iotests/group | 51 | + * See the COPYING file in the top-level directory. |
52 | + * | ||
53 | + */ | ||
54 | + | ||
55 | +#ifndef REMOTE_PCIHOST_H | ||
56 | +#define REMOTE_PCIHOST_H | ||
57 | + | ||
58 | +#include "exec/memory.h" | ||
59 | +#include "hw/pci/pcie_host.h" | ||
60 | + | ||
61 | +#define TYPE_REMOTE_PCIHOST "remote-pcihost" | ||
62 | +OBJECT_DECLARE_SIMPLE_TYPE(RemotePCIHost, REMOTE_PCIHOST) | ||
63 | + | ||
64 | +struct RemotePCIHost { | ||
65 | + /*< private >*/ | ||
66 | + PCIExpressHost parent_obj; | ||
67 | + /*< public >*/ | ||
68 | + | ||
69 | + MemoryRegion *mr_pci_mem; | ||
70 | + MemoryRegion *mr_sys_io; | ||
71 | +}; | ||
72 | + | ||
73 | +#endif | ||
74 | diff --git a/hw/pci-host/remote.c b/hw/pci-host/remote.c | ||
75 | new file mode 100644 | ||
76 | index XXXXXXX..XXXXXXX | ||
77 | --- /dev/null | ||
78 | +++ b/hw/pci-host/remote.c | ||
79 | @@ -XXX,XX +XXX,XX @@ | ||
80 | +/* | ||
81 | + * Remote PCI host device | ||
82 | + * | ||
83 | + * Unlike PCI host devices that model physical hardware, the purpose | ||
84 | + * of this PCI host is to host multi-process QEMU devices. | ||
85 | + * | ||
86 | + * Multi-process QEMU extends the PCI host of a QEMU machine into a | ||
87 | + * remote process. Any PCI device attached to the remote process is | ||
88 | + * visible in the QEMU guest. This allows existing QEMU device models | ||
89 | + * to be reused in the remote process. | ||
90 | + * | ||
91 | + * This PCI host is purely a container for PCI devices. It's fake in the | ||
92 | + * sense that the guest never sees this PCI host and has no way of | ||
93 | + * accessing it. Its job is just to provide the environment that QEMU | ||
94 | + * PCI device models need when running in a remote process. | ||
95 | + * | ||
96 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
97 | + * | ||
98 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
99 | + * See the COPYING file in the top-level directory. | ||
100 | + * | ||
101 | + */ | ||
102 | + | ||
103 | +#include "qemu/osdep.h" | ||
104 | +#include "qemu-common.h" | ||
105 | + | ||
106 | +#include "hw/pci/pci.h" | ||
107 | +#include "hw/pci/pci_host.h" | ||
108 | +#include "hw/pci/pcie_host.h" | ||
109 | +#include "hw/qdev-properties.h" | ||
110 | +#include "hw/pci-host/remote.h" | ||
111 | +#include "exec/memory.h" | ||
112 | + | ||
113 | +static const char *remote_pcihost_root_bus_path(PCIHostState *host_bridge, | ||
114 | + PCIBus *rootbus) | ||
115 | +{ | ||
116 | + return "0000:00"; | ||
117 | +} | ||
118 | + | ||
119 | +static void remote_pcihost_realize(DeviceState *dev, Error **errp) | ||
120 | +{ | ||
121 | + PCIHostState *pci = PCI_HOST_BRIDGE(dev); | ||
122 | + RemotePCIHost *s = REMOTE_PCIHOST(dev); | ||
123 | + | ||
124 | + pci->bus = pci_root_bus_new(DEVICE(s), "remote-pci", | ||
125 | + s->mr_pci_mem, s->mr_sys_io, | ||
126 | + 0, TYPE_PCIE_BUS); | ||
127 | +} | ||
128 | + | ||
129 | +static void remote_pcihost_class_init(ObjectClass *klass, void *data) | ||
130 | +{ | ||
131 | + DeviceClass *dc = DEVICE_CLASS(klass); | ||
132 | + PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass); | ||
133 | + | ||
134 | + hc->root_bus_path = remote_pcihost_root_bus_path; | ||
135 | + dc->realize = remote_pcihost_realize; | ||
136 | + | ||
137 | + dc->user_creatable = false; | ||
138 | + set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); | ||
139 | + dc->fw_name = "pci"; | ||
140 | +} | ||
141 | + | ||
142 | +static const TypeInfo remote_pcihost_info = { | ||
143 | + .name = TYPE_REMOTE_PCIHOST, | ||
144 | + .parent = TYPE_PCIE_HOST_BRIDGE, | ||
145 | + .instance_size = sizeof(RemotePCIHost), | ||
146 | + .class_init = remote_pcihost_class_init, | ||
147 | +}; | ||
148 | + | ||
149 | +static void remote_pcihost_register(void) | ||
150 | +{ | ||
151 | + type_register_static(&remote_pcihost_info); | ||
152 | +} | ||
153 | + | ||
154 | +type_init(remote_pcihost_register) | ||
155 | diff --git a/hw/pci-host/Kconfig b/hw/pci-host/Kconfig | ||
94 | index XXXXXXX..XXXXXXX 100644 | 156 | index XXXXXXX..XXXXXXX 100644 |
95 | --- a/tests/qemu-iotests/group | 157 | --- a/hw/pci-host/Kconfig |
96 | +++ b/tests/qemu-iotests/group | 158 | +++ b/hw/pci-host/Kconfig |
159 | @@ -XXX,XX +XXX,XX @@ config PCI_POWERNV | ||
160 | select PCI_EXPRESS | ||
161 | select MSI_NONBROKEN | ||
162 | select PCIE_PORT | ||
163 | + | ||
164 | +config REMOTE_PCIHOST | ||
165 | + bool | ||
166 | diff --git a/hw/pci-host/meson.build b/hw/pci-host/meson.build | ||
167 | index XXXXXXX..XXXXXXX 100644 | ||
168 | --- a/hw/pci-host/meson.build | ||
169 | +++ b/hw/pci-host/meson.build | ||
170 | @@ -XXX,XX +XXX,XX @@ pci_ss.add(when: 'CONFIG_PCI_EXPRESS_XILINX', if_true: files('xilinx-pcie.c')) | ||
171 | pci_ss.add(when: 'CONFIG_PCI_I440FX', if_true: files('i440fx.c')) | ||
172 | pci_ss.add(when: 'CONFIG_PCI_SABRE', if_true: files('sabre.c')) | ||
173 | pci_ss.add(when: 'CONFIG_XEN_IGD_PASSTHROUGH', if_true: files('xen_igd_pt.c')) | ||
174 | +pci_ss.add(when: 'CONFIG_REMOTE_PCIHOST', if_true: files('remote.c')) | ||
175 | |||
176 | # PPC devices | ||
177 | pci_ss.add(when: 'CONFIG_PREP_PCI', if_true: files('prep.c')) | ||
178 | diff --git a/hw/remote/Kconfig b/hw/remote/Kconfig | ||
179 | index XXXXXXX..XXXXXXX 100644 | ||
180 | --- a/hw/remote/Kconfig | ||
181 | +++ b/hw/remote/Kconfig | ||
97 | @@ -XXX,XX +XXX,XX @@ | 182 | @@ -XXX,XX +XXX,XX @@ |
98 | 198 rw auto | 183 | config MULTIPROCESS |
99 | 200 rw auto | 184 | bool |
100 | 202 rw auto quick | 185 | depends on PCI && KVM |
101 | +203 rw auto | 186 | + select REMOTE_PCIHOST |
102 | -- | 187 | -- |
103 | 2.14.3 | 188 | 2.29.2 |
104 | 189 | ||
105 | diff view generated by jsdifflib |
1 | QMP 'transaction' blockdev-snapshot-sync with multiple disks in an | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | IOThread is an untested code path. Several bugs have been found in | ||
3 | connection with this command. This patch adds a test case to prevent | ||
4 | future regressions. | ||
5 | 2 | ||
6 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 3 | x-remote-machine object sets up various subsystems of the remote |
7 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 4 | device process. Instantiate PCI host bridge object and initialize RAM, IO & |
8 | Reviewed-by: Eric Blake <eblake@redhat.com> | 5 | PCI memory regions. |
9 | Message-id: 20171206144550.22295-10-stefanha@redhat.com | 6 | |
7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
8 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
9 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
11 | Message-id: c537f38d17f90453ca610c6b70cf3480274e0ba1.1611938319.git.jag.raman@oracle.com | ||
10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
11 | --- | 13 | --- |
12 | tests/qemu-iotests/202 | 95 ++++++++++++++++++++++++++++++++++++++++++++++ | 14 | MAINTAINERS | 2 ++ |
13 | tests/qemu-iotests/202.out | 11 ++++++ | 15 | include/hw/pci-host/remote.h | 1 + |
14 | tests/qemu-iotests/group | 1 + | 16 | include/hw/remote/machine.h | 27 ++++++++++++++ |
15 | 3 files changed, 107 insertions(+) | 17 | hw/remote/machine.c | 70 ++++++++++++++++++++++++++++++++++++ |
16 | create mode 100755 tests/qemu-iotests/202 | 18 | hw/meson.build | 1 + |
17 | create mode 100644 tests/qemu-iotests/202.out | 19 | hw/remote/meson.build | 5 +++ |
20 | 6 files changed, 106 insertions(+) | ||
21 | create mode 100644 include/hw/remote/machine.h | ||
22 | create mode 100644 hw/remote/machine.c | ||
23 | create mode 100644 hw/remote/meson.build | ||
18 | 24 | ||
19 | diff --git a/tests/qemu-iotests/202 b/tests/qemu-iotests/202 | 25 | diff --git a/MAINTAINERS b/MAINTAINERS |
20 | new file mode 100755 | 26 | index XXXXXXX..XXXXXXX 100644 |
21 | index XXXXXXX..XXXXXXX | 27 | --- a/MAINTAINERS |
22 | --- /dev/null | 28 | +++ b/MAINTAINERS |
23 | +++ b/tests/qemu-iotests/202 | 29 | @@ -XXX,XX +XXX,XX @@ F: docs/devel/multi-process.rst |
24 | @@ -XXX,XX +XXX,XX @@ | 30 | F: docs/system/multi-process.rst |
25 | +#!/usr/bin/env python | 31 | F: hw/pci-host/remote.c |
26 | +# | 32 | F: include/hw/pci-host/remote.h |
27 | +# Copyright (C) 2017 Red Hat, Inc. | 33 | +F: hw/remote/machine.c |
28 | +# | 34 | +F: include/hw/remote/machine.h |
29 | +# This program is free software; you can redistribute it and/or modify | 35 | |
30 | +# it under the terms of the GNU General Public License as published by | 36 | Build and test automation |
31 | +# the Free Software Foundation; either version 2 of the License, or | 37 | ------------------------- |
32 | +# (at your option) any later version. | 38 | diff --git a/include/hw/pci-host/remote.h b/include/hw/pci-host/remote.h |
33 | +# | 39 | index XXXXXXX..XXXXXXX 100644 |
34 | +# This program is distributed in the hope that it will be useful, | 40 | --- a/include/hw/pci-host/remote.h |
35 | +# but WITHOUT ANY WARRANTY; without even the implied warranty of | 41 | +++ b/include/hw/pci-host/remote.h |
36 | +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | 42 | @@ -XXX,XX +XXX,XX @@ struct RemotePCIHost { |
37 | +# GNU General Public License for more details. | 43 | |
38 | +# | 44 | MemoryRegion *mr_pci_mem; |
39 | +# You should have received a copy of the GNU General Public License | 45 | MemoryRegion *mr_sys_io; |
40 | +# along with this program. If not, see <http://www.gnu.org/licenses/>. | 46 | + MemoryRegion *mr_sys_mem; |
41 | +# | 47 | }; |
42 | +# Creator/Owner: Stefan Hajnoczi <stefanha@redhat.com> | 48 | |
43 | +# | 49 | #endif |
44 | +# Check that QMP 'transaction' blockdev-snapshot-sync with multiple drives on a | 50 | diff --git a/include/hw/remote/machine.h b/include/hw/remote/machine.h |
45 | +# single IOThread completes successfully. This particular command triggered a | ||
46 | +# hang due to recursive AioContext locking and BDRV_POLL_WHILE(). Protect | ||
47 | +# against regressions. | ||
48 | + | ||
49 | +import iotests | ||
50 | + | ||
51 | +iotests.verify_image_format(supported_fmts=['qcow2']) | ||
52 | +iotests.verify_platform(['linux']) | ||
53 | + | ||
54 | +with iotests.FilePath('disk0.img') as disk0_img_path, \ | ||
55 | + iotests.FilePath('disk1.img') as disk1_img_path, \ | ||
56 | + iotests.FilePath('disk0-snap.img') as disk0_snap_img_path, \ | ||
57 | + iotests.FilePath('disk1-snap.img') as disk1_snap_img_path, \ | ||
58 | + iotests.VM() as vm: | ||
59 | + | ||
60 | + img_size = '10M' | ||
61 | + iotests.qemu_img_pipe('create', '-f', iotests.imgfmt, disk0_img_path, img_size) | ||
62 | + iotests.qemu_img_pipe('create', '-f', iotests.imgfmt, disk1_img_path, img_size) | ||
63 | + | ||
64 | + iotests.log('Launching VM...') | ||
65 | + vm.launch() | ||
66 | + | ||
67 | + iotests.log('Adding IOThread...') | ||
68 | + iotests.log(vm.qmp('object-add', | ||
69 | + qom_type='iothread', | ||
70 | + id='iothread0')) | ||
71 | + | ||
72 | + iotests.log('Adding blockdevs...') | ||
73 | + iotests.log(vm.qmp('blockdev-add', | ||
74 | + driver=iotests.imgfmt, | ||
75 | + node_name='disk0', | ||
76 | + file={ | ||
77 | + 'driver': 'file', | ||
78 | + 'filename': disk0_img_path, | ||
79 | + })) | ||
80 | + iotests.log(vm.qmp('blockdev-add', | ||
81 | + driver=iotests.imgfmt, | ||
82 | + node_name='disk1', | ||
83 | + file={ | ||
84 | + 'driver': 'file', | ||
85 | + 'filename': disk1_img_path, | ||
86 | + })) | ||
87 | + | ||
88 | + iotests.log('Setting iothread...') | ||
89 | + iotests.log(vm.qmp('x-blockdev-set-iothread', | ||
90 | + node_name='disk0', | ||
91 | + iothread='iothread0')) | ||
92 | + iotests.log(vm.qmp('x-blockdev-set-iothread', | ||
93 | + node_name='disk1', | ||
94 | + iothread='iothread0')) | ||
95 | + | ||
96 | + iotests.log('Creating external snapshots...') | ||
97 | + iotests.log(vm.qmp( | ||
98 | + 'transaction', | ||
99 | + actions=[ | ||
100 | + { | ||
101 | + 'data': { | ||
102 | + 'node-name': 'disk0', | ||
103 | + 'snapshot-file': disk0_snap_img_path, | ||
104 | + 'snapshot-node-name': 'disk0-snap', | ||
105 | + 'mode': 'absolute-paths', | ||
106 | + 'format': iotests.imgfmt, | ||
107 | + }, | ||
108 | + 'type': 'blockdev-snapshot-sync' | ||
109 | + }, { | ||
110 | + 'data': { | ||
111 | + 'node-name': 'disk1', | ||
112 | + 'snapshot-file': disk1_snap_img_path, | ||
113 | + 'snapshot-node-name': 'disk1-snap', | ||
114 | + 'mode': 'absolute-paths', | ||
115 | + 'format': iotests.imgfmt | ||
116 | + }, | ||
117 | + 'type': 'blockdev-snapshot-sync' | ||
118 | + } | ||
119 | + ])) | ||
120 | diff --git a/tests/qemu-iotests/202.out b/tests/qemu-iotests/202.out | ||
121 | new file mode 100644 | 51 | new file mode 100644 |
122 | index XXXXXXX..XXXXXXX | 52 | index XXXXXXX..XXXXXXX |
123 | --- /dev/null | 53 | --- /dev/null |
124 | +++ b/tests/qemu-iotests/202.out | 54 | +++ b/include/hw/remote/machine.h |
125 | @@ -XXX,XX +XXX,XX @@ | 55 | @@ -XXX,XX +XXX,XX @@ |
126 | +Launching VM... | 56 | +/* |
127 | +Adding IOThread... | 57 | + * Remote machine configuration |
128 | +{u'return': {}} | 58 | + * |
129 | +Adding blockdevs... | 59 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. |
130 | +{u'return': {}} | 60 | + * |
131 | +{u'return': {}} | 61 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. |
132 | +Setting iothread... | 62 | + * See the COPYING file in the top-level directory. |
133 | +{u'return': {}} | 63 | + * |
134 | +{u'return': {}} | 64 | + */ |
135 | +Creating external snapshots... | 65 | + |
136 | +{u'return': {}} | 66 | +#ifndef REMOTE_MACHINE_H |
137 | diff --git a/tests/qemu-iotests/group b/tests/qemu-iotests/group | 67 | +#define REMOTE_MACHINE_H |
68 | + | ||
69 | +#include "qom/object.h" | ||
70 | +#include "hw/boards.h" | ||
71 | +#include "hw/pci-host/remote.h" | ||
72 | + | ||
73 | +struct RemoteMachineState { | ||
74 | + MachineState parent_obj; | ||
75 | + | ||
76 | + RemotePCIHost *host; | ||
77 | +}; | ||
78 | + | ||
79 | +#define TYPE_REMOTE_MACHINE "x-remote-machine" | ||
80 | +OBJECT_DECLARE_SIMPLE_TYPE(RemoteMachineState, REMOTE_MACHINE) | ||
81 | + | ||
82 | +#endif | ||
83 | diff --git a/hw/remote/machine.c b/hw/remote/machine.c | ||
84 | new file mode 100644 | ||
85 | index XXXXXXX..XXXXXXX | ||
86 | --- /dev/null | ||
87 | +++ b/hw/remote/machine.c | ||
88 | @@ -XXX,XX +XXX,XX @@ | ||
89 | +/* | ||
90 | + * Machine for remote device | ||
91 | + * | ||
92 | + * This machine type is used by the remote device process in multi-process | ||
93 | + * QEMU. QEMU device models depend on parent busses, interrupt controllers, | ||
94 | + * memory regions, etc. The remote machine type offers this environment so | ||
95 | + * that QEMU device models can be used as remote devices. | ||
96 | + * | ||
97 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
98 | + * | ||
99 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
100 | + * See the COPYING file in the top-level directory. | ||
101 | + * | ||
102 | + */ | ||
103 | + | ||
104 | +#include "qemu/osdep.h" | ||
105 | +#include "qemu-common.h" | ||
106 | + | ||
107 | +#include "hw/remote/machine.h" | ||
108 | +#include "exec/address-spaces.h" | ||
109 | +#include "exec/memory.h" | ||
110 | +#include "qapi/error.h" | ||
111 | + | ||
112 | +static void remote_machine_init(MachineState *machine) | ||
113 | +{ | ||
114 | + MemoryRegion *system_memory, *system_io, *pci_memory; | ||
115 | + RemoteMachineState *s = REMOTE_MACHINE(machine); | ||
116 | + RemotePCIHost *rem_host; | ||
117 | + | ||
118 | + system_memory = get_system_memory(); | ||
119 | + system_io = get_system_io(); | ||
120 | + | ||
121 | + pci_memory = g_new(MemoryRegion, 1); | ||
122 | + memory_region_init(pci_memory, NULL, "pci", UINT64_MAX); | ||
123 | + | ||
124 | + rem_host = REMOTE_PCIHOST(qdev_new(TYPE_REMOTE_PCIHOST)); | ||
125 | + | ||
126 | + rem_host->mr_pci_mem = pci_memory; | ||
127 | + rem_host->mr_sys_mem = system_memory; | ||
128 | + rem_host->mr_sys_io = system_io; | ||
129 | + | ||
130 | + s->host = rem_host; | ||
131 | + | ||
132 | + object_property_add_child(OBJECT(s), "remote-pcihost", OBJECT(rem_host)); | ||
133 | + memory_region_add_subregion_overlap(system_memory, 0x0, pci_memory, -1); | ||
134 | + | ||
135 | + qdev_realize(DEVICE(rem_host), sysbus_get_default(), &error_fatal); | ||
136 | +} | ||
137 | + | ||
138 | +static void remote_machine_class_init(ObjectClass *oc, void *data) | ||
139 | +{ | ||
140 | + MachineClass *mc = MACHINE_CLASS(oc); | ||
141 | + | ||
142 | + mc->init = remote_machine_init; | ||
143 | + mc->desc = "Experimental remote machine"; | ||
144 | +} | ||
145 | + | ||
146 | +static const TypeInfo remote_machine = { | ||
147 | + .name = TYPE_REMOTE_MACHINE, | ||
148 | + .parent = TYPE_MACHINE, | ||
149 | + .instance_size = sizeof(RemoteMachineState), | ||
150 | + .class_init = remote_machine_class_init, | ||
151 | +}; | ||
152 | + | ||
153 | +static void remote_machine_register_types(void) | ||
154 | +{ | ||
155 | + type_register_static(&remote_machine); | ||
156 | +} | ||
157 | + | ||
158 | +type_init(remote_machine_register_types); | ||
159 | diff --git a/hw/meson.build b/hw/meson.build | ||
138 | index XXXXXXX..XXXXXXX 100644 | 160 | index XXXXXXX..XXXXXXX 100644 |
139 | --- a/tests/qemu-iotests/group | 161 | --- a/hw/meson.build |
140 | +++ b/tests/qemu-iotests/group | 162 | +++ b/hw/meson.build |
163 | @@ -XXX,XX +XXX,XX @@ subdir('moxie') | ||
164 | subdir('nios2') | ||
165 | subdir('openrisc') | ||
166 | subdir('ppc') | ||
167 | +subdir('remote') | ||
168 | subdir('riscv') | ||
169 | subdir('rx') | ||
170 | subdir('s390x') | ||
171 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
172 | new file mode 100644 | ||
173 | index XXXXXXX..XXXXXXX | ||
174 | --- /dev/null | ||
175 | +++ b/hw/remote/meson.build | ||
141 | @@ -XXX,XX +XXX,XX @@ | 176 | @@ -XXX,XX +XXX,XX @@ |
142 | 197 rw auto quick | 177 | +remote_ss = ss.source_set() |
143 | 198 rw auto | 178 | + |
144 | 200 rw auto | 179 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) |
145 | +202 rw auto quick | 180 | + |
181 | +softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
146 | -- | 182 | -- |
147 | 2.14.3 | 183 | 2.29.2 |
148 | 184 | ||
149 | diff view generated by jsdifflib |
1 | From: Paolo Bonzini <pbonzini@redhat.com> | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | BDRV_POLL_WHILE() does not support recursive AioContext locking. It | 3 | Adds qio_channel_writev_full_all() to transmit both data and FDs. |
4 | only releases the AioContext lock once regardless of how many times the | 4 | Refactors existing code to use this helper. |
5 | caller has acquired it. This results in a hang since the IOThread does | ||
6 | not make progress while the AioContext is still locked. | ||
7 | 5 | ||
8 | The following steps trigger the hang: | 6 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
9 | 7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | |
10 | $ qemu-system-x86_64 -M accel=kvm -m 1G -cpu host \ | 8 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> |
11 | -object iothread,id=iothread0 \ | 9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
12 | -device virtio-scsi-pci,iothread=iothread0 \ | 10 | Acked-by: Daniel P. Berrangé <berrange@redhat.com> |
13 | -drive if=none,id=drive0,file=test.img,format=raw \ | 11 | Message-id: 480fbf1fe4152495d60596c9b665124549b426a5.1611938319.git.jag.raman@oracle.com |
14 | -device scsi-hd,drive=drive0 \ | ||
15 | -drive if=none,id=drive1,file=test.img,format=raw \ | ||
16 | -device scsi-hd,drive=drive1 | ||
17 | $ qemu-system-x86_64 ...same options... \ | ||
18 | -incoming tcp::1234 | ||
19 | (qemu) migrate tcp:127.0.0.1:1234 | ||
20 | ...hang... | ||
21 | |||
22 | Tested-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
23 | Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> | ||
24 | Reviewed-by: Eric Blake <eblake@redhat.com> | ||
25 | Message-id: 20171207201320.19284-2-stefanha@redhat.com | ||
26 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
27 | --- | 13 | --- |
28 | block.c | 14 +++++++++++--- | 14 | include/io/channel.h | 25 +++++++++++++++++++++++++ |
29 | 1 file changed, 11 insertions(+), 3 deletions(-) | 15 | io/channel.c | 15 ++++++++++++++- |
16 | 2 files changed, 39 insertions(+), 1 deletion(-) | ||
30 | 17 | ||
31 | diff --git a/block.c b/block.c | 18 | diff --git a/include/io/channel.h b/include/io/channel.h |
32 | index XXXXXXX..XXXXXXX 100644 | 19 | index XXXXXXX..XXXXXXX 100644 |
33 | --- a/block.c | 20 | --- a/include/io/channel.h |
34 | +++ b/block.c | 21 | +++ b/include/io/channel.h |
35 | @@ -XXX,XX +XXX,XX @@ int bdrv_inactivate_all(void) | 22 | @@ -XXX,XX +XXX,XX @@ void qio_channel_set_aio_fd_handler(QIOChannel *ioc, |
36 | BdrvNextIterator it; | 23 | IOHandler *io_write, |
37 | int ret = 0; | 24 | void *opaque); |
38 | int pass; | 25 | |
39 | + GSList *aio_ctxs = NULL, *ctx; | 26 | +/** |
40 | 27 | + * qio_channel_writev_full_all: | |
41 | for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) { | 28 | + * @ioc: the channel object |
42 | - aio_context_acquire(bdrv_get_aio_context(bs)); | 29 | + * @iov: the array of memory regions to write data from |
43 | + AioContext *aio_context = bdrv_get_aio_context(bs); | 30 | + * @niov: the length of the @iov array |
31 | + * @fds: an array of file handles to send | ||
32 | + * @nfds: number of file handles in @fds | ||
33 | + * @errp: pointer to a NULL-initialized error object | ||
34 | + * | ||
35 | + * | ||
36 | + * Behaves like qio_channel_writev_full but will attempt | ||
37 | + * to send all data passed (file handles and memory regions). | ||
38 | + * The function will wait for all requested data | ||
39 | + * to be written, yielding from the current coroutine | ||
40 | + * if required. | ||
41 | + * | ||
42 | + * Returns: 0 if all bytes were written, or -1 on error | ||
43 | + */ | ||
44 | + | 44 | + |
45 | + if (!g_slist_find(aio_ctxs, aio_context)) { | 45 | +int qio_channel_writev_full_all(QIOChannel *ioc, |
46 | + aio_ctxs = g_slist_prepend(aio_ctxs, aio_context); | 46 | + const struct iovec *iov, |
47 | + aio_context_acquire(aio_context); | 47 | + size_t niov, |
48 | + } | 48 | + int *fds, size_t nfds, |
49 | + Error **errp); | ||
50 | + | ||
51 | #endif /* QIO_CHANNEL_H */ | ||
52 | diff --git a/io/channel.c b/io/channel.c | ||
53 | index XXXXXXX..XXXXXXX 100644 | ||
54 | --- a/io/channel.c | ||
55 | +++ b/io/channel.c | ||
56 | @@ -XXX,XX +XXX,XX @@ int qio_channel_writev_all(QIOChannel *ioc, | ||
57 | const struct iovec *iov, | ||
58 | size_t niov, | ||
59 | Error **errp) | ||
60 | +{ | ||
61 | + return qio_channel_writev_full_all(ioc, iov, niov, NULL, 0, errp); | ||
62 | +} | ||
63 | + | ||
64 | +int qio_channel_writev_full_all(QIOChannel *ioc, | ||
65 | + const struct iovec *iov, | ||
66 | + size_t niov, | ||
67 | + int *fds, size_t nfds, | ||
68 | + Error **errp) | ||
69 | { | ||
70 | int ret = -1; | ||
71 | struct iovec *local_iov = g_new(struct iovec, niov); | ||
72 | @@ -XXX,XX +XXX,XX @@ int qio_channel_writev_all(QIOChannel *ioc, | ||
73 | |||
74 | while (nlocal_iov > 0) { | ||
75 | ssize_t len; | ||
76 | - len = qio_channel_writev(ioc, local_iov, nlocal_iov, errp); | ||
77 | + len = qio_channel_writev_full(ioc, local_iov, nlocal_iov, fds, nfds, | ||
78 | + errp); | ||
79 | if (len == QIO_CHANNEL_ERR_BLOCK) { | ||
80 | if (qemu_in_coroutine()) { | ||
81 | qio_channel_yield(ioc, G_IO_OUT); | ||
82 | @@ -XXX,XX +XXX,XX @@ int qio_channel_writev_all(QIOChannel *ioc, | ||
83 | } | ||
84 | |||
85 | iov_discard_front(&local_iov, &nlocal_iov, len); | ||
86 | + | ||
87 | + fds = NULL; | ||
88 | + nfds = 0; | ||
49 | } | 89 | } |
50 | 90 | ||
51 | /* We do two passes of inactivation. The first pass calls to drivers' | 91 | ret = 0; |
52 | @@ -XXX,XX +XXX,XX @@ int bdrv_inactivate_all(void) | ||
53 | } | ||
54 | |||
55 | out: | ||
56 | - for (bs = bdrv_first(&it); bs; bs = bdrv_next(&it)) { | ||
57 | - aio_context_release(bdrv_get_aio_context(bs)); | ||
58 | + for (ctx = aio_ctxs; ctx != NULL; ctx = ctx->next) { | ||
59 | + AioContext *aio_context = ctx->data; | ||
60 | + aio_context_release(aio_context); | ||
61 | } | ||
62 | + g_slist_free(aio_ctxs); | ||
63 | |||
64 | return ret; | ||
65 | } | ||
66 | -- | 92 | -- |
67 | 2.14.3 | 93 | 2.29.2 |
68 | 94 | ||
69 | diff view generated by jsdifflib |
1 | The dirty bitmap actions in qmp_transaction have not used AioContext | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | since the dirty bitmap locking discipline was introduced in commit | 2 | |
3 | 2119882c7eb7e2c612b24fc0c8d86f5887d6f1c3 ("block: introduce | 3 | Adds qio_channel_readv_full_all_eof() and qio_channel_readv_full_all() |
4 | dirty_bitmap_mutex"). Remove the unused field. | 4 | to read both data and FDs. Refactors existing code to use these helpers. |
5 | 5 | ||
6 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 6 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
7 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
8 | Reviewed-by: Eric Blake <eblake@redhat.com> | 8 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> |
9 | Message-id: 20171206144550.22295-7-stefanha@redhat.com | 9 | Acked-by: Daniel P. Berrangé <berrange@redhat.com> |
10 | Message-id: b059c4cc0fb741e794d644c144cc21372cad877d.1611938319.git.jag.raman@oracle.com | ||
10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
11 | --- | 12 | --- |
12 | blockdev.c | 13 ------------- | 13 | include/io/channel.h | 53 +++++++++++++++++++++++ |
13 | 1 file changed, 13 deletions(-) | 14 | io/channel.c | 101 ++++++++++++++++++++++++++++++++++--------- |
14 | 15 | 2 files changed, 134 insertions(+), 20 deletions(-) | |
15 | diff --git a/blockdev.c b/blockdev.c | 16 | |
17 | diff --git a/include/io/channel.h b/include/io/channel.h | ||
16 | index XXXXXXX..XXXXXXX 100644 | 18 | index XXXXXXX..XXXXXXX 100644 |
17 | --- a/blockdev.c | 19 | --- a/include/io/channel.h |
18 | +++ b/blockdev.c | 20 | +++ b/include/io/channel.h |
19 | @@ -XXX,XX +XXX,XX @@ typedef struct BlockDirtyBitmapState { | 21 | @@ -XXX,XX +XXX,XX @@ void qio_channel_set_aio_fd_handler(QIOChannel *ioc, |
20 | BlkActionState common; | 22 | IOHandler *io_write, |
21 | BdrvDirtyBitmap *bitmap; | 23 | void *opaque); |
22 | BlockDriverState *bs; | 24 | |
23 | - AioContext *aio_context; | 25 | +/** |
24 | HBitmap *backup; | 26 | + * qio_channel_readv_full_all_eof: |
25 | bool prepared; | 27 | + * @ioc: the channel object |
26 | } BlockDirtyBitmapState; | 28 | + * @iov: the array of memory regions to read data to |
27 | @@ -XXX,XX +XXX,XX @@ static void block_dirty_bitmap_clear_prepare(BlkActionState *common, | 29 | + * @niov: the length of the @iov array |
30 | + * @fds: an array of file handles to read | ||
31 | + * @nfds: number of file handles in @fds | ||
32 | + * @errp: pointer to a NULL-initialized error object | ||
33 | + * | ||
34 | + * | ||
35 | + * Performs same function as qio_channel_readv_all_eof. | ||
36 | + * Additionally, attempts to read file descriptors shared | ||
37 | + * over the channel. The function will wait for all | ||
38 | + * requested data to be read, yielding from the current | ||
39 | + * coroutine if required. data refers to both file | ||
40 | + * descriptors and the iovs. | ||
41 | + * | ||
42 | + * Returns: 1 if all bytes were read, 0 if end-of-file | ||
43 | + * occurs without data, or -1 on error | ||
44 | + */ | ||
45 | + | ||
46 | +int qio_channel_readv_full_all_eof(QIOChannel *ioc, | ||
47 | + const struct iovec *iov, | ||
48 | + size_t niov, | ||
49 | + int **fds, size_t *nfds, | ||
50 | + Error **errp); | ||
51 | + | ||
52 | +/** | ||
53 | + * qio_channel_readv_full_all: | ||
54 | + * @ioc: the channel object | ||
55 | + * @iov: the array of memory regions to read data to | ||
56 | + * @niov: the length of the @iov array | ||
57 | + * @fds: an array of file handles to read | ||
58 | + * @nfds: number of file handles in @fds | ||
59 | + * @errp: pointer to a NULL-initialized error object | ||
60 | + * | ||
61 | + * | ||
62 | + * Performs same function as qio_channel_readv_all_eof. | ||
63 | + * Additionally, attempts to read file descriptors shared | ||
64 | + * over the channel. The function will wait for all | ||
65 | + * requested data to be read, yielding from the current | ||
66 | + * coroutine if required. data refers to both file | ||
67 | + * descriptors and the iovs. | ||
68 | + * | ||
69 | + * Returns: 0 if all bytes were read, or -1 on error | ||
70 | + */ | ||
71 | + | ||
72 | +int qio_channel_readv_full_all(QIOChannel *ioc, | ||
73 | + const struct iovec *iov, | ||
74 | + size_t niov, | ||
75 | + int **fds, size_t *nfds, | ||
76 | + Error **errp); | ||
77 | + | ||
78 | /** | ||
79 | * qio_channel_writev_full_all: | ||
80 | * @ioc: the channel object | ||
81 | diff --git a/io/channel.c b/io/channel.c | ||
82 | index XXXXXXX..XXXXXXX 100644 | ||
83 | --- a/io/channel.c | ||
84 | +++ b/io/channel.c | ||
85 | @@ -XXX,XX +XXX,XX @@ int qio_channel_readv_all_eof(QIOChannel *ioc, | ||
86 | const struct iovec *iov, | ||
87 | size_t niov, | ||
88 | Error **errp) | ||
89 | +{ | ||
90 | + return qio_channel_readv_full_all_eof(ioc, iov, niov, NULL, NULL, errp); | ||
91 | +} | ||
92 | + | ||
93 | +int qio_channel_readv_all(QIOChannel *ioc, | ||
94 | + const struct iovec *iov, | ||
95 | + size_t niov, | ||
96 | + Error **errp) | ||
97 | +{ | ||
98 | + return qio_channel_readv_full_all(ioc, iov, niov, NULL, NULL, errp); | ||
99 | +} | ||
100 | + | ||
101 | +int qio_channel_readv_full_all_eof(QIOChannel *ioc, | ||
102 | + const struct iovec *iov, | ||
103 | + size_t niov, | ||
104 | + int **fds, size_t *nfds, | ||
105 | + Error **errp) | ||
106 | { | ||
107 | int ret = -1; | ||
108 | struct iovec *local_iov = g_new(struct iovec, niov); | ||
109 | struct iovec *local_iov_head = local_iov; | ||
110 | unsigned int nlocal_iov = niov; | ||
111 | + int **local_fds = fds; | ||
112 | + size_t *local_nfds = nfds; | ||
113 | bool partial = false; | ||
114 | |||
115 | + if (nfds) { | ||
116 | + *nfds = 0; | ||
117 | + } | ||
118 | + | ||
119 | + if (fds) { | ||
120 | + *fds = NULL; | ||
121 | + } | ||
122 | + | ||
123 | nlocal_iov = iov_copy(local_iov, nlocal_iov, | ||
124 | iov, niov, | ||
125 | 0, iov_size(iov, niov)); | ||
126 | |||
127 | - while (nlocal_iov > 0) { | ||
128 | + while ((nlocal_iov > 0) || local_fds) { | ||
129 | ssize_t len; | ||
130 | - len = qio_channel_readv(ioc, local_iov, nlocal_iov, errp); | ||
131 | + len = qio_channel_readv_full(ioc, local_iov, nlocal_iov, local_fds, | ||
132 | + local_nfds, errp); | ||
133 | if (len == QIO_CHANNEL_ERR_BLOCK) { | ||
134 | if (qemu_in_coroutine()) { | ||
135 | qio_channel_yield(ioc, G_IO_IN); | ||
136 | @@ -XXX,XX +XXX,XX @@ int qio_channel_readv_all_eof(QIOChannel *ioc, | ||
137 | qio_channel_wait(ioc, G_IO_IN); | ||
138 | } | ||
139 | continue; | ||
140 | - } else if (len < 0) { | ||
141 | - goto cleanup; | ||
142 | - } else if (len == 0) { | ||
143 | - if (partial) { | ||
144 | - error_setg(errp, | ||
145 | - "Unexpected end-of-file before all bytes were read"); | ||
146 | - } else { | ||
147 | + } | ||
148 | + | ||
149 | + if (len == 0) { | ||
150 | + if (local_nfds && *local_nfds) { | ||
151 | + /* | ||
152 | + * Got some FDs, but no data yet. This isn't an EOF | ||
153 | + * scenario (yet), so carry on to try to read data | ||
154 | + * on next loop iteration | ||
155 | + */ | ||
156 | + goto next_iter; | ||
157 | + } else if (!partial) { | ||
158 | + /* No fds and no data - EOF before any data read */ | ||
159 | ret = 0; | ||
160 | + goto cleanup; | ||
161 | + } else { | ||
162 | + len = -1; | ||
163 | + error_setg(errp, | ||
164 | + "Unexpected end-of-file before all data were read"); | ||
165 | + /* Fallthrough into len < 0 handling */ | ||
166 | + } | ||
167 | + } | ||
168 | + | ||
169 | + if (len < 0) { | ||
170 | + /* Close any FDs we previously received */ | ||
171 | + if (nfds && fds) { | ||
172 | + size_t i; | ||
173 | + for (i = 0; i < (*nfds); i++) { | ||
174 | + close((*fds)[i]); | ||
175 | + } | ||
176 | + g_free(*fds); | ||
177 | + *fds = NULL; | ||
178 | + *nfds = 0; | ||
179 | } | ||
180 | goto cleanup; | ||
181 | } | ||
182 | |||
183 | + if (nlocal_iov) { | ||
184 | + iov_discard_front(&local_iov, &nlocal_iov, len); | ||
185 | + } | ||
186 | + | ||
187 | +next_iter: | ||
188 | partial = true; | ||
189 | - iov_discard_front(&local_iov, &nlocal_iov, len); | ||
190 | + local_fds = NULL; | ||
191 | + local_nfds = NULL; | ||
28 | } | 192 | } |
29 | 193 | ||
30 | bdrv_clear_dirty_bitmap(state->bitmap, &state->backup); | 194 | ret = 1; |
31 | - /* AioContext is released in .clean() */ | 195 | @@ -XXX,XX +XXX,XX @@ int qio_channel_readv_all_eof(QIOChannel *ioc, |
196 | return ret; | ||
32 | } | 197 | } |
33 | 198 | ||
34 | static void block_dirty_bitmap_clear_abort(BlkActionState *common) | 199 | -int qio_channel_readv_all(QIOChannel *ioc, |
35 | @@ -XXX,XX +XXX,XX @@ static void block_dirty_bitmap_clear_commit(BlkActionState *common) | 200 | - const struct iovec *iov, |
36 | hbitmap_free(state->backup); | 201 | - size_t niov, |
202 | - Error **errp) | ||
203 | +int qio_channel_readv_full_all(QIOChannel *ioc, | ||
204 | + const struct iovec *iov, | ||
205 | + size_t niov, | ||
206 | + int **fds, size_t *nfds, | ||
207 | + Error **errp) | ||
208 | { | ||
209 | - int ret = qio_channel_readv_all_eof(ioc, iov, niov, errp); | ||
210 | + int ret = qio_channel_readv_full_all_eof(ioc, iov, niov, fds, nfds, errp); | ||
211 | |||
212 | if (ret == 0) { | ||
213 | - ret = -1; | ||
214 | - error_setg(errp, | ||
215 | - "Unexpected end-of-file before all bytes were read"); | ||
216 | - } else if (ret == 1) { | ||
217 | - ret = 0; | ||
218 | + error_prepend(errp, | ||
219 | + "Unexpected end-of-file before all data were read."); | ||
220 | + return -1; | ||
221 | } | ||
222 | + if (ret == 1) { | ||
223 | + return 0; | ||
224 | + } | ||
225 | + | ||
226 | return ret; | ||
37 | } | 227 | } |
38 | 228 | ||
39 | -static void block_dirty_bitmap_clear_clean(BlkActionState *common) | ||
40 | -{ | ||
41 | - BlockDirtyBitmapState *state = DO_UPCAST(BlockDirtyBitmapState, | ||
42 | - common, common); | ||
43 | - | ||
44 | - if (state->aio_context) { | ||
45 | - aio_context_release(state->aio_context); | ||
46 | - } | ||
47 | -} | ||
48 | - | ||
49 | static void abort_prepare(BlkActionState *common, Error **errp) | ||
50 | { | ||
51 | error_setg(errp, "Transaction aborted using Abort action"); | ||
52 | @@ -XXX,XX +XXX,XX @@ static const BlkActionOps actions[] = { | ||
53 | .prepare = block_dirty_bitmap_clear_prepare, | ||
54 | .commit = block_dirty_bitmap_clear_commit, | ||
55 | .abort = block_dirty_bitmap_clear_abort, | ||
56 | - .clean = block_dirty_bitmap_clear_clean, | ||
57 | } | ||
58 | }; | ||
59 | |||
60 | -- | 229 | -- |
61 | 2.14.3 | 230 | 2.29.2 |
62 | 231 | ||
63 | diff view generated by jsdifflib |
1 | Encapsulate IOThread QOM object lookup so that callers don't need to | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | know how and where IOThread objects live. | ||
3 | 2 | ||
4 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 3 | Defines MPQemuMsg, which is the message that is sent to the remote |
5 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 4 | process. This message is sent over QIOChannel and is used to |
6 | Reviewed-by: Eric Blake <eblake@redhat.com> | 5 | command the remote process to perform various tasks. |
7 | Message-id: 20171206144550.22295-8-stefanha@redhat.com | 6 | Define transmission functions used by proxy and by remote. |
7 | |||
8 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
9 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
10 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
11 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
12 | Message-id: 56ca8bcf95195b2b195b08f6b9565b6d7410bce5.1611938319.git.jag.raman@oracle.com | ||
8 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 13 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
9 | --- | 14 | --- |
10 | include/sysemu/iothread.h | 1 + | 15 | MAINTAINERS | 2 + |
11 | iothread.c | 7 +++++++ | 16 | meson.build | 1 + |
12 | 2 files changed, 8 insertions(+) | 17 | hw/remote/trace.h | 1 + |
18 | include/hw/remote/mpqemu-link.h | 63 ++++++++++ | ||
19 | include/sysemu/iothread.h | 6 + | ||
20 | hw/remote/mpqemu-link.c | 205 ++++++++++++++++++++++++++++++++ | ||
21 | iothread.c | 6 + | ||
22 | hw/remote/meson.build | 1 + | ||
23 | hw/remote/trace-events | 4 + | ||
24 | 9 files changed, 289 insertions(+) | ||
25 | create mode 100644 hw/remote/trace.h | ||
26 | create mode 100644 include/hw/remote/mpqemu-link.h | ||
27 | create mode 100644 hw/remote/mpqemu-link.c | ||
28 | create mode 100644 hw/remote/trace-events | ||
13 | 29 | ||
30 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
31 | index XXXXXXX..XXXXXXX 100644 | ||
32 | --- a/MAINTAINERS | ||
33 | +++ b/MAINTAINERS | ||
34 | @@ -XXX,XX +XXX,XX @@ F: hw/pci-host/remote.c | ||
35 | F: include/hw/pci-host/remote.h | ||
36 | F: hw/remote/machine.c | ||
37 | F: include/hw/remote/machine.h | ||
38 | +F: hw/remote/mpqemu-link.c | ||
39 | +F: include/hw/remote/mpqemu-link.h | ||
40 | |||
41 | Build and test automation | ||
42 | ------------------------- | ||
43 | diff --git a/meson.build b/meson.build | ||
44 | index XXXXXXX..XXXXXXX 100644 | ||
45 | --- a/meson.build | ||
46 | +++ b/meson.build | ||
47 | @@ -XXX,XX +XXX,XX @@ if have_system | ||
48 | 'net', | ||
49 | 'softmmu', | ||
50 | 'ui', | ||
51 | + 'hw/remote', | ||
52 | ] | ||
53 | endif | ||
54 | trace_events_subdirs += [ | ||
55 | diff --git a/hw/remote/trace.h b/hw/remote/trace.h | ||
56 | new file mode 100644 | ||
57 | index XXXXXXX..XXXXXXX | ||
58 | --- /dev/null | ||
59 | +++ b/hw/remote/trace.h | ||
60 | @@ -0,0 +1 @@ | ||
61 | +#include "trace/trace-hw_remote.h" | ||
62 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
63 | new file mode 100644 | ||
64 | index XXXXXXX..XXXXXXX | ||
65 | --- /dev/null | ||
66 | +++ b/include/hw/remote/mpqemu-link.h | ||
67 | @@ -XXX,XX +XXX,XX @@ | ||
68 | +/* | ||
69 | + * Communication channel between QEMU and remote device process | ||
70 | + * | ||
71 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
72 | + * | ||
73 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
74 | + * See the COPYING file in the top-level directory. | ||
75 | + * | ||
76 | + */ | ||
77 | + | ||
78 | +#ifndef MPQEMU_LINK_H | ||
79 | +#define MPQEMU_LINK_H | ||
80 | + | ||
81 | +#include "qom/object.h" | ||
82 | +#include "qemu/thread.h" | ||
83 | +#include "io/channel.h" | ||
84 | + | ||
85 | +#define REMOTE_MAX_FDS 8 | ||
86 | + | ||
87 | +#define MPQEMU_MSG_HDR_SIZE offsetof(MPQemuMsg, data.u64) | ||
88 | + | ||
89 | +/** | ||
90 | + * MPQemuCmd: | ||
91 | + * | ||
92 | + * MPQemuCmd enum type to specify the command to be executed on the remote | ||
93 | + * device. | ||
94 | + * | ||
95 | + * This uses a private protocol between QEMU and the remote process. vfio-user | ||
96 | + * protocol would supersede this in the future. | ||
97 | + * | ||
98 | + */ | ||
99 | +typedef enum { | ||
100 | + MPQEMU_CMD_MAX, | ||
101 | +} MPQemuCmd; | ||
102 | + | ||
103 | +/** | ||
104 | + * MPQemuMsg: | ||
105 | + * @cmd: The remote command | ||
106 | + * @size: Size of the data to be shared | ||
107 | + * @data: Structured data | ||
108 | + * @fds: File descriptors to be shared with remote device | ||
109 | + * | ||
110 | + * MPQemuMsg Format of the message sent to the remote device from QEMU. | ||
111 | + * | ||
112 | + */ | ||
113 | +typedef struct { | ||
114 | + int cmd; | ||
115 | + size_t size; | ||
116 | + | ||
117 | + union { | ||
118 | + uint64_t u64; | ||
119 | + } data; | ||
120 | + | ||
121 | + int fds[REMOTE_MAX_FDS]; | ||
122 | + int num_fds; | ||
123 | +} MPQemuMsg; | ||
124 | + | ||
125 | +bool mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); | ||
126 | +bool mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); | ||
127 | + | ||
128 | +bool mpqemu_msg_valid(MPQemuMsg *msg); | ||
129 | + | ||
130 | +#endif | ||
14 | diff --git a/include/sysemu/iothread.h b/include/sysemu/iothread.h | 131 | diff --git a/include/sysemu/iothread.h b/include/sysemu/iothread.h |
15 | index XXXXXXX..XXXXXXX 100644 | 132 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/include/sysemu/iothread.h | 133 | --- a/include/sysemu/iothread.h |
17 | +++ b/include/sysemu/iothread.h | 134 | +++ b/include/sysemu/iothread.h |
18 | @@ -XXX,XX +XXX,XX @@ typedef struct { | 135 | @@ -XXX,XX +XXX,XX @@ IOThread *iothread_create(const char *id, Error **errp); |
19 | OBJECT_CHECK(IOThread, obj, TYPE_IOTHREAD) | 136 | void iothread_stop(IOThread *iothread); |
20 | 137 | void iothread_destroy(IOThread *iothread); | |
21 | char *iothread_get_id(IOThread *iothread); | 138 | |
22 | +IOThread *iothread_by_id(const char *id); | 139 | +/* |
23 | AioContext *iothread_get_aio_context(IOThread *iothread); | 140 | + * Returns true if executing withing IOThread context, |
24 | void iothread_stop_all(void); | 141 | + * false otherwise. |
25 | GMainContext *iothread_get_g_main_context(IOThread *iothread); | 142 | + */ |
143 | +bool qemu_in_iothread(void); | ||
144 | + | ||
145 | #endif /* IOTHREAD_H */ | ||
146 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
147 | new file mode 100644 | ||
148 | index XXXXXXX..XXXXXXX | ||
149 | --- /dev/null | ||
150 | +++ b/hw/remote/mpqemu-link.c | ||
151 | @@ -XXX,XX +XXX,XX @@ | ||
152 | +/* | ||
153 | + * Communication channel between QEMU and remote device process | ||
154 | + * | ||
155 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
156 | + * | ||
157 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
158 | + * See the COPYING file in the top-level directory. | ||
159 | + * | ||
160 | + */ | ||
161 | + | ||
162 | +#include "qemu/osdep.h" | ||
163 | +#include "qemu-common.h" | ||
164 | + | ||
165 | +#include "qemu/module.h" | ||
166 | +#include "hw/remote/mpqemu-link.h" | ||
167 | +#include "qapi/error.h" | ||
168 | +#include "qemu/iov.h" | ||
169 | +#include "qemu/error-report.h" | ||
170 | +#include "qemu/main-loop.h" | ||
171 | +#include "io/channel.h" | ||
172 | +#include "sysemu/iothread.h" | ||
173 | +#include "trace.h" | ||
174 | + | ||
175 | +/* | ||
176 | + * Send message over the ioc QIOChannel. | ||
177 | + * This function is safe to call from: | ||
178 | + * - main loop in co-routine context. Will block the main loop if not in | ||
179 | + * co-routine context; | ||
180 | + * - vCPU thread with no co-routine context and if the channel is not part | ||
181 | + * of the main loop handling; | ||
182 | + * - IOThread within co-routine context, outside of co-routine context | ||
183 | + * will block IOThread; | ||
184 | + * Returns true if no errors were encountered, false otherwise. | ||
185 | + */ | ||
186 | +bool mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp) | ||
187 | +{ | ||
188 | + ERRP_GUARD(); | ||
189 | + bool iolock = qemu_mutex_iothread_locked(); | ||
190 | + bool iothread = qemu_in_iothread(); | ||
191 | + struct iovec send[2] = {0}; | ||
192 | + int *fds = NULL; | ||
193 | + size_t nfds = 0; | ||
194 | + bool ret = false; | ||
195 | + | ||
196 | + send[0].iov_base = msg; | ||
197 | + send[0].iov_len = MPQEMU_MSG_HDR_SIZE; | ||
198 | + | ||
199 | + send[1].iov_base = (void *)&msg->data; | ||
200 | + send[1].iov_len = msg->size; | ||
201 | + | ||
202 | + if (msg->num_fds) { | ||
203 | + nfds = msg->num_fds; | ||
204 | + fds = msg->fds; | ||
205 | + } | ||
206 | + | ||
207 | + /* | ||
208 | + * Dont use in IOThread out of co-routine context as | ||
209 | + * it will block IOThread. | ||
210 | + */ | ||
211 | + assert(qemu_in_coroutine() || !iothread); | ||
212 | + | ||
213 | + /* | ||
214 | + * Skip unlocking/locking iothread lock when the IOThread is running | ||
215 | + * in co-routine context. Co-routine context is asserted above | ||
216 | + * for IOThread case. | ||
217 | + * Also skip lock handling while in a co-routine in the main context. | ||
218 | + */ | ||
219 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
220 | + qemu_mutex_unlock_iothread(); | ||
221 | + } | ||
222 | + | ||
223 | + if (!qio_channel_writev_full_all(ioc, send, G_N_ELEMENTS(send), | ||
224 | + fds, nfds, errp)) { | ||
225 | + ret = true; | ||
226 | + } else { | ||
227 | + trace_mpqemu_send_io_error(msg->cmd, msg->size, nfds); | ||
228 | + } | ||
229 | + | ||
230 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
231 | + /* See above comment why skip locking here. */ | ||
232 | + qemu_mutex_lock_iothread(); | ||
233 | + } | ||
234 | + | ||
235 | + return ret; | ||
236 | +} | ||
237 | + | ||
238 | +/* | ||
239 | + * Read message from the ioc QIOChannel. | ||
240 | + * This function is safe to call from: | ||
241 | + * - From main loop in co-routine context. Will block the main loop if not in | ||
242 | + * co-routine context; | ||
243 | + * - From vCPU thread with no co-routine context and if the channel is not part | ||
244 | + * of the main loop handling; | ||
245 | + * - From IOThread within co-routine context, outside of co-routine context | ||
246 | + * will block IOThread; | ||
247 | + */ | ||
248 | +static ssize_t mpqemu_read(QIOChannel *ioc, void *buf, size_t len, int **fds, | ||
249 | + size_t *nfds, Error **errp) | ||
250 | +{ | ||
251 | + ERRP_GUARD(); | ||
252 | + struct iovec iov = { .iov_base = buf, .iov_len = len }; | ||
253 | + bool iolock = qemu_mutex_iothread_locked(); | ||
254 | + bool iothread = qemu_in_iothread(); | ||
255 | + int ret = -1; | ||
256 | + | ||
257 | + /* | ||
258 | + * Dont use in IOThread out of co-routine context as | ||
259 | + * it will block IOThread. | ||
260 | + */ | ||
261 | + assert(qemu_in_coroutine() || !iothread); | ||
262 | + | ||
263 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
264 | + qemu_mutex_unlock_iothread(); | ||
265 | + } | ||
266 | + | ||
267 | + ret = qio_channel_readv_full_all_eof(ioc, &iov, 1, fds, nfds, errp); | ||
268 | + | ||
269 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
270 | + qemu_mutex_lock_iothread(); | ||
271 | + } | ||
272 | + | ||
273 | + return (ret <= 0) ? ret : iov.iov_len; | ||
274 | +} | ||
275 | + | ||
276 | +bool mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp) | ||
277 | +{ | ||
278 | + ERRP_GUARD(); | ||
279 | + g_autofree int *fds = NULL; | ||
280 | + size_t nfds = 0; | ||
281 | + ssize_t len; | ||
282 | + bool ret = false; | ||
283 | + | ||
284 | + len = mpqemu_read(ioc, msg, MPQEMU_MSG_HDR_SIZE, &fds, &nfds, errp); | ||
285 | + if (len <= 0) { | ||
286 | + goto fail; | ||
287 | + } else if (len != MPQEMU_MSG_HDR_SIZE) { | ||
288 | + error_setg(errp, "Message header corrupted"); | ||
289 | + goto fail; | ||
290 | + } | ||
291 | + | ||
292 | + if (msg->size > sizeof(msg->data)) { | ||
293 | + error_setg(errp, "Invalid size for message"); | ||
294 | + goto fail; | ||
295 | + } | ||
296 | + | ||
297 | + if (!msg->size) { | ||
298 | + goto copy_fds; | ||
299 | + } | ||
300 | + | ||
301 | + len = mpqemu_read(ioc, &msg->data, msg->size, NULL, NULL, errp); | ||
302 | + if (len <= 0) { | ||
303 | + goto fail; | ||
304 | + } | ||
305 | + if (len != msg->size) { | ||
306 | + error_setg(errp, "Unable to read full message"); | ||
307 | + goto fail; | ||
308 | + } | ||
309 | + | ||
310 | +copy_fds: | ||
311 | + msg->num_fds = nfds; | ||
312 | + if (nfds > G_N_ELEMENTS(msg->fds)) { | ||
313 | + error_setg(errp, | ||
314 | + "Overflow error: received %zu fds, more than max of %d fds", | ||
315 | + nfds, REMOTE_MAX_FDS); | ||
316 | + goto fail; | ||
317 | + } | ||
318 | + if (nfds) { | ||
319 | + memcpy(msg->fds, fds, nfds * sizeof(int)); | ||
320 | + } | ||
321 | + | ||
322 | + ret = true; | ||
323 | + | ||
324 | +fail: | ||
325 | + if (*errp) { | ||
326 | + trace_mpqemu_recv_io_error(msg->cmd, msg->size, nfds); | ||
327 | + } | ||
328 | + while (*errp && nfds) { | ||
329 | + close(fds[nfds - 1]); | ||
330 | + nfds--; | ||
331 | + } | ||
332 | + | ||
333 | + return ret; | ||
334 | +} | ||
335 | + | ||
336 | +bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
337 | +{ | ||
338 | + if (msg->cmd >= MPQEMU_CMD_MAX && msg->cmd < 0) { | ||
339 | + return false; | ||
340 | + } | ||
341 | + | ||
342 | + /* Verify FDs. */ | ||
343 | + if (msg->num_fds >= REMOTE_MAX_FDS) { | ||
344 | + return false; | ||
345 | + } | ||
346 | + | ||
347 | + if (msg->num_fds > 0) { | ||
348 | + for (int i = 0; i < msg->num_fds; i++) { | ||
349 | + if (fcntl(msg->fds[i], F_GETFL) == -1) { | ||
350 | + return false; | ||
351 | + } | ||
352 | + } | ||
353 | + } | ||
354 | + | ||
355 | + return true; | ||
356 | +} | ||
26 | diff --git a/iothread.c b/iothread.c | 357 | diff --git a/iothread.c b/iothread.c |
27 | index XXXXXXX..XXXXXXX 100644 | 358 | index XXXXXXX..XXXXXXX 100644 |
28 | --- a/iothread.c | 359 | --- a/iothread.c |
29 | +++ b/iothread.c | 360 | +++ b/iothread.c |
30 | @@ -XXX,XX +XXX,XX @@ void iothread_destroy(IOThread *iothread) | 361 | @@ -XXX,XX +XXX,XX @@ IOThread *iothread_by_id(const char *id) |
31 | { | 362 | { |
32 | object_unparent(OBJECT(iothread)); | 363 | return IOTHREAD(object_resolve_path_type(id, TYPE_IOTHREAD, NULL)); |
33 | } | 364 | } |
34 | + | 365 | + |
35 | +/* Lookup IOThread by its id. Only finds user-created objects, not internal | 366 | +bool qemu_in_iothread(void) |
36 | + * iothread_create() objects. */ | 367 | +{ |
37 | +IOThread *iothread_by_id(const char *id) | 368 | + return qemu_get_current_aio_context() == qemu_get_aio_context() ? |
38 | +{ | 369 | + false : true; |
39 | + return IOTHREAD(object_resolve_path_type(id, TYPE_IOTHREAD, NULL)); | 370 | +} |
40 | +} | 371 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build |
372 | index XXXXXXX..XXXXXXX 100644 | ||
373 | --- a/hw/remote/meson.build | ||
374 | +++ b/hw/remote/meson.build | ||
375 | @@ -XXX,XX +XXX,XX @@ | ||
376 | remote_ss = ss.source_set() | ||
377 | |||
378 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) | ||
379 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
380 | |||
381 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
382 | diff --git a/hw/remote/trace-events b/hw/remote/trace-events | ||
383 | new file mode 100644 | ||
384 | index XXXXXXX..XXXXXXX | ||
385 | --- /dev/null | ||
386 | +++ b/hw/remote/trace-events | ||
387 | @@ -XXX,XX +XXX,XX @@ | ||
388 | +# multi-process trace events | ||
389 | + | ||
390 | +mpqemu_send_io_error(int cmd, int size, int nfds) "send command %d size %d, %d file descriptors to remote process" | ||
391 | +mpqemu_recv_io_error(int cmd, int size, int nfds) "failed to receive %d size %d, %d file descriptors to remote process" | ||
41 | -- | 392 | -- |
42 | 2.14.3 | 393 | 2.29.2 |
43 | 394 | ||
44 | diff view generated by jsdifflib |
1 | From: Mark Kanda <mark.kanda@oracle.com> | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | virtio-blk logical block size should never be larger than physical block | 3 | Initializes the message handler function in the remote process. It is |
4 | size because it doesn't make sense to have such configurations. QEMU doesn't | 4 | called whenever there's an event pending on QIOChannel that registers |
5 | have a way to effectively express this condition; the best it can do is | 5 | this function. |
6 | report the physical block exponent as 0 - indicating the logical block size | ||
7 | equals the physical block size. | ||
8 | 6 | ||
9 | This is identical to commit 3da023b5827543ee4c022986ea2ad9d1274410b2 | 7 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
10 | but applied to virtio-blk (instead of virtio-scsi). | 8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
11 | 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | |
12 | Signed-off-by: Mark Kanda <mark.kanda@oracle.com> | ||
13 | Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> | ||
14 | Reviewed-by: Ameya More <ameya.more@oracle.com> | ||
15 | Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com> | ||
16 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
17 | Message-id: 773169891f9f2deb4cb7c4ef2655580dbe24c1d1.1513005190.git.mark.kanda@oracle.com | 11 | Message-id: 99d38d8b93753a6409ac2340e858858cda59ab1b.1611938319.git.jag.raman@oracle.com |
18 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
19 | --- | 13 | --- |
20 | hw/block/virtio-blk.c | 7 +++++++ | 14 | MAINTAINERS | 1 + |
21 | 1 file changed, 7 insertions(+) | 15 | include/hw/remote/machine.h | 9 ++++++ |
16 | hw/remote/message.c | 57 +++++++++++++++++++++++++++++++++++++ | ||
17 | hw/remote/meson.build | 1 + | ||
18 | 4 files changed, 68 insertions(+) | ||
19 | create mode 100644 hw/remote/message.c | ||
22 | 20 | ||
23 | diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c | 21 | diff --git a/MAINTAINERS b/MAINTAINERS |
24 | index XXXXXXX..XXXXXXX 100644 | 22 | index XXXXXXX..XXXXXXX 100644 |
25 | --- a/hw/block/virtio-blk.c | 23 | --- a/MAINTAINERS |
26 | +++ b/hw/block/virtio-blk.c | 24 | +++ b/MAINTAINERS |
27 | @@ -XXX,XX +XXX,XX @@ static void virtio_blk_device_realize(DeviceState *dev, Error **errp) | 25 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/machine.c |
28 | 26 | F: include/hw/remote/machine.h | |
29 | blkconf_blocksizes(&conf->conf); | 27 | F: hw/remote/mpqemu-link.c |
30 | 28 | F: include/hw/remote/mpqemu-link.h | |
31 | + if (conf->conf.logical_block_size > | 29 | +F: hw/remote/message.c |
32 | + conf->conf.physical_block_size) { | 30 | |
33 | + error_setg(errp, | 31 | Build and test automation |
34 | + "logical_block_size > physical_block_size not supported"); | 32 | ------------------------- |
35 | + return; | 33 | diff --git a/include/hw/remote/machine.h b/include/hw/remote/machine.h |
34 | index XXXXXXX..XXXXXXX 100644 | ||
35 | --- a/include/hw/remote/machine.h | ||
36 | +++ b/include/hw/remote/machine.h | ||
37 | @@ -XXX,XX +XXX,XX @@ | ||
38 | #include "qom/object.h" | ||
39 | #include "hw/boards.h" | ||
40 | #include "hw/pci-host/remote.h" | ||
41 | +#include "io/channel.h" | ||
42 | |||
43 | struct RemoteMachineState { | ||
44 | MachineState parent_obj; | ||
45 | @@ -XXX,XX +XXX,XX @@ struct RemoteMachineState { | ||
46 | RemotePCIHost *host; | ||
47 | }; | ||
48 | |||
49 | +/* Used to pass to co-routine device and ioc. */ | ||
50 | +typedef struct RemoteCommDev { | ||
51 | + PCIDevice *dev; | ||
52 | + QIOChannel *ioc; | ||
53 | +} RemoteCommDev; | ||
54 | + | ||
55 | #define TYPE_REMOTE_MACHINE "x-remote-machine" | ||
56 | OBJECT_DECLARE_SIMPLE_TYPE(RemoteMachineState, REMOTE_MACHINE) | ||
57 | |||
58 | +void coroutine_fn mpqemu_remote_msg_loop_co(void *data); | ||
59 | + | ||
60 | #endif | ||
61 | diff --git a/hw/remote/message.c b/hw/remote/message.c | ||
62 | new file mode 100644 | ||
63 | index XXXXXXX..XXXXXXX | ||
64 | --- /dev/null | ||
65 | +++ b/hw/remote/message.c | ||
66 | @@ -XXX,XX +XXX,XX @@ | ||
67 | +/* | ||
68 | + * Copyright © 2020, 2021 Oracle and/or its affiliates. | ||
69 | + * | ||
70 | + * This work is licensed under the terms of the GNU GPL-v2, version 2 or later. | ||
71 | + * | ||
72 | + * See the COPYING file in the top-level directory. | ||
73 | + * | ||
74 | + */ | ||
75 | + | ||
76 | +#include "qemu/osdep.h" | ||
77 | +#include "qemu-common.h" | ||
78 | + | ||
79 | +#include "hw/remote/machine.h" | ||
80 | +#include "io/channel.h" | ||
81 | +#include "hw/remote/mpqemu-link.h" | ||
82 | +#include "qapi/error.h" | ||
83 | +#include "sysemu/runstate.h" | ||
84 | + | ||
85 | +void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
86 | +{ | ||
87 | + g_autofree RemoteCommDev *com = (RemoteCommDev *)data; | ||
88 | + PCIDevice *pci_dev = NULL; | ||
89 | + Error *local_err = NULL; | ||
90 | + | ||
91 | + assert(com->ioc); | ||
92 | + | ||
93 | + pci_dev = com->dev; | ||
94 | + for (; !local_err;) { | ||
95 | + MPQemuMsg msg = {0}; | ||
96 | + | ||
97 | + if (!mpqemu_msg_recv(&msg, com->ioc, &local_err)) { | ||
98 | + break; | ||
99 | + } | ||
100 | + | ||
101 | + if (!mpqemu_msg_valid(&msg)) { | ||
102 | + error_setg(&local_err, "Received invalid message from proxy" | ||
103 | + "in remote process pid="FMT_pid"", | ||
104 | + getpid()); | ||
105 | + break; | ||
106 | + } | ||
107 | + | ||
108 | + switch (msg.cmd) { | ||
109 | + default: | ||
110 | + error_setg(&local_err, | ||
111 | + "Unknown command (%d) received for device %s" | ||
112 | + " (pid="FMT_pid")", | ||
113 | + msg.cmd, DEVICE(pci_dev)->id, getpid()); | ||
114 | + } | ||
36 | + } | 115 | + } |
37 | + | 116 | + |
38 | virtio_init(vdev, "virtio-blk", VIRTIO_ID_BLOCK, | 117 | + if (local_err) { |
39 | sizeof(struct virtio_blk_config)); | 118 | + error_report_err(local_err); |
40 | 119 | + qemu_system_shutdown_request(SHUTDOWN_CAUSE_HOST_ERROR); | |
120 | + } else { | ||
121 | + qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN); | ||
122 | + } | ||
123 | +} | ||
124 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
125 | index XXXXXXX..XXXXXXX 100644 | ||
126 | --- a/hw/remote/meson.build | ||
127 | +++ b/hw/remote/meson.build | ||
128 | @@ -XXX,XX +XXX,XX @@ remote_ss = ss.source_set() | ||
129 | |||
130 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) | ||
131 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
132 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) | ||
133 | |||
134 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
41 | -- | 135 | -- |
42 | 2.14.3 | 136 | 2.29.2 |
43 | 137 | ||
44 | diff view generated by jsdifflib |
1 | The AioContext pointer argument to co_aio_sleep_ns() is only used for | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | the sleep timer. It does not affect where the caller coroutine is | 2 | |
3 | resumed. | 3 | Associate the file descriptor for a PCIDevice in remote process with |
4 | 4 | DeviceState object. | |
5 | Due to changes to coroutine and AIO APIs it is now possible to drop the | 5 | |
6 | AioContext pointer argument. This is safe to do since no caller has | 6 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
7 | specific requirements for which AioContext the timer must run in. | 7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
8 | 8 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | |
9 | This patch drops the AioContext pointer argument and renames the | 9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
10 | function to simplify the API. | 10 | Message-id: f405a2ed5d7518b87bea7c59cfdf334d67e5ee51.1611938319.git.jag.raman@oracle.com |
11 | |||
12 | Reported-by: Paolo Bonzini <pbonzini@redhat.com> | ||
13 | Reported-by: Eric Blake <eblake@redhat.com> | ||
14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
15 | Reviewed-by: Eric Blake <eblake@redhat.com> | ||
16 | Message-id: 20171109102652.6360-1-stefanha@redhat.com | ||
17 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
18 | --- | 12 | --- |
19 | include/qemu/coroutine.h | 6 +----- | 13 | MAINTAINERS | 1 + |
20 | block/null.c | 3 +-- | 14 | hw/remote/remote-obj.c | 203 +++++++++++++++++++++++++++++++++++++++++ |
21 | block/sheepdog.c | 3 +-- | 15 | hw/remote/meson.build | 1 + |
22 | util/qemu-coroutine-sleep.c | 4 ++-- | 16 | 3 files changed, 205 insertions(+) |
23 | 4 files changed, 5 insertions(+), 11 deletions(-) | 17 | create mode 100644 hw/remote/remote-obj.c |
24 | 18 | ||
25 | diff --git a/include/qemu/coroutine.h b/include/qemu/coroutine.h | 19 | diff --git a/MAINTAINERS b/MAINTAINERS |
26 | index XXXXXXX..XXXXXXX 100644 | 20 | index XXXXXXX..XXXXXXX 100644 |
27 | --- a/include/qemu/coroutine.h | 21 | --- a/MAINTAINERS |
28 | +++ b/include/qemu/coroutine.h | 22 | +++ b/MAINTAINERS |
29 | @@ -XXX,XX +XXX,XX @@ void qemu_co_rwlock_unlock(CoRwlock *lock); | 23 | @@ -XXX,XX +XXX,XX @@ F: include/hw/remote/machine.h |
30 | 24 | F: hw/remote/mpqemu-link.c | |
31 | /** | 25 | F: include/hw/remote/mpqemu-link.h |
32 | * Yield the coroutine for a given duration | 26 | F: hw/remote/message.c |
33 | - * | 27 | +F: hw/remote/remote-obj.c |
34 | - * Behaves similarly to co_sleep_ns(), but the sleeping coroutine will be | 28 | |
35 | - * resumed when using aio_poll(). | 29 | Build and test automation |
36 | */ | 30 | ------------------------- |
37 | -void coroutine_fn co_aio_sleep_ns(AioContext *ctx, QEMUClockType type, | 31 | diff --git a/hw/remote/remote-obj.c b/hw/remote/remote-obj.c |
38 | - int64_t ns); | 32 | new file mode 100644 |
39 | +void coroutine_fn qemu_co_sleep_ns(QEMUClockType type, int64_t ns); | 33 | index XXXXXXX..XXXXXXX |
40 | 34 | --- /dev/null | |
41 | /** | 35 | +++ b/hw/remote/remote-obj.c |
42 | * Yield until a file descriptor becomes readable | 36 | @@ -XXX,XX +XXX,XX @@ |
43 | diff --git a/block/null.c b/block/null.c | 37 | +/* |
38 | + * Copyright © 2020, 2021 Oracle and/or its affiliates. | ||
39 | + * | ||
40 | + * This work is licensed under the terms of the GNU GPL-v2, version 2 or later. | ||
41 | + * | ||
42 | + * See the COPYING file in the top-level directory. | ||
43 | + * | ||
44 | + */ | ||
45 | + | ||
46 | +#include "qemu/osdep.h" | ||
47 | +#include "qemu-common.h" | ||
48 | + | ||
49 | +#include "qemu/error-report.h" | ||
50 | +#include "qemu/notify.h" | ||
51 | +#include "qom/object_interfaces.h" | ||
52 | +#include "hw/qdev-core.h" | ||
53 | +#include "io/channel.h" | ||
54 | +#include "hw/qdev-core.h" | ||
55 | +#include "hw/remote/machine.h" | ||
56 | +#include "io/channel-util.h" | ||
57 | +#include "qapi/error.h" | ||
58 | +#include "sysemu/sysemu.h" | ||
59 | +#include "hw/pci/pci.h" | ||
60 | +#include "qemu/sockets.h" | ||
61 | +#include "monitor/monitor.h" | ||
62 | + | ||
63 | +#define TYPE_REMOTE_OBJECT "x-remote-object" | ||
64 | +OBJECT_DECLARE_TYPE(RemoteObject, RemoteObjectClass, REMOTE_OBJECT) | ||
65 | + | ||
66 | +struct RemoteObjectClass { | ||
67 | + ObjectClass parent_class; | ||
68 | + | ||
69 | + unsigned int nr_devs; | ||
70 | + unsigned int max_devs; | ||
71 | +}; | ||
72 | + | ||
73 | +struct RemoteObject { | ||
74 | + /* private */ | ||
75 | + Object parent; | ||
76 | + | ||
77 | + Notifier machine_done; | ||
78 | + | ||
79 | + int32_t fd; | ||
80 | + char *devid; | ||
81 | + | ||
82 | + QIOChannel *ioc; | ||
83 | + | ||
84 | + DeviceState *dev; | ||
85 | + DeviceListener listener; | ||
86 | +}; | ||
87 | + | ||
88 | +static void remote_object_set_fd(Object *obj, const char *str, Error **errp) | ||
89 | +{ | ||
90 | + RemoteObject *o = REMOTE_OBJECT(obj); | ||
91 | + int fd = -1; | ||
92 | + | ||
93 | + fd = monitor_fd_param(monitor_cur(), str, errp); | ||
94 | + if (fd == -1) { | ||
95 | + error_prepend(errp, "Could not parse remote object fd %s:", str); | ||
96 | + return; | ||
97 | + } | ||
98 | + | ||
99 | + if (!fd_is_socket(fd)) { | ||
100 | + error_setg(errp, "File descriptor '%s' is not a socket", str); | ||
101 | + close(fd); | ||
102 | + return; | ||
103 | + } | ||
104 | + | ||
105 | + o->fd = fd; | ||
106 | +} | ||
107 | + | ||
108 | +static void remote_object_set_devid(Object *obj, const char *str, Error **errp) | ||
109 | +{ | ||
110 | + RemoteObject *o = REMOTE_OBJECT(obj); | ||
111 | + | ||
112 | + g_free(o->devid); | ||
113 | + | ||
114 | + o->devid = g_strdup(str); | ||
115 | +} | ||
116 | + | ||
117 | +static void remote_object_unrealize_listener(DeviceListener *listener, | ||
118 | + DeviceState *dev) | ||
119 | +{ | ||
120 | + RemoteObject *o = container_of(listener, RemoteObject, listener); | ||
121 | + | ||
122 | + if (o->dev == dev) { | ||
123 | + object_unref(OBJECT(o)); | ||
124 | + } | ||
125 | +} | ||
126 | + | ||
127 | +static void remote_object_machine_done(Notifier *notifier, void *data) | ||
128 | +{ | ||
129 | + RemoteObject *o = container_of(notifier, RemoteObject, machine_done); | ||
130 | + DeviceState *dev = NULL; | ||
131 | + QIOChannel *ioc = NULL; | ||
132 | + Coroutine *co = NULL; | ||
133 | + RemoteCommDev *comdev = NULL; | ||
134 | + Error *err = NULL; | ||
135 | + | ||
136 | + dev = qdev_find_recursive(sysbus_get_default(), o->devid); | ||
137 | + if (!dev || !object_dynamic_cast(OBJECT(dev), TYPE_PCI_DEVICE)) { | ||
138 | + error_report("%s is not a PCI device", o->devid); | ||
139 | + return; | ||
140 | + } | ||
141 | + | ||
142 | + ioc = qio_channel_new_fd(o->fd, &err); | ||
143 | + if (!ioc) { | ||
144 | + error_report_err(err); | ||
145 | + return; | ||
146 | + } | ||
147 | + qio_channel_set_blocking(ioc, false, NULL); | ||
148 | + | ||
149 | + o->dev = dev; | ||
150 | + | ||
151 | + o->listener.unrealize = remote_object_unrealize_listener; | ||
152 | + device_listener_register(&o->listener); | ||
153 | + | ||
154 | + /* co-routine should free this. */ | ||
155 | + comdev = g_new0(RemoteCommDev, 1); | ||
156 | + *comdev = (RemoteCommDev) { | ||
157 | + .ioc = ioc, | ||
158 | + .dev = PCI_DEVICE(dev), | ||
159 | + }; | ||
160 | + | ||
161 | + co = qemu_coroutine_create(mpqemu_remote_msg_loop_co, comdev); | ||
162 | + qemu_coroutine_enter(co); | ||
163 | +} | ||
164 | + | ||
165 | +static void remote_object_init(Object *obj) | ||
166 | +{ | ||
167 | + RemoteObjectClass *k = REMOTE_OBJECT_GET_CLASS(obj); | ||
168 | + RemoteObject *o = REMOTE_OBJECT(obj); | ||
169 | + | ||
170 | + if (k->nr_devs >= k->max_devs) { | ||
171 | + error_report("Reached maximum number of devices: %u", k->max_devs); | ||
172 | + return; | ||
173 | + } | ||
174 | + | ||
175 | + o->ioc = NULL; | ||
176 | + o->fd = -1; | ||
177 | + o->devid = NULL; | ||
178 | + | ||
179 | + k->nr_devs++; | ||
180 | + | ||
181 | + o->machine_done.notify = remote_object_machine_done; | ||
182 | + qemu_add_machine_init_done_notifier(&o->machine_done); | ||
183 | +} | ||
184 | + | ||
185 | +static void remote_object_finalize(Object *obj) | ||
186 | +{ | ||
187 | + RemoteObjectClass *k = REMOTE_OBJECT_GET_CLASS(obj); | ||
188 | + RemoteObject *o = REMOTE_OBJECT(obj); | ||
189 | + | ||
190 | + device_listener_unregister(&o->listener); | ||
191 | + | ||
192 | + if (o->ioc) { | ||
193 | + qio_channel_shutdown(o->ioc, QIO_CHANNEL_SHUTDOWN_BOTH, NULL); | ||
194 | + qio_channel_close(o->ioc, NULL); | ||
195 | + } | ||
196 | + | ||
197 | + object_unref(OBJECT(o->ioc)); | ||
198 | + | ||
199 | + k->nr_devs--; | ||
200 | + g_free(o->devid); | ||
201 | +} | ||
202 | + | ||
203 | +static void remote_object_class_init(ObjectClass *klass, void *data) | ||
204 | +{ | ||
205 | + RemoteObjectClass *k = REMOTE_OBJECT_CLASS(klass); | ||
206 | + | ||
207 | + /* | ||
208 | + * Limit number of supported devices to 1. This is done to avoid devices | ||
209 | + * from one VM accessing the RAM of another VM. This is done until we | ||
210 | + * start using separate address spaces for individual devices. | ||
211 | + */ | ||
212 | + k->max_devs = 1; | ||
213 | + k->nr_devs = 0; | ||
214 | + | ||
215 | + object_class_property_add_str(klass, "fd", NULL, remote_object_set_fd); | ||
216 | + object_class_property_add_str(klass, "devid", NULL, | ||
217 | + remote_object_set_devid); | ||
218 | +} | ||
219 | + | ||
220 | +static const TypeInfo remote_object_info = { | ||
221 | + .name = TYPE_REMOTE_OBJECT, | ||
222 | + .parent = TYPE_OBJECT, | ||
223 | + .instance_size = sizeof(RemoteObject), | ||
224 | + .instance_init = remote_object_init, | ||
225 | + .instance_finalize = remote_object_finalize, | ||
226 | + .class_size = sizeof(RemoteObjectClass), | ||
227 | + .class_init = remote_object_class_init, | ||
228 | + .interfaces = (InterfaceInfo[]) { | ||
229 | + { TYPE_USER_CREATABLE }, | ||
230 | + { } | ||
231 | + } | ||
232 | +}; | ||
233 | + | ||
234 | +static void register_types(void) | ||
235 | +{ | ||
236 | + type_register_static(&remote_object_info); | ||
237 | +} | ||
238 | + | ||
239 | +type_init(register_types); | ||
240 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
44 | index XXXXXXX..XXXXXXX 100644 | 241 | index XXXXXXX..XXXXXXX 100644 |
45 | --- a/block/null.c | 242 | --- a/hw/remote/meson.build |
46 | +++ b/block/null.c | 243 | +++ b/hw/remote/meson.build |
47 | @@ -XXX,XX +XXX,XX @@ static coroutine_fn int null_co_common(BlockDriverState *bs) | 244 | @@ -XXX,XX +XXX,XX @@ remote_ss = ss.source_set() |
48 | BDRVNullState *s = bs->opaque; | 245 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) |
49 | 246 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | |
50 | if (s->latency_ns) { | 247 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) |
51 | - co_aio_sleep_ns(bdrv_get_aio_context(bs), QEMU_CLOCK_REALTIME, | 248 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) |
52 | - s->latency_ns); | 249 | |
53 | + qemu_co_sleep_ns(QEMU_CLOCK_REALTIME, s->latency_ns); | 250 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) |
54 | } | ||
55 | return 0; | ||
56 | } | ||
57 | diff --git a/block/sheepdog.c b/block/sheepdog.c | ||
58 | index XXXXXXX..XXXXXXX 100644 | ||
59 | --- a/block/sheepdog.c | ||
60 | +++ b/block/sheepdog.c | ||
61 | @@ -XXX,XX +XXX,XX @@ static coroutine_fn void reconnect_to_sdog(void *opaque) | ||
62 | if (s->fd < 0) { | ||
63 | DPRINTF("Wait for connection to be established\n"); | ||
64 | error_report_err(local_err); | ||
65 | - co_aio_sleep_ns(bdrv_get_aio_context(s->bs), QEMU_CLOCK_REALTIME, | ||
66 | - 1000000000ULL); | ||
67 | + qemu_co_sleep_ns(QEMU_CLOCK_REALTIME, 1000000000ULL); | ||
68 | } | ||
69 | }; | ||
70 | |||
71 | diff --git a/util/qemu-coroutine-sleep.c b/util/qemu-coroutine-sleep.c | ||
72 | index XXXXXXX..XXXXXXX 100644 | ||
73 | --- a/util/qemu-coroutine-sleep.c | ||
74 | +++ b/util/qemu-coroutine-sleep.c | ||
75 | @@ -XXX,XX +XXX,XX @@ static void co_sleep_cb(void *opaque) | ||
76 | aio_co_wake(sleep_cb->co); | ||
77 | } | ||
78 | |||
79 | -void coroutine_fn co_aio_sleep_ns(AioContext *ctx, QEMUClockType type, | ||
80 | - int64_t ns) | ||
81 | +void coroutine_fn qemu_co_sleep_ns(QEMUClockType type, int64_t ns) | ||
82 | { | ||
83 | + AioContext *ctx = qemu_get_current_aio_context(); | ||
84 | CoSleepCB sleep_cb = { | ||
85 | .co = qemu_coroutine_self(), | ||
86 | }; | ||
87 | -- | 251 | -- |
88 | 2.14.3 | 252 | 2.29.2 |
89 | 253 | ||
90 | diff view generated by jsdifflib |
1 | It is not necessary to hold AioContext across transactions anymore since | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | bdrv_drained_begin/end() is used to keep the nodes quiesced. In fact, | 2 | |
3 | using the AioContext lock for this purpose was always buggy. | 3 | SyncSysMemMsg message format is defined. It is used to send |
4 | 4 | file descriptors of the RAM regions to remote device. | |
5 | This patch reduces the scope of AioContext locked regions. This is not | 5 | RAM on the remote device is configured with a set of file descriptors. |
6 | just a cleanup but also fixes hangs that occur in BDRV_POLL_WHILE() | 6 | Old RAM regions are deleted and new regions, each with an fd, is |
7 | because it is unware of recursive locking and does not release the | 7 | added to the RAM. |
8 | AioContext the necessary number of times to allow progress to be made. | 8 | |
9 | 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | |
10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
11 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 11 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
12 | Reviewed-by: Eric Blake <eblake@redhat.com> | 12 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
13 | Message-id: 20171206144550.22295-3-stefanha@redhat.com | 13 | Message-id: 7d2d1831d812e85f681e7a8ab99e032cf4704689.1611938319.git.jag.raman@oracle.com |
14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
15 | --- | 15 | --- |
16 | blockdev.c | 71 ++++++++++++++++++++++++++++++++++++++++++-------------------- | 16 | MAINTAINERS | 2 + |
17 | 1 file changed, 48 insertions(+), 23 deletions(-) | 17 | include/hw/remote/memory.h | 19 ++++++++++ |
18 | 18 | include/hw/remote/mpqemu-link.h | 10 +++++ | |
19 | diff --git a/blockdev.c b/blockdev.c | 19 | hw/remote/memory.c | 65 +++++++++++++++++++++++++++++++++ |
20 | index XXXXXXX..XXXXXXX 100644 | 20 | hw/remote/mpqemu-link.c | 11 ++++++ |
21 | --- a/blockdev.c | 21 | hw/remote/meson.build | 2 + |
22 | +++ b/blockdev.c | 22 | 6 files changed, 109 insertions(+) |
23 | @@ -XXX,XX +XXX,XX @@ typedef struct ExternalSnapshotState { | 23 | create mode 100644 include/hw/remote/memory.h |
24 | BlkActionState common; | 24 | create mode 100644 hw/remote/memory.c |
25 | BlockDriverState *old_bs; | 25 | |
26 | BlockDriverState *new_bs; | 26 | diff --git a/MAINTAINERS b/MAINTAINERS |
27 | - AioContext *aio_context; | 27 | index XXXXXXX..XXXXXXX 100644 |
28 | bool overlay_appended; | 28 | --- a/MAINTAINERS |
29 | } ExternalSnapshotState; | 29 | +++ b/MAINTAINERS |
30 | 30 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/mpqemu-link.c | |
31 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | 31 | F: include/hw/remote/mpqemu-link.h |
32 | ExternalSnapshotState *state = | 32 | F: hw/remote/message.c |
33 | DO_UPCAST(ExternalSnapshotState, common, common); | 33 | F: hw/remote/remote-obj.c |
34 | TransactionAction *action = common->action; | 34 | +F: include/hw/remote/memory.h |
35 | + AioContext *aio_context; | 35 | +F: hw/remote/memory.c |
36 | 36 | ||
37 | /* 'blockdev-snapshot' and 'blockdev-snapshot-sync' have similar | 37 | Build and test automation |
38 | * purpose but a different set of parameters */ | 38 | ------------------------- |
39 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | 39 | diff --git a/include/hw/remote/memory.h b/include/hw/remote/memory.h |
40 | return; | 40 | new file mode 100644 |
41 | } | 41 | index XXXXXXX..XXXXXXX |
42 | 42 | --- /dev/null | |
43 | - /* Acquire AioContext now so any threads operating on old_bs stop */ | 43 | +++ b/include/hw/remote/memory.h |
44 | - state->aio_context = bdrv_get_aio_context(state->old_bs); | 44 | @@ -XXX,XX +XXX,XX @@ |
45 | - aio_context_acquire(state->aio_context); | 45 | +/* |
46 | + aio_context = bdrv_get_aio_context(state->old_bs); | 46 | + * Memory manager for remote device |
47 | + aio_context_acquire(aio_context); | 47 | + * |
48 | + | 48 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. |
49 | + /* Paired with .clean() */ | 49 | + * |
50 | bdrv_drained_begin(state->old_bs); | 50 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. |
51 | 51 | + * See the COPYING file in the top-level directory. | |
52 | if (!bdrv_is_inserted(state->old_bs)) { | 52 | + * |
53 | error_setg(errp, QERR_DEVICE_HAS_NO_MEDIUM, device); | 53 | + */ |
54 | - return; | 54 | + |
55 | + goto out; | 55 | +#ifndef REMOTE_MEMORY_H |
56 | } | 56 | +#define REMOTE_MEMORY_H |
57 | 57 | + | |
58 | if (bdrv_op_is_blocked(state->old_bs, | 58 | +#include "exec/hwaddr.h" |
59 | BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, errp)) { | 59 | +#include "hw/remote/mpqemu-link.h" |
60 | - return; | 60 | + |
61 | + goto out; | 61 | +void remote_sysmem_reconfig(MPQemuMsg *msg, Error **errp); |
62 | } | 62 | + |
63 | 63 | +#endif | |
64 | if (!bdrv_is_read_only(state->old_bs)) { | 64 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h |
65 | if (bdrv_flush(state->old_bs)) { | 65 | index XXXXXXX..XXXXXXX 100644 |
66 | error_setg(errp, QERR_IO_ERROR); | 66 | --- a/include/hw/remote/mpqemu-link.h |
67 | - return; | 67 | +++ b/include/hw/remote/mpqemu-link.h |
68 | + goto out; | 68 | @@ -XXX,XX +XXX,XX @@ |
69 | #include "qom/object.h" | ||
70 | #include "qemu/thread.h" | ||
71 | #include "io/channel.h" | ||
72 | +#include "exec/hwaddr.h" | ||
73 | |||
74 | #define REMOTE_MAX_FDS 8 | ||
75 | |||
76 | @@ -XXX,XX +XXX,XX @@ | ||
77 | * | ||
78 | */ | ||
79 | typedef enum { | ||
80 | + MPQEMU_CMD_SYNC_SYSMEM, | ||
81 | MPQEMU_CMD_MAX, | ||
82 | } MPQemuCmd; | ||
83 | |||
84 | +typedef struct { | ||
85 | + hwaddr gpas[REMOTE_MAX_FDS]; | ||
86 | + uint64_t sizes[REMOTE_MAX_FDS]; | ||
87 | + off_t offsets[REMOTE_MAX_FDS]; | ||
88 | +} SyncSysmemMsg; | ||
89 | + | ||
90 | /** | ||
91 | * MPQemuMsg: | ||
92 | * @cmd: The remote command | ||
93 | @@ -XXX,XX +XXX,XX @@ typedef enum { | ||
94 | * MPQemuMsg Format of the message sent to the remote device from QEMU. | ||
95 | * | ||
96 | */ | ||
97 | + | ||
98 | typedef struct { | ||
99 | int cmd; | ||
100 | size_t size; | ||
101 | |||
102 | union { | ||
103 | uint64_t u64; | ||
104 | + SyncSysmemMsg sync_sysmem; | ||
105 | } data; | ||
106 | |||
107 | int fds[REMOTE_MAX_FDS]; | ||
108 | diff --git a/hw/remote/memory.c b/hw/remote/memory.c | ||
109 | new file mode 100644 | ||
110 | index XXXXXXX..XXXXXXX | ||
111 | --- /dev/null | ||
112 | +++ b/hw/remote/memory.c | ||
113 | @@ -XXX,XX +XXX,XX @@ | ||
114 | +/* | ||
115 | + * Memory manager for remote device | ||
116 | + * | ||
117 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
118 | + * | ||
119 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
120 | + * See the COPYING file in the top-level directory. | ||
121 | + * | ||
122 | + */ | ||
123 | + | ||
124 | +#include "qemu/osdep.h" | ||
125 | +#include "qemu-common.h" | ||
126 | + | ||
127 | +#include "hw/remote/memory.h" | ||
128 | +#include "exec/address-spaces.h" | ||
129 | +#include "exec/ram_addr.h" | ||
130 | +#include "qapi/error.h" | ||
131 | + | ||
132 | +static void remote_sysmem_reset(void) | ||
133 | +{ | ||
134 | + MemoryRegion *sysmem, *subregion, *next; | ||
135 | + | ||
136 | + sysmem = get_system_memory(); | ||
137 | + | ||
138 | + QTAILQ_FOREACH_SAFE(subregion, &sysmem->subregions, subregions_link, next) { | ||
139 | + if (subregion->ram) { | ||
140 | + memory_region_del_subregion(sysmem, subregion); | ||
141 | + object_unparent(OBJECT(subregion)); | ||
142 | + } | ||
143 | + } | ||
144 | +} | ||
145 | + | ||
146 | +void remote_sysmem_reconfig(MPQemuMsg *msg, Error **errp) | ||
147 | +{ | ||
148 | + ERRP_GUARD(); | ||
149 | + SyncSysmemMsg *sysmem_info = &msg->data.sync_sysmem; | ||
150 | + MemoryRegion *sysmem, *subregion; | ||
151 | + static unsigned int suffix; | ||
152 | + int region; | ||
153 | + | ||
154 | + sysmem = get_system_memory(); | ||
155 | + | ||
156 | + remote_sysmem_reset(); | ||
157 | + | ||
158 | + for (region = 0; region < msg->num_fds; region++) { | ||
159 | + g_autofree char *name; | ||
160 | + subregion = g_new(MemoryRegion, 1); | ||
161 | + name = g_strdup_printf("remote-mem-%u", suffix++); | ||
162 | + memory_region_init_ram_from_fd(subregion, NULL, | ||
163 | + name, sysmem_info->sizes[region], | ||
164 | + true, msg->fds[region], | ||
165 | + sysmem_info->offsets[region], | ||
166 | + errp); | ||
167 | + | ||
168 | + if (*errp) { | ||
169 | + g_free(subregion); | ||
170 | + remote_sysmem_reset(); | ||
171 | + return; | ||
172 | + } | ||
173 | + | ||
174 | + memory_region_add_subregion(sysmem, sysmem_info->gpas[region], | ||
175 | + subregion); | ||
176 | + | ||
177 | + } | ||
178 | +} | ||
179 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
180 | index XXXXXXX..XXXXXXX 100644 | ||
181 | --- a/hw/remote/mpqemu-link.c | ||
182 | +++ b/hw/remote/mpqemu-link.c | ||
183 | @@ -XXX,XX +XXX,XX @@ bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
69 | } | 184 | } |
70 | } | 185 | } |
71 | 186 | ||
72 | if (!bdrv_is_first_non_filter(state->old_bs)) { | 187 | + /* Verify message specific fields. */ |
73 | error_setg(errp, QERR_FEATURE_DISABLED, "snapshot"); | 188 | + switch (msg->cmd) { |
74 | - return; | 189 | + case MPQEMU_CMD_SYNC_SYSMEM: |
75 | + goto out; | 190 | + if (msg->num_fds == 0 || msg->size != sizeof(SyncSysmemMsg)) { |
76 | } | 191 | + return false; |
77 | 192 | + } | |
78 | if (action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_SNAPSHOT_SYNC) { | 193 | + break; |
79 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | 194 | + default: |
80 | 195 | + break; | |
81 | if (node_name && !snapshot_node_name) { | 196 | + } |
82 | error_setg(errp, "New snapshot node name missing"); | 197 | + |
83 | - return; | 198 | return true; |
84 | + goto out; | ||
85 | } | ||
86 | |||
87 | if (snapshot_node_name && | ||
88 | bdrv_lookup_bs(snapshot_node_name, snapshot_node_name, NULL)) { | ||
89 | error_setg(errp, "New snapshot node name already in use"); | ||
90 | - return; | ||
91 | + goto out; | ||
92 | } | ||
93 | |||
94 | flags = state->old_bs->open_flags; | ||
95 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | ||
96 | int64_t size = bdrv_getlength(state->old_bs); | ||
97 | if (size < 0) { | ||
98 | error_setg_errno(errp, -size, "bdrv_getlength failed"); | ||
99 | - return; | ||
100 | + goto out; | ||
101 | } | ||
102 | bdrv_img_create(new_image_file, format, | ||
103 | state->old_bs->filename, | ||
104 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | ||
105 | NULL, size, flags, false, &local_err); | ||
106 | if (local_err) { | ||
107 | error_propagate(errp, local_err); | ||
108 | - return; | ||
109 | + goto out; | ||
110 | } | ||
111 | } | ||
112 | |||
113 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | ||
114 | errp); | ||
115 | /* We will manually add the backing_hd field to the bs later */ | ||
116 | if (!state->new_bs) { | ||
117 | - return; | ||
118 | + goto out; | ||
119 | } | ||
120 | |||
121 | if (bdrv_has_blk(state->new_bs)) { | ||
122 | error_setg(errp, "The snapshot is already in use"); | ||
123 | - return; | ||
124 | + goto out; | ||
125 | } | ||
126 | |||
127 | if (bdrv_op_is_blocked(state->new_bs, BLOCK_OP_TYPE_EXTERNAL_SNAPSHOT, | ||
128 | errp)) { | ||
129 | - return; | ||
130 | + goto out; | ||
131 | } | ||
132 | |||
133 | if (state->new_bs->backing != NULL) { | ||
134 | error_setg(errp, "The snapshot already has a backing image"); | ||
135 | - return; | ||
136 | + goto out; | ||
137 | } | ||
138 | |||
139 | if (!state->new_bs->drv->supports_backing) { | ||
140 | error_setg(errp, "The snapshot does not support backing images"); | ||
141 | - return; | ||
142 | + goto out; | ||
143 | } | ||
144 | |||
145 | - bdrv_set_aio_context(state->new_bs, state->aio_context); | ||
146 | + bdrv_set_aio_context(state->new_bs, aio_context); | ||
147 | |||
148 | /* This removes our old bs and adds the new bs. This is an operation that | ||
149 | * can fail, so we need to do it in .prepare; undoing it for abort is | ||
150 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_prepare(BlkActionState *common, | ||
151 | bdrv_append(state->new_bs, state->old_bs, &local_err); | ||
152 | if (local_err) { | ||
153 | error_propagate(errp, local_err); | ||
154 | - return; | ||
155 | + goto out; | ||
156 | } | ||
157 | state->overlay_appended = true; | ||
158 | + | ||
159 | +out: | ||
160 | + aio_context_release(aio_context); | ||
161 | } | 199 | } |
162 | 200 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | |
163 | static void external_snapshot_commit(BlkActionState *common) | 201 | index XXXXXXX..XXXXXXX 100644 |
164 | { | 202 | --- a/hw/remote/meson.build |
165 | ExternalSnapshotState *state = | 203 | +++ b/hw/remote/meson.build |
166 | DO_UPCAST(ExternalSnapshotState, common, common); | 204 | @@ -XXX,XX +XXX,XX @@ remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) |
167 | + AioContext *aio_context; | 205 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) |
168 | + | 206 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) |
169 | + aio_context = bdrv_get_aio_context(state->old_bs); | 207 | |
170 | + aio_context_acquire(aio_context); | 208 | +specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('memory.c')) |
171 | 209 | + | |
172 | /* We don't need (or want) to use the transactional | 210 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) |
173 | * bdrv_reopen_multiple() across all the entries at once, because we | ||
174 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_commit(BlkActionState *common) | ||
175 | bdrv_reopen(state->old_bs, state->old_bs->open_flags & ~BDRV_O_RDWR, | ||
176 | NULL); | ||
177 | } | ||
178 | + | ||
179 | + aio_context_release(aio_context); | ||
180 | } | ||
181 | |||
182 | static void external_snapshot_abort(BlkActionState *common) | ||
183 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_abort(BlkActionState *common) | ||
184 | DO_UPCAST(ExternalSnapshotState, common, common); | ||
185 | if (state->new_bs) { | ||
186 | if (state->overlay_appended) { | ||
187 | + AioContext *aio_context; | ||
188 | + | ||
189 | + aio_context = bdrv_get_aio_context(state->old_bs); | ||
190 | + aio_context_acquire(aio_context); | ||
191 | + | ||
192 | bdrv_ref(state->old_bs); /* we can't let bdrv_set_backind_hd() | ||
193 | close state->old_bs; we need it */ | ||
194 | bdrv_set_backing_hd(state->new_bs, NULL, &error_abort); | ||
195 | bdrv_replace_node(state->new_bs, state->old_bs, &error_abort); | ||
196 | bdrv_unref(state->old_bs); /* bdrv_replace_node() ref'ed old_bs */ | ||
197 | + | ||
198 | + aio_context_release(aio_context); | ||
199 | } | ||
200 | } | ||
201 | } | ||
202 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_clean(BlkActionState *common) | ||
203 | { | ||
204 | ExternalSnapshotState *state = | ||
205 | DO_UPCAST(ExternalSnapshotState, common, common); | ||
206 | - if (state->aio_context) { | ||
207 | - bdrv_drained_end(state->old_bs); | ||
208 | - bdrv_unref(state->new_bs); | ||
209 | - aio_context_release(state->aio_context); | ||
210 | + AioContext *aio_context; | ||
211 | + | ||
212 | + if (!state->old_bs) { | ||
213 | + return; | ||
214 | } | ||
215 | + | ||
216 | + aio_context = bdrv_get_aio_context(state->old_bs); | ||
217 | + aio_context_acquire(aio_context); | ||
218 | + | ||
219 | + bdrv_drained_end(state->old_bs); | ||
220 | + bdrv_unref(state->new_bs); | ||
221 | + | ||
222 | + aio_context_release(aio_context); | ||
223 | } | ||
224 | |||
225 | typedef struct DriveBackupState { | ||
226 | -- | 211 | -- |
227 | 2.14.3 | 212 | 2.29.2 |
228 | 213 | ||
229 | diff view generated by jsdifflib |
1 | Currently there is no easy way for iotests to ensure that a BDS is bound | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | to a particular IOThread. Normally the virtio-blk device calls | ||
3 | blk_set_aio_context() when dataplane is enabled during guest driver | ||
4 | initialization. This never happens in iotests since -machine | ||
5 | accel=qtest means there is no guest activity (including device driver | ||
6 | initialization). | ||
7 | 2 | ||
8 | This patch adds a QMP command to explicitly assign IOThreads in test | 3 | Defines a PCI Device proxy object as a child of TYPE_PCI_DEVICE. |
9 | cases. See qapi/block-core.json for a description of the command. | ||
10 | 4 | ||
11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 5 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
12 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 6 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> |
13 | Reviewed-by: Eric Blake <eblake@redhat.com> | 7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
14 | Message-id: 20171206144550.22295-9-stefanha@redhat.com | 8 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
9 | Message-id: b5186ebfedf8e557044d09a768846c59230ad3a7.1611938319.git.jag.raman@oracle.com | ||
15 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
16 | --- | 11 | --- |
17 | qapi/block-core.json | 36 ++++++++++++++++++++++++++++++++++++ | 12 | MAINTAINERS | 2 + |
18 | blockdev.c | 41 +++++++++++++++++++++++++++++++++++++++++ | 13 | include/hw/remote/proxy.h | 33 +++++++++++++ |
19 | 2 files changed, 77 insertions(+) | 14 | hw/remote/proxy.c | 99 +++++++++++++++++++++++++++++++++++++++ |
15 | hw/remote/meson.build | 1 + | ||
16 | 4 files changed, 135 insertions(+) | ||
17 | create mode 100644 include/hw/remote/proxy.h | ||
18 | create mode 100644 hw/remote/proxy.c | ||
20 | 19 | ||
21 | diff --git a/qapi/block-core.json b/qapi/block-core.json | 20 | diff --git a/MAINTAINERS b/MAINTAINERS |
22 | index XXXXXXX..XXXXXXX 100644 | 21 | index XXXXXXX..XXXXXXX 100644 |
23 | --- a/qapi/block-core.json | 22 | --- a/MAINTAINERS |
24 | +++ b/qapi/block-core.json | 23 | +++ b/MAINTAINERS |
24 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/message.c | ||
25 | F: hw/remote/remote-obj.c | ||
26 | F: include/hw/remote/memory.h | ||
27 | F: hw/remote/memory.c | ||
28 | +F: hw/remote/proxy.c | ||
29 | +F: include/hw/remote/proxy.h | ||
30 | |||
31 | Build and test automation | ||
32 | ------------------------- | ||
33 | diff --git a/include/hw/remote/proxy.h b/include/hw/remote/proxy.h | ||
34 | new file mode 100644 | ||
35 | index XXXXXXX..XXXXXXX | ||
36 | --- /dev/null | ||
37 | +++ b/include/hw/remote/proxy.h | ||
25 | @@ -XXX,XX +XXX,XX @@ | 38 | @@ -XXX,XX +XXX,XX @@ |
26 | 'data' : { 'parent': 'str', | 39 | +/* |
27 | '*child': 'str', | 40 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. |
28 | '*node': 'str' } } | 41 | + * |
42 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
43 | + * See the COPYING file in the top-level directory. | ||
44 | + * | ||
45 | + */ | ||
29 | + | 46 | + |
30 | +## | 47 | +#ifndef PROXY_H |
31 | +# @x-blockdev-set-iothread: | 48 | +#define PROXY_H |
32 | +# | 49 | + |
33 | +# Move @node and its children into the @iothread. If @iothread is null then | 50 | +#include "hw/pci/pci.h" |
34 | +# move @node and its children into the main loop. | 51 | +#include "io/channel.h" |
35 | +# | 52 | + |
36 | +# The node must not be attached to a BlockBackend. | 53 | +#define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev" |
37 | +# | 54 | +OBJECT_DECLARE_SIMPLE_TYPE(PCIProxyDev, PCI_PROXY_DEV) |
38 | +# @node-name: the name of the block driver node | 55 | + |
39 | +# | 56 | +struct PCIProxyDev { |
40 | +# @iothread: the name of the IOThread object or null for the main loop | 57 | + PCIDevice parent_dev; |
41 | +# | 58 | + char *fd; |
42 | +# Note: this command is experimental and intended for test cases that need | 59 | + |
43 | +# control over IOThreads only. | 60 | + /* |
44 | +# | 61 | + * Mutex used to protect the QIOChannel fd from |
45 | +# Since: 2.12 | 62 | + * the concurrent access by the VCPUs since proxy |
46 | +# | 63 | + * blocks while awaiting for the replies from the |
47 | +# Example: | 64 | + * process remote. |
48 | +# | 65 | + */ |
49 | +# 1. Move a node into an IOThread | 66 | + QemuMutex io_mutex; |
50 | +# -> { "execute": "x-blockdev-set-iothread", | 67 | + QIOChannel *ioc; |
51 | +# "arguments": { "node-name": "disk1", | 68 | + Error *migration_blocker; |
52 | +# "iothread": "iothread0" } } | 69 | +}; |
53 | +# <- { "return": {} } | 70 | + |
54 | +# | 71 | +#endif /* PROXY_H */ |
55 | +# 2. Move a node into the main loop | 72 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c |
56 | +# -> { "execute": "x-blockdev-set-iothread", | 73 | new file mode 100644 |
57 | +# "arguments": { "node-name": "disk1", | 74 | index XXXXXXX..XXXXXXX |
58 | +# "iothread": null } } | 75 | --- /dev/null |
59 | +# <- { "return": {} } | 76 | +++ b/hw/remote/proxy.c |
60 | +# | ||
61 | +## | ||
62 | +{ 'command': 'x-blockdev-set-iothread', | ||
63 | + 'data' : { 'node-name': 'str', | ||
64 | + 'iothread': 'StrOrNull' } } | ||
65 | diff --git a/blockdev.c b/blockdev.c | ||
66 | index XXXXXXX..XXXXXXX 100644 | ||
67 | --- a/blockdev.c | ||
68 | +++ b/blockdev.c | ||
69 | @@ -XXX,XX +XXX,XX @@ | 77 | @@ -XXX,XX +XXX,XX @@ |
70 | #include "qapi/qmp/qerror.h" | 78 | +/* |
71 | #include "qapi/qobject-output-visitor.h" | 79 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. |
72 | #include "sysemu/sysemu.h" | 80 | + * |
73 | +#include "sysemu/iothread.h" | 81 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. |
74 | #include "block/block_int.h" | 82 | + * See the COPYING file in the top-level directory. |
75 | #include "qmp-commands.h" | 83 | + * |
76 | #include "block/trace.h" | 84 | + */ |
77 | @@ -XXX,XX +XXX,XX @@ BlockJobInfoList *qmp_query_block_jobs(Error **errp) | 85 | + |
78 | return head; | 86 | +#include "qemu/osdep.h" |
79 | } | 87 | +#include "qemu-common.h" |
80 | 88 | + | |
81 | +void qmp_x_blockdev_set_iothread(const char *node_name, StrOrNull *iothread, | 89 | +#include "hw/remote/proxy.h" |
82 | + Error **errp) | 90 | +#include "hw/pci/pci.h" |
91 | +#include "qapi/error.h" | ||
92 | +#include "io/channel-util.h" | ||
93 | +#include "hw/qdev-properties.h" | ||
94 | +#include "monitor/monitor.h" | ||
95 | +#include "migration/blocker.h" | ||
96 | +#include "qemu/sockets.h" | ||
97 | + | ||
98 | +static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
83 | +{ | 99 | +{ |
84 | + AioContext *old_context; | 100 | + ERRP_GUARD(); |
85 | + AioContext *new_context; | 101 | + PCIProxyDev *dev = PCI_PROXY_DEV(device); |
86 | + BlockDriverState *bs; | 102 | + int fd; |
87 | + | 103 | + |
88 | + bs = bdrv_find_node(node_name); | 104 | + if (!dev->fd) { |
89 | + if (!bs) { | 105 | + error_setg(errp, "fd parameter not specified for %s", |
90 | + error_setg(errp, "Cannot find node %s", node_name); | 106 | + DEVICE(device)->id); |
91 | + return; | 107 | + return; |
92 | + } | 108 | + } |
93 | + | 109 | + |
94 | + /* If we want to allow more extreme test scenarios this guard could be | 110 | + fd = monitor_fd_param(monitor_cur(), dev->fd, errp); |
95 | + * removed. For now it protects against accidents. */ | 111 | + if (fd == -1) { |
96 | + if (bdrv_has_blk(bs)) { | 112 | + error_prepend(errp, "proxy: unable to parse fd %s: ", dev->fd); |
97 | + error_setg(errp, "Node %s is in use", node_name); | ||
98 | + return; | 113 | + return; |
99 | + } | 114 | + } |
100 | + | 115 | + |
101 | + if (iothread->type == QTYPE_QSTRING) { | 116 | + if (!fd_is_socket(fd)) { |
102 | + IOThread *obj = iothread_by_id(iothread->u.s); | 117 | + error_setg(errp, "proxy: fd %d is not a socket", fd); |
103 | + if (!obj) { | 118 | + close(fd); |
104 | + error_setg(errp, "Cannot find iothread %s", iothread->u.s); | 119 | + return; |
105 | + return; | ||
106 | + } | ||
107 | + | ||
108 | + new_context = iothread_get_aio_context(obj); | ||
109 | + } else { | ||
110 | + new_context = qemu_get_aio_context(); | ||
111 | + } | 120 | + } |
112 | + | 121 | + |
113 | + old_context = bdrv_get_aio_context(bs); | 122 | + dev->ioc = qio_channel_new_fd(fd, errp); |
114 | + aio_context_acquire(old_context); | ||
115 | + | 123 | + |
116 | + bdrv_set_aio_context(bs, new_context); | 124 | + error_setg(&dev->migration_blocker, "%s does not support migration", |
125 | + TYPE_PCI_PROXY_DEV); | ||
126 | + migrate_add_blocker(dev->migration_blocker, errp); | ||
117 | + | 127 | + |
118 | + aio_context_release(old_context); | 128 | + qemu_mutex_init(&dev->io_mutex); |
129 | + qio_channel_set_blocking(dev->ioc, true, NULL); | ||
119 | +} | 130 | +} |
120 | + | 131 | + |
121 | QemuOptsList qemu_common_drive_opts = { | 132 | +static void pci_proxy_dev_exit(PCIDevice *pdev) |
122 | .name = "drive", | 133 | +{ |
123 | .head = QTAILQ_HEAD_INITIALIZER(qemu_common_drive_opts.head), | 134 | + PCIProxyDev *dev = PCI_PROXY_DEV(pdev); |
135 | + | ||
136 | + if (dev->ioc) { | ||
137 | + qio_channel_close(dev->ioc, NULL); | ||
138 | + } | ||
139 | + | ||
140 | + migrate_del_blocker(dev->migration_blocker); | ||
141 | + | ||
142 | + error_free(dev->migration_blocker); | ||
143 | +} | ||
144 | + | ||
145 | +static Property proxy_properties[] = { | ||
146 | + DEFINE_PROP_STRING("fd", PCIProxyDev, fd), | ||
147 | + DEFINE_PROP_END_OF_LIST(), | ||
148 | +}; | ||
149 | + | ||
150 | +static void pci_proxy_dev_class_init(ObjectClass *klass, void *data) | ||
151 | +{ | ||
152 | + DeviceClass *dc = DEVICE_CLASS(klass); | ||
153 | + PCIDeviceClass *k = PCI_DEVICE_CLASS(klass); | ||
154 | + | ||
155 | + k->realize = pci_proxy_dev_realize; | ||
156 | + k->exit = pci_proxy_dev_exit; | ||
157 | + device_class_set_props(dc, proxy_properties); | ||
158 | +} | ||
159 | + | ||
160 | +static const TypeInfo pci_proxy_dev_type_info = { | ||
161 | + .name = TYPE_PCI_PROXY_DEV, | ||
162 | + .parent = TYPE_PCI_DEVICE, | ||
163 | + .instance_size = sizeof(PCIProxyDev), | ||
164 | + .class_init = pci_proxy_dev_class_init, | ||
165 | + .interfaces = (InterfaceInfo[]) { | ||
166 | + { INTERFACE_CONVENTIONAL_PCI_DEVICE }, | ||
167 | + { }, | ||
168 | + }, | ||
169 | +}; | ||
170 | + | ||
171 | +static void pci_proxy_dev_register_types(void) | ||
172 | +{ | ||
173 | + type_register_static(&pci_proxy_dev_type_info); | ||
174 | +} | ||
175 | + | ||
176 | +type_init(pci_proxy_dev_register_types) | ||
177 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
178 | index XXXXXXX..XXXXXXX 100644 | ||
179 | --- a/hw/remote/meson.build | ||
180 | +++ b/hw/remote/meson.build | ||
181 | @@ -XXX,XX +XXX,XX @@ remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) | ||
182 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
183 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) | ||
184 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) | ||
185 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy.c')) | ||
186 | |||
187 | specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('memory.c')) | ||
188 | |||
124 | -- | 189 | -- |
125 | 2.14.3 | 190 | 2.29.2 |
126 | 191 | ||
127 | diff view generated by jsdifflib |
1 | There is a small chance that iothread_stop() hangs as follows: | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | Thread 3 (Thread 0x7f63eba5f700 (LWP 16105)): | 3 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
4 | #0 0x00007f64012c09b6 in ppoll () at /lib64/libc.so.6 | 4 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> |
5 | #1 0x000055959992eac9 in ppoll (__ss=0x0, __timeout=0x0, __nfds=<optimized out>, __fds=<optimized out>) at /usr/include/bits/poll2.h:77 | 5 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
6 | #2 0x000055959992eac9 in qemu_poll_ns (fds=<optimized out>, nfds=<optimized out>, timeout=<optimized out>) at util/qemu-timer.c:322 | 6 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
7 | #3 0x0000559599930711 in aio_poll (ctx=0x55959bdb83c0, blocking=blocking@entry=true) at util/aio-posix.c:629 | 7 | Message-id: d54edb4176361eed86b903e8f27058363b6c83b3.1611938319.git.jag.raman@oracle.com |
8 | #4 0x00005595996806fe in iothread_run (opaque=0x55959bd78400) at iothread.c:59 | ||
9 | #5 0x00007f640159f609 in start_thread () at /lib64/libpthread.so.0 | ||
10 | #6 0x00007f64012cce6f in clone () at /lib64/libc.so.6 | ||
11 | |||
12 | Thread 1 (Thread 0x7f640b45b280 (LWP 16103)): | ||
13 | #0 0x00007f64015a0b6d in pthread_join () at /lib64/libpthread.so.0 | ||
14 | #1 0x00005595999332ef in qemu_thread_join (thread=<optimized out>) at util/qemu-thread-posix.c:547 | ||
15 | #2 0x00005595996808ae in iothread_stop (iothread=<optimized out>) at iothread.c:91 | ||
16 | #3 0x000055959968094d in iothread_stop_iter (object=<optimized out>, opaque=<optimized out>) at iothread.c:102 | ||
17 | #4 0x0000559599857d97 in do_object_child_foreach (obj=obj@entry=0x55959bdb8100, fn=fn@entry=0x559599680930 <iothread_stop_iter>, opaque=opaque@entry=0x0, recurse=recurse@entry=false) at qom/object.c:852 | ||
18 | #5 0x0000559599859477 in object_child_foreach (obj=obj@entry=0x55959bdb8100, fn=fn@entry=0x559599680930 <iothread_stop_iter>, opaque=opaque@entry=0x0) at qom/object.c:867 | ||
19 | #6 0x0000559599680a6e in iothread_stop_all () at iothread.c:341 | ||
20 | #7 0x000055959955b1d5 in main (argc=<optimized out>, argv=<optimized out>, envp=<optimized out>) at vl.c:4913 | ||
21 | |||
22 | The relevant code from iothread_run() is: | ||
23 | |||
24 | while (!atomic_read(&iothread->stopping)) { | ||
25 | aio_poll(iothread->ctx, true); | ||
26 | |||
27 | and iothread_stop(): | ||
28 | |||
29 | iothread->stopping = true; | ||
30 | aio_notify(iothread->ctx); | ||
31 | ... | ||
32 | qemu_thread_join(&iothread->thread); | ||
33 | |||
34 | The following scenario can occur: | ||
35 | |||
36 | 1. IOThread: | ||
37 | while (!atomic_read(&iothread->stopping)) -> stopping=false | ||
38 | |||
39 | 2. Main loop: | ||
40 | iothread->stopping = true; | ||
41 | aio_notify(iothread->ctx); | ||
42 | |||
43 | 3. IOThread: | ||
44 | aio_poll(iothread->ctx, true); -> hang | ||
45 | |||
46 | The bug is explained by the AioContext->notify_me doc comments: | ||
47 | |||
48 | "If this field is 0, everything (file descriptors, bottom halves, | ||
49 | timers) will be re-evaluated before the next blocking poll(), thus the | ||
50 | event_notifier_set call can be skipped." | ||
51 | |||
52 | The problem is that "everything" does not include checking | ||
53 | iothread->stopping. This means iothread_run() will block in aio_poll() | ||
54 | if aio_notify() was called just before aio_poll(). | ||
55 | |||
56 | This patch fixes the hang by replacing aio_notify() with | ||
57 | aio_bh_schedule_oneshot(). This makes aio_poll() or g_main_loop_run() | ||
58 | to return. | ||
59 | |||
60 | Implementing this properly required a new bool running flag. The new | ||
61 | flag prevents races that are tricky if we try to use iothread->stopping. | ||
62 | Now iothread->stopping is purely for iothread_stop() and | ||
63 | iothread->running is purely for the iothread_run() thread. | ||
64 | |||
65 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
66 | Reviewed-by: Eric Blake <eblake@redhat.com> | ||
67 | Message-id: 20171207201320.19284-6-stefanha@redhat.com | ||
68 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 8 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
69 | --- | 9 | --- |
70 | include/sysemu/iothread.h | 3 ++- | 10 | include/hw/remote/mpqemu-link.h | 4 ++++ |
71 | iothread.c | 20 +++++++++++++++----- | 11 | hw/remote/mpqemu-link.c | 34 +++++++++++++++++++++++++++++++++ |
72 | 2 files changed, 17 insertions(+), 6 deletions(-) | 12 | 2 files changed, 38 insertions(+) |
73 | 13 | ||
74 | diff --git a/include/sysemu/iothread.h b/include/sysemu/iothread.h | 14 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h |
75 | index XXXXXXX..XXXXXXX 100644 | 15 | index XXXXXXX..XXXXXXX 100644 |
76 | --- a/include/sysemu/iothread.h | 16 | --- a/include/hw/remote/mpqemu-link.h |
77 | +++ b/include/sysemu/iothread.h | 17 | +++ b/include/hw/remote/mpqemu-link.h |
18 | @@ -XXX,XX +XXX,XX @@ | ||
19 | #include "qemu/thread.h" | ||
20 | #include "io/channel.h" | ||
21 | #include "exec/hwaddr.h" | ||
22 | +#include "io/channel-socket.h" | ||
23 | +#include "hw/remote/proxy.h" | ||
24 | |||
25 | #define REMOTE_MAX_FDS 8 | ||
26 | |||
78 | @@ -XXX,XX +XXX,XX @@ typedef struct { | 27 | @@ -XXX,XX +XXX,XX @@ typedef struct { |
79 | GOnce once; | 28 | bool mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); |
80 | QemuMutex init_done_lock; | 29 | bool mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); |
81 | QemuCond init_done_cond; /* is thread initialization done? */ | 30 | |
82 | - bool stopping; | 31 | +uint64_t mpqemu_msg_send_and_await_reply(MPQemuMsg *msg, PCIProxyDev *pdev, |
83 | + bool stopping; /* has iothread_stop() been called? */ | 32 | + Error **errp); |
84 | + bool running; /* should iothread_run() continue? */ | 33 | bool mpqemu_msg_valid(MPQemuMsg *msg); |
85 | int thread_id; | 34 | |
86 | 35 | #endif | |
87 | /* AioContext poll parameters */ | 36 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c |
88 | diff --git a/iothread.c b/iothread.c | ||
89 | index XXXXXXX..XXXXXXX 100644 | 37 | index XXXXXXX..XXXXXXX 100644 |
90 | --- a/iothread.c | 38 | --- a/hw/remote/mpqemu-link.c |
91 | +++ b/iothread.c | 39 | +++ b/hw/remote/mpqemu-link.c |
92 | @@ -XXX,XX +XXX,XX @@ static void *iothread_run(void *opaque) | 40 | @@ -XXX,XX +XXX,XX @@ fail: |
93 | qemu_cond_signal(&iothread->init_done_cond); | 41 | return ret; |
94 | qemu_mutex_unlock(&iothread->init_done_lock); | ||
95 | |||
96 | - while (!atomic_read(&iothread->stopping)) { | ||
97 | + while (iothread->running) { | ||
98 | aio_poll(iothread->ctx, true); | ||
99 | |||
100 | if (atomic_read(&iothread->worker_context)) { | ||
101 | @@ -XXX,XX +XXX,XX @@ static void *iothread_run(void *opaque) | ||
102 | return NULL; | ||
103 | } | 42 | } |
104 | 43 | ||
105 | +/* Runs in iothread_run() thread */ | 44 | +/* |
106 | +static void iothread_stop_bh(void *opaque) | 45 | + * Send msg and wait for a reply with command code RET_MSG. |
46 | + * Returns the message received of size u64 or UINT64_MAX | ||
47 | + * on error. | ||
48 | + * Called from VCPU thread in non-coroutine context. | ||
49 | + * Used by the Proxy object to communicate to remote processes. | ||
50 | + */ | ||
51 | +uint64_t mpqemu_msg_send_and_await_reply(MPQemuMsg *msg, PCIProxyDev *pdev, | ||
52 | + Error **errp) | ||
107 | +{ | 53 | +{ |
108 | + IOThread *iothread = opaque; | 54 | + ERRP_GUARD(); |
55 | + MPQemuMsg msg_reply = {0}; | ||
56 | + uint64_t ret = UINT64_MAX; | ||
109 | + | 57 | + |
110 | + iothread->running = false; /* stop iothread_run() */ | 58 | + assert(!qemu_in_coroutine()); |
111 | + | 59 | + |
112 | + if (iothread->main_loop) { | 60 | + QEMU_LOCK_GUARD(&pdev->io_mutex); |
113 | + g_main_loop_quit(iothread->main_loop); | 61 | + if (!mpqemu_msg_send(msg, pdev->ioc, errp)) { |
62 | + return ret; | ||
114 | + } | 63 | + } |
64 | + | ||
65 | + if (!mpqemu_msg_recv(&msg_reply, pdev->ioc, errp)) { | ||
66 | + return ret; | ||
67 | + } | ||
68 | + | ||
69 | + if (!mpqemu_msg_valid(&msg_reply)) { | ||
70 | + error_setg(errp, "ERROR: Invalid reply received for command %d", | ||
71 | + msg->cmd); | ||
72 | + return ret; | ||
73 | + } | ||
74 | + | ||
75 | + return msg_reply.data.u64; | ||
115 | +} | 76 | +} |
116 | + | 77 | + |
117 | void iothread_stop(IOThread *iothread) | 78 | bool mpqemu_msg_valid(MPQemuMsg *msg) |
118 | { | 79 | { |
119 | if (!iothread->ctx || iothread->stopping) { | 80 | if (msg->cmd >= MPQEMU_CMD_MAX && msg->cmd < 0) { |
120 | return; | ||
121 | } | ||
122 | iothread->stopping = true; | ||
123 | - aio_notify(iothread->ctx); | ||
124 | - if (atomic_read(&iothread->main_loop)) { | ||
125 | - g_main_loop_quit(iothread->main_loop); | ||
126 | - } | ||
127 | + aio_bh_schedule_oneshot(iothread->ctx, iothread_stop_bh, iothread); | ||
128 | qemu_thread_join(&iothread->thread); | ||
129 | } | ||
130 | |||
131 | @@ -XXX,XX +XXX,XX @@ static void iothread_complete(UserCreatable *obj, Error **errp) | ||
132 | char *name, *thread_name; | ||
133 | |||
134 | iothread->stopping = false; | ||
135 | + iothread->running = true; | ||
136 | iothread->thread_id = -1; | ||
137 | iothread->ctx = aio_context_new(&local_error); | ||
138 | if (!iothread->ctx) { | ||
139 | -- | 81 | -- |
140 | 2.14.3 | 82 | 2.29.2 |
141 | 83 | ||
142 | diff view generated by jsdifflib |
1 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 2 | |
3 | Reviewed-by: Eric Blake <eblake@redhat.com> | 3 | The Proxy Object sends the PCI config space accesses as messages |
4 | Message-id: 20171206144550.22295-5-stefanha@redhat.com | 4 | to the remote process over the communication channel |
5 | |||
6 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
7 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
10 | Message-id: d3c94f4618813234655356c60e6f0d0362ff42d6.1611938319.git.jag.raman@oracle.com | ||
5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
6 | --- | 12 | --- |
7 | blockdev.c | 44 ++++++++++++++++++++++++++++++++++---------- | 13 | include/hw/remote/mpqemu-link.h | 10 ++++++ |
8 | 1 file changed, 34 insertions(+), 10 deletions(-) | 14 | hw/remote/message.c | 60 +++++++++++++++++++++++++++++++++ |
9 | 15 | hw/remote/mpqemu-link.c | 8 ++++- | |
10 | diff --git a/blockdev.c b/blockdev.c | 16 | hw/remote/proxy.c | 55 ++++++++++++++++++++++++++++++ |
11 | index XXXXXXX..XXXXXXX 100644 | 17 | 4 files changed, 132 insertions(+), 1 deletion(-) |
12 | --- a/blockdev.c | 18 | |
13 | +++ b/blockdev.c | 19 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h |
14 | @@ -XXX,XX +XXX,XX @@ typedef struct BlockdevBackupState { | 20 | index XXXXXXX..XXXXXXX 100644 |
15 | BlkActionState common; | 21 | --- a/include/hw/remote/mpqemu-link.h |
16 | BlockDriverState *bs; | 22 | +++ b/include/hw/remote/mpqemu-link.h |
17 | BlockJob *job; | 23 | @@ -XXX,XX +XXX,XX @@ |
18 | - AioContext *aio_context; | 24 | */ |
19 | } BlockdevBackupState; | 25 | typedef enum { |
20 | 26 | MPQEMU_CMD_SYNC_SYSMEM, | |
21 | static BlockJob *do_blockdev_backup(BlockdevBackup *backup, BlockJobTxn *txn, | 27 | + MPQEMU_CMD_RET, |
22 | @@ -XXX,XX +XXX,XX @@ static void blockdev_backup_prepare(BlkActionState *common, Error **errp) | 28 | + MPQEMU_CMD_PCI_CFGWRITE, |
23 | BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common); | 29 | + MPQEMU_CMD_PCI_CFGREAD, |
24 | BlockdevBackup *backup; | 30 | MPQEMU_CMD_MAX, |
25 | BlockDriverState *bs, *target; | 31 | } MPQemuCmd; |
26 | + AioContext *aio_context; | 32 | |
27 | Error *local_err = NULL; | 33 | @@ -XXX,XX +XXX,XX @@ typedef struct { |
28 | 34 | off_t offsets[REMOTE_MAX_FDS]; | |
29 | assert(common->action->type == TRANSACTION_ACTION_KIND_BLOCKDEV_BACKUP); | 35 | } SyncSysmemMsg; |
30 | @@ -XXX,XX +XXX,XX @@ static void blockdev_backup_prepare(BlkActionState *common, Error **errp) | 36 | |
31 | return; | 37 | +typedef struct { |
32 | } | 38 | + uint32_t addr; |
33 | 39 | + uint32_t val; | |
34 | - /* AioContext is released in .clean() */ | 40 | + int len; |
35 | - state->aio_context = bdrv_get_aio_context(bs); | 41 | +} PciConfDataMsg; |
36 | - if (state->aio_context != bdrv_get_aio_context(target)) { | 42 | + |
37 | - state->aio_context = NULL; | 43 | /** |
38 | + aio_context = bdrv_get_aio_context(bs); | 44 | * MPQemuMsg: |
39 | + if (aio_context != bdrv_get_aio_context(target)) { | 45 | * @cmd: The remote command |
40 | error_setg(errp, "Backup between two IO threads is not implemented"); | 46 | @@ -XXX,XX +XXX,XX @@ typedef struct { |
41 | return; | 47 | |
42 | } | 48 | union { |
43 | - aio_context_acquire(state->aio_context); | 49 | uint64_t u64; |
44 | + aio_context_acquire(aio_context); | 50 | + PciConfDataMsg pci_conf_data; |
45 | state->bs = bs; | 51 | SyncSysmemMsg sync_sysmem; |
46 | + | 52 | } data; |
47 | + /* Paired with .clean() */ | 53 | |
48 | bdrv_drained_begin(state->bs); | 54 | diff --git a/hw/remote/message.c b/hw/remote/message.c |
49 | 55 | index XXXXXXX..XXXXXXX 100644 | |
50 | state->job = do_blockdev_backup(backup, common->block_job_txn, &local_err); | 56 | --- a/hw/remote/message.c |
51 | if (local_err) { | 57 | +++ b/hw/remote/message.c |
52 | error_propagate(errp, local_err); | 58 | @@ -XXX,XX +XXX,XX @@ |
53 | - return; | 59 | #include "hw/remote/mpqemu-link.h" |
54 | + goto out; | 60 | #include "qapi/error.h" |
55 | } | 61 | #include "sysemu/runstate.h" |
56 | + | 62 | +#include "hw/pci/pci.h" |
57 | +out: | 63 | + |
58 | + aio_context_release(aio_context); | 64 | +static void process_config_write(QIOChannel *ioc, PCIDevice *dev, |
59 | } | 65 | + MPQemuMsg *msg, Error **errp); |
60 | 66 | +static void process_config_read(QIOChannel *ioc, PCIDevice *dev, | |
61 | static void blockdev_backup_commit(BlkActionState *common) | 67 | + MPQemuMsg *msg, Error **errp); |
68 | |||
69 | void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
62 | { | 70 | { |
63 | BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common); | 71 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) |
64 | + AioContext *aio_context; | 72 | } |
65 | + | 73 | |
66 | + aio_context = bdrv_get_aio_context(state->bs); | 74 | switch (msg.cmd) { |
67 | + aio_context_acquire(aio_context); | 75 | + case MPQEMU_CMD_PCI_CFGWRITE: |
68 | + | 76 | + process_config_write(com->ioc, pci_dev, &msg, &local_err); |
69 | assert(state->job); | 77 | + break; |
70 | block_job_start(state->job); | 78 | + case MPQEMU_CMD_PCI_CFGREAD: |
71 | + | 79 | + process_config_read(com->ioc, pci_dev, &msg, &local_err); |
72 | + aio_context_release(aio_context); | 80 | + break; |
73 | } | 81 | default: |
74 | 82 | error_setg(&local_err, | |
75 | static void blockdev_backup_abort(BlkActionState *common) | 83 | "Unknown command (%d) received for device %s" |
76 | @@ -XXX,XX +XXX,XX @@ static void blockdev_backup_abort(BlkActionState *common) | 84 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) |
77 | BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common); | 85 | qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN); |
78 | |||
79 | if (state->job) { | ||
80 | + AioContext *aio_context; | ||
81 | + | ||
82 | + aio_context = bdrv_get_aio_context(state->bs); | ||
83 | + aio_context_acquire(aio_context); | ||
84 | + | ||
85 | block_job_cancel_sync(state->job); | ||
86 | + | ||
87 | + aio_context_release(aio_context); | ||
88 | } | 86 | } |
89 | } | 87 | } |
90 | 88 | + | |
91 | static void blockdev_backup_clean(BlkActionState *common) | 89 | +static void process_config_write(QIOChannel *ioc, PCIDevice *dev, |
90 | + MPQemuMsg *msg, Error **errp) | ||
91 | +{ | ||
92 | + ERRP_GUARD(); | ||
93 | + PciConfDataMsg *conf = (PciConfDataMsg *)&msg->data.pci_conf_data; | ||
94 | + MPQemuMsg ret = { 0 }; | ||
95 | + | ||
96 | + if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) { | ||
97 | + error_setg(errp, "Bad address for PCI config write, pid "FMT_pid".", | ||
98 | + getpid()); | ||
99 | + ret.data.u64 = UINT64_MAX; | ||
100 | + } else { | ||
101 | + pci_default_write_config(dev, conf->addr, conf->val, conf->len); | ||
102 | + } | ||
103 | + | ||
104 | + ret.cmd = MPQEMU_CMD_RET; | ||
105 | + ret.size = sizeof(ret.data.u64); | ||
106 | + | ||
107 | + if (!mpqemu_msg_send(&ret, ioc, NULL)) { | ||
108 | + error_prepend(errp, "Error returning code to proxy, pid "FMT_pid": ", | ||
109 | + getpid()); | ||
110 | + } | ||
111 | +} | ||
112 | + | ||
113 | +static void process_config_read(QIOChannel *ioc, PCIDevice *dev, | ||
114 | + MPQemuMsg *msg, Error **errp) | ||
115 | +{ | ||
116 | + ERRP_GUARD(); | ||
117 | + PciConfDataMsg *conf = (PciConfDataMsg *)&msg->data.pci_conf_data; | ||
118 | + MPQemuMsg ret = { 0 }; | ||
119 | + | ||
120 | + if ((conf->addr + sizeof(conf->val)) > pci_config_size(dev)) { | ||
121 | + error_setg(errp, "Bad address for PCI config read, pid "FMT_pid".", | ||
122 | + getpid()); | ||
123 | + ret.data.u64 = UINT64_MAX; | ||
124 | + } else { | ||
125 | + ret.data.u64 = pci_default_read_config(dev, conf->addr, conf->len); | ||
126 | + } | ||
127 | + | ||
128 | + ret.cmd = MPQEMU_CMD_RET; | ||
129 | + ret.size = sizeof(ret.data.u64); | ||
130 | + | ||
131 | + if (!mpqemu_msg_send(&ret, ioc, NULL)) { | ||
132 | + error_prepend(errp, "Error returning code to proxy, pid "FMT_pid": ", | ||
133 | + getpid()); | ||
134 | + } | ||
135 | +} | ||
136 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
137 | index XXXXXXX..XXXXXXX 100644 | ||
138 | --- a/hw/remote/mpqemu-link.c | ||
139 | +++ b/hw/remote/mpqemu-link.c | ||
140 | @@ -XXX,XX +XXX,XX @@ uint64_t mpqemu_msg_send_and_await_reply(MPQemuMsg *msg, PCIProxyDev *pdev, | ||
141 | return ret; | ||
142 | } | ||
143 | |||
144 | - if (!mpqemu_msg_valid(&msg_reply)) { | ||
145 | + if (!mpqemu_msg_valid(&msg_reply) || msg_reply.cmd != MPQEMU_CMD_RET) { | ||
146 | error_setg(errp, "ERROR: Invalid reply received for command %d", | ||
147 | msg->cmd); | ||
148 | return ret; | ||
149 | @@ -XXX,XX +XXX,XX @@ bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
150 | return false; | ||
151 | } | ||
152 | break; | ||
153 | + case MPQEMU_CMD_PCI_CFGWRITE: | ||
154 | + case MPQEMU_CMD_PCI_CFGREAD: | ||
155 | + if (msg->size != sizeof(PciConfDataMsg)) { | ||
156 | + return false; | ||
157 | + } | ||
158 | + break; | ||
159 | default: | ||
160 | break; | ||
161 | } | ||
162 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
163 | index XXXXXXX..XXXXXXX 100644 | ||
164 | --- a/hw/remote/proxy.c | ||
165 | +++ b/hw/remote/proxy.c | ||
166 | @@ -XXX,XX +XXX,XX @@ | ||
167 | #include "monitor/monitor.h" | ||
168 | #include "migration/blocker.h" | ||
169 | #include "qemu/sockets.h" | ||
170 | +#include "hw/remote/mpqemu-link.h" | ||
171 | +#include "qemu/error-report.h" | ||
172 | |||
173 | static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
92 | { | 174 | { |
93 | BlockdevBackupState *state = DO_UPCAST(BlockdevBackupState, common, common); | 175 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_exit(PCIDevice *pdev) |
94 | + AioContext *aio_context; | 176 | error_free(dev->migration_blocker); |
95 | |||
96 | - if (state->aio_context) { | ||
97 | - bdrv_drained_end(state->bs); | ||
98 | - aio_context_release(state->aio_context); | ||
99 | + if (!state->bs) { | ||
100 | + return; | ||
101 | } | ||
102 | + | ||
103 | + aio_context = bdrv_get_aio_context(state->bs); | ||
104 | + aio_context_acquire(aio_context); | ||
105 | + | ||
106 | + bdrv_drained_end(state->bs); | ||
107 | + | ||
108 | + aio_context_release(aio_context); | ||
109 | } | 177 | } |
110 | 178 | ||
111 | typedef struct BlockDirtyBitmapState { | 179 | +static void config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val, |
180 | + int len, unsigned int op) | ||
181 | +{ | ||
182 | + MPQemuMsg msg = { 0 }; | ||
183 | + uint64_t ret = -EINVAL; | ||
184 | + Error *local_err = NULL; | ||
185 | + | ||
186 | + msg.cmd = op; | ||
187 | + msg.data.pci_conf_data.addr = addr; | ||
188 | + msg.data.pci_conf_data.val = (op == MPQEMU_CMD_PCI_CFGWRITE) ? *val : 0; | ||
189 | + msg.data.pci_conf_data.len = len; | ||
190 | + msg.size = sizeof(PciConfDataMsg); | ||
191 | + | ||
192 | + ret = mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err); | ||
193 | + if (local_err) { | ||
194 | + error_report_err(local_err); | ||
195 | + } | ||
196 | + | ||
197 | + if (ret == UINT64_MAX) { | ||
198 | + error_report("Failed to perform PCI config %s operation", | ||
199 | + (op == MPQEMU_CMD_PCI_CFGREAD) ? "READ" : "WRITE"); | ||
200 | + } | ||
201 | + | ||
202 | + if (op == MPQEMU_CMD_PCI_CFGREAD) { | ||
203 | + *val = (uint32_t)ret; | ||
204 | + } | ||
205 | +} | ||
206 | + | ||
207 | +static uint32_t pci_proxy_read_config(PCIDevice *d, uint32_t addr, int len) | ||
208 | +{ | ||
209 | + uint32_t val; | ||
210 | + | ||
211 | + config_op_send(PCI_PROXY_DEV(d), addr, &val, len, MPQEMU_CMD_PCI_CFGREAD); | ||
212 | + | ||
213 | + return val; | ||
214 | +} | ||
215 | + | ||
216 | +static void pci_proxy_write_config(PCIDevice *d, uint32_t addr, uint32_t val, | ||
217 | + int len) | ||
218 | +{ | ||
219 | + /* | ||
220 | + * Some of the functions access the copy of remote device's PCI config | ||
221 | + * space which is cached in the proxy device. Therefore, maintain | ||
222 | + * it updated. | ||
223 | + */ | ||
224 | + pci_default_write_config(d, addr, val, len); | ||
225 | + | ||
226 | + config_op_send(PCI_PROXY_DEV(d), addr, &val, len, MPQEMU_CMD_PCI_CFGWRITE); | ||
227 | +} | ||
228 | + | ||
229 | static Property proxy_properties[] = { | ||
230 | DEFINE_PROP_STRING("fd", PCIProxyDev, fd), | ||
231 | DEFINE_PROP_END_OF_LIST(), | ||
232 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_class_init(ObjectClass *klass, void *data) | ||
233 | |||
234 | k->realize = pci_proxy_dev_realize; | ||
235 | k->exit = pci_proxy_dev_exit; | ||
236 | + k->config_read = pci_proxy_read_config; | ||
237 | + k->config_write = pci_proxy_write_config; | ||
238 | + | ||
239 | device_class_set_props(dc, proxy_properties); | ||
240 | } | ||
241 | |||
112 | -- | 242 | -- |
113 | 2.14.3 | 243 | 2.29.2 |
114 | 244 | ||
115 | diff view generated by jsdifflib |
1 | From: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | Convert nvme_init() to realize and rename it to nvme_realize(). | 3 | Proxy device object implements handler for PCI BAR writes and reads. |
4 | 4 | The handler uses BAR_WRITE/BAR_READ message to communicate to the | |
5 | Cc: John Snow <jsnow@redhat.com> | 5 | remote process with the BAR address and value to be written/read. |
6 | Cc: Keith Busch <keith.busch@intel.com> | 6 | The remote process implements handler for BAR_WRITE/BAR_READ |
7 | Cc: Kevin Wolf <kwolf@redhat.com> | 7 | message. |
8 | Cc: Max Reitz <mreitz@redhat.com> | 8 | |
9 | Cc: Markus Armbruster <armbru@redhat.com> | 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> |
10 | 10 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | |
11 | Signed-off-by: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 11 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
12 | Message-id: 2882e72d795e04cbe2120f569d551aef2467ac60.1511317952.git.maozy.fnst@cn.fujitsu.com | 12 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
13 | Message-id: a8b76714a9688be5552c4c92d089bc9e8a4707ff.1611938319.git.jag.raman@oracle.com | ||
13 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
14 | --- | 15 | --- |
15 | hw/block/nvme.c | 18 ++++++++++-------- | 16 | include/hw/remote/mpqemu-link.h | 10 ++++ |
16 | 1 file changed, 10 insertions(+), 8 deletions(-) | 17 | include/hw/remote/proxy.h | 9 ++++ |
17 | 18 | hw/remote/message.c | 83 +++++++++++++++++++++++++++++++++ | |
18 | diff --git a/hw/block/nvme.c b/hw/block/nvme.c | 19 | hw/remote/mpqemu-link.c | 6 +++ |
19 | index XXXXXXX..XXXXXXX 100644 | 20 | hw/remote/proxy.c | 60 ++++++++++++++++++++++++ |
20 | --- a/hw/block/nvme.c | 21 | 5 files changed, 168 insertions(+) |
21 | +++ b/hw/block/nvme.c | 22 | |
22 | @@ -XXX,XX +XXX,XX @@ static const MemoryRegionOps nvme_cmb_ops = { | 23 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h |
23 | }, | 24 | index XXXXXXX..XXXXXXX 100644 |
25 | --- a/include/hw/remote/mpqemu-link.h | ||
26 | +++ b/include/hw/remote/mpqemu-link.h | ||
27 | @@ -XXX,XX +XXX,XX @@ typedef enum { | ||
28 | MPQEMU_CMD_RET, | ||
29 | MPQEMU_CMD_PCI_CFGWRITE, | ||
30 | MPQEMU_CMD_PCI_CFGREAD, | ||
31 | + MPQEMU_CMD_BAR_WRITE, | ||
32 | + MPQEMU_CMD_BAR_READ, | ||
33 | MPQEMU_CMD_MAX, | ||
34 | } MPQemuCmd; | ||
35 | |||
36 | @@ -XXX,XX +XXX,XX @@ typedef struct { | ||
37 | int len; | ||
38 | } PciConfDataMsg; | ||
39 | |||
40 | +typedef struct { | ||
41 | + hwaddr addr; | ||
42 | + uint64_t val; | ||
43 | + unsigned size; | ||
44 | + bool memory; | ||
45 | +} BarAccessMsg; | ||
46 | + | ||
47 | /** | ||
48 | * MPQemuMsg: | ||
49 | * @cmd: The remote command | ||
50 | @@ -XXX,XX +XXX,XX @@ typedef struct { | ||
51 | uint64_t u64; | ||
52 | PciConfDataMsg pci_conf_data; | ||
53 | SyncSysmemMsg sync_sysmem; | ||
54 | + BarAccessMsg bar_access; | ||
55 | } data; | ||
56 | |||
57 | int fds[REMOTE_MAX_FDS]; | ||
58 | diff --git a/include/hw/remote/proxy.h b/include/hw/remote/proxy.h | ||
59 | index XXXXXXX..XXXXXXX 100644 | ||
60 | --- a/include/hw/remote/proxy.h | ||
61 | +++ b/include/hw/remote/proxy.h | ||
62 | @@ -XXX,XX +XXX,XX @@ | ||
63 | #define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev" | ||
64 | OBJECT_DECLARE_SIMPLE_TYPE(PCIProxyDev, PCI_PROXY_DEV) | ||
65 | |||
66 | +typedef struct ProxyMemoryRegion { | ||
67 | + PCIProxyDev *dev; | ||
68 | + MemoryRegion mr; | ||
69 | + bool memory; | ||
70 | + bool present; | ||
71 | + uint8_t type; | ||
72 | +} ProxyMemoryRegion; | ||
73 | + | ||
74 | struct PCIProxyDev { | ||
75 | PCIDevice parent_dev; | ||
76 | char *fd; | ||
77 | @@ -XXX,XX +XXX,XX @@ struct PCIProxyDev { | ||
78 | QemuMutex io_mutex; | ||
79 | QIOChannel *ioc; | ||
80 | Error *migration_blocker; | ||
81 | + ProxyMemoryRegion region[PCI_NUM_REGIONS]; | ||
24 | }; | 82 | }; |
25 | 83 | ||
26 | -static int nvme_init(PCIDevice *pci_dev) | 84 | #endif /* PROXY_H */ |
27 | +static void nvme_realize(PCIDevice *pci_dev, Error **errp) | 85 | diff --git a/hw/remote/message.c b/hw/remote/message.c |
86 | index XXXXXXX..XXXXXXX 100644 | ||
87 | --- a/hw/remote/message.c | ||
88 | +++ b/hw/remote/message.c | ||
89 | @@ -XXX,XX +XXX,XX @@ | ||
90 | #include "qapi/error.h" | ||
91 | #include "sysemu/runstate.h" | ||
92 | #include "hw/pci/pci.h" | ||
93 | +#include "exec/memattrs.h" | ||
94 | |||
95 | static void process_config_write(QIOChannel *ioc, PCIDevice *dev, | ||
96 | MPQemuMsg *msg, Error **errp); | ||
97 | static void process_config_read(QIOChannel *ioc, PCIDevice *dev, | ||
98 | MPQemuMsg *msg, Error **errp); | ||
99 | +static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp); | ||
100 | +static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp); | ||
101 | |||
102 | void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
28 | { | 103 | { |
29 | NvmeCtrl *n = NVME(pci_dev); | 104 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) |
30 | NvmeIdCtrl *id = &n->id_ctrl; | 105 | case MPQEMU_CMD_PCI_CFGREAD: |
31 | @@ -XXX,XX +XXX,XX @@ static int nvme_init(PCIDevice *pci_dev) | 106 | process_config_read(com->ioc, pci_dev, &msg, &local_err); |
32 | Error *local_err = NULL; | 107 | break; |
33 | 108 | + case MPQEMU_CMD_BAR_WRITE: | |
34 | if (!n->conf.blk) { | 109 | + process_bar_write(com->ioc, &msg, &local_err); |
35 | - return -1; | 110 | + break; |
36 | + error_setg(errp, "drive property not set"); | 111 | + case MPQEMU_CMD_BAR_READ: |
37 | + return; | 112 | + process_bar_read(com->ioc, &msg, &local_err); |
113 | + break; | ||
114 | default: | ||
115 | error_setg(&local_err, | ||
116 | "Unknown command (%d) received for device %s" | ||
117 | @@ -XXX,XX +XXX,XX @@ static void process_config_read(QIOChannel *ioc, PCIDevice *dev, | ||
118 | getpid()); | ||
38 | } | 119 | } |
39 | 120 | } | |
40 | bs_size = blk_getlength(n->conf.blk); | 121 | + |
41 | if (bs_size < 0) { | 122 | +static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp) |
42 | - return -1; | 123 | +{ |
43 | + error_setg(errp, "could not get backing file size"); | 124 | + ERRP_GUARD(); |
44 | + return; | 125 | + BarAccessMsg *bar_access = &msg->data.bar_access; |
126 | + AddressSpace *as = | ||
127 | + bar_access->memory ? &address_space_memory : &address_space_io; | ||
128 | + MPQemuMsg ret = { 0 }; | ||
129 | + MemTxResult res; | ||
130 | + uint64_t val; | ||
131 | + | ||
132 | + if (!is_power_of_2(bar_access->size) || | ||
133 | + (bar_access->size > sizeof(uint64_t))) { | ||
134 | + ret.data.u64 = UINT64_MAX; | ||
135 | + goto fail; | ||
136 | + } | ||
137 | + | ||
138 | + val = cpu_to_le64(bar_access->val); | ||
139 | + | ||
140 | + res = address_space_rw(as, bar_access->addr, MEMTXATTRS_UNSPECIFIED, | ||
141 | + (void *)&val, bar_access->size, true); | ||
142 | + | ||
143 | + if (res != MEMTX_OK) { | ||
144 | + error_setg(errp, "Bad address %"PRIx64" for mem write, pid "FMT_pid".", | ||
145 | + bar_access->addr, getpid()); | ||
146 | + ret.data.u64 = -1; | ||
147 | + } | ||
148 | + | ||
149 | +fail: | ||
150 | + ret.cmd = MPQEMU_CMD_RET; | ||
151 | + ret.size = sizeof(ret.data.u64); | ||
152 | + | ||
153 | + if (!mpqemu_msg_send(&ret, ioc, NULL)) { | ||
154 | + error_prepend(errp, "Error returning code to proxy, pid "FMT_pid": ", | ||
155 | + getpid()); | ||
156 | + } | ||
157 | +} | ||
158 | + | ||
159 | +static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp) | ||
160 | +{ | ||
161 | + ERRP_GUARD(); | ||
162 | + BarAccessMsg *bar_access = &msg->data.bar_access; | ||
163 | + MPQemuMsg ret = { 0 }; | ||
164 | + AddressSpace *as; | ||
165 | + MemTxResult res; | ||
166 | + uint64_t val = 0; | ||
167 | + | ||
168 | + as = bar_access->memory ? &address_space_memory : &address_space_io; | ||
169 | + | ||
170 | + if (!is_power_of_2(bar_access->size) || | ||
171 | + (bar_access->size > sizeof(uint64_t))) { | ||
172 | + val = UINT64_MAX; | ||
173 | + goto fail; | ||
174 | + } | ||
175 | + | ||
176 | + res = address_space_rw(as, bar_access->addr, MEMTXATTRS_UNSPECIFIED, | ||
177 | + (void *)&val, bar_access->size, false); | ||
178 | + | ||
179 | + if (res != MEMTX_OK) { | ||
180 | + error_setg(errp, "Bad address %"PRIx64" for mem read, pid "FMT_pid".", | ||
181 | + bar_access->addr, getpid()); | ||
182 | + val = UINT64_MAX; | ||
183 | + } | ||
184 | + | ||
185 | +fail: | ||
186 | + ret.cmd = MPQEMU_CMD_RET; | ||
187 | + ret.data.u64 = le64_to_cpu(val); | ||
188 | + ret.size = sizeof(ret.data.u64); | ||
189 | + | ||
190 | + if (!mpqemu_msg_send(&ret, ioc, NULL)) { | ||
191 | + error_prepend(errp, "Error returning code to proxy, pid "FMT_pid": ", | ||
192 | + getpid()); | ||
193 | + } | ||
194 | +} | ||
195 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
196 | index XXXXXXX..XXXXXXX 100644 | ||
197 | --- a/hw/remote/mpqemu-link.c | ||
198 | +++ b/hw/remote/mpqemu-link.c | ||
199 | @@ -XXX,XX +XXX,XX @@ bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
200 | return false; | ||
201 | } | ||
202 | break; | ||
203 | + case MPQEMU_CMD_BAR_WRITE: | ||
204 | + case MPQEMU_CMD_BAR_READ: | ||
205 | + if ((msg->size != sizeof(BarAccessMsg)) || (msg->num_fds != 0)) { | ||
206 | + return false; | ||
207 | + } | ||
208 | + break; | ||
209 | default: | ||
210 | break; | ||
45 | } | 211 | } |
46 | 212 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | |
47 | blkconf_serial(&n->conf, &n->serial); | 213 | index XXXXXXX..XXXXXXX 100644 |
48 | if (!n->serial) { | 214 | --- a/hw/remote/proxy.c |
49 | - return -1; | 215 | +++ b/hw/remote/proxy.c |
50 | + error_setg(errp, "serial property not set"); | 216 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_register_types(void) |
51 | + return; | ||
52 | } | ||
53 | blkconf_blocksizes(&n->conf); | ||
54 | blkconf_apply_backend_options(&n->conf, blk_is_read_only(n->conf.blk), | ||
55 | false, &local_err); | ||
56 | if (local_err) { | ||
57 | - error_report_err(local_err); | ||
58 | - return -1; | ||
59 | + error_propagate(errp, local_err); | ||
60 | + return; | ||
61 | } | ||
62 | |||
63 | pci_conf = pci_dev->config; | ||
64 | @@ -XXX,XX +XXX,XX @@ static int nvme_init(PCIDevice *pci_dev) | ||
65 | cpu_to_le64(n->ns_size >> | ||
66 | id_ns->lbaf[NVME_ID_NS_FLBAS_INDEX(ns->id_ns.flbas)].ds); | ||
67 | } | ||
68 | - return 0; | ||
69 | } | 217 | } |
70 | 218 | ||
71 | static void nvme_exit(PCIDevice *pci_dev) | 219 | type_init(pci_proxy_dev_register_types) |
72 | @@ -XXX,XX +XXX,XX @@ static void nvme_class_init(ObjectClass *oc, void *data) | 220 | + |
73 | DeviceClass *dc = DEVICE_CLASS(oc); | 221 | +static void send_bar_access_msg(PCIProxyDev *pdev, MemoryRegion *mr, |
74 | PCIDeviceClass *pc = PCI_DEVICE_CLASS(oc); | 222 | + bool write, hwaddr addr, uint64_t *val, |
75 | 223 | + unsigned size, bool memory) | |
76 | - pc->init = nvme_init; | 224 | +{ |
77 | + pc->realize = nvme_realize; | 225 | + MPQemuMsg msg = { 0 }; |
78 | pc->exit = nvme_exit; | 226 | + long ret = -EINVAL; |
79 | pc->class_id = PCI_CLASS_STORAGE_EXPRESS; | 227 | + Error *local_err = NULL; |
80 | pc->vendor_id = PCI_VENDOR_ID_INTEL; | 228 | + |
229 | + msg.size = sizeof(BarAccessMsg); | ||
230 | + msg.data.bar_access.addr = mr->addr + addr; | ||
231 | + msg.data.bar_access.size = size; | ||
232 | + msg.data.bar_access.memory = memory; | ||
233 | + | ||
234 | + if (write) { | ||
235 | + msg.cmd = MPQEMU_CMD_BAR_WRITE; | ||
236 | + msg.data.bar_access.val = *val; | ||
237 | + } else { | ||
238 | + msg.cmd = MPQEMU_CMD_BAR_READ; | ||
239 | + } | ||
240 | + | ||
241 | + ret = mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err); | ||
242 | + if (local_err) { | ||
243 | + error_report_err(local_err); | ||
244 | + } | ||
245 | + | ||
246 | + if (!write) { | ||
247 | + *val = ret; | ||
248 | + } | ||
249 | +} | ||
250 | + | ||
251 | +static void proxy_bar_write(void *opaque, hwaddr addr, uint64_t val, | ||
252 | + unsigned size) | ||
253 | +{ | ||
254 | + ProxyMemoryRegion *pmr = opaque; | ||
255 | + | ||
256 | + send_bar_access_msg(pmr->dev, &pmr->mr, true, addr, &val, size, | ||
257 | + pmr->memory); | ||
258 | +} | ||
259 | + | ||
260 | +static uint64_t proxy_bar_read(void *opaque, hwaddr addr, unsigned size) | ||
261 | +{ | ||
262 | + ProxyMemoryRegion *pmr = opaque; | ||
263 | + uint64_t val; | ||
264 | + | ||
265 | + send_bar_access_msg(pmr->dev, &pmr->mr, false, addr, &val, size, | ||
266 | + pmr->memory); | ||
267 | + | ||
268 | + return val; | ||
269 | +} | ||
270 | + | ||
271 | +const MemoryRegionOps proxy_mr_ops = { | ||
272 | + .read = proxy_bar_read, | ||
273 | + .write = proxy_bar_write, | ||
274 | + .endianness = DEVICE_NATIVE_ENDIAN, | ||
275 | + .impl = { | ||
276 | + .min_access_size = 1, | ||
277 | + .max_access_size = 8, | ||
278 | + }, | ||
279 | +}; | ||
81 | -- | 280 | -- |
82 | 2.14.3 | 281 | 2.29.2 |
83 | 282 | ||
84 | diff view generated by jsdifflib |
1 | From: Mark Kanda <mark.kanda@oracle.com> | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | Depending on the configuration, it can be beneficial to adjust the virtio-blk | 3 | Add ProxyMemoryListener object which is used to keep the view of the RAM |
4 | queue size to something other than the current default of 128. Add a new | 4 | in sync between QEMU and remote process. |
5 | property to make the queue size configurable. | 5 | A MemoryListener is registered for system-memory AddressSpace. The |
6 | listener sends SYNC_SYSMEM message to the remote process when memory | ||
7 | listener commits the changes to memory, the remote process receives | ||
8 | the message and processes it in the handler for SYNC_SYSMEM message. | ||
6 | 9 | ||
7 | Signed-off-by: Mark Kanda <mark.kanda@oracle.com> | 10 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> |
8 | Reviewed-by: Karl Heubaum <karl.heubaum@oracle.com> | 11 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
9 | Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com> | 12 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
10 | Reviewed-by: Ameya More <ameya.more@oracle.com> | 13 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
11 | Message-id: 52e6d742811f10dbd16e996e86cf375b9577c187.1513005190.git.mark.kanda@oracle.com | 14 | Message-id: 04fe4e6a9ca90d4f11ab6f59be7652f5b086a071.1611938319.git.jag.raman@oracle.com |
12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 15 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
13 | --- | 16 | --- |
14 | include/hw/virtio/virtio-blk.h | 1 + | 17 | MAINTAINERS | 2 + |
15 | hw/block/virtio-blk.c | 10 +++++++++- | 18 | include/hw/remote/proxy-memory-listener.h | 28 +++ |
16 | 2 files changed, 10 insertions(+), 1 deletion(-) | 19 | include/hw/remote/proxy.h | 2 + |
20 | hw/remote/message.c | 4 + | ||
21 | hw/remote/proxy-memory-listener.c | 227 ++++++++++++++++++++++ | ||
22 | hw/remote/proxy.c | 6 + | ||
23 | hw/remote/meson.build | 1 + | ||
24 | 7 files changed, 270 insertions(+) | ||
25 | create mode 100644 include/hw/remote/proxy-memory-listener.h | ||
26 | create mode 100644 hw/remote/proxy-memory-listener.c | ||
17 | 27 | ||
18 | diff --git a/include/hw/virtio/virtio-blk.h b/include/hw/virtio/virtio-blk.h | 28 | diff --git a/MAINTAINERS b/MAINTAINERS |
19 | index XXXXXXX..XXXXXXX 100644 | 29 | index XXXXXXX..XXXXXXX 100644 |
20 | --- a/include/hw/virtio/virtio-blk.h | 30 | --- a/MAINTAINERS |
21 | +++ b/include/hw/virtio/virtio-blk.h | 31 | +++ b/MAINTAINERS |
22 | @@ -XXX,XX +XXX,XX @@ struct VirtIOBlkConf | 32 | @@ -XXX,XX +XXX,XX @@ F: include/hw/remote/memory.h |
23 | uint32_t config_wce; | 33 | F: hw/remote/memory.c |
24 | uint32_t request_merging; | 34 | F: hw/remote/proxy.c |
25 | uint16_t num_queues; | 35 | F: include/hw/remote/proxy.h |
26 | + uint16_t queue_size; | 36 | +F: hw/remote/proxy-memory-listener.c |
37 | +F: include/hw/remote/proxy-memory-listener.h | ||
38 | |||
39 | Build and test automation | ||
40 | ------------------------- | ||
41 | diff --git a/include/hw/remote/proxy-memory-listener.h b/include/hw/remote/proxy-memory-listener.h | ||
42 | new file mode 100644 | ||
43 | index XXXXXXX..XXXXXXX | ||
44 | --- /dev/null | ||
45 | +++ b/include/hw/remote/proxy-memory-listener.h | ||
46 | @@ -XXX,XX +XXX,XX @@ | ||
47 | +/* | ||
48 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
49 | + * | ||
50 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
51 | + * See the COPYING file in the top-level directory. | ||
52 | + * | ||
53 | + */ | ||
54 | + | ||
55 | +#ifndef PROXY_MEMORY_LISTENER_H | ||
56 | +#define PROXY_MEMORY_LISTENER_H | ||
57 | + | ||
58 | +#include "exec/memory.h" | ||
59 | +#include "io/channel.h" | ||
60 | + | ||
61 | +typedef struct ProxyMemoryListener { | ||
62 | + MemoryListener listener; | ||
63 | + | ||
64 | + int n_mr_sections; | ||
65 | + MemoryRegionSection *mr_sections; | ||
66 | + | ||
67 | + QIOChannel *ioc; | ||
68 | +} ProxyMemoryListener; | ||
69 | + | ||
70 | +void proxy_memory_listener_configure(ProxyMemoryListener *proxy_listener, | ||
71 | + QIOChannel *ioc); | ||
72 | +void proxy_memory_listener_deconfigure(ProxyMemoryListener *proxy_listener); | ||
73 | + | ||
74 | +#endif | ||
75 | diff --git a/include/hw/remote/proxy.h b/include/hw/remote/proxy.h | ||
76 | index XXXXXXX..XXXXXXX 100644 | ||
77 | --- a/include/hw/remote/proxy.h | ||
78 | +++ b/include/hw/remote/proxy.h | ||
79 | @@ -XXX,XX +XXX,XX @@ | ||
80 | |||
81 | #include "hw/pci/pci.h" | ||
82 | #include "io/channel.h" | ||
83 | +#include "hw/remote/proxy-memory-listener.h" | ||
84 | |||
85 | #define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev" | ||
86 | OBJECT_DECLARE_SIMPLE_TYPE(PCIProxyDev, PCI_PROXY_DEV) | ||
87 | @@ -XXX,XX +XXX,XX @@ struct PCIProxyDev { | ||
88 | QemuMutex io_mutex; | ||
89 | QIOChannel *ioc; | ||
90 | Error *migration_blocker; | ||
91 | + ProxyMemoryListener proxy_listener; | ||
92 | ProxyMemoryRegion region[PCI_NUM_REGIONS]; | ||
27 | }; | 93 | }; |
28 | 94 | ||
29 | struct VirtIOBlockDataPlane; | 95 | diff --git a/hw/remote/message.c b/hw/remote/message.c |
30 | diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c | 96 | index XXXXXXX..XXXXXXX 100644 |
31 | index XXXXXXX..XXXXXXX 100644 | 97 | --- a/hw/remote/message.c |
32 | --- a/hw/block/virtio-blk.c | 98 | +++ b/hw/remote/message.c |
33 | +++ b/hw/block/virtio-blk.c | 99 | @@ -XXX,XX +XXX,XX @@ |
34 | @@ -XXX,XX +XXX,XX @@ static void virtio_blk_device_realize(DeviceState *dev, Error **errp) | 100 | #include "sysemu/runstate.h" |
35 | error_setg(errp, "num-queues property must be larger than 0"); | 101 | #include "hw/pci/pci.h" |
36 | return; | 102 | #include "exec/memattrs.h" |
37 | } | 103 | +#include "hw/remote/memory.h" |
38 | + if (!is_power_of_2(conf->queue_size) || | 104 | |
39 | + conf->queue_size > VIRTQUEUE_MAX_SIZE) { | 105 | static void process_config_write(QIOChannel *ioc, PCIDevice *dev, |
40 | + error_setg(errp, "invalid queue-size property (%" PRIu16 "), " | 106 | MPQemuMsg *msg, Error **errp); |
41 | + "must be a power of 2 (max %d)", | 107 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) |
42 | + conf->queue_size, VIRTQUEUE_MAX_SIZE); | 108 | case MPQEMU_CMD_BAR_READ: |
109 | process_bar_read(com->ioc, &msg, &local_err); | ||
110 | break; | ||
111 | + case MPQEMU_CMD_SYNC_SYSMEM: | ||
112 | + remote_sysmem_reconfig(&msg, &local_err); | ||
113 | + break; | ||
114 | default: | ||
115 | error_setg(&local_err, | ||
116 | "Unknown command (%d) received for device %s" | ||
117 | diff --git a/hw/remote/proxy-memory-listener.c b/hw/remote/proxy-memory-listener.c | ||
118 | new file mode 100644 | ||
119 | index XXXXXXX..XXXXXXX | ||
120 | --- /dev/null | ||
121 | +++ b/hw/remote/proxy-memory-listener.c | ||
122 | @@ -XXX,XX +XXX,XX @@ | ||
123 | +/* | ||
124 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
125 | + * | ||
126 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
127 | + * See the COPYING file in the top-level directory. | ||
128 | + * | ||
129 | + */ | ||
130 | + | ||
131 | +#include "qemu/osdep.h" | ||
132 | +#include "qemu-common.h" | ||
133 | + | ||
134 | +#include "qemu/compiler.h" | ||
135 | +#include "qemu/int128.h" | ||
136 | +#include "qemu/range.h" | ||
137 | +#include "exec/memory.h" | ||
138 | +#include "exec/cpu-common.h" | ||
139 | +#include "cpu.h" | ||
140 | +#include "exec/ram_addr.h" | ||
141 | +#include "exec/address-spaces.h" | ||
142 | +#include "qapi/error.h" | ||
143 | +#include "hw/remote/mpqemu-link.h" | ||
144 | +#include "hw/remote/proxy-memory-listener.h" | ||
145 | + | ||
146 | +/* | ||
147 | + * TODO: get_fd_from_hostaddr(), proxy_mrs_can_merge() and | ||
148 | + * proxy_memory_listener_commit() defined below perform tasks similar to the | ||
149 | + * functions defined in vhost-user.c. These functions are good candidates | ||
150 | + * for refactoring. | ||
151 | + * | ||
152 | + */ | ||
153 | + | ||
154 | +static void proxy_memory_listener_reset(MemoryListener *listener) | ||
155 | +{ | ||
156 | + ProxyMemoryListener *proxy_listener = container_of(listener, | ||
157 | + ProxyMemoryListener, | ||
158 | + listener); | ||
159 | + int mrs; | ||
160 | + | ||
161 | + for (mrs = 0; mrs < proxy_listener->n_mr_sections; mrs++) { | ||
162 | + memory_region_unref(proxy_listener->mr_sections[mrs].mr); | ||
163 | + } | ||
164 | + | ||
165 | + g_free(proxy_listener->mr_sections); | ||
166 | + proxy_listener->mr_sections = NULL; | ||
167 | + proxy_listener->n_mr_sections = 0; | ||
168 | +} | ||
169 | + | ||
170 | +static int get_fd_from_hostaddr(uint64_t host, ram_addr_t *offset) | ||
171 | +{ | ||
172 | + MemoryRegion *mr; | ||
173 | + ram_addr_t off; | ||
174 | + | ||
175 | + /** | ||
176 | + * Assumes that the host address is a valid address as it's | ||
177 | + * coming from the MemoryListener system. In the case host | ||
178 | + * address is not valid, the following call would return | ||
179 | + * the default subregion of "system_memory" region, and | ||
180 | + * not NULL. So it's not possible to check for NULL here. | ||
181 | + */ | ||
182 | + mr = memory_region_from_host((void *)(uintptr_t)host, &off); | ||
183 | + | ||
184 | + if (offset) { | ||
185 | + *offset = off; | ||
186 | + } | ||
187 | + | ||
188 | + return memory_region_get_fd(mr); | ||
189 | +} | ||
190 | + | ||
191 | +static bool proxy_mrs_can_merge(uint64_t host, uint64_t prev_host, size_t size) | ||
192 | +{ | ||
193 | + if (((prev_host + size) != host)) { | ||
194 | + return false; | ||
195 | + } | ||
196 | + | ||
197 | + if (get_fd_from_hostaddr(host, NULL) != | ||
198 | + get_fd_from_hostaddr(prev_host, NULL)) { | ||
199 | + return false; | ||
200 | + } | ||
201 | + | ||
202 | + return true; | ||
203 | +} | ||
204 | + | ||
205 | +static bool try_merge(ProxyMemoryListener *proxy_listener, | ||
206 | + MemoryRegionSection *section) | ||
207 | +{ | ||
208 | + uint64_t mrs_size, mrs_gpa, mrs_page; | ||
209 | + MemoryRegionSection *prev_sec; | ||
210 | + bool merged = false; | ||
211 | + uintptr_t mrs_host; | ||
212 | + RAMBlock *mrs_rb; | ||
213 | + | ||
214 | + if (!proxy_listener->n_mr_sections) { | ||
215 | + return false; | ||
216 | + } | ||
217 | + | ||
218 | + mrs_rb = section->mr->ram_block; | ||
219 | + mrs_page = (uint64_t)qemu_ram_pagesize(mrs_rb); | ||
220 | + mrs_size = int128_get64(section->size); | ||
221 | + mrs_gpa = section->offset_within_address_space; | ||
222 | + mrs_host = (uintptr_t)memory_region_get_ram_ptr(section->mr) + | ||
223 | + section->offset_within_region; | ||
224 | + | ||
225 | + if (get_fd_from_hostaddr(mrs_host, NULL) < 0) { | ||
226 | + return true; | ||
227 | + } | ||
228 | + | ||
229 | + mrs_host = mrs_host & ~(mrs_page - 1); | ||
230 | + mrs_gpa = mrs_gpa & ~(mrs_page - 1); | ||
231 | + mrs_size = ROUND_UP(mrs_size, mrs_page); | ||
232 | + | ||
233 | + prev_sec = proxy_listener->mr_sections + | ||
234 | + (proxy_listener->n_mr_sections - 1); | ||
235 | + uint64_t prev_gpa_start = prev_sec->offset_within_address_space; | ||
236 | + uint64_t prev_size = int128_get64(prev_sec->size); | ||
237 | + uint64_t prev_gpa_end = range_get_last(prev_gpa_start, prev_size); | ||
238 | + uint64_t prev_host_start = | ||
239 | + (uintptr_t)memory_region_get_ram_ptr(prev_sec->mr) + | ||
240 | + prev_sec->offset_within_region; | ||
241 | + uint64_t prev_host_end = range_get_last(prev_host_start, prev_size); | ||
242 | + | ||
243 | + if (mrs_gpa <= (prev_gpa_end + 1)) { | ||
244 | + g_assert(mrs_gpa > prev_gpa_start); | ||
245 | + | ||
246 | + if ((section->mr == prev_sec->mr) && | ||
247 | + proxy_mrs_can_merge(mrs_host, prev_host_start, | ||
248 | + (mrs_gpa - prev_gpa_start))) { | ||
249 | + uint64_t max_end = MAX(prev_host_end, mrs_host + mrs_size); | ||
250 | + merged = true; | ||
251 | + prev_sec->offset_within_address_space = | ||
252 | + MIN(prev_gpa_start, mrs_gpa); | ||
253 | + prev_sec->offset_within_region = | ||
254 | + MIN(prev_host_start, mrs_host) - | ||
255 | + (uintptr_t)memory_region_get_ram_ptr(prev_sec->mr); | ||
256 | + prev_sec->size = int128_make64(max_end - MIN(prev_host_start, | ||
257 | + mrs_host)); | ||
258 | + } | ||
259 | + } | ||
260 | + | ||
261 | + return merged; | ||
262 | +} | ||
263 | + | ||
264 | +static void proxy_memory_listener_region_addnop(MemoryListener *listener, | ||
265 | + MemoryRegionSection *section) | ||
266 | +{ | ||
267 | + ProxyMemoryListener *proxy_listener = container_of(listener, | ||
268 | + ProxyMemoryListener, | ||
269 | + listener); | ||
270 | + | ||
271 | + if (!memory_region_is_ram(section->mr) || | ||
272 | + memory_region_is_rom(section->mr)) { | ||
43 | + return; | 273 | + return; |
44 | + } | 274 | + } |
45 | 275 | + | |
46 | blkconf_serial(&conf->conf, &conf->serial); | 276 | + if (try_merge(proxy_listener, section)) { |
47 | if (!blkconf_apply_backend_options(&conf->conf, | 277 | + return; |
48 | @@ -XXX,XX +XXX,XX @@ static void virtio_blk_device_realize(DeviceState *dev, Error **errp) | 278 | + } |
49 | s->sector_mask = (s->conf.conf.logical_block_size / BDRV_SECTOR_SIZE) - 1; | 279 | + |
50 | 280 | + ++proxy_listener->n_mr_sections; | |
51 | for (i = 0; i < conf->num_queues; i++) { | 281 | + proxy_listener->mr_sections = g_renew(MemoryRegionSection, |
52 | - virtio_add_queue(vdev, 128, virtio_blk_handle_output); | 282 | + proxy_listener->mr_sections, |
53 | + virtio_add_queue(vdev, conf->queue_size, virtio_blk_handle_output); | 283 | + proxy_listener->n_mr_sections); |
54 | } | 284 | + proxy_listener->mr_sections[proxy_listener->n_mr_sections - 1] = *section; |
55 | if (!virtio_blk_data_plane_create(vdev, conf, &s->dataplane, errp)) { | 285 | + proxy_listener->mr_sections[proxy_listener->n_mr_sections - 1].fv = NULL; |
56 | virtio_cleanup(vdev); | 286 | + memory_region_ref(section->mr); |
57 | @@ -XXX,XX +XXX,XX @@ static Property virtio_blk_properties[] = { | 287 | +} |
58 | DEFINE_PROP_BIT("request-merging", VirtIOBlock, conf.request_merging, 0, | 288 | + |
59 | true), | 289 | +static void proxy_memory_listener_commit(MemoryListener *listener) |
60 | DEFINE_PROP_UINT16("num-queues", VirtIOBlock, conf.num_queues, 1), | 290 | +{ |
61 | + DEFINE_PROP_UINT16("queue-size", VirtIOBlock, conf.queue_size, 128), | 291 | + ProxyMemoryListener *proxy_listener = container_of(listener, |
62 | DEFINE_PROP_LINK("iothread", VirtIOBlock, conf.iothread, TYPE_IOTHREAD, | 292 | + ProxyMemoryListener, |
63 | IOThread *), | 293 | + listener); |
64 | DEFINE_PROP_END_OF_LIST(), | 294 | + MPQemuMsg msg; |
295 | + MemoryRegionSection *section; | ||
296 | + ram_addr_t offset; | ||
297 | + uintptr_t host_addr; | ||
298 | + int region; | ||
299 | + Error *local_err = NULL; | ||
300 | + | ||
301 | + memset(&msg, 0, sizeof(MPQemuMsg)); | ||
302 | + | ||
303 | + msg.cmd = MPQEMU_CMD_SYNC_SYSMEM; | ||
304 | + msg.num_fds = proxy_listener->n_mr_sections; | ||
305 | + msg.size = sizeof(SyncSysmemMsg); | ||
306 | + if (msg.num_fds > REMOTE_MAX_FDS) { | ||
307 | + error_report("Number of fds is more than %d", REMOTE_MAX_FDS); | ||
308 | + return; | ||
309 | + } | ||
310 | + | ||
311 | + for (region = 0; region < proxy_listener->n_mr_sections; region++) { | ||
312 | + section = &proxy_listener->mr_sections[region]; | ||
313 | + msg.data.sync_sysmem.gpas[region] = | ||
314 | + section->offset_within_address_space; | ||
315 | + msg.data.sync_sysmem.sizes[region] = int128_get64(section->size); | ||
316 | + host_addr = (uintptr_t)memory_region_get_ram_ptr(section->mr) + | ||
317 | + section->offset_within_region; | ||
318 | + msg.fds[region] = get_fd_from_hostaddr(host_addr, &offset); | ||
319 | + msg.data.sync_sysmem.offsets[region] = offset; | ||
320 | + } | ||
321 | + if (!mpqemu_msg_send(&msg, proxy_listener->ioc, &local_err)) { | ||
322 | + error_report_err(local_err); | ||
323 | + } | ||
324 | +} | ||
325 | + | ||
326 | +void proxy_memory_listener_deconfigure(ProxyMemoryListener *proxy_listener) | ||
327 | +{ | ||
328 | + memory_listener_unregister(&proxy_listener->listener); | ||
329 | + | ||
330 | + proxy_memory_listener_reset(&proxy_listener->listener); | ||
331 | +} | ||
332 | + | ||
333 | +void proxy_memory_listener_configure(ProxyMemoryListener *proxy_listener, | ||
334 | + QIOChannel *ioc) | ||
335 | +{ | ||
336 | + proxy_listener->n_mr_sections = 0; | ||
337 | + proxy_listener->mr_sections = NULL; | ||
338 | + | ||
339 | + proxy_listener->ioc = ioc; | ||
340 | + | ||
341 | + proxy_listener->listener.begin = proxy_memory_listener_reset; | ||
342 | + proxy_listener->listener.commit = proxy_memory_listener_commit; | ||
343 | + proxy_listener->listener.region_add = proxy_memory_listener_region_addnop; | ||
344 | + proxy_listener->listener.region_nop = proxy_memory_listener_region_addnop; | ||
345 | + proxy_listener->listener.priority = 10; | ||
346 | + | ||
347 | + memory_listener_register(&proxy_listener->listener, | ||
348 | + &address_space_memory); | ||
349 | +} | ||
350 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
351 | index XXXXXXX..XXXXXXX 100644 | ||
352 | --- a/hw/remote/proxy.c | ||
353 | +++ b/hw/remote/proxy.c | ||
354 | @@ -XXX,XX +XXX,XX @@ | ||
355 | #include "qemu/sockets.h" | ||
356 | #include "hw/remote/mpqemu-link.h" | ||
357 | #include "qemu/error-report.h" | ||
358 | +#include "hw/remote/proxy-memory-listener.h" | ||
359 | +#include "qom/object.h" | ||
360 | |||
361 | static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
362 | { | ||
363 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
364 | |||
365 | qemu_mutex_init(&dev->io_mutex); | ||
366 | qio_channel_set_blocking(dev->ioc, true, NULL); | ||
367 | + | ||
368 | + proxy_memory_listener_configure(&dev->proxy_listener, dev->ioc); | ||
369 | } | ||
370 | |||
371 | static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
372 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
373 | migrate_del_blocker(dev->migration_blocker); | ||
374 | |||
375 | error_free(dev->migration_blocker); | ||
376 | + | ||
377 | + proxy_memory_listener_deconfigure(&dev->proxy_listener); | ||
378 | } | ||
379 | |||
380 | static void config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val, | ||
381 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
382 | index XXXXXXX..XXXXXXX 100644 | ||
383 | --- a/hw/remote/meson.build | ||
384 | +++ b/hw/remote/meson.build | ||
385 | @@ -XXX,XX +XXX,XX @@ remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) | ||
386 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy.c')) | ||
387 | |||
388 | specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('memory.c')) | ||
389 | +specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy-memory-listener.c')) | ||
390 | |||
391 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
65 | -- | 392 | -- |
66 | 2.14.3 | 393 | 2.29.2 |
67 | 394 | ||
68 | diff view generated by jsdifflib |
1 | From: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | The function name of usb_msd_{realize,unrealize}_*, | 3 | IOHUB object is added to manage PCI IRQs. It uses KVM_IRQFD |
4 | usb_msd_class_initfn_* are unusual. Rename it to | 4 | ioctl to create irqfd to injecting PCI interrupts to the guest. |
5 | usb_msd_*_{realize,unrealize}, usb_msd_class_*_initfn. | 5 | IOHUB object forwards the irqfd to the remote process. Remote process |
6 | uses this fd to directly send interrupts to the guest, bypassing QEMU. | ||
6 | 7 | ||
7 | Cc: Gerd Hoffmann <kraxel@redhat.com> | 8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
8 | 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | |
9 | Signed-off-by: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 10 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
10 | Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org> | 11 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
11 | Message-id: 11e6003433abce35f3f4970e1acc71ee92dbcf51.1511317952.git.maozy.fnst@cn.fujitsu.com | 12 | Message-id: 51d5c3d54e28a68b002e3875c59599c9f5a424a1.1611938319.git.jag.raman@oracle.com |
12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 13 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
13 | --- | 14 | --- |
14 | hw/usb/dev-storage.c | 20 ++++++++++---------- | 15 | MAINTAINERS | 2 + |
15 | 1 file changed, 10 insertions(+), 10 deletions(-) | 16 | include/hw/pci/pci_ids.h | 3 + |
17 | include/hw/remote/iohub.h | 42 +++++++++++ | ||
18 | include/hw/remote/machine.h | 2 + | ||
19 | include/hw/remote/mpqemu-link.h | 1 + | ||
20 | include/hw/remote/proxy.h | 4 ++ | ||
21 | hw/remote/iohub.c | 119 ++++++++++++++++++++++++++++++++ | ||
22 | hw/remote/machine.c | 10 +++ | ||
23 | hw/remote/message.c | 4 ++ | ||
24 | hw/remote/mpqemu-link.c | 5 ++ | ||
25 | hw/remote/proxy.c | 56 +++++++++++++++ | ||
26 | hw/remote/meson.build | 1 + | ||
27 | 12 files changed, 249 insertions(+) | ||
28 | create mode 100644 include/hw/remote/iohub.h | ||
29 | create mode 100644 hw/remote/iohub.c | ||
16 | 30 | ||
17 | diff --git a/hw/usb/dev-storage.c b/hw/usb/dev-storage.c | 31 | diff --git a/MAINTAINERS b/MAINTAINERS |
18 | index XXXXXXX..XXXXXXX 100644 | 32 | index XXXXXXX..XXXXXXX 100644 |
19 | --- a/hw/usb/dev-storage.c | 33 | --- a/MAINTAINERS |
20 | +++ b/hw/usb/dev-storage.c | 34 | +++ b/MAINTAINERS |
21 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_unrealize_storage(USBDevice *dev, Error **errp) | 35 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/proxy.c |
22 | object_unref(OBJECT(&s->bus)); | 36 | F: include/hw/remote/proxy.h |
37 | F: hw/remote/proxy-memory-listener.c | ||
38 | F: include/hw/remote/proxy-memory-listener.h | ||
39 | +F: hw/remote/iohub.c | ||
40 | +F: include/hw/remote/iohub.h | ||
41 | |||
42 | Build and test automation | ||
43 | ------------------------- | ||
44 | diff --git a/include/hw/pci/pci_ids.h b/include/hw/pci/pci_ids.h | ||
45 | index XXXXXXX..XXXXXXX 100644 | ||
46 | --- a/include/hw/pci/pci_ids.h | ||
47 | +++ b/include/hw/pci/pci_ids.h | ||
48 | @@ -XXX,XX +XXX,XX @@ | ||
49 | #define PCI_DEVICE_ID_SUN_SIMBA 0x5000 | ||
50 | #define PCI_DEVICE_ID_SUN_SABRE 0xa000 | ||
51 | |||
52 | +#define PCI_VENDOR_ID_ORACLE 0x108e | ||
53 | +#define PCI_DEVICE_ID_REMOTE_IOHUB 0xb000 | ||
54 | + | ||
55 | #define PCI_VENDOR_ID_CMD 0x1095 | ||
56 | #define PCI_DEVICE_ID_CMD_646 0x0646 | ||
57 | |||
58 | diff --git a/include/hw/remote/iohub.h b/include/hw/remote/iohub.h | ||
59 | new file mode 100644 | ||
60 | index XXXXXXX..XXXXXXX | ||
61 | --- /dev/null | ||
62 | +++ b/include/hw/remote/iohub.h | ||
63 | @@ -XXX,XX +XXX,XX @@ | ||
64 | +/* | ||
65 | + * IO Hub for remote device | ||
66 | + * | ||
67 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
68 | + * | ||
69 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
70 | + * See the COPYING file in the top-level directory. | ||
71 | + * | ||
72 | + */ | ||
73 | + | ||
74 | +#ifndef REMOTE_IOHUB_H | ||
75 | +#define REMOTE_IOHUB_H | ||
76 | + | ||
77 | +#include "hw/pci/pci.h" | ||
78 | +#include "qemu/event_notifier.h" | ||
79 | +#include "qemu/thread-posix.h" | ||
80 | +#include "hw/remote/mpqemu-link.h" | ||
81 | + | ||
82 | +#define REMOTE_IOHUB_NB_PIRQS PCI_DEVFN_MAX | ||
83 | + | ||
84 | +typedef struct ResampleToken { | ||
85 | + void *iohub; | ||
86 | + int pirq; | ||
87 | +} ResampleToken; | ||
88 | + | ||
89 | +typedef struct RemoteIOHubState { | ||
90 | + PCIDevice d; | ||
91 | + EventNotifier irqfds[REMOTE_IOHUB_NB_PIRQS]; | ||
92 | + EventNotifier resamplefds[REMOTE_IOHUB_NB_PIRQS]; | ||
93 | + unsigned int irq_level[REMOTE_IOHUB_NB_PIRQS]; | ||
94 | + ResampleToken token[REMOTE_IOHUB_NB_PIRQS]; | ||
95 | + QemuMutex irq_level_lock[REMOTE_IOHUB_NB_PIRQS]; | ||
96 | +} RemoteIOHubState; | ||
97 | + | ||
98 | +int remote_iohub_map_irq(PCIDevice *pci_dev, int intx); | ||
99 | +void remote_iohub_set_irq(void *opaque, int pirq, int level); | ||
100 | +void process_set_irqfd_msg(PCIDevice *pci_dev, MPQemuMsg *msg); | ||
101 | + | ||
102 | +void remote_iohub_init(RemoteIOHubState *iohub); | ||
103 | +void remote_iohub_finalize(RemoteIOHubState *iohub); | ||
104 | + | ||
105 | +#endif | ||
106 | diff --git a/include/hw/remote/machine.h b/include/hw/remote/machine.h | ||
107 | index XXXXXXX..XXXXXXX 100644 | ||
108 | --- a/include/hw/remote/machine.h | ||
109 | +++ b/include/hw/remote/machine.h | ||
110 | @@ -XXX,XX +XXX,XX @@ | ||
111 | #include "hw/boards.h" | ||
112 | #include "hw/pci-host/remote.h" | ||
113 | #include "io/channel.h" | ||
114 | +#include "hw/remote/iohub.h" | ||
115 | |||
116 | struct RemoteMachineState { | ||
117 | MachineState parent_obj; | ||
118 | |||
119 | RemotePCIHost *host; | ||
120 | + RemoteIOHubState iohub; | ||
121 | }; | ||
122 | |||
123 | /* Used to pass to co-routine device and ioc. */ | ||
124 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
125 | index XXXXXXX..XXXXXXX 100644 | ||
126 | --- a/include/hw/remote/mpqemu-link.h | ||
127 | +++ b/include/hw/remote/mpqemu-link.h | ||
128 | @@ -XXX,XX +XXX,XX @@ typedef enum { | ||
129 | MPQEMU_CMD_PCI_CFGREAD, | ||
130 | MPQEMU_CMD_BAR_WRITE, | ||
131 | MPQEMU_CMD_BAR_READ, | ||
132 | + MPQEMU_CMD_SET_IRQFD, | ||
133 | MPQEMU_CMD_MAX, | ||
134 | } MPQemuCmd; | ||
135 | |||
136 | diff --git a/include/hw/remote/proxy.h b/include/hw/remote/proxy.h | ||
137 | index XXXXXXX..XXXXXXX 100644 | ||
138 | --- a/include/hw/remote/proxy.h | ||
139 | +++ b/include/hw/remote/proxy.h | ||
140 | @@ -XXX,XX +XXX,XX @@ | ||
141 | #include "hw/pci/pci.h" | ||
142 | #include "io/channel.h" | ||
143 | #include "hw/remote/proxy-memory-listener.h" | ||
144 | +#include "qemu/event_notifier.h" | ||
145 | |||
146 | #define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev" | ||
147 | OBJECT_DECLARE_SIMPLE_TYPE(PCIProxyDev, PCI_PROXY_DEV) | ||
148 | @@ -XXX,XX +XXX,XX @@ struct PCIProxyDev { | ||
149 | QIOChannel *ioc; | ||
150 | Error *migration_blocker; | ||
151 | ProxyMemoryListener proxy_listener; | ||
152 | + int virq; | ||
153 | + EventNotifier intr; | ||
154 | + EventNotifier resample; | ||
155 | ProxyMemoryRegion region[PCI_NUM_REGIONS]; | ||
156 | }; | ||
157 | |||
158 | diff --git a/hw/remote/iohub.c b/hw/remote/iohub.c | ||
159 | new file mode 100644 | ||
160 | index XXXXXXX..XXXXXXX | ||
161 | --- /dev/null | ||
162 | +++ b/hw/remote/iohub.c | ||
163 | @@ -XXX,XX +XXX,XX @@ | ||
164 | +/* | ||
165 | + * Remote IO Hub | ||
166 | + * | ||
167 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
168 | + * | ||
169 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
170 | + * See the COPYING file in the top-level directory. | ||
171 | + * | ||
172 | + */ | ||
173 | + | ||
174 | +#include "qemu/osdep.h" | ||
175 | +#include "qemu-common.h" | ||
176 | + | ||
177 | +#include "hw/pci/pci.h" | ||
178 | +#include "hw/pci/pci_ids.h" | ||
179 | +#include "hw/pci/pci_bus.h" | ||
180 | +#include "qemu/thread.h" | ||
181 | +#include "hw/boards.h" | ||
182 | +#include "hw/remote/machine.h" | ||
183 | +#include "hw/remote/iohub.h" | ||
184 | +#include "qemu/main-loop.h" | ||
185 | + | ||
186 | +void remote_iohub_init(RemoteIOHubState *iohub) | ||
187 | +{ | ||
188 | + int pirq; | ||
189 | + | ||
190 | + memset(&iohub->irqfds, 0, sizeof(iohub->irqfds)); | ||
191 | + memset(&iohub->resamplefds, 0, sizeof(iohub->resamplefds)); | ||
192 | + | ||
193 | + for (pirq = 0; pirq < REMOTE_IOHUB_NB_PIRQS; pirq++) { | ||
194 | + qemu_mutex_init(&iohub->irq_level_lock[pirq]); | ||
195 | + iohub->irq_level[pirq] = 0; | ||
196 | + event_notifier_init_fd(&iohub->irqfds[pirq], -1); | ||
197 | + event_notifier_init_fd(&iohub->resamplefds[pirq], -1); | ||
198 | + } | ||
199 | +} | ||
200 | + | ||
201 | +void remote_iohub_finalize(RemoteIOHubState *iohub) | ||
202 | +{ | ||
203 | + int pirq; | ||
204 | + | ||
205 | + for (pirq = 0; pirq < REMOTE_IOHUB_NB_PIRQS; pirq++) { | ||
206 | + qemu_set_fd_handler(event_notifier_get_fd(&iohub->resamplefds[pirq]), | ||
207 | + NULL, NULL, NULL); | ||
208 | + event_notifier_cleanup(&iohub->irqfds[pirq]); | ||
209 | + event_notifier_cleanup(&iohub->resamplefds[pirq]); | ||
210 | + qemu_mutex_destroy(&iohub->irq_level_lock[pirq]); | ||
211 | + } | ||
212 | +} | ||
213 | + | ||
214 | +int remote_iohub_map_irq(PCIDevice *pci_dev, int intx) | ||
215 | +{ | ||
216 | + return pci_dev->devfn; | ||
217 | +} | ||
218 | + | ||
219 | +void remote_iohub_set_irq(void *opaque, int pirq, int level) | ||
220 | +{ | ||
221 | + RemoteIOHubState *iohub = opaque; | ||
222 | + | ||
223 | + assert(pirq >= 0); | ||
224 | + assert(pirq < PCI_DEVFN_MAX); | ||
225 | + | ||
226 | + QEMU_LOCK_GUARD(&iohub->irq_level_lock[pirq]); | ||
227 | + | ||
228 | + if (level) { | ||
229 | + if (++iohub->irq_level[pirq] == 1) { | ||
230 | + event_notifier_set(&iohub->irqfds[pirq]); | ||
231 | + } | ||
232 | + } else if (iohub->irq_level[pirq] > 0) { | ||
233 | + iohub->irq_level[pirq]--; | ||
234 | + } | ||
235 | +} | ||
236 | + | ||
237 | +static void intr_resample_handler(void *opaque) | ||
238 | +{ | ||
239 | + ResampleToken *token = opaque; | ||
240 | + RemoteIOHubState *iohub = token->iohub; | ||
241 | + int pirq, s; | ||
242 | + | ||
243 | + pirq = token->pirq; | ||
244 | + | ||
245 | + s = event_notifier_test_and_clear(&iohub->resamplefds[pirq]); | ||
246 | + | ||
247 | + assert(s >= 0); | ||
248 | + | ||
249 | + QEMU_LOCK_GUARD(&iohub->irq_level_lock[pirq]); | ||
250 | + | ||
251 | + if (iohub->irq_level[pirq]) { | ||
252 | + event_notifier_set(&iohub->irqfds[pirq]); | ||
253 | + } | ||
254 | +} | ||
255 | + | ||
256 | +void process_set_irqfd_msg(PCIDevice *pci_dev, MPQemuMsg *msg) | ||
257 | +{ | ||
258 | + RemoteMachineState *machine = REMOTE_MACHINE(current_machine); | ||
259 | + RemoteIOHubState *iohub = &machine->iohub; | ||
260 | + int pirq, intx; | ||
261 | + | ||
262 | + intx = pci_get_byte(pci_dev->config + PCI_INTERRUPT_PIN) - 1; | ||
263 | + | ||
264 | + pirq = remote_iohub_map_irq(pci_dev, intx); | ||
265 | + | ||
266 | + if (event_notifier_get_fd(&iohub->irqfds[pirq]) != -1) { | ||
267 | + qemu_set_fd_handler(event_notifier_get_fd(&iohub->resamplefds[pirq]), | ||
268 | + NULL, NULL, NULL); | ||
269 | + event_notifier_cleanup(&iohub->irqfds[pirq]); | ||
270 | + event_notifier_cleanup(&iohub->resamplefds[pirq]); | ||
271 | + memset(&iohub->token[pirq], 0, sizeof(ResampleToken)); | ||
272 | + } | ||
273 | + | ||
274 | + event_notifier_init_fd(&iohub->irqfds[pirq], msg->fds[0]); | ||
275 | + event_notifier_init_fd(&iohub->resamplefds[pirq], msg->fds[1]); | ||
276 | + | ||
277 | + iohub->token[pirq].iohub = iohub; | ||
278 | + iohub->token[pirq].pirq = pirq; | ||
279 | + | ||
280 | + qemu_set_fd_handler(msg->fds[1], intr_resample_handler, NULL, | ||
281 | + &iohub->token[pirq]); | ||
282 | +} | ||
283 | diff --git a/hw/remote/machine.c b/hw/remote/machine.c | ||
284 | index XXXXXXX..XXXXXXX 100644 | ||
285 | --- a/hw/remote/machine.c | ||
286 | +++ b/hw/remote/machine.c | ||
287 | @@ -XXX,XX +XXX,XX @@ | ||
288 | #include "exec/address-spaces.h" | ||
289 | #include "exec/memory.h" | ||
290 | #include "qapi/error.h" | ||
291 | +#include "hw/pci/pci_host.h" | ||
292 | +#include "hw/remote/iohub.h" | ||
293 | |||
294 | static void remote_machine_init(MachineState *machine) | ||
295 | { | ||
296 | MemoryRegion *system_memory, *system_io, *pci_memory; | ||
297 | RemoteMachineState *s = REMOTE_MACHINE(machine); | ||
298 | RemotePCIHost *rem_host; | ||
299 | + PCIHostState *pci_host; | ||
300 | |||
301 | system_memory = get_system_memory(); | ||
302 | system_io = get_system_io(); | ||
303 | @@ -XXX,XX +XXX,XX @@ static void remote_machine_init(MachineState *machine) | ||
304 | memory_region_add_subregion_overlap(system_memory, 0x0, pci_memory, -1); | ||
305 | |||
306 | qdev_realize(DEVICE(rem_host), sysbus_get_default(), &error_fatal); | ||
307 | + | ||
308 | + pci_host = PCI_HOST_BRIDGE(rem_host); | ||
309 | + | ||
310 | + remote_iohub_init(&s->iohub); | ||
311 | + | ||
312 | + pci_bus_irqs(pci_host->bus, remote_iohub_set_irq, remote_iohub_map_irq, | ||
313 | + &s->iohub, REMOTE_IOHUB_NB_PIRQS); | ||
23 | } | 314 | } |
24 | 315 | ||
25 | -static void usb_msd_realize_storage(USBDevice *dev, Error **errp) | 316 | static void remote_machine_class_init(ObjectClass *oc, void *data) |
26 | +static void usb_msd_storage_realize(USBDevice *dev, Error **errp) | 317 | diff --git a/hw/remote/message.c b/hw/remote/message.c |
318 | index XXXXXXX..XXXXXXX 100644 | ||
319 | --- a/hw/remote/message.c | ||
320 | +++ b/hw/remote/message.c | ||
321 | @@ -XXX,XX +XXX,XX @@ | ||
322 | #include "hw/pci/pci.h" | ||
323 | #include "exec/memattrs.h" | ||
324 | #include "hw/remote/memory.h" | ||
325 | +#include "hw/remote/iohub.h" | ||
326 | |||
327 | static void process_config_write(QIOChannel *ioc, PCIDevice *dev, | ||
328 | MPQemuMsg *msg, Error **errp); | ||
329 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
330 | case MPQEMU_CMD_SYNC_SYSMEM: | ||
331 | remote_sysmem_reconfig(&msg, &local_err); | ||
332 | break; | ||
333 | + case MPQEMU_CMD_SET_IRQFD: | ||
334 | + process_set_irqfd_msg(pci_dev, &msg); | ||
335 | + break; | ||
336 | default: | ||
337 | error_setg(&local_err, | ||
338 | "Unknown command (%d) received for device %s" | ||
339 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
340 | index XXXXXXX..XXXXXXX 100644 | ||
341 | --- a/hw/remote/mpqemu-link.c | ||
342 | +++ b/hw/remote/mpqemu-link.c | ||
343 | @@ -XXX,XX +XXX,XX @@ bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
344 | return false; | ||
345 | } | ||
346 | break; | ||
347 | + case MPQEMU_CMD_SET_IRQFD: | ||
348 | + if (msg->size || (msg->num_fds != 2)) { | ||
349 | + return false; | ||
350 | + } | ||
351 | + break; | ||
352 | default: | ||
353 | break; | ||
354 | } | ||
355 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
356 | index XXXXXXX..XXXXXXX 100644 | ||
357 | --- a/hw/remote/proxy.c | ||
358 | +++ b/hw/remote/proxy.c | ||
359 | @@ -XXX,XX +XXX,XX @@ | ||
360 | #include "qemu/error-report.h" | ||
361 | #include "hw/remote/proxy-memory-listener.h" | ||
362 | #include "qom/object.h" | ||
363 | +#include "qemu/event_notifier.h" | ||
364 | +#include "sysemu/kvm.h" | ||
365 | +#include "util/event_notifier-posix.c" | ||
366 | + | ||
367 | +static void proxy_intx_update(PCIDevice *pci_dev) | ||
368 | +{ | ||
369 | + PCIProxyDev *dev = PCI_PROXY_DEV(pci_dev); | ||
370 | + PCIINTxRoute route; | ||
371 | + int pin = pci_get_byte(pci_dev->config + PCI_INTERRUPT_PIN) - 1; | ||
372 | + | ||
373 | + if (dev->virq != -1) { | ||
374 | + kvm_irqchip_remove_irqfd_notifier_gsi(kvm_state, &dev->intr, dev->virq); | ||
375 | + dev->virq = -1; | ||
376 | + } | ||
377 | + | ||
378 | + route = pci_device_route_intx_to_irq(pci_dev, pin); | ||
379 | + | ||
380 | + dev->virq = route.irq; | ||
381 | + | ||
382 | + if (dev->virq != -1) { | ||
383 | + kvm_irqchip_add_irqfd_notifier_gsi(kvm_state, &dev->intr, | ||
384 | + &dev->resample, dev->virq); | ||
385 | + } | ||
386 | +} | ||
387 | + | ||
388 | +static void setup_irqfd(PCIProxyDev *dev) | ||
389 | +{ | ||
390 | + PCIDevice *pci_dev = PCI_DEVICE(dev); | ||
391 | + MPQemuMsg msg; | ||
392 | + Error *local_err = NULL; | ||
393 | + | ||
394 | + event_notifier_init(&dev->intr, 0); | ||
395 | + event_notifier_init(&dev->resample, 0); | ||
396 | + | ||
397 | + memset(&msg, 0, sizeof(MPQemuMsg)); | ||
398 | + msg.cmd = MPQEMU_CMD_SET_IRQFD; | ||
399 | + msg.num_fds = 2; | ||
400 | + msg.fds[0] = event_notifier_get_fd(&dev->intr); | ||
401 | + msg.fds[1] = event_notifier_get_fd(&dev->resample); | ||
402 | + msg.size = 0; | ||
403 | + | ||
404 | + if (!mpqemu_msg_send(&msg, dev->ioc, &local_err)) { | ||
405 | + error_report_err(local_err); | ||
406 | + } | ||
407 | + | ||
408 | + dev->virq = -1; | ||
409 | + | ||
410 | + proxy_intx_update(pci_dev); | ||
411 | + | ||
412 | + pci_device_set_intx_routing_notifier(pci_dev, proxy_intx_update); | ||
413 | +} | ||
414 | |||
415 | static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
27 | { | 416 | { |
28 | MSDState *s = USB_STORAGE_DEV(dev); | 417 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) |
29 | BlockBackend *blk = s->conf.blk; | 418 | qio_channel_set_blocking(dev->ioc, true, NULL); |
30 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_realize_storage(USBDevice *dev, Error **errp) | 419 | |
31 | s->scsi_dev = scsi_dev; | 420 | proxy_memory_listener_configure(&dev->proxy_listener, dev->ioc); |
421 | + | ||
422 | + setup_irqfd(dev); | ||
32 | } | 423 | } |
33 | 424 | ||
34 | -static void usb_msd_unrealize_bot(USBDevice *dev, Error **errp) | 425 | static void pci_proxy_dev_exit(PCIDevice *pdev) |
35 | +static void usb_msd_bot_unrealize(USBDevice *dev, Error **errp) | 426 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_exit(PCIDevice *pdev) |
36 | { | 427 | error_free(dev->migration_blocker); |
37 | MSDState *s = USB_STORAGE_DEV(dev); | 428 | |
38 | 429 | proxy_memory_listener_deconfigure(&dev->proxy_listener); | |
39 | object_unref(OBJECT(&s->bus)); | 430 | + |
431 | + event_notifier_cleanup(&dev->intr); | ||
432 | + event_notifier_cleanup(&dev->resample); | ||
40 | } | 433 | } |
41 | 434 | ||
42 | -static void usb_msd_realize_bot(USBDevice *dev, Error **errp) | 435 | static void config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val, |
43 | +static void usb_msd_bot_realize(USBDevice *dev, Error **errp) | 436 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build |
44 | { | 437 | index XXXXXXX..XXXXXXX 100644 |
45 | MSDState *s = USB_STORAGE_DEV(dev); | 438 | --- a/hw/remote/meson.build |
46 | DeviceState *d = DEVICE(dev); | 439 | +++ b/hw/remote/meson.build |
47 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_class_initfn_common(ObjectClass *klass, void *data) | 440 | @@ -XXX,XX +XXX,XX @@ remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) |
48 | dc->vmsd = &vmstate_usb_msd; | 441 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) |
49 | } | 442 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) |
50 | 443 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy.c')) | |
51 | -static void usb_msd_class_initfn_storage(ObjectClass *klass, void *data) | 444 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('iohub.c')) |
52 | +static void usb_msd_class_storage_initfn(ObjectClass *klass, void *data) | 445 | |
53 | { | 446 | specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('memory.c')) |
54 | DeviceClass *dc = DEVICE_CLASS(klass); | 447 | specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy-memory-listener.c')) |
55 | USBDeviceClass *uc = USB_DEVICE_CLASS(klass); | ||
56 | |||
57 | - uc->realize = usb_msd_realize_storage; | ||
58 | + uc->realize = usb_msd_storage_realize; | ||
59 | uc->unrealize = usb_msd_unrealize_storage; | ||
60 | dc->props = msd_properties; | ||
61 | } | ||
62 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_instance_init(Object *obj) | ||
63 | object_property_set_int(obj, -1, "bootindex", NULL); | ||
64 | } | ||
65 | |||
66 | -static void usb_msd_class_initfn_bot(ObjectClass *klass, void *data) | ||
67 | +static void usb_msd_class_bot_initfn(ObjectClass *klass, void *data) | ||
68 | { | ||
69 | USBDeviceClass *uc = USB_DEVICE_CLASS(klass); | ||
70 | |||
71 | - uc->realize = usb_msd_realize_bot; | ||
72 | - uc->unrealize = usb_msd_unrealize_bot; | ||
73 | + uc->realize = usb_msd_bot_realize; | ||
74 | + uc->unrealize = usb_msd_bot_unrealize; | ||
75 | uc->attached_settable = true; | ||
76 | } | ||
77 | |||
78 | static const TypeInfo msd_info = { | ||
79 | .name = "usb-storage", | ||
80 | .parent = TYPE_USB_STORAGE, | ||
81 | - .class_init = usb_msd_class_initfn_storage, | ||
82 | + .class_init = usb_msd_class_storage_initfn, | ||
83 | .instance_init = usb_msd_instance_init, | ||
84 | }; | ||
85 | |||
86 | static const TypeInfo bot_info = { | ||
87 | .name = "usb-bot", | ||
88 | .parent = TYPE_USB_STORAGE, | ||
89 | - .class_init = usb_msd_class_initfn_bot, | ||
90 | + .class_init = usb_msd_class_bot_initfn, | ||
91 | }; | ||
92 | |||
93 | static void usb_msd_register_types(void) | ||
94 | -- | 448 | -- |
95 | 2.14.3 | 449 | 2.29.2 |
96 | 450 | ||
97 | diff view generated by jsdifflib |
1 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 1 | From: Jagannathan Raman <jag.raman@oracle.com> |
---|---|---|---|
2 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | 2 | |
3 | Reviewed-by: Eric Blake <eblake@redhat.com> | 3 | Retrieve PCI configuration info about the remote device and |
4 | Message-id: 20171206144550.22295-4-stefanha@redhat.com | 4 | configure the Proxy PCI object based on the returned information |
5 | |||
6 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
8 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
10 | Message-id: 85ee367bbb993aa23699b44cfedd83b4ea6d5221.1611938319.git.jag.raman@oracle.com | ||
5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
6 | --- | 12 | --- |
7 | blockdev.c | 42 ++++++++++++++++++++++++++++++++++-------- | 13 | hw/remote/proxy.c | 84 +++++++++++++++++++++++++++++++++++++++++++++++ |
8 | 1 file changed, 34 insertions(+), 8 deletions(-) | 14 | 1 file changed, 84 insertions(+) |
9 | 15 | ||
10 | diff --git a/blockdev.c b/blockdev.c | 16 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c |
11 | index XXXXXXX..XXXXXXX 100644 | 17 | index XXXXXXX..XXXXXXX 100644 |
12 | --- a/blockdev.c | 18 | --- a/hw/remote/proxy.c |
13 | +++ b/blockdev.c | 19 | +++ b/hw/remote/proxy.c |
14 | @@ -XXX,XX +XXX,XX @@ static void external_snapshot_clean(BlkActionState *common) | 20 | @@ -XXX,XX +XXX,XX @@ |
15 | typedef struct DriveBackupState { | 21 | #include "sysemu/kvm.h" |
16 | BlkActionState common; | 22 | #include "util/event_notifier-posix.c" |
17 | BlockDriverState *bs; | 23 | |
18 | - AioContext *aio_context; | 24 | +static void probe_pci_info(PCIDevice *dev, Error **errp); |
19 | BlockJob *job; | ||
20 | } DriveBackupState; | ||
21 | |||
22 | @@ -XXX,XX +XXX,XX @@ static void drive_backup_prepare(BlkActionState *common, Error **errp) | ||
23 | DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common); | ||
24 | BlockDriverState *bs; | ||
25 | DriveBackup *backup; | ||
26 | + AioContext *aio_context; | ||
27 | Error *local_err = NULL; | ||
28 | |||
29 | assert(common->action->type == TRANSACTION_ACTION_KIND_DRIVE_BACKUP); | ||
30 | @@ -XXX,XX +XXX,XX @@ static void drive_backup_prepare(BlkActionState *common, Error **errp) | ||
31 | return; | ||
32 | } | ||
33 | |||
34 | - /* AioContext is released in .clean() */ | ||
35 | - state->aio_context = bdrv_get_aio_context(bs); | ||
36 | - aio_context_acquire(state->aio_context); | ||
37 | + aio_context = bdrv_get_aio_context(bs); | ||
38 | + aio_context_acquire(aio_context); | ||
39 | + | 25 | + |
40 | + /* Paired with .clean() */ | 26 | static void proxy_intx_update(PCIDevice *pci_dev) |
41 | bdrv_drained_begin(bs); | 27 | { |
28 | PCIProxyDev *dev = PCI_PROXY_DEV(pci_dev); | ||
29 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
30 | { | ||
31 | ERRP_GUARD(); | ||
32 | PCIProxyDev *dev = PCI_PROXY_DEV(device); | ||
33 | + uint8_t *pci_conf = device->config; | ||
34 | int fd; | ||
35 | |||
36 | if (!dev->fd) { | ||
37 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
38 | qemu_mutex_init(&dev->io_mutex); | ||
39 | qio_channel_set_blocking(dev->ioc, true, NULL); | ||
40 | |||
41 | + pci_conf[PCI_LATENCY_TIMER] = 0xff; | ||
42 | + pci_conf[PCI_INTERRUPT_PIN] = 0x01; | ||
42 | + | 43 | + |
43 | state->bs = bs; | 44 | proxy_memory_listener_configure(&dev->proxy_listener, dev->ioc); |
44 | 45 | ||
45 | state->job = do_drive_backup(backup, common->block_job_txn, &local_err); | 46 | setup_irqfd(dev); |
46 | if (local_err) { | ||
47 | error_propagate(errp, local_err); | ||
48 | - return; | ||
49 | + goto out; | ||
50 | } | ||
51 | + | 47 | + |
52 | +out: | 48 | + probe_pci_info(PCI_DEVICE(dev), errp); |
53 | + aio_context_release(aio_context); | ||
54 | } | 49 | } |
55 | 50 | ||
56 | static void drive_backup_commit(BlkActionState *common) | 51 | static void pci_proxy_dev_exit(PCIDevice *pdev) |
57 | { | 52 | @@ -XXX,XX +XXX,XX @@ const MemoryRegionOps proxy_mr_ops = { |
58 | DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common); | 53 | .max_access_size = 8, |
59 | + AioContext *aio_context; | 54 | }, |
55 | }; | ||
60 | + | 56 | + |
61 | + aio_context = bdrv_get_aio_context(state->bs); | 57 | +static void probe_pci_info(PCIDevice *dev, Error **errp) |
62 | + aio_context_acquire(aio_context); | 58 | +{ |
59 | + PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(dev); | ||
60 | + uint32_t orig_val, new_val, base_class, val; | ||
61 | + PCIProxyDev *pdev = PCI_PROXY_DEV(dev); | ||
62 | + DeviceClass *dc = DEVICE_CLASS(pc); | ||
63 | + uint8_t type; | ||
64 | + int i, size; | ||
63 | + | 65 | + |
64 | assert(state->job); | 66 | + config_op_send(pdev, PCI_VENDOR_ID, &val, 2, MPQEMU_CMD_PCI_CFGREAD); |
65 | block_job_start(state->job); | 67 | + pc->vendor_id = (uint16_t)val; |
66 | + | 68 | + |
67 | + aio_context_release(aio_context); | 69 | + config_op_send(pdev, PCI_DEVICE_ID, &val, 2, MPQEMU_CMD_PCI_CFGREAD); |
68 | } | 70 | + pc->device_id = (uint16_t)val; |
69 | |||
70 | static void drive_backup_abort(BlkActionState *common) | ||
71 | @@ -XXX,XX +XXX,XX @@ static void drive_backup_abort(BlkActionState *common) | ||
72 | DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common); | ||
73 | |||
74 | if (state->job) { | ||
75 | + AioContext *aio_context; | ||
76 | + | 71 | + |
77 | + aio_context = bdrv_get_aio_context(state->bs); | 72 | + config_op_send(pdev, PCI_CLASS_DEVICE, &val, 2, MPQEMU_CMD_PCI_CFGREAD); |
78 | + aio_context_acquire(aio_context); | 73 | + pc->class_id = (uint16_t)val; |
79 | + | 74 | + |
80 | block_job_cancel_sync(state->job); | 75 | + config_op_send(pdev, PCI_SUBSYSTEM_ID, &val, 2, MPQEMU_CMD_PCI_CFGREAD); |
76 | + pc->subsystem_id = (uint16_t)val; | ||
81 | + | 77 | + |
82 | + aio_context_release(aio_context); | 78 | + base_class = pc->class_id >> 4; |
83 | } | 79 | + switch (base_class) { |
84 | } | 80 | + case PCI_BASE_CLASS_BRIDGE: |
85 | 81 | + set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); | |
86 | static void drive_backup_clean(BlkActionState *common) | 82 | + break; |
87 | { | 83 | + case PCI_BASE_CLASS_STORAGE: |
88 | DriveBackupState *state = DO_UPCAST(DriveBackupState, common, common); | 84 | + set_bit(DEVICE_CATEGORY_STORAGE, dc->categories); |
89 | + AioContext *aio_context; | 85 | + break; |
90 | 86 | + case PCI_BASE_CLASS_NETWORK: | |
91 | - if (state->aio_context) { | 87 | + set_bit(DEVICE_CATEGORY_NETWORK, dc->categories); |
92 | - bdrv_drained_end(state->bs); | 88 | + break; |
93 | - aio_context_release(state->aio_context); | 89 | + case PCI_BASE_CLASS_INPUT: |
94 | + if (!state->bs) { | 90 | + set_bit(DEVICE_CATEGORY_INPUT, dc->categories); |
95 | + return; | 91 | + break; |
96 | } | 92 | + case PCI_BASE_CLASS_DISPLAY: |
93 | + set_bit(DEVICE_CATEGORY_DISPLAY, dc->categories); | ||
94 | + break; | ||
95 | + case PCI_BASE_CLASS_PROCESSOR: | ||
96 | + set_bit(DEVICE_CATEGORY_CPU, dc->categories); | ||
97 | + break; | ||
98 | + default: | ||
99 | + set_bit(DEVICE_CATEGORY_MISC, dc->categories); | ||
100 | + break; | ||
101 | + } | ||
97 | + | 102 | + |
98 | + aio_context = bdrv_get_aio_context(state->bs); | 103 | + for (i = 0; i < PCI_NUM_REGIONS; i++) { |
99 | + aio_context_acquire(aio_context); | 104 | + config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &orig_val, 4, |
105 | + MPQEMU_CMD_PCI_CFGREAD); | ||
106 | + new_val = 0xffffffff; | ||
107 | + config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &new_val, 4, | ||
108 | + MPQEMU_CMD_PCI_CFGWRITE); | ||
109 | + config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &new_val, 4, | ||
110 | + MPQEMU_CMD_PCI_CFGREAD); | ||
111 | + size = (~(new_val & 0xFFFFFFF0)) + 1; | ||
112 | + config_op_send(pdev, PCI_BASE_ADDRESS_0 + (4 * i), &orig_val, 4, | ||
113 | + MPQEMU_CMD_PCI_CFGWRITE); | ||
114 | + type = (new_val & 0x1) ? | ||
115 | + PCI_BASE_ADDRESS_SPACE_IO : PCI_BASE_ADDRESS_SPACE_MEMORY; | ||
100 | + | 116 | + |
101 | + bdrv_drained_end(state->bs); | 117 | + if (size) { |
102 | + | 118 | + g_autofree char *name; |
103 | + aio_context_release(aio_context); | 119 | + pdev->region[i].dev = pdev; |
104 | } | 120 | + pdev->region[i].present = true; |
105 | 121 | + if (type == PCI_BASE_ADDRESS_SPACE_MEMORY) { | |
106 | typedef struct BlockdevBackupState { | 122 | + pdev->region[i].memory = true; |
123 | + } | ||
124 | + name = g_strdup_printf("bar-region-%d", i); | ||
125 | + memory_region_init_io(&pdev->region[i].mr, OBJECT(pdev), | ||
126 | + &proxy_mr_ops, &pdev->region[i], | ||
127 | + name, size); | ||
128 | + pci_register_bar(dev, i, type, &pdev->region[i].mr); | ||
129 | + } | ||
130 | + } | ||
131 | +} | ||
107 | -- | 132 | -- |
108 | 2.14.3 | 133 | 2.29.2 |
109 | 134 | ||
110 | diff view generated by jsdifflib |
1 | From: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
---|---|---|---|
2 | 2 | ||
3 | When the function no success value to transmit, it usually make the | 3 | Perform device reset in the remote process when QEMU performs |
4 | function return void. It has turned out not to be a success, because | 4 | device reset. This is required to reset the internal state |
5 | it means that the extra local_err variable and error_propagate() will | 5 | (like registers, etc...) of emulated devices |
6 | be needed. It leads to cumbersome code, therefore, transmit success/ | ||
7 | failure in the return value is worth. | ||
8 | 6 | ||
9 | So fix the return type of blkconf_apply_backend_options(), | 7 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
10 | blkconf_geometry() and virtio_blk_data_plane_create() to avoid it. | 8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
11 | 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | |
12 | Cc: John Snow <jsnow@redhat.com> | ||
13 | Cc: Kevin Wolf <kwolf@redhat.com> | ||
14 | Cc: Max Reitz <mreitz@redhat.com> | ||
15 | Cc: Stefan Hajnoczi <stefanha@redhat.com> | ||
16 | |||
17 | Signed-off-by: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | ||
18 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | 10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
19 | Message-id: ac0edc1fc70c4457e5cec94405eb7d1f89f9c2c1.1511317952.git.maozy.fnst@cn.fujitsu.com | 11 | Message-id: 7cb220a51f565dc0817bd76e2f540e89c2d2b850.1611938319.git.jag.raman@oracle.com |
20 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
21 | --- | 13 | --- |
22 | hw/block/dataplane/virtio-blk.h | 2 +- | 14 | include/hw/remote/mpqemu-link.h | 1 + |
23 | include/hw/block/block.h | 4 ++-- | 15 | hw/remote/message.c | 22 ++++++++++++++++++++++ |
24 | hw/block/block.c | 15 +++++++++------ | 16 | hw/remote/proxy.c | 19 +++++++++++++++++++ |
25 | hw/block/dataplane/virtio-blk.c | 12 +++++++----- | 17 | 3 files changed, 42 insertions(+) |
26 | 4 files changed, 19 insertions(+), 14 deletions(-) | ||
27 | 18 | ||
28 | diff --git a/hw/block/dataplane/virtio-blk.h b/hw/block/dataplane/virtio-blk.h | 19 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h |
29 | index XXXXXXX..XXXXXXX 100644 | 20 | index XXXXXXX..XXXXXXX 100644 |
30 | --- a/hw/block/dataplane/virtio-blk.h | 21 | --- a/include/hw/remote/mpqemu-link.h |
31 | +++ b/hw/block/dataplane/virtio-blk.h | 22 | +++ b/include/hw/remote/mpqemu-link.h |
23 | @@ -XXX,XX +XXX,XX @@ typedef enum { | ||
24 | MPQEMU_CMD_BAR_WRITE, | ||
25 | MPQEMU_CMD_BAR_READ, | ||
26 | MPQEMU_CMD_SET_IRQFD, | ||
27 | + MPQEMU_CMD_DEVICE_RESET, | ||
28 | MPQEMU_CMD_MAX, | ||
29 | } MPQemuCmd; | ||
30 | |||
31 | diff --git a/hw/remote/message.c b/hw/remote/message.c | ||
32 | index XXXXXXX..XXXXXXX 100644 | ||
33 | --- a/hw/remote/message.c | ||
34 | +++ b/hw/remote/message.c | ||
32 | @@ -XXX,XX +XXX,XX @@ | 35 | @@ -XXX,XX +XXX,XX @@ |
33 | 36 | #include "exec/memattrs.h" | |
34 | typedef struct VirtIOBlockDataPlane VirtIOBlockDataPlane; | 37 | #include "hw/remote/memory.h" |
35 | 38 | #include "hw/remote/iohub.h" | |
36 | -void virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | 39 | +#include "sysemu/reset.h" |
37 | +bool virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | 40 | |
38 | VirtIOBlockDataPlane **dataplane, | 41 | static void process_config_write(QIOChannel *ioc, PCIDevice *dev, |
39 | Error **errp); | 42 | MPQemuMsg *msg, Error **errp); |
40 | void virtio_blk_data_plane_destroy(VirtIOBlockDataPlane *s); | 43 | @@ -XXX,XX +XXX,XX @@ static void process_config_read(QIOChannel *ioc, PCIDevice *dev, |
41 | diff --git a/include/hw/block/block.h b/include/hw/block/block.h | 44 | MPQemuMsg *msg, Error **errp); |
42 | index XXXXXXX..XXXXXXX 100644 | 45 | static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp); |
43 | --- a/include/hw/block/block.h | 46 | static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp); |
44 | +++ b/include/hw/block/block.h | 47 | +static void process_device_reset_msg(QIOChannel *ioc, PCIDevice *dev, |
45 | @@ -XXX,XX +XXX,XX @@ static inline unsigned int get_physical_block_exp(BlockConf *conf) | 48 | + Error **errp); |
46 | /* Configuration helpers */ | 49 | |
47 | 50 | void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | |
48 | void blkconf_serial(BlockConf *conf, char **serial); | 51 | { |
49 | -void blkconf_geometry(BlockConf *conf, int *trans, | 52 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) |
50 | +bool blkconf_geometry(BlockConf *conf, int *trans, | 53 | case MPQEMU_CMD_SET_IRQFD: |
51 | unsigned cyls_max, unsigned heads_max, unsigned secs_max, | 54 | process_set_irqfd_msg(pci_dev, &msg); |
52 | Error **errp); | 55 | break; |
53 | void blkconf_blocksizes(BlockConf *conf); | 56 | + case MPQEMU_CMD_DEVICE_RESET: |
54 | -void blkconf_apply_backend_options(BlockConf *conf, bool readonly, | 57 | + process_device_reset_msg(com->ioc, pci_dev, &local_err); |
55 | +bool blkconf_apply_backend_options(BlockConf *conf, bool readonly, | 58 | + break; |
56 | bool resizable, Error **errp); | 59 | default: |
57 | 60 | error_setg(&local_err, | |
58 | /* Hard disk geometry */ | 61 | "Unknown command (%d) received for device %s" |
59 | diff --git a/hw/block/block.c b/hw/block/block.c | 62 | @@ -XXX,XX +XXX,XX @@ fail: |
60 | index XXXXXXX..XXXXXXX 100644 | 63 | getpid()); |
61 | --- a/hw/block/block.c | ||
62 | +++ b/hw/block/block.c | ||
63 | @@ -XXX,XX +XXX,XX @@ void blkconf_blocksizes(BlockConf *conf) | ||
64 | } | 64 | } |
65 | } | 65 | } |
66 | 66 | + | |
67 | -void blkconf_apply_backend_options(BlockConf *conf, bool readonly, | 67 | +static void process_device_reset_msg(QIOChannel *ioc, PCIDevice *dev, |
68 | +bool blkconf_apply_backend_options(BlockConf *conf, bool readonly, | 68 | + Error **errp) |
69 | bool resizable, Error **errp) | 69 | +{ |
70 | + DeviceClass *dc = DEVICE_GET_CLASS(dev); | ||
71 | + DeviceState *s = DEVICE(dev); | ||
72 | + MPQemuMsg ret = { 0 }; | ||
73 | + | ||
74 | + if (dc->reset) { | ||
75 | + dc->reset(s); | ||
76 | + } | ||
77 | + | ||
78 | + ret.cmd = MPQEMU_CMD_RET; | ||
79 | + | ||
80 | + mpqemu_msg_send(&ret, ioc, errp); | ||
81 | +} | ||
82 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
83 | index XXXXXXX..XXXXXXX 100644 | ||
84 | --- a/hw/remote/proxy.c | ||
85 | +++ b/hw/remote/proxy.c | ||
86 | @@ -XXX,XX +XXX,XX @@ | ||
87 | #include "util/event_notifier-posix.c" | ||
88 | |||
89 | static void probe_pci_info(PCIDevice *dev, Error **errp); | ||
90 | +static void proxy_device_reset(DeviceState *dev); | ||
91 | |||
92 | static void proxy_intx_update(PCIDevice *pci_dev) | ||
70 | { | 93 | { |
71 | BlockBackend *blk = conf->blk; | 94 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_class_init(ObjectClass *klass, void *data) |
72 | @@ -XXX,XX +XXX,XX @@ void blkconf_apply_backend_options(BlockConf *conf, bool readonly, | 95 | k->config_read = pci_proxy_read_config; |
73 | 96 | k->config_write = pci_proxy_write_config; | |
74 | ret = blk_set_perm(blk, perm, shared_perm, errp); | 97 | |
75 | if (ret < 0) { | 98 | + dc->reset = proxy_device_reset; |
76 | - return; | ||
77 | + return false; | ||
78 | } | ||
79 | |||
80 | switch (conf->wce) { | ||
81 | @@ -XXX,XX +XXX,XX @@ void blkconf_apply_backend_options(BlockConf *conf, bool readonly, | ||
82 | |||
83 | blk_set_enable_write_cache(blk, wce); | ||
84 | blk_set_on_error(blk, rerror, werror); | ||
85 | + | 99 | + |
86 | + return true; | 100 | device_class_set_props(dc, proxy_properties); |
87 | } | 101 | } |
88 | 102 | ||
89 | -void blkconf_geometry(BlockConf *conf, int *ptrans, | 103 | @@ -XXX,XX +XXX,XX @@ static void probe_pci_info(PCIDevice *dev, Error **errp) |
90 | +bool blkconf_geometry(BlockConf *conf, int *ptrans, | ||
91 | unsigned cyls_max, unsigned heads_max, unsigned secs_max, | ||
92 | Error **errp) | ||
93 | { | ||
94 | @@ -XXX,XX +XXX,XX @@ void blkconf_geometry(BlockConf *conf, int *ptrans, | ||
95 | if (conf->cyls || conf->heads || conf->secs) { | ||
96 | if (conf->cyls < 1 || conf->cyls > cyls_max) { | ||
97 | error_setg(errp, "cyls must be between 1 and %u", cyls_max); | ||
98 | - return; | ||
99 | + return false; | ||
100 | } | ||
101 | if (conf->heads < 1 || conf->heads > heads_max) { | ||
102 | error_setg(errp, "heads must be between 1 and %u", heads_max); | ||
103 | - return; | ||
104 | + return false; | ||
105 | } | ||
106 | if (conf->secs < 1 || conf->secs > secs_max) { | ||
107 | error_setg(errp, "secs must be between 1 and %u", secs_max); | ||
108 | - return; | ||
109 | + return false; | ||
110 | } | 104 | } |
111 | } | 105 | } |
112 | + return true; | ||
113 | } | 106 | } |
114 | diff --git a/hw/block/dataplane/virtio-blk.c b/hw/block/dataplane/virtio-blk.c | ||
115 | index XXXXXXX..XXXXXXX 100644 | ||
116 | --- a/hw/block/dataplane/virtio-blk.c | ||
117 | +++ b/hw/block/dataplane/virtio-blk.c | ||
118 | @@ -XXX,XX +XXX,XX @@ static void notify_guest_bh(void *opaque) | ||
119 | } | ||
120 | |||
121 | /* Context: QEMU global mutex held */ | ||
122 | -void virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | ||
123 | +bool virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | ||
124 | VirtIOBlockDataPlane **dataplane, | ||
125 | Error **errp) | ||
126 | { | ||
127 | @@ -XXX,XX +XXX,XX @@ void virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | ||
128 | error_setg(errp, | ||
129 | "device is incompatible with iothread " | ||
130 | "(transport does not support notifiers)"); | ||
131 | - return; | ||
132 | + return false; | ||
133 | } | ||
134 | if (!virtio_device_ioeventfd_enabled(vdev)) { | ||
135 | error_setg(errp, "ioeventfd is required for iothread"); | ||
136 | - return; | ||
137 | + return false; | ||
138 | } | ||
139 | |||
140 | /* If dataplane is (re-)enabled while the guest is running there could | ||
141 | @@ -XXX,XX +XXX,XX @@ void virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | ||
142 | */ | ||
143 | if (blk_op_is_blocked(conf->conf.blk, BLOCK_OP_TYPE_DATAPLANE, errp)) { | ||
144 | error_prepend(errp, "cannot start virtio-blk dataplane: "); | ||
145 | - return; | ||
146 | + return false; | ||
147 | } | ||
148 | } | ||
149 | /* Don't try if transport does not support notifiers. */ | ||
150 | if (!virtio_device_ioeventfd_enabled(vdev)) { | ||
151 | - return; | ||
152 | + return false; | ||
153 | } | ||
154 | |||
155 | s = g_new0(VirtIOBlockDataPlane, 1); | ||
156 | @@ -XXX,XX +XXX,XX @@ void virtio_blk_data_plane_create(VirtIODevice *vdev, VirtIOBlkConf *conf, | ||
157 | s->batch_notify_vqs = bitmap_new(conf->num_queues); | ||
158 | |||
159 | *dataplane = s; | ||
160 | + | 107 | + |
161 | + return true; | 108 | +static void proxy_device_reset(DeviceState *dev) |
162 | } | 109 | +{ |
163 | 110 | + PCIProxyDev *pdev = PCI_PROXY_DEV(dev); | |
164 | /* Context: QEMU global mutex held */ | 111 | + MPQemuMsg msg = { 0 }; |
112 | + Error *local_err = NULL; | ||
113 | + | ||
114 | + msg.cmd = MPQEMU_CMD_DEVICE_RESET; | ||
115 | + msg.size = 0; | ||
116 | + | ||
117 | + mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err); | ||
118 | + if (local_err) { | ||
119 | + error_report_err(local_err); | ||
120 | + } | ||
121 | + | ||
122 | +} | ||
165 | -- | 123 | -- |
166 | 2.14.3 | 124 | 2.29.2 |
167 | 125 | ||
168 | diff view generated by jsdifflib |
1 | From: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 1 | From: "Denis V. Lunev" <den@openvz.org> |
---|---|---|---|
2 | 2 | ||
3 | Cc: John Snow <jsnow@redhat.com> | 3 | Original specification says that l1 table size if 64 * l1_size, which |
4 | Cc: Kevin Wolf <kwolf@redhat.com> | 4 | is obviously wrong. The size of the l1 entry is 64 _bits_, not bytes. |
5 | Cc: Max Reitz <mreitz@redhat.com> | 5 | Thus 64 is to be replaces with 8 as specification says about bytes. |
6 | Cc: Keith Busch <keith.busch@intel.com> | ||
7 | Cc: Stefan Hajnoczi <stefanha@redhat.com> | ||
8 | Cc: "Michael S. Tsirkin" <mst@redhat.com> | ||
9 | Cc: Paolo Bonzini <pbonzini@redhat.com> | ||
10 | Cc: Gerd Hoffmann <kraxel@redhat.com> | ||
11 | Cc: Markus Armbruster <armbru@redhat.com> | ||
12 | 6 | ||
13 | Signed-off-by: Mao Zhongyi <maozy.fnst@cn.fujitsu.com> | 7 | There is also minor tweak, field name is renamed from l1 to l1_table, |
14 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | 8 | which matches with the later text. |
15 | Message-id: e77848d3735ba590f23ffbf8094379c646c33d79.1511317952.git.maozy.fnst@cn.fujitsu.com | 9 | |
10 | Signed-off-by: Denis V. Lunev <den@openvz.org> | ||
11 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | ||
12 | Message-id: 20210128171313.2210947-1-den@openvz.org | ||
13 | CC: Stefan Hajnoczi <stefanha@redhat.com> | ||
14 | CC: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | ||
15 | |||
16 | [Replace the original commit message "docs: fix mistake in dirty bitmap | ||
17 | feature description" as suggested by Eric Blake. | ||
18 | --Stefan] | ||
19 | |||
16 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 20 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
17 | --- | 21 | --- |
18 | hw/block/fdc.c | 17 ++++++----------- | 22 | docs/interop/parallels.txt | 2 +- |
19 | hw/block/nvme.c | 7 ++----- | 23 | 1 file changed, 1 insertion(+), 1 deletion(-) |
20 | hw/block/virtio-blk.c | 18 ++++++------------ | ||
21 | hw/ide/qdev.c | 12 ++++-------- | ||
22 | hw/scsi/scsi-disk.c | 13 ++++--------- | ||
23 | hw/usb/dev-storage.c | 9 +++------ | ||
24 | 6 files changed, 25 insertions(+), 51 deletions(-) | ||
25 | 24 | ||
26 | diff --git a/hw/block/fdc.c b/hw/block/fdc.c | 25 | diff --git a/docs/interop/parallels.txt b/docs/interop/parallels.txt |
27 | index XXXXXXX..XXXXXXX 100644 | 26 | index XXXXXXX..XXXXXXX 100644 |
28 | --- a/hw/block/fdc.c | 27 | --- a/docs/interop/parallels.txt |
29 | +++ b/hw/block/fdc.c | 28 | +++ b/docs/interop/parallels.txt |
30 | @@ -XXX,XX +XXX,XX @@ static void fd_revalidate(FDrive *drv) | 29 | @@ -XXX,XX +XXX,XX @@ of its data area are: |
31 | static void fd_change_cb(void *opaque, bool load, Error **errp) | 30 | 28 - 31: l1_size |
32 | { | 31 | The number of entries in the L1 table of the bitmap. |
33 | FDrive *drive = opaque; | 32 | |
34 | - Error *local_err = NULL; | 33 | - variable: l1 (64 * l1_size bytes) |
35 | 34 | + variable: l1_table (8 * l1_size bytes) | |
36 | if (!load) { | 35 | L1 offset table (in bytes) |
37 | blk_set_perm(drive->blk, 0, BLK_PERM_ALL, &error_abort); | 36 | |
38 | } else { | 37 | A dirty bitmap is stored using a one-level structure for the mapping to host |
39 | - blkconf_apply_backend_options(drive->conf, | ||
40 | - blk_is_read_only(drive->blk), false, | ||
41 | - &local_err); | ||
42 | - if (local_err) { | ||
43 | - error_propagate(errp, local_err); | ||
44 | + if (!blkconf_apply_backend_options(drive->conf, | ||
45 | + blk_is_read_only(drive->blk), false, | ||
46 | + errp)) { | ||
47 | return; | ||
48 | } | ||
49 | } | ||
50 | @@ -XXX,XX +XXX,XX @@ static void floppy_drive_realize(DeviceState *qdev, Error **errp) | ||
51 | FloppyDrive *dev = FLOPPY_DRIVE(qdev); | ||
52 | FloppyBus *bus = FLOPPY_BUS(qdev->parent_bus); | ||
53 | FDrive *drive; | ||
54 | - Error *local_err = NULL; | ||
55 | int ret; | ||
56 | |||
57 | if (dev->unit == -1) { | ||
58 | @@ -XXX,XX +XXX,XX @@ static void floppy_drive_realize(DeviceState *qdev, Error **errp) | ||
59 | dev->conf.rerror = BLOCKDEV_ON_ERROR_AUTO; | ||
60 | dev->conf.werror = BLOCKDEV_ON_ERROR_AUTO; | ||
61 | |||
62 | - blkconf_apply_backend_options(&dev->conf, blk_is_read_only(dev->conf.blk), | ||
63 | - false, &local_err); | ||
64 | - if (local_err) { | ||
65 | - error_propagate(errp, local_err); | ||
66 | + if (!blkconf_apply_backend_options(&dev->conf, | ||
67 | + blk_is_read_only(dev->conf.blk), | ||
68 | + false, errp)) { | ||
69 | return; | ||
70 | } | ||
71 | |||
72 | diff --git a/hw/block/nvme.c b/hw/block/nvme.c | ||
73 | index XXXXXXX..XXXXXXX 100644 | ||
74 | --- a/hw/block/nvme.c | ||
75 | +++ b/hw/block/nvme.c | ||
76 | @@ -XXX,XX +XXX,XX @@ static void nvme_realize(PCIDevice *pci_dev, Error **errp) | ||
77 | int i; | ||
78 | int64_t bs_size; | ||
79 | uint8_t *pci_conf; | ||
80 | - Error *local_err = NULL; | ||
81 | |||
82 | if (!n->conf.blk) { | ||
83 | error_setg(errp, "drive property not set"); | ||
84 | @@ -XXX,XX +XXX,XX @@ static void nvme_realize(PCIDevice *pci_dev, Error **errp) | ||
85 | return; | ||
86 | } | ||
87 | blkconf_blocksizes(&n->conf); | ||
88 | - blkconf_apply_backend_options(&n->conf, blk_is_read_only(n->conf.blk), | ||
89 | - false, &local_err); | ||
90 | - if (local_err) { | ||
91 | - error_propagate(errp, local_err); | ||
92 | + if (!blkconf_apply_backend_options(&n->conf, blk_is_read_only(n->conf.blk), | ||
93 | + false, errp)) { | ||
94 | return; | ||
95 | } | ||
96 | |||
97 | diff --git a/hw/block/virtio-blk.c b/hw/block/virtio-blk.c | ||
98 | index XXXXXXX..XXXXXXX 100644 | ||
99 | --- a/hw/block/virtio-blk.c | ||
100 | +++ b/hw/block/virtio-blk.c | ||
101 | @@ -XXX,XX +XXX,XX @@ static void virtio_blk_device_realize(DeviceState *dev, Error **errp) | ||
102 | VirtIODevice *vdev = VIRTIO_DEVICE(dev); | ||
103 | VirtIOBlock *s = VIRTIO_BLK(dev); | ||
104 | VirtIOBlkConf *conf = &s->conf; | ||
105 | - Error *err = NULL; | ||
106 | unsigned i; | ||
107 | |||
108 | if (!conf->conf.blk) { | ||
109 | @@ -XXX,XX +XXX,XX @@ static void virtio_blk_device_realize(DeviceState *dev, Error **errp) | ||
110 | } | ||
111 | |||
112 | blkconf_serial(&conf->conf, &conf->serial); | ||
113 | - blkconf_apply_backend_options(&conf->conf, | ||
114 | - blk_is_read_only(conf->conf.blk), true, | ||
115 | - &err); | ||
116 | - if (err) { | ||
117 | - error_propagate(errp, err); | ||
118 | + if (!blkconf_apply_backend_options(&conf->conf, | ||
119 | + blk_is_read_only(conf->conf.blk), true, | ||
120 | + errp)) { | ||
121 | return; | ||
122 | } | ||
123 | s->original_wce = blk_enable_write_cache(conf->conf.blk); | ||
124 | - blkconf_geometry(&conf->conf, NULL, 65535, 255, 255, &err); | ||
125 | - if (err) { | ||
126 | - error_propagate(errp, err); | ||
127 | + if (!blkconf_geometry(&conf->conf, NULL, 65535, 255, 255, errp)) { | ||
128 | return; | ||
129 | } | ||
130 | + | ||
131 | blkconf_blocksizes(&conf->conf); | ||
132 | |||
133 | virtio_init(vdev, "virtio-blk", VIRTIO_ID_BLOCK, | ||
134 | @@ -XXX,XX +XXX,XX @@ static void virtio_blk_device_realize(DeviceState *dev, Error **errp) | ||
135 | for (i = 0; i < conf->num_queues; i++) { | ||
136 | virtio_add_queue(vdev, 128, virtio_blk_handle_output); | ||
137 | } | ||
138 | - virtio_blk_data_plane_create(vdev, conf, &s->dataplane, &err); | ||
139 | - if (err != NULL) { | ||
140 | - error_propagate(errp, err); | ||
141 | + if (!virtio_blk_data_plane_create(vdev, conf, &s->dataplane, errp)) { | ||
142 | virtio_cleanup(vdev); | ||
143 | return; | ||
144 | } | ||
145 | diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c | ||
146 | index XXXXXXX..XXXXXXX 100644 | ||
147 | --- a/hw/ide/qdev.c | ||
148 | +++ b/hw/ide/qdev.c | ||
149 | @@ -XXX,XX +XXX,XX @@ static void ide_dev_initfn(IDEDevice *dev, IDEDriveKind kind, Error **errp) | ||
150 | { | ||
151 | IDEBus *bus = DO_UPCAST(IDEBus, qbus, dev->qdev.parent_bus); | ||
152 | IDEState *s = bus->ifs + dev->unit; | ||
153 | - Error *err = NULL; | ||
154 | int ret; | ||
155 | |||
156 | if (!dev->conf.blk) { | ||
157 | @@ -XXX,XX +XXX,XX @@ static void ide_dev_initfn(IDEDevice *dev, IDEDriveKind kind, Error **errp) | ||
158 | |||
159 | blkconf_serial(&dev->conf, &dev->serial); | ||
160 | if (kind != IDE_CD) { | ||
161 | - blkconf_geometry(&dev->conf, &dev->chs_trans, 65535, 16, 255, &err); | ||
162 | - if (err) { | ||
163 | - error_propagate(errp, err); | ||
164 | + if (!blkconf_geometry(&dev->conf, &dev->chs_trans, 65535, 16, 255, | ||
165 | + errp)) { | ||
166 | return; | ||
167 | } | ||
168 | } | ||
169 | - blkconf_apply_backend_options(&dev->conf, kind == IDE_CD, kind != IDE_CD, | ||
170 | - &err); | ||
171 | - if (err) { | ||
172 | - error_propagate(errp, err); | ||
173 | + if (!blkconf_apply_backend_options(&dev->conf, kind == IDE_CD, | ||
174 | + kind != IDE_CD, errp)) { | ||
175 | return; | ||
176 | } | ||
177 | |||
178 | diff --git a/hw/scsi/scsi-disk.c b/hw/scsi/scsi-disk.c | ||
179 | index XXXXXXX..XXXXXXX 100644 | ||
180 | --- a/hw/scsi/scsi-disk.c | ||
181 | +++ b/hw/scsi/scsi-disk.c | ||
182 | @@ -XXX,XX +XXX,XX @@ static void scsi_disk_unit_attention_reported(SCSIDevice *dev) | ||
183 | static void scsi_realize(SCSIDevice *dev, Error **errp) | ||
184 | { | ||
185 | SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, dev); | ||
186 | - Error *err = NULL; | ||
187 | |||
188 | if (!s->qdev.conf.blk) { | ||
189 | error_setg(errp, "drive property not set"); | ||
190 | @@ -XXX,XX +XXX,XX @@ static void scsi_realize(SCSIDevice *dev, Error **errp) | ||
191 | } | ||
192 | |||
193 | if (dev->type == TYPE_DISK) { | ||
194 | - blkconf_geometry(&dev->conf, NULL, 65535, 255, 255, &err); | ||
195 | - if (err) { | ||
196 | - error_propagate(errp, err); | ||
197 | + if (!blkconf_geometry(&dev->conf, NULL, 65535, 255, 255, errp)) { | ||
198 | return; | ||
199 | } | ||
200 | } | ||
201 | - blkconf_apply_backend_options(&dev->conf, | ||
202 | - blk_is_read_only(s->qdev.conf.blk), | ||
203 | - dev->type == TYPE_DISK, &err); | ||
204 | - if (err) { | ||
205 | - error_propagate(errp, err); | ||
206 | + if (!blkconf_apply_backend_options(&dev->conf, | ||
207 | + blk_is_read_only(s->qdev.conf.blk), | ||
208 | + dev->type == TYPE_DISK, errp)) { | ||
209 | return; | ||
210 | } | ||
211 | |||
212 | diff --git a/hw/usb/dev-storage.c b/hw/usb/dev-storage.c | ||
213 | index XXXXXXX..XXXXXXX 100644 | ||
214 | --- a/hw/usb/dev-storage.c | ||
215 | +++ b/hw/usb/dev-storage.c | ||
216 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_realize_storage(USBDevice *dev, Error **errp) | ||
217 | MSDState *s = USB_STORAGE_DEV(dev); | ||
218 | BlockBackend *blk = s->conf.blk; | ||
219 | SCSIDevice *scsi_dev; | ||
220 | - Error *err = NULL; | ||
221 | |||
222 | if (!blk) { | ||
223 | error_setg(errp, "drive property not set"); | ||
224 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_realize_storage(USBDevice *dev, Error **errp) | ||
225 | |||
226 | blkconf_serial(&s->conf, &dev->serial); | ||
227 | blkconf_blocksizes(&s->conf); | ||
228 | - blkconf_apply_backend_options(&s->conf, blk_is_read_only(blk), true, &err); | ||
229 | - if (err) { | ||
230 | - error_propagate(errp, err); | ||
231 | + if (!blkconf_apply_backend_options(&s->conf, blk_is_read_only(blk), true, | ||
232 | + errp)) { | ||
233 | return; | ||
234 | } | ||
235 | |||
236 | @@ -XXX,XX +XXX,XX @@ static void usb_msd_realize_storage(USBDevice *dev, Error **errp) | ||
237 | &usb_msd_scsi_info_storage, NULL); | ||
238 | scsi_dev = scsi_bus_legacy_add_drive(&s->bus, blk, 0, !!s->removable, | ||
239 | s->conf.bootindex, dev->serial, | ||
240 | - &err); | ||
241 | + errp); | ||
242 | blk_unref(blk); | ||
243 | if (!scsi_dev) { | ||
244 | - error_propagate(errp, err); | ||
245 | return; | ||
246 | } | ||
247 | usb_msd_handle_reset(dev); | ||
248 | -- | 38 | -- |
249 | 2.14.3 | 39 | 2.29.2 |
250 | 40 | ||
251 | diff view generated by jsdifflib |