| 1 | The following changes since commit 6c769690ac845fa62642a5f93b4e4bd906adab95: | 1 | The following changes since commit 1214d55d1c41fbab3a9973a05085b8760647e411: |
|---|---|---|---|
| 2 | 2 | ||
| 3 | Merge remote-tracking branch 'remotes/vsementsov/tags/pull-simplebench-2021-05-04' into staging (2021-05-21 12:02:34 +0100) | 3 | Merge remote-tracking branch 'remotes/nvme/tags/nvme-next-pull-request' into staging (2021-02-09 13:24:37 +0000) |
| 4 | 4 | ||
| 5 | are available in the Git repository at: | 5 | are available in the Git repository at: |
| 6 | 6 | ||
| 7 | https://gitlab.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 0a6f0c76a030710780ce10d6347a70f098024d21: | 9 | for you to fetch changes up to eb847c42296497978942f738cd41dc29a35a49b2: |
| 10 | 10 | ||
| 11 | coroutine-sleep: introduce qemu_co_sleep (2021-05-21 18:22:33 +0100) | 11 | docs: fix Parallels Image "dirty bitmap" section (2021-02-10 09:23:28 +0000) |
| 12 | 12 | ||
| 13 | ---------------------------------------------------------------- | 13 | ---------------------------------------------------------------- |
| 14 | Pull request | 14 | Pull request |
| 15 | 15 | ||
| 16 | v4: | ||
| 17 | * Add PCI_EXPRESS Kconfig dependency to fix s390x in "multi-process: setup PCI | ||
| 18 | host bridge for remote device" [Philippe and Thomas] | ||
| 19 | |||
| 16 | ---------------------------------------------------------------- | 20 | ---------------------------------------------------------------- |
| 17 | 21 | ||
| 18 | Paolo Bonzini (6): | 22 | Denis V. Lunev (1): |
| 19 | coroutine-sleep: use a stack-allocated timer | 23 | docs: fix Parallels Image "dirty bitmap" section |
| 20 | coroutine-sleep: disallow NULL QemuCoSleepState** argument | ||
| 21 | coroutine-sleep: allow qemu_co_sleep_wake that wakes nothing | ||
| 22 | coroutine-sleep: move timer out of QemuCoSleepState | ||
| 23 | coroutine-sleep: replace QemuCoSleepState pointer with struct in the | ||
| 24 | API | ||
| 25 | coroutine-sleep: introduce qemu_co_sleep | ||
| 26 | 24 | ||
| 27 | Philippe Mathieu-Daudé (1): | 25 | Elena Ufimtseva (8): |
| 28 | bitops.h: Improve find_xxx_bit() documentation | 26 | multi-process: add configure and usage information |
| 27 | io: add qio_channel_writev_full_all helper | ||
| 28 | io: add qio_channel_readv_full_all_eof & qio_channel_readv_full_all | ||
| 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 | ||
| 29 | 35 | ||
| 30 | Zenghui Yu (1): | 36 | Jagannathan Raman (11): |
| 31 | multi-process: Initialize variables declared with g_auto* | 37 | memory: alloc RAM from file at offset |
| 38 | multi-process: Add config option for multi-process QEMU | ||
| 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 | ||
| 32 | 49 | ||
| 33 | include/qemu/bitops.h | 15 ++++++-- | 50 | John G Johnson (1): |
| 34 | include/qemu/coroutine.h | 27 ++++++++----- | 51 | multi-process: add the concept description to |
| 35 | block/block-copy.c | 10 ++--- | 52 | docs/devel/qemu-multiprocess |
| 36 | block/nbd.c | 14 +++---- | 53 | |
| 37 | hw/remote/memory.c | 5 +-- | 54 | Stefan Hajnoczi (6): |
| 38 | hw/remote/proxy.c | 3 +- | 55 | .github: point Repo Lockdown bot to GitLab repo |
| 39 | util/qemu-coroutine-sleep.c | 75 +++++++++++++++++++------------------ | 56 | gitmodules: use GitLab repos instead of qemu.org |
| 40 | 7 files changed, 79 insertions(+), 70 deletions(-) | 57 | gitlab-ci: remove redundant GitLab repo URL command |
| 58 | docs: update README to use GitLab repo URLs | ||
| 59 | pc-bios: update mirror URLs to GitLab | ||
| 60 | get_maintainer: update repo URL to GitLab | ||
| 61 | |||
| 62 | MAINTAINERS | 24 + | ||
| 63 | README.rst | 4 +- | ||
| 64 | docs/devel/index.rst | 1 + | ||
| 65 | docs/devel/multi-process.rst | 966 ++++++++++++++++++++++ | ||
| 66 | docs/system/index.rst | 1 + | ||
| 67 | docs/system/multi-process.rst | 64 ++ | ||
| 68 | docs/interop/parallels.txt | 2 +- | ||
| 69 | configure | 10 + | ||
| 70 | meson.build | 5 +- | ||
| 71 | hw/remote/trace.h | 1 + | ||
| 72 | include/exec/memory.h | 2 + | ||
| 73 | include/exec/ram_addr.h | 4 +- | ||
| 74 | include/hw/pci-host/remote.h | 30 + | ||
| 75 | include/hw/pci/pci_ids.h | 3 + | ||
| 76 | include/hw/remote/iohub.h | 42 + | ||
| 77 | include/hw/remote/machine.h | 38 + | ||
| 78 | include/hw/remote/memory.h | 19 + | ||
| 79 | include/hw/remote/mpqemu-link.h | 99 +++ | ||
| 80 | include/hw/remote/proxy-memory-listener.h | 28 + | ||
| 81 | include/hw/remote/proxy.h | 48 ++ | ||
| 82 | include/io/channel.h | 78 ++ | ||
| 83 | include/qemu/mmap-alloc.h | 4 +- | ||
| 84 | include/sysemu/iothread.h | 6 + | ||
| 85 | backends/hostmem-memfd.c | 2 +- | ||
| 86 | hw/misc/ivshmem.c | 3 +- | ||
| 87 | hw/pci-host/remote.c | 75 ++ | ||
| 88 | hw/remote/iohub.c | 119 +++ | ||
| 89 | hw/remote/machine.c | 80 ++ | ||
| 90 | hw/remote/memory.c | 65 ++ | ||
| 91 | hw/remote/message.c | 230 ++++++ | ||
| 92 | hw/remote/mpqemu-link.c | 267 ++++++ | ||
| 93 | hw/remote/proxy-memory-listener.c | 227 +++++ | ||
| 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 | ||
| 41 | 138 | ||
| 42 | -- | 139 | -- |
| 43 | 2.31.1 | 140 | 2.29.2 |
| 44 | 141 | 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 |
New patch | |||
|---|---|---|---|
| 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). | ||
| 1 | 4 | ||
| 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@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 | ||
| 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 11 | --- | ||
| 12 | .gitmodules | 44 ++++++++++++++++++++++---------------------- | ||
| 13 | 1 file changed, 22 insertions(+), 22 deletions(-) | ||
| 14 | |||
| 15 | diff --git a/.gitmodules b/.gitmodules | ||
| 16 | index XXXXXXX..XXXXXXX 100644 | ||
| 17 | --- a/.gitmodules | ||
| 18 | +++ b/.gitmodules | ||
| 19 | @@ -XXX,XX +XXX,XX @@ | ||
| 20 | [submodule "roms/seabios"] | ||
| 21 | path = roms/seabios | ||
| 22 | - url = https://git.qemu.org/git/seabios.git/ | ||
| 23 | + url = https://gitlab.com/qemu-project/seabios.git/ | ||
| 24 | [submodule "roms/SLOF"] | ||
| 25 | path = roms/SLOF | ||
| 26 | - url = https://git.qemu.org/git/SLOF.git | ||
| 27 | + url = https://gitlab.com/qemu-project/SLOF.git | ||
| 28 | [submodule "roms/ipxe"] | ||
| 29 | path = roms/ipxe | ||
| 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 | ||
| 108 | -- | ||
| 109 | 2.29.2 | ||
| 110 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | It is no longer necessary to point .gitmodules at GitLab repos when | ||
| 2 | running in GitLab CI since they are now used all the time. | ||
| 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 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
| 8 | Message-id: 20210111115017.156802-4-stefanha@redhat.com | ||
| 9 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 10 | --- | ||
| 11 | .gitlab-ci.yml | 1 - | ||
| 12 | 1 file changed, 1 deletion(-) | ||
| 13 | |||
| 14 | diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml | ||
| 15 | index XXXXXXX..XXXXXXX 100644 | ||
| 16 | --- a/.gitlab-ci.yml | ||
| 17 | +++ b/.gitlab-ci.yml | ||
| 18 | @@ -XXX,XX +XXX,XX @@ include: | ||
| 19 | image: $CI_REGISTRY_IMAGE/qemu/$IMAGE:latest | ||
| 20 | before_script: | ||
| 21 | - JOBS=$(expr $(nproc) + 1) | ||
| 22 | - - sed -i s,git.qemu.org/git,gitlab.com/qemu-project, .gitmodules | ||
| 23 | script: | ||
| 24 | - mkdir build | ||
| 25 | - cd build | ||
| 26 | -- | ||
| 27 | 2.29.2 | ||
| 28 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 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). | ||
| 1 | 4 | ||
| 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@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 | ||
| 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 11 | --- | ||
| 12 | README.rst | 4 ++-- | ||
| 13 | 1 file changed, 2 insertions(+), 2 deletions(-) | ||
| 14 | |||
| 15 | diff --git a/README.rst b/README.rst | ||
| 16 | index XXXXXXX..XXXXXXX 100644 | ||
| 17 | --- a/README.rst | ||
| 18 | +++ b/README.rst | ||
| 19 | @@ -XXX,XX +XXX,XX @@ The QEMU source code is maintained under the GIT version control system. | ||
| 20 | |||
| 21 | .. code-block:: shell | ||
| 22 | |||
| 23 | - git clone https://git.qemu.org/git/qemu.git | ||
| 24 | + git clone https://gitlab.com/qemu-project/qemu.git | ||
| 25 | |||
| 26 | When submitting patches, one common approach is to use 'git | ||
| 27 | format-patch' and/or 'git send-email' to format & send the mail to the | ||
| 28 | @@ -XXX,XX +XXX,XX @@ The QEMU website is also maintained under source control. | ||
| 29 | |||
| 30 | .. code-block:: shell | ||
| 31 | |||
| 32 | - git clone https://git.qemu.org/git/qemu-web.git | ||
| 33 | + git clone https://gitlab.com/qemu-project/qemu-web.git | ||
| 34 | |||
| 35 | * `<https://www.qemu.org/2017/02/04/the-new-qemu-website-is-up/>`_ | ||
| 36 | |||
| 37 | -- | ||
| 38 | 2.29.2 | ||
| 39 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 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). | ||
| 1 | 4 | ||
| 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@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 | ||
| 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 11 | --- | ||
| 12 | pc-bios/README | 4 ++-- | ||
| 13 | 1 file changed, 2 insertions(+), 2 deletions(-) | ||
| 14 | |||
| 15 | diff --git a/pc-bios/README b/pc-bios/README | ||
| 16 | index XXXXXXX..XXXXXXX 100644 | ||
| 17 | --- a/pc-bios/README | ||
| 18 | +++ b/pc-bios/README | ||
| 19 | @@ -XXX,XX +XXX,XX @@ | ||
| 20 | legacy x86 software to communicate with an attached serial console as | ||
| 21 | if a video card were attached. The master sources reside in a subversion | ||
| 22 | repository at http://sgabios.googlecode.com/svn/trunk. A git mirror is | ||
| 23 | - available at https://git.qemu.org/git/sgabios.git. | ||
| 24 | + available at https://gitlab.com/qemu-project/sgabios.git. | ||
| 25 | |||
| 26 | - The PXE roms come from the iPXE project. Built with BANNER_TIME 0. | ||
| 27 | Sources available at http://ipxe.org. Vendor:Device ID -> ROM mapping: | ||
| 28 | @@ -XXX,XX +XXX,XX @@ | ||
| 29 | |||
| 30 | - The u-boot binary for e500 comes from the upstream denx u-boot project where | ||
| 31 | it was compiled using the qemu-ppce500 target. | ||
| 32 | - A git mirror is available at: https://git.qemu.org/git/u-boot.git | ||
| 33 | + A git mirror is available at: https://gitlab.com/qemu-project/u-boot.git | ||
| 34 | The hash used to compile the current version is: 2072e72 | ||
| 35 | |||
| 36 | - Skiboot (https://github.com/open-power/skiboot/) is an OPAL | ||
| 37 | -- | ||
| 38 | 2.29.2 | ||
| 39 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 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). | ||
| 1 | 4 | ||
| 5 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 6 | Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com> | ||
| 7 | Reviewed-by: Thomas Huth <thuth@redhat.com> | ||
| 8 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
| 9 | Message-id: 20210111115017.156802-7-stefanha@redhat.com | ||
| 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 11 | --- | ||
| 12 | scripts/get_maintainer.pl | 2 +- | ||
| 13 | 1 file changed, 1 insertion(+), 1 deletion(-) | ||
| 14 | |||
| 15 | diff --git a/scripts/get_maintainer.pl b/scripts/get_maintainer.pl | ||
| 16 | index XXXXXXX..XXXXXXX 100755 | ||
| 17 | --- a/scripts/get_maintainer.pl | ||
| 18 | +++ b/scripts/get_maintainer.pl | ||
| 19 | @@ -XXX,XX +XXX,XX @@ sub vcs_exists { | ||
| 20 | warn("$P: No supported VCS found. Add --nogit to options?\n"); | ||
| 21 | warn("Using a git repository produces better results.\n"); | ||
| 22 | warn("Try latest git repository using:\n"); | ||
| 23 | - warn("git clone https://git.qemu.org/git/qemu.git\n"); | ||
| 24 | + warn("git clone https://gitlab.com/qemu-project/qemu.git\n"); | ||
| 25 | $printed_novcs = 1; | ||
| 26 | } | ||
| 27 | return 0; | ||
| 28 | -- | ||
| 29 | 2.29.2 | ||
| 30 | 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 |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Jagannathan Raman <jag.raman@oracle.com> | |
| 2 | |||
| 3 | Allow RAM MemoryRegion to be created from an offset in a file, instead | ||
| 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 | ||
| 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 13 | --- | ||
| 14 | include/exec/memory.h | 2 ++ | ||
| 15 | include/exec/ram_addr.h | 4 ++-- | ||
| 16 | include/qemu/mmap-alloc.h | 4 +++- | ||
| 17 | backends/hostmem-memfd.c | 2 +- | ||
| 18 | hw/misc/ivshmem.c | 3 ++- | ||
| 19 | softmmu/memory.c | 3 ++- | ||
| 20 | softmmu/physmem.c | 12 +++++++----- | ||
| 21 | util/mmap-alloc.c | 8 +++++--- | ||
| 22 | util/oslib-posix.c | 2 +- | ||
| 23 | 9 files changed, 25 insertions(+), 15 deletions(-) | ||
| 24 | |||
| 25 | diff --git a/include/exec/memory.h b/include/exec/memory.h | ||
| 26 | index XXXXXXX..XXXXXXX 100644 | ||
| 27 | --- a/include/exec/memory.h | ||
| 28 | +++ b/include/exec/memory.h | ||
| 29 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_file(MemoryRegion *mr, | ||
| 30 | * @size: size of the region. | ||
| 31 | * @share: %true if memory must be mmaped with the MAP_SHARED flag | ||
| 32 | * @fd: the fd to mmap. | ||
| 33 | + * @offset: offset within the file referenced by fd | ||
| 34 | * @errp: pointer to Error*, to store an error if it happens. | ||
| 35 | * | ||
| 36 | * Note that this function does not do anything to cause the data in the | ||
| 37 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_fd(MemoryRegion *mr, | ||
| 38 | uint64_t size, | ||
| 39 | bool share, | ||
| 40 | int fd, | ||
| 41 | + ram_addr_t offset, | ||
| 42 | Error **errp); | ||
| 43 | #endif | ||
| 44 | |||
| 45 | diff --git a/include/exec/ram_addr.h b/include/exec/ram_addr.h | ||
| 46 | index XXXXXXX..XXXXXXX 100644 | ||
| 47 | --- a/include/exec/ram_addr.h | ||
| 48 | +++ b/include/exec/ram_addr.h | ||
| 49 | @@ -XXX,XX +XXX,XX @@ RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr, | ||
| 50 | uint32_t ram_flags, const char *mem_path, | ||
| 51 | bool readonly, Error **errp); | ||
| 52 | RAMBlock *qemu_ram_alloc_from_fd(ram_addr_t size, MemoryRegion *mr, | ||
| 53 | - uint32_t ram_flags, int fd, bool readonly, | ||
| 54 | - Error **errp); | ||
| 55 | + uint32_t ram_flags, int fd, off_t offset, | ||
| 56 | + bool readonly, Error **errp); | ||
| 57 | |||
| 58 | RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host, | ||
| 59 | MemoryRegion *mr, Error **errp); | ||
| 60 | diff --git a/include/qemu/mmap-alloc.h b/include/qemu/mmap-alloc.h | ||
| 61 | index XXXXXXX..XXXXXXX 100644 | ||
| 62 | --- a/include/qemu/mmap-alloc.h | ||
| 63 | +++ b/include/qemu/mmap-alloc.h | ||
| 64 | @@ -XXX,XX +XXX,XX @@ size_t qemu_mempath_getpagesize(const char *mem_path); | ||
| 65 | * @readonly: true for a read-only mapping, false for read/write. | ||
| 66 | * @shared: map has RAM_SHARED flag. | ||
| 67 | * @is_pmem: map has RAM_PMEM flag. | ||
| 68 | + * @map_offset: map starts at offset of map_offset from the start of fd | ||
| 69 | * | ||
| 70 | * Return: | ||
| 71 | * On success, return a pointer to the mapped area. | ||
| 72 | @@ -XXX,XX +XXX,XX @@ void *qemu_ram_mmap(int fd, | ||
| 73 | size_t align, | ||
| 74 | bool readonly, | ||
| 75 | bool shared, | ||
| 76 | - bool is_pmem); | ||
| 77 | + bool is_pmem, | ||
| 78 | + off_t map_offset); | ||
| 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); | ||
| 106 | if (local_err) { | ||
| 107 | error_propagate(errp, local_err); | ||
| 108 | return; | ||
| 109 | diff --git a/softmmu/memory.c b/softmmu/memory.c | ||
| 110 | index XXXXXXX..XXXXXXX 100644 | ||
| 111 | --- a/softmmu/memory.c | ||
| 112 | +++ b/softmmu/memory.c | ||
| 113 | @@ -XXX,XX +XXX,XX @@ void memory_region_init_ram_from_fd(MemoryRegion *mr, | ||
| 114 | uint64_t size, | ||
| 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, | ||
| 143 | } | ||
| 144 | |||
| 145 | area = qemu_ram_mmap(fd, memory, block->mr->align, readonly, | ||
| 146 | - block->flags & RAM_SHARED, block->flags & RAM_PMEM); | ||
| 147 | + block->flags & RAM_SHARED, block->flags & RAM_PMEM, | ||
| 148 | + offset); | ||
| 149 | if (area == MAP_FAILED) { | ||
| 150 | error_setg_errno(errp, errno, | ||
| 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; | ||
| 174 | } | ||
| 175 | |||
| 176 | - block = qemu_ram_alloc_from_fd(size, mr, ram_flags, fd, readonly, errp); | ||
| 177 | + block = qemu_ram_alloc_from_fd(size, mr, ram_flags, fd, 0, readonly, errp); | ||
| 178 | if (!block) { | ||
| 179 | if (created) { | ||
| 180 | unlink(mem_path); | ||
| 181 | diff --git a/util/mmap-alloc.c b/util/mmap-alloc.c | ||
| 182 | index XXXXXXX..XXXXXXX 100644 | ||
| 183 | --- a/util/mmap-alloc.c | ||
| 184 | +++ b/util/mmap-alloc.c | ||
| 185 | @@ -XXX,XX +XXX,XX @@ void *qemu_ram_mmap(int fd, | ||
| 186 | size_t align, | ||
| 187 | bool readonly, | ||
| 188 | bool shared, | ||
| 189 | - bool is_pmem) | ||
| 190 | + bool is_pmem, | ||
| 191 | + off_t map_offset) | ||
| 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); | ||
| 211 | } | ||
| 212 | |||
| 213 | if (ptr == MAP_FAILED) { | ||
| 214 | diff --git a/util/oslib-posix.c b/util/oslib-posix.c | ||
| 215 | index XXXXXXX..XXXXXXX 100644 | ||
| 216 | --- a/util/oslib-posix.c | ||
| 217 | +++ b/util/oslib-posix.c | ||
| 218 | @@ -XXX,XX +XXX,XX @@ void *qemu_memalign(size_t alignment, size_t size) | ||
| 219 | void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment, bool shared) | ||
| 220 | { | ||
| 221 | size_t align = QEMU_VMALLOC_ALIGN; | ||
| 222 | - void *ptr = qemu_ram_mmap(-1, size, align, false, shared, false); | ||
| 223 | + void *ptr = qemu_ram_mmap(-1, size, align, false, shared, false, 0); | ||
| 224 | |||
| 225 | if (ptr == MAP_FAILED) { | ||
| 226 | return NULL; | ||
| 227 | -- | ||
| 228 | 2.29.2 | ||
| 229 | 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 |
New patch | |||
|---|---|---|---|
| 1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
| 1 | 2 | ||
| 3 | PCI host bridge is setup for the remote device process. It is | ||
| 4 | implemented using remote-pcihost object. It is an extension of the PCI | ||
| 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 | ||
| 14 | |||
| 15 | [Added PCI_EXPRESS condition in hw/remote/Kconfig since remote-pcihost | ||
| 16 | needs PCIe. This solves "make check" failure on s390x. Fix suggested by | ||
| 17 | Philippe Mathieu-Daudé <philmd@redhat.com> and Thomas Huth | ||
| 18 | <thuth@redhat.com>. | ||
| 19 | --Stefan] | ||
| 20 | |||
| 21 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 22 | --- | ||
| 23 | MAINTAINERS | 2 + | ||
| 24 | include/hw/pci-host/remote.h | 29 ++++++++++++++ | ||
| 25 | hw/pci-host/remote.c | 75 ++++++++++++++++++++++++++++++++++++ | ||
| 26 | hw/pci-host/Kconfig | 3 ++ | ||
| 27 | hw/pci-host/meson.build | 1 + | ||
| 28 | hw/remote/Kconfig | 3 +- | ||
| 29 | 6 files changed, 112 insertions(+), 1 deletion(-) | ||
| 30 | create mode 100644 include/hw/pci-host/remote.h | ||
| 31 | create mode 100644 hw/pci-host/remote.c | ||
| 32 | |||
| 33 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 34 | index XXXXXXX..XXXXXXX 100644 | ||
| 35 | --- a/MAINTAINERS | ||
| 36 | +++ b/MAINTAINERS | ||
| 37 | @@ -XXX,XX +XXX,XX @@ M: John G Johnson <john.g.johnson@oracle.com> | ||
| 38 | S: Maintained | ||
| 39 | F: docs/devel/multi-process.rst | ||
| 40 | F: docs/system/multi-process.rst | ||
| 41 | +F: hw/pci-host/remote.c | ||
| 42 | +F: include/hw/pci-host/remote.h | ||
| 43 | |||
| 44 | Build and test automation | ||
| 45 | ------------------------- | ||
| 46 | diff --git a/include/hw/pci-host/remote.h b/include/hw/pci-host/remote.h | ||
| 47 | new file mode 100644 | ||
| 48 | index XXXXXXX..XXXXXXX | ||
| 49 | --- /dev/null | ||
| 50 | +++ b/include/hw/pci-host/remote.h | ||
| 51 | @@ -XXX,XX +XXX,XX @@ | ||
| 52 | +/* | ||
| 53 | + * PCI Host for remote device | ||
| 54 | + * | ||
| 55 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 56 | + * | ||
| 57 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
| 58 | + * See the COPYING file in the top-level directory. | ||
| 59 | + * | ||
| 60 | + */ | ||
| 61 | + | ||
| 62 | +#ifndef REMOTE_PCIHOST_H | ||
| 63 | +#define REMOTE_PCIHOST_H | ||
| 64 | + | ||
| 65 | +#include "exec/memory.h" | ||
| 66 | +#include "hw/pci/pcie_host.h" | ||
| 67 | + | ||
| 68 | +#define TYPE_REMOTE_PCIHOST "remote-pcihost" | ||
| 69 | +OBJECT_DECLARE_SIMPLE_TYPE(RemotePCIHost, REMOTE_PCIHOST) | ||
| 70 | + | ||
| 71 | +struct RemotePCIHost { | ||
| 72 | + /*< private >*/ | ||
| 73 | + PCIExpressHost parent_obj; | ||
| 74 | + /*< public >*/ | ||
| 75 | + | ||
| 76 | + MemoryRegion *mr_pci_mem; | ||
| 77 | + MemoryRegion *mr_sys_io; | ||
| 78 | +}; | ||
| 79 | + | ||
| 80 | +#endif | ||
| 81 | diff --git a/hw/pci-host/remote.c b/hw/pci-host/remote.c | ||
| 82 | new file mode 100644 | ||
| 83 | index XXXXXXX..XXXXXXX | ||
| 84 | --- /dev/null | ||
| 85 | +++ b/hw/pci-host/remote.c | ||
| 86 | @@ -XXX,XX +XXX,XX @@ | ||
| 87 | +/* | ||
| 88 | + * Remote PCI host device | ||
| 89 | + * | ||
| 90 | + * Unlike PCI host devices that model physical hardware, the purpose | ||
| 91 | + * of this PCI host is to host multi-process QEMU devices. | ||
| 92 | + * | ||
| 93 | + * Multi-process QEMU extends the PCI host of a QEMU machine into a | ||
| 94 | + * remote process. Any PCI device attached to the remote process is | ||
| 95 | + * visible in the QEMU guest. This allows existing QEMU device models | ||
| 96 | + * to be reused in the remote process. | ||
| 97 | + * | ||
| 98 | + * This PCI host is purely a container for PCI devices. It's fake in the | ||
| 99 | + * sense that the guest never sees this PCI host and has no way of | ||
| 100 | + * accessing it. Its job is just to provide the environment that QEMU | ||
| 101 | + * PCI device models need when running in a remote process. | ||
| 102 | + * | ||
| 103 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 104 | + * | ||
| 105 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
| 106 | + * See the COPYING file in the top-level directory. | ||
| 107 | + * | ||
| 108 | + */ | ||
| 109 | + | ||
| 110 | +#include "qemu/osdep.h" | ||
| 111 | +#include "qemu-common.h" | ||
| 112 | + | ||
| 113 | +#include "hw/pci/pci.h" | ||
| 114 | +#include "hw/pci/pci_host.h" | ||
| 115 | +#include "hw/pci/pcie_host.h" | ||
| 116 | +#include "hw/qdev-properties.h" | ||
| 117 | +#include "hw/pci-host/remote.h" | ||
| 118 | +#include "exec/memory.h" | ||
| 119 | + | ||
| 120 | +static const char *remote_pcihost_root_bus_path(PCIHostState *host_bridge, | ||
| 121 | + PCIBus *rootbus) | ||
| 122 | +{ | ||
| 123 | + return "0000:00"; | ||
| 124 | +} | ||
| 125 | + | ||
| 126 | +static void remote_pcihost_realize(DeviceState *dev, Error **errp) | ||
| 127 | +{ | ||
| 128 | + PCIHostState *pci = PCI_HOST_BRIDGE(dev); | ||
| 129 | + RemotePCIHost *s = REMOTE_PCIHOST(dev); | ||
| 130 | + | ||
| 131 | + pci->bus = pci_root_bus_new(DEVICE(s), "remote-pci", | ||
| 132 | + s->mr_pci_mem, s->mr_sys_io, | ||
| 133 | + 0, TYPE_PCIE_BUS); | ||
| 134 | +} | ||
| 135 | + | ||
| 136 | +static void remote_pcihost_class_init(ObjectClass *klass, void *data) | ||
| 137 | +{ | ||
| 138 | + DeviceClass *dc = DEVICE_CLASS(klass); | ||
| 139 | + PCIHostBridgeClass *hc = PCI_HOST_BRIDGE_CLASS(klass); | ||
| 140 | + | ||
| 141 | + hc->root_bus_path = remote_pcihost_root_bus_path; | ||
| 142 | + dc->realize = remote_pcihost_realize; | ||
| 143 | + | ||
| 144 | + dc->user_creatable = false; | ||
| 145 | + set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); | ||
| 146 | + dc->fw_name = "pci"; | ||
| 147 | +} | ||
| 148 | + | ||
| 149 | +static const TypeInfo remote_pcihost_info = { | ||
| 150 | + .name = TYPE_REMOTE_PCIHOST, | ||
| 151 | + .parent = TYPE_PCIE_HOST_BRIDGE, | ||
| 152 | + .instance_size = sizeof(RemotePCIHost), | ||
| 153 | + .class_init = remote_pcihost_class_init, | ||
| 154 | +}; | ||
| 155 | + | ||
| 156 | +static void remote_pcihost_register(void) | ||
| 157 | +{ | ||
| 158 | + type_register_static(&remote_pcihost_info); | ||
| 159 | +} | ||
| 160 | + | ||
| 161 | +type_init(remote_pcihost_register) | ||
| 162 | diff --git a/hw/pci-host/Kconfig b/hw/pci-host/Kconfig | ||
| 163 | index XXXXXXX..XXXXXXX 100644 | ||
| 164 | --- a/hw/pci-host/Kconfig | ||
| 165 | +++ b/hw/pci-host/Kconfig | ||
| 166 | @@ -XXX,XX +XXX,XX @@ config PCI_POWERNV | ||
| 167 | select PCI_EXPRESS | ||
| 168 | select MSI_NONBROKEN | ||
| 169 | select PCIE_PORT | ||
| 170 | + | ||
| 171 | +config REMOTE_PCIHOST | ||
| 172 | + bool | ||
| 173 | diff --git a/hw/pci-host/meson.build b/hw/pci-host/meson.build | ||
| 174 | index XXXXXXX..XXXXXXX 100644 | ||
| 175 | --- a/hw/pci-host/meson.build | ||
| 176 | +++ b/hw/pci-host/meson.build | ||
| 177 | @@ -XXX,XX +XXX,XX @@ pci_ss.add(when: 'CONFIG_PCI_EXPRESS_XILINX', if_true: files('xilinx-pcie.c')) | ||
| 178 | pci_ss.add(when: 'CONFIG_PCI_I440FX', if_true: files('i440fx.c')) | ||
| 179 | pci_ss.add(when: 'CONFIG_PCI_SABRE', if_true: files('sabre.c')) | ||
| 180 | pci_ss.add(when: 'CONFIG_XEN_IGD_PASSTHROUGH', if_true: files('xen_igd_pt.c')) | ||
| 181 | +pci_ss.add(when: 'CONFIG_REMOTE_PCIHOST', if_true: files('remote.c')) | ||
| 182 | |||
| 183 | # PPC devices | ||
| 184 | pci_ss.add(when: 'CONFIG_PREP_PCI', if_true: files('prep.c')) | ||
| 185 | diff --git a/hw/remote/Kconfig b/hw/remote/Kconfig | ||
| 186 | index XXXXXXX..XXXXXXX 100644 | ||
| 187 | --- a/hw/remote/Kconfig | ||
| 188 | +++ b/hw/remote/Kconfig | ||
| 189 | @@ -XXX,XX +XXX,XX @@ | ||
| 190 | config MULTIPROCESS | ||
| 191 | bool | ||
| 192 | - depends on PCI && KVM | ||
| 193 | + depends on PCI && PCI_EXPRESS && KVM | ||
| 194 | + select REMOTE_PCIHOST | ||
| 195 | -- | ||
| 196 | 2.29.2 | ||
| 197 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
| 1 | 2 | ||
| 3 | x-remote-machine object sets up various subsystems of the remote | ||
| 4 | device process. Instantiate PCI host bridge object and initialize RAM, IO & | ||
| 5 | PCI memory regions. | ||
| 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 | ||
| 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 13 | --- | ||
| 14 | MAINTAINERS | 2 ++ | ||
| 15 | include/hw/pci-host/remote.h | 1 + | ||
| 16 | include/hw/remote/machine.h | 27 ++++++++++++++ | ||
| 17 | hw/remote/machine.c | 70 ++++++++++++++++++++++++++++++++++++ | ||
| 18 | hw/meson.build | 1 + | ||
| 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 | ||
| 24 | |||
| 25 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 26 | index XXXXXXX..XXXXXXX 100644 | ||
| 27 | --- a/MAINTAINERS | ||
| 28 | +++ b/MAINTAINERS | ||
| 29 | @@ -XXX,XX +XXX,XX @@ F: docs/devel/multi-process.rst | ||
| 30 | F: docs/system/multi-process.rst | ||
| 31 | F: hw/pci-host/remote.c | ||
| 32 | F: include/hw/pci-host/remote.h | ||
| 33 | +F: hw/remote/machine.c | ||
| 34 | +F: include/hw/remote/machine.h | ||
| 35 | |||
| 36 | Build and test automation | ||
| 37 | ------------------------- | ||
| 38 | diff --git a/include/hw/pci-host/remote.h b/include/hw/pci-host/remote.h | ||
| 39 | index XXXXXXX..XXXXXXX 100644 | ||
| 40 | --- a/include/hw/pci-host/remote.h | ||
| 41 | +++ b/include/hw/pci-host/remote.h | ||
| 42 | @@ -XXX,XX +XXX,XX @@ struct RemotePCIHost { | ||
| 43 | |||
| 44 | MemoryRegion *mr_pci_mem; | ||
| 45 | MemoryRegion *mr_sys_io; | ||
| 46 | + MemoryRegion *mr_sys_mem; | ||
| 47 | }; | ||
| 48 | |||
| 49 | #endif | ||
| 50 | diff --git a/include/hw/remote/machine.h b/include/hw/remote/machine.h | ||
| 51 | new file mode 100644 | ||
| 52 | index XXXXXXX..XXXXXXX | ||
| 53 | --- /dev/null | ||
| 54 | +++ b/include/hw/remote/machine.h | ||
| 55 | @@ -XXX,XX +XXX,XX @@ | ||
| 56 | +/* | ||
| 57 | + * Remote machine configuration | ||
| 58 | + * | ||
| 59 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 60 | + * | ||
| 61 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
| 62 | + * See the COPYING file in the top-level directory. | ||
| 63 | + * | ||
| 64 | + */ | ||
| 65 | + | ||
| 66 | +#ifndef REMOTE_MACHINE_H | ||
| 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 | ||
| 160 | index XXXXXXX..XXXXXXX 100644 | ||
| 161 | --- a/hw/meson.build | ||
| 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 | ||
| 176 | @@ -XXX,XX +XXX,XX @@ | ||
| 177 | +remote_ss = ss.source_set() | ||
| 178 | + | ||
| 179 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) | ||
| 180 | + | ||
| 181 | +softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
| 182 | -- | ||
| 183 | 2.29.2 | ||
| 184 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 1 | 2 | ||
| 3 | Adds qio_channel_writev_full_all() to transmit both data and FDs. | ||
| 4 | Refactors existing code to use this helper. | ||
| 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 | Acked-by: Daniel P. Berrangé <berrange@redhat.com> | ||
| 11 | Message-id: 480fbf1fe4152495d60596c9b665124549b426a5.1611938319.git.jag.raman@oracle.com | ||
| 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 13 | --- | ||
| 14 | include/io/channel.h | 25 +++++++++++++++++++++++++ | ||
| 15 | io/channel.c | 15 ++++++++++++++- | ||
| 16 | 2 files changed, 39 insertions(+), 1 deletion(-) | ||
| 17 | |||
| 18 | diff --git a/include/io/channel.h b/include/io/channel.h | ||
| 19 | index XXXXXXX..XXXXXXX 100644 | ||
| 20 | --- a/include/io/channel.h | ||
| 21 | +++ b/include/io/channel.h | ||
| 22 | @@ -XXX,XX +XXX,XX @@ void qio_channel_set_aio_fd_handler(QIOChannel *ioc, | ||
| 23 | IOHandler *io_write, | ||
| 24 | void *opaque); | ||
| 25 | |||
| 26 | +/** | ||
| 27 | + * qio_channel_writev_full_all: | ||
| 28 | + * @ioc: the channel object | ||
| 29 | + * @iov: the array of memory regions to write data from | ||
| 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 | + | ||
| 45 | +int qio_channel_writev_full_all(QIOChannel *ioc, | ||
| 46 | + const struct iovec *iov, | ||
| 47 | + size_t niov, | ||
| 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; | ||
| 89 | } | ||
| 90 | |||
| 91 | ret = 0; | ||
| 92 | -- | ||
| 93 | 2.29.2 | ||
| 94 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | |
| 2 | |||
| 3 | Adds qio_channel_readv_full_all_eof() and qio_channel_readv_full_all() | ||
| 4 | to read both data and FDs. Refactors existing code to use these helpers. | ||
| 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 | Acked-by: Daniel P. Berrangé <berrange@redhat.com> | ||
| 10 | Message-id: b059c4cc0fb741e794d644c144cc21372cad877d.1611938319.git.jag.raman@oracle.com | ||
| 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 12 | --- | ||
| 13 | include/io/channel.h | 53 +++++++++++++++++++++++ | ||
| 14 | io/channel.c | 101 ++++++++++++++++++++++++++++++++++--------- | ||
| 15 | 2 files changed, 134 insertions(+), 20 deletions(-) | ||
| 16 | |||
| 17 | diff --git a/include/io/channel.h b/include/io/channel.h | ||
| 18 | index XXXXXXX..XXXXXXX 100644 | ||
| 19 | --- a/include/io/channel.h | ||
| 20 | +++ b/include/io/channel.h | ||
| 21 | @@ -XXX,XX +XXX,XX @@ void qio_channel_set_aio_fd_handler(QIOChannel *ioc, | ||
| 22 | IOHandler *io_write, | ||
| 23 | void *opaque); | ||
| 24 | |||
| 25 | +/** | ||
| 26 | + * qio_channel_readv_full_all_eof: | ||
| 27 | + * @ioc: the channel object | ||
| 28 | + * @iov: the array of memory regions to read data to | ||
| 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; | ||
| 192 | } | ||
| 193 | |||
| 194 | ret = 1; | ||
| 195 | @@ -XXX,XX +XXX,XX @@ int qio_channel_readv_all_eof(QIOChannel *ioc, | ||
| 196 | return ret; | ||
| 197 | } | ||
| 198 | |||
| 199 | -int qio_channel_readv_all(QIOChannel *ioc, | ||
| 200 | - const struct iovec *iov, | ||
| 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; | ||
| 227 | } | ||
| 228 | |||
| 229 | -- | ||
| 230 | 2.29.2 | ||
| 231 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | |
| 2 | |||
| 3 | Defines MPQemuMsg, which is the message that is sent to the remote | ||
| 4 | process. This message is sent over QIOChannel and is used to | ||
| 5 | command the remote process to perform various tasks. | ||
| 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 | ||
| 13 | |||
| 14 | [Replace struct iovec send[2] = {0} with {} to make clang happy as | ||
| 15 | suggested by Peter Maydell <peter.maydell@linaro.org>. | ||
| 16 | --Stefan] | ||
| 17 | |||
| 18 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 19 | --- | ||
| 20 | MAINTAINERS | 2 + | ||
| 21 | meson.build | 1 + | ||
| 22 | hw/remote/trace.h | 1 + | ||
| 23 | include/hw/remote/mpqemu-link.h | 63 ++++++++++ | ||
| 24 | include/sysemu/iothread.h | 6 + | ||
| 25 | hw/remote/mpqemu-link.c | 205 ++++++++++++++++++++++++++++++++ | ||
| 26 | iothread.c | 6 + | ||
| 27 | hw/remote/meson.build | 1 + | ||
| 28 | hw/remote/trace-events | 4 + | ||
| 29 | 9 files changed, 289 insertions(+) | ||
| 30 | create mode 100644 hw/remote/trace.h | ||
| 31 | create mode 100644 include/hw/remote/mpqemu-link.h | ||
| 32 | create mode 100644 hw/remote/mpqemu-link.c | ||
| 33 | create mode 100644 hw/remote/trace-events | ||
| 34 | |||
| 35 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 36 | index XXXXXXX..XXXXXXX 100644 | ||
| 37 | --- a/MAINTAINERS | ||
| 38 | +++ b/MAINTAINERS | ||
| 39 | @@ -XXX,XX +XXX,XX @@ F: hw/pci-host/remote.c | ||
| 40 | F: include/hw/pci-host/remote.h | ||
| 41 | F: hw/remote/machine.c | ||
| 42 | F: include/hw/remote/machine.h | ||
| 43 | +F: hw/remote/mpqemu-link.c | ||
| 44 | +F: include/hw/remote/mpqemu-link.h | ||
| 45 | |||
| 46 | Build and test automation | ||
| 47 | ------------------------- | ||
| 48 | diff --git a/meson.build b/meson.build | ||
| 49 | index XXXXXXX..XXXXXXX 100644 | ||
| 50 | --- a/meson.build | ||
| 51 | +++ b/meson.build | ||
| 52 | @@ -XXX,XX +XXX,XX @@ if have_system | ||
| 53 | 'net', | ||
| 54 | 'softmmu', | ||
| 55 | 'ui', | ||
| 56 | + 'hw/remote', | ||
| 57 | ] | ||
| 58 | endif | ||
| 59 | if have_system or have_user | ||
| 60 | diff --git a/hw/remote/trace.h b/hw/remote/trace.h | ||
| 61 | new file mode 100644 | ||
| 62 | index XXXXXXX..XXXXXXX | ||
| 63 | --- /dev/null | ||
| 64 | +++ b/hw/remote/trace.h | ||
| 65 | @@ -0,0 +1 @@ | ||
| 66 | +#include "trace/trace-hw_remote.h" | ||
| 67 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
| 68 | new file mode 100644 | ||
| 69 | index XXXXXXX..XXXXXXX | ||
| 70 | --- /dev/null | ||
| 71 | +++ b/include/hw/remote/mpqemu-link.h | ||
| 72 | @@ -XXX,XX +XXX,XX @@ | ||
| 73 | +/* | ||
| 74 | + * Communication channel between QEMU and remote device process | ||
| 75 | + * | ||
| 76 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 77 | + * | ||
| 78 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
| 79 | + * See the COPYING file in the top-level directory. | ||
| 80 | + * | ||
| 81 | + */ | ||
| 82 | + | ||
| 83 | +#ifndef MPQEMU_LINK_H | ||
| 84 | +#define MPQEMU_LINK_H | ||
| 85 | + | ||
| 86 | +#include "qom/object.h" | ||
| 87 | +#include "qemu/thread.h" | ||
| 88 | +#include "io/channel.h" | ||
| 89 | + | ||
| 90 | +#define REMOTE_MAX_FDS 8 | ||
| 91 | + | ||
| 92 | +#define MPQEMU_MSG_HDR_SIZE offsetof(MPQemuMsg, data.u64) | ||
| 93 | + | ||
| 94 | +/** | ||
| 95 | + * MPQemuCmd: | ||
| 96 | + * | ||
| 97 | + * MPQemuCmd enum type to specify the command to be executed on the remote | ||
| 98 | + * device. | ||
| 99 | + * | ||
| 100 | + * This uses a private protocol between QEMU and the remote process. vfio-user | ||
| 101 | + * protocol would supersede this in the future. | ||
| 102 | + * | ||
| 103 | + */ | ||
| 104 | +typedef enum { | ||
| 105 | + MPQEMU_CMD_MAX, | ||
| 106 | +} MPQemuCmd; | ||
| 107 | + | ||
| 108 | +/** | ||
| 109 | + * MPQemuMsg: | ||
| 110 | + * @cmd: The remote command | ||
| 111 | + * @size: Size of the data to be shared | ||
| 112 | + * @data: Structured data | ||
| 113 | + * @fds: File descriptors to be shared with remote device | ||
| 114 | + * | ||
| 115 | + * MPQemuMsg Format of the message sent to the remote device from QEMU. | ||
| 116 | + * | ||
| 117 | + */ | ||
| 118 | +typedef struct { | ||
| 119 | + int cmd; | ||
| 120 | + size_t size; | ||
| 121 | + | ||
| 122 | + union { | ||
| 123 | + uint64_t u64; | ||
| 124 | + } data; | ||
| 125 | + | ||
| 126 | + int fds[REMOTE_MAX_FDS]; | ||
| 127 | + int num_fds; | ||
| 128 | +} MPQemuMsg; | ||
| 129 | + | ||
| 130 | +bool mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); | ||
| 131 | +bool mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); | ||
| 132 | + | ||
| 133 | +bool mpqemu_msg_valid(MPQemuMsg *msg); | ||
| 134 | + | ||
| 135 | +#endif | ||
| 136 | diff --git a/include/sysemu/iothread.h b/include/sysemu/iothread.h | ||
| 137 | index XXXXXXX..XXXXXXX 100644 | ||
| 138 | --- a/include/sysemu/iothread.h | ||
| 139 | +++ b/include/sysemu/iothread.h | ||
| 140 | @@ -XXX,XX +XXX,XX @@ IOThread *iothread_create(const char *id, Error **errp); | ||
| 141 | void iothread_stop(IOThread *iothread); | ||
| 142 | void iothread_destroy(IOThread *iothread); | ||
| 143 | |||
| 144 | +/* | ||
| 145 | + * Returns true if executing withing IOThread context, | ||
| 146 | + * false otherwise. | ||
| 147 | + */ | ||
| 148 | +bool qemu_in_iothread(void); | ||
| 149 | + | ||
| 150 | #endif /* IOTHREAD_H */ | ||
| 151 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
| 152 | new file mode 100644 | ||
| 153 | index XXXXXXX..XXXXXXX | ||
| 154 | --- /dev/null | ||
| 155 | +++ b/hw/remote/mpqemu-link.c | ||
| 156 | @@ -XXX,XX +XXX,XX @@ | ||
| 157 | +/* | ||
| 158 | + * Communication channel between QEMU and remote device process | ||
| 159 | + * | ||
| 160 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 161 | + * | ||
| 162 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
| 163 | + * See the COPYING file in the top-level directory. | ||
| 164 | + * | ||
| 165 | + */ | ||
| 166 | + | ||
| 167 | +#include "qemu/osdep.h" | ||
| 168 | +#include "qemu-common.h" | ||
| 169 | + | ||
| 170 | +#include "qemu/module.h" | ||
| 171 | +#include "hw/remote/mpqemu-link.h" | ||
| 172 | +#include "qapi/error.h" | ||
| 173 | +#include "qemu/iov.h" | ||
| 174 | +#include "qemu/error-report.h" | ||
| 175 | +#include "qemu/main-loop.h" | ||
| 176 | +#include "io/channel.h" | ||
| 177 | +#include "sysemu/iothread.h" | ||
| 178 | +#include "trace.h" | ||
| 179 | + | ||
| 180 | +/* | ||
| 181 | + * Send message over the ioc QIOChannel. | ||
| 182 | + * This function is safe to call from: | ||
| 183 | + * - main loop in co-routine context. Will block the main loop if not in | ||
| 184 | + * co-routine context; | ||
| 185 | + * - vCPU thread with no co-routine context and if the channel is not part | ||
| 186 | + * of the main loop handling; | ||
| 187 | + * - IOThread within co-routine context, outside of co-routine context | ||
| 188 | + * will block IOThread; | ||
| 189 | + * Returns true if no errors were encountered, false otherwise. | ||
| 190 | + */ | ||
| 191 | +bool mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp) | ||
| 192 | +{ | ||
| 193 | + ERRP_GUARD(); | ||
| 194 | + bool iolock = qemu_mutex_iothread_locked(); | ||
| 195 | + bool iothread = qemu_in_iothread(); | ||
| 196 | + struct iovec send[2] = {}; | ||
| 197 | + int *fds = NULL; | ||
| 198 | + size_t nfds = 0; | ||
| 199 | + bool ret = false; | ||
| 200 | + | ||
| 201 | + send[0].iov_base = msg; | ||
| 202 | + send[0].iov_len = MPQEMU_MSG_HDR_SIZE; | ||
| 203 | + | ||
| 204 | + send[1].iov_base = (void *)&msg->data; | ||
| 205 | + send[1].iov_len = msg->size; | ||
| 206 | + | ||
| 207 | + if (msg->num_fds) { | ||
| 208 | + nfds = msg->num_fds; | ||
| 209 | + fds = msg->fds; | ||
| 210 | + } | ||
| 211 | + | ||
| 212 | + /* | ||
| 213 | + * Dont use in IOThread out of co-routine context as | ||
| 214 | + * it will block IOThread. | ||
| 215 | + */ | ||
| 216 | + assert(qemu_in_coroutine() || !iothread); | ||
| 217 | + | ||
| 218 | + /* | ||
| 219 | + * Skip unlocking/locking iothread lock when the IOThread is running | ||
| 220 | + * in co-routine context. Co-routine context is asserted above | ||
| 221 | + * for IOThread case. | ||
| 222 | + * Also skip lock handling while in a co-routine in the main context. | ||
| 223 | + */ | ||
| 224 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
| 225 | + qemu_mutex_unlock_iothread(); | ||
| 226 | + } | ||
| 227 | + | ||
| 228 | + if (!qio_channel_writev_full_all(ioc, send, G_N_ELEMENTS(send), | ||
| 229 | + fds, nfds, errp)) { | ||
| 230 | + ret = true; | ||
| 231 | + } else { | ||
| 232 | + trace_mpqemu_send_io_error(msg->cmd, msg->size, nfds); | ||
| 233 | + } | ||
| 234 | + | ||
| 235 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
| 236 | + /* See above comment why skip locking here. */ | ||
| 237 | + qemu_mutex_lock_iothread(); | ||
| 238 | + } | ||
| 239 | + | ||
| 240 | + return ret; | ||
| 241 | +} | ||
| 242 | + | ||
| 243 | +/* | ||
| 244 | + * Read message from the ioc QIOChannel. | ||
| 245 | + * This function is safe to call from: | ||
| 246 | + * - From main loop in co-routine context. Will block the main loop if not in | ||
| 247 | + * co-routine context; | ||
| 248 | + * - From vCPU thread with no co-routine context and if the channel is not part | ||
| 249 | + * of the main loop handling; | ||
| 250 | + * - From IOThread within co-routine context, outside of co-routine context | ||
| 251 | + * will block IOThread; | ||
| 252 | + */ | ||
| 253 | +static ssize_t mpqemu_read(QIOChannel *ioc, void *buf, size_t len, int **fds, | ||
| 254 | + size_t *nfds, Error **errp) | ||
| 255 | +{ | ||
| 256 | + ERRP_GUARD(); | ||
| 257 | + struct iovec iov = { .iov_base = buf, .iov_len = len }; | ||
| 258 | + bool iolock = qemu_mutex_iothread_locked(); | ||
| 259 | + bool iothread = qemu_in_iothread(); | ||
| 260 | + int ret = -1; | ||
| 261 | + | ||
| 262 | + /* | ||
| 263 | + * Dont use in IOThread out of co-routine context as | ||
| 264 | + * it will block IOThread. | ||
| 265 | + */ | ||
| 266 | + assert(qemu_in_coroutine() || !iothread); | ||
| 267 | + | ||
| 268 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
| 269 | + qemu_mutex_unlock_iothread(); | ||
| 270 | + } | ||
| 271 | + | ||
| 272 | + ret = qio_channel_readv_full_all_eof(ioc, &iov, 1, fds, nfds, errp); | ||
| 273 | + | ||
| 274 | + if (iolock && !iothread && !qemu_in_coroutine()) { | ||
| 275 | + qemu_mutex_lock_iothread(); | ||
| 276 | + } | ||
| 277 | + | ||
| 278 | + return (ret <= 0) ? ret : iov.iov_len; | ||
| 279 | +} | ||
| 280 | + | ||
| 281 | +bool mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp) | ||
| 282 | +{ | ||
| 283 | + ERRP_GUARD(); | ||
| 284 | + g_autofree int *fds = NULL; | ||
| 285 | + size_t nfds = 0; | ||
| 286 | + ssize_t len; | ||
| 287 | + bool ret = false; | ||
| 288 | + | ||
| 289 | + len = mpqemu_read(ioc, msg, MPQEMU_MSG_HDR_SIZE, &fds, &nfds, errp); | ||
| 290 | + if (len <= 0) { | ||
| 291 | + goto fail; | ||
| 292 | + } else if (len != MPQEMU_MSG_HDR_SIZE) { | ||
| 293 | + error_setg(errp, "Message header corrupted"); | ||
| 294 | + goto fail; | ||
| 295 | + } | ||
| 296 | + | ||
| 297 | + if (msg->size > sizeof(msg->data)) { | ||
| 298 | + error_setg(errp, "Invalid size for message"); | ||
| 299 | + goto fail; | ||
| 300 | + } | ||
| 301 | + | ||
| 302 | + if (!msg->size) { | ||
| 303 | + goto copy_fds; | ||
| 304 | + } | ||
| 305 | + | ||
| 306 | + len = mpqemu_read(ioc, &msg->data, msg->size, NULL, NULL, errp); | ||
| 307 | + if (len <= 0) { | ||
| 308 | + goto fail; | ||
| 309 | + } | ||
| 310 | + if (len != msg->size) { | ||
| 311 | + error_setg(errp, "Unable to read full message"); | ||
| 312 | + goto fail; | ||
| 313 | + } | ||
| 314 | + | ||
| 315 | +copy_fds: | ||
| 316 | + msg->num_fds = nfds; | ||
| 317 | + if (nfds > G_N_ELEMENTS(msg->fds)) { | ||
| 318 | + error_setg(errp, | ||
| 319 | + "Overflow error: received %zu fds, more than max of %d fds", | ||
| 320 | + nfds, REMOTE_MAX_FDS); | ||
| 321 | + goto fail; | ||
| 322 | + } | ||
| 323 | + if (nfds) { | ||
| 324 | + memcpy(msg->fds, fds, nfds * sizeof(int)); | ||
| 325 | + } | ||
| 326 | + | ||
| 327 | + ret = true; | ||
| 328 | + | ||
| 329 | +fail: | ||
| 330 | + if (*errp) { | ||
| 331 | + trace_mpqemu_recv_io_error(msg->cmd, msg->size, nfds); | ||
| 332 | + } | ||
| 333 | + while (*errp && nfds) { | ||
| 334 | + close(fds[nfds - 1]); | ||
| 335 | + nfds--; | ||
| 336 | + } | ||
| 337 | + | ||
| 338 | + return ret; | ||
| 339 | +} | ||
| 340 | + | ||
| 341 | +bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
| 342 | +{ | ||
| 343 | + if (msg->cmd >= MPQEMU_CMD_MAX && msg->cmd < 0) { | ||
| 344 | + return false; | ||
| 345 | + } | ||
| 346 | + | ||
| 347 | + /* Verify FDs. */ | ||
| 348 | + if (msg->num_fds >= REMOTE_MAX_FDS) { | ||
| 349 | + return false; | ||
| 350 | + } | ||
| 351 | + | ||
| 352 | + if (msg->num_fds > 0) { | ||
| 353 | + for (int i = 0; i < msg->num_fds; i++) { | ||
| 354 | + if (fcntl(msg->fds[i], F_GETFL) == -1) { | ||
| 355 | + return false; | ||
| 356 | + } | ||
| 357 | + } | ||
| 358 | + } | ||
| 359 | + | ||
| 360 | + return true; | ||
| 361 | +} | ||
| 362 | diff --git a/iothread.c b/iothread.c | ||
| 363 | index XXXXXXX..XXXXXXX 100644 | ||
| 364 | --- a/iothread.c | ||
| 365 | +++ b/iothread.c | ||
| 366 | @@ -XXX,XX +XXX,XX @@ IOThread *iothread_by_id(const char *id) | ||
| 367 | { | ||
| 368 | return IOTHREAD(object_resolve_path_type(id, TYPE_IOTHREAD, NULL)); | ||
| 369 | } | ||
| 370 | + | ||
| 371 | +bool qemu_in_iothread(void) | ||
| 372 | +{ | ||
| 373 | + return qemu_get_current_aio_context() == qemu_get_aio_context() ? | ||
| 374 | + false : true; | ||
| 375 | +} | ||
| 376 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
| 377 | index XXXXXXX..XXXXXXX 100644 | ||
| 378 | --- a/hw/remote/meson.build | ||
| 379 | +++ b/hw/remote/meson.build | ||
| 380 | @@ -XXX,XX +XXX,XX @@ | ||
| 381 | remote_ss = ss.source_set() | ||
| 382 | |||
| 383 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) | ||
| 384 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
| 385 | |||
| 386 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
| 387 | diff --git a/hw/remote/trace-events b/hw/remote/trace-events | ||
| 388 | new file mode 100644 | ||
| 389 | index XXXXXXX..XXXXXXX | ||
| 390 | --- /dev/null | ||
| 391 | +++ b/hw/remote/trace-events | ||
| 392 | @@ -XXX,XX +XXX,XX @@ | ||
| 393 | +# multi-process trace events | ||
| 394 | + | ||
| 395 | +mpqemu_send_io_error(int cmd, int size, int nfds) "send command %d size %d, %d file descriptors to remote process" | ||
| 396 | +mpqemu_recv_io_error(int cmd, int size, int nfds) "failed to receive %d size %d, %d file descriptors to remote process" | ||
| 397 | -- | ||
| 398 | 2.29.2 | ||
| 399 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
| 1 | 2 | ||
| 3 | Initializes the message handler function in the remote process. It is | ||
| 4 | called whenever there's an event pending on QIOChannel that registers | ||
| 5 | this function. | ||
| 6 | |||
| 7 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
| 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 11 | Message-id: 99d38d8b93753a6409ac2340e858858cda59ab1b.1611938319.git.jag.raman@oracle.com | ||
| 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 13 | --- | ||
| 14 | MAINTAINERS | 1 + | ||
| 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 | ||
| 20 | |||
| 21 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 22 | index XXXXXXX..XXXXXXX 100644 | ||
| 23 | --- a/MAINTAINERS | ||
| 24 | +++ b/MAINTAINERS | ||
| 25 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/machine.c | ||
| 26 | F: include/hw/remote/machine.h | ||
| 27 | F: hw/remote/mpqemu-link.c | ||
| 28 | F: include/hw/remote/mpqemu-link.h | ||
| 29 | +F: hw/remote/message.c | ||
| 30 | |||
| 31 | Build and test automation | ||
| 32 | ------------------------- | ||
| 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 | + } | ||
| 115 | + } | ||
| 116 | + | ||
| 117 | + if (local_err) { | ||
| 118 | + error_report_err(local_err); | ||
| 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) | ||
| 135 | -- | ||
| 136 | 2.29.2 | ||
| 137 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Jagannathan Raman <jag.raman@oracle.com> | |
| 2 | |||
| 3 | Associate the file descriptor for a PCIDevice in remote process with | ||
| 4 | DeviceState object. | ||
| 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: f405a2ed5d7518b87bea7c59cfdf334d67e5ee51.1611938319.git.jag.raman@oracle.com | ||
| 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 12 | --- | ||
| 13 | MAINTAINERS | 1 + | ||
| 14 | hw/remote/remote-obj.c | 203 +++++++++++++++++++++++++++++++++++++++++ | ||
| 15 | hw/remote/meson.build | 1 + | ||
| 16 | 3 files changed, 205 insertions(+) | ||
| 17 | create mode 100644 hw/remote/remote-obj.c | ||
| 18 | |||
| 19 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 20 | index XXXXXXX..XXXXXXX 100644 | ||
| 21 | --- a/MAINTAINERS | ||
| 22 | +++ b/MAINTAINERS | ||
| 23 | @@ -XXX,XX +XXX,XX @@ F: include/hw/remote/machine.h | ||
| 24 | F: hw/remote/mpqemu-link.c | ||
| 25 | F: include/hw/remote/mpqemu-link.h | ||
| 26 | F: hw/remote/message.c | ||
| 27 | +F: hw/remote/remote-obj.c | ||
| 28 | |||
| 29 | Build and test automation | ||
| 30 | ------------------------- | ||
| 31 | diff --git a/hw/remote/remote-obj.c b/hw/remote/remote-obj.c | ||
| 32 | new file mode 100644 | ||
| 33 | index XXXXXXX..XXXXXXX | ||
| 34 | --- /dev/null | ||
| 35 | +++ b/hw/remote/remote-obj.c | ||
| 36 | @@ -XXX,XX +XXX,XX @@ | ||
| 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 | ||
| 241 | index XXXXXXX..XXXXXXX 100644 | ||
| 242 | --- a/hw/remote/meson.build | ||
| 243 | +++ b/hw/remote/meson.build | ||
| 244 | @@ -XXX,XX +XXX,XX @@ remote_ss = ss.source_set() | ||
| 245 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('machine.c')) | ||
| 246 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
| 247 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) | ||
| 248 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) | ||
| 249 | |||
| 250 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
| 251 | -- | ||
| 252 | 2.29.2 | ||
| 253 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Jagannathan Raman <jag.raman@oracle.com> | |
| 2 | |||
| 3 | SyncSysMemMsg message format is defined. It is used to send | ||
| 4 | file descriptors of the RAM regions to remote device. | ||
| 5 | RAM on the remote device is configured with a set of file descriptors. | ||
| 6 | Old RAM regions are deleted and new regions, each with an fd, is | ||
| 7 | added to the RAM. | ||
| 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: 7d2d1831d812e85f681e7a8ab99e032cf4704689.1611938319.git.jag.raman@oracle.com | ||
| 14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 15 | --- | ||
| 16 | MAINTAINERS | 2 + | ||
| 17 | include/hw/remote/memory.h | 19 ++++++++++ | ||
| 18 | include/hw/remote/mpqemu-link.h | 10 +++++ | ||
| 19 | hw/remote/memory.c | 65 +++++++++++++++++++++++++++++++++ | ||
| 20 | hw/remote/mpqemu-link.c | 11 ++++++ | ||
| 21 | hw/remote/meson.build | 2 + | ||
| 22 | 6 files changed, 109 insertions(+) | ||
| 23 | create mode 100644 include/hw/remote/memory.h | ||
| 24 | create mode 100644 hw/remote/memory.c | ||
| 25 | |||
| 26 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 27 | index XXXXXXX..XXXXXXX 100644 | ||
| 28 | --- a/MAINTAINERS | ||
| 29 | +++ b/MAINTAINERS | ||
| 30 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/mpqemu-link.c | ||
| 31 | F: include/hw/remote/mpqemu-link.h | ||
| 32 | F: hw/remote/message.c | ||
| 33 | F: hw/remote/remote-obj.c | ||
| 34 | +F: include/hw/remote/memory.h | ||
| 35 | +F: hw/remote/memory.c | ||
| 36 | |||
| 37 | Build and test automation | ||
| 38 | ------------------------- | ||
| 39 | diff --git a/include/hw/remote/memory.h b/include/hw/remote/memory.h | ||
| 40 | new file mode 100644 | ||
| 41 | index XXXXXXX..XXXXXXX | ||
| 42 | --- /dev/null | ||
| 43 | +++ b/include/hw/remote/memory.h | ||
| 44 | @@ -XXX,XX +XXX,XX @@ | ||
| 45 | +/* | ||
| 46 | + * Memory manager for remote device | ||
| 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 REMOTE_MEMORY_H | ||
| 56 | +#define REMOTE_MEMORY_H | ||
| 57 | + | ||
| 58 | +#include "exec/hwaddr.h" | ||
| 59 | +#include "hw/remote/mpqemu-link.h" | ||
| 60 | + | ||
| 61 | +void remote_sysmem_reconfig(MPQemuMsg *msg, Error **errp); | ||
| 62 | + | ||
| 63 | +#endif | ||
| 64 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
| 65 | index XXXXXXX..XXXXXXX 100644 | ||
| 66 | --- a/include/hw/remote/mpqemu-link.h | ||
| 67 | +++ b/include/hw/remote/mpqemu-link.h | ||
| 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) | ||
| 184 | } | ||
| 185 | } | ||
| 186 | |||
| 187 | + /* Verify message specific fields. */ | ||
| 188 | + switch (msg->cmd) { | ||
| 189 | + case MPQEMU_CMD_SYNC_SYSMEM: | ||
| 190 | + if (msg->num_fds == 0 || msg->size != sizeof(SyncSysmemMsg)) { | ||
| 191 | + return false; | ||
| 192 | + } | ||
| 193 | + break; | ||
| 194 | + default: | ||
| 195 | + break; | ||
| 196 | + } | ||
| 197 | + | ||
| 198 | return true; | ||
| 199 | } | ||
| 200 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
| 201 | index XXXXXXX..XXXXXXX 100644 | ||
| 202 | --- a/hw/remote/meson.build | ||
| 203 | +++ b/hw/remote/meson.build | ||
| 204 | @@ -XXX,XX +XXX,XX @@ remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
| 205 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) | ||
| 206 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) | ||
| 207 | |||
| 208 | +specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('memory.c')) | ||
| 209 | + | ||
| 210 | softmmu_ss.add_all(when: 'CONFIG_MULTIPROCESS', if_true: remote_ss) | ||
| 211 | -- | ||
| 212 | 2.29.2 | ||
| 213 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 1 | 2 | ||
| 3 | Defines a PCI Device proxy object as a child of TYPE_PCI_DEVICE. | ||
| 4 | |||
| 5 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 6 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 7 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
| 8 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 9 | Message-id: b5186ebfedf8e557044d09a768846c59230ad3a7.1611938319.git.jag.raman@oracle.com | ||
| 10 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 11 | --- | ||
| 12 | MAINTAINERS | 2 + | ||
| 13 | include/hw/remote/proxy.h | 33 +++++++++++++ | ||
| 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 | ||
| 19 | |||
| 20 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 21 | index XXXXXXX..XXXXXXX 100644 | ||
| 22 | --- a/MAINTAINERS | ||
| 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 | ||
| 38 | @@ -XXX,XX +XXX,XX @@ | ||
| 39 | +/* | ||
| 40 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 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 | + */ | ||
| 46 | + | ||
| 47 | +#ifndef PROXY_H | ||
| 48 | +#define PROXY_H | ||
| 49 | + | ||
| 50 | +#include "hw/pci/pci.h" | ||
| 51 | +#include "io/channel.h" | ||
| 52 | + | ||
| 53 | +#define TYPE_PCI_PROXY_DEV "x-pci-proxy-dev" | ||
| 54 | +OBJECT_DECLARE_SIMPLE_TYPE(PCIProxyDev, PCI_PROXY_DEV) | ||
| 55 | + | ||
| 56 | +struct PCIProxyDev { | ||
| 57 | + PCIDevice parent_dev; | ||
| 58 | + char *fd; | ||
| 59 | + | ||
| 60 | + /* | ||
| 61 | + * Mutex used to protect the QIOChannel fd from | ||
| 62 | + * the concurrent access by the VCPUs since proxy | ||
| 63 | + * blocks while awaiting for the replies from the | ||
| 64 | + * process remote. | ||
| 65 | + */ | ||
| 66 | + QemuMutex io_mutex; | ||
| 67 | + QIOChannel *ioc; | ||
| 68 | + Error *migration_blocker; | ||
| 69 | +}; | ||
| 70 | + | ||
| 71 | +#endif /* PROXY_H */ | ||
| 72 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
| 73 | new file mode 100644 | ||
| 74 | index XXXXXXX..XXXXXXX | ||
| 75 | --- /dev/null | ||
| 76 | +++ b/hw/remote/proxy.c | ||
| 77 | @@ -XXX,XX +XXX,XX @@ | ||
| 78 | +/* | ||
| 79 | + * Copyright © 2018, 2021 Oracle and/or its affiliates. | ||
| 80 | + * | ||
| 81 | + * This work is licensed under the terms of the GNU GPL, version 2 or later. | ||
| 82 | + * See the COPYING file in the top-level directory. | ||
| 83 | + * | ||
| 84 | + */ | ||
| 85 | + | ||
| 86 | +#include "qemu/osdep.h" | ||
| 87 | +#include "qemu-common.h" | ||
| 88 | + | ||
| 89 | +#include "hw/remote/proxy.h" | ||
| 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) | ||
| 99 | +{ | ||
| 100 | + ERRP_GUARD(); | ||
| 101 | + PCIProxyDev *dev = PCI_PROXY_DEV(device); | ||
| 102 | + int fd; | ||
| 103 | + | ||
| 104 | + if (!dev->fd) { | ||
| 105 | + error_setg(errp, "fd parameter not specified for %s", | ||
| 106 | + DEVICE(device)->id); | ||
| 107 | + return; | ||
| 108 | + } | ||
| 109 | + | ||
| 110 | + fd = monitor_fd_param(monitor_cur(), dev->fd, errp); | ||
| 111 | + if (fd == -1) { | ||
| 112 | + error_prepend(errp, "proxy: unable to parse fd %s: ", dev->fd); | ||
| 113 | + return; | ||
| 114 | + } | ||
| 115 | + | ||
| 116 | + if (!fd_is_socket(fd)) { | ||
| 117 | + error_setg(errp, "proxy: fd %d is not a socket", fd); | ||
| 118 | + close(fd); | ||
| 119 | + return; | ||
| 120 | + } | ||
| 121 | + | ||
| 122 | + dev->ioc = qio_channel_new_fd(fd, errp); | ||
| 123 | + | ||
| 124 | + error_setg(&dev->migration_blocker, "%s does not support migration", | ||
| 125 | + TYPE_PCI_PROXY_DEV); | ||
| 126 | + migrate_add_blocker(dev->migration_blocker, errp); | ||
| 127 | + | ||
| 128 | + qemu_mutex_init(&dev->io_mutex); | ||
| 129 | + qio_channel_set_blocking(dev->ioc, true, NULL); | ||
| 130 | +} | ||
| 131 | + | ||
| 132 | +static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
| 133 | +{ | ||
| 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 | |||
| 189 | -- | ||
| 190 | 2.29.2 | ||
| 191 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 1 | 2 | ||
| 3 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 4 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 5 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
| 6 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 7 | Message-id: d54edb4176361eed86b903e8f27058363b6c83b3.1611938319.git.jag.raman@oracle.com | ||
| 8 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 9 | --- | ||
| 10 | include/hw/remote/mpqemu-link.h | 4 ++++ | ||
| 11 | hw/remote/mpqemu-link.c | 34 +++++++++++++++++++++++++++++++++ | ||
| 12 | 2 files changed, 38 insertions(+) | ||
| 13 | |||
| 14 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
| 15 | index XXXXXXX..XXXXXXX 100644 | ||
| 16 | --- a/include/hw/remote/mpqemu-link.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 | |||
| 27 | @@ -XXX,XX +XXX,XX @@ typedef struct { | ||
| 28 | bool mpqemu_msg_send(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); | ||
| 29 | bool mpqemu_msg_recv(MPQemuMsg *msg, QIOChannel *ioc, Error **errp); | ||
| 30 | |||
| 31 | +uint64_t mpqemu_msg_send_and_await_reply(MPQemuMsg *msg, PCIProxyDev *pdev, | ||
| 32 | + Error **errp); | ||
| 33 | bool mpqemu_msg_valid(MPQemuMsg *msg); | ||
| 34 | |||
| 35 | #endif | ||
| 36 | diff --git a/hw/remote/mpqemu-link.c b/hw/remote/mpqemu-link.c | ||
| 37 | index XXXXXXX..XXXXXXX 100644 | ||
| 38 | --- a/hw/remote/mpqemu-link.c | ||
| 39 | +++ b/hw/remote/mpqemu-link.c | ||
| 40 | @@ -XXX,XX +XXX,XX @@ fail: | ||
| 41 | return ret; | ||
| 42 | } | ||
| 43 | |||
| 44 | +/* | ||
| 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) | ||
| 53 | +{ | ||
| 54 | + ERRP_GUARD(); | ||
| 55 | + MPQemuMsg msg_reply = {0}; | ||
| 56 | + uint64_t ret = UINT64_MAX; | ||
| 57 | + | ||
| 58 | + assert(!qemu_in_coroutine()); | ||
| 59 | + | ||
| 60 | + QEMU_LOCK_GUARD(&pdev->io_mutex); | ||
| 61 | + if (!mpqemu_msg_send(msg, pdev->ioc, errp)) { | ||
| 62 | + return ret; | ||
| 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; | ||
| 76 | +} | ||
| 77 | + | ||
| 78 | bool mpqemu_msg_valid(MPQemuMsg *msg) | ||
| 79 | { | ||
| 80 | if (msg->cmd >= MPQEMU_CMD_MAX && msg->cmd < 0) { | ||
| 81 | -- | ||
| 82 | 2.29.2 | ||
| 83 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> | |
| 2 | |||
| 3 | The Proxy Object sends the PCI config space accesses as messages | ||
| 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 | ||
| 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 12 | --- | ||
| 13 | include/hw/remote/mpqemu-link.h | 10 ++++++ | ||
| 14 | hw/remote/message.c | 60 +++++++++++++++++++++++++++++++++ | ||
| 15 | hw/remote/mpqemu-link.c | 8 ++++- | ||
| 16 | hw/remote/proxy.c | 55 ++++++++++++++++++++++++++++++ | ||
| 17 | 4 files changed, 132 insertions(+), 1 deletion(-) | ||
| 18 | |||
| 19 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
| 20 | index XXXXXXX..XXXXXXX 100644 | ||
| 21 | --- a/include/hw/remote/mpqemu-link.h | ||
| 22 | +++ b/include/hw/remote/mpqemu-link.h | ||
| 23 | @@ -XXX,XX +XXX,XX @@ | ||
| 24 | */ | ||
| 25 | typedef enum { | ||
| 26 | MPQEMU_CMD_SYNC_SYSMEM, | ||
| 27 | + MPQEMU_CMD_RET, | ||
| 28 | + MPQEMU_CMD_PCI_CFGWRITE, | ||
| 29 | + MPQEMU_CMD_PCI_CFGREAD, | ||
| 30 | MPQEMU_CMD_MAX, | ||
| 31 | } MPQemuCmd; | ||
| 32 | |||
| 33 | @@ -XXX,XX +XXX,XX @@ typedef struct { | ||
| 34 | off_t offsets[REMOTE_MAX_FDS]; | ||
| 35 | } SyncSysmemMsg; | ||
| 36 | |||
| 37 | +typedef struct { | ||
| 38 | + uint32_t addr; | ||
| 39 | + uint32_t val; | ||
| 40 | + int len; | ||
| 41 | +} PciConfDataMsg; | ||
| 42 | + | ||
| 43 | /** | ||
| 44 | * MPQemuMsg: | ||
| 45 | * @cmd: The remote command | ||
| 46 | @@ -XXX,XX +XXX,XX @@ typedef struct { | ||
| 47 | |||
| 48 | union { | ||
| 49 | uint64_t u64; | ||
| 50 | + PciConfDataMsg pci_conf_data; | ||
| 51 | SyncSysmemMsg sync_sysmem; | ||
| 52 | } data; | ||
| 53 | |||
| 54 | diff --git a/hw/remote/message.c b/hw/remote/message.c | ||
| 55 | index XXXXXXX..XXXXXXX 100644 | ||
| 56 | --- a/hw/remote/message.c | ||
| 57 | +++ b/hw/remote/message.c | ||
| 58 | @@ -XXX,XX +XXX,XX @@ | ||
| 59 | #include "hw/remote/mpqemu-link.h" | ||
| 60 | #include "qapi/error.h" | ||
| 61 | #include "sysemu/runstate.h" | ||
| 62 | +#include "hw/pci/pci.h" | ||
| 63 | + | ||
| 64 | +static void process_config_write(QIOChannel *ioc, PCIDevice *dev, | ||
| 65 | + MPQemuMsg *msg, Error **errp); | ||
| 66 | +static void process_config_read(QIOChannel *ioc, PCIDevice *dev, | ||
| 67 | + MPQemuMsg *msg, Error **errp); | ||
| 68 | |||
| 69 | void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 70 | { | ||
| 71 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 72 | } | ||
| 73 | |||
| 74 | switch (msg.cmd) { | ||
| 75 | + case MPQEMU_CMD_PCI_CFGWRITE: | ||
| 76 | + process_config_write(com->ioc, pci_dev, &msg, &local_err); | ||
| 77 | + break; | ||
| 78 | + case MPQEMU_CMD_PCI_CFGREAD: | ||
| 79 | + process_config_read(com->ioc, pci_dev, &msg, &local_err); | ||
| 80 | + break; | ||
| 81 | default: | ||
| 82 | error_setg(&local_err, | ||
| 83 | "Unknown command (%d) received for device %s" | ||
| 84 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 85 | qemu_system_shutdown_request(SHUTDOWN_CAUSE_GUEST_SHUTDOWN); | ||
| 86 | } | ||
| 87 | } | ||
| 88 | + | ||
| 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) | ||
| 174 | { | ||
| 175 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
| 176 | error_free(dev->migration_blocker); | ||
| 177 | } | ||
| 178 | |||
| 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 | |||
| 242 | -- | ||
| 243 | 2.29.2 | ||
| 244 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | 1 | From: Jagannathan Raman <jag.raman@oracle.com> | |
| 2 | |||
| 3 | Proxy device object implements handler for PCI BAR writes and reads. | ||
| 4 | The handler uses BAR_WRITE/BAR_READ message to communicate to the | ||
| 5 | remote process with the BAR address and value to be written/read. | ||
| 6 | The remote process implements handler for BAR_WRITE/BAR_READ | ||
| 7 | message. | ||
| 8 | |||
| 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 10 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 11 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
| 12 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 13 | Message-id: a8b76714a9688be5552c4c92d089bc9e8a4707ff.1611938319.git.jag.raman@oracle.com | ||
| 14 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 15 | --- | ||
| 16 | include/hw/remote/mpqemu-link.h | 10 ++++ | ||
| 17 | include/hw/remote/proxy.h | 9 ++++ | ||
| 18 | hw/remote/message.c | 83 +++++++++++++++++++++++++++++++++ | ||
| 19 | hw/remote/mpqemu-link.c | 6 +++ | ||
| 20 | hw/remote/proxy.c | 60 ++++++++++++++++++++++++ | ||
| 21 | 5 files changed, 168 insertions(+) | ||
| 22 | |||
| 23 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h | ||
| 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]; | ||
| 82 | }; | ||
| 83 | |||
| 84 | #endif /* PROXY_H */ | ||
| 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) | ||
| 103 | { | ||
| 104 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 105 | case MPQEMU_CMD_PCI_CFGREAD: | ||
| 106 | process_config_read(com->ioc, pci_dev, &msg, &local_err); | ||
| 107 | break; | ||
| 108 | + case MPQEMU_CMD_BAR_WRITE: | ||
| 109 | + process_bar_write(com->ioc, &msg, &local_err); | ||
| 110 | + break; | ||
| 111 | + case MPQEMU_CMD_BAR_READ: | ||
| 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()); | ||
| 119 | } | ||
| 120 | } | ||
| 121 | + | ||
| 122 | +static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp) | ||
| 123 | +{ | ||
| 124 | + ERRP_GUARD(); | ||
| 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; | ||
| 211 | } | ||
| 212 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
| 213 | index XXXXXXX..XXXXXXX 100644 | ||
| 214 | --- a/hw/remote/proxy.c | ||
| 215 | +++ b/hw/remote/proxy.c | ||
| 216 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_register_types(void) | ||
| 217 | } | ||
| 218 | |||
| 219 | type_init(pci_proxy_dev_register_types) | ||
| 220 | + | ||
| 221 | +static void send_bar_access_msg(PCIProxyDev *pdev, MemoryRegion *mr, | ||
| 222 | + bool write, hwaddr addr, uint64_t *val, | ||
| 223 | + unsigned size, bool memory) | ||
| 224 | +{ | ||
| 225 | + MPQemuMsg msg = { 0 }; | ||
| 226 | + long ret = -EINVAL; | ||
| 227 | + Error *local_err = NULL; | ||
| 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 | +}; | ||
| 280 | -- | ||
| 281 | 2.29.2 | ||
| 282 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
| 1 | 2 | ||
| 3 | Add ProxyMemoryListener object which is used to keep the view of the RAM | ||
| 4 | in sync between QEMU and remote process. | ||
| 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. | ||
| 9 | |||
| 10 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 11 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
| 12 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 13 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 14 | Message-id: 04fe4e6a9ca90d4f11ab6f59be7652f5b086a071.1611938319.git.jag.raman@oracle.com | ||
| 15 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 16 | --- | ||
| 17 | MAINTAINERS | 2 + | ||
| 18 | include/hw/remote/proxy-memory-listener.h | 28 +++ | ||
| 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 | ||
| 27 | |||
| 28 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 29 | index XXXXXXX..XXXXXXX 100644 | ||
| 30 | --- a/MAINTAINERS | ||
| 31 | +++ b/MAINTAINERS | ||
| 32 | @@ -XXX,XX +XXX,XX @@ F: include/hw/remote/memory.h | ||
| 33 | F: hw/remote/memory.c | ||
| 34 | F: hw/remote/proxy.c | ||
| 35 | F: include/hw/remote/proxy.h | ||
| 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]; | ||
| 93 | }; | ||
| 94 | |||
| 95 | diff --git a/hw/remote/message.c b/hw/remote/message.c | ||
| 96 | index XXXXXXX..XXXXXXX 100644 | ||
| 97 | --- a/hw/remote/message.c | ||
| 98 | +++ b/hw/remote/message.c | ||
| 99 | @@ -XXX,XX +XXX,XX @@ | ||
| 100 | #include "sysemu/runstate.h" | ||
| 101 | #include "hw/pci/pci.h" | ||
| 102 | #include "exec/memattrs.h" | ||
| 103 | +#include "hw/remote/memory.h" | ||
| 104 | |||
| 105 | static void process_config_write(QIOChannel *ioc, PCIDevice *dev, | ||
| 106 | MPQemuMsg *msg, Error **errp); | ||
| 107 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 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)) { | ||
| 273 | + return; | ||
| 274 | + } | ||
| 275 | + | ||
| 276 | + if (try_merge(proxy_listener, section)) { | ||
| 277 | + return; | ||
| 278 | + } | ||
| 279 | + | ||
| 280 | + ++proxy_listener->n_mr_sections; | ||
| 281 | + proxy_listener->mr_sections = g_renew(MemoryRegionSection, | ||
| 282 | + proxy_listener->mr_sections, | ||
| 283 | + proxy_listener->n_mr_sections); | ||
| 284 | + proxy_listener->mr_sections[proxy_listener->n_mr_sections - 1] = *section; | ||
| 285 | + proxy_listener->mr_sections[proxy_listener->n_mr_sections - 1].fv = NULL; | ||
| 286 | + memory_region_ref(section->mr); | ||
| 287 | +} | ||
| 288 | + | ||
| 289 | +static void proxy_memory_listener_commit(MemoryListener *listener) | ||
| 290 | +{ | ||
| 291 | + ProxyMemoryListener *proxy_listener = container_of(listener, | ||
| 292 | + ProxyMemoryListener, | ||
| 293 | + listener); | ||
| 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) | ||
| 392 | -- | ||
| 393 | 2.29.2 | ||
| 394 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
| 1 | 2 | ||
| 3 | IOHUB object is added to manage PCI IRQs. It uses KVM_IRQFD | ||
| 4 | ioctl to create irqfd to injecting PCI interrupts to the guest. | ||
| 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. | ||
| 7 | |||
| 8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> | ||
| 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 10 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> | ||
| 11 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 12 | Message-id: 51d5c3d54e28a68b002e3875c59599c9f5a424a1.1611938319.git.jag.raman@oracle.com | ||
| 13 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 14 | --- | ||
| 15 | MAINTAINERS | 2 + | ||
| 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 | ||
| 30 | |||
| 31 | diff --git a/MAINTAINERS b/MAINTAINERS | ||
| 32 | index XXXXXXX..XXXXXXX 100644 | ||
| 33 | --- a/MAINTAINERS | ||
| 34 | +++ b/MAINTAINERS | ||
| 35 | @@ -XXX,XX +XXX,XX @@ F: hw/remote/proxy.c | ||
| 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); | ||
| 314 | } | ||
| 315 | |||
| 316 | static void remote_machine_class_init(ObjectClass *oc, void *data) | ||
| 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) | ||
| 416 | { | ||
| 417 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_realize(PCIDevice *device, Error **errp) | ||
| 418 | qio_channel_set_blocking(dev->ioc, true, NULL); | ||
| 419 | |||
| 420 | proxy_memory_listener_configure(&dev->proxy_listener, dev->ioc); | ||
| 421 | + | ||
| 422 | + setup_irqfd(dev); | ||
| 423 | } | ||
| 424 | |||
| 425 | static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
| 426 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
| 427 | error_free(dev->migration_blocker); | ||
| 428 | |||
| 429 | proxy_memory_listener_deconfigure(&dev->proxy_listener); | ||
| 430 | + | ||
| 431 | + event_notifier_cleanup(&dev->intr); | ||
| 432 | + event_notifier_cleanup(&dev->resample); | ||
| 433 | } | ||
| 434 | |||
| 435 | static void config_op_send(PCIProxyDev *pdev, uint32_t addr, uint32_t *val, | ||
| 436 | diff --git a/hw/remote/meson.build b/hw/remote/meson.build | ||
| 437 | index XXXXXXX..XXXXXXX 100644 | ||
| 438 | --- a/hw/remote/meson.build | ||
| 439 | +++ b/hw/remote/meson.build | ||
| 440 | @@ -XXX,XX +XXX,XX @@ remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('mpqemu-link.c')) | ||
| 441 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('message.c')) | ||
| 442 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('remote-obj.c')) | ||
| 443 | remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy.c')) | ||
| 444 | +remote_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('iohub.c')) | ||
| 445 | |||
| 446 | specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('memory.c')) | ||
| 447 | specific_ss.add(when: 'CONFIG_MULTIPROCESS', if_true: files('proxy-memory-listener.c')) | ||
| 448 | -- | ||
| 449 | 2.29.2 | ||
| 450 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: Jagannathan Raman <jag.raman@oracle.com> | ||
| 1 | 2 | ||
| 3 | Retrieve PCI configuration info about the remote device and | ||
| 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 | ||
| 11 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 12 | --- | ||
| 13 | hw/remote/proxy.c | 84 +++++++++++++++++++++++++++++++++++++++++++++++ | ||
| 14 | 1 file changed, 84 insertions(+) | ||
| 15 | |||
| 16 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | ||
| 17 | index XXXXXXX..XXXXXXX 100644 | ||
| 18 | --- a/hw/remote/proxy.c | ||
| 19 | +++ b/hw/remote/proxy.c | ||
| 20 | @@ -XXX,XX +XXX,XX @@ | ||
| 21 | #include "sysemu/kvm.h" | ||
| 22 | #include "util/event_notifier-posix.c" | ||
| 23 | |||
| 24 | +static void probe_pci_info(PCIDevice *dev, Error **errp); | ||
| 25 | + | ||
| 26 | static void proxy_intx_update(PCIDevice *pci_dev) | ||
| 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; | ||
| 43 | + | ||
| 44 | proxy_memory_listener_configure(&dev->proxy_listener, dev->ioc); | ||
| 45 | |||
| 46 | setup_irqfd(dev); | ||
| 47 | + | ||
| 48 | + probe_pci_info(PCI_DEVICE(dev), errp); | ||
| 49 | } | ||
| 50 | |||
| 51 | static void pci_proxy_dev_exit(PCIDevice *pdev) | ||
| 52 | @@ -XXX,XX +XXX,XX @@ const MemoryRegionOps proxy_mr_ops = { | ||
| 53 | .max_access_size = 8, | ||
| 54 | }, | ||
| 55 | }; | ||
| 56 | + | ||
| 57 | +static void probe_pci_info(PCIDevice *dev, Error **errp) | ||
| 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; | ||
| 65 | + | ||
| 66 | + config_op_send(pdev, PCI_VENDOR_ID, &val, 2, MPQEMU_CMD_PCI_CFGREAD); | ||
| 67 | + pc->vendor_id = (uint16_t)val; | ||
| 68 | + | ||
| 69 | + config_op_send(pdev, PCI_DEVICE_ID, &val, 2, MPQEMU_CMD_PCI_CFGREAD); | ||
| 70 | + pc->device_id = (uint16_t)val; | ||
| 71 | + | ||
| 72 | + config_op_send(pdev, PCI_CLASS_DEVICE, &val, 2, MPQEMU_CMD_PCI_CFGREAD); | ||
| 73 | + pc->class_id = (uint16_t)val; | ||
| 74 | + | ||
| 75 | + config_op_send(pdev, PCI_SUBSYSTEM_ID, &val, 2, MPQEMU_CMD_PCI_CFGREAD); | ||
| 76 | + pc->subsystem_id = (uint16_t)val; | ||
| 77 | + | ||
| 78 | + base_class = pc->class_id >> 4; | ||
| 79 | + switch (base_class) { | ||
| 80 | + case PCI_BASE_CLASS_BRIDGE: | ||
| 81 | + set_bit(DEVICE_CATEGORY_BRIDGE, dc->categories); | ||
| 82 | + break; | ||
| 83 | + case PCI_BASE_CLASS_STORAGE: | ||
| 84 | + set_bit(DEVICE_CATEGORY_STORAGE, dc->categories); | ||
| 85 | + break; | ||
| 86 | + case PCI_BASE_CLASS_NETWORK: | ||
| 87 | + set_bit(DEVICE_CATEGORY_NETWORK, dc->categories); | ||
| 88 | + break; | ||
| 89 | + case PCI_BASE_CLASS_INPUT: | ||
| 90 | + set_bit(DEVICE_CATEGORY_INPUT, dc->categories); | ||
| 91 | + break; | ||
| 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 | + } | ||
| 102 | + | ||
| 103 | + for (i = 0; i < PCI_NUM_REGIONS; i++) { | ||
| 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; | ||
| 116 | + | ||
| 117 | + if (size) { | ||
| 118 | + g_autofree char *name; | ||
| 119 | + pdev->region[i].dev = pdev; | ||
| 120 | + pdev->region[i].present = true; | ||
| 121 | + if (type == PCI_BASE_ADDRESS_SPACE_MEMORY) { | ||
| 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 | +} | ||
| 132 | -- | ||
| 133 | 2.29.2 | ||
| 134 | diff view generated by jsdifflib |
| 1 | From: Zenghui Yu <yuzenghui@huawei.com> | 1 | From: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
|---|---|---|---|
| 2 | 2 | ||
| 3 | Quote docs/devel/style.rst (section "Automatic memory deallocation"): | 3 | Perform device reset in the remote process when QEMU performs |
| 4 | device reset. This is required to reset the internal state | ||
| 5 | (like registers, etc...) of emulated devices | ||
| 4 | 6 | ||
| 5 | * Variables declared with g_auto* MUST always be initialized, | 7 | Signed-off-by: Elena Ufimtseva <elena.ufimtseva@oracle.com> |
| 6 | otherwise the cleanup function will use uninitialized stack memory | 8 | Signed-off-by: John G Johnson <john.g.johnson@oracle.com> |
| 7 | 9 | Signed-off-by: Jagannathan Raman <jag.raman@oracle.com> | |
| 8 | Initialize @name properly to get rid of the compilation error (using | 10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
| 9 | gcc-7.3.0 on CentOS): | 11 | Message-id: 7cb220a51f565dc0817bd76e2f540e89c2d2b850.1611938319.git.jag.raman@oracle.com |
| 10 | |||
| 11 | ../hw/remote/proxy.c: In function 'pci_proxy_dev_realize': | ||
| 12 | /usr/include/glib-2.0/glib/glib-autocleanups.h:28:3: error: 'name' may be used uninitialized in this function [-Werror=maybe-uninitialized] | ||
| 13 | g_free (*pp); | ||
| 14 | ^~~~~~~~~~~~ | ||
| 15 | ../hw/remote/proxy.c:350:30: note: 'name' was declared here | ||
| 16 | g_autofree char *name; | ||
| 17 | ^~~~ | ||
| 18 | |||
| 19 | Signed-off-by: Zenghui Yu <yuzenghui@huawei.com> | ||
| 20 | Reviewed-by: Jagannathan Raman <jag.raman@oracle.com> | ||
| 21 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
| 22 | Reviewed-by: Miroslav Rezanina <mrezanin@redhat.com> | ||
| 23 | Message-id: 20210312112143.1369-1-yuzenghui@huawei.com | ||
| 24 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | 12 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> |
| 25 | --- | 13 | --- |
| 26 | hw/remote/memory.c | 5 ++--- | 14 | include/hw/remote/mpqemu-link.h | 1 + |
| 27 | hw/remote/proxy.c | 3 +-- | 15 | hw/remote/message.c | 22 ++++++++++++++++++++++ |
| 28 | 2 files changed, 3 insertions(+), 5 deletions(-) | 16 | hw/remote/proxy.c | 19 +++++++++++++++++++ |
| 17 | 3 files changed, 42 insertions(+) | ||
| 29 | 18 | ||
| 30 | diff --git a/hw/remote/memory.c b/hw/remote/memory.c | 19 | diff --git a/include/hw/remote/mpqemu-link.h b/include/hw/remote/mpqemu-link.h |
| 31 | index XXXXXXX..XXXXXXX 100644 | 20 | index XXXXXXX..XXXXXXX 100644 |
| 32 | --- a/hw/remote/memory.c | 21 | --- a/include/hw/remote/mpqemu-link.h |
| 33 | +++ b/hw/remote/memory.c | 22 | +++ b/include/hw/remote/mpqemu-link.h |
| 34 | @@ -XXX,XX +XXX,XX @@ void remote_sysmem_reconfig(MPQemuMsg *msg, Error **errp) | 23 | @@ -XXX,XX +XXX,XX @@ typedef enum { |
| 35 | 24 | MPQEMU_CMD_BAR_WRITE, | |
| 36 | remote_sysmem_reset(); | 25 | MPQEMU_CMD_BAR_READ, |
| 37 | 26 | MPQEMU_CMD_SET_IRQFD, | |
| 38 | - for (region = 0; region < msg->num_fds; region++) { | 27 | + MPQEMU_CMD_DEVICE_RESET, |
| 39 | - g_autofree char *name; | 28 | MPQEMU_CMD_MAX, |
| 40 | + for (region = 0; region < msg->num_fds; region++, suffix++) { | 29 | } MPQemuCmd; |
| 41 | + g_autofree char *name = g_strdup_printf("remote-mem-%u", suffix); | 30 | |
| 42 | subregion = g_new(MemoryRegion, 1); | 31 | diff --git a/hw/remote/message.c b/hw/remote/message.c |
| 43 | - name = g_strdup_printf("remote-mem-%u", suffix++); | 32 | index XXXXXXX..XXXXXXX 100644 |
| 44 | memory_region_init_ram_from_fd(subregion, NULL, | 33 | --- a/hw/remote/message.c |
| 45 | name, sysmem_info->sizes[region], | 34 | +++ b/hw/remote/message.c |
| 46 | true, msg->fds[region], | 35 | @@ -XXX,XX +XXX,XX @@ |
| 36 | #include "exec/memattrs.h" | ||
| 37 | #include "hw/remote/memory.h" | ||
| 38 | #include "hw/remote/iohub.h" | ||
| 39 | +#include "sysemu/reset.h" | ||
| 40 | |||
| 41 | static void process_config_write(QIOChannel *ioc, PCIDevice *dev, | ||
| 42 | MPQemuMsg *msg, Error **errp); | ||
| 43 | @@ -XXX,XX +XXX,XX @@ static void process_config_read(QIOChannel *ioc, PCIDevice *dev, | ||
| 44 | MPQemuMsg *msg, Error **errp); | ||
| 45 | static void process_bar_write(QIOChannel *ioc, MPQemuMsg *msg, Error **errp); | ||
| 46 | static void process_bar_read(QIOChannel *ioc, MPQemuMsg *msg, Error **errp); | ||
| 47 | +static void process_device_reset_msg(QIOChannel *ioc, PCIDevice *dev, | ||
| 48 | + Error **errp); | ||
| 49 | |||
| 50 | void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 51 | { | ||
| 52 | @@ -XXX,XX +XXX,XX @@ void coroutine_fn mpqemu_remote_msg_loop_co(void *data) | ||
| 53 | case MPQEMU_CMD_SET_IRQFD: | ||
| 54 | process_set_irqfd_msg(pci_dev, &msg); | ||
| 55 | break; | ||
| 56 | + case MPQEMU_CMD_DEVICE_RESET: | ||
| 57 | + process_device_reset_msg(com->ioc, pci_dev, &local_err); | ||
| 58 | + break; | ||
| 59 | default: | ||
| 60 | error_setg(&local_err, | ||
| 61 | "Unknown command (%d) received for device %s" | ||
| 62 | @@ -XXX,XX +XXX,XX @@ fail: | ||
| 63 | getpid()); | ||
| 64 | } | ||
| 65 | } | ||
| 66 | + | ||
| 67 | +static void process_device_reset_msg(QIOChannel *ioc, PCIDevice *dev, | ||
| 68 | + 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 | +} | ||
| 47 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c | 82 | diff --git a/hw/remote/proxy.c b/hw/remote/proxy.c |
| 48 | index XXXXXXX..XXXXXXX 100644 | 83 | index XXXXXXX..XXXXXXX 100644 |
| 49 | --- a/hw/remote/proxy.c | 84 | --- a/hw/remote/proxy.c |
| 50 | +++ b/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) | ||
| 93 | { | ||
| 94 | @@ -XXX,XX +XXX,XX @@ static void pci_proxy_dev_class_init(ObjectClass *klass, void *data) | ||
| 95 | k->config_read = pci_proxy_read_config; | ||
| 96 | k->config_write = pci_proxy_write_config; | ||
| 97 | |||
| 98 | + dc->reset = proxy_device_reset; | ||
| 99 | + | ||
| 100 | device_class_set_props(dc, proxy_properties); | ||
| 101 | } | ||
| 102 | |||
| 51 | @@ -XXX,XX +XXX,XX @@ static void probe_pci_info(PCIDevice *dev, Error **errp) | 103 | @@ -XXX,XX +XXX,XX @@ static void probe_pci_info(PCIDevice *dev, Error **errp) |
| 52 | PCI_BASE_ADDRESS_SPACE_IO : PCI_BASE_ADDRESS_SPACE_MEMORY; | 104 | } |
| 53 | 105 | } | |
| 54 | if (size) { | 106 | } |
| 55 | - g_autofree char *name; | 107 | + |
| 56 | + g_autofree char *name = g_strdup_printf("bar-region-%d", i); | 108 | +static void proxy_device_reset(DeviceState *dev) |
| 57 | pdev->region[i].dev = pdev; | 109 | +{ |
| 58 | pdev->region[i].present = true; | 110 | + PCIProxyDev *pdev = PCI_PROXY_DEV(dev); |
| 59 | if (type == PCI_BASE_ADDRESS_SPACE_MEMORY) { | 111 | + MPQemuMsg msg = { 0 }; |
| 60 | pdev->region[i].memory = true; | 112 | + Error *local_err = NULL; |
| 61 | } | 113 | + |
| 62 | - name = g_strdup_printf("bar-region-%d", i); | 114 | + msg.cmd = MPQEMU_CMD_DEVICE_RESET; |
| 63 | memory_region_init_io(&pdev->region[i].mr, OBJECT(pdev), | 115 | + msg.size = 0; |
| 64 | &proxy_mr_ops, &pdev->region[i], | 116 | + |
| 65 | name, size); | 117 | + mpqemu_msg_send_and_await_reply(&msg, pdev, &local_err); |
| 118 | + if (local_err) { | ||
| 119 | + error_report_err(local_err); | ||
| 120 | + } | ||
| 121 | + | ||
| 122 | +} | ||
| 66 | -- | 123 | -- |
| 67 | 2.31.1 | 124 | 2.29.2 |
| 68 | 125 | diff view generated by jsdifflib |
New patch | |||
|---|---|---|---|
| 1 | From: "Denis V. Lunev" <den@openvz.org> | ||
| 1 | 2 | ||
| 3 | Original specification says that l1 table size if 64 * l1_size, which | ||
| 4 | is obviously wrong. The size of the l1 entry is 64 _bits_, not bytes. | ||
| 5 | Thus 64 is to be replaces with 8 as specification says about bytes. | ||
| 6 | |||
| 7 | There is also minor tweak, field name is renamed from l1 to l1_table, | ||
| 8 | which matches with the later text. | ||
| 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 | |||
| 20 | Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
| 21 | --- | ||
| 22 | docs/interop/parallels.txt | 2 +- | ||
| 23 | 1 file changed, 1 insertion(+), 1 deletion(-) | ||
| 24 | |||
| 25 | diff --git a/docs/interop/parallels.txt b/docs/interop/parallels.txt | ||
| 26 | index XXXXXXX..XXXXXXX 100644 | ||
| 27 | --- a/docs/interop/parallels.txt | ||
| 28 | +++ b/docs/interop/parallels.txt | ||
| 29 | @@ -XXX,XX +XXX,XX @@ of its data area are: | ||
| 30 | 28 - 31: l1_size | ||
| 31 | The number of entries in the L1 table of the bitmap. | ||
| 32 | |||
| 33 | - variable: l1 (64 * l1_size bytes) | ||
| 34 | + variable: l1_table (8 * l1_size bytes) | ||
| 35 | L1 offset table (in bytes) | ||
| 36 | |||
| 37 | A dirty bitmap is stored using a one-level structure for the mapping to host | ||
| 38 | -- | ||
| 39 | 2.29.2 | ||
| 40 | diff view generated by jsdifflib |