1 | The following changes since commit 16aaacb307ed607b9780c12702c44f0fe52edc7e: | 1 | The following changes since commit 8cb41fda78c7ebde0dd248c6afe1d336efb0de50: |
---|---|---|---|
2 | 2 | ||
3 | Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20200430' into staging (2020-04-30 14:00:36 +0100) | 3 | Merge remote-tracking branch 'remotes/philmd/tags/machine-20211101' into staging (2021-11-02 05:53:45 -0400) |
4 | 4 | ||
5 | are available in the Git repository at: | 5 | are available in the Git repository at: |
6 | 6 | ||
7 | git://repo.or.cz/qemu/kevin.git tags/for-upstream | 7 | git://repo.or.cz/qemu/kevin.git tags/for-upstream |
8 | 8 | ||
9 | for you to fetch changes up to eaae29ef89d498d0eac553c77b554f310a47f809: | 9 | for you to fetch changes up to a8951438946d72d74c9bdbdb38fce95aa2973a88: |
10 | 10 | ||
11 | qemu-storage-daemon: Fix non-string --object properties (2020-04-30 17:51:07 +0200) | 11 | block/nvme: Extract nvme_free_queue() from nvme_free_queue_pair() (2021-11-02 15:49:13 +0100) |
12 | 12 | ||
13 | ---------------------------------------------------------------- | 13 | ---------------------------------------------------------------- |
14 | Block layer patches: | 14 | Block layer patches |
15 | 15 | ||
16 | - Fix resize (extending) of short overlays | 16 | - Fail gracefully when blockdev-snapshot creates loops |
17 | - nvme: introduce PMR support from NVMe 1.4 spec | 17 | - ide: Fix IDENTIFY DEVICE for disks > 128 GiB |
18 | - qemu-storage-daemon: Fix non-string --object properties | 18 | - file-posix: Fix return value translation for AIO discards |
19 | - file-posix: add 'aio-max-batch' option | ||
20 | - rbd: implement bdrv_co_block_status | ||
21 | - Code cleanups and build fixes | ||
19 | 22 | ||
20 | ---------------------------------------------------------------- | 23 | ---------------------------------------------------------------- |
21 | Alberto Garcia (1): | 24 | Ari Sundholm (1): |
22 | qcow2: Add incompatibility note between backing files and raw external data files | 25 | block/file-posix: Fix return value translation for AIO discards |
23 | 26 | ||
24 | Andrzej Jakowski (1): | 27 | Fabrice Fontaine (1): |
25 | nvme: introduce PMR support from NVMe 1.4 spec | 28 | block/export/fuse.c: fix musl build |
26 | 29 | ||
27 | Kevin Wolf (12): | 30 | Hanna Reitz (1): |
28 | block: Add flags to BlockDriver.bdrv_co_truncate() | 31 | block-backend: Silence clang -m32 compiler warning |
29 | block: Add flags to bdrv(_co)_truncate() | ||
30 | block-backend: Add flags to blk_truncate() | ||
31 | qcow2: Support BDRV_REQ_ZERO_WRITE for truncate | ||
32 | raw-format: Support BDRV_REQ_ZERO_WRITE for truncate | ||
33 | file-posix: Support BDRV_REQ_ZERO_WRITE for truncate | ||
34 | block: truncate: Don't make backing file data visible | ||
35 | iotests: Filter testfiles out in filter_img_info() | ||
36 | iotests: Test committing to short backing file | ||
37 | qcow2: Forward ZERO_WRITE flag for full preallocation | ||
38 | qom: Factor out user_creatable_add_dict() | ||
39 | qemu-storage-daemon: Fix non-string --object properties | ||
40 | 32 | ||
41 | Paolo Bonzini (1): | 33 | Kevin Wolf (1): |
42 | qemu-iotests: allow qcow2 external discarded clusters to contain stale data | 34 | block: Fail gracefully when blockdev-snapshot creates loops |
43 | 35 | ||
44 | docs/interop/qcow2.txt | 3 + | 36 | Peter Lieven (1): |
45 | hw/block/nvme.h | 2 + | 37 | block/rbd: implement bdrv_co_block_status |
46 | include/block/block.h | 5 +- | 38 | |
47 | include/block/block_int.h | 10 +- | 39 | Philippe Mathieu-Daudé (3): |
48 | include/block/nvme.h | 172 ++++++++++++++++++++++++++ | 40 | block/nvme: Automatically free qemu_memalign() with QEMU_AUTO_VFREE |
49 | include/qom/object_interfaces.h | 16 +++ | 41 | block/nvme: Display CQ/SQ pointer in nvme_free_queue_pair() |
50 | include/sysemu/block-backend.h | 2 +- | 42 | block/nvme: Extract nvme_free_queue() from nvme_free_queue_pair() |
51 | block.c | 3 +- | 43 | |
52 | block/block-backend.c | 4 +- | 44 | Samuel Thibault (1): |
53 | block/commit.c | 4 +- | 45 | ide: Cap LBA28 capacity announcement to 2^28-1 |
54 | block/crypto.c | 7 +- | 46 | |
55 | block/file-posix.c | 6 +- | 47 | Stefano Garzarella (3): |
56 | block/file-win32.c | 2 +- | 48 | file-posix: add `aio-max-batch` option |
57 | block/gluster.c | 1 + | 49 | linux-aio: add `dev_max_batch` parameter to laio_co_submit() |
58 | block/io.c | 43 ++++++- | 50 | linux-aio: add `dev_max_batch` parameter to laio_io_unplug() |
59 | block/iscsi.c | 2 +- | 51 | |
60 | block/mirror.c | 2 +- | 52 | qapi/block-core.json | 7 +++ |
61 | block/nfs.c | 3 +- | 53 | include/block/raw-aio.h | 6 ++- |
62 | block/parallels.c | 6 +- | 54 | block.c | 10 ++++ |
63 | block/qcow.c | 4 +- | 55 | block/block-backend.c | 2 +- |
64 | block/qcow2-cluster.c | 2 +- | 56 | block/export/fuse.c | 4 ++ |
65 | block/qcow2-refcount.c | 2 +- | 57 | block/file-posix.c | 18 ++++++-- |
66 | block/qcow2.c | 73 +++++++++-- | 58 | block/linux-aio.c | 38 ++++++++++----- |
67 | block/qed.c | 3 +- | 59 | block/nvme.c | 22 +++++---- |
68 | block/raw-format.c | 6 +- | 60 | block/rbd.c | 112 +++++++++++++++++++++++++++++++++++++++++++++ |
69 | block/rbd.c | 1 + | 61 | hw/ide/core.c | 8 +++- |
70 | block/sheepdog.c | 4 +- | 62 | block/trace-events | 2 +- |
71 | block/ssh.c | 2 +- | 63 | tests/qemu-iotests/085 | 31 ++++++++++++- |
72 | block/vdi.c | 2 +- | 64 | tests/qemu-iotests/085.out | 33 +++++++++++-- |
73 | block/vhdx-log.c | 2 +- | 65 | 13 files changed, 258 insertions(+), 35 deletions(-) |
74 | block/vhdx.c | 6 +- | ||
75 | block/vmdk.c | 8 +- | ||
76 | block/vpc.c | 2 +- | ||
77 | blockdev.c | 2 +- | ||
78 | hw/block/nvme.c | 109 ++++++++++++++++ | ||
79 | qemu-img.c | 2 +- | ||
80 | qemu-io-cmds.c | 2 +- | ||
81 | qemu-storage-daemon.c | 4 +- | ||
82 | qom/object_interfaces.c | 31 +++++ | ||
83 | qom/qom-qmp-cmds.c | 24 +--- | ||
84 | tests/test-block-iothread.c | 9 +- | ||
85 | tests/qemu-iotests/iotests.py | 5 +- | ||
86 | hw/block/Makefile.objs | 2 +- | ||
87 | hw/block/trace-events | 4 + | ||
88 | tests/qemu-iotests/244 | 10 +- | ||
89 | tests/qemu-iotests/244.out | 9 +- | ||
90 | tests/qemu-iotests/274 | 155 +++++++++++++++++++++++ | ||
91 | tests/qemu-iotests/274.out | 268 ++++++++++++++++++++++++++++++++++++++++ | ||
92 | tests/qemu-iotests/group | 1 + | ||
93 | 49 files changed, 951 insertions(+), 96 deletions(-) | ||
94 | create mode 100755 tests/qemu-iotests/274 | ||
95 | create mode 100644 tests/qemu-iotests/274.out | ||
96 | 66 | ||
97 | 67 | diff view generated by jsdifflib |
Deleted patch | |||
---|---|---|---|
1 | From: Alberto Garcia <berto@igalia.com> | ||
2 | 1 | ||
3 | Backing files and raw external data files are mutually exclusive. | ||
4 | The documentation of the raw external data bit (in autoclear_features) | ||
5 | already indicates that, but we should also mention it on the other | ||
6 | side. | ||
7 | |||
8 | Suggested-by: Eric Blake <eblake@redhat.com> | ||
9 | Signed-off-by: Alberto Garcia <berto@igalia.com> | ||
10 | Message-Id: <20200410121816.8334-1-berto@igalia.com> | ||
11 | Reviewed-by: Eric Blake <eblake@redhat.com> | ||
12 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | ||
13 | --- | ||
14 | docs/interop/qcow2.txt | 3 +++ | ||
15 | 1 file changed, 3 insertions(+) | ||
16 | |||
17 | diff --git a/docs/interop/qcow2.txt b/docs/interop/qcow2.txt | ||
18 | index XXXXXXX..XXXXXXX 100644 | ||
19 | --- a/docs/interop/qcow2.txt | ||
20 | +++ b/docs/interop/qcow2.txt | ||
21 | @@ -XXX,XX +XXX,XX @@ The first cluster of a qcow2 image contains the file header: | ||
22 | is stored (NB: The string is not null terminated). 0 if the | ||
23 | image doesn't have a backing file. | ||
24 | |||
25 | + Note: backing files are incompatible with raw external data | ||
26 | + files (auto-clear feature bit 1). | ||
27 | + | ||
28 | 16 - 19: backing_file_size | ||
29 | Length of the backing file name in bytes. Must not be | ||
30 | longer than 1023 bytes. Undefined if the image doesn't have | ||
31 | -- | ||
32 | 2.25.3 | ||
33 | |||
34 | diff view generated by jsdifflib |
Deleted patch | |||
---|---|---|---|
1 | From: Paolo Bonzini <pbonzini@redhat.com> | ||
2 | 1 | ||
3 | Test 244 checks the expected behavior of qcow2 external data files | ||
4 | with respect to zero and discarded clusters. Filesystems however | ||
5 | are free to ignore discard requests, and this seems to be the | ||
6 | case for overlayfs. Relax the tests to skip checks on the | ||
7 | external data file for discarded areas, which implies not using | ||
8 | qemu-img compare in the data_file_raw=on case. | ||
9 | |||
10 | This fixes docker tests on RHEL8. | ||
11 | |||
12 | Cc: Kevin Wolf <kwolf@redhat.com> | ||
13 | Cc: qemu-block@nongnu.org | ||
14 | Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> | ||
15 | Message-Id: <20200409191006.24429-1-pbonzini@redhat.com> | ||
16 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | ||
17 | --- | ||
18 | tests/qemu-iotests/244 | 10 ++++++++-- | ||
19 | tests/qemu-iotests/244.out | 9 ++++++--- | ||
20 | 2 files changed, 14 insertions(+), 5 deletions(-) | ||
21 | |||
22 | diff --git a/tests/qemu-iotests/244 b/tests/qemu-iotests/244 | ||
23 | index XXXXXXX..XXXXXXX 100755 | ||
24 | --- a/tests/qemu-iotests/244 | ||
25 | +++ b/tests/qemu-iotests/244 | ||
26 | @@ -XXX,XX +XXX,XX @@ $QEMU_IO -c 'read -P 0 0 1M' \ | ||
27 | echo | ||
28 | $QEMU_IO -c 'read -P 0 0 1M' \ | ||
29 | -c 'read -P 0x11 1M 1M' \ | ||
30 | - -c 'read -P 0 2M 2M' \ | ||
31 | -c 'read -P 0x11 4M 1M' \ | ||
32 | -c 'read -P 0 5M 1M' \ | ||
33 | -f raw "$TEST_IMG.data" | | ||
34 | @@ -XXX,XX +XXX,XX @@ $QEMU_IO -c 'read -P 0 0 1M' \ | ||
35 | -f $IMGFMT "$TEST_IMG" | | ||
36 | _filter_qemu_io | ||
37 | |||
38 | +# Discarded clusters are only marked as such in the qcow2 metadata, but | ||
39 | +# they can contain stale data in the external data file. Instead, zero | ||
40 | +# clusters must be zeroed in the external data file too. | ||
41 | echo | ||
42 | -$QEMU_IMG compare "$TEST_IMG" "$TEST_IMG.data" | ||
43 | +$QEMU_IO -c 'read -P 0 0 1M' \ | ||
44 | + -c 'read -P 0x11 1M 1M' \ | ||
45 | + -c 'read -P 0 3M 3M' \ | ||
46 | + -f raw "$TEST_IMG".data | | ||
47 | + _filter_qemu_io | ||
48 | |||
49 | echo -n "qcow2 file size after I/O: " | ||
50 | du -b $TEST_IMG | cut -f1 | ||
51 | diff --git a/tests/qemu-iotests/244.out b/tests/qemu-iotests/244.out | ||
52 | index XXXXXXX..XXXXXXX 100644 | ||
53 | --- a/tests/qemu-iotests/244.out | ||
54 | +++ b/tests/qemu-iotests/244.out | ||
55 | @@ -XXX,XX +XXX,XX @@ read 1048576/1048576 bytes at offset 0 | ||
56 | 1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
57 | read 1048576/1048576 bytes at offset 1048576 | ||
58 | 1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
59 | -read 2097152/2097152 bytes at offset 2097152 | ||
60 | -2 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
61 | read 1048576/1048576 bytes at offset 4194304 | ||
62 | 1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
63 | read 1048576/1048576 bytes at offset 5242880 | ||
64 | @@ -XXX,XX +XXX,XX @@ read 1048576/1048576 bytes at offset 1048576 | ||
65 | read 4194304/4194304 bytes at offset 2097152 | ||
66 | 4 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
67 | |||
68 | -Images are identical. | ||
69 | +read 1048576/1048576 bytes at offset 0 | ||
70 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
71 | +read 1048576/1048576 bytes at offset 1048576 | ||
72 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
73 | +read 3145728/3145728 bytes at offset 3145728 | ||
74 | +3 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
75 | qcow2 file size after I/O: 327680 | ||
76 | |||
77 | === bdrv_co_block_status test for file and offset=0 === | ||
78 | -- | ||
79 | 2.25.3 | ||
80 | |||
81 | diff view generated by jsdifflib |
1 | After processing the option string with the keyval parser, we get a | 1 | From: Ari Sundholm <ari@tuxera.com> |
---|---|---|---|
2 | QDict that contains only strings. This QDict must be fed to a keyval | ||
3 | visitor which converts the strings into the right data types. | ||
4 | 2 | ||
5 | qmp_object_add(), however, uses the normal QObject input visitor, which | 3 | AIO discards regressed as a result of the following commit: |
6 | expects a QDict where all properties already have the QType that matches | 4 | 0dfc7af2 block/file-posix: Optimize for macOS |
7 | the data type required by the QOM object type. | ||
8 | 5 | ||
9 | Change the --object implementation in qemu-storage-daemon so that it | 6 | When trying to run blkdiscard within a Linux guest, the request would |
10 | doesn't call qmp_object_add(), but calls user_creatable_add_dict() | 7 | fail, with some errors in dmesg: |
11 | directly instead and pass it a new keyval boolean that decides which | ||
12 | visitor must be used. | ||
13 | 8 | ||
14 | Reported-by: Coiby Xu <coiby.xu@gmail.com> | 9 | ---- [ snip ] ---- |
10 | [ 4.010070] sd 2:0:0:0: [sda] tag#0 FAILED Result: hostbyte=DID_OK | ||
11 | driverbyte=DRIVER_SENSE | ||
12 | [ 4.011061] sd 2:0:0:0: [sda] tag#0 Sense Key : Aborted Command | ||
13 | [current] | ||
14 | [ 4.011061] sd 2:0:0:0: [sda] tag#0 Add. Sense: I/O process | ||
15 | terminated | ||
16 | [ 4.011061] sd 2:0:0:0: [sda] tag#0 CDB: Unmap/Read sub-channel 42 | ||
17 | 00 00 00 00 00 00 00 18 00 | ||
18 | [ 4.011061] blk_update_request: I/O error, dev sda, sector 0 | ||
19 | ---- [ snip ] ---- | ||
20 | |||
21 | This turns out to be a result of a flaw in changes to the error value | ||
22 | translation logic in handle_aiocb_discard(). The default return value | ||
23 | may be left untranslated in some configurations, and the wrong variable | ||
24 | is used in one translation. | ||
25 | |||
26 | Fix both issues. | ||
27 | |||
28 | Fixes: 0dfc7af2b28 ("block/file-posix: Optimize for macOS") | ||
29 | Cc: qemu-stable@nongnu.org | ||
30 | Signed-off-by: Ari Sundholm <ari@tuxera.com> | ||
31 | Signed-off-by: Emil Karlson <jkarlson@tuxera.com> | ||
32 | Reviewed-by: Akihiko Odaki <akihiko.odaki@gmail.com> | ||
33 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
34 | Message-Id: <20211019110954.4170931-1-ari@tuxera.com> | ||
15 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 35 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
16 | --- | 36 | --- |
17 | include/qom/object_interfaces.h | 6 +++++- | 37 | block/file-posix.c | 4 ++-- |
18 | qemu-storage-daemon.c | 4 +--- | 38 | 1 file changed, 2 insertions(+), 2 deletions(-) |
19 | qom/object_interfaces.c | 8 ++++++-- | ||
20 | qom/qom-qmp-cmds.c | 2 +- | ||
21 | 4 files changed, 13 insertions(+), 7 deletions(-) | ||
22 | 39 | ||
23 | diff --git a/include/qom/object_interfaces.h b/include/qom/object_interfaces.h | 40 | diff --git a/block/file-posix.c b/block/file-posix.c |
24 | index XXXXXXX..XXXXXXX 100644 | 41 | index XXXXXXX..XXXXXXX 100644 |
25 | --- a/include/qom/object_interfaces.h | 42 | --- a/block/file-posix.c |
26 | +++ b/include/qom/object_interfaces.h | 43 | +++ b/block/file-posix.c |
27 | @@ -XXX,XX +XXX,XX @@ Object *user_creatable_add_type(const char *type, const char *id, | 44 | @@ -XXX,XX +XXX,XX @@ static int handle_aiocb_copy_range(void *opaque) |
28 | /** | 45 | static int handle_aiocb_discard(void *opaque) |
29 | * user_creatable_add_dict: | ||
30 | * @qdict: the object definition | ||
31 | + * @keyval: if true, use a keyval visitor for processing @qdict (i.e. | ||
32 | + * assume that all @qdict values are strings); otherwise, use | ||
33 | + * the normal QObject visitor (i.e. assume all @qdict values | ||
34 | + * have the QType expected by the QOM object type) | ||
35 | * @errp: if an error occurs, a pointer to an area to store the error | ||
36 | * | ||
37 | * Create an instance of the user creatable object that is defined by | ||
38 | @@ -XXX,XX +XXX,XX @@ Object *user_creatable_add_type(const char *type, const char *id, | ||
39 | * ID from the key 'id'. The remaining entries in @qdict are used to | ||
40 | * initialize the object properties. | ||
41 | */ | ||
42 | -void user_creatable_add_dict(QDict *qdict, Error **errp); | ||
43 | +void user_creatable_add_dict(QDict *qdict, bool keyval, Error **errp); | ||
44 | |||
45 | /** | ||
46 | * user_creatable_add_opts: | ||
47 | diff --git a/qemu-storage-daemon.c b/qemu-storage-daemon.c | ||
48 | index XXXXXXX..XXXXXXX 100644 | ||
49 | --- a/qemu-storage-daemon.c | ||
50 | +++ b/qemu-storage-daemon.c | ||
51 | @@ -XXX,XX +XXX,XX @@ static void process_options(int argc, char *argv[]) | ||
52 | QemuOpts *opts; | ||
53 | const char *type; | ||
54 | QDict *args; | ||
55 | - QObject *ret_data = NULL; | ||
56 | |||
57 | /* FIXME The keyval parser rejects 'help' arguments, so we must | ||
58 | * unconditionall try QemuOpts first. */ | ||
59 | @@ -XXX,XX +XXX,XX @@ static void process_options(int argc, char *argv[]) | ||
60 | qemu_opts_del(opts); | ||
61 | |||
62 | args = keyval_parse(optarg, "qom-type", &error_fatal); | ||
63 | - qmp_object_add(args, &ret_data, &error_fatal); | ||
64 | + user_creatable_add_dict(args, true, &error_fatal); | ||
65 | qobject_unref(args); | ||
66 | - qobject_unref(ret_data); | ||
67 | break; | ||
68 | } | ||
69 | default: | ||
70 | diff --git a/qom/object_interfaces.c b/qom/object_interfaces.c | ||
71 | index XXXXXXX..XXXXXXX 100644 | ||
72 | --- a/qom/object_interfaces.c | ||
73 | +++ b/qom/object_interfaces.c | ||
74 | @@ -XXX,XX +XXX,XX @@ out: | ||
75 | return obj; | ||
76 | } | ||
77 | |||
78 | -void user_creatable_add_dict(QDict *qdict, Error **errp) | ||
79 | +void user_creatable_add_dict(QDict *qdict, bool keyval, Error **errp) | ||
80 | { | 46 | { |
81 | Visitor *v; | 47 | RawPosixAIOData *aiocb = opaque; |
82 | Object *obj; | 48 | - int ret = -EOPNOTSUPP; |
83 | @@ -XXX,XX +XXX,XX @@ void user_creatable_add_dict(QDict *qdict, Error **errp) | 49 | + int ret = -ENOTSUP; |
84 | } | 50 | BDRVRawState *s = aiocb->bs->opaque; |
85 | qdict_del(qdict, "id"); | 51 | |
86 | 52 | if (!s->has_discard) { | |
87 | - v = qobject_input_visitor_new(QOBJECT(qdict)); | 53 | @@ -XXX,XX +XXX,XX @@ static int handle_aiocb_discard(void *opaque) |
88 | + if (keyval) { | 54 | #ifdef CONFIG_FALLOCATE_PUNCH_HOLE |
89 | + v = qobject_input_visitor_new_keyval(QOBJECT(qdict)); | 55 | ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, |
90 | + } else { | 56 | aiocb->aio_offset, aiocb->aio_nbytes); |
91 | + v = qobject_input_visitor_new(QOBJECT(qdict)); | 57 | - ret = translate_err(-errno); |
92 | + } | 58 | + ret = translate_err(ret); |
93 | obj = user_creatable_add_type(type, id, qdict, v, errp); | 59 | #elif defined(__APPLE__) && (__MACH__) |
94 | visit_free(v); | 60 | fpunchhole_t fpunchhole; |
95 | object_unref(obj); | 61 | fpunchhole.fp_flags = 0; |
96 | diff --git a/qom/qom-qmp-cmds.c b/qom/qom-qmp-cmds.c | ||
97 | index XXXXXXX..XXXXXXX 100644 | ||
98 | --- a/qom/qom-qmp-cmds.c | ||
99 | +++ b/qom/qom-qmp-cmds.c | ||
100 | @@ -XXX,XX +XXX,XX @@ void qmp_object_add(QDict *qdict, QObject **ret_data, Error **errp) | ||
101 | qobject_unref(pdict); | ||
102 | } | ||
103 | |||
104 | - user_creatable_add_dict(qdict, errp); | ||
105 | + user_creatable_add_dict(qdict, false, errp); | ||
106 | } | ||
107 | |||
108 | void qmp_object_del(const char *id, Error **errp) | ||
109 | -- | 62 | -- |
110 | 2.25.3 | 63 | 2.31.1 |
111 | 64 | ||
112 | 65 | diff view generated by jsdifflib |
1 | The BDRV_REQ_ZERO_WRITE is currently implemented in a way that first the | 1 | Using blockdev-snapshot to append a node as an overlay to itself, or to |
---|---|---|---|
2 | image is possibly preallocated and then the zero flag is added to all | 2 | any of its parents, causes crashes. Catch the condition and return an |
3 | clusters. This means that a copy-on-write operation may be needed when | 3 | error for these cases instead. |
4 | writing to these clusters, despite having used preallocation, negating | ||
5 | one of the major benefits of preallocation. | ||
6 | 4 | ||
7 | Instead, try to forward the BDRV_REQ_ZERO_WRITE to the protocol driver, | 5 | Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1824363 |
8 | and if the protocol driver can ensure that the new area reads as zeros, | ||
9 | we can skip setting the zero flag in the qcow2 layer. | ||
10 | |||
11 | Unfortunately, the same approach doesn't work for metadata | ||
12 | preallocation, so we'll still set the zero flag there. | ||
13 | |||
14 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 6 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
15 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 7 | Message-Id: <20211018134714.48438-1-kwolf@redhat.com> |
16 | Message-Id: <20200424142701.67053-1-kwolf@redhat.com> | ||
17 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 8 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> |
18 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 9 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
19 | --- | 10 | --- |
20 | block/qcow2.c | 22 +++++++++++++++++++--- | 11 | block.c | 10 ++++++++++ |
21 | tests/qemu-iotests/274.out | 4 ++-- | 12 | tests/qemu-iotests/085 | 31 ++++++++++++++++++++++++++++++- |
22 | 2 files changed, 21 insertions(+), 5 deletions(-) | 13 | tests/qemu-iotests/085.out | 33 ++++++++++++++++++++++++++++++--- |
14 | 3 files changed, 70 insertions(+), 4 deletions(-) | ||
23 | 15 | ||
24 | diff --git a/block/qcow2.c b/block/qcow2.c | 16 | diff --git a/block.c b/block.c |
25 | index XXXXXXX..XXXXXXX 100644 | 17 | index XXXXXXX..XXXXXXX 100644 |
26 | --- a/block/qcow2.c | 18 | --- a/block.c |
27 | +++ b/block/qcow2.c | 19 | +++ b/block.c |
28 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, | 20 | @@ -XXX,XX +XXX,XX @@ static BlockDriverState *bdrv_open_inherit(const char *filename, |
29 | /* Allocate the data area */ | 21 | BdrvChildRole child_role, |
30 | new_file_size = allocation_start + | 22 | Error **errp); |
31 | nb_new_data_clusters * s->cluster_size; | 23 | |
32 | - /* Image file grows, so @exact does not matter */ | 24 | +static bool bdrv_recurse_has_child(BlockDriverState *bs, |
33 | - ret = bdrv_co_truncate(bs->file, new_file_size, false, prealloc, 0, | 25 | + BlockDriverState *child); |
34 | - errp); | 26 | + |
35 | + /* | 27 | static void bdrv_replace_child_noperm(BdrvChild *child, |
36 | + * Image file grows, so @exact does not matter. | 28 | BlockDriverState *new_bs); |
37 | + * | 29 | static void bdrv_remove_file_or_backing_child(BlockDriverState *bs, |
38 | + * If we need to zero out the new area, try first whether the protocol | 30 | @@ -XXX,XX +XXX,XX @@ static void bdrv_replace_child_noperm(BdrvChild *child, |
39 | + * driver can already take care of this. | 31 | int drain_saldo; |
40 | + */ | 32 | |
41 | + if (flags & BDRV_REQ_ZERO_WRITE) { | 33 | assert(!child->frozen); |
42 | + ret = bdrv_co_truncate(bs->file, new_file_size, false, prealloc, | 34 | + assert(old_bs != new_bs); |
43 | + BDRV_REQ_ZERO_WRITE, NULL); | 35 | |
44 | + if (ret >= 0) { | 36 | if (old_bs && new_bs) { |
45 | + flags &= ~BDRV_REQ_ZERO_WRITE; | 37 | assert(bdrv_get_aio_context(old_bs) == bdrv_get_aio_context(new_bs)); |
46 | + } | 38 | @@ -XXX,XX +XXX,XX @@ static int bdrv_attach_child_noperm(BlockDriverState *parent_bs, |
47 | + } else { | 39 | |
48 | + ret = -1; | 40 | assert(parent_bs->drv); |
49 | + } | 41 | |
50 | + if (ret < 0) { | 42 | + if (bdrv_recurse_has_child(child_bs, parent_bs)) { |
51 | + ret = bdrv_co_truncate(bs->file, new_file_size, false, prealloc, 0, | 43 | + error_setg(errp, "Making '%s' a %s child of '%s' would create a cycle", |
52 | + errp); | 44 | + child_bs->node_name, child_name, parent_bs->node_name); |
53 | + } | 45 | + return -EINVAL; |
54 | if (ret < 0) { | 46 | + } |
55 | error_prepend(errp, "Failed to resize underlying file: "); | 47 | + |
56 | qcow2_free_clusters(bs, allocation_start, | 48 | bdrv_get_cumulative_perm(parent_bs, &perm, &shared_perm); |
57 | diff --git a/tests/qemu-iotests/274.out b/tests/qemu-iotests/274.out | 49 | bdrv_child_perm(parent_bs, child_bs, NULL, child_role, NULL, |
50 | perm, shared_perm, &perm, &shared_perm); | ||
51 | diff --git a/tests/qemu-iotests/085 b/tests/qemu-iotests/085 | ||
52 | index XXXXXXX..XXXXXXX 100755 | ||
53 | --- a/tests/qemu-iotests/085 | ||
54 | +++ b/tests/qemu-iotests/085 | ||
55 | @@ -XXX,XX +XXX,XX @@ do_blockdev_add() | ||
56 | } | ||
57 | |||
58 | # ${1}: unique identifier for the snapshot filename | ||
59 | -add_snapshot_image() | ||
60 | +create_snapshot_image() | ||
61 | { | ||
62 | base_image="${TEST_DIR}/$((${1}-1))-${snapshot_virt0}" | ||
63 | snapshot_file="${TEST_DIR}/${1}-${snapshot_virt0}" | ||
64 | TEST_IMG=$snapshot_file _make_test_img -u -b "${base_image}" -F $IMGFMT "$size" | ||
65 | +} | ||
66 | + | ||
67 | +# ${1}: unique identifier for the snapshot filename | ||
68 | +add_snapshot_image() | ||
69 | +{ | ||
70 | + snapshot_file="${TEST_DIR}/${1}-${snapshot_virt0}" | ||
71 | + create_snapshot_image "$1" | ||
72 | do_blockdev_add "$1" "'backing': null, " "${snapshot_file}" | ||
73 | } | ||
74 | |||
75 | @@ -XXX,XX +XXX,XX @@ _make_test_img -b "${TEST_IMG}.base" -F $IMGFMT "$size" | ||
76 | do_blockdev_add ${SNAPSHOTS} "" "${TEST_IMG}" | ||
77 | blockdev_snapshot ${SNAPSHOTS} error | ||
78 | |||
79 | +echo | ||
80 | +echo === Invalid command - creating loops === | ||
81 | +echo | ||
82 | + | ||
83 | +SNAPSHOTS=$((${SNAPSHOTS}+1)) | ||
84 | +add_snapshot_image ${SNAPSHOTS} | ||
85 | + | ||
86 | +_send_qemu_cmd $h "{ 'execute': 'blockdev-snapshot', | ||
87 | + 'arguments': { 'node':'snap_${SNAPSHOTS}', | ||
88 | + 'overlay':'snap_${SNAPSHOTS}' } | ||
89 | + }" "error" | ||
90 | + | ||
91 | +SNAPSHOTS=$((${SNAPSHOTS}+1)) | ||
92 | +create_snapshot_image ${SNAPSHOTS} | ||
93 | +do_blockdev_add ${SNAPSHOTS} "'backing': 'snap_$((${SNAPSHOTS}-1))', " \ | ||
94 | + "${TEST_DIR}/${SNAPSHOTS}-${snapshot_virt0}" | ||
95 | + | ||
96 | +_send_qemu_cmd $h "{ 'execute': 'blockdev-snapshot', | ||
97 | + 'arguments': { 'node':'snap_${SNAPSHOTS}', | ||
98 | + 'overlay':'snap_$((${SNAPSHOTS}-1))' } | ||
99 | + }" "error" | ||
100 | + | ||
101 | echo | ||
102 | echo === Invalid command - The node does not exist === | ||
103 | echo | ||
104 | diff --git a/tests/qemu-iotests/085.out b/tests/qemu-iotests/085.out | ||
58 | index XXXXXXX..XXXXXXX 100644 | 105 | index XXXXXXX..XXXXXXX 100644 |
59 | --- a/tests/qemu-iotests/274.out | 106 | --- a/tests/qemu-iotests/085.out |
60 | +++ b/tests/qemu-iotests/274.out | 107 | +++ b/tests/qemu-iotests/085.out |
61 | @@ -XXX,XX +XXX,XX @@ read 65536/65536 bytes at offset 9437184 | 108 | @@ -XXX,XX +XXX,XX @@ Formatting 'TEST_DIR/t.IMGFMT', fmt=IMGFMT size=134217728 backing_file=TEST_DIR/ |
62 | 10 MiB (0xa00000) bytes allocated at offset 5 MiB (0x500000) | 109 | 'overlay':'snap_13' } } |
63 | 110 | {"error": {"class": "GenericError", "desc": "The overlay already has a backing image"}} | |
64 | [{ "start": 0, "length": 5242880, "depth": 1, "zero": true, "data": false}, | 111 | |
65 | -{ "start": 5242880, "length": 10485760, "depth": 0, "zero": true, "data": false, "offset": 327680}] | 112 | +=== Invalid command - creating loops === |
66 | +{ "start": 5242880, "length": 10485760, "depth": 0, "zero": false, "data": true, "offset": 327680}] | 113 | + |
67 | 114 | +Formatting 'TEST_DIR/14-snapshot-v0.IMGFMT', fmt=IMGFMT size=134217728 backing_file=TEST_DIR/13-snapshot-v0.IMGFMT backing_fmt=IMGFMT | |
68 | === preallocation=full === | 115 | +{ 'execute': 'blockdev-add', 'arguments': |
69 | Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=16777216 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | 116 | + { 'driver': 'IMGFMT', 'node-name': 'snap_14', 'backing': null, |
70 | @@ -XXX,XX +XXX,XX @@ read 65536/65536 bytes at offset 11534336 | 117 | + 'file': |
71 | 4 MiB (0x400000) bytes allocated at offset 8 MiB (0x800000) | 118 | + { 'driver': 'file', 'filename': 'TEST_DIR/14-snapshot-v0.IMGFMT', |
72 | 119 | + 'node-name': 'file_14' } } } | |
73 | [{ "start": 0, "length": 8388608, "depth": 1, "zero": true, "data": false}, | 120 | +{"return": {}} |
74 | -{ "start": 8388608, "length": 4194304, "depth": 0, "zero": true, "data": false, "offset": 327680}] | 121 | +{ 'execute': 'blockdev-snapshot', |
75 | +{ "start": 8388608, "length": 4194304, "depth": 0, "zero": false, "data": true, "offset": 327680}] | 122 | + 'arguments': { 'node':'snap_14', |
76 | 123 | + 'overlay':'snap_14' } | |
77 | === preallocation=off === | 124 | + } |
78 | Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=393216 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | 125 | +{"error": {"class": "GenericError", "desc": "Making 'snap_14' a backing child of 'snap_14' would create a cycle"}} |
126 | +Formatting 'TEST_DIR/15-snapshot-v0.IMGFMT', fmt=IMGFMT size=134217728 backing_file=TEST_DIR/14-snapshot-v0.IMGFMT backing_fmt=IMGFMT | ||
127 | +{ 'execute': 'blockdev-add', 'arguments': | ||
128 | + { 'driver': 'IMGFMT', 'node-name': 'snap_15', 'backing': 'snap_14', | ||
129 | + 'file': | ||
130 | + { 'driver': 'file', 'filename': 'TEST_DIR/15-snapshot-v0.IMGFMT', | ||
131 | + 'node-name': 'file_15' } } } | ||
132 | +{"return": {}} | ||
133 | +{ 'execute': 'blockdev-snapshot', | ||
134 | + 'arguments': { 'node':'snap_15', | ||
135 | + 'overlay':'snap_14' } | ||
136 | + } | ||
137 | +{"error": {"class": "GenericError", "desc": "Making 'snap_15' a backing child of 'snap_14' would create a cycle"}} | ||
138 | + | ||
139 | === Invalid command - The node does not exist === | ||
140 | |||
141 | { 'execute': 'blockdev-snapshot', | ||
142 | 'arguments': { 'node': 'virtio0', | ||
143 | - 'overlay':'snap_14' } } | ||
144 | -{"error": {"class": "GenericError", "desc": "Cannot find device='snap_14' nor node-name='snap_14'"}} | ||
145 | + 'overlay':'snap_16' } } | ||
146 | +{"error": {"class": "GenericError", "desc": "Cannot find device='snap_16' nor node-name='snap_16'"}} | ||
147 | { 'execute': 'blockdev-snapshot', | ||
148 | 'arguments': { 'node':'nodevice', | ||
149 | - 'overlay':'snap_13' } | ||
150 | + 'overlay':'snap_15' } | ||
151 | } | ||
152 | {"error": {"class": "GenericError", "desc": "Cannot find device='nodevice' nor node-name='nodevice'"}} | ||
153 | *** done | ||
79 | -- | 154 | -- |
80 | 2.25.3 | 155 | 2.31.1 |
81 | 156 | ||
82 | 157 | diff view generated by jsdifflib |
1 | When extending the size of an image that has a backing file larger than | 1 | From: Peter Lieven <pl@kamp.de> |
---|---|---|---|
2 | its old size, make sure that the backing file data doesn't become | ||
3 | visible in the guest, but the added area is properly zeroed out. | ||
4 | 2 | ||
5 | Consider the following scenario where the overlay is shorter than its | 3 | the qemu rbd driver currently lacks support for bdrv_co_block_status. |
6 | backing file: | 4 | This results mainly in incorrect progress during block operations (e.g. |
5 | qemu-img convert with an rbd image as source). | ||
7 | 6 | ||
8 | base.qcow2: AAAAAAAA | 7 | This patch utilizes the rbd_diff_iterate2 call from librbd to detect |
9 | overlay.qcow2: BBBB | 8 | allocated and unallocated (all zero areas). |
10 | 9 | ||
11 | When resizing (extending) overlay.qcow2, the new blocks should not stay | 10 | To avoid querying the ceph OSDs for the answer this is only done if |
12 | unallocated and make the additional As from base.qcow2 visible like | 11 | the image has the fast-diff feature which depends on the object-map and |
13 | before this patch, but zeros should be read. | 12 | exclusive-lock features. In this case it is guaranteed that the information |
13 | is present in memory in the librbd client and thus very fast. | ||
14 | 14 | ||
15 | A similar case happens with the various variants of a commit job when an | 15 | If fast-diff is not available all areas are reported to be allocated |
16 | intermediate file is short (- for unallocated): | 16 | which is the current behaviour if bdrv_co_block_status is not implemented. |
17 | 17 | ||
18 | base.qcow2: A-A-AAAA | 18 | Signed-off-by: Peter Lieven <pl@kamp.de> |
19 | mid.qcow2: BB-B | 19 | Message-Id: <20211012152231.24868-1-pl@kamp.de> |
20 | top.qcow2: C--C--C- | 20 | Reviewed-by: Ilya Dryomov <idryomov@gmail.com> |
21 | |||
22 | After commit top.qcow2 to mid.qcow2, the following happens: | ||
23 | |||
24 | mid.qcow2: CB-C00C0 (correct result) | ||
25 | mid.qcow2: CB-C--C- (before this fix) | ||
26 | |||
27 | Without the fix, blocks that previously read as zeros on top.qcow2 | ||
28 | suddenly turn into A. | ||
29 | |||
30 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | ||
31 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | ||
32 | Message-Id: <20200424125448.63318-8-kwolf@redhat.com> | ||
33 | Reviewed-by: Max Reitz <mreitz@redhat.com> | ||
34 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 21 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
35 | --- | 22 | --- |
36 | block/io.c | 25 +++++++++++++++++++++++++ | 23 | block/rbd.c | 112 ++++++++++++++++++++++++++++++++++++++++++++++++++++ |
37 | 1 file changed, 25 insertions(+) | 24 | 1 file changed, 112 insertions(+) |
38 | 25 | ||
39 | diff --git a/block/io.c b/block/io.c | 26 | diff --git a/block/rbd.c b/block/rbd.c |
40 | index XXXXXXX..XXXXXXX 100644 | 27 | index XXXXXXX..XXXXXXX 100644 |
41 | --- a/block/io.c | 28 | --- a/block/rbd.c |
42 | +++ b/block/io.c | 29 | +++ b/block/rbd.c |
43 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, | 30 | @@ -XXX,XX +XXX,XX @@ typedef struct RBDTask { |
44 | goto out; | 31 | int64_t ret; |
45 | } | 32 | } RBDTask; |
46 | 33 | ||
34 | +typedef struct RBDDiffIterateReq { | ||
35 | + uint64_t offs; | ||
36 | + uint64_t bytes; | ||
37 | + bool exists; | ||
38 | +} RBDDiffIterateReq; | ||
39 | + | ||
40 | static int qemu_rbd_connect(rados_t *cluster, rados_ioctx_t *io_ctx, | ||
41 | BlockdevOptionsRbd *opts, bool cache, | ||
42 | const char *keypairs, const char *secretid, | ||
43 | @@ -XXX,XX +XXX,XX @@ static ImageInfoSpecific *qemu_rbd_get_specific_info(BlockDriverState *bs, | ||
44 | return spec_info; | ||
45 | } | ||
46 | |||
47 | +/* | ||
48 | + * rbd_diff_iterate2 allows to interrupt the exection by returning a negative | ||
49 | + * value in the callback routine. Choose a value that does not conflict with | ||
50 | + * an existing exitcode and return it if we want to prematurely stop the | ||
51 | + * execution because we detected a change in the allocation status. | ||
52 | + */ | ||
53 | +#define QEMU_RBD_EXIT_DIFF_ITERATE2 -9000 | ||
54 | + | ||
55 | +static int qemu_rbd_diff_iterate_cb(uint64_t offs, size_t len, | ||
56 | + int exists, void *opaque) | ||
57 | +{ | ||
58 | + RBDDiffIterateReq *req = opaque; | ||
59 | + | ||
60 | + assert(req->offs + req->bytes <= offs); | ||
47 | + /* | 61 | + /* |
48 | + * If the image has a backing file that is large enough that it would | 62 | + * we do not diff against a snapshot so we should never receive a callback |
49 | + * provide data for the new area, we cannot leave it unallocated because | 63 | + * for a hole. |
50 | + * then the backing file content would become visible. Instead, zero-fill | ||
51 | + * the new area. | ||
52 | + * | ||
53 | + * Note that if the image has a backing file, but was opened without the | ||
54 | + * backing file, taking care of keeping things consistent with that backing | ||
55 | + * file is the user's responsibility. | ||
56 | + */ | 64 | + */ |
57 | + if (new_bytes && bs->backing) { | 65 | + assert(exists); |
58 | + int64_t backing_len; | ||
59 | + | 66 | + |
60 | + backing_len = bdrv_getlength(backing_bs(bs)); | 67 | + if (!req->exists && offs > req->offs) { |
61 | + if (backing_len < 0) { | 68 | + /* |
62 | + ret = backing_len; | 69 | + * we started in an unallocated area and hit the first allocated |
63 | + error_setg_errno(errp, -ret, "Could not get backing file size"); | 70 | + * block. req->bytes must be set to the length of the unallocated area |
64 | + goto out; | 71 | + * before the allocated area. stop further processing. |
65 | + } | 72 | + */ |
66 | + | 73 | + req->bytes = offs - req->offs; |
67 | + if (backing_len > old_size) { | 74 | + return QEMU_RBD_EXIT_DIFF_ITERATE2; |
68 | + flags |= BDRV_REQ_ZERO_WRITE; | ||
69 | + } | ||
70 | + } | 75 | + } |
71 | + | 76 | + |
72 | if (drv->bdrv_co_truncate) { | 77 | + if (req->exists && offs > req->offs + req->bytes) { |
73 | if (flags & ~bs->supported_truncate_flags) { | 78 | + /* |
74 | error_setg(errp, "Block driver does not support requested flags"); | 79 | + * we started in an allocated area and jumped over an unallocated area, |
80 | + * req->bytes contains the length of the allocated area before the | ||
81 | + * unallocated area. stop further processing. | ||
82 | + */ | ||
83 | + return QEMU_RBD_EXIT_DIFF_ITERATE2; | ||
84 | + } | ||
85 | + | ||
86 | + req->bytes += len; | ||
87 | + req->exists = true; | ||
88 | + | ||
89 | + return 0; | ||
90 | +} | ||
91 | + | ||
92 | +static int coroutine_fn qemu_rbd_co_block_status(BlockDriverState *bs, | ||
93 | + bool want_zero, int64_t offset, | ||
94 | + int64_t bytes, int64_t *pnum, | ||
95 | + int64_t *map, | ||
96 | + BlockDriverState **file) | ||
97 | +{ | ||
98 | + BDRVRBDState *s = bs->opaque; | ||
99 | + int status, r; | ||
100 | + RBDDiffIterateReq req = { .offs = offset }; | ||
101 | + uint64_t features, flags; | ||
102 | + | ||
103 | + assert(offset + bytes <= s->image_size); | ||
104 | + | ||
105 | + /* default to all sectors allocated */ | ||
106 | + status = BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID; | ||
107 | + *map = offset; | ||
108 | + *file = bs; | ||
109 | + *pnum = bytes; | ||
110 | + | ||
111 | + /* check if RBD image supports fast-diff */ | ||
112 | + r = rbd_get_features(s->image, &features); | ||
113 | + if (r < 0) { | ||
114 | + return status; | ||
115 | + } | ||
116 | + if (!(features & RBD_FEATURE_FAST_DIFF)) { | ||
117 | + return status; | ||
118 | + } | ||
119 | + | ||
120 | + /* check if RBD fast-diff result is valid */ | ||
121 | + r = rbd_get_flags(s->image, &flags); | ||
122 | + if (r < 0) { | ||
123 | + return status; | ||
124 | + } | ||
125 | + if (flags & RBD_FLAG_FAST_DIFF_INVALID) { | ||
126 | + return status; | ||
127 | + } | ||
128 | + | ||
129 | + r = rbd_diff_iterate2(s->image, NULL, offset, bytes, true, true, | ||
130 | + qemu_rbd_diff_iterate_cb, &req); | ||
131 | + if (r < 0 && r != QEMU_RBD_EXIT_DIFF_ITERATE2) { | ||
132 | + return status; | ||
133 | + } | ||
134 | + assert(req.bytes <= bytes); | ||
135 | + if (!req.exists) { | ||
136 | + if (r == 0) { | ||
137 | + /* | ||
138 | + * rbd_diff_iterate2 does not invoke callbacks for unallocated | ||
139 | + * areas. This here catches the case where no callback was | ||
140 | + * invoked at all (req.bytes == 0). | ||
141 | + */ | ||
142 | + assert(req.bytes == 0); | ||
143 | + req.bytes = bytes; | ||
144 | + } | ||
145 | + status = BDRV_BLOCK_ZERO | BDRV_BLOCK_OFFSET_VALID; | ||
146 | + } | ||
147 | + | ||
148 | + *pnum = req.bytes; | ||
149 | + return status; | ||
150 | +} | ||
151 | + | ||
152 | static int64_t qemu_rbd_getlength(BlockDriverState *bs) | ||
153 | { | ||
154 | BDRVRBDState *s = bs->opaque; | ||
155 | @@ -XXX,XX +XXX,XX @@ static BlockDriver bdrv_rbd = { | ||
156 | #ifdef LIBRBD_SUPPORTS_WRITE_ZEROES | ||
157 | .bdrv_co_pwrite_zeroes = qemu_rbd_co_pwrite_zeroes, | ||
158 | #endif | ||
159 | + .bdrv_co_block_status = qemu_rbd_co_block_status, | ||
160 | |||
161 | .bdrv_snapshot_create = qemu_rbd_snap_create, | ||
162 | .bdrv_snapshot_delete = qemu_rbd_snap_remove, | ||
75 | -- | 163 | -- |
76 | 2.25.3 | 164 | 2.31.1 |
77 | 165 | ||
78 | 166 | diff view generated by jsdifflib |
1 | The QMP handler qmp_object_add() and the implementation of --object in | 1 | From: Samuel Thibault <samuel.thibault@ens-lyon.org> |
---|---|---|---|
2 | qemu-storage-daemon can share most of the code. Currently, | ||
3 | qemu-storage-daemon calls qmp_object_add(), but this is not correct | ||
4 | because different visitors need to be used. | ||
5 | 2 | ||
6 | As a first step towards a fix, make qmp_object_add() a wrapper around a | 3 | The LBA28 capacity (at offsets 60/61 of identification) is supposed to |
7 | new function user_creatable_add_dict() that can get an additional | 4 | express the maximum size supported by LBA28 commands. If the device is |
8 | parameter. The handling of "props" is only required for compatibility | 5 | larger than this, we have to cap it to 2^28-1. |
9 | and not required for the qemu-storage-daemon command line, so it stays | ||
10 | in qmp_object_add(). | ||
11 | 6 | ||
7 | At least NetBSD happens to be using this value to determine whether to use | ||
8 | LBA28 or LBA48 for its commands, using LBA28 for sectors that don't need | ||
9 | LBA48. This commit thus fixes NetBSD access to disks larger than 128GiB. | ||
10 | |||
11 | Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org> | ||
12 | Message-Id: <20210824104344.3878849-1-samuel.thibault@ens-lyon.org> | ||
12 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 13 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
13 | --- | 14 | --- |
14 | include/qom/object_interfaces.h | 12 ++++++++++++ | 15 | hw/ide/core.c | 8 ++++++-- |
15 | qom/object_interfaces.c | 27 +++++++++++++++++++++++++++ | 16 | 1 file changed, 6 insertions(+), 2 deletions(-) |
16 | qom/qom-qmp-cmds.c | 24 +----------------------- | ||
17 | 3 files changed, 40 insertions(+), 23 deletions(-) | ||
18 | 17 | ||
19 | diff --git a/include/qom/object_interfaces.h b/include/qom/object_interfaces.h | 18 | diff --git a/hw/ide/core.c b/hw/ide/core.c |
20 | index XXXXXXX..XXXXXXX 100644 | 19 | index XXXXXXX..XXXXXXX 100644 |
21 | --- a/include/qom/object_interfaces.h | 20 | --- a/hw/ide/core.c |
22 | +++ b/include/qom/object_interfaces.h | 21 | +++ b/hw/ide/core.c |
23 | @@ -XXX,XX +XXX,XX @@ Object *user_creatable_add_type(const char *type, const char *id, | 22 | @@ -XXX,XX +XXX,XX @@ static void put_le16(uint16_t *p, unsigned int v) |
24 | const QDict *qdict, | 23 | static void ide_identify_size(IDEState *s) |
25 | Visitor *v, Error **errp); | 24 | { |
26 | 25 | uint16_t *p = (uint16_t *)s->identify_data; | |
27 | +/** | 26 | - put_le16(p + 60, s->nb_sectors); |
28 | + * user_creatable_add_dict: | 27 | - put_le16(p + 61, s->nb_sectors >> 16); |
29 | + * @qdict: the object definition | 28 | + int64_t nb_sectors_lba28 = s->nb_sectors; |
30 | + * @errp: if an error occurs, a pointer to an area to store the error | 29 | + if (nb_sectors_lba28 >= 1 << 28) { |
31 | + * | 30 | + nb_sectors_lba28 = (1 << 28) - 1; |
32 | + * Create an instance of the user creatable object that is defined by | ||
33 | + * @qdict. The object type is taken from the QDict key 'qom-type', its | ||
34 | + * ID from the key 'id'. The remaining entries in @qdict are used to | ||
35 | + * initialize the object properties. | ||
36 | + */ | ||
37 | +void user_creatable_add_dict(QDict *qdict, Error **errp); | ||
38 | + | ||
39 | /** | ||
40 | * user_creatable_add_opts: | ||
41 | * @opts: the object definition | ||
42 | diff --git a/qom/object_interfaces.c b/qom/object_interfaces.c | ||
43 | index XXXXXXX..XXXXXXX 100644 | ||
44 | --- a/qom/object_interfaces.c | ||
45 | +++ b/qom/object_interfaces.c | ||
46 | @@ -XXX,XX +XXX,XX @@ | ||
47 | #include "qapi/qmp/qerror.h" | ||
48 | #include "qapi/qmp/qjson.h" | ||
49 | #include "qapi/qmp/qstring.h" | ||
50 | +#include "qapi/qobject-input-visitor.h" | ||
51 | #include "qom/object_interfaces.h" | ||
52 | #include "qemu/help_option.h" | ||
53 | #include "qemu/module.h" | ||
54 | @@ -XXX,XX +XXX,XX @@ out: | ||
55 | return obj; | ||
56 | } | ||
57 | |||
58 | +void user_creatable_add_dict(QDict *qdict, Error **errp) | ||
59 | +{ | ||
60 | + Visitor *v; | ||
61 | + Object *obj; | ||
62 | + g_autofree char *type = NULL; | ||
63 | + g_autofree char *id = NULL; | ||
64 | + | ||
65 | + type = g_strdup(qdict_get_try_str(qdict, "qom-type")); | ||
66 | + if (!type) { | ||
67 | + error_setg(errp, QERR_MISSING_PARAMETER, "qom-type"); | ||
68 | + return; | ||
69 | + } | 31 | + } |
70 | + qdict_del(qdict, "qom-type"); | 32 | + put_le16(p + 60, nb_sectors_lba28); |
71 | + | 33 | + put_le16(p + 61, nb_sectors_lba28 >> 16); |
72 | + id = g_strdup(qdict_get_try_str(qdict, "id")); | 34 | put_le16(p + 100, s->nb_sectors); |
73 | + if (!id) { | 35 | put_le16(p + 101, s->nb_sectors >> 16); |
74 | + error_setg(errp, QERR_MISSING_PARAMETER, "id"); | 36 | put_le16(p + 102, s->nb_sectors >> 32); |
75 | + return; | ||
76 | + } | ||
77 | + qdict_del(qdict, "id"); | ||
78 | + | ||
79 | + v = qobject_input_visitor_new(QOBJECT(qdict)); | ||
80 | + obj = user_creatable_add_type(type, id, qdict, v, errp); | ||
81 | + visit_free(v); | ||
82 | + object_unref(obj); | ||
83 | +} | ||
84 | |||
85 | Object *user_creatable_add_opts(QemuOpts *opts, Error **errp) | ||
86 | { | ||
87 | diff --git a/qom/qom-qmp-cmds.c b/qom/qom-qmp-cmds.c | ||
88 | index XXXXXXX..XXXXXXX 100644 | ||
89 | --- a/qom/qom-qmp-cmds.c | ||
90 | +++ b/qom/qom-qmp-cmds.c | ||
91 | @@ -XXX,XX +XXX,XX @@ | ||
92 | #include "qapi/qapi-commands-qom.h" | ||
93 | #include "qapi/qmp/qdict.h" | ||
94 | #include "qapi/qmp/qerror.h" | ||
95 | -#include "qapi/qobject-input-visitor.h" | ||
96 | #include "qemu/cutils.h" | ||
97 | #include "qom/object_interfaces.h" | ||
98 | #include "qom/qom-qobject.h" | ||
99 | @@ -XXX,XX +XXX,XX @@ void qmp_object_add(QDict *qdict, QObject **ret_data, Error **errp) | ||
100 | { | ||
101 | QObject *props; | ||
102 | QDict *pdict; | ||
103 | - Visitor *v; | ||
104 | - Object *obj; | ||
105 | - g_autofree char *type = NULL; | ||
106 | - g_autofree char *id = NULL; | ||
107 | - | ||
108 | - type = g_strdup(qdict_get_try_str(qdict, "qom-type")); | ||
109 | - if (!type) { | ||
110 | - error_setg(errp, QERR_MISSING_PARAMETER, "qom-type"); | ||
111 | - return; | ||
112 | - } | ||
113 | - qdict_del(qdict, "qom-type"); | ||
114 | - | ||
115 | - id = g_strdup(qdict_get_try_str(qdict, "id")); | ||
116 | - if (!id) { | ||
117 | - error_setg(errp, QERR_MISSING_PARAMETER, "id"); | ||
118 | - return; | ||
119 | - } | ||
120 | - qdict_del(qdict, "id"); | ||
121 | |||
122 | props = qdict_get(qdict, "props"); | ||
123 | if (props) { | ||
124 | @@ -XXX,XX +XXX,XX @@ void qmp_object_add(QDict *qdict, QObject **ret_data, Error **errp) | ||
125 | qobject_unref(pdict); | ||
126 | } | ||
127 | |||
128 | - v = qobject_input_visitor_new(QOBJECT(qdict)); | ||
129 | - obj = user_creatable_add_type(type, id, qdict, v, errp); | ||
130 | - visit_free(v); | ||
131 | - object_unref(obj); | ||
132 | + user_creatable_add_dict(qdict, errp); | ||
133 | } | ||
134 | |||
135 | void qmp_object_del(const char *id, Error **errp) | ||
136 | -- | 37 | -- |
137 | 2.25.3 | 38 | 2.31.1 |
138 | 39 | ||
139 | 40 | diff view generated by jsdifflib |
1 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 1 | From: Fabrice Fontaine <fontaine.fabrice@gmail.com> |
---|---|---|---|
2 | Message-Id: <20200424125448.63318-10-kwolf@redhat.com> | 2 | |
3 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 3 | Include linux/falloc.h if CONFIG_FALLOCATE_ZERO_RANGE is defined to fix |
4 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 4 | https://gitlab.com/qemu-project/qemu/-/commit/50482fda98bd62e072c30b7ea73c985c4e9d9bbb |
5 | and avoid the following build failure on musl: | ||
6 | |||
7 | ../block/export/fuse.c: In function 'fuse_fallocate': | ||
8 | ../block/export/fuse.c:643:21: error: 'FALLOC_FL_ZERO_RANGE' undeclared (first use in this function) | ||
9 | 643 | else if (mode & FALLOC_FL_ZERO_RANGE) { | ||
10 | | ^~~~~~~~~~~~~~~~~~~~ | ||
11 | |||
12 | Fixes: | ||
13 | - http://autobuild.buildroot.org/results/be24433a429fda681fb66698160132c1c99bc53b | ||
14 | |||
15 | Fixes: 50482fda98b ("block/export/fuse.c: fix musl build") | ||
16 | Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com> | ||
17 | Message-Id: <20211022095209.1319671-1-fontaine.fabrice@gmail.com> | ||
18 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
5 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 19 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
6 | --- | 20 | --- |
7 | tests/qemu-iotests/274 | 155 +++++++++++++++++++++ | 21 | block/export/fuse.c | 4 ++++ |
8 | tests/qemu-iotests/274.out | 268 +++++++++++++++++++++++++++++++++++++ | 22 | 1 file changed, 4 insertions(+) |
9 | tests/qemu-iotests/group | 1 + | ||
10 | 3 files changed, 424 insertions(+) | ||
11 | create mode 100755 tests/qemu-iotests/274 | ||
12 | create mode 100644 tests/qemu-iotests/274.out | ||
13 | 23 | ||
14 | diff --git a/tests/qemu-iotests/274 b/tests/qemu-iotests/274 | 24 | diff --git a/block/export/fuse.c b/block/export/fuse.c |
15 | new file mode 100755 | 25 | index XXXXXXX..XXXXXXX 100644 |
16 | index XXXXXXX..XXXXXXX | 26 | --- a/block/export/fuse.c |
17 | --- /dev/null | 27 | +++ b/block/export/fuse.c |
18 | +++ b/tests/qemu-iotests/274 | ||
19 | @@ -XXX,XX +XXX,XX @@ | 28 | @@ -XXX,XX +XXX,XX @@ |
20 | +#!/usr/bin/env python3 | 29 | #include <fuse.h> |
21 | +# | 30 | #include <fuse_lowlevel.h> |
22 | +# Copyright (C) 2019 Red Hat, Inc. | 31 | |
23 | +# | 32 | +#if defined(CONFIG_FALLOCATE_ZERO_RANGE) |
24 | +# This program is free software; you can redistribute it and/or modify | 33 | +#include <linux/falloc.h> |
25 | +# it under the terms of the GNU General Public License as published by | 34 | +#endif |
26 | +# the Free Software Foundation; either version 2 of the License, or | ||
27 | +# (at your option) any later version. | ||
28 | +# | ||
29 | +# This program is distributed in the hope that it will be useful, | ||
30 | +# but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
31 | +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
32 | +# GNU General Public License for more details. | ||
33 | +# | ||
34 | +# You should have received a copy of the GNU General Public License | ||
35 | +# along with this program. If not, see <http://www.gnu.org/licenses/>. | ||
36 | +# | ||
37 | +# Creator/Owner: Kevin Wolf <kwolf@redhat.com> | ||
38 | +# | ||
39 | +# Some tests for short backing files and short overlays | ||
40 | + | 35 | + |
41 | +import iotests | 36 | #ifdef __linux__ |
42 | + | 37 | #include <linux/fs.h> |
43 | +iotests.verify_image_format(supported_fmts=['qcow2']) | 38 | #endif |
44 | +iotests.verify_platform(['linux']) | ||
45 | + | ||
46 | +size_short = 1 * 1024 * 1024 | ||
47 | +size_long = 2 * 1024 * 1024 | ||
48 | +size_diff = size_long - size_short | ||
49 | + | ||
50 | +def create_chain() -> None: | ||
51 | + iotests.qemu_img_log('create', '-f', iotests.imgfmt, base, | ||
52 | + str(size_long)) | ||
53 | + iotests.qemu_img_log('create', '-f', iotests.imgfmt, '-b', base, mid, | ||
54 | + str(size_short)) | ||
55 | + iotests.qemu_img_log('create', '-f', iotests.imgfmt, '-b', mid, top, | ||
56 | + str(size_long)) | ||
57 | + | ||
58 | + iotests.qemu_io_log('-c', 'write -P 1 0 %d' % size_long, base) | ||
59 | + | ||
60 | +def create_vm() -> iotests.VM: | ||
61 | + vm = iotests.VM() | ||
62 | + vm.add_blockdev('file,filename=%s,node-name=base-file' % base) | ||
63 | + vm.add_blockdev('%s,file=base-file,node-name=base' % iotests.imgfmt) | ||
64 | + vm.add_blockdev('file,filename=%s,node-name=mid-file' % mid) | ||
65 | + vm.add_blockdev('%s,file=mid-file,node-name=mid,backing=base' | ||
66 | + % iotests.imgfmt) | ||
67 | + vm.add_drive(top, 'backing=mid,node-name=top') | ||
68 | + return vm | ||
69 | + | ||
70 | +with iotests.FilePath('base') as base, \ | ||
71 | + iotests.FilePath('mid') as mid, \ | ||
72 | + iotests.FilePath('top') as top: | ||
73 | + | ||
74 | + iotests.log('== Commit tests ==') | ||
75 | + | ||
76 | + create_chain() | ||
77 | + | ||
78 | + iotests.log('=== Check visible data ===') | ||
79 | + | ||
80 | + iotests.qemu_io_log('-c', 'read -P 1 0 %d' % size_short, top) | ||
81 | + iotests.qemu_io_log('-c', 'read -P 0 %d %d' % (size_short, size_diff), top) | ||
82 | + | ||
83 | + iotests.log('=== Checking allocation status ===') | ||
84 | + | ||
85 | + iotests.qemu_io_log('-c', 'alloc 0 %d' % size_short, | ||
86 | + '-c', 'alloc %d %d' % (size_short, size_diff), | ||
87 | + base) | ||
88 | + | ||
89 | + iotests.qemu_io_log('-c', 'alloc 0 %d' % size_short, | ||
90 | + '-c', 'alloc %d %d' % (size_short, size_diff), | ||
91 | + mid) | ||
92 | + | ||
93 | + iotests.qemu_io_log('-c', 'alloc 0 %d' % size_short, | ||
94 | + '-c', 'alloc %d %d' % (size_short, size_diff), | ||
95 | + top) | ||
96 | + | ||
97 | + iotests.log('=== Checking map ===') | ||
98 | + | ||
99 | + iotests.qemu_img_log('map', '--output=json', base) | ||
100 | + iotests.qemu_img_log('map', '--output=human', base) | ||
101 | + iotests.qemu_img_log('map', '--output=json', mid) | ||
102 | + iotests.qemu_img_log('map', '--output=human', mid) | ||
103 | + iotests.qemu_img_log('map', '--output=json', top) | ||
104 | + iotests.qemu_img_log('map', '--output=human', top) | ||
105 | + | ||
106 | + iotests.log('=== Testing qemu-img commit (top -> mid) ===') | ||
107 | + | ||
108 | + iotests.qemu_img_log('commit', top) | ||
109 | + iotests.img_info_log(mid) | ||
110 | + iotests.qemu_io_log('-c', 'read -P 1 0 %d' % size_short, mid) | ||
111 | + iotests.qemu_io_log('-c', 'read -P 0 %d %d' % (size_short, size_diff), mid) | ||
112 | + | ||
113 | + iotests.log('=== Testing HMP commit (top -> mid) ===') | ||
114 | + | ||
115 | + create_chain() | ||
116 | + with create_vm() as vm: | ||
117 | + vm.launch() | ||
118 | + vm.qmp_log('human-monitor-command', command_line='commit drive0') | ||
119 | + | ||
120 | + iotests.img_info_log(mid) | ||
121 | + iotests.qemu_io_log('-c', 'read -P 1 0 %d' % size_short, mid) | ||
122 | + iotests.qemu_io_log('-c', 'read -P 0 %d %d' % (size_short, size_diff), mid) | ||
123 | + | ||
124 | + iotests.log('=== Testing QMP active commit (top -> mid) ===') | ||
125 | + | ||
126 | + create_chain() | ||
127 | + with create_vm() as vm: | ||
128 | + vm.launch() | ||
129 | + vm.qmp_log('block-commit', device='top', base_node='mid', | ||
130 | + job_id='job0', auto_dismiss=False) | ||
131 | + vm.run_job('job0', wait=5) | ||
132 | + | ||
133 | + iotests.img_info_log(mid) | ||
134 | + iotests.qemu_io_log('-c', 'read -P 1 0 %d' % size_short, mid) | ||
135 | + iotests.qemu_io_log('-c', 'read -P 0 %d %d' % (size_short, size_diff), mid) | ||
136 | + | ||
137 | + | ||
138 | + iotests.log('== Resize tests ==') | ||
139 | + | ||
140 | + # Use different sizes for different allocation modes: | ||
141 | + # | ||
142 | + # We want to have at least one test where 32 bit truncation in the size of | ||
143 | + # the overlapping area becomes visible. This is covered by the | ||
144 | + # prealloc='off' case (1G to 6G is an overlap of 5G). | ||
145 | + # | ||
146 | + # However, we can only do this for modes that don't preallocate data | ||
147 | + # because otherwise we might run out of space on the test host. | ||
148 | + # | ||
149 | + # We also want to test some unaligned combinations. | ||
150 | + for (prealloc, base_size, top_size_old, top_size_new, off) in [ | ||
151 | + ('off', '6G', '1G', '8G', '5G'), | ||
152 | + ('metadata', '32G', '30G', '33G', '31G'), | ||
153 | + ('falloc', '10M', '5M', '15M', '9M'), | ||
154 | + ('full', '16M', '8M', '12M', '11M'), | ||
155 | + ('off', '384k', '253k', '512k', '253k'), | ||
156 | + ('off', '400k', '256k', '512k', '336k'), | ||
157 | + ('off', '512k', '256k', '500k', '436k')]: | ||
158 | + | ||
159 | + iotests.log('=== preallocation=%s ===' % prealloc) | ||
160 | + iotests.qemu_img_log('create', '-f', iotests.imgfmt, base, base_size) | ||
161 | + iotests.qemu_img_log('create', '-f', iotests.imgfmt, '-b', base, top, | ||
162 | + top_size_old) | ||
163 | + iotests.qemu_io_log('-c', 'write -P 1 %s 64k' % off, base) | ||
164 | + | ||
165 | + # After this, top_size_old to base_size should be allocated/zeroed. | ||
166 | + # | ||
167 | + # In theory, leaving base_size to top_size_new unallocated would be | ||
168 | + # correct, but in practice, if we zero out anything, we zero out | ||
169 | + # everything up to top_size_new. | ||
170 | + iotests.qemu_img_log('resize', '-f', iotests.imgfmt, | ||
171 | + '--preallocation', prealloc, top, top_size_new) | ||
172 | + iotests.qemu_io_log('-c', 'read -P 0 %s 64k' % off, top) | ||
173 | + iotests.qemu_io_log('-c', 'map', top) | ||
174 | + iotests.qemu_img_log('map', '--output=json', top) | ||
175 | diff --git a/tests/qemu-iotests/274.out b/tests/qemu-iotests/274.out | ||
176 | new file mode 100644 | ||
177 | index XXXXXXX..XXXXXXX | ||
178 | --- /dev/null | ||
179 | +++ b/tests/qemu-iotests/274.out | ||
180 | @@ -XXX,XX +XXX,XX @@ | ||
181 | +== Commit tests == | ||
182 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=2097152 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
183 | + | ||
184 | +Formatting 'TEST_DIR/PID-mid', fmt=qcow2 size=1048576 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
185 | + | ||
186 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=2097152 backing_file=TEST_DIR/PID-mid cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
187 | + | ||
188 | +wrote 2097152/2097152 bytes at offset 0 | ||
189 | +2 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
190 | + | ||
191 | +=== Check visible data === | ||
192 | +read 1048576/1048576 bytes at offset 0 | ||
193 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
194 | + | ||
195 | +read 1048576/1048576 bytes at offset 1048576 | ||
196 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
197 | + | ||
198 | +=== Checking allocation status === | ||
199 | +1048576/1048576 bytes allocated at offset 0 bytes | ||
200 | +1048576/1048576 bytes allocated at offset 1 MiB | ||
201 | + | ||
202 | +0/1048576 bytes allocated at offset 0 bytes | ||
203 | +0/0 bytes allocated at offset 1 MiB | ||
204 | + | ||
205 | +0/1048576 bytes allocated at offset 0 bytes | ||
206 | +0/1048576 bytes allocated at offset 1 MiB | ||
207 | + | ||
208 | +=== Checking map === | ||
209 | +[{ "start": 0, "length": 2097152, "depth": 0, "zero": false, "data": true, "offset": 327680}] | ||
210 | + | ||
211 | +Offset Length Mapped to File | ||
212 | +0 0x200000 0x50000 TEST_DIR/PID-base | ||
213 | + | ||
214 | +[{ "start": 0, "length": 1048576, "depth": 1, "zero": false, "data": true, "offset": 327680}] | ||
215 | + | ||
216 | +Offset Length Mapped to File | ||
217 | +0 0x100000 0x50000 TEST_DIR/PID-base | ||
218 | + | ||
219 | +[{ "start": 0, "length": 1048576, "depth": 2, "zero": false, "data": true, "offset": 327680}, | ||
220 | +{ "start": 1048576, "length": 1048576, "depth": 0, "zero": true, "data": false}] | ||
221 | + | ||
222 | +Offset Length Mapped to File | ||
223 | +0 0x100000 0x50000 TEST_DIR/PID-base | ||
224 | + | ||
225 | +=== Testing qemu-img commit (top -> mid) === | ||
226 | +Image committed. | ||
227 | + | ||
228 | +image: TEST_IMG | ||
229 | +file format: IMGFMT | ||
230 | +virtual size: 2 MiB (2097152 bytes) | ||
231 | +cluster_size: 65536 | ||
232 | +backing file: TEST_DIR/PID-base | ||
233 | +Format specific information: | ||
234 | + compat: 1.1 | ||
235 | + lazy refcounts: false | ||
236 | + refcount bits: 16 | ||
237 | + corrupt: false | ||
238 | + | ||
239 | +read 1048576/1048576 bytes at offset 0 | ||
240 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
241 | + | ||
242 | +read 1048576/1048576 bytes at offset 1048576 | ||
243 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
244 | + | ||
245 | +=== Testing HMP commit (top -> mid) === | ||
246 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=2097152 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
247 | + | ||
248 | +Formatting 'TEST_DIR/PID-mid', fmt=qcow2 size=1048576 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
249 | + | ||
250 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=2097152 backing_file=TEST_DIR/PID-mid cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
251 | + | ||
252 | +wrote 2097152/2097152 bytes at offset 0 | ||
253 | +2 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
254 | + | ||
255 | +{"execute": "human-monitor-command", "arguments": {"command-line": "commit drive0"}} | ||
256 | +{"return": ""} | ||
257 | +image: TEST_IMG | ||
258 | +file format: IMGFMT | ||
259 | +virtual size: 2 MiB (2097152 bytes) | ||
260 | +cluster_size: 65536 | ||
261 | +backing file: TEST_DIR/PID-base | ||
262 | +Format specific information: | ||
263 | + compat: 1.1 | ||
264 | + lazy refcounts: false | ||
265 | + refcount bits: 16 | ||
266 | + corrupt: false | ||
267 | + | ||
268 | +read 1048576/1048576 bytes at offset 0 | ||
269 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
270 | + | ||
271 | +read 1048576/1048576 bytes at offset 1048576 | ||
272 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
273 | + | ||
274 | +=== Testing QMP active commit (top -> mid) === | ||
275 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=2097152 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
276 | + | ||
277 | +Formatting 'TEST_DIR/PID-mid', fmt=qcow2 size=1048576 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
278 | + | ||
279 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=2097152 backing_file=TEST_DIR/PID-mid cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
280 | + | ||
281 | +wrote 2097152/2097152 bytes at offset 0 | ||
282 | +2 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
283 | + | ||
284 | +{"execute": "block-commit", "arguments": {"auto-dismiss": false, "base-node": "mid", "device": "top", "job-id": "job0"}} | ||
285 | +{"return": {}} | ||
286 | +{"execute": "job-complete", "arguments": {"id": "job0"}} | ||
287 | +{"return": {}} | ||
288 | +{"data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}, "event": "BLOCK_JOB_READY", "timestamp": {"microseconds": "USECS", "seconds": "SECS"}} | ||
289 | +{"data": {"device": "job0", "len": 0, "offset": 0, "speed": 0, "type": "commit"}, "event": "BLOCK_JOB_COMPLETED", "timestamp": {"microseconds": "USECS", "seconds": "SECS"}} | ||
290 | +{"execute": "job-dismiss", "arguments": {"id": "job0"}} | ||
291 | +{"return": {}} | ||
292 | +image: TEST_IMG | ||
293 | +file format: IMGFMT | ||
294 | +virtual size: 2 MiB (2097152 bytes) | ||
295 | +cluster_size: 65536 | ||
296 | +backing file: TEST_DIR/PID-base | ||
297 | +Format specific information: | ||
298 | + compat: 1.1 | ||
299 | + lazy refcounts: false | ||
300 | + refcount bits: 16 | ||
301 | + corrupt: false | ||
302 | + | ||
303 | +read 1048576/1048576 bytes at offset 0 | ||
304 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
305 | + | ||
306 | +read 1048576/1048576 bytes at offset 1048576 | ||
307 | +1 MiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
308 | + | ||
309 | +== Resize tests == | ||
310 | +=== preallocation=off === | ||
311 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=6442450944 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
312 | + | ||
313 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=1073741824 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
314 | + | ||
315 | +wrote 65536/65536 bytes at offset 5368709120 | ||
316 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
317 | + | ||
318 | +Image resized. | ||
319 | + | ||
320 | +read 65536/65536 bytes at offset 5368709120 | ||
321 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
322 | + | ||
323 | +1 GiB (0x40000000) bytes not allocated at offset 0 bytes (0x0) | ||
324 | +7 GiB (0x1c0000000) bytes allocated at offset 1 GiB (0x40000000) | ||
325 | + | ||
326 | +[{ "start": 0, "length": 1073741824, "depth": 1, "zero": true, "data": false}, | ||
327 | +{ "start": 1073741824, "length": 7516192768, "depth": 0, "zero": true, "data": false}] | ||
328 | + | ||
329 | +=== preallocation=metadata === | ||
330 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=34359738368 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
331 | + | ||
332 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=32212254720 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
333 | + | ||
334 | +wrote 65536/65536 bytes at offset 33285996544 | ||
335 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
336 | + | ||
337 | +Image resized. | ||
338 | + | ||
339 | +read 65536/65536 bytes at offset 33285996544 | ||
340 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
341 | + | ||
342 | +30 GiB (0x780000000) bytes not allocated at offset 0 bytes (0x0) | ||
343 | +3 GiB (0xc0000000) bytes allocated at offset 30 GiB (0x780000000) | ||
344 | + | ||
345 | +[{ "start": 0, "length": 32212254720, "depth": 1, "zero": true, "data": false}, | ||
346 | +{ "start": 32212254720, "length": 536870912, "depth": 0, "zero": true, "data": false, "offset": 327680}, | ||
347 | +{ "start": 32749125632, "length": 536870912, "depth": 0, "zero": true, "data": false, "offset": 537264128}, | ||
348 | +{ "start": 33285996544, "length": 536870912, "depth": 0, "zero": true, "data": false, "offset": 1074200576}, | ||
349 | +{ "start": 33822867456, "length": 536870912, "depth": 0, "zero": true, "data": false, "offset": 1611137024}, | ||
350 | +{ "start": 34359738368, "length": 536870912, "depth": 0, "zero": true, "data": false, "offset": 2148139008}, | ||
351 | +{ "start": 34896609280, "length": 536870912, "depth": 0, "zero": true, "data": false, "offset": 2685075456}] | ||
352 | + | ||
353 | +=== preallocation=falloc === | ||
354 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=10485760 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
355 | + | ||
356 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=5242880 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
357 | + | ||
358 | +wrote 65536/65536 bytes at offset 9437184 | ||
359 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
360 | + | ||
361 | +Image resized. | ||
362 | + | ||
363 | +read 65536/65536 bytes at offset 9437184 | ||
364 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
365 | + | ||
366 | +5 MiB (0x500000) bytes not allocated at offset 0 bytes (0x0) | ||
367 | +10 MiB (0xa00000) bytes allocated at offset 5 MiB (0x500000) | ||
368 | + | ||
369 | +[{ "start": 0, "length": 5242880, "depth": 1, "zero": true, "data": false}, | ||
370 | +{ "start": 5242880, "length": 10485760, "depth": 0, "zero": true, "data": false, "offset": 327680}] | ||
371 | + | ||
372 | +=== preallocation=full === | ||
373 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=16777216 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
374 | + | ||
375 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=8388608 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
376 | + | ||
377 | +wrote 65536/65536 bytes at offset 11534336 | ||
378 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
379 | + | ||
380 | +Image resized. | ||
381 | + | ||
382 | +read 65536/65536 bytes at offset 11534336 | ||
383 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
384 | + | ||
385 | +8 MiB (0x800000) bytes not allocated at offset 0 bytes (0x0) | ||
386 | +4 MiB (0x400000) bytes allocated at offset 8 MiB (0x800000) | ||
387 | + | ||
388 | +[{ "start": 0, "length": 8388608, "depth": 1, "zero": true, "data": false}, | ||
389 | +{ "start": 8388608, "length": 4194304, "depth": 0, "zero": true, "data": false, "offset": 327680}] | ||
390 | + | ||
391 | +=== preallocation=off === | ||
392 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=393216 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
393 | + | ||
394 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=259072 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
395 | + | ||
396 | +wrote 65536/65536 bytes at offset 259072 | ||
397 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
398 | + | ||
399 | +Image resized. | ||
400 | + | ||
401 | +read 65536/65536 bytes at offset 259072 | ||
402 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
403 | + | ||
404 | +192 KiB (0x30000) bytes not allocated at offset 0 bytes (0x0) | ||
405 | +320 KiB (0x50000) bytes allocated at offset 192 KiB (0x30000) | ||
406 | + | ||
407 | +[{ "start": 0, "length": 196608, "depth": 1, "zero": true, "data": false}, | ||
408 | +{ "start": 196608, "length": 65536, "depth": 0, "zero": false, "data": true, "offset": 327680}, | ||
409 | +{ "start": 262144, "length": 262144, "depth": 0, "zero": true, "data": false}] | ||
410 | + | ||
411 | +=== preallocation=off === | ||
412 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=409600 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
413 | + | ||
414 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=262144 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
415 | + | ||
416 | +wrote 65536/65536 bytes at offset 344064 | ||
417 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
418 | + | ||
419 | +Image resized. | ||
420 | + | ||
421 | +read 65536/65536 bytes at offset 344064 | ||
422 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
423 | + | ||
424 | +256 KiB (0x40000) bytes not allocated at offset 0 bytes (0x0) | ||
425 | +256 KiB (0x40000) bytes allocated at offset 256 KiB (0x40000) | ||
426 | + | ||
427 | +[{ "start": 0, "length": 262144, "depth": 1, "zero": true, "data": false}, | ||
428 | +{ "start": 262144, "length": 262144, "depth": 0, "zero": true, "data": false}] | ||
429 | + | ||
430 | +=== preallocation=off === | ||
431 | +Formatting 'TEST_DIR/PID-base', fmt=qcow2 size=524288 cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
432 | + | ||
433 | +Formatting 'TEST_DIR/PID-top', fmt=qcow2 size=262144 backing_file=TEST_DIR/PID-base cluster_size=65536 lazy_refcounts=off refcount_bits=16 | ||
434 | + | ||
435 | +wrote 65536/65536 bytes at offset 446464 | ||
436 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
437 | + | ||
438 | +Image resized. | ||
439 | + | ||
440 | +read 65536/65536 bytes at offset 446464 | ||
441 | +64 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec) | ||
442 | + | ||
443 | +256 KiB (0x40000) bytes not allocated at offset 0 bytes (0x0) | ||
444 | +244 KiB (0x3d000) bytes allocated at offset 256 KiB (0x40000) | ||
445 | + | ||
446 | +[{ "start": 0, "length": 262144, "depth": 1, "zero": true, "data": false}, | ||
447 | +{ "start": 262144, "length": 249856, "depth": 0, "zero": true, "data": false}] | ||
448 | + | ||
449 | diff --git a/tests/qemu-iotests/group b/tests/qemu-iotests/group | ||
450 | index XXXXXXX..XXXXXXX 100644 | ||
451 | --- a/tests/qemu-iotests/group | ||
452 | +++ b/tests/qemu-iotests/group | ||
453 | @@ -XXX,XX +XXX,XX @@ | ||
454 | 270 rw backing quick | ||
455 | 272 rw | ||
456 | 273 backing quick | ||
457 | +274 rw backing | ||
458 | 277 rw quick | ||
459 | 279 rw backing quick | ||
460 | 280 rw migration quick | ||
461 | -- | 39 | -- |
462 | 2.25.3 | 40 | 2.31.1 |
463 | 41 | ||
464 | 42 | diff view generated by jsdifflib |
1 | For regular files, we always get BDRV_REQ_ZERO_WRITE behaviour from the | 1 | From: Stefano Garzarella <sgarzare@redhat.com> |
---|---|---|---|
2 | OS, so we can advertise the flag and just ignore it. | ||
3 | 2 | ||
4 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 3 | Commit d7ddd0a161 ("linux-aio: limit the batch size using |
5 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 4 | `aio-max-batch` parameter") added a way to limit the batch size |
6 | Reviewed-by: Alberto Garcia <berto@igalia.com> | 5 | of Linux AIO backend for the entire AIO context. |
7 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 6 | |
8 | Message-Id: <20200424125448.63318-7-kwolf@redhat.com> | 7 | The same AIO context can be shared by multiple devices, so |
8 | latency-sensitive devices may want to limit the batch size even | ||
9 | more to avoid increasing latency. | ||
10 | |||
11 | For this reason we add the `aio-max-batch` option to the file | ||
12 | backend, which will be used by the next commits to limit the size of | ||
13 | batches including requests generated by this device. | ||
14 | |||
15 | Suggested-by: Kevin Wolf <kwolf@redhat.com> | ||
16 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | ||
17 | Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> | ||
18 | Message-Id: <20211026162346.253081-2-sgarzare@redhat.com> | ||
19 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
9 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 20 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
10 | --- | 21 | --- |
11 | block/file-posix.c | 4 ++++ | 22 | qapi/block-core.json | 7 +++++++ |
12 | 1 file changed, 4 insertions(+) | 23 | block/file-posix.c | 9 +++++++++ |
24 | 2 files changed, 16 insertions(+) | ||
13 | 25 | ||
26 | diff --git a/qapi/block-core.json b/qapi/block-core.json | ||
27 | index XXXXXXX..XXXXXXX 100644 | ||
28 | --- a/qapi/block-core.json | ||
29 | +++ b/qapi/block-core.json | ||
30 | @@ -XXX,XX +XXX,XX @@ | ||
31 | # for this device (default: none, forward the commands via SG_IO; | ||
32 | # since 2.11) | ||
33 | # @aio: AIO backend (default: threads) (since: 2.8) | ||
34 | +# @aio-max-batch: maximum number of requests to batch together into a single | ||
35 | +# submission in the AIO backend. The smallest value between | ||
36 | +# this and the aio-max-batch value of the IOThread object is | ||
37 | +# chosen. | ||
38 | +# 0 means that the AIO backend will handle it automatically. | ||
39 | +# (default: 0, since 6.2) | ||
40 | # @locking: whether to enable file locking. If set to 'auto', only enable | ||
41 | # when Open File Descriptor (OFD) locking API is available | ||
42 | # (default: auto, since 2.10) | ||
43 | @@ -XXX,XX +XXX,XX @@ | ||
44 | '*pr-manager': 'str', | ||
45 | '*locking': 'OnOffAuto', | ||
46 | '*aio': 'BlockdevAioOptions', | ||
47 | + '*aio-max-batch': 'int', | ||
48 | '*drop-cache': {'type': 'bool', | ||
49 | 'if': 'CONFIG_LINUX'}, | ||
50 | '*x-check-cache-dropped': { 'type': 'bool', | ||
14 | diff --git a/block/file-posix.c b/block/file-posix.c | 51 | diff --git a/block/file-posix.c b/block/file-posix.c |
15 | index XXXXXXX..XXXXXXX 100644 | 52 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/block/file-posix.c | 53 | --- a/block/file-posix.c |
17 | +++ b/block/file-posix.c | 54 | +++ b/block/file-posix.c |
55 | @@ -XXX,XX +XXX,XX @@ typedef struct BDRVRawState { | ||
56 | uint64_t locked_perm; | ||
57 | uint64_t locked_shared_perm; | ||
58 | |||
59 | + uint64_t aio_max_batch; | ||
60 | + | ||
61 | int perm_change_fd; | ||
62 | int perm_change_flags; | ||
63 | BDRVReopenState *reopen_state; | ||
64 | @@ -XXX,XX +XXX,XX @@ static QemuOptsList raw_runtime_opts = { | ||
65 | .type = QEMU_OPT_STRING, | ||
66 | .help = "host AIO implementation (threads, native, io_uring)", | ||
67 | }, | ||
68 | + { | ||
69 | + .name = "aio-max-batch", | ||
70 | + .type = QEMU_OPT_NUMBER, | ||
71 | + .help = "AIO max batch size (0 = auto handled by AIO backend, default: 0)", | ||
72 | + }, | ||
73 | { | ||
74 | .name = "locking", | ||
75 | .type = QEMU_OPT_STRING, | ||
18 | @@ -XXX,XX +XXX,XX @@ static int raw_open_common(BlockDriverState *bs, QDict *options, | 76 | @@ -XXX,XX +XXX,XX @@ static int raw_open_common(BlockDriverState *bs, QDict *options, |
77 | s->use_linux_io_uring = (aio == BLOCKDEV_AIO_OPTIONS_IO_URING); | ||
19 | #endif | 78 | #endif |
20 | 79 | ||
21 | bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK; | 80 | + s->aio_max_batch = qemu_opt_get_number(opts, "aio-max-batch", 0); |
22 | + if (S_ISREG(st.st_mode)) { | 81 | + |
23 | + /* When extending regular files, we get zeros from the OS */ | 82 | locking = qapi_enum_parse(&OnOffAuto_lookup, |
24 | + bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE; | 83 | qemu_opt_get(opts, "locking"), |
25 | + } | 84 | ON_OFF_AUTO_AUTO, &local_err); |
26 | ret = 0; | ||
27 | fail: | ||
28 | if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) { | ||
29 | -- | 85 | -- |
30 | 2.25.3 | 86 | 2.31.1 |
31 | 87 | ||
32 | 88 | diff view generated by jsdifflib |
1 | Now that block drivers can support flags for .bdrv_co_truncate, expose | 1 | From: Stefano Garzarella <sgarzare@redhat.com> |
---|---|---|---|
2 | the parameter in the node level interfaces bdrv_co_truncate() and | ||
3 | bdrv_truncate(). | ||
4 | 2 | ||
5 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 3 | This new parameter can be used by block devices to limit the |
6 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 4 | Linux AIO batch size more than the limit set by the AIO context. |
7 | Reviewed-by: Alberto Garcia <berto@igalia.com> | 5 | |
8 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 6 | file-posix backend supports this, passing its `aio-max-batch` option |
9 | Message-Id: <20200424125448.63318-3-kwolf@redhat.com> | 7 | previously added. |
8 | |||
9 | Add an helper function to calculate the maximum batch size. | ||
10 | |||
11 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
12 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> | ||
13 | Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> | ||
14 | Message-Id: <20211026162346.253081-3-sgarzare@redhat.com> | ||
10 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 15 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
11 | --- | 16 | --- |
12 | include/block/block.h | 5 +++-- | 17 | include/block/raw-aio.h | 3 ++- |
13 | block/block-backend.c | 2 +- | 18 | block/file-posix.c | 3 ++- |
14 | block/crypto.c | 2 +- | 19 | block/linux-aio.c | 30 ++++++++++++++++++++++-------- |
15 | block/io.c | 12 +++++++----- | 20 | 3 files changed, 26 insertions(+), 10 deletions(-) |
16 | block/parallels.c | 6 +++--- | ||
17 | block/qcow.c | 4 ++-- | ||
18 | block/qcow2-refcount.c | 2 +- | ||
19 | block/qcow2.c | 15 +++++++++------ | ||
20 | block/raw-format.c | 2 +- | ||
21 | block/vhdx-log.c | 2 +- | ||
22 | block/vhdx.c | 2 +- | ||
23 | block/vmdk.c | 2 +- | ||
24 | tests/test-block-iothread.c | 6 +++--- | ||
25 | 13 files changed, 34 insertions(+), 28 deletions(-) | ||
26 | 21 | ||
27 | diff --git a/include/block/block.h b/include/block/block.h | 22 | diff --git a/include/block/raw-aio.h b/include/block/raw-aio.h |
28 | index XXXXXXX..XXXXXXX 100644 | 23 | index XXXXXXX..XXXXXXX 100644 |
29 | --- a/include/block/block.h | 24 | --- a/include/block/raw-aio.h |
30 | +++ b/include/block/block.h | 25 | +++ b/include/block/raw-aio.h |
31 | @@ -XXX,XX +XXX,XX @@ BlockDriverState *bdrv_find_backing_image(BlockDriverState *bs, | 26 | @@ -XXX,XX +XXX,XX @@ typedef struct LinuxAioState LinuxAioState; |
32 | void bdrv_refresh_filename(BlockDriverState *bs); | 27 | LinuxAioState *laio_init(Error **errp); |
33 | 28 | void laio_cleanup(LinuxAioState *s); | |
34 | int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, | 29 | int coroutine_fn laio_co_submit(BlockDriverState *bs, LinuxAioState *s, int fd, |
35 | - PreallocMode prealloc, Error **errp); | 30 | - uint64_t offset, QEMUIOVector *qiov, int type); |
36 | + PreallocMode prealloc, BdrvRequestFlags flags, | 31 | + uint64_t offset, QEMUIOVector *qiov, int type, |
37 | + Error **errp); | 32 | + uint64_t dev_max_batch); |
38 | int bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, | 33 | void laio_detach_aio_context(LinuxAioState *s, AioContext *old_context); |
39 | - PreallocMode prealloc, Error **errp); | 34 | void laio_attach_aio_context(LinuxAioState *s, AioContext *new_context); |
40 | + PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); | 35 | void laio_io_plug(BlockDriverState *bs, LinuxAioState *s); |
41 | 36 | diff --git a/block/file-posix.c b/block/file-posix.c | |
42 | int64_t bdrv_nb_sectors(BlockDriverState *bs); | ||
43 | int64_t bdrv_getlength(BlockDriverState *bs); | ||
44 | diff --git a/block/block-backend.c b/block/block-backend.c | ||
45 | index XXXXXXX..XXXXXXX 100644 | 37 | index XXXXXXX..XXXXXXX 100644 |
46 | --- a/block/block-backend.c | 38 | --- a/block/file-posix.c |
47 | +++ b/block/block-backend.c | 39 | +++ b/block/file-posix.c |
48 | @@ -XXX,XX +XXX,XX @@ int blk_truncate(BlockBackend *blk, int64_t offset, bool exact, | 40 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn raw_co_prw(BlockDriverState *bs, uint64_t offset, |
49 | return -ENOMEDIUM; | 41 | } else if (s->use_linux_aio) { |
42 | LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs)); | ||
43 | assert(qiov->size == bytes); | ||
44 | - return laio_co_submit(bs, aio, s->fd, offset, qiov, type); | ||
45 | + return laio_co_submit(bs, aio, s->fd, offset, qiov, type, | ||
46 | + s->aio_max_batch); | ||
47 | #endif | ||
50 | } | 48 | } |
51 | 49 | ||
52 | - return bdrv_truncate(blk->root, offset, exact, prealloc, errp); | 50 | diff --git a/block/linux-aio.c b/block/linux-aio.c |
53 | + return bdrv_truncate(blk->root, offset, exact, prealloc, 0, errp); | 51 | index XXXXXXX..XXXXXXX 100644 |
52 | --- a/block/linux-aio.c | ||
53 | +++ b/block/linux-aio.c | ||
54 | @@ -XXX,XX +XXX,XX @@ static void ioq_submit(LinuxAioState *s) | ||
55 | } | ||
54 | } | 56 | } |
55 | 57 | ||
56 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, | 58 | +static uint64_t laio_max_batch(LinuxAioState *s, uint64_t dev_max_batch) |
57 | diff --git a/block/crypto.c b/block/crypto.c | 59 | +{ |
58 | index XXXXXXX..XXXXXXX 100644 | 60 | + uint64_t max_batch = s->aio_context->aio_max_batch ?: DEFAULT_MAX_BATCH; |
59 | --- a/block/crypto.c | 61 | + |
60 | +++ b/block/crypto.c | 62 | + /* |
61 | @@ -XXX,XX +XXX,XX @@ block_crypto_co_truncate(BlockDriverState *bs, int64_t offset, bool exact, | 63 | + * AIO context can be shared between multiple block devices, so |
62 | 64 | + * `dev_max_batch` allows reducing the batch size for latency-sensitive | |
63 | offset += payload_offset; | 65 | + * devices. |
64 | 66 | + */ | |
65 | - return bdrv_co_truncate(bs->file, offset, exact, prealloc, errp); | 67 | + max_batch = MIN_NON_ZERO(dev_max_batch, max_batch); |
66 | + return bdrv_co_truncate(bs->file, offset, exact, prealloc, 0, errp); | 68 | + |
69 | + /* limit the batch with the number of available events */ | ||
70 | + max_batch = MIN_NON_ZERO(MAX_EVENTS - s->io_q.in_flight, max_batch); | ||
71 | + | ||
72 | + return max_batch; | ||
73 | +} | ||
74 | + | ||
75 | void laio_io_plug(BlockDriverState *bs, LinuxAioState *s) | ||
76 | { | ||
77 | s->io_q.plugged++; | ||
78 | @@ -XXX,XX +XXX,XX @@ void laio_io_unplug(BlockDriverState *bs, LinuxAioState *s) | ||
67 | } | 79 | } |
68 | 80 | ||
69 | static void block_crypto_close(BlockDriverState *bs) | 81 | static int laio_do_submit(int fd, struct qemu_laiocb *laiocb, off_t offset, |
70 | diff --git a/block/io.c b/block/io.c | 82 | - int type) |
71 | index XXXXXXX..XXXXXXX 100644 | 83 | + int type, uint64_t dev_max_batch) |
72 | --- a/block/io.c | ||
73 | +++ b/block/io.c | ||
74 | @@ -XXX,XX +XXX,XX @@ static void bdrv_parent_cb_resize(BlockDriverState *bs) | ||
75 | * 'offset' bytes in length. | ||
76 | */ | ||
77 | int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, | ||
78 | - PreallocMode prealloc, Error **errp) | ||
79 | + PreallocMode prealloc, BdrvRequestFlags flags, | ||
80 | + Error **errp) | ||
81 | { | 84 | { |
82 | BlockDriverState *bs = child->bs; | 85 | LinuxAioState *s = laiocb->ctx; |
83 | BlockDriver *drv = bs->drv; | 86 | struct iocb *iocbs = &laiocb->iocb; |
84 | BdrvTrackedRequest req; | 87 | QEMUIOVector *qiov = laiocb->qiov; |
85 | - BdrvRequestFlags flags = 0; | 88 | - int64_t max_batch = s->aio_context->aio_max_batch ?: DEFAULT_MAX_BATCH; |
86 | int64_t old_size, new_bytes; | 89 | - |
90 | - /* limit the batch with the number of available events */ | ||
91 | - max_batch = MIN_NON_ZERO(MAX_EVENTS - s->io_q.in_flight, max_batch); | ||
92 | |||
93 | switch (type) { | ||
94 | case QEMU_AIO_WRITE: | ||
95 | @@ -XXX,XX +XXX,XX @@ static int laio_do_submit(int fd, struct qemu_laiocb *laiocb, off_t offset, | ||
96 | s->io_q.in_queue++; | ||
97 | if (!s->io_q.blocked && | ||
98 | (!s->io_q.plugged || | ||
99 | - s->io_q.in_queue >= max_batch)) { | ||
100 | + s->io_q.in_queue >= laio_max_batch(s, dev_max_batch))) { | ||
101 | ioq_submit(s); | ||
102 | } | ||
103 | |||
104 | @@ -XXX,XX +XXX,XX @@ static int laio_do_submit(int fd, struct qemu_laiocb *laiocb, off_t offset, | ||
105 | } | ||
106 | |||
107 | int coroutine_fn laio_co_submit(BlockDriverState *bs, LinuxAioState *s, int fd, | ||
108 | - uint64_t offset, QEMUIOVector *qiov, int type) | ||
109 | + uint64_t offset, QEMUIOVector *qiov, int type, | ||
110 | + uint64_t dev_max_batch) | ||
111 | { | ||
87 | int ret; | 112 | int ret; |
88 | 113 | struct qemu_laiocb laiocb = { | |
89 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, | 114 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn laio_co_submit(BlockDriverState *bs, LinuxAioState *s, int fd, |
90 | } | 115 | .qiov = qiov, |
91 | ret = drv->bdrv_co_truncate(bs, offset, exact, prealloc, flags, errp); | ||
92 | } else if (bs->file && drv->is_filter) { | ||
93 | - ret = bdrv_co_truncate(bs->file, offset, exact, prealloc, errp); | ||
94 | + ret = bdrv_co_truncate(bs->file, offset, exact, prealloc, flags, errp); | ||
95 | } else { | ||
96 | error_setg(errp, "Image format driver does not support resize"); | ||
97 | ret = -ENOTSUP; | ||
98 | @@ -XXX,XX +XXX,XX @@ typedef struct TruncateCo { | ||
99 | int64_t offset; | ||
100 | bool exact; | ||
101 | PreallocMode prealloc; | ||
102 | + BdrvRequestFlags flags; | ||
103 | Error **errp; | ||
104 | int ret; | ||
105 | } TruncateCo; | ||
106 | @@ -XXX,XX +XXX,XX @@ static void coroutine_fn bdrv_truncate_co_entry(void *opaque) | ||
107 | { | ||
108 | TruncateCo *tco = opaque; | ||
109 | tco->ret = bdrv_co_truncate(tco->child, tco->offset, tco->exact, | ||
110 | - tco->prealloc, tco->errp); | ||
111 | + tco->prealloc, tco->flags, tco->errp); | ||
112 | aio_wait_kick(); | ||
113 | } | ||
114 | |||
115 | int bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, | ||
116 | - PreallocMode prealloc, Error **errp) | ||
117 | + PreallocMode prealloc, BdrvRequestFlags flags, Error **errp) | ||
118 | { | ||
119 | Coroutine *co; | ||
120 | TruncateCo tco = { | ||
121 | @@ -XXX,XX +XXX,XX @@ int bdrv_truncate(BdrvChild *child, int64_t offset, bool exact, | ||
122 | .offset = offset, | ||
123 | .exact = exact, | ||
124 | .prealloc = prealloc, | ||
125 | + .flags = flags, | ||
126 | .errp = errp, | ||
127 | .ret = NOT_DONE, | ||
128 | }; | 116 | }; |
129 | diff --git a/block/parallels.c b/block/parallels.c | 117 | |
130 | index XXXXXXX..XXXXXXX 100644 | 118 | - ret = laio_do_submit(fd, &laiocb, offset, type); |
131 | --- a/block/parallels.c | 119 | + ret = laio_do_submit(fd, &laiocb, offset, type, dev_max_batch); |
132 | +++ b/block/parallels.c | 120 | if (ret < 0) { |
133 | @@ -XXX,XX +XXX,XX @@ static int64_t allocate_clusters(BlockDriverState *bs, int64_t sector_num, | 121 | return ret; |
134 | } else { | ||
135 | ret = bdrv_truncate(bs->file, | ||
136 | (s->data_end + space) << BDRV_SECTOR_BITS, | ||
137 | - false, PREALLOC_MODE_OFF, NULL); | ||
138 | + false, PREALLOC_MODE_OFF, 0, NULL); | ||
139 | } | ||
140 | if (ret < 0) { | ||
141 | return ret; | ||
142 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn parallels_co_check(BlockDriverState *bs, | ||
143 | * That means we have to pass exact=true. | ||
144 | */ | ||
145 | ret = bdrv_truncate(bs->file, res->image_end_offset, true, | ||
146 | - PREALLOC_MODE_OFF, &local_err); | ||
147 | + PREALLOC_MODE_OFF, 0, &local_err); | ||
148 | if (ret < 0) { | ||
149 | error_report_err(local_err); | ||
150 | res->check_errors++; | ||
151 | @@ -XXX,XX +XXX,XX @@ static void parallels_close(BlockDriverState *bs) | ||
152 | |||
153 | /* errors are ignored, so we might as well pass exact=true */ | ||
154 | bdrv_truncate(bs->file, s->data_end << BDRV_SECTOR_BITS, true, | ||
155 | - PREALLOC_MODE_OFF, NULL); | ||
156 | + PREALLOC_MODE_OFF, 0, NULL); | ||
157 | } | 122 | } |
158 | |||
159 | g_free(s->bat_dirty_bmap); | ||
160 | diff --git a/block/qcow.c b/block/qcow.c | ||
161 | index XXXXXXX..XXXXXXX 100644 | ||
162 | --- a/block/qcow.c | ||
163 | +++ b/block/qcow.c | ||
164 | @@ -XXX,XX +XXX,XX @@ static int get_cluster_offset(BlockDriverState *bs, | ||
165 | return -E2BIG; | ||
166 | } | ||
167 | ret = bdrv_truncate(bs->file, cluster_offset + s->cluster_size, | ||
168 | - false, PREALLOC_MODE_OFF, NULL); | ||
169 | + false, PREALLOC_MODE_OFF, 0, NULL); | ||
170 | if (ret < 0) { | ||
171 | return ret; | ||
172 | } | ||
173 | @@ -XXX,XX +XXX,XX @@ static int qcow_make_empty(BlockDriverState *bs) | ||
174 | l1_length) < 0) | ||
175 | return -1; | ||
176 | ret = bdrv_truncate(bs->file, s->l1_table_offset + l1_length, false, | ||
177 | - PREALLOC_MODE_OFF, NULL); | ||
178 | + PREALLOC_MODE_OFF, 0, NULL); | ||
179 | if (ret < 0) | ||
180 | return ret; | ||
181 | |||
182 | diff --git a/block/qcow2-refcount.c b/block/qcow2-refcount.c | ||
183 | index XXXXXXX..XXXXXXX 100644 | ||
184 | --- a/block/qcow2-refcount.c | ||
185 | +++ b/block/qcow2-refcount.c | ||
186 | @@ -XXX,XX +XXX,XX @@ static int check_refblocks(BlockDriverState *bs, BdrvCheckResult *res, | ||
187 | } | ||
188 | |||
189 | ret = bdrv_truncate(bs->file, offset + s->cluster_size, false, | ||
190 | - PREALLOC_MODE_OFF, &local_err); | ||
191 | + PREALLOC_MODE_OFF, 0, &local_err); | ||
192 | if (ret < 0) { | ||
193 | error_report_err(local_err); | ||
194 | goto resize_fail; | ||
195 | diff --git a/block/qcow2.c b/block/qcow2.c | ||
196 | index XXXXXXX..XXXXXXX 100644 | ||
197 | --- a/block/qcow2.c | ||
198 | +++ b/block/qcow2.c | ||
199 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn preallocate_co(BlockDriverState *bs, uint64_t offset, | ||
200 | mode = PREALLOC_MODE_OFF; | ||
201 | } | ||
202 | ret = bdrv_co_truncate(s->data_file, host_offset + cur_bytes, false, | ||
203 | - mode, errp); | ||
204 | + mode, 0, errp); | ||
205 | if (ret < 0) { | ||
206 | return ret; | ||
207 | } | ||
208 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, | ||
209 | * always fulfilled, so there is no need to pass it on.) | ||
210 | */ | ||
211 | bdrv_co_truncate(bs->file, (last_cluster + 1) * s->cluster_size, | ||
212 | - false, PREALLOC_MODE_OFF, &local_err); | ||
213 | + false, PREALLOC_MODE_OFF, 0, &local_err); | ||
214 | if (local_err) { | ||
215 | warn_reportf_err(local_err, | ||
216 | "Failed to truncate the tail of the image: "); | ||
217 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, | ||
218 | * file should be resized to the exact target size, too, | ||
219 | * so we pass @exact here. | ||
220 | */ | ||
221 | - ret = bdrv_co_truncate(s->data_file, offset, exact, prealloc, errp); | ||
222 | + ret = bdrv_co_truncate(s->data_file, offset, exact, prealloc, 0, | ||
223 | + errp); | ||
224 | if (ret < 0) { | ||
225 | goto fail; | ||
226 | } | ||
227 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, | ||
228 | new_file_size = allocation_start + | ||
229 | nb_new_data_clusters * s->cluster_size; | ||
230 | /* Image file grows, so @exact does not matter */ | ||
231 | - ret = bdrv_co_truncate(bs->file, new_file_size, false, prealloc, errp); | ||
232 | + ret = bdrv_co_truncate(bs->file, new_file_size, false, prealloc, 0, | ||
233 | + errp); | ||
234 | if (ret < 0) { | ||
235 | error_prepend(errp, "Failed to resize underlying file: "); | ||
236 | qcow2_free_clusters(bs, allocation_start, | ||
237 | @@ -XXX,XX +XXX,XX @@ qcow2_co_pwritev_compressed_part(BlockDriverState *bs, | ||
238 | if (len < 0) { | ||
239 | return len; | ||
240 | } | ||
241 | - return bdrv_co_truncate(bs->file, len, false, PREALLOC_MODE_OFF, NULL); | ||
242 | + return bdrv_co_truncate(bs->file, len, false, PREALLOC_MODE_OFF, 0, | ||
243 | + NULL); | ||
244 | } | ||
245 | |||
246 | if (offset_into_cluster(s, offset)) { | ||
247 | @@ -XXX,XX +XXX,XX @@ static int make_completely_empty(BlockDriverState *bs) | ||
248 | } | ||
249 | |||
250 | ret = bdrv_truncate(bs->file, (3 + l1_clusters) * s->cluster_size, false, | ||
251 | - PREALLOC_MODE_OFF, &local_err); | ||
252 | + PREALLOC_MODE_OFF, 0, &local_err); | ||
253 | if (ret < 0) { | ||
254 | error_report_err(local_err); | ||
255 | goto fail; | ||
256 | diff --git a/block/raw-format.c b/block/raw-format.c | ||
257 | index XXXXXXX..XXXXXXX 100644 | ||
258 | --- a/block/raw-format.c | ||
259 | +++ b/block/raw-format.c | ||
260 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, | ||
261 | |||
262 | s->size = offset; | ||
263 | offset += s->offset; | ||
264 | - return bdrv_co_truncate(bs->file, offset, exact, prealloc, errp); | ||
265 | + return bdrv_co_truncate(bs->file, offset, exact, prealloc, 0, errp); | ||
266 | } | ||
267 | |||
268 | static void raw_eject(BlockDriverState *bs, bool eject_flag) | ||
269 | diff --git a/block/vhdx-log.c b/block/vhdx-log.c | ||
270 | index XXXXXXX..XXXXXXX 100644 | ||
271 | --- a/block/vhdx-log.c | ||
272 | +++ b/block/vhdx-log.c | ||
273 | @@ -XXX,XX +XXX,XX @@ static int vhdx_log_flush(BlockDriverState *bs, BDRVVHDXState *s, | ||
274 | goto exit; | ||
275 | } | ||
276 | ret = bdrv_truncate(bs->file, new_file_size, false, | ||
277 | - PREALLOC_MODE_OFF, NULL); | ||
278 | + PREALLOC_MODE_OFF, 0, NULL); | ||
279 | if (ret < 0) { | ||
280 | goto exit; | ||
281 | } | ||
282 | diff --git a/block/vhdx.c b/block/vhdx.c | ||
283 | index XXXXXXX..XXXXXXX 100644 | ||
284 | --- a/block/vhdx.c | ||
285 | +++ b/block/vhdx.c | ||
286 | @@ -XXX,XX +XXX,XX @@ static int vhdx_allocate_block(BlockDriverState *bs, BDRVVHDXState *s, | ||
287 | } | ||
288 | |||
289 | return bdrv_truncate(bs->file, *new_offset + s->block_size, false, | ||
290 | - PREALLOC_MODE_OFF, NULL); | ||
291 | + PREALLOC_MODE_OFF, 0, NULL); | ||
292 | } | ||
293 | |||
294 | /* | ||
295 | diff --git a/block/vmdk.c b/block/vmdk.c | ||
296 | index XXXXXXX..XXXXXXX 100644 | ||
297 | --- a/block/vmdk.c | ||
298 | +++ b/block/vmdk.c | ||
299 | @@ -XXX,XX +XXX,XX @@ vmdk_co_pwritev_compressed(BlockDriverState *bs, uint64_t offset, | ||
300 | } | ||
301 | length = QEMU_ALIGN_UP(length, BDRV_SECTOR_SIZE); | ||
302 | ret = bdrv_truncate(s->extents[i].file, length, false, | ||
303 | - PREALLOC_MODE_OFF, NULL); | ||
304 | + PREALLOC_MODE_OFF, 0, NULL); | ||
305 | if (ret < 0) { | ||
306 | return ret; | ||
307 | } | ||
308 | diff --git a/tests/test-block-iothread.c b/tests/test-block-iothread.c | ||
309 | index XXXXXXX..XXXXXXX 100644 | ||
310 | --- a/tests/test-block-iothread.c | ||
311 | +++ b/tests/test-block-iothread.c | ||
312 | @@ -XXX,XX +XXX,XX @@ static void test_sync_op_truncate(BdrvChild *c) | ||
313 | int ret; | ||
314 | |||
315 | /* Normal success path */ | ||
316 | - ret = bdrv_truncate(c, 65536, false, PREALLOC_MODE_OFF, NULL); | ||
317 | + ret = bdrv_truncate(c, 65536, false, PREALLOC_MODE_OFF, 0, NULL); | ||
318 | g_assert_cmpint(ret, ==, 0); | ||
319 | |||
320 | /* Early error: Negative offset */ | ||
321 | - ret = bdrv_truncate(c, -2, false, PREALLOC_MODE_OFF, NULL); | ||
322 | + ret = bdrv_truncate(c, -2, false, PREALLOC_MODE_OFF, 0, NULL); | ||
323 | g_assert_cmpint(ret, ==, -EINVAL); | ||
324 | |||
325 | /* Error: Read-only image */ | ||
326 | c->bs->read_only = true; | ||
327 | c->bs->open_flags &= ~BDRV_O_RDWR; | ||
328 | |||
329 | - ret = bdrv_truncate(c, 65536, false, PREALLOC_MODE_OFF, NULL); | ||
330 | + ret = bdrv_truncate(c, 65536, false, PREALLOC_MODE_OFF, 0, NULL); | ||
331 | g_assert_cmpint(ret, ==, -EACCES); | ||
332 | |||
333 | c->bs->read_only = false; | ||
334 | -- | 123 | -- |
335 | 2.25.3 | 124 | 2.31.1 |
336 | 125 | ||
337 | 126 | diff view generated by jsdifflib |
1 | This adds a new BdrvRequestFlags parameter to the .bdrv_co_truncate() | 1 | From: Stefano Garzarella <sgarzare@redhat.com> |
---|---|---|---|
2 | driver callbacks, and a supported_truncate_flags field in | ||
3 | BlockDriverState that allows drivers to advertise support for request | ||
4 | flags in the context of truncate. | ||
5 | 2 | ||
6 | For now, we always pass 0 and no drivers declare support for any flag. | 3 | Between the submission of a request and the unplug, other devices |
4 | with larger limits may have been queued new requests without flushing | ||
5 | the batch. | ||
7 | 6 | ||
8 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 7 | Using the new `dev_max_batch` parameter, laio_io_unplug() can check |
9 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 8 | if the batch exceeds the device limit to flush the current batch. |
10 | Reviewed-by: Alberto Garcia <berto@igalia.com> | 9 | |
11 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 10 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
12 | Message-Id: <20200424125448.63318-2-kwolf@redhat.com> | 11 | Reviewed-by: Kevin Wolf <kwolf@redhat.com> |
12 | Signed-off-by: Stefano Garzarella <sgarzare@redhat.com> | ||
13 | Message-Id: <20211026162346.253081-4-sgarzare@redhat.com> | ||
13 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 14 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
14 | --- | 15 | --- |
15 | include/block/block_int.h | 10 +++++++++- | 16 | include/block/raw-aio.h | 3 ++- |
16 | block/crypto.c | 3 ++- | 17 | block/file-posix.c | 2 +- |
17 | block/file-posix.c | 2 +- | 18 | block/linux-aio.c | 8 +++++--- |
18 | block/file-win32.c | 2 +- | 19 | 3 files changed, 8 insertions(+), 5 deletions(-) |
19 | block/gluster.c | 1 + | ||
20 | block/io.c | 8 +++++++- | ||
21 | block/iscsi.c | 2 +- | ||
22 | block/nfs.c | 3 ++- | ||
23 | block/qcow2.c | 2 +- | ||
24 | block/qed.c | 1 + | ||
25 | block/raw-format.c | 2 +- | ||
26 | block/rbd.c | 1 + | ||
27 | block/sheepdog.c | 4 ++-- | ||
28 | block/ssh.c | 2 +- | ||
29 | tests/test-block-iothread.c | 3 ++- | ||
30 | 15 files changed, 33 insertions(+), 13 deletions(-) | ||
31 | 20 | ||
32 | diff --git a/include/block/block_int.h b/include/block/block_int.h | 21 | diff --git a/include/block/raw-aio.h b/include/block/raw-aio.h |
33 | index XXXXXXX..XXXXXXX 100644 | 22 | index XXXXXXX..XXXXXXX 100644 |
34 | --- a/include/block/block_int.h | 23 | --- a/include/block/raw-aio.h |
35 | +++ b/include/block/block_int.h | 24 | +++ b/include/block/raw-aio.h |
36 | @@ -XXX,XX +XXX,XX @@ struct BlockDriver { | 25 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn laio_co_submit(BlockDriverState *bs, LinuxAioState *s, int fd, |
37 | */ | 26 | void laio_detach_aio_context(LinuxAioState *s, AioContext *old_context); |
38 | int coroutine_fn (*bdrv_co_truncate)(BlockDriverState *bs, int64_t offset, | 27 | void laio_attach_aio_context(LinuxAioState *s, AioContext *new_context); |
39 | bool exact, PreallocMode prealloc, | 28 | void laio_io_plug(BlockDriverState *bs, LinuxAioState *s); |
40 | - Error **errp); | 29 | -void laio_io_unplug(BlockDriverState *bs, LinuxAioState *s); |
41 | + BdrvRequestFlags flags, Error **errp); | 30 | +void laio_io_unplug(BlockDriverState *bs, LinuxAioState *s, |
42 | 31 | + uint64_t dev_max_batch); | |
43 | int64_t (*bdrv_getlength)(BlockDriverState *bs); | 32 | #endif |
44 | bool has_variable_length; | 33 | /* io_uring.c - Linux io_uring implementation */ |
45 | @@ -XXX,XX +XXX,XX @@ struct BlockDriverState { | 34 | #ifdef CONFIG_LINUX_IO_URING |
46 | /* Flags honored during pwrite_zeroes (so far: BDRV_REQ_FUA, | ||
47 | * BDRV_REQ_MAY_UNMAP, BDRV_REQ_WRITE_UNCHANGED) */ | ||
48 | unsigned int supported_zero_flags; | ||
49 | + /* | ||
50 | + * Flags honoured during truncate (so far: BDRV_REQ_ZERO_WRITE). | ||
51 | + * | ||
52 | + * If BDRV_REQ_ZERO_WRITE is given, the truncate operation must make sure | ||
53 | + * that any added space reads as all zeros. If this can't be guaranteed, | ||
54 | + * the operation must fail. | ||
55 | + */ | ||
56 | + unsigned int supported_truncate_flags; | ||
57 | |||
58 | /* the following member gives a name to every node on the bs graph. */ | ||
59 | char node_name[32]; | ||
60 | diff --git a/block/crypto.c b/block/crypto.c | ||
61 | index XXXXXXX..XXXXXXX 100644 | ||
62 | --- a/block/crypto.c | ||
63 | +++ b/block/crypto.c | ||
64 | @@ -XXX,XX +XXX,XX @@ static int block_crypto_co_create_generic(BlockDriverState *bs, | ||
65 | |||
66 | static int coroutine_fn | ||
67 | block_crypto_co_truncate(BlockDriverState *bs, int64_t offset, bool exact, | ||
68 | - PreallocMode prealloc, Error **errp) | ||
69 | + PreallocMode prealloc, BdrvRequestFlags flags, | ||
70 | + Error **errp) | ||
71 | { | ||
72 | BlockCrypto *crypto = bs->opaque; | ||
73 | uint64_t payload_offset = | ||
74 | diff --git a/block/file-posix.c b/block/file-posix.c | 35 | diff --git a/block/file-posix.c b/block/file-posix.c |
75 | index XXXXXXX..XXXXXXX 100644 | 36 | index XXXXXXX..XXXXXXX 100644 |
76 | --- a/block/file-posix.c | 37 | --- a/block/file-posix.c |
77 | +++ b/block/file-posix.c | 38 | +++ b/block/file-posix.c |
78 | @@ -XXX,XX +XXX,XX @@ raw_regular_truncate(BlockDriverState *bs, int fd, int64_t offset, | 39 | @@ -XXX,XX +XXX,XX @@ static void raw_aio_unplug(BlockDriverState *bs) |
79 | 40 | #ifdef CONFIG_LINUX_AIO | |
80 | static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, | 41 | if (s->use_linux_aio) { |
81 | bool exact, PreallocMode prealloc, | 42 | LinuxAioState *aio = aio_get_linux_aio(bdrv_get_aio_context(bs)); |
82 | - Error **errp) | 43 | - laio_io_unplug(bs, aio); |
83 | + BdrvRequestFlags flags, Error **errp) | 44 | + laio_io_unplug(bs, aio, s->aio_max_batch); |
45 | } | ||
46 | #endif | ||
47 | #ifdef CONFIG_LINUX_IO_URING | ||
48 | diff --git a/block/linux-aio.c b/block/linux-aio.c | ||
49 | index XXXXXXX..XXXXXXX 100644 | ||
50 | --- a/block/linux-aio.c | ||
51 | +++ b/block/linux-aio.c | ||
52 | @@ -XXX,XX +XXX,XX @@ void laio_io_plug(BlockDriverState *bs, LinuxAioState *s) | ||
53 | s->io_q.plugged++; | ||
54 | } | ||
55 | |||
56 | -void laio_io_unplug(BlockDriverState *bs, LinuxAioState *s) | ||
57 | +void laio_io_unplug(BlockDriverState *bs, LinuxAioState *s, | ||
58 | + uint64_t dev_max_batch) | ||
84 | { | 59 | { |
85 | BDRVRawState *s = bs->opaque; | 60 | assert(s->io_q.plugged); |
86 | struct stat st; | 61 | - if (--s->io_q.plugged == 0 && |
87 | diff --git a/block/file-win32.c b/block/file-win32.c | 62 | - !s->io_q.blocked && !QSIMPLEQ_EMPTY(&s->io_q.pending)) { |
88 | index XXXXXXX..XXXXXXX 100644 | 63 | + if (s->io_q.in_queue >= laio_max_batch(s, dev_max_batch) || |
89 | --- a/block/file-win32.c | 64 | + (--s->io_q.plugged == 0 && |
90 | +++ b/block/file-win32.c | 65 | + !s->io_q.blocked && !QSIMPLEQ_EMPTY(&s->io_q.pending))) { |
91 | @@ -XXX,XX +XXX,XX @@ static void raw_close(BlockDriverState *bs) | 66 | ioq_submit(s); |
92 | |||
93 | static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, | ||
94 | bool exact, PreallocMode prealloc, | ||
95 | - Error **errp) | ||
96 | + BdrvRequestFlags flags, Error **errp) | ||
97 | { | ||
98 | BDRVRawState *s = bs->opaque; | ||
99 | LONG low, high; | ||
100 | diff --git a/block/gluster.c b/block/gluster.c | ||
101 | index XXXXXXX..XXXXXXX 100644 | ||
102 | --- a/block/gluster.c | ||
103 | +++ b/block/gluster.c | ||
104 | @@ -XXX,XX +XXX,XX @@ static coroutine_fn int qemu_gluster_co_truncate(BlockDriverState *bs, | ||
105 | int64_t offset, | ||
106 | bool exact, | ||
107 | PreallocMode prealloc, | ||
108 | + BdrvRequestFlags flags, | ||
109 | Error **errp) | ||
110 | { | ||
111 | BDRVGlusterState *s = bs->opaque; | ||
112 | diff --git a/block/io.c b/block/io.c | ||
113 | index XXXXXXX..XXXXXXX 100644 | ||
114 | --- a/block/io.c | ||
115 | +++ b/block/io.c | ||
116 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, | ||
117 | BlockDriverState *bs = child->bs; | ||
118 | BlockDriver *drv = bs->drv; | ||
119 | BdrvTrackedRequest req; | ||
120 | + BdrvRequestFlags flags = 0; | ||
121 | int64_t old_size, new_bytes; | ||
122 | int ret; | ||
123 | |||
124 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn bdrv_co_truncate(BdrvChild *child, int64_t offset, bool exact, | ||
125 | } | 67 | } |
126 | |||
127 | if (drv->bdrv_co_truncate) { | ||
128 | - ret = drv->bdrv_co_truncate(bs, offset, exact, prealloc, errp); | ||
129 | + if (flags & ~bs->supported_truncate_flags) { | ||
130 | + error_setg(errp, "Block driver does not support requested flags"); | ||
131 | + ret = -ENOTSUP; | ||
132 | + goto out; | ||
133 | + } | ||
134 | + ret = drv->bdrv_co_truncate(bs, offset, exact, prealloc, flags, errp); | ||
135 | } else if (bs->file && drv->is_filter) { | ||
136 | ret = bdrv_co_truncate(bs->file, offset, exact, prealloc, errp); | ||
137 | } else { | ||
138 | diff --git a/block/iscsi.c b/block/iscsi.c | ||
139 | index XXXXXXX..XXXXXXX 100644 | ||
140 | --- a/block/iscsi.c | ||
141 | +++ b/block/iscsi.c | ||
142 | @@ -XXX,XX +XXX,XX @@ static void iscsi_reopen_commit(BDRVReopenState *reopen_state) | ||
143 | |||
144 | static int coroutine_fn iscsi_co_truncate(BlockDriverState *bs, int64_t offset, | ||
145 | bool exact, PreallocMode prealloc, | ||
146 | - Error **errp) | ||
147 | + BdrvRequestFlags flags, Error **errp) | ||
148 | { | ||
149 | IscsiLun *iscsilun = bs->opaque; | ||
150 | int64_t cur_length; | ||
151 | diff --git a/block/nfs.c b/block/nfs.c | ||
152 | index XXXXXXX..XXXXXXX 100644 | ||
153 | --- a/block/nfs.c | ||
154 | +++ b/block/nfs.c | ||
155 | @@ -XXX,XX +XXX,XX @@ static int64_t nfs_get_allocated_file_size(BlockDriverState *bs) | ||
156 | |||
157 | static int coroutine_fn | ||
158 | nfs_file_co_truncate(BlockDriverState *bs, int64_t offset, bool exact, | ||
159 | - PreallocMode prealloc, Error **errp) | ||
160 | + PreallocMode prealloc, BdrvRequestFlags flags, | ||
161 | + Error **errp) | ||
162 | { | ||
163 | NFSClient *client = bs->opaque; | ||
164 | int ret; | ||
165 | diff --git a/block/qcow2.c b/block/qcow2.c | ||
166 | index XXXXXXX..XXXXXXX 100644 | ||
167 | --- a/block/qcow2.c | ||
168 | +++ b/block/qcow2.c | ||
169 | @@ -XXX,XX +XXX,XX @@ fail: | ||
170 | |||
171 | static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, | ||
172 | bool exact, PreallocMode prealloc, | ||
173 | - Error **errp) | ||
174 | + BdrvRequestFlags flags, Error **errp) | ||
175 | { | ||
176 | BDRVQcow2State *s = bs->opaque; | ||
177 | uint64_t old_length; | ||
178 | diff --git a/block/qed.c b/block/qed.c | ||
179 | index XXXXXXX..XXXXXXX 100644 | ||
180 | --- a/block/qed.c | ||
181 | +++ b/block/qed.c | ||
182 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn bdrv_qed_co_truncate(BlockDriverState *bs, | ||
183 | int64_t offset, | ||
184 | bool exact, | ||
185 | PreallocMode prealloc, | ||
186 | + BdrvRequestFlags flags, | ||
187 | Error **errp) | ||
188 | { | ||
189 | BDRVQEDState *s = bs->opaque; | ||
190 | diff --git a/block/raw-format.c b/block/raw-format.c | ||
191 | index XXXXXXX..XXXXXXX 100644 | ||
192 | --- a/block/raw-format.c | ||
193 | +++ b/block/raw-format.c | ||
194 | @@ -XXX,XX +XXX,XX @@ static void raw_refresh_limits(BlockDriverState *bs, Error **errp) | ||
195 | |||
196 | static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, | ||
197 | bool exact, PreallocMode prealloc, | ||
198 | - Error **errp) | ||
199 | + BdrvRequestFlags flags, Error **errp) | ||
200 | { | ||
201 | BDRVRawState *s = bs->opaque; | ||
202 | |||
203 | diff --git a/block/rbd.c b/block/rbd.c | ||
204 | index XXXXXXX..XXXXXXX 100644 | ||
205 | --- a/block/rbd.c | ||
206 | +++ b/block/rbd.c | ||
207 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qemu_rbd_co_truncate(BlockDriverState *bs, | ||
208 | int64_t offset, | ||
209 | bool exact, | ||
210 | PreallocMode prealloc, | ||
211 | + BdrvRequestFlags flags, | ||
212 | Error **errp) | ||
213 | { | ||
214 | int r; | ||
215 | diff --git a/block/sheepdog.c b/block/sheepdog.c | ||
216 | index XXXXXXX..XXXXXXX 100644 | ||
217 | --- a/block/sheepdog.c | ||
218 | +++ b/block/sheepdog.c | ||
219 | @@ -XXX,XX +XXX,XX @@ static int64_t sd_getlength(BlockDriverState *bs) | ||
220 | |||
221 | static int coroutine_fn sd_co_truncate(BlockDriverState *bs, int64_t offset, | ||
222 | bool exact, PreallocMode prealloc, | ||
223 | - Error **errp) | ||
224 | + BdrvRequestFlags flags, Error **errp) | ||
225 | { | ||
226 | BDRVSheepdogState *s = bs->opaque; | ||
227 | int ret, fd; | ||
228 | @@ -XXX,XX +XXX,XX @@ static coroutine_fn int sd_co_writev(BlockDriverState *bs, int64_t sector_num, | ||
229 | |||
230 | assert(!flags); | ||
231 | if (offset > s->inode.vdi_size) { | ||
232 | - ret = sd_co_truncate(bs, offset, false, PREALLOC_MODE_OFF, NULL); | ||
233 | + ret = sd_co_truncate(bs, offset, false, PREALLOC_MODE_OFF, 0, NULL); | ||
234 | if (ret < 0) { | ||
235 | return ret; | ||
236 | } | ||
237 | diff --git a/block/ssh.c b/block/ssh.c | ||
238 | index XXXXXXX..XXXXXXX 100644 | ||
239 | --- a/block/ssh.c | ||
240 | +++ b/block/ssh.c | ||
241 | @@ -XXX,XX +XXX,XX @@ static int64_t ssh_getlength(BlockDriverState *bs) | ||
242 | |||
243 | static int coroutine_fn ssh_co_truncate(BlockDriverState *bs, int64_t offset, | ||
244 | bool exact, PreallocMode prealloc, | ||
245 | - Error **errp) | ||
246 | + BdrvRequestFlags flags, Error **errp) | ||
247 | { | ||
248 | BDRVSSHState *s = bs->opaque; | ||
249 | |||
250 | diff --git a/tests/test-block-iothread.c b/tests/test-block-iothread.c | ||
251 | index XXXXXXX..XXXXXXX 100644 | ||
252 | --- a/tests/test-block-iothread.c | ||
253 | +++ b/tests/test-block-iothread.c | ||
254 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn bdrv_test_co_pdiscard(BlockDriverState *bs, | ||
255 | |||
256 | static int coroutine_fn | ||
257 | bdrv_test_co_truncate(BlockDriverState *bs, int64_t offset, bool exact, | ||
258 | - PreallocMode prealloc, Error **errp) | ||
259 | + PreallocMode prealloc, BdrvRequestFlags flags, | ||
260 | + Error **errp) | ||
261 | { | ||
262 | return 0; | ||
263 | } | 68 | } |
264 | -- | 69 | -- |
265 | 2.25.3 | 70 | 2.31.1 |
266 | 71 | ||
267 | 72 | diff view generated by jsdifflib |
1 | Now that node level interface bdrv_truncate() supports passing request | 1 | From: Hanna Reitz <hreitz@redhat.com> |
---|---|---|---|
2 | flags to the block driver, expose this on the BlockBackend level, too. | ||
3 | 2 | ||
4 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 3 | Similarly to e7e588d432d31ecebc26358e47201dd108db964c, there is a |
5 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 4 | warning in block/block-backend.c that qiov->size <= INT64_MAX is always |
6 | Reviewed-by: Alberto Garcia <berto@igalia.com> | 5 | true on machines where size_t is narrower than a uint64_t. In said |
7 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 6 | commit, we silenced this warning by casting to uint64_t. |
8 | Message-Id: <20200424125448.63318-4-kwolf@redhat.com> | 7 | |
8 | The commit introducing this warning here | ||
9 | (a93d81c84afa717b0a1a6947524d8d1fbfd6bbf5) anticipated it and so tried | ||
10 | to address it the same way. However, it only did so in one of two | ||
11 | places where this comparison occurs, and so we still need to fix up the | ||
12 | other one. | ||
13 | |||
14 | Fixes: a93d81c84afa717b0a1a6947524d8d1fbfd6bbf5 | ||
15 | ("block-backend: convert blk_aio_ functions to int64_t bytes | ||
16 | paramter") | ||
17 | Signed-off-by: Hanna Reitz <hreitz@redhat.com> | ||
18 | Message-Id: <20211026090745.30800-1-hreitz@redhat.com> | ||
19 | Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com> | ||
9 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 20 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
10 | --- | 21 | --- |
11 | include/sysemu/block-backend.h | 2 +- | 22 | block/block-backend.c | 2 +- |
12 | block.c | 3 ++- | 23 | 1 file changed, 1 insertion(+), 1 deletion(-) |
13 | block/block-backend.c | 4 ++-- | ||
14 | block/commit.c | 4 ++-- | ||
15 | block/crypto.c | 2 +- | ||
16 | block/mirror.c | 2 +- | ||
17 | block/qcow2.c | 4 ++-- | ||
18 | block/qed.c | 2 +- | ||
19 | block/vdi.c | 2 +- | ||
20 | block/vhdx.c | 4 ++-- | ||
21 | block/vmdk.c | 6 +++--- | ||
22 | block/vpc.c | 2 +- | ||
23 | blockdev.c | 2 +- | ||
24 | qemu-img.c | 2 +- | ||
25 | qemu-io-cmds.c | 2 +- | ||
26 | 15 files changed, 22 insertions(+), 21 deletions(-) | ||
27 | 24 | ||
28 | diff --git a/include/sysemu/block-backend.h b/include/sysemu/block-backend.h | ||
29 | index XXXXXXX..XXXXXXX 100644 | ||
30 | --- a/include/sysemu/block-backend.h | ||
31 | +++ b/include/sysemu/block-backend.h | ||
32 | @@ -XXX,XX +XXX,XX @@ int coroutine_fn blk_co_pwrite_zeroes(BlockBackend *blk, int64_t offset, | ||
33 | int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, | ||
34 | int bytes); | ||
35 | int blk_truncate(BlockBackend *blk, int64_t offset, bool exact, | ||
36 | - PreallocMode prealloc, Error **errp); | ||
37 | + PreallocMode prealloc, BdrvRequestFlags flags, Error **errp); | ||
38 | int blk_pdiscard(BlockBackend *blk, int64_t offset, int bytes); | ||
39 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, | ||
40 | int64_t pos, int size); | ||
41 | diff --git a/block.c b/block.c | ||
42 | index XXXXXXX..XXXXXXX 100644 | ||
43 | --- a/block.c | ||
44 | +++ b/block.c | ||
45 | @@ -XXX,XX +XXX,XX @@ static int64_t create_file_fallback_truncate(BlockBackend *blk, | ||
46 | int64_t size; | ||
47 | int ret; | ||
48 | |||
49 | - ret = blk_truncate(blk, minimum_size, false, PREALLOC_MODE_OFF, &local_err); | ||
50 | + ret = blk_truncate(blk, minimum_size, false, PREALLOC_MODE_OFF, 0, | ||
51 | + &local_err); | ||
52 | if (ret < 0 && ret != -ENOTSUP) { | ||
53 | error_propagate(errp, local_err); | ||
54 | return ret; | ||
55 | diff --git a/block/block-backend.c b/block/block-backend.c | 25 | diff --git a/block/block-backend.c b/block/block-backend.c |
56 | index XXXXXXX..XXXXXXX 100644 | 26 | index XXXXXXX..XXXXXXX 100644 |
57 | --- a/block/block-backend.c | 27 | --- a/block/block-backend.c |
58 | +++ b/block/block-backend.c | 28 | +++ b/block/block-backend.c |
59 | @@ -XXX,XX +XXX,XX @@ int blk_pwrite_compressed(BlockBackend *blk, int64_t offset, const void *buf, | 29 | @@ -XXX,XX +XXX,XX @@ BlockAIOCB *blk_aio_pwritev(BlockBackend *blk, int64_t offset, |
30 | QEMUIOVector *qiov, BdrvRequestFlags flags, | ||
31 | BlockCompletionFunc *cb, void *opaque) | ||
32 | { | ||
33 | - assert(qiov->size <= INT64_MAX); | ||
34 | + assert((uint64_t)qiov->size <= INT64_MAX); | ||
35 | return blk_aio_prwv(blk, offset, qiov->size, qiov, | ||
36 | blk_aio_write_entry, flags, cb, opaque); | ||
60 | } | 37 | } |
61 | |||
62 | int blk_truncate(BlockBackend *blk, int64_t offset, bool exact, | ||
63 | - PreallocMode prealloc, Error **errp) | ||
64 | + PreallocMode prealloc, BdrvRequestFlags flags, Error **errp) | ||
65 | { | ||
66 | if (!blk_is_available(blk)) { | ||
67 | error_setg(errp, "No medium inserted"); | ||
68 | return -ENOMEDIUM; | ||
69 | } | ||
70 | |||
71 | - return bdrv_truncate(blk->root, offset, exact, prealloc, 0, errp); | ||
72 | + return bdrv_truncate(blk->root, offset, exact, prealloc, flags, errp); | ||
73 | } | ||
74 | |||
75 | int blk_save_vmstate(BlockBackend *blk, const uint8_t *buf, | ||
76 | diff --git a/block/commit.c b/block/commit.c | ||
77 | index XXXXXXX..XXXXXXX 100644 | ||
78 | --- a/block/commit.c | ||
79 | +++ b/block/commit.c | ||
80 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn commit_run(Job *job, Error **errp) | ||
81 | } | ||
82 | |||
83 | if (base_len < len) { | ||
84 | - ret = blk_truncate(s->base, len, false, PREALLOC_MODE_OFF, NULL); | ||
85 | + ret = blk_truncate(s->base, len, false, PREALLOC_MODE_OFF, 0, NULL); | ||
86 | if (ret) { | ||
87 | goto out; | ||
88 | } | ||
89 | @@ -XXX,XX +XXX,XX @@ int bdrv_commit(BlockDriverState *bs) | ||
90 | * grow the backing file image if possible. If not possible, | ||
91 | * we must return an error */ | ||
92 | if (length > backing_length) { | ||
93 | - ret = blk_truncate(backing, length, false, PREALLOC_MODE_OFF, | ||
94 | + ret = blk_truncate(backing, length, false, PREALLOC_MODE_OFF, 0, | ||
95 | &local_err); | ||
96 | if (ret < 0) { | ||
97 | error_report_err(local_err); | ||
98 | diff --git a/block/crypto.c b/block/crypto.c | ||
99 | index XXXXXXX..XXXXXXX 100644 | ||
100 | --- a/block/crypto.c | ||
101 | +++ b/block/crypto.c | ||
102 | @@ -XXX,XX +XXX,XX @@ static ssize_t block_crypto_init_func(QCryptoBlock *block, | ||
103 | * which will be used by the crypto header | ||
104 | */ | ||
105 | return blk_truncate(data->blk, data->size + headerlen, false, | ||
106 | - data->prealloc, errp); | ||
107 | + data->prealloc, 0, errp); | ||
108 | } | ||
109 | |||
110 | |||
111 | diff --git a/block/mirror.c b/block/mirror.c | ||
112 | index XXXXXXX..XXXXXXX 100644 | ||
113 | --- a/block/mirror.c | ||
114 | +++ b/block/mirror.c | ||
115 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn mirror_run(Job *job, Error **errp) | ||
116 | |||
117 | if (s->bdev_length > base_length) { | ||
118 | ret = blk_truncate(s->target, s->bdev_length, false, | ||
119 | - PREALLOC_MODE_OFF, NULL); | ||
120 | + PREALLOC_MODE_OFF, 0, NULL); | ||
121 | if (ret < 0) { | ||
122 | goto immediate_exit; | ||
123 | } | ||
124 | diff --git a/block/qcow2.c b/block/qcow2.c | ||
125 | index XXXXXXX..XXXXXXX 100644 | ||
126 | --- a/block/qcow2.c | ||
127 | +++ b/block/qcow2.c | ||
128 | @@ -XXX,XX +XXX,XX @@ qcow2_co_create(BlockdevCreateOptions *create_options, Error **errp) | ||
129 | |||
130 | /* Okay, now that we have a valid image, let's give it the right size */ | ||
131 | ret = blk_truncate(blk, qcow2_opts->size, false, qcow2_opts->preallocation, | ||
132 | - errp); | ||
133 | + 0, errp); | ||
134 | if (ret < 0) { | ||
135 | error_prepend(errp, "Could not resize image: "); | ||
136 | goto out; | ||
137 | @@ -XXX,XX +XXX,XX @@ static int qcow2_amend_options(BlockDriverState *bs, QemuOpts *opts, | ||
138 | * Amending image options should ensure that the image has | ||
139 | * exactly the given new values, so pass exact=true here. | ||
140 | */ | ||
141 | - ret = blk_truncate(blk, new_size, true, PREALLOC_MODE_OFF, errp); | ||
142 | + ret = blk_truncate(blk, new_size, true, PREALLOC_MODE_OFF, 0, errp); | ||
143 | blk_unref(blk); | ||
144 | if (ret < 0) { | ||
145 | return ret; | ||
146 | diff --git a/block/qed.c b/block/qed.c | ||
147 | index XXXXXXX..XXXXXXX 100644 | ||
148 | --- a/block/qed.c | ||
149 | +++ b/block/qed.c | ||
150 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn bdrv_qed_co_create(BlockdevCreateOptions *opts, | ||
151 | * The QED format associates file length with allocation status, | ||
152 | * so a new file (which is empty) must have a length of 0. | ||
153 | */ | ||
154 | - ret = blk_truncate(blk, 0, true, PREALLOC_MODE_OFF, errp); | ||
155 | + ret = blk_truncate(blk, 0, true, PREALLOC_MODE_OFF, 0, errp); | ||
156 | if (ret < 0) { | ||
157 | goto out; | ||
158 | } | ||
159 | diff --git a/block/vdi.c b/block/vdi.c | ||
160 | index XXXXXXX..XXXXXXX 100644 | ||
161 | --- a/block/vdi.c | ||
162 | +++ b/block/vdi.c | ||
163 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn vdi_co_do_create(BlockdevCreateOptions *create_options, | ||
164 | |||
165 | if (image_type == VDI_TYPE_STATIC) { | ||
166 | ret = blk_truncate(blk, offset + blocks * block_size, false, | ||
167 | - PREALLOC_MODE_OFF, errp); | ||
168 | + PREALLOC_MODE_OFF, 0, errp); | ||
169 | if (ret < 0) { | ||
170 | error_prepend(errp, "Failed to statically allocate file"); | ||
171 | goto exit; | ||
172 | diff --git a/block/vhdx.c b/block/vhdx.c | ||
173 | index XXXXXXX..XXXXXXX 100644 | ||
174 | --- a/block/vhdx.c | ||
175 | +++ b/block/vhdx.c | ||
176 | @@ -XXX,XX +XXX,XX @@ static int vhdx_create_bat(BlockBackend *blk, BDRVVHDXState *s, | ||
177 | /* All zeroes, so we can just extend the file - the end of the BAT | ||
178 | * is the furthest thing we have written yet */ | ||
179 | ret = blk_truncate(blk, data_file_offset, false, PREALLOC_MODE_OFF, | ||
180 | - errp); | ||
181 | + 0, errp); | ||
182 | if (ret < 0) { | ||
183 | goto exit; | ||
184 | } | ||
185 | } else if (type == VHDX_TYPE_FIXED) { | ||
186 | ret = blk_truncate(blk, data_file_offset + image_size, false, | ||
187 | - PREALLOC_MODE_OFF, errp); | ||
188 | + PREALLOC_MODE_OFF, 0, errp); | ||
189 | if (ret < 0) { | ||
190 | goto exit; | ||
191 | } | ||
192 | diff --git a/block/vmdk.c b/block/vmdk.c | ||
193 | index XXXXXXX..XXXXXXX 100644 | ||
194 | --- a/block/vmdk.c | ||
195 | +++ b/block/vmdk.c | ||
196 | @@ -XXX,XX +XXX,XX @@ static int vmdk_init_extent(BlockBackend *blk, | ||
197 | int gd_buf_size; | ||
198 | |||
199 | if (flat) { | ||
200 | - ret = blk_truncate(blk, filesize, false, PREALLOC_MODE_OFF, errp); | ||
201 | + ret = blk_truncate(blk, filesize, false, PREALLOC_MODE_OFF, 0, errp); | ||
202 | goto exit; | ||
203 | } | ||
204 | magic = cpu_to_be32(VMDK4_MAGIC); | ||
205 | @@ -XXX,XX +XXX,XX @@ static int vmdk_init_extent(BlockBackend *blk, | ||
206 | } | ||
207 | |||
208 | ret = blk_truncate(blk, le64_to_cpu(header.grain_offset) << 9, false, | ||
209 | - PREALLOC_MODE_OFF, errp); | ||
210 | + PREALLOC_MODE_OFF, 0, errp); | ||
211 | if (ret < 0) { | ||
212 | goto exit; | ||
213 | } | ||
214 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn vmdk_co_do_create(int64_t size, | ||
215 | /* bdrv_pwrite write padding zeros to align to sector, we don't need that | ||
216 | * for description file */ | ||
217 | if (desc_offset == 0) { | ||
218 | - ret = blk_truncate(blk, desc_len, false, PREALLOC_MODE_OFF, errp); | ||
219 | + ret = blk_truncate(blk, desc_len, false, PREALLOC_MODE_OFF, 0, errp); | ||
220 | if (ret < 0) { | ||
221 | goto exit; | ||
222 | } | ||
223 | diff --git a/block/vpc.c b/block/vpc.c | ||
224 | index XXXXXXX..XXXXXXX 100644 | ||
225 | --- a/block/vpc.c | ||
226 | +++ b/block/vpc.c | ||
227 | @@ -XXX,XX +XXX,XX @@ static int create_fixed_disk(BlockBackend *blk, uint8_t *buf, | ||
228 | /* Add footer to total size */ | ||
229 | total_size += HEADER_SIZE; | ||
230 | |||
231 | - ret = blk_truncate(blk, total_size, false, PREALLOC_MODE_OFF, errp); | ||
232 | + ret = blk_truncate(blk, total_size, false, PREALLOC_MODE_OFF, 0, errp); | ||
233 | if (ret < 0) { | ||
234 | return ret; | ||
235 | } | ||
236 | diff --git a/blockdev.c b/blockdev.c | ||
237 | index XXXXXXX..XXXXXXX 100644 | ||
238 | --- a/blockdev.c | ||
239 | +++ b/blockdev.c | ||
240 | @@ -XXX,XX +XXX,XX @@ void qmp_block_resize(bool has_device, const char *device, | ||
241 | } | ||
242 | |||
243 | bdrv_drained_begin(bs); | ||
244 | - ret = blk_truncate(blk, size, false, PREALLOC_MODE_OFF, errp); | ||
245 | + ret = blk_truncate(blk, size, false, PREALLOC_MODE_OFF, 0, errp); | ||
246 | bdrv_drained_end(bs); | ||
247 | |||
248 | out: | ||
249 | diff --git a/qemu-img.c b/qemu-img.c | ||
250 | index XXXXXXX..XXXXXXX 100644 | ||
251 | --- a/qemu-img.c | ||
252 | +++ b/qemu-img.c | ||
253 | @@ -XXX,XX +XXX,XX @@ static int img_resize(int argc, char **argv) | ||
254 | * resizing, so pass @exact=true. It is of no use to report | ||
255 | * success when the image has not actually been resized. | ||
256 | */ | ||
257 | - ret = blk_truncate(blk, total_size, true, prealloc, &err); | ||
258 | + ret = blk_truncate(blk, total_size, true, prealloc, 0, &err); | ||
259 | if (!ret) { | ||
260 | qprintf(quiet, "Image resized.\n"); | ||
261 | } else { | ||
262 | diff --git a/qemu-io-cmds.c b/qemu-io-cmds.c | ||
263 | index XXXXXXX..XXXXXXX 100644 | ||
264 | --- a/qemu-io-cmds.c | ||
265 | +++ b/qemu-io-cmds.c | ||
266 | @@ -XXX,XX +XXX,XX @@ static int truncate_f(BlockBackend *blk, int argc, char **argv) | ||
267 | * exact=true. It is better to err on the "emit more errors" side | ||
268 | * than to be overly permissive. | ||
269 | */ | ||
270 | - ret = blk_truncate(blk, offset, true, PREALLOC_MODE_OFF, &local_err); | ||
271 | + ret = blk_truncate(blk, offset, true, PREALLOC_MODE_OFF, 0, &local_err); | ||
272 | if (ret < 0) { | ||
273 | error_report_err(local_err); | ||
274 | return ret; | ||
275 | -- | 38 | -- |
276 | 2.25.3 | 39 | 2.31.1 |
277 | 40 | ||
278 | 41 | diff view generated by jsdifflib |
Deleted patch | |||
---|---|---|---|
1 | If BDRV_REQ_ZERO_WRITE is set and we're extending the image, calling | ||
2 | qcow2_cluster_zeroize() with flags=0 does the right thing: It doesn't | ||
3 | undo any previous preallocation, but just adds the zero flag to all | ||
4 | relevant L2 entries. If an external data file is in use, a write_zeroes | ||
5 | request to the data file is made instead. | ||
6 | 1 | ||
7 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | ||
8 | Message-Id: <20200424125448.63318-5-kwolf@redhat.com> | ||
9 | Reviewed-by: Eric Blake <eblake@redhat.com> | ||
10 | Reviewed-by: Max Reitz <mreitz@redhat.com> | ||
11 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | ||
12 | --- | ||
13 | block/qcow2-cluster.c | 2 +- | ||
14 | block/qcow2.c | 34 ++++++++++++++++++++++++++++++++++ | ||
15 | 2 files changed, 35 insertions(+), 1 deletion(-) | ||
16 | |||
17 | diff --git a/block/qcow2-cluster.c b/block/qcow2-cluster.c | ||
18 | index XXXXXXX..XXXXXXX 100644 | ||
19 | --- a/block/qcow2-cluster.c | ||
20 | +++ b/block/qcow2-cluster.c | ||
21 | @@ -XXX,XX +XXX,XX @@ int qcow2_cluster_zeroize(BlockDriverState *bs, uint64_t offset, | ||
22 | /* Caller must pass aligned values, except at image end */ | ||
23 | assert(QEMU_IS_ALIGNED(offset, s->cluster_size)); | ||
24 | assert(QEMU_IS_ALIGNED(end_offset, s->cluster_size) || | ||
25 | - end_offset == bs->total_sectors << BDRV_SECTOR_BITS); | ||
26 | + end_offset >= bs->total_sectors << BDRV_SECTOR_BITS); | ||
27 | |||
28 | /* The zero flag is only supported by version 3 and newer */ | ||
29 | if (s->qcow_version < 3) { | ||
30 | diff --git a/block/qcow2.c b/block/qcow2.c | ||
31 | index XXXXXXX..XXXXXXX 100644 | ||
32 | --- a/block/qcow2.c | ||
33 | +++ b/block/qcow2.c | ||
34 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow2_do_open(BlockDriverState *bs, QDict *options, | ||
35 | |||
36 | bs->supported_zero_flags = header.version >= 3 ? | ||
37 | BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK : 0; | ||
38 | + bs->supported_truncate_flags = BDRV_REQ_ZERO_WRITE; | ||
39 | |||
40 | /* Repair image if dirty */ | ||
41 | if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only && | ||
42 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn qcow2_co_truncate(BlockDriverState *bs, int64_t offset, | ||
43 | g_assert_not_reached(); | ||
44 | } | ||
45 | |||
46 | + if ((flags & BDRV_REQ_ZERO_WRITE) && offset > old_length) { | ||
47 | + uint64_t zero_start = QEMU_ALIGN_UP(old_length, s->cluster_size); | ||
48 | + | ||
49 | + /* | ||
50 | + * Use zero clusters as much as we can. qcow2_cluster_zeroize() | ||
51 | + * requires a cluster-aligned start. The end may be unaligned if it is | ||
52 | + * at the end of the image (which it is here). | ||
53 | + */ | ||
54 | + ret = qcow2_cluster_zeroize(bs, zero_start, offset - zero_start, 0); | ||
55 | + if (ret < 0) { | ||
56 | + error_setg_errno(errp, -ret, "Failed to zero out new clusters"); | ||
57 | + goto fail; | ||
58 | + } | ||
59 | + | ||
60 | + /* Write explicit zeros for the unaligned head */ | ||
61 | + if (zero_start > old_length) { | ||
62 | + uint64_t len = zero_start - old_length; | ||
63 | + uint8_t *buf = qemu_blockalign0(bs, len); | ||
64 | + QEMUIOVector qiov; | ||
65 | + qemu_iovec_init_buf(&qiov, buf, len); | ||
66 | + | ||
67 | + qemu_co_mutex_unlock(&s->lock); | ||
68 | + ret = qcow2_co_pwritev_part(bs, old_length, len, &qiov, 0, 0); | ||
69 | + qemu_co_mutex_lock(&s->lock); | ||
70 | + | ||
71 | + qemu_vfree(buf); | ||
72 | + if (ret < 0) { | ||
73 | + error_setg_errno(errp, -ret, "Failed to zero out the new area"); | ||
74 | + goto fail; | ||
75 | + } | ||
76 | + } | ||
77 | + } | ||
78 | + | ||
79 | if (prealloc != PREALLOC_MODE_OFF) { | ||
80 | /* Flush metadata before actually changing the image size */ | ||
81 | ret = qcow2_write_caches(bs); | ||
82 | -- | ||
83 | 2.25.3 | ||
84 | |||
85 | diff view generated by jsdifflib |
1 | From: Andrzej Jakowski <andrzej.jakowski@linux.intel.com> | 1 | From: Philippe Mathieu-Daudé <philmd@redhat.com> |
---|---|---|---|
2 | 2 | ||
3 | This patch introduces support for PMR that has been defined as part of NVMe 1.4 | 3 | Since commit 4d324c0bf65 ("introduce QEMU_AUTO_VFREE") buffers |
4 | spec. User can now specify a pmrdev option that should point to HostMemoryBackend. | 4 | allocated by qemu_memalign() can automatically freed when using |
5 | pmrdev memory region will subsequently be exposed as PCI BAR 2 in emulated NVMe | 5 | the QEMU_AUTO_VFREE macro. Use it to simplify a bit. |
6 | device. Guest OS can perform mmio read and writes to the PMR region that will stay | ||
7 | persistent across system reboot. | ||
8 | 6 | ||
9 | Signed-off-by: Andrzej Jakowski <andrzej.jakowski@linux.intel.com> | 7 | Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> |
10 | Reviewed-by: Klaus Jensen <k.jensen@samsung.com> | 8 | Message-Id: <20211006164931.172349-2-philmd@redhat.com> |
11 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | 9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
12 | Message-Id: <20200330164656.9348-1-andrzej.jakowski@linux.intel.com> | ||
13 | Reviewed-by: Keith Busch <kbusch@kernel.org> | ||
14 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 10 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
15 | --- | 11 | --- |
16 | hw/block/nvme.h | 2 + | 12 | block/nvme.c | 11 ++++------- |
17 | include/block/nvme.h | 172 +++++++++++++++++++++++++++++++++++++++++ | 13 | 1 file changed, 4 insertions(+), 7 deletions(-) |
18 | hw/block/nvme.c | 109 ++++++++++++++++++++++++++ | ||
19 | hw/block/Makefile.objs | 2 +- | ||
20 | hw/block/trace-events | 4 + | ||
21 | 5 files changed, 288 insertions(+), 1 deletion(-) | ||
22 | 14 | ||
23 | diff --git a/hw/block/nvme.h b/hw/block/nvme.h | 15 | diff --git a/block/nvme.c b/block/nvme.c |
24 | index XXXXXXX..XXXXXXX 100644 | 16 | index XXXXXXX..XXXXXXX 100644 |
25 | --- a/hw/block/nvme.h | 17 | --- a/block/nvme.c |
26 | +++ b/hw/block/nvme.h | 18 | +++ b/block/nvme.c |
27 | @@ -XXX,XX +XXX,XX @@ typedef struct NvmeCtrl { | 19 | @@ -XXX,XX +XXX,XX @@ static bool nvme_identify(BlockDriverState *bs, int namespace, Error **errp) |
28 | uint64_t timestamp_set_qemu_clock_ms; /* QEMU clock time */ | 20 | { |
29 | 21 | BDRVNVMeState *s = bs->opaque; | |
30 | char *serial; | 22 | bool ret = false; |
31 | + HostMemoryBackend *pmrdev; | 23 | - union { |
32 | + | 24 | + QEMU_AUTO_VFREE union { |
33 | NvmeNamespace *namespaces; | 25 | NvmeIdCtrl ctrl; |
34 | NvmeSQueue **sq; | 26 | NvmeIdNs ns; |
35 | NvmeCQueue **cq; | 27 | - } *id; |
36 | diff --git a/include/block/nvme.h b/include/block/nvme.h | 28 | + } *id = NULL; |
37 | index XXXXXXX..XXXXXXX 100644 | 29 | NvmeLBAF *lbaf; |
38 | --- a/include/block/nvme.h | 30 | uint16_t oncs; |
39 | +++ b/include/block/nvme.h | 31 | int r; |
40 | @@ -XXX,XX +XXX,XX @@ typedef struct NvmeBar { | 32 | @@ -XXX,XX +XXX,XX @@ static bool nvme_identify(BlockDriverState *bs, int namespace, Error **errp) |
41 | uint64_t acq; | 33 | s->blkshift = lbaf->ds; |
42 | uint32_t cmbloc; | 34 | out: |
43 | uint32_t cmbsz; | 35 | qemu_vfio_dma_unmap(s->vfio, id); |
44 | + uint8_t padding[3520]; /* not used by QEMU */ | 36 | - qemu_vfree(id); |
45 | + uint32_t pmrcap; | 37 | |
46 | + uint32_t pmrctl; | 38 | return ret; |
47 | + uint32_t pmrsts; | 39 | } |
48 | + uint32_t pmrebs; | 40 | @@ -XXX,XX +XXX,XX @@ static int nvme_co_prw(BlockDriverState *bs, uint64_t offset, uint64_t bytes, |
49 | + uint32_t pmrswtp; | 41 | { |
50 | + uint32_t pmrmsc; | 42 | BDRVNVMeState *s = bs->opaque; |
51 | } NvmeBar; | 43 | int r; |
52 | 44 | - uint8_t *buf = NULL; | |
53 | enum NvmeCapShift { | 45 | + QEMU_AUTO_VFREE uint8_t *buf = NULL; |
54 | @@ -XXX,XX +XXX,XX @@ enum NvmeCapShift { | 46 | QEMUIOVector local_qiov; |
55 | CAP_CSS_SHIFT = 37, | 47 | size_t len = QEMU_ALIGN_UP(bytes, qemu_real_host_page_size); |
56 | CAP_MPSMIN_SHIFT = 48, | 48 | assert(QEMU_IS_ALIGNED(offset, s->page_size)); |
57 | CAP_MPSMAX_SHIFT = 52, | 49 | @@ -XXX,XX +XXX,XX @@ static int nvme_co_prw(BlockDriverState *bs, uint64_t offset, uint64_t bytes, |
58 | + CAP_PMR_SHIFT = 56, | 50 | if (!r && !is_write) { |
59 | }; | 51 | qemu_iovec_from_buf(qiov, 0, buf, bytes); |
60 | |||
61 | enum NvmeCapMask { | ||
62 | @@ -XXX,XX +XXX,XX @@ enum NvmeCapMask { | ||
63 | CAP_CSS_MASK = 0xff, | ||
64 | CAP_MPSMIN_MASK = 0xf, | ||
65 | CAP_MPSMAX_MASK = 0xf, | ||
66 | + CAP_PMR_MASK = 0x1, | ||
67 | }; | ||
68 | |||
69 | #define NVME_CAP_MQES(cap) (((cap) >> CAP_MQES_SHIFT) & CAP_MQES_MASK) | ||
70 | @@ -XXX,XX +XXX,XX @@ enum NvmeCapMask { | ||
71 | << CAP_MPSMIN_SHIFT) | ||
72 | #define NVME_CAP_SET_MPSMAX(cap, val) (cap |= (uint64_t)(val & CAP_MPSMAX_MASK)\ | ||
73 | << CAP_MPSMAX_SHIFT) | ||
74 | +#define NVME_CAP_SET_PMRS(cap, val) (cap |= (uint64_t)(val & CAP_PMR_MASK)\ | ||
75 | + << CAP_PMR_SHIFT) | ||
76 | |||
77 | enum NvmeCcShift { | ||
78 | CC_EN_SHIFT = 0, | ||
79 | @@ -XXX,XX +XXX,XX @@ enum NvmeCmbszMask { | ||
80 | #define NVME_CMBSZ_GETSIZE(cmbsz) \ | ||
81 | (NVME_CMBSZ_SZ(cmbsz) * (1 << (12 + 4 * NVME_CMBSZ_SZU(cmbsz)))) | ||
82 | |||
83 | +enum NvmePmrcapShift { | ||
84 | + PMRCAP_RDS_SHIFT = 3, | ||
85 | + PMRCAP_WDS_SHIFT = 4, | ||
86 | + PMRCAP_BIR_SHIFT = 5, | ||
87 | + PMRCAP_PMRTU_SHIFT = 8, | ||
88 | + PMRCAP_PMRWBM_SHIFT = 10, | ||
89 | + PMRCAP_PMRTO_SHIFT = 16, | ||
90 | + PMRCAP_CMSS_SHIFT = 24, | ||
91 | +}; | ||
92 | + | ||
93 | +enum NvmePmrcapMask { | ||
94 | + PMRCAP_RDS_MASK = 0x1, | ||
95 | + PMRCAP_WDS_MASK = 0x1, | ||
96 | + PMRCAP_BIR_MASK = 0x7, | ||
97 | + PMRCAP_PMRTU_MASK = 0x3, | ||
98 | + PMRCAP_PMRWBM_MASK = 0xf, | ||
99 | + PMRCAP_PMRTO_MASK = 0xff, | ||
100 | + PMRCAP_CMSS_MASK = 0x1, | ||
101 | +}; | ||
102 | + | ||
103 | +#define NVME_PMRCAP_RDS(pmrcap) \ | ||
104 | + ((pmrcap >> PMRCAP_RDS_SHIFT) & PMRCAP_RDS_MASK) | ||
105 | +#define NVME_PMRCAP_WDS(pmrcap) \ | ||
106 | + ((pmrcap >> PMRCAP_WDS_SHIFT) & PMRCAP_WDS_MASK) | ||
107 | +#define NVME_PMRCAP_BIR(pmrcap) \ | ||
108 | + ((pmrcap >> PMRCAP_BIR_SHIFT) & PMRCAP_BIR_MASK) | ||
109 | +#define NVME_PMRCAP_PMRTU(pmrcap) \ | ||
110 | + ((pmrcap >> PMRCAP_PMRTU_SHIFT) & PMRCAP_PMRTU_MASK) | ||
111 | +#define NVME_PMRCAP_PMRWBM(pmrcap) \ | ||
112 | + ((pmrcap >> PMRCAP_PMRWBM_SHIFT) & PMRCAP_PMRWBM_MASK) | ||
113 | +#define NVME_PMRCAP_PMRTO(pmrcap) \ | ||
114 | + ((pmrcap >> PMRCAP_PMRTO_SHIFT) & PMRCAP_PMRTO_MASK) | ||
115 | +#define NVME_PMRCAP_CMSS(pmrcap) \ | ||
116 | + ((pmrcap >> PMRCAP_CMSS_SHIFT) & PMRCAP_CMSS_MASK) | ||
117 | + | ||
118 | +#define NVME_PMRCAP_SET_RDS(pmrcap, val) \ | ||
119 | + (pmrcap |= (uint64_t)(val & PMRCAP_RDS_MASK) << PMRCAP_RDS_SHIFT) | ||
120 | +#define NVME_PMRCAP_SET_WDS(pmrcap, val) \ | ||
121 | + (pmrcap |= (uint64_t)(val & PMRCAP_WDS_MASK) << PMRCAP_WDS_SHIFT) | ||
122 | +#define NVME_PMRCAP_SET_BIR(pmrcap, val) \ | ||
123 | + (pmrcap |= (uint64_t)(val & PMRCAP_BIR_MASK) << PMRCAP_BIR_SHIFT) | ||
124 | +#define NVME_PMRCAP_SET_PMRTU(pmrcap, val) \ | ||
125 | + (pmrcap |= (uint64_t)(val & PMRCAP_PMRTU_MASK) << PMRCAP_PMRTU_SHIFT) | ||
126 | +#define NVME_PMRCAP_SET_PMRWBM(pmrcap, val) \ | ||
127 | + (pmrcap |= (uint64_t)(val & PMRCAP_PMRWBM_MASK) << PMRCAP_PMRWBM_SHIFT) | ||
128 | +#define NVME_PMRCAP_SET_PMRTO(pmrcap, val) \ | ||
129 | + (pmrcap |= (uint64_t)(val & PMRCAP_PMRTO_MASK) << PMRCAP_PMRTO_SHIFT) | ||
130 | +#define NVME_PMRCAP_SET_CMSS(pmrcap, val) \ | ||
131 | + (pmrcap |= (uint64_t)(val & PMRCAP_CMSS_MASK) << PMRCAP_CMSS_SHIFT) | ||
132 | + | ||
133 | +enum NvmePmrctlShift { | ||
134 | + PMRCTL_EN_SHIFT = 0, | ||
135 | +}; | ||
136 | + | ||
137 | +enum NvmePmrctlMask { | ||
138 | + PMRCTL_EN_MASK = 0x1, | ||
139 | +}; | ||
140 | + | ||
141 | +#define NVME_PMRCTL_EN(pmrctl) ((pmrctl >> PMRCTL_EN_SHIFT) & PMRCTL_EN_MASK) | ||
142 | + | ||
143 | +#define NVME_PMRCTL_SET_EN(pmrctl, val) \ | ||
144 | + (pmrctl |= (uint64_t)(val & PMRCTL_EN_MASK) << PMRCTL_EN_SHIFT) | ||
145 | + | ||
146 | +enum NvmePmrstsShift { | ||
147 | + PMRSTS_ERR_SHIFT = 0, | ||
148 | + PMRSTS_NRDY_SHIFT = 8, | ||
149 | + PMRSTS_HSTS_SHIFT = 9, | ||
150 | + PMRSTS_CBAI_SHIFT = 12, | ||
151 | +}; | ||
152 | + | ||
153 | +enum NvmePmrstsMask { | ||
154 | + PMRSTS_ERR_MASK = 0xff, | ||
155 | + PMRSTS_NRDY_MASK = 0x1, | ||
156 | + PMRSTS_HSTS_MASK = 0x7, | ||
157 | + PMRSTS_CBAI_MASK = 0x1, | ||
158 | +}; | ||
159 | + | ||
160 | +#define NVME_PMRSTS_ERR(pmrsts) \ | ||
161 | + ((pmrsts >> PMRSTS_ERR_SHIFT) & PMRSTS_ERR_MASK) | ||
162 | +#define NVME_PMRSTS_NRDY(pmrsts) \ | ||
163 | + ((pmrsts >> PMRSTS_NRDY_SHIFT) & PMRSTS_NRDY_MASK) | ||
164 | +#define NVME_PMRSTS_HSTS(pmrsts) \ | ||
165 | + ((pmrsts >> PMRSTS_HSTS_SHIFT) & PMRSTS_HSTS_MASK) | ||
166 | +#define NVME_PMRSTS_CBAI(pmrsts) \ | ||
167 | + ((pmrsts >> PMRSTS_CBAI_SHIFT) & PMRSTS_CBAI_MASK) | ||
168 | + | ||
169 | +#define NVME_PMRSTS_SET_ERR(pmrsts, val) \ | ||
170 | + (pmrsts |= (uint64_t)(val & PMRSTS_ERR_MASK) << PMRSTS_ERR_SHIFT) | ||
171 | +#define NVME_PMRSTS_SET_NRDY(pmrsts, val) \ | ||
172 | + (pmrsts |= (uint64_t)(val & PMRSTS_NRDY_MASK) << PMRSTS_NRDY_SHIFT) | ||
173 | +#define NVME_PMRSTS_SET_HSTS(pmrsts, val) \ | ||
174 | + (pmrsts |= (uint64_t)(val & PMRSTS_HSTS_MASK) << PMRSTS_HSTS_SHIFT) | ||
175 | +#define NVME_PMRSTS_SET_CBAI(pmrsts, val) \ | ||
176 | + (pmrsts |= (uint64_t)(val & PMRSTS_CBAI_MASK) << PMRSTS_CBAI_SHIFT) | ||
177 | + | ||
178 | +enum NvmePmrebsShift { | ||
179 | + PMREBS_PMRSZU_SHIFT = 0, | ||
180 | + PMREBS_RBB_SHIFT = 4, | ||
181 | + PMREBS_PMRWBZ_SHIFT = 8, | ||
182 | +}; | ||
183 | + | ||
184 | +enum NvmePmrebsMask { | ||
185 | + PMREBS_PMRSZU_MASK = 0xf, | ||
186 | + PMREBS_RBB_MASK = 0x1, | ||
187 | + PMREBS_PMRWBZ_MASK = 0xffffff, | ||
188 | +}; | ||
189 | + | ||
190 | +#define NVME_PMREBS_PMRSZU(pmrebs) \ | ||
191 | + ((pmrebs >> PMREBS_PMRSZU_SHIFT) & PMREBS_PMRSZU_MASK) | ||
192 | +#define NVME_PMREBS_RBB(pmrebs) \ | ||
193 | + ((pmrebs >> PMREBS_RBB_SHIFT) & PMREBS_RBB_MASK) | ||
194 | +#define NVME_PMREBS_PMRWBZ(pmrebs) \ | ||
195 | + ((pmrebs >> PMREBS_PMRWBZ_SHIFT) & PMREBS_PMRWBZ_MASK) | ||
196 | + | ||
197 | +#define NVME_PMREBS_SET_PMRSZU(pmrebs, val) \ | ||
198 | + (pmrebs |= (uint64_t)(val & PMREBS_PMRSZU_MASK) << PMREBS_PMRSZU_SHIFT) | ||
199 | +#define NVME_PMREBS_SET_RBB(pmrebs, val) \ | ||
200 | + (pmrebs |= (uint64_t)(val & PMREBS_RBB_MASK) << PMREBS_RBB_SHIFT) | ||
201 | +#define NVME_PMREBS_SET_PMRWBZ(pmrebs, val) \ | ||
202 | + (pmrebs |= (uint64_t)(val & PMREBS_PMRWBZ_MASK) << PMREBS_PMRWBZ_SHIFT) | ||
203 | + | ||
204 | +enum NvmePmrswtpShift { | ||
205 | + PMRSWTP_PMRSWTU_SHIFT = 0, | ||
206 | + PMRSWTP_PMRSWTV_SHIFT = 8, | ||
207 | +}; | ||
208 | + | ||
209 | +enum NvmePmrswtpMask { | ||
210 | + PMRSWTP_PMRSWTU_MASK = 0xf, | ||
211 | + PMRSWTP_PMRSWTV_MASK = 0xffffff, | ||
212 | +}; | ||
213 | + | ||
214 | +#define NVME_PMRSWTP_PMRSWTU(pmrswtp) \ | ||
215 | + ((pmrswtp >> PMRSWTP_PMRSWTU_SHIFT) & PMRSWTP_PMRSWTU_MASK) | ||
216 | +#define NVME_PMRSWTP_PMRSWTV(pmrswtp) \ | ||
217 | + ((pmrswtp >> PMRSWTP_PMRSWTV_SHIFT) & PMRSWTP_PMRSWTV_MASK) | ||
218 | + | ||
219 | +#define NVME_PMRSWTP_SET_PMRSWTU(pmrswtp, val) \ | ||
220 | + (pmrswtp |= (uint64_t)(val & PMRSWTP_PMRSWTU_MASK) << PMRSWTP_PMRSWTU_SHIFT) | ||
221 | +#define NVME_PMRSWTP_SET_PMRSWTV(pmrswtp, val) \ | ||
222 | + (pmrswtp |= (uint64_t)(val & PMRSWTP_PMRSWTV_MASK) << PMRSWTP_PMRSWTV_SHIFT) | ||
223 | + | ||
224 | +enum NvmePmrmscShift { | ||
225 | + PMRMSC_CMSE_SHIFT = 1, | ||
226 | + PMRMSC_CBA_SHIFT = 12, | ||
227 | +}; | ||
228 | + | ||
229 | +enum NvmePmrmscMask { | ||
230 | + PMRMSC_CMSE_MASK = 0x1, | ||
231 | + PMRMSC_CBA_MASK = 0xfffffffffffff, | ||
232 | +}; | ||
233 | + | ||
234 | +#define NVME_PMRMSC_CMSE(pmrmsc) \ | ||
235 | + ((pmrmsc >> PMRMSC_CMSE_SHIFT) & PMRMSC_CMSE_MASK) | ||
236 | +#define NVME_PMRMSC_CBA(pmrmsc) \ | ||
237 | + ((pmrmsc >> PMRMSC_CBA_SHIFT) & PMRMSC_CBA_MASK) | ||
238 | + | ||
239 | +#define NVME_PMRMSC_SET_CMSE(pmrmsc, val) \ | ||
240 | + (pmrmsc |= (uint64_t)(val & PMRMSC_CMSE_MASK) << PMRMSC_CMSE_SHIFT) | ||
241 | +#define NVME_PMRMSC_SET_CBA(pmrmsc, val) \ | ||
242 | + (pmrmsc |= (uint64_t)(val & PMRMSC_CBA_MASK) << PMRMSC_CBA_SHIFT) | ||
243 | + | ||
244 | typedef struct NvmeCmd { | ||
245 | uint8_t opcode; | ||
246 | uint8_t fuse; | ||
247 | diff --git a/hw/block/nvme.c b/hw/block/nvme.c | ||
248 | index XXXXXXX..XXXXXXX 100644 | ||
249 | --- a/hw/block/nvme.c | ||
250 | +++ b/hw/block/nvme.c | ||
251 | @@ -XXX,XX +XXX,XX @@ | ||
252 | * -drive file=<file>,if=none,id=<drive_id> | ||
253 | * -device nvme,drive=<drive_id>,serial=<serial>,id=<id[optional]>, \ | ||
254 | * cmb_size_mb=<cmb_size_mb[optional]>, \ | ||
255 | + * [pmrdev=<mem_backend_file_id>,] \ | ||
256 | * num_queues=<N[optional]> | ||
257 | * | ||
258 | * Note cmb_size_mb denotes size of CMB in MB. CMB is assumed to be at | ||
259 | * offset 0 in BAR2 and supports only WDS, RDS and SQS for now. | ||
260 | + * | ||
261 | + * cmb_size_mb= and pmrdev= options are mutually exclusive due to limitation | ||
262 | + * in available BAR's. cmb_size_mb= will take precedence over pmrdev= when | ||
263 | + * both provided. | ||
264 | + * Enabling pmr emulation can be achieved by pointing to memory-backend-file. | ||
265 | + * For example: | ||
266 | + * -object memory-backend-file,id=<mem_id>,share=on,mem-path=<file_path>, \ | ||
267 | + * size=<size> .... -device nvme,...,pmrdev=<mem_id> | ||
268 | */ | ||
269 | |||
270 | #include "qemu/osdep.h" | ||
271 | @@ -XXX,XX +XXX,XX @@ | ||
272 | #include "sysemu/sysemu.h" | ||
273 | #include "qapi/error.h" | ||
274 | #include "qapi/visitor.h" | ||
275 | +#include "sysemu/hostmem.h" | ||
276 | #include "sysemu/block-backend.h" | ||
277 | +#include "exec/ram_addr.h" | ||
278 | |||
279 | #include "qemu/log.h" | ||
280 | #include "qemu/module.h" | ||
281 | @@ -XXX,XX +XXX,XX @@ static void nvme_write_bar(NvmeCtrl *n, hwaddr offset, uint64_t data, | ||
282 | NVME_GUEST_ERR(nvme_ub_mmiowr_cmbsz_readonly, | ||
283 | "invalid write to read only CMBSZ, ignored"); | ||
284 | return; | ||
285 | + case 0xE00: /* PMRCAP */ | ||
286 | + NVME_GUEST_ERR(nvme_ub_mmiowr_pmrcap_readonly, | ||
287 | + "invalid write to PMRCAP register, ignored"); | ||
288 | + return; | ||
289 | + case 0xE04: /* TODO PMRCTL */ | ||
290 | + break; | ||
291 | + case 0xE08: /* PMRSTS */ | ||
292 | + NVME_GUEST_ERR(nvme_ub_mmiowr_pmrsts_readonly, | ||
293 | + "invalid write to PMRSTS register, ignored"); | ||
294 | + return; | ||
295 | + case 0xE0C: /* PMREBS */ | ||
296 | + NVME_GUEST_ERR(nvme_ub_mmiowr_pmrebs_readonly, | ||
297 | + "invalid write to PMREBS register, ignored"); | ||
298 | + return; | ||
299 | + case 0xE10: /* PMRSWTP */ | ||
300 | + NVME_GUEST_ERR(nvme_ub_mmiowr_pmrswtp_readonly, | ||
301 | + "invalid write to PMRSWTP register, ignored"); | ||
302 | + return; | ||
303 | + case 0xE14: /* TODO PMRMSC */ | ||
304 | + break; | ||
305 | default: | ||
306 | NVME_GUEST_ERR(nvme_ub_mmiowr_invalid, | ||
307 | "invalid MMIO write," | ||
308 | @@ -XXX,XX +XXX,XX @@ static uint64_t nvme_mmio_read(void *opaque, hwaddr addr, unsigned size) | ||
309 | } | 52 | } |
310 | 53 | - qemu_vfree(buf); | |
311 | if (addr < sizeof(n->bar)) { | 54 | return r; |
312 | + /* | ||
313 | + * When PMRWBM bit 1 is set then read from | ||
314 | + * from PMRSTS should ensure prior writes | ||
315 | + * made it to persistent media | ||
316 | + */ | ||
317 | + if (addr == 0xE08 && | ||
318 | + (NVME_PMRCAP_PMRWBM(n->bar.pmrcap) & 0x02)) { | ||
319 | + qemu_ram_writeback(n->pmrdev->mr.ram_block, | ||
320 | + 0, n->pmrdev->size); | ||
321 | + } | ||
322 | memcpy(&val, ptr + addr, size); | ||
323 | } else { | ||
324 | NVME_GUEST_ERR(nvme_ub_mmiord_invalid_ofs, | ||
325 | @@ -XXX,XX +XXX,XX @@ static void nvme_realize(PCIDevice *pci_dev, Error **errp) | ||
326 | error_setg(errp, "serial property not set"); | ||
327 | return; | ||
328 | } | ||
329 | + | ||
330 | + if (!n->cmb_size_mb && n->pmrdev) { | ||
331 | + if (host_memory_backend_is_mapped(n->pmrdev)) { | ||
332 | + char *path = object_get_canonical_path_component(OBJECT(n->pmrdev)); | ||
333 | + error_setg(errp, "can't use already busy memdev: %s", path); | ||
334 | + g_free(path); | ||
335 | + return; | ||
336 | + } | ||
337 | + | ||
338 | + if (!is_power_of_2(n->pmrdev->size)) { | ||
339 | + error_setg(errp, "pmr backend size needs to be power of 2 in size"); | ||
340 | + return; | ||
341 | + } | ||
342 | + | ||
343 | + host_memory_backend_set_mapped(n->pmrdev, true); | ||
344 | + } | ||
345 | + | ||
346 | blkconf_blocksizes(&n->conf); | ||
347 | if (!blkconf_apply_backend_options(&n->conf, blk_is_read_only(n->conf.blk), | ||
348 | false, errp)) { | ||
349 | @@ -XXX,XX +XXX,XX @@ static void nvme_realize(PCIDevice *pci_dev, Error **errp) | ||
350 | PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64 | | ||
351 | PCI_BASE_ADDRESS_MEM_PREFETCH, &n->ctrl_mem); | ||
352 | |||
353 | + } else if (n->pmrdev) { | ||
354 | + /* Controller Capabilities register */ | ||
355 | + NVME_CAP_SET_PMRS(n->bar.cap, 1); | ||
356 | + | ||
357 | + /* PMR Capabities register */ | ||
358 | + n->bar.pmrcap = 0; | ||
359 | + NVME_PMRCAP_SET_RDS(n->bar.pmrcap, 0); | ||
360 | + NVME_PMRCAP_SET_WDS(n->bar.pmrcap, 0); | ||
361 | + NVME_PMRCAP_SET_BIR(n->bar.pmrcap, 2); | ||
362 | + NVME_PMRCAP_SET_PMRTU(n->bar.pmrcap, 0); | ||
363 | + /* Turn on bit 1 support */ | ||
364 | + NVME_PMRCAP_SET_PMRWBM(n->bar.pmrcap, 0x02); | ||
365 | + NVME_PMRCAP_SET_PMRTO(n->bar.pmrcap, 0); | ||
366 | + NVME_PMRCAP_SET_CMSS(n->bar.pmrcap, 0); | ||
367 | + | ||
368 | + /* PMR Control register */ | ||
369 | + n->bar.pmrctl = 0; | ||
370 | + NVME_PMRCTL_SET_EN(n->bar.pmrctl, 0); | ||
371 | + | ||
372 | + /* PMR Status register */ | ||
373 | + n->bar.pmrsts = 0; | ||
374 | + NVME_PMRSTS_SET_ERR(n->bar.pmrsts, 0); | ||
375 | + NVME_PMRSTS_SET_NRDY(n->bar.pmrsts, 0); | ||
376 | + NVME_PMRSTS_SET_HSTS(n->bar.pmrsts, 0); | ||
377 | + NVME_PMRSTS_SET_CBAI(n->bar.pmrsts, 0); | ||
378 | + | ||
379 | + /* PMR Elasticity Buffer Size register */ | ||
380 | + n->bar.pmrebs = 0; | ||
381 | + NVME_PMREBS_SET_PMRSZU(n->bar.pmrebs, 0); | ||
382 | + NVME_PMREBS_SET_RBB(n->bar.pmrebs, 0); | ||
383 | + NVME_PMREBS_SET_PMRWBZ(n->bar.pmrebs, 0); | ||
384 | + | ||
385 | + /* PMR Sustained Write Throughput register */ | ||
386 | + n->bar.pmrswtp = 0; | ||
387 | + NVME_PMRSWTP_SET_PMRSWTU(n->bar.pmrswtp, 0); | ||
388 | + NVME_PMRSWTP_SET_PMRSWTV(n->bar.pmrswtp, 0); | ||
389 | + | ||
390 | + /* PMR Memory Space Control register */ | ||
391 | + n->bar.pmrmsc = 0; | ||
392 | + NVME_PMRMSC_SET_CMSE(n->bar.pmrmsc, 0); | ||
393 | + NVME_PMRMSC_SET_CBA(n->bar.pmrmsc, 0); | ||
394 | + | ||
395 | + pci_register_bar(pci_dev, NVME_PMRCAP_BIR(n->bar.pmrcap), | ||
396 | + PCI_BASE_ADDRESS_SPACE_MEMORY | PCI_BASE_ADDRESS_MEM_TYPE_64 | | ||
397 | + PCI_BASE_ADDRESS_MEM_PREFETCH, &n->pmrdev->mr); | ||
398 | } | ||
399 | |||
400 | for (i = 0; i < n->num_namespaces; i++) { | ||
401 | @@ -XXX,XX +XXX,XX @@ static void nvme_exit(PCIDevice *pci_dev) | ||
402 | if (n->cmb_size_mb) { | ||
403 | g_free(n->cmbuf); | ||
404 | } | ||
405 | + | ||
406 | + if (n->pmrdev) { | ||
407 | + host_memory_backend_set_mapped(n->pmrdev, false); | ||
408 | + } | ||
409 | msix_uninit_exclusive_bar(pci_dev); | ||
410 | } | 55 | } |
411 | 56 | ||
412 | static Property nvme_props[] = { | 57 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nvme_co_pdiscard(BlockDriverState *bs, |
413 | DEFINE_BLOCK_PROPERTIES(NvmeCtrl, conf), | 58 | BDRVNVMeState *s = bs->opaque; |
414 | + DEFINE_PROP_LINK("pmrdev", NvmeCtrl, pmrdev, TYPE_MEMORY_BACKEND, | 59 | NVMeQueuePair *ioq = s->queues[INDEX_IO(0)]; |
415 | + HostMemoryBackend *), | 60 | NVMeRequest *req; |
416 | DEFINE_PROP_STRING("serial", NvmeCtrl, serial), | 61 | - NvmeDsmRange *buf; |
417 | DEFINE_PROP_UINT32("cmb_size_mb", NvmeCtrl, cmb_size_mb, 0), | 62 | + QEMU_AUTO_VFREE NvmeDsmRange *buf = NULL; |
418 | DEFINE_PROP_UINT32("num_queues", NvmeCtrl, num_queues, 64), | 63 | QEMUIOVector local_qiov; |
419 | diff --git a/hw/block/Makefile.objs b/hw/block/Makefile.objs | 64 | int ret; |
420 | index XXXXXXX..XXXXXXX 100644 | 65 | |
421 | --- a/hw/block/Makefile.objs | 66 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn nvme_co_pdiscard(BlockDriverState *bs, |
422 | +++ b/hw/block/Makefile.objs | 67 | trace_nvme_dsm_done(s, offset, bytes, ret); |
423 | @@ -XXX,XX +XXX,XX @@ common-obj-$(CONFIG_PFLASH_CFI02) += pflash_cfi02.o | 68 | out: |
424 | common-obj-$(CONFIG_XEN) += xen-block.o | 69 | qemu_iovec_destroy(&local_qiov); |
425 | common-obj-$(CONFIG_ECC) += ecc.o | 70 | - qemu_vfree(buf); |
426 | common-obj-$(CONFIG_ONENAND) += onenand.o | 71 | return ret; |
427 | -common-obj-$(CONFIG_NVME_PCI) += nvme.o | 72 | |
428 | common-obj-$(CONFIG_SWIM) += swim.o | 73 | } |
429 | |||
430 | common-obj-$(CONFIG_SH4) += tc58128.o | ||
431 | |||
432 | obj-$(CONFIG_VIRTIO_BLK) += virtio-blk.o | ||
433 | obj-$(CONFIG_VHOST_USER_BLK) += vhost-user-blk.o | ||
434 | +obj-$(CONFIG_NVME_PCI) += nvme.o | ||
435 | |||
436 | obj-y += dataplane/ | ||
437 | diff --git a/hw/block/trace-events b/hw/block/trace-events | ||
438 | index XXXXXXX..XXXXXXX 100644 | ||
439 | --- a/hw/block/trace-events | ||
440 | +++ b/hw/block/trace-events | ||
441 | @@ -XXX,XX +XXX,XX @@ nvme_ub_mmiowr_ssreset_w1c_unsupported(void) "attempted to W1C CSTS.NSSRO but CA | ||
442 | nvme_ub_mmiowr_ssreset_unsupported(void) "attempted NVM subsystem reset but CAP.NSSRS is zero (not supported)" | ||
443 | nvme_ub_mmiowr_cmbloc_reserved(void) "invalid write to reserved CMBLOC when CMBSZ is zero, ignored" | ||
444 | nvme_ub_mmiowr_cmbsz_readonly(void) "invalid write to read only CMBSZ, ignored" | ||
445 | +nvme_ub_mmiowr_pmrcap_readonly(void) "invalid write to read only PMRCAP, ignored" | ||
446 | +nvme_ub_mmiowr_pmrsts_readonly(void) "invalid write to read only PMRSTS, ignored" | ||
447 | +nvme_ub_mmiowr_pmrebs_readonly(void) "invalid write to read only PMREBS, ignored" | ||
448 | +nvme_ub_mmiowr_pmrswtp_readonly(void) "invalid write to read only PMRSWTP, ignored" | ||
449 | nvme_ub_mmiowr_invalid(uint64_t offset, uint64_t data) "invalid MMIO write, offset=0x%"PRIx64", data=0x%"PRIx64"" | ||
450 | nvme_ub_mmiord_misaligned32(uint64_t offset) "MMIO read not 32-bit aligned, offset=0x%"PRIx64"" | ||
451 | nvme_ub_mmiord_toosmall(uint64_t offset) "MMIO read smaller than 32-bits, offset=0x%"PRIx64"" | ||
452 | -- | 74 | -- |
453 | 2.25.3 | 75 | 2.31.1 |
454 | 76 | ||
455 | 77 | diff view generated by jsdifflib |
1 | We want to keep TEST_IMG for the full path of the main test image, but | 1 | From: Philippe Mathieu-Daudé <philmd@redhat.com> |
---|---|---|---|
2 | filter_testfiles() must be called for other test images before replacing | ||
3 | other things like the image format because the test directory path could | ||
4 | contain the format as a substring. | ||
5 | 2 | ||
6 | Insert a filter_testfiles() call between both. | 3 | For debugging purpose it is helpful to know the CQ/SQ pointers. |
4 | We already have a trace event in nvme_free_queue_pair(), extend | ||
5 | it to report these pointer addresses. | ||
7 | 6 | ||
8 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 7 | Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> |
9 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 8 | Message-Id: <20211006164931.172349-3-philmd@redhat.com> |
10 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 9 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> |
11 | Message-Id: <20200424125448.63318-9-kwolf@redhat.com> | ||
12 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 10 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
13 | --- | 11 | --- |
14 | tests/qemu-iotests/iotests.py | 5 +++-- | 12 | block/nvme.c | 2 +- |
15 | 1 file changed, 3 insertions(+), 2 deletions(-) | 13 | block/trace-events | 2 +- |
14 | 2 files changed, 2 insertions(+), 2 deletions(-) | ||
16 | 15 | ||
17 | diff --git a/tests/qemu-iotests/iotests.py b/tests/qemu-iotests/iotests.py | 16 | diff --git a/block/nvme.c b/block/nvme.c |
18 | index XXXXXXX..XXXXXXX 100644 | 17 | index XXXXXXX..XXXXXXX 100644 |
19 | --- a/tests/qemu-iotests/iotests.py | 18 | --- a/block/nvme.c |
20 | +++ b/tests/qemu-iotests/iotests.py | 19 | +++ b/block/nvme.c |
21 | @@ -XXX,XX +XXX,XX @@ def filter_img_info(output, filename): | 20 | @@ -XXX,XX +XXX,XX @@ static bool nvme_init_queue(BDRVNVMeState *s, NVMeQueue *q, |
22 | for line in output.split('\n'): | 21 | |
23 | if 'disk size' in line or 'actual-size' in line: | 22 | static void nvme_free_queue_pair(NVMeQueuePair *q) |
24 | continue | 23 | { |
25 | - line = line.replace(filename, 'TEST_IMG') \ | 24 | - trace_nvme_free_queue_pair(q->index, q); |
26 | - .replace(imgfmt, 'IMGFMT') | 25 | + trace_nvme_free_queue_pair(q->index, q, &q->cq, &q->sq); |
27 | + line = line.replace(filename, 'TEST_IMG') | 26 | if (q->completion_bh) { |
28 | + line = filter_testfiles(line) | 27 | qemu_bh_delete(q->completion_bh); |
29 | + line = line.replace(imgfmt, 'IMGFMT') | 28 | } |
30 | line = re.sub('iters: [0-9]+', 'iters: XXX', line) | 29 | diff --git a/block/trace-events b/block/trace-events |
31 | line = re.sub('uuid: [-a-f0-9]+', 'uuid: XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX', line) | 30 | index XXXXXXX..XXXXXXX 100644 |
32 | line = re.sub('cid: [0-9]+', 'cid: XXXXXXXXXX', line) | 31 | --- a/block/trace-events |
32 | +++ b/block/trace-events | ||
33 | @@ -XXX,XX +XXX,XX @@ nvme_dsm_done(void *s, int64_t offset, int64_t bytes, int ret) "s %p offset 0x%" | ||
34 | nvme_dma_map_flush(void *s) "s %p" | ||
35 | nvme_free_req_queue_wait(void *s, unsigned q_index) "s %p q #%u" | ||
36 | nvme_create_queue_pair(unsigned q_index, void *q, size_t size, void *aio_context, int fd) "index %u q %p size %zu aioctx %p fd %d" | ||
37 | -nvme_free_queue_pair(unsigned q_index, void *q) "index %u q %p" | ||
38 | +nvme_free_queue_pair(unsigned q_index, void *q, void *cq, void *sq) "index %u q %p cq %p sq %p" | ||
39 | nvme_cmd_map_qiov(void *s, void *cmd, void *req, void *qiov, int entries) "s %p cmd %p req %p qiov %p entries %d" | ||
40 | nvme_cmd_map_qiov_pages(void *s, int i, uint64_t page) "s %p page[%d] 0x%"PRIx64 | ||
41 | nvme_cmd_map_qiov_iov(void *s, int i, void *page, int pages) "s %p iov[%d] %p pages %d" | ||
33 | -- | 42 | -- |
34 | 2.25.3 | 43 | 2.31.1 |
35 | 44 | ||
36 | 45 | diff view generated by jsdifflib |
1 | The raw format driver can simply forward the flag and let its bs->file | 1 | From: Philippe Mathieu-Daudé <philmd@redhat.com> |
---|---|---|---|
2 | child take care of actually providing the zeros. | ||
3 | 2 | ||
4 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 3 | Instead of duplicating code, extract the common helper to free |
5 | Reviewed-by: Max Reitz <mreitz@redhat.com> | 4 | a single queue. |
6 | Reviewed-by: Eric Blake <eblake@redhat.com> | 5 | |
7 | Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com> | 6 | Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com> |
8 | Message-Id: <20200424125448.63318-6-kwolf@redhat.com> | 7 | Message-Id: <20211006164931.172349-4-philmd@redhat.com> |
8 | Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com> | ||
9 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> | 9 | Signed-off-by: Kevin Wolf <kwolf@redhat.com> |
10 | --- | 10 | --- |
11 | block/raw-format.c | 4 +++- | 11 | block/nvme.c | 9 +++++++-- |
12 | 1 file changed, 3 insertions(+), 1 deletion(-) | 12 | 1 file changed, 7 insertions(+), 2 deletions(-) |
13 | 13 | ||
14 | diff --git a/block/raw-format.c b/block/raw-format.c | 14 | diff --git a/block/nvme.c b/block/nvme.c |
15 | index XXXXXXX..XXXXXXX 100644 | 15 | index XXXXXXX..XXXXXXX 100644 |
16 | --- a/block/raw-format.c | 16 | --- a/block/nvme.c |
17 | +++ b/block/raw-format.c | 17 | +++ b/block/nvme.c |
18 | @@ -XXX,XX +XXX,XX @@ static int coroutine_fn raw_co_truncate(BlockDriverState *bs, int64_t offset, | 18 | @@ -XXX,XX +XXX,XX @@ static bool nvme_init_queue(BDRVNVMeState *s, NVMeQueue *q, |
19 | 19 | return r == 0; | |
20 | s->size = offset; | ||
21 | offset += s->offset; | ||
22 | - return bdrv_co_truncate(bs->file, offset, exact, prealloc, 0, errp); | ||
23 | + return bdrv_co_truncate(bs->file, offset, exact, prealloc, flags, errp); | ||
24 | } | 20 | } |
25 | 21 | ||
26 | static void raw_eject(BlockDriverState *bs, bool eject_flag) | 22 | +static void nvme_free_queue(NVMeQueue *q) |
27 | @@ -XXX,XX +XXX,XX @@ static int raw_open(BlockDriverState *bs, QDict *options, int flags, | 23 | +{ |
28 | bs->supported_zero_flags = BDRV_REQ_WRITE_UNCHANGED | | 24 | + qemu_vfree(q->queue); |
29 | ((BDRV_REQ_FUA | BDRV_REQ_MAY_UNMAP | BDRV_REQ_NO_FALLBACK) & | 25 | +} |
30 | bs->file->bs->supported_zero_flags); | 26 | + |
31 | + bs->supported_truncate_flags = bs->file->bs->supported_truncate_flags & | 27 | static void nvme_free_queue_pair(NVMeQueuePair *q) |
32 | + BDRV_REQ_ZERO_WRITE; | 28 | { |
33 | 29 | trace_nvme_free_queue_pair(q->index, q, &q->cq, &q->sq); | |
34 | if (bs->probed && !bdrv_is_read_only(bs)) { | 30 | if (q->completion_bh) { |
35 | bdrv_refresh_filename(bs->file->bs); | 31 | qemu_bh_delete(q->completion_bh); |
32 | } | ||
33 | + nvme_free_queue(&q->sq); | ||
34 | + nvme_free_queue(&q->cq); | ||
35 | qemu_vfree(q->prp_list_pages); | ||
36 | - qemu_vfree(q->sq.queue); | ||
37 | - qemu_vfree(q->cq.queue); | ||
38 | qemu_mutex_destroy(&q->lock); | ||
39 | g_free(q); | ||
40 | } | ||
36 | -- | 41 | -- |
37 | 2.25.3 | 42 | 2.31.1 |
38 | 43 | ||
39 | 44 | diff view generated by jsdifflib |