Message ID | 20171012185916.22776-1-eblake@redhat.com |
---|---|
Headers | show |
Series | add byte-based block_status driver callbacks | expand |
Hi! Is it a latest portion of moving to byte-based? All other parts are already merged? Are you going to update it? (or this version if ok for review?) 12.10.2017 21:58, Eric Blake wrote: > There are patches floating around to add NBD_CMD_BLOCK_STATUS, > but NBD wants to report status on byte granularity (even if the > reporting will probably be naturally aligned to sectors or even > much higher levels). I've therefore started the task of > converting our block status code to report at a byte granularity > rather than sectors. > > Now that 2.11 is open, I'm rebasing/reposting the remaining patches. > > The overall conversion currently looks like: > part 1: bdrv_is_allocated (merged, commit 51b0a488) > part 2: dirty-bitmap (merged, commit ca759622) > part 3: bdrv_get_block_status (v6 is posted [1], partially reviewed) > part 4: .bdrv_co_block_status (this series, v3 was here [2]) > > Available as a tag at: > git fetch git://repo.or.cz/qemu/ericb.git nbd-byte-callback-v4 > > Based-on: <20171012034720.11947-1-eblake@redhat.com> > ([PATCH v6 00/24] make bdrv_get_block_status byte-based) > > Since v3: > - rebase to series 3 tweak in get_status interface > - further simplify qed code > - better documentation of optimization flag (s/mapping/want_zero/) > > 001/20:[0033] [FC] 'block: Add .bdrv_co_block_status() callback' > 002/20:[0077] [FC] 'block: Switch passthrough drivers to .bdrv_co_block_status()' > 003/20:[0020] [FC] 'file-posix: Switch to .bdrv_co_block_status()' > 004/20:[0019] [FC] 'gluster: Switch to .bdrv_co_block_status()' > 005/20:[----] [--] 'iscsi: Switch cluster_sectors to byte-based' > 006/20:[----] [--] 'iscsi: Switch iscsi_allocmap_update() to byte-based' > 007/20:[0022] [FC] 'iscsi: Switch to .bdrv_co_block_status()' > 008/20:[0013] [FC] 'null: Switch to .bdrv_co_block_status()' > 009/20:[0017] [FC] 'parallels: Switch to .bdrv_co_block_status()' > 010/20:[0014] [FC] 'qcow: Switch to .bdrv_co_block_status()' > 011/20:[0019] [FC] 'qcow2: Switch to .bdrv_co_block_status()' > 012/20:[0086] [FC] 'qed: Switch to .bdrv_co_block_status()' > 013/20:[0015] [FC] 'raw: Switch to .bdrv_co_block_status()' > 014/20:[0011] [FC] 'sheepdog: Switch to .bdrv_co_block_status()' > 015/20:[----] [--] 'vdi: Avoid bitrot of debugging code' > 016/20:[0017] [FC] 'vdi: Switch to .bdrv_co_block_status()' > 017/20:[0013] [FC] 'vmdk: Switch to .bdrv_co_block_status()' > 018/20:[0019] [FC] 'vpc: Switch to .bdrv_co_block_status()' > 019/20:[0010] [FC] 'vvfat: Switch to .bdrv_co_block_status()' > 020/20:[0019] [FC] 'block: Drop unused .bdrv_co_get_block_status()' > > Eric Blake (20): > block: Add .bdrv_co_block_status() callback > block: Switch passthrough drivers to .bdrv_co_block_status() > file-posix: Switch to .bdrv_co_block_status() > gluster: Switch to .bdrv_co_block_status() > iscsi: Switch cluster_sectors to byte-based > iscsi: Switch iscsi_allocmap_update() to byte-based > iscsi: Switch to .bdrv_co_block_status() > null: Switch to .bdrv_co_block_status() > parallels: Switch to .bdrv_co_block_status() > qcow: Switch to .bdrv_co_block_status() > qcow2: Switch to .bdrv_co_block_status() > qed: Switch to .bdrv_co_block_status() > raw: Switch to .bdrv_co_block_status() > sheepdog: Switch to .bdrv_co_block_status() > vdi: Avoid bitrot of debugging code > vdi: Switch to .bdrv_co_block_status() > vmdk: Switch to .bdrv_co_block_status() > vpc: Switch to .bdrv_co_block_status() > vvfat: Switch to .bdrv_co_block_status() > block: Drop unused .bdrv_co_get_block_status() > > include/block/block.h | 9 ++- > include/block/block_int.h | 43 +++++++------ > block/io.c | 80 +++++++++++------------- > block/blkdebug.c | 20 +++--- > block/commit.c | 2 +- > block/file-posix.c | 59 ++++++++++-------- > block/gluster.c | 67 +++++++++++--------- > block/iscsi.c | 154 +++++++++++++++++++++++++--------------------- > block/mirror.c | 2 +- > block/null.c | 23 +++---- > block/parallels.c | 22 ++++--- > block/qcow.c | 27 ++++---- > block/qcow2.c | 24 ++++---- > block/qed.c | 84 +++++++++---------------- > block/raw-format.c | 16 ++--- > block/sheepdog.c | 26 ++++---- > block/throttle.c | 2 +- > block/vdi.c | 45 +++++++------- > block/vmdk.c | 28 ++++----- > block/vpc.c | 42 +++++++------ > block/vvfat.c | 16 +++-- > 21 files changed, 404 insertions(+), 387 deletions(-) >
On 11/21/2017 05:27 AM, Vladimir Sementsov-Ogievskiy wrote: > Hi! > > Is it a latest portion of moving to byte-based? All other parts are > already merged? Are you going to update it? (or this version if ok for > review?) This version is okay to review; although it will need a v5 respin once 2.12 is reopened for some minor merge conflicts that have crept in during the meantime (regarding bs->drv NULL checks).
Most of conversions looks fine, but it is not simple to prove the correctness, because we start to use internal driver logic on offsets and lengths, not aligned to sectors. And we can't imagine the consequences (at least, I can't and my r-b doesn't give the guarantee) of such change. It is like take some function and expand its scope of parameters. May be I'm too paranoiac. May be it would be good to align requests in bdrv_co_block_status to sectors at least for drivers which do not provide request_alignment. May be each patch should be reviewed by person, knowing that particular driver. 12.10.2017 21:58, Eric Blake wrote: > There are patches floating around to add NBD_CMD_BLOCK_STATUS, > but NBD wants to report status on byte granularity (even if the > reporting will probably be naturally aligned to sectors or even > much higher levels). I've therefore started the task of > converting our block status code to report at a byte granularity > rather than sectors. > > Now that 2.11 is open, I'm rebasing/reposting the remaining patches. > > The overall conversion currently looks like: > part 1: bdrv_is_allocated (merged, commit 51b0a488) > part 2: dirty-bitmap (merged, commit ca759622) > part 3: bdrv_get_block_status (v6 is posted [1], partially reviewed) > part 4: .bdrv_co_block_status (this series, v3 was here [2]) > > Available as a tag at: > git fetch git://repo.or.cz/qemu/ericb.git nbd-byte-callback-v4 > > Based-on: <20171012034720.11947-1-eblake@redhat.com> > ([PATCH v6 00/24] make bdrv_get_block_status byte-based) > > Since v3: > - rebase to series 3 tweak in get_status interface > - further simplify qed code > - better documentation of optimization flag (s/mapping/want_zero/) > > 001/20:[0033] [FC] 'block: Add .bdrv_co_block_status() callback' > 002/20:[0077] [FC] 'block: Switch passthrough drivers to .bdrv_co_block_status()' > 003/20:[0020] [FC] 'file-posix: Switch to .bdrv_co_block_status()' > 004/20:[0019] [FC] 'gluster: Switch to .bdrv_co_block_status()' > 005/20:[----] [--] 'iscsi: Switch cluster_sectors to byte-based' > 006/20:[----] [--] 'iscsi: Switch iscsi_allocmap_update() to byte-based' > 007/20:[0022] [FC] 'iscsi: Switch to .bdrv_co_block_status()' > 008/20:[0013] [FC] 'null: Switch to .bdrv_co_block_status()' > 009/20:[0017] [FC] 'parallels: Switch to .bdrv_co_block_status()' > 010/20:[0014] [FC] 'qcow: Switch to .bdrv_co_block_status()' > 011/20:[0019] [FC] 'qcow2: Switch to .bdrv_co_block_status()' > 012/20:[0086] [FC] 'qed: Switch to .bdrv_co_block_status()' > 013/20:[0015] [FC] 'raw: Switch to .bdrv_co_block_status()' > 014/20:[0011] [FC] 'sheepdog: Switch to .bdrv_co_block_status()' > 015/20:[----] [--] 'vdi: Avoid bitrot of debugging code' > 016/20:[0017] [FC] 'vdi: Switch to .bdrv_co_block_status()' > 017/20:[0013] [FC] 'vmdk: Switch to .bdrv_co_block_status()' > 018/20:[0019] [FC] 'vpc: Switch to .bdrv_co_block_status()' > 019/20:[0010] [FC] 'vvfat: Switch to .bdrv_co_block_status()' > 020/20:[0019] [FC] 'block: Drop unused .bdrv_co_get_block_status()' > > Eric Blake (20): > block: Add .bdrv_co_block_status() callback > block: Switch passthrough drivers to .bdrv_co_block_status() > file-posix: Switch to .bdrv_co_block_status() > gluster: Switch to .bdrv_co_block_status() > iscsi: Switch cluster_sectors to byte-based > iscsi: Switch iscsi_allocmap_update() to byte-based > iscsi: Switch to .bdrv_co_block_status() > null: Switch to .bdrv_co_block_status() > parallels: Switch to .bdrv_co_block_status() > qcow: Switch to .bdrv_co_block_status() > qcow2: Switch to .bdrv_co_block_status() > qed: Switch to .bdrv_co_block_status() > raw: Switch to .bdrv_co_block_status() > sheepdog: Switch to .bdrv_co_block_status() > vdi: Avoid bitrot of debugging code > vdi: Switch to .bdrv_co_block_status() > vmdk: Switch to .bdrv_co_block_status() > vpc: Switch to .bdrv_co_block_status() > vvfat: Switch to .bdrv_co_block_status() > block: Drop unused .bdrv_co_get_block_status() > > include/block/block.h | 9 ++- > include/block/block_int.h | 43 +++++++------ > block/io.c | 80 +++++++++++------------- > block/blkdebug.c | 20 +++--- > block/commit.c | 2 +- > block/file-posix.c | 59 ++++++++++-------- > block/gluster.c | 67 +++++++++++--------- > block/iscsi.c | 154 +++++++++++++++++++++++++--------------------- > block/mirror.c | 2 +- > block/null.c | 23 +++---- > block/parallels.c | 22 ++++--- > block/qcow.c | 27 ++++---- > block/qcow2.c | 24 ++++---- > block/qed.c | 84 +++++++++---------------- > block/raw-format.c | 16 ++--- > block/sheepdog.c | 26 ++++---- > block/throttle.c | 2 +- > block/vdi.c | 45 +++++++------- > block/vmdk.c | 28 ++++----- > block/vpc.c | 42 +++++++------ > block/vvfat.c | 16 +++-- > 21 files changed, 404 insertions(+), 387 deletions(-) >
On 11/30/2017 07:04 AM, Vladimir Sementsov-Ogievskiy wrote: > Most of conversions looks fine, but it is not simple to prove the > correctness, because we start to use internal driver logic on offsets > and lengths, not aligned to sectors. The block layer guarantees that it will not pass unaligned data to the drivers, given the driver's definition of request_alignment. For drivers that have a request_alignment of 1, you are correct that the driver now sees requests at a smaller alignment than before, so it needs to be carefully reviewed per-driver. But for drivers that are still sector-based (where request_alignment is 512 for other reasons), there is no change in behavior. > And we can't imagine the > consequences (at least, I can't and my r-b doesn't give the guarantee) > of such change. It is like take some function and expand its scope of > parameters. > > May be I'm too paranoiac. May be it would be good to align requests > in bdrv_co_block_status to sectors at least for drivers which do not > provide request_alignment. May be each patch should be reviewed by > person, knowing that particular driver. ALL drivers have request_alignment set to a non-zero value; it defaults to 1 if the driver provides modern interfaces and does not override the value, and defaults to 512 if the driver still uses older sector-based interfaces. But I do welcome additional review; I have to post a v5 anyway.