64bit block-layer: part II
v6 cover 64bit block-layer: part II
Hi all!
Sorry for a long delay :( Finally, here is v6.
v6: rebase on new rbd handlers and backup-top renamed to copy-before-write. Also:
01: add Eric's r-b
tweak commit msg to not mention sheepdog
02: add Eric's r-b
03: tweak commit msg
drop extra type conversion in raw_co_pwrite_zeroes
04: vvfat: no write_target_commit() anymore
nvme: fix over-80 line
type conversion in raw_co_pwrite_zeroes() now dropped in 03
05: tweak commit msg
add Eric's r-b
06: update comment
07: tweak commit msg, grammar and typos in comments
don't add extra empty line
handle also qemu_rbd_co_pwrite_zeroes
08: tweak commit msg, add Eric's r-b 09: tweak commit msg, add Eric's r-b 10: add rbd, drop sheepdog
add assertion to iscsi_co_pdiscard()
11: tweak commit msg, add Eric's r-b
Vladimir Sementsov-Ogievskiy (11):
block/io: bring request check to bdrv_co_(read,write)v_vmstate
qcow2: check request on vmstate save/load path
block: use int64_t instead of uint64_t in driver read handlers
block: use int64_t instead of uint64_t in driver write handlers
block: use int64_t instead of uint64_t in copy_range driver handlers
block: make BlockLimits::max_pwrite_zeroes 64bit
block: use int64_t instead of int in driver write_zeroes handlers block/io: allow 64bit write-zeroes requests
block: make BlockLimits::max_pdiscard 64bit
block: use int64_t instead of int in driver discard handlers block/io: allow 64bit discard requests
include/block/block_int.h | 66 ++++++++++++++++-------------- block/blkdebug.c | 12 +++--- block/blklogwrites.c | 16 ++++---- block/blkreplay.c | 8 ++-- block/blkverify.c | 8 ++-- block/bochs.c | 4 +- block/cloop.c | 4 +- block/commit.c | 2 +- block/copy-before-write.c | 15 +++---- block/copy-on-read.c | 19 +++++---- block/crypto.c | 8 ++-- block/curl.c | 3 +- block/dmg.c | 4 +- block/file-posix.c | 35 ++++++++-------- block/file-win32.c | 8 ++-- block/filter-compress.c | 15 +++---- block/gluster.c | 13 +++--- block/io.c | 44 +++++++++++++++----- block/iscsi.c | 58 ++++++++++++++++---------- block/mirror.c | 8 ++-- block/nbd.c | 22 ++++++---- block/nfs.c | 12 +++--- block/null.c | 18 ++++---- block/nvme.c | 48 ++++++++++++++++++---- block/preallocate.c | 14 +++---- block/qcow.c | 16 ++++---- block/qcow2-cluster.c | 14 ++++++- block/qcow2.c | 70 +++++++++++++++++++++++--------- block/qed.c | 9 +++- block/quorum.c | 11 ++--- block/raw-format.c | 36 ++++++++-------- block/rbd.c | 20 ++++----- block/throttle.c | 18 ++++---- block/vdi.c | 8 ++-- block/vmdk.c | 14 +++---- block/vpc.c | 8 ++-- block/vvfat.c | 8 ++-- tests/unit/test-bdrv-drain.c | 16 ++++---- tests/unit/test-block-iothread.c | 21 +++++++--- block/trace-events | 10 ++--- slirp | 2 +- ui/keycodemapdb | 2 +- 42 files changed, 452 insertions(+), 295 deletions(-)
Signature · 1 line
2.29.2
PATCH v6 1/11 block/io: bring request check to bdrv_co_(read, write)v_vmstate
Only qcow2 driver supports vmstate.
In qcow2 these requests go through .bdrv_co_p{read,write}v_part handlers.
So, let's do our basic check for the request on vmstate generic handlers.
--- block/io.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-)
block/io.c +16 -2
@@ -2751,7 +2751,12 @@ bdrv_co_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) {BlockDriver *drv = bs->drv;BlockDriverState *child_bs =bdrv_primary_bs (bs); -int ret = -ENOTSUP ; +int ret; + + ret =bdrv_check_qiov_request (pos, qiov->size, qiov,0 ,NULL ); +if (ret <0 ) { +return ret; + }if (!drv) {return -ENOMEDIUM ;
@@ -2763,6 +2768,8 @@ bdrv_co_readv_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) ret = drv->bdrv_load_vmstate (bs, qiov, pos); }else if (child_bs) { ret =bdrv_co_readv_vmstate (child_bs, qiov, pos); + }else { + ret = -ENOTSUP ; }bdrv_dec_in_flight (bs);
@@ -2775,7 +2782,12 @@ bdrv_co_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) {BlockDriver *drv = bs->drv;BlockDriverState *child_bs =bdrv_primary_bs (bs); -int ret = -ENOTSUP ; +int ret; + + ret =bdrv_check_qiov_request (pos, qiov->size, qiov,0 ,NULL ); +if (ret <0 ) { +return ret; + }if (!drv) {return -ENOMEDIUM ;
@@ -2787,6 +2799,8 @@ bdrv_co_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos) ret = drv->bdrv_save_vmstate (bs, qiov, pos); }else if (child_bs) { ret =bdrv_co_writev_vmstate (child_bs, qiov, pos); + }else { + ret = -ENOTSUP ; }bdrv_dec_in_flight (bs);
Signature · 1 line
2.29.2
PATCH v6 2/11 qcow2: check request on vmstate save/load path
In reply to Vladimir Sementsov-OgievskiyWe modify the request by adding an offset to vmstate. Let's check the modified request. It will help us to safely move .bdrv_co_preadv_part and .bdrv_co_pwritev_part to int64_t type of offset and bytes.
--- include/block/block_int.h | 3 +++ block/io.c | 6 +++--- block/qcow2.c | 43 +++++++++++++++++++++++++++++++++------ 3 files changed, 43 insertions(+), 9 deletions(-)
include/block/block_int.h +3 -0
@@ -93,6 +93,9 @@ typedef struct BdrvTrackedRequest {struct BdrvTrackedRequest *waiting_for; } BdrvTrackedRequest; +int bdrv_check_qiov_request (int64_t offset,int64_t bytes, +QEMUIOVector *qiov,size_t qiov_offset, +Error **errp);int bdrv_check_request (int64_t offset,int64_t bytes,Error **errp);struct BlockDriver {
block/io.c +3 -3
@@ -956,9 +956,9 @@ bool coroutine_fn bdrv_make_request_serialising(BdrvTrackedRequest *req,return waited; } -static int bdrv_check_qiov_request (int64_t offset,int64_t bytes, -QEMUIOVector *qiov,size_t qiov_offset, -Error **errp) +int bdrv_check_qiov_request (int64_t offset,int64_t bytes, +QEMUIOVector *qiov,size_t qiov_offset, +Error **errp) {/* * Check generic offset/bytes correctness
block/qcow2.c +37 -6
61 lines not on this page · raw message
Signature · 1 line
2.29.2
PATCH v6 3/11 block: use int64_t instead of uint64_t in driver read handlers
In reply to Vladimir Sementsov-OgievskiyWe are generally moving to int64_t for both offset and bytes parameters on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and with possibility for signed return type (where negative value means error).
So, convert driver read handlers parameters which are already 64bit to signed type.
While being here, convert also flags parameter to be BdrvRequestFlags.
Now let's consider all callers. Simple
git grep '\->bdrv_\(aio\|co\)_preadv\(_part\)\?'
shows that's there three callers of driver function:
bdrv_driver_preadv() in block/io.c, passes int64_t, checked by
bdrv_check_qiov_request() to be non-negative.
qcow2_load_vmstate() does bdrv_check_qiov_request().
do_perform_cow_read() has uint64_t argument. And a lot of things in qcow2 driver are uint64_t, so converting it is big job. But we must not work with requests that don't satisfy bdrv_check_qiov_request(), so let's just assert it here.
Still, the functions may be called directly, not only by drv->... Let's check:
git grep '\.bdrv_\(aio\|co\)_preadv\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
The only one such caller:
QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, &data, 1);
...
ret = bdrv_replace_test_co_preadv(bs, 0, 1, &qiov, 0);in tesTS/unit/test-bdrv-drain.c, and it's OK obviously.
--- include/block/block_int.h | 11 ++++++----- block/blkdebug.c | 4 ++-- block/blklogwrites.c | 4 ++-- block/blkreplay.c | 2 +- block/blkverify.c | 4 ++-- block/bochs.c | 4 ++-- block/cloop.c | 4 ++-- block/commit.c | 2 +- block/copy-before-write.c | 4 ++-- block/copy-on-read.c | 4 ++-- block/crypto.c | 4 ++-- block/curl.c | 3 ++- block/dmg.c | 4 ++-- block/file-posix.c | 6 +++--- block/file-win32.c | 4 ++-- block/filter-compress.c | 4 ++-- block/mirror.c | 2 +- block/nbd.c | 5 +++-- block/nfs.c | 6 +++--- block/null.c | 9 +++++---- block/nvme.c | 5 +++-- block/preallocate.c | 4 ++-- block/qcow.c | 6 +++--- block/qcow2-cluster.c | 14 +++++++++++++- block/qcow2.c | 5 +++-- block/quorum.c | 4 ++-- block/raw-format.c | 20 ++++++++++---------- block/rbd.c | 6 +++--- block/throttle.c | 5 +++-- block/vdi.c | 4 ++-- block/vmdk.c | 4 ++-- block/vpc.c | 4 ++-- block/vvfat.c | 4 ++-- tests/unit/test-bdrv-drain.c | 16 +++++++++------- tests/unit/test-block-iothread.c | 19 ++++++++++++++----- slirp | 2 +- ui/keycodemapdb | 2 +- 37 files changed, 122 insertions(+), 92 deletions(-)
include/block/block_int.h +6 -5
@@ -234,8 +234,8 @@ struct BlockDriver {/* aio */ BlockAIOCB *(*bdrv_aio_preadv)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags, -BlockCompletionFunc *cb,void *opaque); +int64_t offset,int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags,BlockCompletionFunc *cb,void *opaque);BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs,uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags,BlockCompletionFunc *cb,void *opaque);
@@ -264,10 +264,11 @@ struct BlockDriver {* The buffer in @qiov may point directly to guest memory. */ int coroutine_fn (*bdrv_co_preadv)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags); +int64_t offset,int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags); int coroutine_fn (*bdrv_co_preadv_part)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,size_t qiov_offset,int flags); +int64_t offset,int64_t bytes, +QEMUIOVector *qiov,size_t qiov_offset,BdrvRequestFlags flags); int coroutine_fn (*bdrv_co_writev)(BlockDriverState *bs,int64_t sector_num,int nb_sectors,QEMUIOVector *qiov,int flags);/**
block/blkdebug.c +2 -2
@@ -631,8 +631,8 @@ static int rule_check(BlockDriverState *bs, uint64_t offset, uint64_t bytes, } static int coroutine_fn -blkdebug_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +blkdebug_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {int err;
block/blklogwrites.c +2 -2
@@ -301,8 +301,8 @@ static void blk_log_writes_refresh_limits(BlockDriverState *bs, Error **errp) } static int coroutine_fn -blk_log_writes_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +blk_log_writes_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {return bdrv_co_preadv (bs->file, offset, bytes, qiov, flags); }
block/blkreplay.c +1 -1
@@ -72,7 +72,7 @@ static void block_request_create(uint64_t reqid, BlockDriverState *bs, } static int coroutine_fn blkreplay_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes,QEMUIOVector *qiov,BdrvRequestFlags flags) {uint64_t reqid =blkreplay_next_id ();int ret =bdrv_co_preadv (bs->file, offset, bytes, qiov, flags);
block/blkverify.c +2 -2
@@ -221,8 +221,8 @@ blkverify_co_prwv(BlockDriverState *bs, BlkverifyRequest *r, uint64_t offset, } static int coroutine_fn -blkverify_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +blkverify_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BlkverifyRequest r;QEMUIOVector raw_qiov;
block/bochs.c +2 -2
@@ -238,8 +238,8 @@ static int64_t seek_to_sector(BlockDriverState *bs, int64_t sector_num) } static int coroutine_fn -bochs_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +bochs_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVBochsState *s = bs->opaque;uint64_t sector_num = offset >>BDRV_SECTOR_BITS ;
block/cloop.c +2 -2
@@ -245,8 +245,8 @@ static inline int cloop_read_block(BlockDriverState *bs, int block_num) } static int coroutine_fn -cloop_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +cloop_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVCloopState *s = bs->opaque;uint64_t sector_num = offset >>BDRV_SECTOR_BITS ;
block/commit.c +1 -1
@@ -207,7 +207,7 @@ static const BlockJobDriver commit_job_driver = { }; static int coroutine_fn bdrv_commit_top_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes,QEMUIOVector *qiov,BdrvRequestFlags flags) {return bdrv_co_preadv (bs->backing, offset, bytes, qiov, flags); }
block/copy-before-write.c +2 -2
@@ -40,8 +40,8 @@ typedef struct BDRVCopyBeforeWriteState { } BDRVCopyBeforeWriteState; static coroutine_fn int cbw_co_preadv( -BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {return bdrv_co_preadv (bs->file, offset, bytes, qiov, flags); }
block/copy-on-read.c +2 -2
@@ -128,10 +128,10 @@ static int64_t cor_getlength(BlockDriverState *bs) static int coroutine_fn cor_co_preadv_part(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, +int64_t offset,int64_t bytes,QEMUIOVector *qiov,size_t qiov_offset, -int flags) +BdrvRequestFlags flags) {int64_t n;int local_flags;
block/crypto.c +2 -2
@@ -397,8 +397,8 @@ static int block_crypto_reopen_prepare(BDRVReopenState *state,#define BLOCK_CRYPTO_MAX_IO_SIZE (1024 * 1024) static coroutine_fn int -block_crypto_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +block_crypto_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BlockCrypto *crypto = bs->opaque;uint64_t cur_bytes;/* number of bytes in current iteration */
block/curl.c +2 -1
@@ -896,7 +896,8 @@ out: } static int coroutine_fn curl_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {CURLAIOCB acb = { .co =qemu_coroutine_self (),
block/dmg.c +2 -2
@@ -689,8 +689,8 @@ static inline int dmg_read_chunk(BlockDriverState *bs, uint64_t sector_num) } static int coroutine_fn -dmg_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +dmg_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVDMGState *s = bs->opaque;uint64_t sector_num = offset >>BDRV_SECTOR_BITS ;
block/file-posix.c +3 -3
@@ -2077,9 +2077,9 @@ static int coroutine_fn raw_co_prw(BlockDriverState *bs, uint64_t offset,return raw_thread_pool_submit (bs, handle_aiocb_rw, &acb); } -static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +static int coroutine_fn raw_co_preadv(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {return raw_co_prw (bs, offset, bytes, qiov,QEMU_AIO_READ ); }
block/file-win32.c +2 -2
@@ -440,8 +440,8 @@ fail: }static BlockAIOCB *raw_aio_preadv (BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags, +int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags,BlockCompletionFunc *cb,void *opaque) {BDRVRawState *s = bs->opaque;
block/filter-compress.c +2 -2
@@ -63,10 +63,10 @@ static int64_t compress_getlength(BlockDriverState *bs) static int coroutine_fn compress_co_preadv_part(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, +int64_t offset,int64_t bytes,QEMUIOVector *qiov,size_t qiov_offset, -int flags) +BdrvRequestFlags flags) {return bdrv_co_preadv_part (bs->file, offset, bytes, qiov, qiov_offset, flags);
block/mirror.c +1 -1
@@ -1395,7 +1395,7 @@ static void coroutine_fn active_write_settle(MirrorOp *op) } static int coroutine_fn bdrv_mirror_top_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes,QEMUIOVector *qiov,BdrvRequestFlags flags) {return bdrv_co_preadv (bs->backing, offset, bytes, qiov, flags); }
block/nbd.c +3 -2
@@ -1322,8 +1322,9 @@ static int nbd_co_request(BlockDriverState *bs, NBDRequest *request,return ret ? ret : request_ret; } -static int nbd_client_co_preadv (BlockDriverState *bs,uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov,int flags) +static int nbd_client_co_preadv (BlockDriverState *bs,int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {int ret, request_ret;Error *local_err =NULL ;
block/nfs.c +3 -3
@@ -262,9 +262,9 @@ nfs_co_generic_cb(int ret, struct nfs_context *nfs, void *data, nfs_co_generic_bh_cb, task); } -static int coroutine_fn nfs_co_preadv(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *iov, -int flags) +static int coroutine_fn nfs_co_preadv(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *iov, +BdrvRequestFlags flags) {NFSClient *client = bs->opaque;NFSRPC task;
block/null.c +5 -4
@@ -116,8 +116,9 @@ static coroutine_fn int null_co_common(BlockDriverState *bs) } static coroutine_fn int null_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVNullState *s = bs->opaque;
@@ -187,8 +188,8 @@ static inline BlockAIOCB *null_aio_common(BlockDriverState *bs, }static BlockAIOCB *null_aio_preadv (BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags, +int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags,BlockCompletionFunc *cb,void *opaque) {
block/nvme.c +3 -2
@@ -1243,8 +1243,9 @@ static int nvme_co_prw(BlockDriverState *bs, uint64_t offset, uint64_t bytes, } static coroutine_fn int nvme_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {return nvme_co_prw (bs, offset, bytes, qiov, false, flags); }
block/preallocate.c +2 -2
@@ -227,8 +227,8 @@ static void preallocate_reopen_abort(BDRVReopenState *state) } static coroutine_fn int preallocate_co_preadv_part( -BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,size_t qiov_offset,int flags) +BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,size_t qiov_offset,BdrvRequestFlags flags) {return bdrv_co_preadv_part (bs->file, offset, bytes, qiov, qiov_offset, flags);
block/qcow.c +3 -3
@@ -617,9 +617,9 @@ static void qcow_refresh_limits(BlockDriverState *bs, Error **errp) bs->bl.request_alignment =BDRV_SECTOR_SIZE ; } -static coroutine_fn int qcow_co_preadv(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +static coroutine_fn int qcow_co_preadv(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVQcowState *s = bs->opaque;int offset_in_cluster;
block/qcow2-cluster.c +13 -1
@@ -505,7 +505,19 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs,return -ENOMEDIUM ; } -/* Call .bdrv_co_readv() directly instead of using the public block-layer +/* +* We never deal with requests that doesn't satisfy +* bdrv_check_qiov_request(), and aligning requests to clusters never break +* this condition. So, do some assertions before calling +* bs->drv->bdrv_co_preadv_part() which has int64_t arguments. +*/ +assert (src_cluster_offset <=INT64_MAX ); +assert (src_cluster_offset + offset_in_cluster <=INT64_MAX ); +assert (qiov->size <=INT64_MAX ); +bdrv_check_qiov_request (src_cluster_offset + offset_in_cluster, qiov->size, + qiov,0 , &error_abort); +/* +* Call .bdrv_co_readv() directly instead of using the public block-layer * interface. This avoids double I/O throttling and request tracking, * which can lead to deadlock when block layer copy-on-read is enabled. */
block/qcow2.c +3 -2
@@ -2310,9 +2310,10 @@ static coroutine_fn int qcow2_co_preadv_task_entry(AioTask *task) } static coroutine_fn int qcow2_co_preadv_part(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, +int64_t offset,int64_t bytes,QEMUIOVector *qiov, -size_t qiov_offset,int flags) +size_t qiov_offset, +BdrvRequestFlags flags) {BDRVQcow2State *s = bs->opaque;int ret =0 ;
block/quorum.c +2 -2
@@ -663,8 +663,8 @@ static int read_fifo_child(QuorumAIOCB *acb)return ret; } -static int quorum_co_preadv (BlockDriverState *bs,uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov,int flags) +static int quorum_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVQuorumState *s = bs->opaque;QuorumAIOCB *acb =quorum_aio_get (bs, qiov, offset, bytes, flags);
block/raw-format.c +10 -10
@@ -181,8 +181,8 @@ static void raw_reopen_abort(BDRVReopenState *state) }/* Check and adjust the offset, against 'offset' and 'size' options. */ -static inline int raw_adjust_offset (BlockDriverState *bs,uint64_t *offset, -uint64_t bytes,bool is_write) +static inline int raw_adjust_offset (BlockDriverState *bs,int64_t *offset, +int64_t bytes,bool is_write) {BDRVRawState *s = bs->opaque;
@@ -201,9 +201,9 @@ static inline int raw_adjust_offset(BlockDriverState *bs, uint64_t *offset,return 0 ; } -static int coroutine_fn raw_co_preadv(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +static int coroutine_fn raw_co_preadv(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {int ret;
@@ -259,7 +259,7 @@ static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, qiov = &local_qiov; } - ret =raw_adjust_offset (bs, &offset, bytes, true); + ret =raw_adjust_offset (bs, (int64_t *)&offset, bytes, true);if (ret) { goto fail; }
@@ -294,7 +294,7 @@ static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs, {int ret; - ret =raw_adjust_offset (bs, (uint64_t *)&offset, bytes, true); + ret =raw_adjust_offset (bs, &offset, bytes, true);if (ret) {return ret; }
@@ -306,7 +306,7 @@ static int coroutine_fn raw_co_pdiscard(BlockDriverState *bs, {int ret; - ret =raw_adjust_offset (bs, (uint64_t *)&offset, bytes, true); + ret =raw_adjust_offset (bs, &offset, bytes, true);if (ret) {return ret; }
@@ -541,7 +541,7 @@ static int coroutine_fn raw_co_copy_range_from(BlockDriverState *bs, {int ret; - ret =raw_adjust_offset (bs, &src_offset, bytes, false); + ret =raw_adjust_offset (bs, (int64_t *)&src_offset, bytes, false);if (ret) {return ret; }
@@ -560,7 +560,7 @@ static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs, {int ret; - ret =raw_adjust_offset (bs, &dst_offset, bytes, true); + ret =raw_adjust_offset (bs, (int64_t *)&dst_offset, bytes, true);if (ret) {return ret; }
block/rbd.c +3 -3
@@ -1164,9 +1164,9 @@ static int coroutine_fn qemu_rbd_start_co(BlockDriverState *bs, }static int -coroutine_fn qemu_rbd_co_preadv(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +coroutine_fn qemu_rbd_co_preadv(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {return qemu_rbd_start_co (bs, offset, bytes, qiov, flags,RBD_AIO_READ ); }
block/throttle.c +3 -2
@@ -112,8 +112,9 @@ static int64_t throttle_getlength(BlockDriverState *bs) } static int coroutine_fn throttle_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {ThrottleGroupMember *tgm = bs->opaque;
block/vdi.c +2 -2
@@ -544,8 +544,8 @@ static int coroutine_fn vdi_co_block_status(BlockDriverState *bs, } static int coroutine_fn -vdi_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vdi_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVVdiState *s = bs->opaque;QEMUIOVector local_qiov;
block/vmdk.c +2 -2
@@ -1888,8 +1888,8 @@ static int vmdk_read_extent(VmdkExtent *extent, int64_t cluster_offset, } static int coroutine_fn -vmdk_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vmdk_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVVmdkState *s = bs->opaque;int ret;
block/vpc.c +2 -2
@@ -608,8 +608,8 @@ static int vpc_get_info(BlockDriverState *bs, BlockDriverInfo *bdi) } static int coroutine_fn -vpc_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vpc_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVVPCState *s = bs->opaque;int ret;
block/vvfat.c +2 -2
@@ -1522,8 +1522,8 @@ static int vvfat_read(BlockDriverState *bs, int64_t sector_num, } static int coroutine_fn -vvfat_co_preadv (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vvfat_co_preadv (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {int ret;BDRVVVFATState *s = bs->opaque;
tests/unit/test-bdrv-drain.c +9 -7
@@ -65,8 +65,9 @@ static void co_reenter_bh(void *opaque) } static int coroutine_fn bdrv_test_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVTestState *s = bs->opaque;
@@ -1106,8 +1107,9 @@ static void bdrv_test_top_close(BlockDriverState *bs) } static int coroutine_fn bdrv_test_top_co_preadv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVTestTopState *tts = bs->opaque;return bdrv_co_preadv (tts->wait_child, offset, bytes, qiov, flags);
@@ -1855,10 +1857,10 @@ static void bdrv_replace_test_close(BlockDriverState *bs)* Set .has_read to true and return success. */ static int coroutine_fn bdrv_replace_test_co_preadv(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, +int64_t offset, +int64_t bytes,QEMUIOVector *qiov, -int flags) +BdrvRequestFlags flags) {BDRVReplaceTestState *s = bs->opaque;
tests/unit/test-block-iothread.c +14 -5
@@ -31,9 +31,18 @@#include "qemu/main-loop.h" #include "iothread.h" -static int coroutine_fn bdrv_test_co_prwv(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +static int coroutine_fn bdrv_test_co_preadv(BlockDriverState *bs, +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) +{ +return 0 ; +} + +static int coroutine_fn bdrv_test_co_pwritev(BlockDriverState *bs, +uint64_t offset,uint64_t bytes, +QEMUIOVector *qiov, +int flags) {return 0 ; }
@@ -66,8 +75,8 @@ static BlockDriver bdrv_test = { .format_name = "test", .instance_size = 1, - .bdrv_co_preadv = bdrv_test_co_prwv, - .bdrv_co_pwritev = bdrv_test_co_prwv, + .bdrv_co_preadv = bdrv_test_co_preadv, + .bdrv_co_pwritev = bdrv_test_co_pwritev, .bdrv_co_pdiscard = bdrv_test_co_pdiscard, .bdrv_co_truncate = bdrv_test_co_truncate, .bdrv_co_block_status = bdrv_test_co_block_status,
slirp +1 -1
@@ -1 +1 @@ -Subproject commit a88d9ace234a24ce1c17189642ef9104799425e0 +Subproject commit 8f43a99191afb47ca3f3c6972f6306209f367ece
ui/keycodemapdb +1 -1
@@ -1 +1 @@ -Subproject commit d21009b1c9f94b740ea66be8e48a1d8ad8124023 +Subproject commit 6119e6e19a050df847418de7babe5166779955e4
Signature · 1 line
2.29.2
Re: [PATCH v6 03/11] block: use int64_t instead of uint64_t in driver read handlers
Quoted 39 lines · Vladimir Sementsov-Ogievskiy
We are generally moving to
int64_tfor both offset and bytes parameters on all io paths.Main motivation is realization of 64-bit
write_zeroesoperation for fast zeroing large disk chunks, up to the whole disk.We chose signed type, to be consistent with
off_t(which is signed) and with possibility for signed return type (where negative value means error).So, convert driver read handlers parameters which are already 64bit to signed type.
While being here, convert also flags parameter to be BdrvRequestFlags.
Now let's consider all callers. Simple
git grep '\->bdrv_\(aio\|co\)_preadv\(_part\)\?'
shows that's there three callers of driver function:
bdrv_driver_preadv()inblock/io.c, passesint64_t, checked by
bdrv_check_qiov_request()to be non-negative.
qcow2_load_vmstate()doesbdrv_check_qiov_request().
do_perform_cow_read()hasuint64_targument. And a lot of things in qcow2 driver areuint64_t, so converting it is big job. But we must not work with requests that don't satisfybdrv_check_qiov_request(), so let's just assert it here.Still, the functions may be called directly, not only by drv->... Let's check:
git grep '\.bdrv_\(aio\|co\)_preadv\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
Quoted 7 lines
The only one such caller:
QEMUIOVector qiov = QEMU_IOVEC_INIT_BUF(qiov, &data, 1); ... ret = bdrv_replace_test_co_preadv(bs, 0, 1, &qiov, 0);in
tesTS/unit/test-bdrv-drain.c, and it's OK obviously.
Odd capitalization.
block/qcow2-cluster.c static int coroutine_fn do_perform_cow_read(BlockDriverState *bs, · line 510
Quoted 13 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>---
+++
b/block/qcow2-cluster.cblock/qcow2-cluster.c +3
-1@@ -505,7 +505,19 @@ static int coroutine_fn do_perform_cow_read(BlockDriverState *bs,return -ENOMEDIUM ; } -/* Call .bdrv_co_readv() directly instead of using the public block-layer +/* +* We never deal with requests that doesn't satisfy +* bdrv_check_qiov_request(), and aligning requests to clusters never break
never breaks
Quoted 18 lines
block/qcow2-cluster.c +10
-1+* this condition. So, do some assertions before calling +* bs->drv->bdrv_co_preadv_part() which has int64_t arguments. +*/ +assert (src_cluster_offset <=INT64_MAX ); +assert (src_cluster_offset + offset_in_cluster <=INT64_MAX ); +assert (qiov->size <=INT64_MAX ); +bdrv_check_qiov_request (src_cluster_offset + offset_in_cluster, qiov->size, + qiov,0 , &error_abort); +++ b/slirp@@ -1 +1 @@ -Subproject commit a88d9ace234a24ce1c17189642ef9104799425e0 +Subproject commit 8f43a99191afb47ca3f3c6972f6306209f367eceui/keycodemapdb +0
-0
ui/keycodemapdb line 1
Quoted 2 lines
ui/keycodemapdb +1
-1@@ -1 +1 @@ -Subproject commit d21009b1c9f94b740ea66be8e48a1d8ad8124023 +Subproject commit 6119e6e19a050df847418de7babe5166779955e4
Oops. Fix that (or I can do it while staging, if the rest of the series is okay), and you have:
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v6 4/11 block: use int64_t instead of uint64_t in driver write handlers
In reply to Vladimir Sementsov-OgievskiyWe are generally moving to int64_t for both offset and bytes parameters on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and with possibility for signed return type (where negative value means error).
So, convert driver write handlers parameters which are already 64bit to signed type.
While being here, convert also flags parameter to be BdrvRequestFlags.
Now let's consider all callers. Simple
git grep '\->bdrv_\(aio\|co\)_pwritev\(_part\)\?'
shows that's there three callers of driver function:
bdrv_driver_pwritev() and bdrv_driver_pwritev_compressed() in block/io.c, both pass int64_t, checked by bdrv_check_qiov_request() to be non-negative.
qcow2_save_vmstate() does bdrv_check_qiov_request().
Still, the functions may be called directly, not only by drv->... Let's check:
git grep '\.bdrv_\(aio\|co\)_pwritev\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
shows several callers:
qcow2:
qcow2_co_truncate() write at most up to @offset, which is checked in
generic qcow2_co_truncate() by bdrv_check_request().
qcow2_co_pwritev_compressed_task() pass the request (or part of the
request) that already went through normal write path, so it should be OK
qcow:
qcow_co_pwritev_compressed() pass int64_t, it's updated by this patch
quorum:
quorum_co_pwrite_zeroes() pass int64_t and int - OK
throttle:
throttle_co_pwritev_compressed() pass int64_t, it's updated by this patch
vmdk:
vmdk_co_pwritev_compressed() pass int64_t, it's updated by this patch
--- include/block/block_int.h | 16 ++++++++-------- block/blkdebug.c | 4 ++-- block/blklogwrites.c | 4 ++-- block/blkreplay.c | 2 +- block/blkverify.c | 4 ++-- block/copy-before-write.c | 7 ++++--- block/copy-on-read.c | 11 ++++++----- block/crypto.c | 4 ++-- block/file-posix.c | 6 +++--- block/file-win32.c | 4 ++-- block/filter-compress.c | 7 ++++--- block/io.c | 6 ++++-- block/mirror.c | 2 +- block/nbd.c | 5 +++-- block/nfs.c | 6 +++--- block/null.c | 9 +++++---- block/nvme.c | 5 +++-- block/preallocate.c | 6 +++--- block/qcow.c | 10 +++++----- block/qcow2.c | 6 +++--- block/quorum.c | 5 +++-- block/raw-format.c | 8 ++++---- block/rbd.c | 6 +++--- block/throttle.c | 9 +++++---- block/vdi.c | 4 ++-- block/vmdk.c | 8 ++++---- block/vpc.c | 4 ++-- block/vvfat.c | 4 ++-- tests/unit/test-block-iothread.c | 4 ++-- block/trace-events | 2 +- 30 files changed, 94 insertions(+), 84 deletions(-)
include/block/block_int.h +8 -8
@@ -237,8 +237,8 @@ struct BlockDriver { int64_t offset, int64_t bytes, QEMUIOVector *qiov, BdrvRequestFlags flags, BlockCompletionFunc *cb, void *opaque);BlockAIOCB *(*bdrv_aio_pwritev)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags, -BlockCompletionFunc *cb,void *opaque); +int64_t offset,int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags,BlockCompletionFunc *cb,void *opaque); BlockAIOCB *(*bdrv_aio_flush)(BlockDriverState *bs, BlockCompletionFunc *cb, void *opaque); BlockAIOCB *(*bdrv_aio_pdiscard)(BlockDriverState *bs,
@@ -287,10 +287,11 @@ struct BlockDriver {* The buffer in @qiov may point directly to guest memory. */ int coroutine_fn (*bdrv_co_pwritev)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags); +int64_t offset,int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags); int coroutine_fn (*bdrv_co_pwritev_part)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,size_t qiov_offset,int flags); +int64_t offset,int64_t bytes,QEMUIOVector *qiov,size_t qiov_offset, +BdrvRequestFlags flags);/* * Efficiently zero a region of the disk image. Typically an image format
@@ -428,10 +429,9 @@ struct BlockDriver { Error **errp); int coroutine_fn (*bdrv_co_pwritev_compressed)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov); +int64_t offset,int64_t bytes,QEMUIOVector *qiov); int coroutine_fn (*bdrv_co_pwritev_compressed_part)(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov, -size_t qiov_offset); +int64_t offset,int64_t bytes,QEMUIOVector *qiov,size_t qiov_offset);int (*bdrv_snapshot_create)(BlockDriverState *bs,QEMUSnapshotInfo *sn_info);
block/blkdebug.c +2 -2
@@ -652,8 +652,8 @@ blkdebug_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, } static int coroutine_fn -blkdebug_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +blkdebug_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {int err;
block/blklogwrites.c +2 -2
@@ -460,8 +460,8 @@ blk_log_writes_co_do_file_pdiscard(BlkLogWritesFileReq *fr) } static int coroutine_fn -blk_log_writes_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +blk_log_writes_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {return blk_log_writes_co_log (bs, offset, bytes, qiov, flags, blk_log_writes_co_do_file_pwritev,0 , false);
block/blkreplay.c +1 -1
@@ -83,7 +83,7 @@ static int coroutine_fn blkreplay_co_preadv(BlockDriverState *bs, } static int coroutine_fn blkreplay_co_pwritev(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes,QEMUIOVector *qiov,BdrvRequestFlags flags) {uint64_t reqid =blkreplay_next_id ();int ret =bdrv_co_pwritev (bs->file, offset, bytes, qiov, flags);
block/blkverify.c +2 -2
@@ -250,8 +250,8 @@ blkverify_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, } static int coroutine_fn -blkverify_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +blkverify_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BlkverifyRequest r;return blkverify_co_prwv (bs, &r, offset, bytes, qiov, qiov, flags, true);
block/copy-before-write.c +4 -3
@@ -86,9 +86,10 @@ static int coroutine_fn cbw_co_pwrite_zeroes(BlockDriverState *bs, } static coroutine_fn int cbw_co_pwritev(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset, +int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {int ret =cbw_do_copy_before_write (bs, offset, bytes, flags);if (ret <0 ) {
block/copy-on-read.c +6 -5
@@ -181,10 +181,11 @@ static int coroutine_fn cor_co_preadv_part(BlockDriverState *bs, static int coroutine_fn cor_co_pwritev_part(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, +int64_t offset, +int64_t bytes,QEMUIOVector *qiov, -size_t qiov_offset,int flags) +size_t qiov_offset, +BdrvRequestFlags flags) {return bdrv_co_pwritev_part (bs->file, offset, bytes, qiov, qiov_offset, flags);
@@ -207,8 +208,8 @@ static int coroutine_fn cor_co_pdiscard(BlockDriverState *bs, static int coroutine_fn cor_co_pwritev_compressed(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, +int64_t offset, +int64_t bytes,QEMUIOVector *qiov) { return bdrv_co_pwritev(bs->file, offset, bytes, qiov,
block/crypto.c +2 -2
@@ -460,8 +460,8 @@ block_crypto_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, static coroutine_fn int -block_crypto_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +block_crypto_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BlockCrypto *crypto = bs->opaque;uint64_t cur_bytes;/* number of bytes in current iteration */
block/file-posix.c +3 -3
@@ -2084,9 +2084,9 @@ static int coroutine_fn raw_co_preadv(BlockDriverState *bs, int64_t offset,return raw_co_prw (bs, offset, bytes, qiov,QEMU_AIO_READ ); } -static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {assert (flags ==0 );return raw_co_prw (bs, offset, bytes, qiov,QEMU_AIO_WRITE );
block/file-win32.c +2 -2
@@ -455,8 +455,8 @@ static BlockAIOCB *raw_aio_preadv(BlockDriverState *bs, }static BlockAIOCB *raw_aio_pwritev (BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags, +int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags,BlockCompletionFunc *cb,void *opaque) {BDRVRawState *s = bs->opaque;
block/filter-compress.c +4 -3
@@ -74,10 +74,11 @@ static int coroutine_fn compress_co_preadv_part(BlockDriverState *bs, static int coroutine_fn compress_co_pwritev_part(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, +int64_t offset, +int64_t bytes,QEMUIOVector *qiov, -size_t qiov_offset,int flags) +size_t qiov_offset, +BdrvRequestFlags flags) {return bdrv_co_pwritev_part (bs->file, offset, bytes, qiov, qiov_offset, flags |BDRV_REQ_WRITE_COMPRESSED );
block/io.c +4 -2
@@ -1230,7 +1230,8 @@ out: static int coroutine_fn bdrv_driver_pwritev(BlockDriverState *bs,int64_t offset,int64_t bytes,QEMUIOVector *qiov, -size_t qiov_offset,int flags) +size_t qiov_offset, +BdrvRequestFlags flags) {BlockDriver *drv = bs->drv;int64_t sector_num;
@@ -2070,7 +2071,8 @@ bdrv_co_write_req_finish(BdrvChild *child, int64_t offset, int64_t bytes,*/ static int coroutine_fn bdrv_aligned_pwritev(BdrvChild *child,BdrvTrackedRequest *req,int64_t offset,int64_t bytes, -int64_t align,QEMUIOVector *qiov,size_t qiov_offset,int flags) +int64_t align,QEMUIOVector *qiov,size_t qiov_offset, +BdrvRequestFlags flags) {BlockDriverState *bs = child->bs;BlockDriver *drv = bs->drv;
block/mirror.c +1 -1
@@ -1449,7 +1449,7 @@ out: } static int coroutine_fn bdrv_mirror_top_pwritev(BlockDriverState *bs, -uint64_t offset,uint64_t bytes,QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes,QEMUIOVector *qiov,BdrvRequestFlags flags) {MirrorBDSOpaque *s = bs->opaque;QEMUIOVector bounce_qiov;
block/nbd.c +3 -2
@@ -1381,8 +1381,9 @@ static int nbd_client_co_preadv(BlockDriverState *bs, int64_t offset,return ret ? ret : request_ret; } -static int nbd_client_co_pwritev (BlockDriverState *bs,uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov,int flags) +static int nbd_client_co_pwritev (BlockDriverState *bs,int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = {
block/nfs.c +3 -3
@@ -296,9 +296,9 @@ static int coroutine_fn nfs_co_preadv(BlockDriverState *bs, int64_t offset,return 0 ; } -static int coroutine_fn nfs_co_pwritev(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *iov, -int flags) +static int coroutine_fn nfs_co_pwritev(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *iov, +BdrvRequestFlags flags) {NFSClient *client = bs->opaque;NFSRPC task;
block/null.c +5 -4
@@ -130,8 +130,9 @@ static coroutine_fn int null_co_preadv(BlockDriverState *bs, } static coroutine_fn int null_co_pwritev(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {return null_co_common (bs); }
@@ -203,8 +204,8 @@ static BlockAIOCB *null_aio_preadv(BlockDriverState *bs, }static BlockAIOCB *null_aio_pwritev (BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags, +int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags,BlockCompletionFunc *cb,void *opaque) {
block/nvme.c +3 -2
@@ -1251,8 +1251,9 @@ static coroutine_fn int nvme_co_preadv(BlockDriverState *bs, } static coroutine_fn int nvme_co_pwritev(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {return nvme_co_prw (bs, offset, bytes, qiov, true, flags); }
block/preallocate.c +3 -3
@@ -349,11 +349,11 @@ static int coroutine_fn preallocate_co_pwrite_zeroes(BlockDriverState *bs, } static coroutine_fn int preallocate_co_pwritev_part(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, +int64_t offset, +int64_t bytes,QEMUIOVector *qiov,size_t qiov_offset, -int flags) +BdrvRequestFlags flags) {handle_write (bs, offset, bytes, false);
block/qcow.c +5 -5
@@ -714,9 +714,9 @@ static coroutine_fn int qcow_co_preadv(BlockDriverState *bs, int64_t offset,return ret; } -static coroutine_fn int qcow_co_pwritev(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +static coroutine_fn int qcow_co_pwritev(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVQcowState *s = bs->opaque;int offset_in_cluster;
@@ -1047,8 +1047,8 @@ static int qcow_make_empty(BlockDriverState *bs)/* XXX: put compressed sectors first, then all the cluster aligned tables to avoid losing bytes in alignment */ static coroutine_fn int -qcow_co_pwritev_compressed (BlockDriverState *bs,uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov) +qcow_co_pwritev_compressed (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov) {BDRVQcowState *s = bs->opaque;z_stream strm;
block/qcow2.c +3 -3
@@ -2597,8 +2597,8 @@ static coroutine_fn int qcow2_co_pwritev_task_entry(AioTask *task) } static coroutine_fn int qcow2_co_pwritev_part( -BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,size_t qiov_offset,int flags) +BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,size_t qiov_offset,BdrvRequestFlags flags) {BDRVQcow2State *s = bs->opaque;int offset_in_cluster;
@@ -4631,7 +4631,7 @@ static coroutine_fn int qcow2_co_pwritev_compressed_task_entry(AioTask *task)*/ static coroutine_fn intqcow2_co_pwritev_compressed_part (BlockDriverState *bs, -uint64_t offset,uint64_t bytes, +int64_t offset,int64_t bytes,QEMUIOVector *qiov,size_t qiov_offset) {BDRVQcow2State *s = bs->opaque;
block/quorum.c +3 -2
@@ -714,8 +714,9 @@ static void write_quorum_entry(void *opaque) } } -static int quorum_co_pwritev (BlockDriverState *bs,uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov,int flags) +static int quorum_co_pwritev (BlockDriverState *bs,int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVQuorumState *s = bs->opaque;QuorumAIOCB *acb =quorum_aio_get (bs, qiov, offset, bytes, flags);
block/raw-format.c +4 -4
@@ -216,9 +216,9 @@ static int coroutine_fn raw_co_preadv(BlockDriverState *bs, int64_t offset,return bdrv_co_preadv (bs->file, offset, bytes, qiov, flags); } -static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {void *buf =NULL ;BlockDriver *drv;
@@ -259,7 +259,7 @@ static int coroutine_fn raw_co_pwritev(BlockDriverState *bs, uint64_t offset, qiov = &local_qiov; } - ret =raw_adjust_offset (bs, (int64_t *)&offset, bytes, true); + ret =raw_adjust_offset (bs, &offset, bytes, true);if (ret) { goto fail; }
block/rbd.c +3 -3
@@ -1172,9 +1172,9 @@ coroutine_fn qemu_rbd_co_preadv(BlockDriverState *bs, int64_t offset, }static int -coroutine_fn qemu_rbd_co_pwritev(BlockDriverState *bs, uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov, -int flags) +coroutine_fn qemu_rbd_co_pwritev(BlockDriverState *bs, int64_t offset, +int64_t bytes,QEMUIOVector *qiov, +BdrvRequestFlags flags) {BDRVRBDState *s = bs->opaque;/*
block/throttle.c +5 -4
@@ -124,8 +124,9 @@ static int coroutine_fn throttle_co_preadv(BlockDriverState *bs, } static int coroutine_fn throttle_co_pwritev(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +int64_t offset,int64_t bytes, +QEMUIOVector *qiov, +BdrvRequestFlags flags) {ThrottleGroupMember *tgm = bs->opaque;throttle_group_co_io_limits_intercept (tgm, bytes, true);
@@ -153,8 +154,8 @@ static int coroutine_fn throttle_co_pdiscard(BlockDriverState *bs, } static int coroutine_fn throttle_co_pwritev_compressed(BlockDriverState *bs, -uint64_t offset, -uint64_t bytes, +int64_t offset, +int64_t bytes,QEMUIOVector *qiov) { return throttle_co_pwritev(bs, offset, bytes, qiov,
block/vdi.c +2 -2
@@ -600,8 +600,8 @@ vdi_co_preadv(BlockDriverState *bs, int64_t offset, int64_t bytes, } static int coroutine_fn -vdi_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vdi_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVVdiState *s = bs->opaque;QEMUIOVector local_qiov;
block/vmdk.c +4 -4
@@ -2068,8 +2068,8 @@ static int vmdk_pwritev(BlockDriverState *bs, uint64_t offset, } static int coroutine_fn -vmdk_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vmdk_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {int ret;BDRVVmdkState *s = bs->opaque;
@@ -2080,8 +2080,8 @@ vmdk_co_pwritev(BlockDriverState *bs, uint64_t offset, uint64_t bytes, } static int coroutine_fn -vmdk_co_pwritev_compressed (BlockDriverState *bs,uint64_t offset, -uint64_t bytes,QEMUIOVector *qiov) +vmdk_co_pwritev_compressed (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov) {if (bytes ==0 ) {/* The caller will write bytes 0 to signal EOF.
block/vpc.c +2 -2
@@ -658,8 +658,8 @@ fail: } static int coroutine_fn -vpc_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vpc_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {BDRVVPCState *s = bs->opaque;int64_t image_offset;
block/vvfat.c +2 -2
@@ -3061,8 +3061,8 @@ DLOG(checkpoint()); } static int coroutine_fn -vvfat_co_pwritev (BlockDriverState *bs,uint64_t offset,uint64_t bytes, -QEMUIOVector *qiov,int flags) +vvfat_co_pwritev (BlockDriverState *bs,int64_t offset,int64_t bytes, +QEMUIOVector *qiov,BdrvRequestFlags flags) {int ret;BDRVVVFATState *s = bs->opaque;
tests/unit/test-block-iothread.c +2 -2
@@ -40,9 +40,9 @@ static int coroutine_fn bdrv_test_co_preadv(BlockDriverState *bs, } static int coroutine_fn bdrv_test_co_pwritev(BlockDriverState *bs, -uint64_t offset,uint64_t bytes, +int64_t offset,int64_t bytes,QEMUIOVector *qiov, -int flags) +BdrvRequestFlags flags) {return 0 ; }
block/trace-events +1 -1
@@ -75,7 +75,7 @@ luring_resubmit_short_read(void *s, void *luringcb, int nread) "LuringState %p l # qcow2.c qcow2_add_task(void *co, void *bs, void *pool, const char *action, int cluster_type, uint64_t host_offset, uint64_t offset, uint64_t bytes, void *qiov, size_t qiov_offset) "co %p bs %p pool %p: %s: cluster_type %d file_cluster_offset %" PRIu64 " offset %" PRIu64 " bytes %" PRIu64 " qiov %p qiov_offset %zu" -qcow2_writev_start_req(void *co, int64_t offset, int bytes) "co %p offset 0x%" PRIx64 " bytes %d" +qcow2_writev_start_req(void *co, int64_t offset, int64_t bytes) "co %p offset 0x%" PRIx64 " bytes %" PRId64 qcow2_writev_done_req(void *co, int ret) "co %p ret %d" qcow2_writev_start_part(void *co) "co %p" qcow2_writev_done_part(void *co, int cur_bytes) "co %p cur_bytes %d"
Signature · 1 line
2.29.2
Re: [PATCH v6 04/11] block: use int64_t instead of uint64_t in driver write handlers
Quoted 2 lines · Vladimir Sementsov-Ogievskiy
We are generally moving to
int64_tfor both offset and bytes parameters on all io paths.
...
Quoted 27 lines
Still, the functions may be called directly, not only by drv->... Let's check:
git grep '\.bdrv_\(aio\|co\)_pwritev\(_part\)\?\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; doneshows several callers:
qcow2:
qcow2_co_truncate()write at most up to @offset, which is checked in
genericqcow2_co_truncate()bybdrv_check_request().
qcow2_co_pwritev_compressed_task()pass the request (or part of the
request) that already went through normal write path, so it should be OKqcow:
qcow_co_pwritev_compressed()passint64_t, it's updated by this patchquorum:
quorum_co_pwrite_zeroes()passint64_tand int - OKthrottle:
throttle_co_pwritev_compressed()passint64_t, it's updated by this patch
Quoted 6 lines
vmdk:
vmdk_co_pwritev_compressed()passint64_t, it's updated by this patchSigned-off-by Vladimir Sementsov-Ogievskiy <[email protected]>---
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v6 5/11 block: use int64_t instead of uint64_t in copy_range driver handlers
In reply to Vladimir Sementsov-OgievskiyWe are generally moving to int64_t for both offset and bytes parameters on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and with possibility for signed return type (where negative value means error).
So, convert driver copy_range handlers parameters which are already 64bit to signed type.
Now let's consider all callers. Simple
git grep '\->bdrv_co_copy_range'
shows the only caller:
bdrv_co_copy_range_internal(), which does bdrv_check_request32(), so everything is OK.
Still, the functions may be called directly, not only by drv->... Let's check:
git grep '\.bdrv_co_copy_range_\(from\|to\)\s*=' | \
awk '{print $4}' | sed 's/,//' | sed 's/&//' | sort | uniq | \
while read func; do git grep "$func(" | \
grep -v "$func(BlockDriverState"; done
shows no more callers. So, we are done.
--- include/block/block_int.h | 12 ++++++------ block/file-posix.c | 10 +++++----- block/iscsi.c | 12 ++++++------ block/qcow2.c | 12 ++++++------ block/raw-format.c | 16 ++++++++-------- 5 files changed, 31 insertions(+), 31 deletions(-)
include/block/block_int.h +6 -6
@@ -313,10 +313,10 @@ struct BlockDriver {*/ int coroutine_fn (*bdrv_co_copy_range_from)(BlockDriverState *bs,BdrvChild *src, -uint64_t offset, +int64_t offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags);
@@ -330,10 +330,10 @@ struct BlockDriver {*/ int coroutine_fn (*bdrv_co_copy_range_to)(BlockDriverState *bs,BdrvChild *src, -uint64_t src_offset, +int64_t src_offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags);
block/file-posix.c +5 -5
@@ -3202,8 +3202,8 @@ static void raw_abort_perm_update(BlockDriverState *bs) } static int coroutine_fn raw_co_copy_range_from( -BlockDriverState *bs,BdrvChild *src,uint64_t src_offset, -BdrvChild *dst,uint64_t dst_offset,uint64_t bytes, +BlockDriverState *bs,BdrvChild *src,int64_t src_offset, +BdrvChild *dst,int64_t dst_offset,int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) { return bdrv_co_copy_range_to(src, src_offset, dst, dst_offset, bytes,
@@ -3212,10 +3212,10 @@ static int coroutine_fn raw_co_copy_range_from( static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs,BdrvChild *src, -uint64_t src_offset, +int64_t src_offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {
block/iscsi.c +6 -6
@@ -2172,10 +2172,10 @@ static void coroutine_fn iscsi_co_invalidate_cache(BlockDriverState *bs, static int coroutine_fn iscsi_co_copy_range_from(BlockDriverState *bs,BdrvChild *src, -uint64_t src_offset, +int64_t src_offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {
@@ -2313,10 +2313,10 @@ static void iscsi_xcopy_data(struct iscsi_data *data, static int coroutine_fn iscsi_co_copy_range_to(BlockDriverState *bs,BdrvChild *src, -uint64_t src_offset, +int64_t src_offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {
block/qcow2.c +6 -6
@@ -4026,9 +4026,9 @@ static coroutine_fn int qcow2_co_pdiscard(BlockDriverState *bs, static int coroutine_fnqcow2_co_copy_range_from (BlockDriverState *bs, -BdrvChild *src,uint64_t src_offset, -BdrvChild *dst,uint64_t dst_offset, -uint64_t bytes,BdrvRequestFlags read_flags, +BdrvChild *src,int64_t src_offset, +BdrvChild *dst,int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {BDRVQcow2State *s = bs->opaque;
@@ -4109,9 +4109,9 @@ out: static int coroutine_fnqcow2_co_copy_range_to (BlockDriverState *bs, -BdrvChild *src,uint64_t src_offset, -BdrvChild *dst,uint64_t dst_offset, -uint64_t bytes,BdrvRequestFlags read_flags, +BdrvChild *src,int64_t src_offset, +BdrvChild *dst,int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {BDRVQcow2State *s = bs->opaque;
block/raw-format.c +8 -8
@@ -532,16 +532,16 @@ static int raw_probe_geometry(BlockDriverState *bs, HDGeometry *geo) static int coroutine_fn raw_co_copy_range_from(BlockDriverState *bs,BdrvChild *src, -uint64_t src_offset, +int64_t src_offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {int ret; - ret =raw_adjust_offset (bs, (int64_t *)&src_offset, bytes, false); + ret =raw_adjust_offset (bs, &src_offset, bytes, false);if (ret) {return ret; }
@@ -551,16 +551,16 @@ static int coroutine_fn raw_co_copy_range_from(BlockDriverState *bs, static int coroutine_fn raw_co_copy_range_to(BlockDriverState *bs,BdrvChild *src, -uint64_t src_offset, +int64_t src_offset,BdrvChild *dst, -uint64_t dst_offset, -uint64_t bytes, +int64_t dst_offset, +int64_t bytes,BdrvRequestFlags read_flags,BdrvRequestFlags write_flags) {int ret; - ret =raw_adjust_offset (bs, (int64_t *)&dst_offset, bytes, true); + ret =raw_adjust_offset (bs, &dst_offset, bytes, true);if (ret) {return ret; }
Signature · 1 line
2.29.2
PATCH v6 6/11 block: make BlockLimits::max_pwrite_zeroes 64bit
In reply to Vladimir Sementsov-OgievskiyWe are going to support 64 bit write-zeroes requests. Now update the limit variable. It's absolutely safe. The variable is set in some drivers, and used in bdrv_co_do_pwrite_zeroes().
Update also max_write_zeroes variable in bdrv_co_do_pwrite_zeroes(), so that bdrv_co_do_pwrite_zeroes() is now prepared to 64bit requests. The remaining logic including num, offset and bytes variables is already supporting 64bit requests.
So the only thing that prevents 64 bit requests is limiting max_write_zeroes variable to INT_MAX in bdrv_co_do_pwrite_zeroes(). We'll drop this limitation after updating all block drivers.
Ah, we also have bdrv_check_request32() in bdrv_co_pwritev_part(). It will be modified to do bdrv_check_request() for write-zeroes path.
--- include/block/block_int.h | 9 +++++---- block/io.c | 2 +- 2 files changed, 6 insertions(+), 5 deletions(-)
include/block/block_int.h +5 -4
@@ -676,10 +676,11 @@ typedef struct BlockLimits {* that is set. May be 0 if bl.request_alignment is good enough */ uint32_t pdiscard_alignment; -/* Maximum number of bytes that can zeroized at once (since it is -* signed, it must be < 2G, if set). Must be multiple of -* pwrite_zeroes_alignment. May be 0 if no inherent 32-bit limit */ -int32_t max_pwrite_zeroes; +/* +* Maximum number of bytes that can zeroized at once. Must be multiple of +* pwrite_zeroes_alignment. 0 means no limit. +*/ +int64_t max_pwrite_zeroes;/* Optimal alignment for write zeroes requests in bytes. A power * of 2 is best but not mandatory. Must be a multiple of
block/io.c +1 -1
@@ -1869,7 +1869,7 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs,int head =0 ;int tail =0 ; -int max_write_zeroes =MIN_NON_ZERO (bs->bl.max_pwrite_zeroes,INT_MAX ); +int64_t max_write_zeroes =MIN_NON_ZERO (bs->bl.max_pwrite_zeroes,INT_MAX );int alignment =MAX (bs->bl.pwrite_zeroes_alignment, bs->bl.request_alignment);int max_transfer =MIN_NON_ZERO (bs->bl.max_transfer,MAX_BOUNCE_BUFFER );
Signature · 1 line
2.29.2
Re: [PATCH v6 06/11] block: make BlockLimits::max_pwrite_zeroes 64bit
Quoted 13 lines · Vladimir Sementsov-Ogievskiy
We are going to support 64 bit write-zeroes requests. Now update the limit variable. It's absolutely safe. The variable is set in some drivers, and used in
bdrv_co_do_pwrite_zeroes().Update also
max_write_zeroesvariable inbdrv_co_do_pwrite_zeroes(), so thatbdrv_co_do_pwrite_zeroes()is now prepared to 64bit requests. The remaining logic including num, offset and bytes variables is already supporting 64bit requests.So the only thing that prevents 64 bit requests is limiting
max_write_zeroesvariable toINT_MAXinbdrv_co_do_pwrite_zeroes(). We'll drop this limitation after updating all block drivers.
Quoted 8 lines
Ah, we also have
bdrv_check_request32()inbdrv_co_pwritev_part(). It will be modified to dobdrv_check_request()for write-zeroes path.Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- include/block/block_int.h | 9 +++++---- block/io.c | 2 +- 2 files changed, 6 insertions(+), 5 deletions(-)
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v6 7/11 block: use int64_t instead of int in driver write_zeroes handlers
In reply to Vladimir Sementsov-OgievskiyWe are generally moving to int64_t for both offset and bytes parameters on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and with possibility for signed return type (where negative value means error).
So, convert driver write_zeroes handlers bytes parameter to int64_t.
The only caller of all updated function is bdrv_co_do_pwrite_zeroes().
bdrv_co_do_pwrite_zeroes() itself is of course OK with widening of callee parameter type. Also, bdrv_co_do_pwrite_zeroes()'s max_write_zeroes is limited to INT_MAX. So, updated functions all are safe, they will not get "bytes" larger than before.
Still, let's look through all updated functions, and add assertions to the ones which are actually unprepared to values larger than INT_MAX. For these drivers also set explicit max_pwrite_zeroes limit.
Let's go:
blkdebug: calculations can't overflow, thanks to
bdrv_check_qiov_request() in generic layer. rule_check() and bdrv_co_pwrite_zeroes() both have 64bit argument.
blklogwrites: pass to blk_log_writes_co_log() with 64bit argument.
blkreplay, copy-on-read, filter-compress: pass to
bdrv_co_pwrite_zeroes() which is OK
copy-before-write: Calls cbw_do_copy_before_write() and
bdrv_co_pwrite_zeroes, both have 64bit argument.
file-posix: both handler calls raw_do_pwrite_zeroes, which is updated.
In raw_do_pwrite_zeroes() calculations are OK due to bdrv_check_qiov_request(), bytes go to RawPosixAIOData::aio_nbytes which is uint64_t.
Check also where that uint64_t gets handed:
handle_aiocb_write_zeroes_block() passes a uint64_t[2] to ioctl(BLKZEROOUT), handle_aiocb_write_zeroes() calls do_fallocate() which takes off_t (and we compile to always have 64-bit off_t), as does handle_aiocb_write_zeroes_unmap. All look safe.
gluster: bytes go to GlusterAIOCB::size which is int64_t and to
glfs_zerofill_async works with off_t.
iscsi: Aha, here we deal with iscsi_writesame16_task() that has
uint32_t num_blocks argument and iscsi_writesame16_task() has uint16_t argument. Make comments, add assertions and clarify max_pwrite_zeroes calculation.
iscsi_allocmap_() functions already has int64_t argument is_byte_request_lun_aligned is simple to update, do it.
mirror_top: pass to bdrv_mirror_top_do_write which has uint64_t
argument
nbd: Aha, here we have protocol limitation, and NBDRequest::len is
uint32_t. max_pwrite_zeroes is cleanly set to 32bit value, so we are OK for now.
nvme: Again, protocol limitation. And no inherent limit for
write-zeroes at all. But from code that calculates cdw12 it's obvious that we do have limit and alignment. Let's clarify it. Also, obviously the code is not prepared to handle bytes=0. Let's handle this case too.
trace events already 64bit
preallocate: pass to handle_write() and bdrv_co_pwrite_zeroes(), both
64bit.
rbd: pass to qemu_rbd_start_co() which is 64bit.
qcow2: offset + bytes and alignment still works good (thanks to
bdrv_check_qiov_request()), so tail calculation is OK qcow2_subcluster_zeroize() has 64bit argument, should be OK trace events updated
qed: qed_co_request wants int nb_sectors. Also in code we have size_t
used for request length which may be 32bit. So, let's just keep INT_MAX as a limit (aligning it down to pwrite_zeroes_alignment) and don't care.
raw-format: Is OK. raw_adjust_offset and bdrv_co_pwrite_zeroes are both
64bit.
throttle: Both throttle_group_co_io_limits_intercept() and
bdrv_co_pwrite_zeroes() are 64bit.
vmdk: pass to vmdk_pwritev which is 64bit
quorum: pass to quorum_co_pwritev() which is 64bit
Hooray!
At this point all block drivers are prepared to support 64bit write-zero requests, or have explicitly set max_pwrite_zeroes.
--- include/block/block_int.h | 2 +- block/blkdebug.c | 2 +- block/blklogwrites.c | 4 ++-- block/blkreplay.c | 2 +- block/copy-before-write.c | 2 +- block/copy-on-read.c | 2 +- block/file-posix.c | 6 +++--- block/filter-compress.c | 2 +- block/gluster.c | 6 +++--- block/iscsi.c | 30 ++++++++++++++++++++---------- block/mirror.c | 2 +- block/nbd.c | 6 ++++-- block/nvme.c | 24 +++++++++++++++++++++--- block/preallocate.c | 2 +- block/qcow2.c | 2 +- block/qed.c | 9 ++++++++- block/quorum.c | 2 +- block/raw-format.c | 2 +- block/rbd.c | 4 ++-- block/throttle.c | 2 +- block/vmdk.c | 2 +- block/trace-events | 4 ++-- 22 files changed, 78 insertions(+), 41 deletions(-)
include/block/block_int.h +1 -1
@@ -300,7 +300,7 @@ struct BlockDriver {* will be called instead. */ int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags); +int64_t offset,int64_t bytes,BdrvRequestFlags flags); int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs,int64_t offset,int bytes);
block/blkdebug.c +1 -1
@@ -684,7 +684,7 @@ static int blkdebug_co_flush(BlockDriverState *bs) } static int coroutine_fn blkdebug_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes, +int64_t offset,int64_t bytes,BdrvRequestFlags flags) { uint32_t align = MAX(bs->bl.request_alignment,
block/blklogwrites.c +2 -2
@@ -468,8 +468,8 @@ blk_log_writes_co_pwritev(BlockDriverState *bs, int64_t offset, int64_t bytes, } static int coroutine_fn -blk_log_writes_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset,int bytes, -BdrvRequestFlags flags) +blk_log_writes_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset, +int64_t bytes,BdrvRequestFlags flags) { return blk_log_writes_co_log(bs, offset, bytes, NULL, flags, blk_log_writes_co_do_file_pwrite_zeroes, 0,
block/blkreplay.c +1 -1
@@ -94,7 +94,7 @@ static int coroutine_fn blkreplay_co_pwritev(BlockDriverState *bs, } static int coroutine_fn blkreplay_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags) +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {uint64_t reqid =blkreplay_next_id ();int ret =bdrv_co_pwrite_zeroes (bs->file, offset, bytes, flags);
block/copy-before-write.c +1 -1
@@ -75,7 +75,7 @@ static int coroutine_fn cbw_co_pdiscard(BlockDriverState *bs, } static int coroutine_fn cbw_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags) +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {int ret =cbw_do_copy_before_write (bs, offset, bytes, flags);if (ret <0 ) {
block/copy-on-read.c +1 -1
@@ -193,7 +193,7 @@ static int coroutine_fn cor_co_pwritev_part(BlockDriverState *bs, static int coroutine_fn cor_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes, +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {return bdrv_co_pwrite_zeroes (bs->file, offset, bytes, flags);
block/file-posix.c +3 -3
@@ -2971,7 +2971,7 @@ raw_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) } static int coroutine_fn -raw_do_pwrite_zeroes (BlockDriverState *bs,int64_t offset,int bytes, +raw_do_pwrite_zeroes (BlockDriverState *bs,int64_t offset,int64_t bytes,BdrvRequestFlags flags,bool blkdev) {BDRVRawState *s = bs->opaque;
@@ -3039,7 +3039,7 @@ raw_do_pwrite_zeroes(BlockDriverState *bs, int64_t offset, int bytes, static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs,int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags) {return raw_do_pwrite_zeroes (bs, offset, bytes, flags, false); }
@@ -3604,7 +3604,7 @@ hdev_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) } static coroutine_fn int hdev_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags) +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {int rc;
block/filter-compress.c +1 -1
@@ -86,7 +86,7 @@ static int coroutine_fn compress_co_pwritev_part(BlockDriverState *bs, static int coroutine_fn compress_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes, +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {return bdrv_co_pwrite_zeroes (bs->file, offset, bytes, flags);
block/gluster.c +3 -3
@@ -1003,19 +1003,19 @@ static void qemu_gluster_reopen_abort(BDRVReopenState *state)#ifdef CONFIG_GLUSTERFS_ZEROFILL static coroutine_fn int qemu_gluster_co_pwrite_zeroes(BlockDriverState *bs,int64_t offset, -int size, +int64_t bytes,BdrvRequestFlags flags) {int ret;GlusterAIOCB acb;BDRVGlusterState *s = bs->opaque; - acb.size = size; + acb.size = bytes; acb.ret =0 ; acb.coroutine =qemu_coroutine_self (); acb.aio_context =bdrv_get_aio_context (bs); - ret =glfs_zerofill_async (s->fd, offset, size, gluster_finish_aiocb, &acb); + ret =glfs_zerofill_async (s->fd, offset, bytes, gluster_finish_aiocb, &acb);if (ret <0 ) {return -errno; }
block/iscsi.c +20 -10
@@ -427,14 +427,14 @@ static int64_t sector_qemu2lun(int64_t sector, IscsiLun *iscsilun)return sector *BDRV_SECTOR_SIZE / iscsilun->block_size; } -static bool is_byte_request_lun_aligned (int64_t offset,int count, +static bool is_byte_request_lun_aligned (int64_t offset,int64_t bytes,IscsiLun *iscsilun) { -if (offset % iscsilun->block_size || count % iscsilun->block_size) { +if (offset % iscsilun->block_size || bytes % iscsilun->block_size) {error_report ("iSCSI misaligned request: " "iscsilun->block_size %u, offset %" PRIi64 -", count %d" , - iscsilun->block_size, offset, count); +", bytes %" PRIi64, + iscsilun->block_size, offset, bytes);return false; }return true;
@@ -1205,12 +1205,12 @@ out_unlock:static int coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags) {IscsiLun *iscsilun = bs->opaque;struct IscsiTask iTask;uint64_t lba; -uint32_t nb_blocks; +uint64_t nb_blocks;bool use_16_for_ws = iscsilun->use_16_for_rw;int r =0 ;
@@ -1250,11 +1250,21 @@ coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,iscsi_co_init_iscsitask (iscsilun, &iTask); retry:if (use_16_for_ws) { +/* +* iscsi_writesame16_task num_blocks argument is uint32_t. We rely here +* on our max_pwrite_zeroes limit. +*/ +assert (nb_blocks <UINT32_MAX ); iTask.task =iscsi_writesame16_task (iscsilun->iscsi, iscsilun->lun, lba, iscsilun->zeroblock, iscsilun->block_size, nb_blocks,0 , !!(flags &BDRV_REQ_MAY_UNMAP ),0 ,0 , iscsi_co_generic_cb, &iTask); }else { +/* +* iscsi_writesame10_task num_blocks argument is uint16_t. We rely here +* on our max_pwrite_zeroes limit. +*/ +assert (nb_blocks <UINT16_MAX ); iTask.task = iscsi_writesame10_task(iscsilun->iscsi, iscsilun->lun, lba, iscsilun->zeroblock, iscsilun->block_size, nb_blocks, 0, !!(flags & BDRV_REQ_MAY_UNMAP),
@@ -2074,10 +2084,10 @@ static void iscsi_refresh_limits(BlockDriverState *bs, Error **errp) bs->bl.pdiscard_alignment = iscsilun->block_size; } -if (iscsilun->bl.max_ws_len <0xffffffff / block_size) { - bs->bl.max_pwrite_zeroes = - iscsilun->bl.max_ws_len * iscsilun->block_size; - } + bs->bl.max_pwrite_zeroes = +MIN_NON_ZERO (iscsilun->bl.max_ws_len * iscsilun->block_size, + max_xfer_len * iscsilun->block_size); +if (iscsilun->lbp.lbpws) { bs->bl.pwrite_zeroes_alignment = iscsilun->bl.opt_unmap_gran * iscsilun->block_size;
block/mirror.c +1 -1
@@ -1494,7 +1494,7 @@ static int coroutine_fn bdrv_mirror_top_flush(BlockDriverState *bs) } static int coroutine_fn bdrv_mirror_top_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags) +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {return bdrv_mirror_top_do_write (bs,MIRROR_METHOD_ZERO , offset, bytes,NULL , flags);
block/nbd.c +4 -2
@@ -1407,15 +1407,17 @@ static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, }static int nbd_client_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags) {BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = { .type =NBD_CMD_WRITE_ZEROES , .from = offset, - .len = bytes, + .len = bytes,/* .len is uint32_t actually */ }; +assert (bytes <UINT32_MAX );/* relay on max_pwrite_zeroes */ +assert (!(s->info.flags &NBD_FLAG_READ_ONLY ));if (!(s->info.flags &NBD_FLAG_SEND_WRITE_ZEROES )) {return -ENOTSUP ;
block/nvme.c +21 -3
@@ -1288,19 +1288,29 @@ static coroutine_fn int nvme_co_flush(BlockDriverState *bs) static coroutine_fn int nvme_co_pwrite_zeroes(BlockDriverState *bs,int64_t offset, -int bytes, +int64_t bytes,BdrvRequestFlags flags) {BDRVNVMeState *s = bs->opaque;NVMeQueuePair *ioq = s->queues[INDEX_IO (0 )];NVMeRequest *req; - -uint32_t cdw12 = ((bytes >> s->blkshift) -1 ) &0xFFFF ; +uint32_t cdw12;if (!s->supports_write_zeroes) {return -ENOTSUP ; } +if (bytes ==0 ) { +return 0 ; + } + + cdw12 = ((bytes >> s->blkshift) -1 ) &0xFFFF ; +/* +* We should not lose information. pwrite_zeroes_alignment and +* max_pwrite_zeroes guarantees it. +*/ +assert (((cdw12 +1 ) << s->blkshift) == bytes); +NvmeCmd cmd = { .opcode =NVME_CMD_WRITE_ZEROES , .nsid =cpu_to_le32 (s->nsid),
@@ -1464,6 +1474,14 @@ static void nvme_refresh_limits(BlockDriverState *bs, Error **errp) bs->bl.opt_mem_alignment = s->page_size; bs->bl.request_alignment = s->page_size; bs->bl.max_transfer = s->max_transfer; + +/* +* Look at nvme_co_pwrite_zeroes: after shift and decrement we should get +* at most 0xFFFF +*/ + bs->bl.max_pwrite_zeroes =1ULL << (s->blkshift +16 ); + bs->bl.pwrite_zeroes_alignment =MAX (bs->bl.request_alignment, +1UL << s->blkshift); }static void nvme_detach_aio_context (BlockDriverState *bs)
block/preallocate.c +1 -1
@@ -337,7 +337,7 @@ static bool coroutine_fn handle_write(BlockDriverState *bs, int64_t offset, } static int coroutine_fn preallocate_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags) +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {bool want_merge_zero = !(flags & ~(BDRV_REQ_ZERO_WRITE |BDRV_REQ_NO_FALLBACK ));
block/qcow2.c +1 -1
@@ -3941,7 +3941,7 @@ static bool is_zero(BlockDriverState *bs, int64_t offset, int64_t bytes) } static coroutine_fn int qcow2_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes,BdrvRequestFlags flags) +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {int ret;BDRVQcow2State *s = bs->opaque;
block/qed.c +8 -1
@@ -582,6 +582,7 @@ static void bdrv_qed_refresh_limits(BlockDriverState *bs, Error **errp)BDRVQEDState *s = bs->opaque; bs->bl.pwrite_zeroes_alignment = s->header.cluster_size; + bs->bl.max_pwrite_zeroes =QEMU_ALIGN_DOWN (INT_MAX , s->header.cluster_size); }/* We have nothing to do for QED reopen, stubs just return
@@ -1397,7 +1398,7 @@ static int coroutine_fn bdrv_qed_co_writev(BlockDriverState *bs, static int coroutine_fn bdrv_qed_co_pwrite_zeroes(BlockDriverState *bs,int64_t offset, -int bytes, +int64_t bytes,BdrvRequestFlags flags) {BDRVQEDState *s = bs->opaque;
@@ -1408,6 +1409,12 @@ static int coroutine_fn bdrv_qed_co_pwrite_zeroes(BlockDriverState *bs,*/ QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov,NULL , bytes); +/* +* QED is not prepared for 63bit write-zero requests, so rely on +* max_pwrite_zeroes. +*/ +assert (bytes <=INT_MAX ); +/* Fall back if the request is not aligned */ if (qed_offset_into_cluster (s, offset) ||qed_offset_into_cluster (s, bytes)) {
block/quorum.c +1 -1
@@ -746,7 +746,7 @@ static int quorum_co_pwritev(BlockDriverState *bs, int64_t offset, }static int quorum_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags) { return quorum_co_pwritev(bs, offset, bytes, NULL,
block/raw-format.c +1 -1
@@ -289,7 +289,7 @@ static int coroutine_fn raw_co_block_status(BlockDriverState *bs, } static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes, +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {int ret;
block/rbd.c +2 -2
@@ -1205,9 +1205,9 @@ static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs,#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES static int coroutine_fn qemu_rbd_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, -int count,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags) { -return qemu_rbd_start_co (bs, offset, count,NULL , flags, +return qemu_rbd_start_co (bs, offset, bytes,NULL , flags,RBD_AIO_WRITE_ZEROES ); }#endif
block/throttle.c +1 -1
@@ -135,7 +135,7 @@ static int coroutine_fn throttle_co_pwritev(BlockDriverState *bs, } static int coroutine_fn throttle_co_pwrite_zeroes(BlockDriverState *bs, -int64_t offset,int bytes, +int64_t offset,int64_t bytes,BdrvRequestFlags flags) {ThrottleGroupMember *tgm = bs->opaque;
block/vmdk.c +1 -1
@@ -2109,7 +2109,7 @@ vmdk_co_pwritev_compressed(BlockDriverState *bs, int64_t offset, int64_t bytes, static int coroutine_fn vmdk_co_pwrite_zeroes(BlockDriverState *bs,int64_t offset, -int bytes, +int64_t bytes,BdrvRequestFlags flags) {int ret;
block/trace-events +2 -2
@@ -80,8 +80,8 @@ qcow2_writev_done_req(void *co, int ret) "co %p ret %d" qcow2_writev_start_part(void *co) "co %p" qcow2_writev_done_part(void *co, int cur_bytes) "co %p cur_bytes %d" qcow2_writev_data(void *co, uint64_t offset) "co %p offset 0x%" PRIx64 -qcow2_pwrite_zeroes_start_req(void *co, int64_t offset, int count) "co %p offset 0x%" PRIx64 " count %d" -qcow2_pwrite_zeroes(void *co, int64_t offset, int count) "co %p offset 0x%" PRIx64 " count %d" +qcow2_pwrite_zeroes_start_req(void *co, int64_t offset, int64_t bytes) "co %p offset 0x%" PRIx64 " bytes %" PRId64 +qcow2_pwrite_zeroes(void *co, int64_t offset, int64_t bytes) "co %p offset 0x%" PRIx64 " bytes %" PRId64 qcow2_skip_cow(void *co, uint64_t offset, int nb_clusters) "co %p offset 0x%" PRIx64 " nb_clusters %d" # qcow2-cluster.c
Signature · 1 line
2.29.2
Re: [PATCH v6 07/11] block: use int64_t instead of int in driver write_zeroes handlers
Quoted 23 lines · Vladimir Sementsov-Ogievskiy
We are generally moving to
int64_tfor both offset and bytes parameters on all io paths.Main motivation is realization of 64-bit
write_zeroesoperation for fast zeroing large disk chunks, up to the whole disk.We chose signed type, to be consistent with
off_t(which is signed) and with possibility for signed return type (where negative value means error).So, convert driver
write_zeroeshandlers bytes parameter toint64_t.The only caller of all updated function is
bdrv_co_do_pwrite_zeroes().
bdrv_co_do_pwrite_zeroes()itself is of course OK with widening of callee parameter type. Also, bdrv_co_do_pwrite_zeroes()'smax_write_zeroesis limited toINT_MAX. So, updated functions all are safe, they will not get "bytes" larger than before.Still, let's look through all updated functions, and add assertions to the ones which are actually unprepared to values larger than
INT_MAX. For these drivers also set explicitmax_pwrite_zeroeslimit.
[snip]
Quoted 3 lines
At this point all block drivers are prepared to support 64bit write-zero requests, or have explicitly set
max_pwrite_zeroes.
The long commit message is essential, but the analysis looks sane.
Quoted 9 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>---
+++
b/block/iscsi.cblock/iscsi.c +0
-0@@ -1250,11 +1250,21 @@ coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,iscsi_co_init_iscsitask (iscsilun, &iTask); retry:
block/iscsi.c coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, · line 1261
Quoted 10 lines
block/iscsi.c +5
-0if (use_16_for_ws) { +/* +* iscsi_writesame16_task num_blocks argument is uint32_t. We rely here +* on our max_pwrite_zeroes limit. +*/ +assert (nb_blocks <UINT32_MAX ); iTask.task =iscsi_writesame16_task (iscsilun->iscsi, iscsilun->lun, lba, iscsilun->zeroblock, iscsilun->block_size, nb_blocks,0 , !!(flags &BDRV_REQ_MAY_UNMAP ),0 ,0 , iscsi_co_generic_cb, &iTask);
Should this be <= instead of < ?
block/iscsi.c coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, · line 1270
Quoted 9 lines
block/iscsi.c +5
-0}else { +/* +* iscsi_writesame10_task num_blocks argument is uint16_t. We rely here +* on our max_pwrite_zeroes limit. +*/ +assert (nb_blocks <UINT16_MAX ); iTask.task = iscsi_writesame10_task(iscsilun->iscsi, iscsilun->lun, lba, iscsilun->zeroblock, iscsilun->block_size, nb_blocks, 0, !!(flags & BDRV_REQ_MAY_UNMAP),
here too. The 16-bit limit is where we're most likely to run into someone actually trying to zeroize that much at once.
Quoted 7 lines
+++
b/block/nbd.cblock/nbd.c +1
-1@@ -1407,15 +1407,17 @@ static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, }static int nbd_client_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags)
block/nbd.c static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, · line 1419
Quoted 10 lines
block/nbd.c +2
-1{BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = { .type =NBD_CMD_WRITE_ZEROES , .from = offset, - .len = bytes, + .len = bytes,/* .len is uint32_t actually */ }; +assert (bytes <UINT32_MAX );/* relay on max_pwrite_zeroes */
And again. Here, you happen to get by with < because we clamped bl.max_pwrite_zeroes at BDRV_REQUEST_MAX_BYTES, which is INT_MAX rounded down. But I had to check; whereas using <= would be less worrisome, even if we never get a request that large.
If you agree with my analysis, I can make that change while preparing my pull request.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [PATCH v6 07/11] block: use int64_t instead of int in driver write_zeroes handlers
Quoted 16 lines · Eric Blake
Quoted 16 lines
+++
b/block/nbd.cblock/nbd.c +2
-2@@ -1407,15 +1407,17 @@ static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, }static int nbd_client_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags) {BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = { .type =NBD_CMD_WRITE_ZEROES , .from = offset, - .len = bytes, + .len = bytes,/* .len is uint32_t actually */ };
Quoted 6 lines
block/nbd.c static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, · line 1419
Quoted 1 line
block/nbd.c +1
-0+assert (bytes <UINT32_MAX );/* relay on max_pwrite_zeroes */ And again. Here, you happen to get by with < because we clamped bl.max_pwrite_zeroes at
BDRV_REQUEST_MAX_BYTES, which isINT_MAXrounded down. But I had to check; whereas using <= would be less worrisome, even if we never get a request that large.
Whoops, I was reading a local patch of mine. Upstream has merely:
uint32_t max = MIN_NON_ZERO(NBD_MAX_BUFFER_SIZE, s->info.max_block);
bs->bl.max_pdiscard = QEMU_ALIGN_DOWN(INT_MAX, min);
bs->bl.max_pwrite_zeroes = max;which is an even smaller limit than BDRV_REQUEST_MAX_BYTES (and obviously one we're trying to raise). But the point remains that using <= rather than < will make it easier to review the code where we raise the limits (either up to the 4G-1 limit of the current protocol, or with protocol extensions to finally get to 64-bit requests).
Quoted 4 lines
If you agree with my analysis, I can make that change while preparing my pull request.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
Re: [PATCH v6 07/11] block: use int64_t instead of int in driver write_zeroes handlers
In reply to Eric BlakeQuoted 89 lines · Eric Blake
Quoted 23 lines · Vladimir Sementsov-Ogievskiy
We are generally moving to
int64_tfor both offset and bytes parameters on all io paths.Main motivation is realization of 64-bit
write_zeroesoperation for fast zeroing large disk chunks, up to the whole disk.We chose signed type, to be consistent with
off_t(which is signed) and with possibility for signed return type (where negative value means error).So, convert driver
write_zeroeshandlers bytes parameter toint64_t.The only caller of all updated function is
bdrv_co_do_pwrite_zeroes().
bdrv_co_do_pwrite_zeroes()itself is of course OK with widening of callee parameter type. Also, bdrv_co_do_pwrite_zeroes()'smax_write_zeroesis limited toINT_MAX. So, updated functions all are safe, they will not get "bytes" larger than before.Still, let's look through all updated functions, and add assertions to the ones which are actually unprepared to values larger than
INT_MAX. For these drivers also set explicitmax_pwrite_zeroeslimit.[snip]
Quoted 3 lines
At this point all block drivers are prepared to support 64bit write-zero requests, or have explicitly set
max_pwrite_zeroes.The long commit message is essential, but the analysis looks sane.
Quoted 9 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>---
+++
b/block/iscsi.cblock/iscsi.c +0
-0@@ -1250,11 +1250,21 @@ coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset,iscsi_co_init_iscsitask (iscsilun, &iTask); retry:block/iscsi.c coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, · line 1261
Quoted 10 lines
block/iscsi.c +5
-0if (use_16_for_ws) { +/* +* iscsi_writesame16_task num_blocks argument is uint32_t. We rely here +* on our max_pwrite_zeroes limit. +*/ +assert (nb_blocks <UINT32_MAX ); iTask.task =iscsi_writesame16_task (iscsilun->iscsi, iscsilun->lun, lba, iscsilun->zeroblock, iscsilun->block_size, nb_blocks,0 , !!(flags &BDRV_REQ_MAY_UNMAP ),0 ,0 , iscsi_co_generic_cb, &iTask);Should this be <= instead of < ?
block/iscsi.c coroutine_fn iscsi_co_pwrite_zeroes(BlockDriverState *bs, int64_t offset, · line 1270
Quoted 9 lines
block/iscsi.c +5
-0}else { +/* +* iscsi_writesame10_task num_blocks argument is uint16_t. We rely here +* on our max_pwrite_zeroes limit. +*/ +assert (nb_blocks <UINT16_MAX ); iTask.task = iscsi_writesame10_task(iscsilun->iscsi, iscsilun->lun, lba, iscsilun->zeroblock, iscsilun->block_size, nb_blocks, 0, !!(flags & BDRV_REQ_MAY_UNMAP),here too. The 16-bit limit is where we're most likely to run into someone actually trying to zeroize that much at once.
Quoted 7 lines
+++
b/block/nbd.cblock/nbd.c +1
-1@@ -1407,15 +1407,17 @@ static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, }static int nbd_client_co_pwrite_zeroes (BlockDriverState *bs,int64_t offset, -int bytes,BdrvRequestFlags flags) +int64_t bytes,BdrvRequestFlags flags)block/nbd.c static int nbd_client_co_pwritev(BlockDriverState *bs, int64_t offset, · line 1419
Quoted 10 lines
block/nbd.c +2
-1{BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = { .type =NBD_CMD_WRITE_ZEROES , .from = offset, - .len = bytes, + .len = bytes,/* .len is uint32_t actually */ }; +assert (bytes <UINT32_MAX );/* relay on max_pwrite_zeroes */ And again. Here, you happen to get by with < because we clamped bl.max_pwrite_zeroes at
BDRV_REQUEST_MAX_BYTES, which isINT_MAXrounded down. But I had to check; whereas using <= would be less worrisome, even if we never get a request that large.
Quoted 2 lines
If you agree with my analysis, I can make that change while preparing my pull request.
I agree, <= should be right thing, thanks!
Quoted 3 lines
Reviewed-by Eric Blake <[email protected]>
Signature · 2 lines
Best regards, Vladimir
PATCH v6 8/11 block/io: allow 64bit write-zeroes requests
In reply to Vladimir Sementsov-OgievskiyNow that all drivers are updated by previous commit, we can drop two last limiters on write-zeroes path: INT_MAX in bdrv_co_do_pwrite_zeroes() and bdrv_check_request32() in bdrv_co_pwritev_part().
Now everything is prepared for implementing incredibly cool and fast big-write-zeroes in NBD and qcow2. And any other driver which wants it of course.
--- block/io.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-)
block/io.c +7 -2
@@ -1869,7 +1869,8 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs,int head =0 ;int tail =0 ; -int64_t max_write_zeroes =MIN_NON_ZERO (bs->bl.max_pwrite_zeroes,INT_MAX ); +int64_t max_write_zeroes =MIN_NON_ZERO (bs->bl.max_pwrite_zeroes, +INT64_MAX );int alignment =MAX (bs->bl.pwrite_zeroes_alignment, bs->bl.request_alignment);int max_transfer =MIN_NON_ZERO (bs->bl.max_transfer,MAX_BOUNCE_BUFFER );
@@ -2245,7 +2246,11 @@ int coroutine_fn bdrv_co_pwritev_part(BdrvChild *child,return -ENOMEDIUM ; } - ret =bdrv_check_request32 (offset, bytes, qiov, qiov_offset); +if (flags &BDRV_REQ_ZERO_WRITE ) { + ret =bdrv_check_qiov_request (offset, bytes, qiov, qiov_offset,NULL ); + }else { + ret =bdrv_check_request32 (offset, bytes, qiov, qiov_offset); + }if (ret <0 ) {return ret; }
Signature · 1 line
2.29.2
PATCH v6 9/11 block: make BlockLimits::max_pdiscard 64bit
In reply to Vladimir Sementsov-OgievskiyWe are going to support 64 bit discard requests. Now update the limit variable. It's absolutely safe. The variable is set in some drivers, and used in bdrv_co_pdiscard().
Update also max_pdiscard variable in bdrv_co_pdiscard(), so that bdrv_co_pdiscard() is now prepared for 64bit requests. The remaining logic including num, offset and bytes variables is already supporting 64bit requests.
So the only thing that prevents 64 bit requests is limiting max_pdiscard variable to INT_MAX in bdrv_co_pdiscard(). We'll drop this limitation after updating all block drivers.
--- include/block/block_int.h | 11 ++++++----- block/io.c | 3 ++- 2 files changed, 8 insertions(+), 6 deletions(-)
include/block/block_int.h +6 -5
@@ -664,11 +664,12 @@ typedef struct BlockLimits {* otherwise. */ uint32_t request_alignment; -/* Maximum number of bytes that can be discarded at once (since it -* is signed, it must be < 2G, if set). Must be multiple of -* pdiscard_alignment, but need not be power of 2. May be 0 if no -* inherent 32-bit limit */ -int32_t max_pdiscard; +/* +* Maximum number of bytes that can be discarded at once. Must be multiple +* of pdiscard_alignment, but need not be power of 2. May be 0 if no +* inherent 64-bit limit. +*/ +int64_t max_pdiscard;/* Optimal alignment for discard requests in bytes. A power of 2 * is best but not mandatory. Must be a multiple of
block/io.c +2 -1
@@ -2997,7 +2997,8 @@ int coroutine_fn bdrv_co_pdiscard(BdrvChild *child, int64_t offset, int64_t bytes) {BdrvTrackedRequest req; -int max_pdiscard, ret; +int ret; +int64_t max_pdiscard;int head, tail, align;BlockDriverState *bs = child->bs;
Signature · 1 line
2.29.2
PATCH v6 10/11 block: use int64_t instead of int in driver discard handlers
In reply to Vladimir Sementsov-OgievskiyWe are generally moving to int64_t for both offset and bytes parameters on all io paths.
Main motivation is realization of 64-bit write_zeroes operation for fast zeroing large disk chunks, up to the whole disk.
We chose signed type, to be consistent with off_t (which is signed) and with possibility for signed return type (where negative value means error).
So, convert driver discard handlers bytes parameter to int64_t.
The only caller of all updated function is bdrv_co_pdiscard in block/io.c. It is already prepared to work with 64bit requests, but pass at most max(bs->bl.max_pdiscard, INT_MAX) to the driver.
Let's look at all updated functions:
blkdebug: all calculations are still OK, thanks to
bdrv_check_qiov_request().
both rule_check and bdrv_co_pdiscard are 64bit
blklogwrites: pass to blk_loc_writes_co_log which is 64bit
blkreplay, copy-on-read, filter-compress: pass to bdrv_co_pdiscard, OK
copy-before-write: pass to bdrv_co_pdiscard which is 64bit and to
cbw_do_copy_before_write which is 64bit
file-posix: one handler calls raw_account_discard() is 64bit and both
handlers calls raw_do_pdiscard(). Update raw_do_pdiscard, which pass to RawPosixAIOData::aio_nbytes, which is 64bit (and calls raw_account_discard())
gluster: somehow, third argument of glfs_discard_async is size_t.
Let's set max_pdiscard accordingly.
iscsi: iscsi_allocmap_set_invalid is 64bit,
!is_byte_request_lun_aligned is 64bit. list.num is uint32_t. Let's clarify max_pdiscard and pdiscard_alignment.
mirror_top: pass to bdrv_mirror_top_do_write() which is
64bit
nbd: protocol limitation. max_pdiscard is alredy set strict enough,
keep it as is for now.
nvme: buf.nlb is uint32_t and we do shift. So, add corresponding limits
to nvme_refresh_limits().
preallocate: pass to bdrv_co_pdiscard() which is 64bit.
rbd: pass to qemu_rbd_start_co() which is 64bit.
qcow2: calculations are still OK, thanks to bdrv_check_qiov_request(),
qcow2_cluster_discard() is 64bit.
raw-format: raw_adjust_offset() is 64bit, bdrv_co_pdiscard too.
throttle: pass to bdrv_co_pdiscard() which is 64bit and to
throttle_group_co_io_limits_intercept() which is 64bit as well.
test-block-iothread: bytes argument is unused
Great! Now all drivers are prepared to handle 64bit discard requests, or else have explicit max_pdiscard limits.
--- include/block/block_int.h | 2 +- block/blkdebug.c | 2 +- block/blklogwrites.c | 4 ++-- block/blkreplay.c | 2 +- block/copy-before-write.c | 2 +- block/copy-on-read.c | 2 +- block/file-posix.c | 7 ++++--- block/filter-compress.c | 2 +- block/gluster.c | 7 +++++-- block/iscsi.c | 16 +++++++++++----- block/mirror.c | 2 +- block/nbd.c | 6 ++++-- block/nvme.c | 14 +++++++++++++- block/preallocate.c | 2 +- block/qcow2.c | 2 +- block/raw-format.c | 2 +- block/rbd.c | 4 ++-- block/throttle.c | 2 +- tests/unit/test-block-iothread.c | 2 +- block/trace-events | 4 ++-- 20 files changed, 55 insertions(+), 31 deletions(-)
include/block/block_int.h +1 -1
@@ -302,7 +302,7 @@ struct BlockDriver { int coroutine_fn (*bdrv_co_pwrite_zeroes)(BlockDriverState *bs,int64_t offset,int64_t bytes,BdrvRequestFlags flags); int coroutine_fn (*bdrv_co_pdiscard)(BlockDriverState *bs, -int64_t offset,int bytes); +int64_t offset,int64_t bytes);/* Map [offset, offset + nbytes) range onto a child of @bs to copy from, * and invoke bdrv_co_copy_range_from(child, ...), or invoke
block/blkdebug.c +1 -1
@@ -717,7 +717,7 @@ static int coroutine_fn blkdebug_co_pwrite_zeroes(BlockDriverState *bs, } static int coroutine_fn blkdebug_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {uint32_t align = bs->bl.pdiscard_alignment;int err;
block/blklogwrites.c +2 -2
@@ -484,9 +484,9 @@ static int coroutine_fn blk_log_writes_co_flush_to_disk(BlockDriverState *bs) } static int coroutine_fn -blk_log_writes_co_pdiscard (BlockDriverState *bs,int64_t offset,int count) +blk_log_writes_co_pdiscard (BlockDriverState *bs,int64_t offset,int64_t bytes) { -return blk_log_writes_co_log (bs, offset, count,NULL ,0 , +return blk_log_writes_co_log (bs, offset, bytes,NULL ,0 , blk_log_writes_co_do_file_pdiscard,LOG_DISCARD_FLAG , false); }
block/blkreplay.c +1 -1
@@ -105,7 +105,7 @@ static int coroutine_fn blkreplay_co_pwrite_zeroes(BlockDriverState *bs, } static int coroutine_fn blkreplay_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {uint64_t reqid =blkreplay_next_id ();int ret =bdrv_co_pdiscard (bs->file, offset, bytes);
block/copy-before-write.c +1 -1
@@ -64,7 +64,7 @@ static coroutine_fn int cbw_do_copy_before_write(BlockDriverState *bs, } static int coroutine_fn cbw_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {int ret =cbw_do_copy_before_write (bs, offset, bytes,0 );if (ret <0 ) {
block/copy-on-read.c +1 -1
@@ -201,7 +201,7 @@ static int coroutine_fn cor_co_pwrite_zeroes(BlockDriverState *bs, static int coroutine_fn cor_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {return bdrv_co_pdiscard (bs->file, offset, bytes); }
block/file-posix.c +4 -3
@@ -2941,7 +2941,8 @@ static void raw_account_discard(BDRVRawState *s, uint64_t nbytes, int ret) } static coroutine_fn int -raw_do_pdiscard (BlockDriverState *bs,int64_t offset,int bytes,bool blkdev) +raw_do_pdiscard (BlockDriverState *bs,int64_t offset,int64_t bytes, +bool blkdev) {BDRVRawState *s = bs->opaque;RawPosixAIOData acb;
@@ -2965,7 +2966,7 @@ raw_do_pdiscard(BlockDriverState *bs, int64_t offset, int bytes, bool blkdev) } static coroutine_fn int -raw_co_pdiscard (BlockDriverState *bs,int64_t offset,int bytes) +raw_co_pdiscard (BlockDriverState *bs,int64_t offset,int64_t bytes) {return raw_do_pdiscard (bs, offset, bytes, false); }
@@ -3590,7 +3591,7 @@ hdev_co_ioctl(BlockDriverState *bs, unsigned long int req, void *buf)#endif /* linux */ static coroutine_fn int -hdev_co_pdiscard (BlockDriverState *bs,int64_t offset,int bytes) +hdev_co_pdiscard (BlockDriverState *bs,int64_t offset,int64_t bytes) {BDRVRawState *s = bs->opaque;int ret;
block/filter-compress.c +1 -1
@@ -94,7 +94,7 @@ static int coroutine_fn compress_co_pwrite_zeroes(BlockDriverState *bs, static int coroutine_fn compress_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {return bdrv_co_pdiscard (bs->file, offset, bytes); }
block/gluster.c +5 -2
@@ -891,6 +891,7 @@ out:static void qemu_gluster_refresh_limits (BlockDriverState *bs,Error **errp) { bs->bl.max_transfer =GLUSTER_MAX_TRANSFER ; + bs->bl.max_pdiscard =SIZE_MAX ; } static int qemu_gluster_reopen_prepare(BDRVReopenState *state,
@@ -1297,18 +1298,20 @@ error:#ifdef CONFIG_GLUSTERFS_DISCARD static coroutine_fn int qemu_gluster_co_pdiscard(BlockDriverState *bs, -int64_t offset,int size) +int64_t offset,int64_t bytes) {int ret;GlusterAIOCB acb;BDRVGlusterState *s = bs->opaque; +assert (bytes <=SIZE_MAX );/* rely on max_pdiscard */ + acb.size =0 ; acb.ret =0 ; acb.coroutine =qemu_coroutine_self (); acb.aio_context =bdrv_get_aio_context (bs); - ret =glfs_discard_async (s->fd, offset, size, gluster_finish_aiocb, &acb); + ret =glfs_discard_async (s->fd, offset, bytes, gluster_finish_aiocb, &acb);if (ret <0 ) {return -errno; }
block/iscsi.c +11 -5
@@ -1141,7 +1141,8 @@ iscsi_getlength(BlockDriverState *bs) }static int -coroutine_fn iscsi_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes) +coroutine_fn iscsi_co_pdiscard(BlockDriverState *bs, int64_t offset, +int64_t bytes) {IscsiLun *iscsilun = bs->opaque;struct IscsiTask iTask;
@@ -1157,6 +1158,12 @@ coroutine_fn iscsi_co_pdiscard(BlockDriverState *bs, int64_t offset, int bytes)return 0 ; } +/* +* We don't want to overflow list.num which is uint32_t. +* We rely on our max_pdiscard. +*/ +assert (bytes / iscsilun->block_size <=UINT32_MAX ); + list.lba = offset / iscsilun->block_size; list.num = bytes / iscsilun->block_size;
@@ -2074,10 +2081,9 @@ static void iscsi_refresh_limits(BlockDriverState *bs, Error **errp) }if (iscsilun->lbp.lbpu) { -if (iscsilun->bl.max_unmap <0xffffffff / block_size) { - bs->bl.max_pdiscard = - iscsilun->bl.max_unmap * iscsilun->block_size; - } + bs->bl.max_pdiscard = +MIN_NON_ZERO (iscsilun->bl.max_unmap * iscsilun->block_size, + (uint64_t )UINT32_MAX * iscsilun->block_size); bs->bl.pdiscard_alignment = iscsilun->bl.opt_unmap_gran * iscsilun->block_size; }else {
block/mirror.c +1 -1
@@ -1501,7 +1501,7 @@ static int coroutine_fn bdrv_mirror_top_pwrite_zeroes(BlockDriverState *bs, } static int coroutine_fn bdrv_mirror_top_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {return bdrv_mirror_top_do_write (bs,MIRROR_METHOD_DISCARD , offset, bytes,NULL ,0 );
block/nbd.c +4 -2
@@ -1457,15 +1457,17 @@ static int nbd_client_co_flush(BlockDriverState *bs) }static int nbd_client_co_pdiscard (BlockDriverState *bs,int64_t offset, -int bytes) +int64_t bytes) {BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = { .type =NBD_CMD_TRIM , .from = offset, - .len = bytes, + .len = bytes,/* len is uint32_t */ }; +assert (bytes <=UINT32_MAX );/* rely on max_pdiscard */ +assert (!(s->info.flags &NBD_FLAG_READ_ONLY ));if (!(s->info.flags &NBD_FLAG_SEND_TRIM ) || !bytes) {return 0 ;
block/nvme.c +13 -1
@@ -1352,7 +1352,7 @@ static coroutine_fn int nvme_co_pwrite_zeroes(BlockDriverState *bs, static int coroutine_fn nvme_co_pdiscard(BlockDriverState *bs,int64_t offset, -int bytes) +int64_t bytes) {BDRVNVMeState *s = bs->opaque;NVMeQueuePair *ioq = s->queues[INDEX_IO (0 )];
@@ -1379,6 +1379,14 @@ static int coroutine_fn nvme_co_pdiscard(BlockDriverState *bs,assert (s->queue_count >1 ); +/* +* Filling the @buf requires @offset and @bytes to satisfy restrictions +* defined in nvme_refresh_limits(). +*/ +assert (QEMU_IS_ALIGNED (bytes,1UL << s->blkshift)); +assert (QEMU_IS_ALIGNED (offset,1UL << s->blkshift)); +assert ((bytes >> s->blkshift) <=UINT32_MAX ); + buf =qemu_try_memalign (s->page_size, s->page_size);if (!buf) {return -ENOMEM ;
@@ -1482,6 +1490,10 @@ static void nvme_refresh_limits(BlockDriverState *bs, Error **errp) bs->bl.max_pwrite_zeroes =1ULL << (s->blkshift +16 ); bs->bl.pwrite_zeroes_alignment =MAX (bs->bl.request_alignment,1UL << s->blkshift); + + bs->bl.max_pdiscard = (uint64_t )UINT32_MAX << s->blkshift; + bs->bl.pdiscard_alignment =MAX (bs->bl.request_alignment, +1UL << s->blkshift); }static void nvme_detach_aio_context (BlockDriverState *bs)
block/preallocate.c +1 -1
@@ -235,7 +235,7 @@ static coroutine_fn int preallocate_co_preadv_part( } static int coroutine_fn preallocate_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {return bdrv_co_pdiscard (bs->file, offset, bytes); }
block/qcow2.c +1 -1
@@ -3996,7 +3996,7 @@ static coroutine_fn int qcow2_co_pwrite_zeroes(BlockDriverState *bs, } static coroutine_fn int qcow2_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {int ret;BDRVQcow2State *s = bs->opaque;
block/raw-format.c +1 -1
@@ -302,7 +302,7 @@ static int coroutine_fn raw_co_pwrite_zeroes(BlockDriverState *bs, } static int coroutine_fn raw_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {int ret;
block/rbd.c +2 -2
@@ -1197,9 +1197,9 @@ static int coroutine_fn qemu_rbd_co_flush(BlockDriverState *bs) } static int coroutine_fn qemu_rbd_co_pdiscard(BlockDriverState *bs, -int64_t offset,int count) +int64_t offset,int64_t bytes) { -return qemu_rbd_start_co (bs, offset, count,NULL ,0 ,RBD_AIO_DISCARD ); +return qemu_rbd_start_co (bs, offset, bytes,NULL ,0 ,RBD_AIO_DISCARD ); }#ifdef LIBRBD_SUPPORTS_WRITE_ZEROES
block/throttle.c +1 -1
@@ -145,7 +145,7 @@ static int coroutine_fn throttle_co_pwrite_zeroes(BlockDriverState *bs, } static int coroutine_fn throttle_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {ThrottleGroupMember *tgm = bs->opaque;throttle_group_co_io_limits_intercept (tgm, bytes, true);
tests/unit/test-block-iothread.c +1 -1
@@ -48,7 +48,7 @@ static int coroutine_fn bdrv_test_co_pwritev(BlockDriverState *bs, } static int coroutine_fn bdrv_test_co_pdiscard(BlockDriverState *bs, -int64_t offset,int bytes) +int64_t offset,int64_t bytes) {return 0 ; }
block/trace-events +2 -2
@@ -152,8 +152,8 @@ nvme_write_zeroes(void *s, uint64_t offset, uint64_t bytes, int flags) "s %p off nvme_qiov_unaligned(const void *qiov, int n, void *base, size_t size, int align) "qiov %p n %d base %p size 0x%zx align 0x%x" nvme_prw_buffered(void *s, uint64_t offset, uint64_t bytes, int niov, int is_write) "s %p offset 0x%"PRIx64" bytes %"PRId64" niov %d is_write %d" nvme_rw_done(void *s, int is_write, uint64_t offset, uint64_t bytes, int ret) "s %p is_write %d offset 0x%"PRIx64" bytes %"PRId64" ret %d" -nvme_dsm(void *s, uint64_t offset, uint64_t bytes) "s %p offset 0x%"PRIx64" bytes %"PRId64"" -nvme_dsm_done(void *s, uint64_t offset, uint64_t bytes, int ret) "s %p offset 0x%"PRIx64" bytes %"PRId64" ret %d" +nvme_dsm(void *s, int64_t offset, int64_t bytes) "s %p offset 0x%"PRIx64" bytes %"PRId64"" +nvme_dsm_done(void *s, int64_t offset, int64_t bytes, int ret) "s %p offset 0x%"PRIx64" bytes %"PRId64" ret %d" nvme_dma_map_flush(void *s) "s %p" nvme_free_req_queue_wait(void *s, unsigned q_index) "s %p q #%u" nvme_create_queue_pair(unsigned q_index, void *q, unsigned size, void *aio_context, int fd) "index %u q %p size %u aioctx %p fd %d"
Signature · 1 line
2.29.2
Re: [PATCH v6 10/11] block: use int64_t instead of int in driver discard handlers
Quoted 16 lines · Vladimir Sementsov-Ogievskiy
We are generally moving to
int64_tfor both offset and bytes parameters on all io paths.Main motivation is realization of 64-bit
write_zeroesoperation for fast zeroing large disk chunks, up to the whole disk.We chose signed type, to be consistent with
off_t(which is signed) and with possibility for signed return type (where negative value means error).So, convert driver discard handlers bytes parameter to
int64_t.The only caller of all updated function is
bdrv_co_pdiscardinblock/io.c. It is already prepared to work with 64bit requests, but pass at mostmax(bs->bl.max_pdiscard, INT_MAX)to the driver.
Quoted 5 lines
Let's look at all updated functions:
Great! Now all drivers are prepared to handle 64bit discard requests, or else have explicit
max_pdiscardlimits.
Very similar analysis to the write-zeroes.
Quoted 11 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>---
+++
b/block/nbd.cblock/nbd.c +1
-1@@ -1457,15 +1457,17 @@ static int nbd_client_co_flush(BlockDriverState *bs) }static int nbd_client_co_pdiscard (BlockDriverState *bs,int64_t offset, -int bytes) +int64_t bytes)
block/nbd.c static int nbd_client_co_flush(BlockDriverState *bs) · line 1469
Quoted 10 lines
block/nbd.c +2
-1{BDRVNBDState *s = (BDRVNBDState *)bs->opaque;NBDRequest request = { .type =NBD_CMD_TRIM , .from = offset, - .len = bytes, + .len = bytes,/* len is uint32_t */ }; +assert (bytes <=UINT32_MAX );/* rely on max_pdiscard */
Aha - you used <= here, compared to < in 7/11 on write zeroes. Consistency says we want <= in both places.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
PATCH v6 11/11 block/io: allow 64bit discard requests
In reply to Vladimir Sementsov-OgievskiyNow that all drivers are updated by the previous commit, we can drop the last limiter on pdiscard path: INT_MAX in bdrv_co_pdiscard().
Now everything is prepared for implementing incredibly cool and fast big-discard requests in NBD and qcow2. And any other driver which wants it of course.
--- block/io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
block/io.c +1 -1
@@ -3042,7 +3042,7 @@ int coroutine_fn bdrv_co_pdiscard(BdrvChild *child, int64_t offset, goto out; } - max_pdiscard =QEMU_ALIGN_DOWN (MIN_NON_ZERO (bs->bl.max_pdiscard,INT_MAX ), + max_pdiscard =QEMU_ALIGN_DOWN (MIN_NON_ZERO (bs->bl.max_pdiscard,INT64_MAX ), align);assert (max_pdiscard >= bs->bl.request_alignment);
Signature · 1 line
2.29.2
Ping)
Not reviewed: 6,7,10
Quoted 43 lines · Vladimir Sementsov-Ogievskiy
Hi all!
Sorry for a long delay:( Finally, here is v6.
v6: rebase on new rbd handlers and backup-top renamed to copy-before-write. Also:
01: add Eric's r-b
tweak commit msg to not mention sheepdog
02: add Eric's r-b
03: tweak commit msg
drop extra type conversion inraw_co_pwrite_zeroes04: vvfat: no
write_target_commit()anymore
nvme: fix over-80 line
type conversion inraw_co_pwrite_zeroes()now dropped in 0305: tweak commit msg
add Eric's r-b06: update comment
07: tweak commit msg, grammar and typos in comments
don't add extra empty line
handle alsoqemu_rbd_co_pwrite_zeroes08: tweak commit msg, add Eric's r-b 09: tweak commit msg, add Eric's r-b 10: add rbd, drop sheepdog
add assertion toiscsi_co_pdiscard()
11: tweak commit msg, add Eric's r-bVladimir Sementsov-Ogievskiy (11):
block/io: bring request check to bdrv_co_(read,write)v_vmstate
qcow2: check request on vmstate save/load path
block: useint64_tinstead ofuint64_tin driver read handlers
block: useint64_tinstead ofuint64_tin driver write handlers
block: useint64_tinstead ofuint64_tincopy_rangedriver handlers
block: makeBlockLimits::max_pwrite_zeroes64bit
block: useint64_tinstead of int in driverwrite_zeroeshandlers block/io: allow 64bit write-zeroes requests
block: makeBlockLimits::max_pdiscard64bit
block: useint64_tinstead of int in driver discard handlers block/io: allow 64bit discard requests
Signature · 2 lines
Best regards, Vladimir
Quoted 3 lines · Vladimir Sementsov-Ogievskiy
Ping)
Not reviewed: 6,7,10
Also on my list to see what I can get in this week's NBD pull request.
Quoted 25 lines
Quoted 17 lines · Vladimir Sementsov-Ogievskiy
Hi all!
Sorry for a long delay:( Finally, here is v6.
Vladimir Sementsov-Ogievskiy (11):
block/io: bring request check to bdrv_co_(read,write)v_vmstate
qcow2: check request on vmstate save/load path
block: useint64_tinstead ofuint64_tin driver read handlers
block: useint64_tinstead ofuint64_tin driver write handlers
block: useint64_tinstead ofuint64_tincopy_rangedriver handlers
block: makeBlockLimits::max_pwrite_zeroes64bit
block: useint64_tinstead of int in driverwrite_zeroeshandlers block/io: allow 64bit write-zeroes requests
block: makeBlockLimits::max_pdiscard64bit
block: useint64_tinstead of int in driver discard handlers block/io: allow 64bit discard requestsSignature · 2 lines
Best regards, Vladimir
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org