block: local qiov helper: part I
v2 cover block: local qiov helper: part I
Hi all!
Here is a proposal for a new simple helper for a very often patter around qemu_iovec_init_external, when we need simple qiov with only one iov, initialized from external buffer.
Here only block/io.c updated to use new helper, I'll update other things on top of this separately.
v2:
- smarter padding for @size and changed structure
- other fixes (described in each patch in Notes)
Vladimir Sementsov-Ogievskiy (2):
block: enhance QEMUIOVector structure
block/io: use qemu_iovec_init_buf
include/qemu/iov.h | 47 ++++++++++++++++++++++-- block/io.c | 90 +++++++++++----------------------------------- 2 files changed, 66 insertions(+), 71 deletions(-)
Signature · 1 line
2.18.0
PATCH v2 1/2 block: enhance QEMUIOVector structure
Add a possibility of embedded iovec, for cases when we need only one local iov.
--- include/qemu/iov.h | 47 ++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 45 insertions(+), 2 deletions(-)
include/qemu/iov.h +45 -2
@@ -133,10 +133,53 @@ size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt,typedef struct QEMUIOVector {struct iovec *iov;int niov; -int nalloc; -size_t size; + +/* +* For external @iov (qemu_iovec_init_external()) or allocated @iov +* (qemu_iovec_init()) @size is the cumulative size of iovecs and +* @local_iov is invalid and unused. +* +* For embedded @iov (QEMU_IOVEC_INIT_BUF() or qemu_iovec_init_buf()), +* @iov is equal to &@local_iov, and @size is valid, as it has same +* offset and type as @local_iov.iov_len, which is guaranteed by +* static assertions below. +* +* @nalloc is valid always and is -1 both for embedded and external +* cases. It included into union only to make appropriate padding for +* @size field avoiding creation of 0-length array in the worst case. +*/ +union { +struct { +int nalloc; +struct iovec local_iov; + }; +struct { +char __pad[sizeof (int ) + offsetof(struct iovec , iov_len)]; +size_t size; + }; + }; }QEMUIOVector ; +QEMU_BUILD_BUG_ON (offsetof(QEMUIOVector , size) != + offsetof(QEMUIOVector, local_iov.iov_len)); + +#define QEMU_IOVEC_INIT_BUF(self, buf, len) \ +{ \ + .iov = &(self).local_iov, \ + .niov = 1, \ + .nalloc = -1, \ + .local_iov = { \ + .iov_base = (void *)(buf), /* cast away const */ \ + .iov_len = (len), \ + }, \ +} + +static inline void qemu_iovec_init_buf (QEMUIOVector *qiov, +void *buf,size_t len) +{ + *qiov = (QEMUIOVector )QEMU_IOVEC_INIT_BUF (*qiov, buf, len); +} +void qemu_iovec_init (QEMUIOVector *qiov,int alloc_hint);void qemu_iovec_init_external (QEMUIOVector *qiov,struct iovec *iov,int niov);void qemu_iovec_add (QEMUIOVector *qiov,void *base,size_t len);
Signature · 1 line
2.18.0
Re: [Qemu-devel] [PATCH v2 1/2] block: enhance QEMUIOVector structure
Quoted 13 lines · Vladimir Sementsov-Ogievskiy
Add a possibility of embedded iovec, for cases when we need only one local iov.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- include/qemu/iov.h | 47 ++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 45 insertions(+), 2 deletions(-)include/qemu/iov.h +0
-0
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 139
Quoted 9 lines
include/qemu/iov.h +4
-2@@ -133,10 +133,53 @@ size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, typedef struct QEMUIOVector { struct iovec *iov; int niov; - int nalloc; - size_t size; + + /* + * For external @iov (qemu_iovec_init_external()) or allocated @iov + * (qemu_iovec_init()) @size is the cumulative size of iovecs and
s/ @size/, @size/
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 145
Quoted 6 lines
include/qemu/iov.h +6
-0+* @local_iov is invalid and unused. +* +* For embedded @iov (QEMU_IOVEC_INIT_BUF() or qemu_iovec_init_buf()), +* @iov is equal to &@local_iov, and @size is valid, as it has same +* offset and type as @local_iov.iov_len, which is guaranteed by +* static assertions below.
Only one static assertion below (s/assertions/assertion/), but even that one could perhaps be dropped and this wording changed to "which is guaranteed by the layout below"; or you could restore the assertion in the earlier patch that sizeof(size) and sizeof(struct iovec.iov_len) are equal) to make the plural correct.
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 147
Quoted 2 lines
include/qemu/iov.h +2
-0+* +* @nalloc is valid always and is -1 both for embedded and external
s/valid always/always valid/
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 149
Quoted 2 lines
include/qemu/iov.h +2
-0+* cases. It included into union only to make appropriate padding for +* @size field avoiding creation of 0-length array in the worst case.
It is included in the union only to ensure the padding prior to the @size field will not result in a 0-length array.
Quoted 5 lines
include/qemu/iov.h +5
-0+*/ + union { + struct { + int nalloc; + struct iovec local_iov;
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 164
Quoted 10 lines
include/qemu/iov.h +8
-0+ }; +struct { +char __pad[sizeof (int ) + offsetof(struct iovec , iov_len)]; +size_t size; + }; + }; } QEMUIOVector; +QEMU_BUILD_BUG_ON (offsetof(QEMUIOVector , size) != + offsetof(QEMUIOVector, local_iov.iov_len));
I'm not sure this assertion adds any value; I don't see any leeway on how a compiler could lay out the struct based on the declaration of the padding. However, I'm not opposed to keeping it in the patch if someone else finds it useful.
Quoted 4 lines
include/qemu/iov.h +4
-0+ +#define QEMU_IOVEC_INIT_BUF (self, buf, len) \ +{ \ + .iov = &(self).local_iov, \
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 178
Quoted 10 lines
include/qemu/iov.h +10
-0+ .niov = 1, \ + .nalloc = -1, \ + .local_iov = { \ + .iov_base = (void *)(buf), /* cast away const */ \ + .iov_len = (len), \ + }, \ +} + +static inline void qemu_iovec_init_buf (QEMUIOVector *qiov, +void *buf,size_t len)
Should this be 'const void *buf', to make it easier to initialize a qiov used for writes from an incoming const pointer? That, and having const here would make the 'cast away const' comment above all the more obvious (I know it is necessary based on code in patch 2, but having it be worthwhile in patch 1 makes it all the more obvious as a standalone patch).
Quoted 8 lines
include/qemu/iov.h +4
-0+{ + *qiov = (QEMUIOVector )QEMU_IOVEC_INIT_BUF (*qiov, buf, len); +} +void qemu_iovec_init (QEMUIOVector *qiov,int alloc_hint);void qemu_iovec_init_external (QEMUIOVector *qiov,struct iovec *iov,int niov);void qemu_iovec_add (QEMUIOVector *qiov,void *base,size_t len);
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v2 1/2] block: enhance QEMUIOVector structure
Quoted 63 lines · Eric Blake
Quoted 13 lines · Vladimir Sementsov-Ogievskiy
Add a possibility of embedded iovec, for cases when we need only one local iov.
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- include/qemu/iov.h | 47 ++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 45 insertions(+), 2 deletions(-)include/qemu/iov.h +0
-0include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 139
Quoted 9 lines
include/qemu/iov.h +4
-2@@ -133,10 +133,53 @@ size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, typedef struct QEMUIOVector { struct iovec *iov; int niov; - int nalloc; - size_t size; + + /* + * For external @iov (qemu_iovec_init_external()) or allocated @iov + * (qemu_iovec_init()) @size is the cumulative size of iovecs ands/ @size/, @size/
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 145
Quoted 6 lines
include/qemu/iov.h +6
-0+* @local_iov is invalid and unused. +* +* For embedded @iov (QEMU_IOVEC_INIT_BUF() or qemu_iovec_init_buf()), +* @iov is equal to &@local_iov, and @size is valid, as it has same +* offset and type as @local_iov.iov_len, which is guaranteed by +* static assertions below. Only one static assertion below (s/assertions/assertion/), but even that one could perhaps be dropped and this wording changed to "which is guaranteed by the layout below"; or you could restore the assertion in the earlier patch that sizeof(size) and
sizeof(struct iovec.iov_len)are equal) to make the plural correct.include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 147
Quoted 2 lines
include/qemu/iov.h +2
-0+* +* @nalloc is valid always and is -1 both for embedded and external s/valid always/always valid/
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 149
Quoted 2 lines
include/qemu/iov.h +2
-0+* cases. It included into union only to make appropriate padding for +* @size field avoiding creation of 0-length array in the worst case. It is included in the union only to ensure the padding prior to the @size field will not result in a 0-length array.
Quoted 13 lines
include/qemu/iov.h +11
-0+*/ +union { +struct { +int nalloc; +struct iovec local_iov; + }; +struct { +char __pad[sizeof (int ) + offsetof(struct iovec , iov_len)]; +size_t size; + }; + }; } QEMUIOVector;
Quoted 7 lines
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 164
Quoted 2 lines
include/qemu/iov.h +2
-0+QEMU_BUILD_BUG_ON (offsetof(QEMUIOVector , size) != + offsetof(QEMUIOVector, local_iov.iov_len));I'm not sure this assertion adds any value; I don't see any leeway on how a compiler could lay out the struct based on the declaration of the padding. However, I'm not opposed to keeping it in the patch if someone else finds it useful.
Assertion is a bit simpler to understand than structure layout. And it exactly asserts what the comment say about @size...
Quoted 13 lines
Quoted 13 lines
include/qemu/iov.h +12
-0+ +#define QEMU_IOVEC_INIT_BUF(self, buf, len) \ +{ \ + .iov = &(self).local_iov, \ + .niov = 1, \ + .nalloc = -1, \ + .local_iov = { \ + .iov_base = (void *)(buf), /* cast away const */ \ + .iov_len = (len), \ + }, \ +} +
Quoted 8 lines
include/qemu/iov.h size_t iov_discard_back(struct iovec *iov, unsigned int *iov_cnt, · line 178
Quoted 2 lines
include/qemu/iov.h +2
-0+static inline void qemu_iovec_init_buf (QEMUIOVector *qiov, +void *buf,size_t len)Should this be 'const void *buf', to make it easier to initialize a qiov used for writes from an incoming const pointer? That, and having const here would make the 'cast away const' comment above all the more obvious (I know it is necessary based on code in patch 2, but having it be worthwhile in patch 1 makes it all the more obvious as a standalone patch).
I think, yes. So, we'll be able to use both macro and function for const buffers in the same way
Quoted 10 lines
Quoted 9 lines
include/qemu/iov.h +4
-0+{ + *qiov = (QEMUIOVector )QEMU_IOVEC_INIT_BUF (*qiov, buf, len); +} +void qemu_iovec_init (QEMUIOVector *qiov,int alloc_hint);void qemu_iovec_init_external (QEMUIOVector *qiov,struct iovec *iov,int niov);void qemu_iovec_add (QEMUIOVector *qiov,void *base,size_t len);
Signature · 2 lines
Best regards, Vladimir
PATCH v2 2/2 block/io: use qemu_iovec_init_buf
In reply to Vladimir Sementsov-OgievskiyUse new qemu_iovec_init_buf() instead of qemu_iovec_init_external( ... , 1), which simplifies the code.
--- block/io.c | 90 +++++++++++++----------------------------------------- 1 file changed, 21 insertions(+), 69 deletions(-)
block/io.c +21 -69
@@ -842,17 +842,13 @@ static int bdrv_prwv_co(BdrvChild *child, int64_t offset,static int bdrv_rw_co (BdrvChild *child,int64_t sector_num,uint8_t *buf,int nb_sectors,bool is_write,BdrvRequestFlags flags) { -QEMUIOVector qiov; -struct iovec iov = { - .iov_base = (void *)buf, - .iov_len = nb_sectors *BDRV_SECTOR_SIZE , - }; +QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov, buf, + nb_sectors *BDRV_SECTOR_SIZE );if (nb_sectors <0 || nb_sectors >BDRV_REQUEST_MAX_SECTORS ) {return -EINVAL ; } -qemu_iovec_init_external (&qiov, &iov,1 );return bdrv_prwv_co (child, sector_num <<BDRV_SECTOR_BITS , &qiov, is_write, flags); }
@@ -879,13 +875,8 @@ int bdrv_write(BdrvChild *child, int64_t sector_num,int bdrv_pwrite_zeroes (BdrvChild *child,int64_t offset,int bytes,BdrvRequestFlags flags) { -QEMUIOVector qiov; -struct iovec iov = { - .iov_base =NULL , - .iov_len = bytes, - }; +QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov,NULL , bytes); -qemu_iovec_init_external (&qiov, &iov,1 );return bdrv_prwv_co (child, offset, &qiov, true,BDRV_REQ_ZERO_WRITE | flags); }
@@ -949,17 +940,12 @@ int bdrv_preadv(BdrvChild *child, int64_t offset, QEMUIOVector *qiov)int bdrv_pread (BdrvChild *child,int64_t offset,void *buf,int bytes) { -QEMUIOVector qiov; -struct iovec iov = { - .iov_base = (void *)buf, - .iov_len = bytes, - }; +QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov, buf, bytes);if (bytes <0 ) {return -EINVAL ; } -qemu_iovec_init_external (&qiov, &iov,1 );return bdrv_preadv (child, offset, &qiov); }
@@ -977,17 +963,12 @@ int bdrv_pwritev(BdrvChild *child, int64_t offset, QEMUIOVector *qiov)int bdrv_pwrite (BdrvChild *child,int64_t offset,const void *buf,int bytes) { -QEMUIOVector qiov; -struct iovec iov = { - .iov_base = (void *) buf, - .iov_len = bytes, - }; +QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov, buf, bytes);if (bytes <0 ) {return -EINVAL ; } -qemu_iovec_init_external (&qiov, &iov,1 );return bdrv_pwritev (child, offset, &qiov); }
@@ -1164,7 +1145,6 @@ static int coroutine_fn bdrv_co_do_copy_on_readv(BdrvChild *child,void *bounce_buffer;BlockDriver *drv = bs->drv; -struct iovec iov;QEMUIOVector local_qiov;int64_t cluster_offset;int64_t cluster_bytes;
@@ -1229,9 +1209,8 @@ static int coroutine_fn bdrv_co_do_copy_on_readv(BdrvChild *child,if (ret <=0 ) {/* Must copy-on-read; use the bounce buffer */ - iov.iov_base = bounce_buffer; - iov.iov_len = pnum =MIN (pnum,MAX_BOUNCE_BUFFER ); -qemu_iovec_init_external (&local_qiov, &iov,1 ); + pnum =MIN (pnum,MAX_BOUNCE_BUFFER ); +qemu_iovec_init_buf (&local_qiov, bounce_buffer, pnum); ret =bdrv_driver_preadv (bs, cluster_offset, pnum, &local_qiov,0 );
@@ -1476,7 +1455,7 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs, {BlockDriver *drv = bs->drv;QEMUIOVector qiov; -struct iovec iov = {0 }; +void *buf =NULL ;int ret =0 ;bool need_flush = false;int head =0 ;
@@ -1546,16 +1525,15 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs, need_flush = true; } num =MIN (num, max_transfer); - iov.iov_len = num; -if (iov.iov_base ==NULL ) { - iov.iov_base =qemu_try_blockalign (bs, num); -if (iov.iov_base ==NULL ) { +if (buf ==NULL ) { + buf =qemu_try_blockalign (bs, num); +if (buf ==NULL ) { ret = -ENOMEM ; goto fail; } -memset (iov.iov_base,0 , num); +memset (buf,0 , num); } -qemu_iovec_init_external (&qiov, &iov,1 ); +qemu_iovec_init_buf (&qiov, buf, num); ret =bdrv_driver_pwritev (bs, offset, num, &qiov, write_flags);
@@ -1563,8 +1541,8 @@ static int coroutine_fn bdrv_co_do_pwrite_zeroes(BlockDriverState *bs,* all future requests. */ if (num < max_transfer) { -qemu_vfree (iov.iov_base); - iov.iov_base =NULL ; +qemu_vfree (buf); + buf =NULL ; } }
@@ -1576,7 +1554,7 @@ fail:if (ret ==0 && need_flush) { ret =bdrv_co_flush (bs); } -qemu_vfree (iov.iov_base); +qemu_vfree (buf);return ret; }
@@ -1762,7 +1740,6 @@ static int coroutine_fn bdrv_co_do_zero_pwritev(BdrvChild *child,BlockDriverState *bs = child->bs;uint8_t *buf =NULL ;QEMUIOVector local_qiov; -struct iovec iov;uint64_t align = bs->bl.request_alignment;unsigned int head_padding_bytes, tail_padding_bytes;int ret =0 ;
@@ -1774,11 +1751,7 @@ static int coroutine_fn bdrv_co_do_zero_pwritev(BdrvChild *child,assert (flags &BDRV_REQ_ZERO_WRITE );if (head_padding_bytes || tail_padding_bytes) { buf =qemu_blockalign (bs, align); - iov = (struct iovec ) { - .iov_base = buf, - .iov_len = align, - }; -qemu_iovec_init_external (&local_qiov, &iov,1 ); +qemu_iovec_init_buf (&local_qiov, buf, align); }if (head_padding_bytes) {uint64_t zero_bytes =MIN (bytes, align - head_padding_bytes);
@@ -1884,17 +1857,12 @@ int coroutine_fn bdrv_co_pwritev(BdrvChild *child,if (offset & (align -1 )) {QEMUIOVector head_qiov; -struct iovec head_iov;mark_request_serialising (&req, align);wait_serialising_requests (&req); head_buf =qemu_blockalign (bs, align); - head_iov = (struct iovec ) { - .iov_base = head_buf, - .iov_len = align, - }; -qemu_iovec_init_external (&head_qiov, &head_iov,1 ); +qemu_iovec_init_buf (&head_qiov, head_buf, align);bdrv_debug_event (bs,BLKDBG_PWRITEV_RMW_HEAD ); ret = bdrv_aligned_preadv(child, &req, offset & ~(align - 1), align,
@@ -1923,7 +1891,6 @@ int coroutine_fn bdrv_co_pwritev(BdrvChild *child,if ((offset + bytes) & (align -1 )) {QEMUIOVector tail_qiov; -struct iovec tail_iov;size_t tail_bytes;bool waited;
@@ -1932,11 +1899,7 @@ int coroutine_fn bdrv_co_pwritev(BdrvChild *child,assert (!waited || !use_local_qiov); tail_buf =qemu_blockalign (bs, align); - tail_iov = (struct iovec ) { - .iov_base = tail_buf, - .iov_len = align, - }; -qemu_iovec_init_external (&tail_qiov, &tail_iov,1 ); +qemu_iovec_init_buf (&tail_qiov, tail_buf, align);bdrv_debug_event (bs,BLKDBG_PWRITEV_RMW_TAIL ); ret = bdrv_aligned_preadv(child, &req, (offset + bytes) & ~(align - 1),
@@ -2465,15 +2428,9 @@ bdrv_rw_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos,int bdrv_save_vmstate (BlockDriverState *bs,const uint8_t *buf,int64_t pos,int size) { -QEMUIOVector qiov; -struct iovec iov = { - .iov_base = (void *) buf, - .iov_len = size, - }; +QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov, buf, size);int ret; -qemu_iovec_init_external (&qiov, &iov,1 ); - ret =bdrv_writev_vmstate (bs, &qiov, pos);if (ret <0 ) {return ret;
@@ -2490,14 +2447,9 @@ int bdrv_writev_vmstate(BlockDriverState *bs, QEMUIOVector *qiov, int64_t pos)int bdrv_load_vmstate (BlockDriverState *bs,uint8_t *buf,int64_t pos,int size) { -QEMUIOVector qiov; -struct iovec iov = { - .iov_base = buf, - .iov_len = size, - }; +QEMUIOVector qiov =QEMU_IOVEC_INIT_BUF (qiov, buf, size);int ret; -qemu_iovec_init_external (&qiov, &iov,1 ); ret =bdrv_readv_vmstate (bs, &qiov, pos);if (ret <0 ) {return ret;
Signature · 1 line
2.18.0
Re: [Qemu-devel] [PATCH v2 2/2] block/io: use qemu_iovec_init_buf
Quoted 2 lines · Vladimir Sementsov-Ogievskiy
Use new
qemu_iovec_init_buf()instead of qemu_iovec_init_external( ... , 1), which simplifies the code.
Did you just do a manual search for candidate callers?
As you said in the cover letter, there are other files that can benefit as well; are you planning on making v3 of the series longer?
Quoted 5 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/io.c | 90 +++++++++++++----------------------------------------- 1 file changed, 21 insertions(+), 69 deletions(-)
But I'm loving the diffstat - it is definitely a nice change.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v2 2/2] block/io: use qemu_iovec_init_buf
Quoted 8 lines · Eric Blake
Quoted 2 lines · Vladimir Sementsov-Ogievskiy
Use new
qemu_iovec_init_buf()instead of qemu_iovec_init_external( ... , 1), which simplifies the code.Did you just do a manual search for candidate callers?
As you said in the cover letter, there are other files that can benefit as well; are you planning on making v3 of the series longer?
git grep qemu_iovec_init_external | grep 1
shows a lot of, exactly 34 after io.c already updated.
They are in different subsystems, so I think it should be simpler to push this one as a precedent and example, and then send separate patches (or series) per-maintainer.
hm, in other words:
# git grep -l 'qemu_iovec_init_external.*1' block/backup.c block/block-backend.c block/commit.c block/parallels.c block/qcow.c block/qcow2.c block/qed-table.c block/qed.c block/stream.c block/vmdk.c hw/ide/atapi.c hw/ide/core.c hw/scsi/scsi-disk.c migration/block.c qemu-img.c tests/test-bdrv-drain.c
Quoted 11 lines
Quoted 6 lines
Signed-off-by Vladimir Sementsov-Ogievskiy <[email protected]>--- block/io.c | 90 +++++++++++++----------------------------------------- 1 file changed, 21 insertions(+), 69 deletions(-)But I'm loving the diffstat - it is definitely a nice change.
Reviewed-by Eric Blake <[email protected]>
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v2 2/2] block/io: use qemu_iovec_init_buf
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Quoted 8 lines · Eric Blake
Quoted 2 lines · Vladimir Sementsov-Ogievskiy
Use new
qemu_iovec_init_buf()instead of qemu_iovec_init_external( ... , 1), which simplifies the code.Did you just do a manual search for candidate callers?
As you said in the cover letter, there are other files that can benefit as well; are you planning on making v3 of the series longer?
Quoted 6 lines
git grep
qemu_iovec_init_external| grep 1shows a lot of, exactly 34 after
io.calready updated.
They are in different subsystems, so I think it should be simpler to push this one as a precedent and example, and then send separate patches (or series) per-maintainer.
Most are block related, so getting it in through the block maintainers is probably the easiest - but you ARE right that one patch per one or two files or two is going to be smartest (otherwise, it gets too big).
Quoted 11 lines
hm, in other words:
# git grep -l 'qemu_iovec_init_external.*1' block/backup.c block/block-backend.c block/commit.c block/parallels.c block/qcow.c block/qcow2.c block/qed-table.c block/qed.c
Quoted 8 lines
block/stream.c block/vmdk.c hw/ide/atapi.c hw/ide/core.c hw/scsi/scsi-disk.c migration/block.c qemu-img.c tests/test-bdrv-drain.c
I'd group qed-table.c and qed.c; and the two hw/ide/ files; resulting in 14 more patches to go.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org
Re: [Qemu-devel] [PATCH v2 2/2] block/io: use qemu_iovec_init_buf
Quoted 42 lines · Eric Blake
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Quoted 8 lines · Eric Blake
Quoted 2 lines · Vladimir Sementsov-Ogievskiy
Use new
qemu_iovec_init_buf()instead of qemu_iovec_init_external( ... , 1), which simplifies the code.Did you just do a manual search for candidate callers?
As you said in the cover letter, there are other files that can benefit as well; are you planning on making v3 of the series longer?
Quoted 6 lines
git grep
qemu_iovec_init_external| grep 1shows a lot of, exactly 34 after
io.calready updated.
They are in different subsystems, so I think it should be simpler to push this one as a precedent and example, and then send separate patches (or series) per-maintainer.Most are block related, so getting it in through the block maintainers is probably the easiest - but you ARE right that one patch per one or two files or two is going to be smartest (otherwise, it gets too big).
Quoted 11 lines
hm, in other words:
# git grep -l 'qemu_iovec_init_external.*1' block/backup.c block/block-backend.c block/commit.c block/parallels.c block/qcow.c block/qcow2.c block/qed-table.c block/qed.cQuoted 8 lines
block/stream.c block/vmdk.c hw/ide/atapi.c hw/ide/core.c hw/scsi/scsi-disk.c migration/block.c qemu-img.c tests/test-bdrv-drain.c
Quoted 3 lines
I'd group
qed-table.candqed.c; and the two hw/ide/ files; resulting in 14 more patches to go.
So, you, think better is to make one common patch set? Ok, I'll do (hmm or start doing) it tomorrow as v3 if no other opinions.
Signature · 2 lines
Best regards, Vladimir
Re: [Qemu-devel] [PATCH v2 2/2] block/io: use qemu_iovec_init_buf
Quoted 10 lines · Vladimir Sementsov-Ogievskiy
Quoted 9 lines
Quoted 4 lines
shows a lot of, exactly 34 after
io.calready updated.
They are in different subsystems, so I think it should be simpler to push this one as a precedent and example, and then send separate patches (or series) per-maintainer.Most are block related, so getting it in through the block maintainers is probably the easiest - but you ARE right that one patch per one or two files or two is going to be smartest (otherwise, it gets too big).
Quoted 6 lines
Quoted 3 lines
I'd group
qed-table.candqed.c; and the two hw/ide/ files; resulting in 14 more patches to go.So, you, think better is to make one common patch set? Ok, I'll do (hmm or start doing) it tomorrow as v3 if no other opinions.
Yes, a v3 series of 16 patches for the entire conversion, and cc:
qemu-block, should be sufficient. Looking forward to seeing v3.
Signature · 3 lines
Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3226 Virtualization: qemu.org | libvirt.org