* [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-21 6:50 ` jazeltq
0 siblings, 0 replies; 12+ messages in thread
From: jazeltq @ 2017-02-21 6:50 UTC (permalink / raw)
To: jdurgin, jcody, dillaman, kwolf, mreitz, qemu-block
Cc: tianqing, ceph-devel, jazeltq, qemu-devel
From: tianqing <tianqing@unitedstack.com>
Rbd can do readv and writev directly, so wo do not need to transform
iov to buf or vice versa any more.
Signed-off-by: tianqing <tianqing@unitedstack.com>
---
block/rbd.c | 80 ++++++++++++++++++++++++++++++++++++++++++-------------------
1 file changed, 56 insertions(+), 24 deletions(-)
diff --git a/block/rbd.c b/block/rbd.c
index a57b3e3..22e8e69 100644
--- a/block/rbd.c
+++ b/block/rbd.c
@@ -62,6 +62,13 @@
#define RBD_MAX_SNAP_NAME_SIZE 128
#define RBD_MAX_SNAPS 100
+/* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */
+#ifdef LIBRBD_SUPPORTS_IOVEC
+#define LIBRBD_USE_IOVEC 1
+#else
+#define LIBRBD_USE_IOVEC 0
+#endif
+
typedef enum {
RBD_AIO_READ,
RBD_AIO_WRITE,
@@ -310,6 +317,17 @@ static int qemu_rbd_set_conf(rados_t cluster, const char *conf,
return ret;
}
+static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
+{
+ if (LIBRBD_USE_IOVEC) {
+ RBDAIOCB *acb = rcb->acb;
+ iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
+ acb->qiov->size - offs);
+ } else {
+ memset(rcb->buf + offs, 0, rcb->size - offs);
+ }
+}
+
static int qemu_rbd_create(const char *filename, QemuOpts *opts, Error **errp)
{
Error *local_err = NULL;
@@ -426,11 +444,11 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
}
} else {
if (r < 0) {
- memset(rcb->buf, 0, rcb->size);
+ qemu_rbd_memset(rcb, 0);
acb->ret = r;
acb->error = 1;
} else if (r < rcb->size) {
- memset(rcb->buf + r, 0, rcb->size - r);
+ qemu_rbd_memset(rcb, r);
if (!acb->error) {
acb->ret = rcb->size;
}
@@ -441,10 +459,13 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
g_free(rcb);
- if (acb->cmd == RBD_AIO_READ) {
- qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
+ if (!LIBRBD_USE_IOVEC) {
+ if (acb->cmd == RBD_AIO_READ) {
+ qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
+ }
+ qemu_vfree(acb->bounce);
}
- qemu_vfree(acb->bounce);
+
acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
qemu_aio_unref(acb);
@@ -655,7 +676,6 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
RBDAIOCB *acb;
RADOSCB *rcb = NULL;
rbd_completion_t c;
- char *buf;
int r;
BDRVRBDState *s = bs->opaque;
@@ -664,27 +684,29 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
acb->cmd = cmd;
acb->qiov = qiov;
assert(!qiov || qiov->size == size);
- if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
- acb->bounce = NULL;
- } else {
- acb->bounce = qemu_try_blockalign(bs, qiov->size);
- if (acb->bounce == NULL) {
- goto failed;
+
+ rcb = g_new(RADOSCB, 1);
+
+ if (!LIBRBD_USE_IOVEC) {
+ if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
+ acb->bounce = NULL;
+ } else {
+ acb->bounce = qemu_try_blockalign(bs, qiov->size);
+ if (acb->bounce == NULL) {
+ goto failed;
+ }
}
+ if (cmd == RBD_AIO_WRITE) {
+ qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
+ }
+ rcb->buf = acb->bounce;
}
+
acb->ret = 0;
acb->error = 0;
acb->s = s;
- if (cmd == RBD_AIO_WRITE) {
- qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
- }
-
- buf = acb->bounce;
-
- rcb = g_new(RADOSCB, 1);
rcb->acb = acb;
- rcb->buf = buf;
rcb->s = acb->s;
rcb->size = size;
r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
@@ -694,10 +716,18 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
switch (cmd) {
case RBD_AIO_WRITE:
- r = rbd_aio_write(s->image, off, size, buf, c);
+#ifdef LIBRBD_SUPPORTS_IOVEC
+ r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
+#else
+ r = rbd_aio_write(s->image, off, size, rcb->buf, c);
+#endif
break;
case RBD_AIO_READ:
- r = rbd_aio_read(s->image, off, size, buf, c);
+#ifdef LIBRBD_SUPPORTS_IOVEC
+ r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
+#else
+ r = rbd_aio_read(s->image, off, size, rcb->buf, c);
+#endif
break;
case RBD_AIO_DISCARD:
r = rbd_aio_discard_wrapper(s->image, off, size, c);
@@ -712,14 +742,16 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
if (r < 0) {
goto failed_completion;
}
-
return &acb->common;
failed_completion:
rbd_aio_release(c);
failed:
g_free(rcb);
- qemu_vfree(acb->bounce);
+ if (!LIBRBD_USE_IOVEC) {
+ qemu_vfree(acb->bounce);
+ }
+
qemu_aio_unref(acb);
return NULL;
}
--
2.10.2
^ permalink raw reply related [flat|nested] 12+ messages in thread
* [Qemu-devel] [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-21 6:50 ` jazeltq
0 siblings, 0 replies; 12+ messages in thread
From: jazeltq @ 2017-02-21 6:50 UTC (permalink / raw)
To: jdurgin, jcody, dillaman, kwolf, mreitz, qemu-block
Cc: qemu-devel, ceph-devel, jazeltq, tianqing
From: tianqing <tianqing@unitedstack.com>
Rbd can do readv and writev directly, so wo do not need to transform
iov to buf or vice versa any more.
Signed-off-by: tianqing <tianqing@unitedstack.com>
---
block/rbd.c | 80 ++++++++++++++++++++++++++++++++++++++++++-------------------
1 file changed, 56 insertions(+), 24 deletions(-)
diff --git a/block/rbd.c b/block/rbd.c
index a57b3e3..22e8e69 100644
--- a/block/rbd.c
+++ b/block/rbd.c
@@ -62,6 +62,13 @@
#define RBD_MAX_SNAP_NAME_SIZE 128
#define RBD_MAX_SNAPS 100
+/* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */
+#ifdef LIBRBD_SUPPORTS_IOVEC
+#define LIBRBD_USE_IOVEC 1
+#else
+#define LIBRBD_USE_IOVEC 0
+#endif
+
typedef enum {
RBD_AIO_READ,
RBD_AIO_WRITE,
@@ -310,6 +317,17 @@ static int qemu_rbd_set_conf(rados_t cluster, const char *conf,
return ret;
}
+static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
+{
+ if (LIBRBD_USE_IOVEC) {
+ RBDAIOCB *acb = rcb->acb;
+ iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
+ acb->qiov->size - offs);
+ } else {
+ memset(rcb->buf + offs, 0, rcb->size - offs);
+ }
+}
+
static int qemu_rbd_create(const char *filename, QemuOpts *opts, Error **errp)
{
Error *local_err = NULL;
@@ -426,11 +444,11 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
}
} else {
if (r < 0) {
- memset(rcb->buf, 0, rcb->size);
+ qemu_rbd_memset(rcb, 0);
acb->ret = r;
acb->error = 1;
} else if (r < rcb->size) {
- memset(rcb->buf + r, 0, rcb->size - r);
+ qemu_rbd_memset(rcb, r);
if (!acb->error) {
acb->ret = rcb->size;
}
@@ -441,10 +459,13 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
g_free(rcb);
- if (acb->cmd == RBD_AIO_READ) {
- qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
+ if (!LIBRBD_USE_IOVEC) {
+ if (acb->cmd == RBD_AIO_READ) {
+ qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
+ }
+ qemu_vfree(acb->bounce);
}
- qemu_vfree(acb->bounce);
+
acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
qemu_aio_unref(acb);
@@ -655,7 +676,6 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
RBDAIOCB *acb;
RADOSCB *rcb = NULL;
rbd_completion_t c;
- char *buf;
int r;
BDRVRBDState *s = bs->opaque;
@@ -664,27 +684,29 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
acb->cmd = cmd;
acb->qiov = qiov;
assert(!qiov || qiov->size == size);
- if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
- acb->bounce = NULL;
- } else {
- acb->bounce = qemu_try_blockalign(bs, qiov->size);
- if (acb->bounce == NULL) {
- goto failed;
+
+ rcb = g_new(RADOSCB, 1);
+
+ if (!LIBRBD_USE_IOVEC) {
+ if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
+ acb->bounce = NULL;
+ } else {
+ acb->bounce = qemu_try_blockalign(bs, qiov->size);
+ if (acb->bounce == NULL) {
+ goto failed;
+ }
}
+ if (cmd == RBD_AIO_WRITE) {
+ qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
+ }
+ rcb->buf = acb->bounce;
}
+
acb->ret = 0;
acb->error = 0;
acb->s = s;
- if (cmd == RBD_AIO_WRITE) {
- qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
- }
-
- buf = acb->bounce;
-
- rcb = g_new(RADOSCB, 1);
rcb->acb = acb;
- rcb->buf = buf;
rcb->s = acb->s;
rcb->size = size;
r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
@@ -694,10 +716,18 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
switch (cmd) {
case RBD_AIO_WRITE:
- r = rbd_aio_write(s->image, off, size, buf, c);
+#ifdef LIBRBD_SUPPORTS_IOVEC
+ r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
+#else
+ r = rbd_aio_write(s->image, off, size, rcb->buf, c);
+#endif
break;
case RBD_AIO_READ:
- r = rbd_aio_read(s->image, off, size, buf, c);
+#ifdef LIBRBD_SUPPORTS_IOVEC
+ r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
+#else
+ r = rbd_aio_read(s->image, off, size, rcb->buf, c);
+#endif
break;
case RBD_AIO_DISCARD:
r = rbd_aio_discard_wrapper(s->image, off, size, c);
@@ -712,14 +742,16 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
if (r < 0) {
goto failed_completion;
}
-
return &acb->common;
failed_completion:
rbd_aio_release(c);
failed:
g_free(rcb);
- qemu_vfree(acb->bounce);
+ if (!LIBRBD_USE_IOVEC) {
+ qemu_vfree(acb->bounce);
+ }
+
qemu_aio_unref(acb);
return NULL;
}
--
2.10.2
^ permalink raw reply related [flat|nested] 12+ messages in thread
* Re: [RFC v6] RBD: Add support readv,writev for rbd
2017-02-21 6:50 ` [Qemu-devel] " jazeltq
@ 2017-02-24 3:52 ` Jeff Cody
-1 siblings, 0 replies; 12+ messages in thread
From: Jeff Cody @ 2017-02-24 3:52 UTC (permalink / raw)
To: jazeltq
Cc: jdurgin, dillaman, kwolf, mreitz, qemu-block, qemu-devel,
ceph-devel, tianqing
On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
> From: tianqing <tianqing@unitedstack.com>
>
> Rbd can do readv and writev directly, so wo do not need to transform
> iov to buf or vice versa any more.
>
> Signed-off-by: tianqing <tianqing@unitedstack.com>
> ---
This is marked as an RFC still - is this a series you would like to see in
2.9?
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Qemu-devel] [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-24 3:52 ` Jeff Cody
0 siblings, 0 replies; 12+ messages in thread
From: Jeff Cody @ 2017-02-24 3:52 UTC (permalink / raw)
To: jazeltq
Cc: jdurgin, dillaman, kwolf, mreitz, qemu-block, qemu-devel,
ceph-devel, tianqing
On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
> From: tianqing <tianqing@unitedstack.com>
>
> Rbd can do readv and writev directly, so wo do not need to transform
> iov to buf or vice versa any more.
>
> Signed-off-by: tianqing <tianqing@unitedstack.com>
> ---
This is marked as an RFC still - is this a series you would like to see in
2.9?
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [RFC v6] RBD: Add support readv,writev for rbd
2017-02-24 3:52 ` [Qemu-devel] " Jeff Cody
@ 2017-02-24 6:09 ` Jaze Lee
-1 siblings, 0 replies; 12+ messages in thread
From: Jaze Lee @ 2017-02-24 6:09 UTC (permalink / raw)
To: Jeff Cody
Cc: Josh Durgin, Jason Dillaman, Kevin Wolf, Max Reitz, qemu-block,
qemu-devel, ceph-devel, tianqing
2017-02-24 11:52 GMT+08:00 Jeff Cody <jcody@redhat.com>:
> On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
>> From: tianqing <tianqing@unitedstack.com>
>>
>> Rbd can do readv and writev directly, so wo do not need to transform
>> iov to buf or vice versa any more.
>>
>> Signed-off-by: tianqing <tianqing@unitedstack.com>
>> ---
>
>
> This is marked as an RFC still - is this a series you would like to see in
> 2.9?
Yes. What should i do if i like it in 2.9
--
谦谦君子
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Qemu-devel] [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-24 6:09 ` Jaze Lee
0 siblings, 0 replies; 12+ messages in thread
From: Jaze Lee @ 2017-02-24 6:09 UTC (permalink / raw)
To: Jeff Cody
Cc: Josh Durgin, Jason Dillaman, Kevin Wolf, Max Reitz, qemu-block,
qemu-devel, ceph-devel, tianqing
2017-02-24 11:52 GMT+08:00 Jeff Cody <jcody@redhat.com>:
> On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
>> From: tianqing <tianqing@unitedstack.com>
>>
>> Rbd can do readv and writev directly, so wo do not need to transform
>> iov to buf or vice versa any more.
>>
>> Signed-off-by: tianqing <tianqing@unitedstack.com>
>> ---
>
>
> This is marked as an RFC still - is this a series you would like to see in
> 2.9?
Yes. What should i do if i like it in 2.9
--
谦谦君子
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [RFC v6] RBD: Add support readv,writev for rbd
2017-02-24 6:09 ` [Qemu-devel] " Jaze Lee
@ 2017-02-24 11:42 ` Jeff Cody
-1 siblings, 0 replies; 12+ messages in thread
From: Jeff Cody @ 2017-02-24 11:42 UTC (permalink / raw)
To: Jaze Lee
Cc: Josh Durgin, Jason Dillaman, Kevin Wolf, Max Reitz, qemu-block,
qemu-devel, ceph-devel, tianqing
On Fri, Feb 24, 2017 at 02:09:31PM +0800, Jaze Lee wrote:
> 2017-02-24 11:52 GMT+08:00 Jeff Cody <jcody@redhat.com>:
> > On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
> >> From: tianqing <tianqing@unitedstack.com>
> >>
> >> Rbd can do readv and writev directly, so wo do not need to transform
> >> iov to buf or vice versa any more.
> >>
> >> Signed-off-by: tianqing <tianqing@unitedstack.com>
> >> ---
> >
> >
> > This is marked as an RFC still - is this a series you would like to see in
> > 2.9?
>
> Yes. What should i do if i like it in 2.9
>
Ideally you would submit the series as a non-RFC patch (that is, it would
have progressed to a normal patch away from RFC).
But in this case, it seems to me that this patch has progressed beyond RFC;
is there any reason it is still marked as RFC instead of just a patch? It
looks OK to me.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Qemu-devel] [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-24 11:42 ` Jeff Cody
0 siblings, 0 replies; 12+ messages in thread
From: Jeff Cody @ 2017-02-24 11:42 UTC (permalink / raw)
To: Jaze Lee
Cc: Josh Durgin, Jason Dillaman, Kevin Wolf, Max Reitz, qemu-block,
qemu-devel, ceph-devel, tianqing
On Fri, Feb 24, 2017 at 02:09:31PM +0800, Jaze Lee wrote:
> 2017-02-24 11:52 GMT+08:00 Jeff Cody <jcody@redhat.com>:
> > On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
> >> From: tianqing <tianqing@unitedstack.com>
> >>
> >> Rbd can do readv and writev directly, so wo do not need to transform
> >> iov to buf or vice versa any more.
> >>
> >> Signed-off-by: tianqing <tianqing@unitedstack.com>
> >> ---
> >
> >
> > This is marked as an RFC still - is this a series you would like to see in
> > 2.9?
>
> Yes. What should i do if i like it in 2.9
>
Ideally you would submit the series as a non-RFC patch (that is, it would
have progressed to a normal patch away from RFC).
But in this case, it seems to me that this patch has progressed beyond RFC;
is there any reason it is still marked as RFC instead of just a patch? It
looks OK to me.
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [RFC v6] RBD: Add support readv,writev for rbd
2017-02-24 11:42 ` [Qemu-devel] " Jeff Cody
@ 2017-02-24 12:51 ` Jaze Lee
-1 siblings, 0 replies; 12+ messages in thread
From: Jaze Lee @ 2017-02-24 12:51 UTC (permalink / raw)
To: Jeff Cody
Cc: Josh Durgin, Jason Dillaman, Kevin Wolf, Max Reitz, qemu-block,
qemu-devel, ceph-devel, tianqing
Sorry, I do not know. I thought we should always add RFC all though
the patch's life...
What is the prefix i should add? PATH 0 ?
2017-02-24 19:42 GMT+08:00 Jeff Cody <jcody@redhat.com>:
> On Fri, Feb 24, 2017 at 02:09:31PM +0800, Jaze Lee wrote:
>> 2017-02-24 11:52 GMT+08:00 Jeff Cody <jcody@redhat.com>:
>> > On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
>> >> From: tianqing <tianqing@unitedstack.com>
>> >>
>> >> Rbd can do readv and writev directly, so wo do not need to transform
>> >> iov to buf or vice versa any more.
>> >>
>> >> Signed-off-by: tianqing <tianqing@unitedstack.com>
>> >> ---
>> >
>> >
>> > This is marked as an RFC still - is this a series you would like to see in
>> > 2.9?
>>
>> Yes. What should i do if i like it in 2.9
>>
>
> Ideally you would submit the series as a non-RFC patch (that is, it would
> have progressed to a normal patch away from RFC).
>
> But in this case, it seems to me that this patch has progressed beyond RFC;
> is there any reason it is still marked as RFC instead of just a patch? It
> looks OK to me.
>
--
谦谦君子
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [Qemu-devel] [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-24 12:51 ` Jaze Lee
0 siblings, 0 replies; 12+ messages in thread
From: Jaze Lee @ 2017-02-24 12:51 UTC (permalink / raw)
To: Jeff Cody
Cc: Josh Durgin, Jason Dillaman, Kevin Wolf, Max Reitz, qemu-block,
qemu-devel, ceph-devel, tianqing
Sorry, I do not know. I thought we should always add RFC all though
the patch's life...
What is the prefix i should add? PATH 0 ?
2017-02-24 19:42 GMT+08:00 Jeff Cody <jcody@redhat.com>:
> On Fri, Feb 24, 2017 at 02:09:31PM +0800, Jaze Lee wrote:
>> 2017-02-24 11:52 GMT+08:00 Jeff Cody <jcody@redhat.com>:
>> > On Tue, Feb 21, 2017 at 02:50:03PM +0800, jazeltq@gmail.com wrote:
>> >> From: tianqing <tianqing@unitedstack.com>
>> >>
>> >> Rbd can do readv and writev directly, so wo do not need to transform
>> >> iov to buf or vice versa any more.
>> >>
>> >> Signed-off-by: tianqing <tianqing@unitedstack.com>
>> >> ---
>> >
>> >
>> > This is marked as an RFC still - is this a series you would like to see in
>> > 2.9?
>>
>> Yes. What should i do if i like it in 2.9
>>
>
> Ideally you would submit the series as a non-RFC patch (that is, it would
> have progressed to a normal patch away from RFC).
>
> But in this case, it seems to me that this patch has progressed beyond RFC;
> is there any reason it is still marked as RFC instead of just a patch? It
> looks OK to me.
>
--
谦谦君子
^ permalink raw reply [flat|nested] 12+ messages in thread
* Re: [RFC v6] RBD: Add support readv,writev for rbd
2017-02-21 3:43 jazeltq
@ 2017-02-21 4:50 ` Jeff Cody
0 siblings, 0 replies; 12+ messages in thread
From: Jeff Cody @ 2017-02-21 4:50 UTC (permalink / raw)
To: jazeltq
Cc: kwolf, qemu-block, qemu-devel, mreitz, tianqing, ceph-devel, dillaman
On Tue, Feb 21, 2017 at 11:43:36AM +0800, jazeltq@gmail.com wrote:
> From: tianqing <tianqing@unitedstack.com>
>
> Rbd can do readv and writev directly, so wo do not need to transform
> iov to buf or vice versa any more.
>
> Signed-off-by: tianqing <tianqing@unitedstack.com>
> ---
> block/rbd.c | 79 ++++++++++++++++++++++++++++++++++++++++++-------------------
> 1 file changed, 54 insertions(+), 25 deletions(-)
>
[...]
> case RBD_AIO_WRITE:
> - r = rbd_aio_write(s->image, off, size, buf, c);
> + if(!LIBRBD_USE_IOVEC)
> + r = rbd_aio_write(s->image, off, size, rcb->buf, c);
> + else
> + r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
> break;
> case RBD_AIO_READ:
> - r = rbd_aio_read(s->image, off, size, buf, c);
> + if(!LIBRBD_USE_IOVEC)
> + r = rbd_aio_read(s->image, off, size, rcb->buf, c);
> + else
> + r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
These will need to stay #ifdef's since they are new symbols.
> break;
> case RBD_AIO_DISCARD:
> r = rbd_aio_discard_wrapper(s->image, off, size, c);
> @@ -712,14 +740,15 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
> if (r < 0) {
> goto failed_completion;
> }
> -
> return &acb->common;
>
> failed_completion:
> rbd_aio_release(c);
> failed:
> g_free(rcb);
> - qemu_vfree(acb->bounce);
> + if(!LIBRBD_USE_IOVEC)
> + qemu_vfree(acb->bounce);
> +
> qemu_aio_unref(acb);
> return NULL;
> }
> --
> 2.10.2
>
(Also code formatting as pointed out by patchew)
^ permalink raw reply [flat|nested] 12+ messages in thread
* [RFC v6] RBD: Add support readv,writev for rbd
@ 2017-02-21 3:43 jazeltq
2017-02-21 4:50 ` Jeff Cody
0 siblings, 1 reply; 12+ messages in thread
From: jazeltq @ 2017-02-21 3:43 UTC (permalink / raw)
To: jdurgin, jcody, dillaman, kwolf, mreitz, qemu-block
Cc: qemu-devel, ceph-devel, jazeltq, tianqing
From: tianqing <tianqing@unitedstack.com>
Rbd can do readv and writev directly, so wo do not need to transform
iov to buf or vice versa any more.
Signed-off-by: tianqing <tianqing@unitedstack.com>
---
block/rbd.c | 79 ++++++++++++++++++++++++++++++++++++++++++-------------------
1 file changed, 54 insertions(+), 25 deletions(-)
diff --git a/block/rbd.c b/block/rbd.c
index a57b3e3..5373680 100644
--- a/block/rbd.c
+++ b/block/rbd.c
@@ -62,6 +62,13 @@
#define RBD_MAX_SNAP_NAME_SIZE 128
#define RBD_MAX_SNAPS 100
+/* The LIBRBD_SUPPORTS_IOVEC is defined in librbd.h */
+#ifdef LIBRBD_SUPPORTS_IOVEC
+#define LIBRBD_USE_IOVEC 1
+#else
+#define LIBRBD_USE_IOVEC 0
+#endif
+
typedef enum {
RBD_AIO_READ,
RBD_AIO_WRITE,
@@ -310,6 +317,17 @@ static int qemu_rbd_set_conf(rados_t cluster, const char *conf,
return ret;
}
+static void qemu_rbd_memset(RADOSCB *rcb, int64_t offs)
+{
+ if(LIBRBD_USE_IOVEC){
+ RBDAIOCB *acb = rcb->acb;
+ iov_memset(acb->qiov->iov, acb->qiov->niov, offs, 0,
+ acb->qiov->size - offs);
+ } else {
+ memset(rcb->buf + offs, 0, rcb->size - offs);
+ }
+}
+
static int qemu_rbd_create(const char *filename, QemuOpts *opts, Error **errp)
{
Error *local_err = NULL;
@@ -426,11 +444,11 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
}
} else {
if (r < 0) {
- memset(rcb->buf, 0, rcb->size);
+ qemu_rbd_memset(rcb, 0);
acb->ret = r;
acb->error = 1;
} else if (r < rcb->size) {
- memset(rcb->buf + r, 0, rcb->size - r);
+ qemu_rbd_memset(rcb, r);
if (!acb->error) {
acb->ret = rcb->size;
}
@@ -440,11 +458,14 @@ static void qemu_rbd_complete_aio(RADOSCB *rcb)
}
g_free(rcb);
-
- if (acb->cmd == RBD_AIO_READ) {
- qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
+
+ if(!LIBRBD_USE_IOVEC){
+ if (acb->cmd == RBD_AIO_READ) {
+ qemu_iovec_from_buf(acb->qiov, 0, acb->bounce, acb->qiov->size);
+ }
+ qemu_vfree(acb->bounce);
}
- qemu_vfree(acb->bounce);
+
acb->common.cb(acb->common.opaque, (acb->ret > 0 ? 0 : acb->ret));
qemu_aio_unref(acb);
@@ -655,7 +676,6 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
RBDAIOCB *acb;
RADOSCB *rcb = NULL;
rbd_completion_t c;
- char *buf;
int r;
BDRVRBDState *s = bs->opaque;
@@ -664,27 +684,29 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
acb->cmd = cmd;
acb->qiov = qiov;
assert(!qiov || qiov->size == size);
- if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
- acb->bounce = NULL;
- } else {
- acb->bounce = qemu_try_blockalign(bs, qiov->size);
- if (acb->bounce == NULL) {
- goto failed;
+
+ rcb = g_new(RADOSCB, 1);
+
+ if(!LIBRBD_USE_IOVEC){
+ if (cmd == RBD_AIO_DISCARD || cmd == RBD_AIO_FLUSH) {
+ acb->bounce = NULL;
+ } else {
+ acb->bounce = qemu_try_blockalign(bs, qiov->size);
+ if (acb->bounce == NULL) {
+ goto failed;
+ }
}
+ if (cmd == RBD_AIO_WRITE) {
+ qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
+ }
+ rcb->buf = acb->bounce;
}
+
acb->ret = 0;
acb->error = 0;
acb->s = s;
- if (cmd == RBD_AIO_WRITE) {
- qemu_iovec_to_buf(acb->qiov, 0, acb->bounce, qiov->size);
- }
-
- buf = acb->bounce;
-
- rcb = g_new(RADOSCB, 1);
rcb->acb = acb;
- rcb->buf = buf;
rcb->s = acb->s;
rcb->size = size;
r = rbd_aio_create_completion(rcb, (rbd_callback_t) rbd_finish_aiocb, &c);
@@ -694,10 +716,16 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
switch (cmd) {
case RBD_AIO_WRITE:
- r = rbd_aio_write(s->image, off, size, buf, c);
+ if(!LIBRBD_USE_IOVEC)
+ r = rbd_aio_write(s->image, off, size, rcb->buf, c);
+ else
+ r = rbd_aio_writev(s->image, qiov->iov, qiov->niov, off, c);
break;
case RBD_AIO_READ:
- r = rbd_aio_read(s->image, off, size, buf, c);
+ if(!LIBRBD_USE_IOVEC)
+ r = rbd_aio_read(s->image, off, size, rcb->buf, c);
+ else
+ r = rbd_aio_readv(s->image, qiov->iov, qiov->niov, off, c);
break;
case RBD_AIO_DISCARD:
r = rbd_aio_discard_wrapper(s->image, off, size, c);
@@ -712,14 +740,15 @@ static BlockAIOCB *rbd_start_aio(BlockDriverState *bs,
if (r < 0) {
goto failed_completion;
}
-
return &acb->common;
failed_completion:
rbd_aio_release(c);
failed:
g_free(rcb);
- qemu_vfree(acb->bounce);
+ if(!LIBRBD_USE_IOVEC)
+ qemu_vfree(acb->bounce);
+
qemu_aio_unref(acb);
return NULL;
}
--
2.10.2
^ permalink raw reply related [flat|nested] 12+ messages in thread
end of thread, other threads:[~2017-02-24 12:51 UTC | newest]
Thread overview: 12+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-02-21 6:50 [RFC v6] RBD: Add support readv,writev for rbd jazeltq
2017-02-21 6:50 ` [Qemu-devel] " jazeltq
2017-02-24 3:52 ` Jeff Cody
2017-02-24 3:52 ` [Qemu-devel] " Jeff Cody
2017-02-24 6:09 ` Jaze Lee
2017-02-24 6:09 ` [Qemu-devel] " Jaze Lee
2017-02-24 11:42 ` Jeff Cody
2017-02-24 11:42 ` [Qemu-devel] " Jeff Cody
2017-02-24 12:51 ` Jaze Lee
2017-02-24 12:51 ` [Qemu-devel] " Jaze Lee
-- strict thread matches above, loose matches on Subject: below --
2017-02-21 3:43 jazeltq
2017-02-21 4:50 ` Jeff Cody
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.