[Qemu-devel] [PATCH] block: Simplify bdrv_can_write_zeroes_with_unmap()

Eric Blake posted 1 patch 6 years, 3 months ago
Patches applied successfully (tree, apply log)
git fetch https://github.com/patchew-project/qemu tags/patchew/20180126193439.20219-1-eblake@redhat.com
Test checkpatch passed
Test docker-build@min-glib passed
Test docker-mingw@fedora passed
Test docker-quick@centos6 passed
Test ppc passed
Test s390x passed
include/block/block.h | 11 -----------
block.c               |  8 +-------
block/crypto.c        |  1 -
block/file-posix.c    |  3 +--
block/iscsi.c         |  6 ++++--
block/nbd.c           | 11 -----------
block/qcow2.c         |  3 +--
block/qed.c           |  1 -
8 files changed, 7 insertions(+), 37 deletions(-)
[Qemu-devel] [PATCH] block: Simplify bdrv_can_write_zeroes_with_unmap()
Posted by Eric Blake 6 years, 3 months ago
We don't need the can_write_zeroes_with_unmap field in
BlockDriverInfo, because it is redundant information with
supported_zero_flags & BDRV_REQ_MAY_UNMAP.  Note that
BlockDriverInfo and supported_zero_flags are both per-device
settings, rather than global state about the driver as a
whole, which means one or both of these bits of information
can already be conditional.  Let's audit how they were set:

crypto: always setting can_write_ to false is pointless (the
struct starts life zero-initialized), no use of supported_

nbd: just recently fixed to set can_write_ if supported_
includes MAY_UNMAP (thus this commit effectively reverts
bca80059e and solves the problem mentioned there in a more
global way)

file-posix, iscsi, qcow2: can_write_ is conditional, while
supported_ was unconditional; but passing MAY_UNMAP would
fail with ENOTSUP if the condition wasn't met

qed: can_write_ is unconditional, but pwrite_zeroes lacks
support for MAY_UNMAP and supported_ is not set. Perhaps
support can be added later (since it would be similar to
qcow2), but for now claiming false is no real loss

all other drivers: can_write_ is not set, and supported_ is
either unset or a passthrough

Simplify the code by moving the conditional into
supported_zero_flags for all drivers, then dropping the
now-unused BDI field.  For callers that relied on
bdrv_can_write_zeroes_with_unmap(), we return the same
per-device settings for drivers that had conditions (no
observable change in behavior there); and can now return
true (instead of false) for drivers that support passthrough
(for example, the commit driver) which gives those drivers
the same fix as nbd just got in bca80059e.  For callers that
relied on supported_zero_flags, we now have a few more places
that can avoid a wasted call to pwrite_zeroes() that will
just fail with ENOTSUP.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>

---
The commit id mentioned above is dependent on me not having to respin
my latest NBD pull request:

Based-on: <20180126160411.4033-1-eblake@redhat.com>
([PULL 0/8] NBD patches through 26 Jan)
---
 include/block/block.h | 11 -----------
 block.c               |  8 +-------
 block/crypto.c        |  1 -
 block/file-posix.c    |  3 +--
 block/iscsi.c         |  6 ++++--
 block/nbd.c           | 11 -----------
 block/qcow2.c         |  3 +--
 block/qed.c           |  1 -
 8 files changed, 7 insertions(+), 37 deletions(-)

diff --git a/include/block/block.h b/include/block/block.h
index 9b12774ddf2..d808849784c 100644
--- a/include/block/block.h
+++ b/include/block/block.h
@@ -28,17 +28,6 @@ typedef struct BlockDriverInfo {
      * to the LBPRZ flag in the SCSI logical block provisioning page.
      */
     bool unallocated_blocks_are_zero;
-    /*
-     * True if the driver can optimize writing zeroes by unmapping
-     * sectors. This is equivalent to the BLKDISCARDZEROES ioctl in Linux
-     * with the difference that in qemu a discard is allowed to silently
-     * fail. Therefore we have to use bdrv_pwrite_zeroes with the
-     * BDRV_REQ_MAY_UNMAP flag for an optimized zero write with unmapping.
-     * After this call the driver has to guarantee that the contents read
-     * back as zero. It is additionally required that the block device is
-     * opened with BDRV_O_UNMAP flag for this to work.
-     */
-    bool can_write_zeroes_with_unmap;
     /*
      * True if this block driver only supports compressed writes
      */
diff --git a/block.c b/block.c
index a8da4f2b255..ebbb907b580 100644
--- a/block.c
+++ b/block.c
@@ -4007,17 +4007,11 @@ bool bdrv_unallocated_blocks_are_zero(BlockDriverState *bs)

 bool bdrv_can_write_zeroes_with_unmap(BlockDriverState *bs)
 {
-    BlockDriverInfo bdi;
-
     if (!(bs->open_flags & BDRV_O_UNMAP)) {
         return false;
     }

-    if (bdrv_get_info(bs, &bdi) == 0) {
-        return bdi.can_write_zeroes_with_unmap;
-    }
-
-    return false;
+    return bs->supported_zero_flags & BDRV_REQ_MAY_UNMAP;
 }

 const char *bdrv_get_encrypted_filename(BlockDriverState *bs)
diff --git a/block/crypto.c b/block/crypto.c
index 60ddf8623ee..5d18f5dc98a 100644
--- a/block/crypto.c
+++ b/block/crypto.c
@@ -574,7 +574,6 @@ static int block_crypto_get_info_luks(BlockDriverState *bs,
     }

     bdi->unallocated_blocks_are_zero = false;
-    bdi->can_write_zeroes_with_unmap = false;
     bdi->cluster_size = subbdi.cluster_size;

     return 0;
diff --git a/block/file-posix.c b/block/file-posix.c
index 36ee89e9402..2ede2b0f392 100644
--- a/block/file-posix.c
+++ b/block/file-posix.c
@@ -546,7 +546,6 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,

     s->has_discard = true;
     s->has_write_zeroes = true;
-    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
     if ((bs->open_flags & BDRV_O_NOCACHE) != 0) {
         s->needs_alignment = true;
     }
@@ -596,6 +595,7 @@ static int raw_open_common(BlockDriverState *bs, QDict *options,
     }
 #endif

+    bs->supported_zero_flags = s->discard_zeroes ? BDRV_REQ_MAY_UNMAP : 0;
     ret = 0;
 fail:
     if (filename && (bdrv_flags & BDRV_O_TEMPORARY)) {
@@ -2220,7 +2220,6 @@ static int raw_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
     BDRVRawState *s = bs->opaque;

     bdi->unallocated_blocks_are_zero = s->discard_zeroes;
-    bdi->can_write_zeroes_with_unmap = s->discard_zeroes;
     return 0;
 }

diff --git a/block/iscsi.c b/block/iscsi.c
index 6a1c53711a7..673196b87e5 100644
--- a/block/iscsi.c
+++ b/block/iscsi.c
@@ -1875,7 +1875,6 @@ static int iscsi_open(BlockDriverState *bs, QDict *options, int flags,
     if (iscsilun->dpofua) {
         bs->supported_write_flags = BDRV_REQ_FUA;
     }
-    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;

     /* Check the write protect flag of the LUN if we want to write */
     if (iscsilun->type == TYPE_DISK && (flags & BDRV_O_RDWR) &&
@@ -1959,6 +1958,10 @@ static int iscsi_open(BlockDriverState *bs, QDict *options, int flags,
         }
     }

+    if (iscsilun->lbprz && iscsilun->lbp.lbpws) {
+        bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+    }
+
 out:
     qemu_opts_del(opts);
     g_free(initiator_name);
@@ -2158,7 +2161,6 @@ static int iscsi_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     IscsiLun *iscsilun = bs->opaque;
     bdi->unallocated_blocks_are_zero = iscsilun->lbprz;
-    bdi->can_write_zeroes_with_unmap = iscsilun->lbprz && iscsilun->lbp.lbpws;
     bdi->cluster_size = iscsilun->cluster_sectors * BDRV_SECTOR_SIZE;
     return 0;
 }
diff --git a/block/nbd.c b/block/nbd.c
index 94220f6d143..8b8ba56cdd0 100644
--- a/block/nbd.c
+++ b/block/nbd.c
@@ -566,14 +566,6 @@ static void nbd_refresh_filename(BlockDriverState *bs, QDict *options)
     bs->full_open_options = opts;
 }

-static int nbd_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
-{
-    if (bs->supported_zero_flags & BDRV_REQ_MAY_UNMAP) {
-        bdi->can_write_zeroes_with_unmap = true;
-    }
-    return 0;
-}
-
 static BlockDriver bdrv_nbd = {
     .format_name                = "nbd",
     .protocol_name              = "nbd",
@@ -591,7 +583,6 @@ static BlockDriver bdrv_nbd = {
     .bdrv_detach_aio_context    = nbd_detach_aio_context,
     .bdrv_attach_aio_context    = nbd_attach_aio_context,
     .bdrv_refresh_filename      = nbd_refresh_filename,
-    .bdrv_get_info              = nbd_get_info,
 };

 static BlockDriver bdrv_nbd_tcp = {
@@ -611,7 +602,6 @@ static BlockDriver bdrv_nbd_tcp = {
     .bdrv_detach_aio_context    = nbd_detach_aio_context,
     .bdrv_attach_aio_context    = nbd_attach_aio_context,
     .bdrv_refresh_filename      = nbd_refresh_filename,
-    .bdrv_get_info              = nbd_get_info,
 };

 static BlockDriver bdrv_nbd_unix = {
@@ -631,7 +621,6 @@ static BlockDriver bdrv_nbd_unix = {
     .bdrv_detach_aio_context    = nbd_detach_aio_context,
     .bdrv_attach_aio_context    = nbd_attach_aio_context,
     .bdrv_refresh_filename      = nbd_refresh_filename,
-    .bdrv_get_info              = nbd_get_info,
 };

 static void bdrv_nbd_init(void)
diff --git a/block/qcow2.c b/block/qcow2.c
index 1f80961e1b9..fe4b2135087 100644
--- a/block/qcow2.c
+++ b/block/qcow2.c
@@ -1477,7 +1477,7 @@ static int qcow2_do_open(BlockDriverState *bs, QDict *options, int flags,

     /* Initialise locks */
     qemu_co_mutex_init(&s->lock);
-    bs->supported_zero_flags = BDRV_REQ_MAY_UNMAP;
+    bs->supported_zero_flags = header.version >= 3 ? BDRV_REQ_MAY_UNMAP : 0;

     /* Repair image if dirty */
     if (!(flags & (BDRV_O_CHECK | BDRV_O_INACTIVE)) && !bs->read_only &&
@@ -3769,7 +3769,6 @@ static int qcow2_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
 {
     BDRVQcow2State *s = bs->opaque;
     bdi->unallocated_blocks_are_zero = true;
-    bdi->can_write_zeroes_with_unmap = (s->qcow_version >= 3);
     bdi->cluster_size = s->cluster_size;
     bdi->vm_state_offset = qcow2_vm_state_offset(s);
     return 0;
diff --git a/block/qed.c b/block/qed.c
index 821dcaa0559..61acb87f2ab 100644
--- a/block/qed.c
+++ b/block/qed.c
@@ -1438,7 +1438,6 @@ static int bdrv_qed_get_info(BlockDriverState *bs, BlockDriverInfo *bdi)
     bdi->cluster_size = s->header.cluster_size;
     bdi->is_dirty = s->header.features & QED_F_NEED_CHECK;
     bdi->unallocated_blocks_are_zero = true;
-    bdi->can_write_zeroes_with_unmap = true;
     return 0;
 }

-- 
2.14.3


Re: [Qemu-devel] [Qemu-block] [PATCH] block: Simplify bdrv_can_write_zeroes_with_unmap()
Posted by Stefan Hajnoczi 6 years, 3 months ago
On Fri, Jan 26, 2018 at 01:34:39PM -0600, Eric Blake wrote:
> We don't need the can_write_zeroes_with_unmap field in
> BlockDriverInfo, because it is redundant information with
> supported_zero_flags & BDRV_REQ_MAY_UNMAP.  Note that
> BlockDriverInfo and supported_zero_flags are both per-device
> settings, rather than global state about the driver as a
> whole, which means one or both of these bits of information
> can already be conditional.  Let's audit how they were set:
> 
> crypto: always setting can_write_ to false is pointless (the
> struct starts life zero-initialized), no use of supported_
> 
> nbd: just recently fixed to set can_write_ if supported_
> includes MAY_UNMAP (thus this commit effectively reverts
> bca80059e and solves the problem mentioned there in a more
> global way)
> 
> file-posix, iscsi, qcow2: can_write_ is conditional, while
> supported_ was unconditional; but passing MAY_UNMAP would
> fail with ENOTSUP if the condition wasn't met
> 
> qed: can_write_ is unconditional, but pwrite_zeroes lacks
> support for MAY_UNMAP and supported_ is not set. Perhaps
> support can be added later (since it would be similar to
> qcow2), but for now claiming false is no real loss
> 
> all other drivers: can_write_ is not set, and supported_ is
> either unset or a passthrough
> 
> Simplify the code by moving the conditional into
> supported_zero_flags for all drivers, then dropping the
> now-unused BDI field.  For callers that relied on
> bdrv_can_write_zeroes_with_unmap(), we return the same
> per-device settings for drivers that had conditions (no
> observable change in behavior there); and can now return
> true (instead of false) for drivers that support passthrough
> (for example, the commit driver) which gives those drivers
> the same fix as nbd just got in bca80059e.  For callers that
> relied on supported_zero_flags, we now have a few more places
> that can avoid a wasted call to pwrite_zeroes() that will
> just fail with ENOTSUP.
> 
> Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
> Signed-off-by: Eric Blake <eblake@redhat.com>
> 
> ---
> The commit id mentioned above is dependent on me not having to respin
> my latest NBD pull request:
> 
> Based-on: <20180126160411.4033-1-eblake@redhat.com>
> ([PULL 0/8] NBD patches through 26 Jan)
> ---
>  include/block/block.h | 11 -----------
>  block.c               |  8 +-------
>  block/crypto.c        |  1 -
>  block/file-posix.c    |  3 +--
>  block/iscsi.c         |  6 ++++--
>  block/nbd.c           | 11 -----------
>  block/qcow2.c         |  3 +--
>  block/qed.c           |  1 -
>  8 files changed, 7 insertions(+), 37 deletions(-)

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Re: [Qemu-devel] [Qemu-block] [PATCH] block: Simplify bdrv_can_write_zeroes_with_unmap()
Posted by Eric Blake 6 years, 3 months ago
On 01/29/2018 05:08 AM, Stefan Hajnoczi wrote:
> On Fri, Jan 26, 2018 at 01:34:39PM -0600, Eric Blake wrote:
>> We don't need the can_write_zeroes_with_unmap field in
>> BlockDriverInfo, because it is redundant information with
>> supported_zero_flags & BDRV_REQ_MAY_UNMAP.  Note that
>> BlockDriverInfo and supported_zero_flags are both per-device
>> settings, rather than global state about the driver as a
>> whole, which means one or both of these bits of information
>> can already be conditional.  Let's audit how they were set:
>>
...
>> Simplify the code by moving the conditional into
>> supported_zero_flags for all drivers, then dropping the
>> now-unused BDI field.  For callers that relied on
>> bdrv_can_write_zeroes_with_unmap(), we return the same
>> per-device settings for drivers that had conditions (no
>> observable change in behavior there); and can now return
>> true (instead of false) for drivers that support passthrough
>> (for example, the commit driver) which gives those drivers
>> the same fix as nbd just got in bca80059e.  For callers that
>> relied on supported_zero_flags, we now have a few more places
>> that can avoid a wasted call to pwrite_zeroes() that will
>> just fail with ENOTSUP.
>>
>> Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
>> Signed-off-by: Eric Blake <eblake@redhat.com>
>>
>> ---
>> The commit id mentioned above is dependent on me not having to respin
>> my latest NBD pull request:
>>
>> Based-on: <20180126160411.4033-1-eblake@redhat.com>
>> ([PULL 0/8] NBD patches through 26 Jan)

> Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
> 

Thanks.

Since this patch was discovered in relation to NBD code, I'm fine taking 
it through my NBD queue; but it's also more generic to the block layer, 
so I'm also fine if one of the other block maintainers grabs it first 
through their tree.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org