[Qemu-devel] [PATCH] rbd: shift byte count as a 64-bit value

[Qemu-devel] [PATCH] rbd: shift byte count as a 64-bit value

[Paolo Bonzini]

Otherwise, reads of more than 2GB fail.  Until commit
7bbca9e290a9c7c217b5a24fc6094e91e54bd05d, reads of 2^41
bytes succeeded at least theoretically.

In fact, pdiscard ought to receive a 64-bit integer as the
count for the same reason.

Reported by Coverity.

Fixes: 7bbca9e290a9c7c217b5a24fc6094e91e54bd05d
Cc: qemu-stable@nongnu.org
Cc: kwolf@redhat.com
Cc: eblake@redhat.com
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 block/rbd.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/block/rbd.c b/block/rbd.c
index 6f9eb6f..f6e1d4b 100644
--- a/block/rbd.c
+++ b/block/rbd.c
@@ -733,7 +733,7 @@ static BlockAIOCB *qemu_rbd_aio_readv(BlockDriverState *bs,
                                       void *opaque)
 {
     return rbd_start_aio(bs, sector_num << BDRV_SECTOR_BITS, qiov,
-                         nb_sectors << BDRV_SECTOR_BITS, cb, opaque,
+                         (int64_t) nb_sectors << BDRV_SECTOR_BITS, cb, opaque,
                          RBD_AIO_READ);
 }
 
@@ -745,7 +745,7 @@ static BlockAIOCB *qemu_rbd_aio_writev(BlockDriverState *bs,
                                        void *opaque)
 {
     return rbd_start_aio(bs, sector_num << BDRV_SECTOR_BITS, qiov,
-                         nb_sectors << BDRV_SECTOR_BITS, cb, opaque,
+                         (int64_t) nb_sectors << BDRV_SECTOR_BITS, cb, opaque,
                          RBD_AIO_WRITE);
 }
 
-- 
2.7.4

Re: [Qemu-devel] [PATCH] rbd: shift byte count as a 64-bit value

[Eric Blake]

[-- Attachment #1: Type: text/plain, Size: 1271 bytes --]

On 10/10/2016 03:00 PM, Paolo Bonzini wrote:
> Otherwise, reads of more than 2GB fail.  Until commit
> 7bbca9e290a9c7c217b5a24fc6094e91e54bd05d, reads of 2^41
> bytes succeeded at least theoretically.

I don't think the block layer ever passes a read request down to drivers
that large in the first place; we have a lot of code in io.c that
clamps/asserts things to be smaller than 2^31 (ie.
BDRV_REQUEST_MAX_SECTORS).  So this may just be cleaning up something
that would become latent with a future block change to support larger
sizes, rather than an actual bug fix today.  But it is worth making.

> In fact, pdiscard ought to receive a 64-bit integer as the
> count for the same reason.

Yes, auditing ALL of the block code for eliminating arbitrary 32-bit
limits would be a big project.

> 
> Reported by Coverity.
> 
> Fixes: 7bbca9e290a9c7c217b5a24fc6094e91e54bd05d
> Cc: qemu-stable@nongnu.org
> Cc: kwolf@redhat.com
> Cc: eblake@redhat.com
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  block/rbd.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

Reviewed-by: Eric Blake <eblake@redhat.com>

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 604 bytes --]