[v3,1/1] block/vmdk: Fix the endian problem of buf_len and lba

Commit Message

Hao QingFeng Dec. 16, 2016, 5:20 a.m. UTC

The problem was triggered by qemu-iotests case 055. It failed when it
was comparing the compressed vmdk image with original test.img.

The cause is that buf_len in vmdk_write_extent wasn't converted to
little-endian before it was stored to disk. But later vmdk_read_extent
read it and converted it from little-endian to cpu endian.
If the cpu is big-endian like s390, the problem will happen and
the data length read by vmdk_read_extent will become invalid!
The fix is to add the conversion in vmdk_write_extent, meanwhile,
repair the endianness problem of lba field which shall also be converted
to little-endian before storing to disk.

Cc: [e-mail]
Signed-off-by: QingFeng Hao <[e-mail]>
Signed-off-by: Jing Liu <[e-mail]>
Signed-off-by: Kevin Wolf <[e-mail]>
Reviewed-by: Fam Zheng <[e-mail]>
---
 block/vmdk.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Jing Liu Jan. 16, 2017, 2:38 a.m. UTC | #1

Dears,

We would like to know if this patch will be pulled

into upstream or what else we need to do for it?

Because for upstream, the qemu-iotests case 055 still failed.

Thanks.

Jing


On 12/16/2016 01:20 PM, QingFeng Hao wrote:
> The problem was triggered by qemu-iotests case 055. It failed when it
> was comparing the compressed vmdk image with original test.img.
>
> The cause is that buf_len in vmdk_write_extent wasn't converted to
> little-endian before it was stored to disk. But later vmdk_read_extent
> read it and converted it from little-endian to cpu endian.
> If the cpu is big-endian like s390, the problem will happen and
> the data length read by vmdk_read_extent will become invalid!
> The fix is to add the conversion in vmdk_write_extent, meanwhile,
> repair the endianness problem of lba field which shall also be converted
> to little-endian before storing to disk.
>
> Cc: [e-mail]
> Signed-off-by: QingFeng Hao <[e-mail]>
> Signed-off-by: Jing Liu <[e-mail]>
> Signed-off-by: Kevin Wolf <[e-mail]>
> Reviewed-by: Fam Zheng <[e-mail]>
> ---
>   block/vmdk.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/block/vmdk.c b/block/vmdk.c
> index a11c27a..26e5f95 100644
> --- a/block/vmdk.c
> +++ b/block/vmdk.c
> @@ -1354,8 +1354,8 @@ static int vmdk_write_extent(VmdkExtent *extent, int64_t cluster_offset,
>               goto out;
>           }
>
> -        data->lba = offset >> BDRV_SECTOR_BITS;
> -        data->size = buf_len;
> +        data->lba = cpu_to_le64(offset >> BDRV_SECTOR_BITS);
> +        data->size = cpu_to_le32(buf_len);
>
>           n_bytes = buf_len + sizeof(VmdkGrainMarker);
>           iov = (struct iovec) {

Fam Zheng Jan. 16, 2017, 5:26 a.m. UTC | #2

On Mon, 01/16 10:38, liujing wrote:
> Dears,
> 
> We would like to know if this patch will be pulled
> 
> into upstream or what else we need to do for it?
> 
> Because for upstream, the qemu-iotests case 055 still failed.

Kevin, would you like to take this or should I do it?

Fam

> 
> Thanks.
> 
> Jing
> 
> 
> On 12/16/2016 01:20 PM, QingFeng Hao wrote:
> > The problem was triggered by qemu-iotests case 055. It failed when it
> > was comparing the compressed vmdk image with original test.img.
> > 
> > The cause is that buf_len in vmdk_write_extent wasn't converted to
> > little-endian before it was stored to disk. But later vmdk_read_extent
> > read it and converted it from little-endian to cpu endian.
> > If the cpu is big-endian like s390, the problem will happen and
> > the data length read by vmdk_read_extent will become invalid!
> > The fix is to add the conversion in vmdk_write_extent, meanwhile,
> > repair the endianness problem of lba field which shall also be converted
> > to little-endian before storing to disk.
> > 
> > Cc: [e-mail]
> > Signed-off-by: QingFeng Hao <[e-mail]>
> > Signed-off-by: Jing Liu <[e-mail]>
> > Signed-off-by: Kevin Wolf <[e-mail]>
> > Reviewed-by: Fam Zheng <[e-mail]>
> > ---
> >   block/vmdk.c | 4 ++--
> >   1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/block/vmdk.c b/block/vmdk.c
> > index a11c27a..26e5f95 100644
> > --- a/block/vmdk.c
> > +++ b/block/vmdk.c
> > @@ -1354,8 +1354,8 @@ static int vmdk_write_extent(VmdkExtent *extent, int64_t cluster_offset,
> >               goto out;
> >           }
> > 
> > -        data->lba = offset >> BDRV_SECTOR_BITS;
> > -        data->size = buf_len;
> > +        data->lba = cpu_to_le64(offset >> BDRV_SECTOR_BITS);
> > +        data->size = cpu_to_le32(buf_len);
> > 
> >           n_bytes = buf_len + sizeof(VmdkGrainMarker);
> >           iov = (struct iovec) {
> 
>

Fam Zheng Jan. 23, 2017, 10:51 a.m. UTC | #3

On Mon, 01/16 13:26, Fam Zheng wrote:
> On Mon, 01/16 10:38, liujing wrote:
> > Dears,
> > 
> > We would like to know if this patch will be pulled
> > 
> > into upstream or what else we need to do for it?
> > 
> > Because for upstream, the qemu-iotests case 055 still failed.
> 
> Kevin, would you like to take this or should I do it?

Adding Max since he's taking care of the other VMDK fix I posted.

Fam

Max Reitz Jan. 25, 2017, 4:25 p.m. UTC | #4

On 16.12.2016 06:20, QingFeng Hao wrote:
> The problem was triggered by qemu-iotests case 055. It failed when it
> was comparing the compressed vmdk image with original test.img.
>
> The cause is that buf_len in vmdk_write_extent wasn't converted to
> little-endian before it was stored to disk. But later vmdk_read_extent
> read it and converted it from little-endian to cpu endian.
> If the cpu is big-endian like s390, the problem will happen and
> the data length read by vmdk_read_extent will become invalid!
> The fix is to add the conversion in vmdk_write_extent, meanwhile,
> repair the endianness problem of lba field which shall also be converted
> to little-endian before storing to disk.
>
> Cc: [e-mail]
> Signed-off-by: QingFeng Hao <[e-mail]>
> Signed-off-by: Jing Liu <[e-mail]>
> Signed-off-by: Kevin Wolf <[e-mail]>
> Reviewed-by: Fam Zheng <[e-mail]>
> ---
>  block/vmdk.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)

Thanks, I've applied the patch to my block tree:

https://github.com/XanClic/qemu/commits/block

Max

Hao QingFeng Feb. 13, 2017, 9:17 a.m. UTC | #5

在 2017-01-26 0:25, Max Reitz 写道:
> On 16.12.2016 06:20, QingFeng Hao wrote:
>> The problem was triggered by qemu-iotests case 055. It failed when it
>> was comparing the compressed vmdk image with original test.img.
>>
>> The cause is that buf_len in vmdk_write_extent wasn't converted to
>> little-endian before it was stored to disk. But later vmdk_read_extent
>> read it and converted it from little-endian to cpu endian.
>> If the cpu is big-endian like s390, the problem will happen and
>> the data length read by vmdk_read_extent will become invalid!
>> The fix is to add the conversion in vmdk_write_extent, meanwhile,
>> repair the endianness problem of lba field which shall also be converted
>> to little-endian before storing to disk.
>>
>> Cc: [e-mail]
>> Signed-off-by: QingFeng Hao <[e-mail]>
>> Signed-off-by: Jing Liu <[e-mail]>
>> Signed-off-by: Kevin Wolf <[e-mail]>
>> Reviewed-by: Fam Zheng <[e-mail]>
>> ---
>>   block/vmdk.c | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
> Thanks, I've applied the patch to my block tree:
>
> https://github.com/XanClic/qemu/commits/block
Thanks Max!
>
> Max
>

Patch

diff --git a/block/vmdk.c b/block/vmdk.c
index a11c27a..26e5f95 100644
--- a/block/vmdk.c
+++ b/block/vmdk.c
@@ -1354,8 +1354,8 @@  static int vmdk_write_extent(VmdkExtent *extent, int64_t cluster_offset,
             goto out;
         }
 
-        data->lba = offset >> BDRV_SECTOR_BITS;
-        data->size = buf_len;
+        data->lba = cpu_to_le64(offset >> BDRV_SECTOR_BITS);
+        data->size = cpu_to_le32(buf_len);
 
         n_bytes = buf_len + sizeof(VmdkGrainMarker);
         iov = (struct iovec) {