Re: [Qemu-block] [PATCH for-2.8 v1 1/1] block/vmdk: Fix the endian problem of buf_len

[Hao QingFeng]

在 2016-11-25 20:05, Kevin Wolf 写道:
> Am 25.11.2016 um 11:48 hat Hao QingFeng geschrieben:
> > 在 2016-11-25 18:21, Kevin Wolf 写道:
> > > [ Cc: Fam, qemu-stable ]
> > >
> > > Am 25.11.2016 um 11:06 hat QingFeng Hao geschrieben:
> > > > 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.
> > > >
> > > > Signed-off-by: QingFeng Hao <address@hidden>
> > > > Signed-off-by: Jing Liu <address@hidden>
> > > Sounds like something that should still be fixed for 2.8 and in the
> > > stable branches.
> > I didn't find the stable branch on upstream, the latest is 2.6 maybe
> > it's a private one? :-)
> I think maybe it's only created once work for the 2.7.1 release begins.
>
> Anyway, this is not your problem, just make sure to CC the qemu-stable
> mailing list for bug fixes that affect previous versions as well. It's
> also helpful if you put a line like this immediately before the
> signoffs:
>
> Cc: address@hidden
>
> This makes sure that the stable release maintainers see the patch,
> and everything else will be taken care of by them.
Thanks and I'll add it in the next patch.
> > > > diff --git a/block/vmdk.c b/block/vmdk.c
> > > > index a11c27a..bf6667f 100644
> > > > --- a/block/vmdk.c
> > > > +++ b/block/vmdk.c
> > > > @@ -1355,7 +1355,7 @@ static int vmdk_write_extent(VmdkExtent *extent, int64_t 
> > > > cluster_offset,
> > > >           }
> > > >           data->lba = offset >> BDRV_SECTOR_BITS;
> > > > -        data->size = buf_len;
> > > > +        data->size = cpu_to_le32(buf_len);
> > > At least data->lba needs to be fixed, too, both here and in
> > > vmdk_read_extent(). Host endianness in an image file is always wrong.
> > Good detection!
> Are you going to send a v2 that adds this fix?
Yes, I think so.  :-)
> > > Maybe we should audit the whole driver for endianness problems.
> > Good sight! maybe we can encapsulate a suite of endianness functions
> > for the structures to avoid missing some elements of them like lba?
> > Also will be better for reuse. When I am checking the endianness calls
> > in vmdk_create_extent, I am just afraid of missing one. :-)
> Sounds like a good move in general, but let's make sure to keep it
> separate from the fixes. The reason is that the fixes can still be
> included in the 2.8 release, but code refactoring would only be for 2.9.
>
> You can work on both at the same time, of course, but organise things so
> that we have a patch series where all the bug fixes come first (for 2.8)
> and only then the type-safe refactoring (for 2.9).
Thanks for your advice! I'll ensure the bug fixes come first and prepare 
other stuff meanwhile.
> Kevin

--
QingFeng Hao(Robin)