Re: [GIT PULL] for-2.6.32/bug-fixes

["Jan Beulich" <JBeulich@novell.com>]

>>> On 17.05.11 at 17:57, Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> wrote:
>> > No attaching of data to the barrier.
>> 
>> Sure, this direction we agree about. But your change is enforcing
>> it the other way around (if barrier then no data), which wasn't the
>> case so far.
> 
> OK, even if the code that actually does the bio submission does
> not attach any data to the bio? The end result is the same - no
> data with barriers.

My problem is that I can't see where attaching data would be
skipped. The only thing I see is the BUG_ON() you pointed at
earlier, checking that if there is no data, then this must be a
barrier request.

>> >> Additionally, looking at the check in vbd_translate(), wouldn't you
>> >> think there ought to be overflow checking for the addition, too?
>> > 
>> > Sure, could add that in. Albeit it seems incorrect to do it in that
>> > function. It checks to see if the sector is correct, and -1 is definitly
>> > wrong.
>> 
>> Hmm, depends on your perspective - I'd say that any sector_number
>> is valid when nr_sects is zero.
> 
> I concur. The value that is passed by the frontend is not zero. It is -1.

Oh, you say both sector_number and nr_sects are -1? Looking
again... No, that can't be the case, the value starts out at zero
in dispatch_rw_block_io().

Jan