Data Corruption bug with Samba's vfs_iouring and Linux 5.6.7/5.7rc3

[Jens Axboe]

On 5/7/20 12:55 PM, Jeremy Allison wrote:
> On Thu, May 07, 2020 at 12:35:42PM -0600, Jens Axboe wrote:
>> On 5/7/20 12:31 PM, Jeremy Allison wrote:
>>>
>>> Look at how quickly someone spotted disk corruption
>>> because of the change in userspace-visible behavior
>>> of the io_uring interface. We only shipped that code
>>> 03 March 2020 and someone *already* found it.
>>
>> I _think_ that will only happen on regular files if you use RWF_NOWAIT
>> or similar, for regular blocking it should not happen. So I don't think
>> you're at risk there, though I do think that anyone should write
>> applications with short IOs in mind or they will run into surprises down
>> the line. Should have been more clear!
> 
> Well we definitely considered short IOs writing the
> server code, but as the protocol allows that to be
> visible to the clients (in fact it has explicit
> fields meant to deal with it) it wasn't considered
> vital to hide them from clients.

Yes, and in case my reply wasn't totally clear, it was more of a general
observation, not directed at Samba specifically!

> We'll certainly fix up short reads for the iouring
> module, but it's less clear we should mess with
> our existing blocking threaded pread/pwrite code
> to deal with them. Possibly goes into the bucket
> of "belt and braces, couldn't possibly hurt" :-).

Agree, belts and suspenders for the regular pread/pwrite, that's a fair
position.

> Thanks for the clarification !

Thanks for getting this fleshed out! Impressed with the speed at which
we got to the bottom of this.

-- 
Jens Axboe