Re: Fwd: Ext4 bug with fallocate

[Eric Sandeen <sandeen@redhat.com>]

Fredrik Andersson wrote:
>> Not trying to ignore you, but trying to sort out some corruption issues that
>> I'm afraid I have to rank higher than a hang for now...
> 
> I understand. Good to know someone is thinking about it anyway :-)
> 
>> From the traces, things are hung up on the i_data_sem; several pdflushes
>> want it for read in ext4_get_blocks_wrap, and drbdmakes presumably have it
>> downed for write in ext4_ext_truncate.  I'm not immediately seeing the
>> deadlock tonight, but then it's getting late...
>>
>> Is the application that's triggering this something you can share?
>>
> 
> Sorry, I can't provide source code for these applications.

Ok, I understand

> But drdbmake does what is described above: Preallocates a very big
> chunk on disk and then
> truncates it to its final length once it's done.

To try to emulate, how does it write into the preallocated space; large 
or small IOs?  Sequential streaming?  mmap writes?  It may not be 
relevant but would be nice to try to match it as closely as possible.

Thanks,
-Eric

> Please let me know if you need more details.
> 
> /Fredrik