[PATCH 0/1] [Karmic] LP#420423 -- virtio barriers fix

[Stefan Bader]

Tim Gardner wrote:
> Andy Whitcroft wrote:
>> The following patch fixes virtio based log spam reporting discarded
>> empty barriers.  This is basically a pull in of the patch which was proposed
>> for this issue as below:
>>
>>     http://patchwork.kernel.org/patch/39589/
>>
>> As yet this has not been applied upstream but it does appear to be receiving
>> positive responses.  This closes a couple of karmic regressions.
>>
>> Proposing for Karmic.
>>
>> -apw
>>
>> Andy Whitcroft (1):
>>   UBUNTU: [Upstream] block: silently error unsupported empty barriers
>>     too
>>
>>  block/blk-core.c |    2 +-
>>  1 files changed, 1 insertions(+), 1 deletions(-)
>>
>>
> 
> 
> I'll defer to the other expert opinions since I've not the faintest idea 
> what a bio_barrier() or bio_has_data() does.
> 
> rtg

Without looking at the code, but bio_barrier() should just be a bit test which 
checks whether a bio (atomic part of a request) has the barrier flag set. If 
that is true, devices/block layers that support barriers must ensure that 
everything sent up to and including this part of I/O has been physically placed 
into a non-volatile storage before proceeding. Which is nice if you directly 
use a device supporting things like tagged commands and partial flushes but in 
most cases of soft-RAIDs would require additional code which is not there anyways.
And bio_has_data() just checks whether the bio used has payload data attached 
(which it does dot need to).

Stefan