[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [dm-devel] [PATCHSET #upstream] block, fs: replace HARDBARRIER with FLUSH/FUA, take#2



On 09/03/2010 12:29 PM, Tejun Heo wrote:
> Hello,
> 
> This patchset is combination of the following two patchsets.
> 
>  [1] block, fs: replace HARDBARRIER with FLUSH/FUA
>  [2] block, dm: finish REQ_FLUSH/FUA conversion, take#2
> 
> Changes from the previous postings are
> 
> * Rebased on top of v2.6.36-rc3.
> 
> * 0010-block-implement-REQ_FLUSH-FUA-based-interface-for-FL.patch
>   description updated to include why special -EOPNOTSUPP handling is
>   no longer necessary as suggested by Mike Snitzer.
> 
> * 0025-dm-relax-ordering-of-bio-based-flush-implementation.patch
>   description updated to include why flush_error and special error
>   handling logic are no longer necessary as suggested by Mike
>   Snitzer.
> 
> * 0013-block-initialize-flush-request-with-WRITE_FLUSH-inst.patch and
>   0014-block-kick-queue-after-sequencing-REQ_FLUSH-FUA.patch added.
>   These fixes aren't strictly required by the current users but remove
>   fishy corner cases.
> 
> * 0015-block-make-sure-FSEQ_DATA-request-has-the-same-rq_di.patch
>   added.  This is needed to make requested based dm work.
> 
> * 0026-dm-fix-locking-context-in-queue_io.patch added.  This fixes
>   locking context problem introduced by
>   0025-dm-relax-ordering-of-bio-based-flush-implementation.patch.
> 
> * Tested/Reviewed-by's added and patches reordered such that block
>   layer conversion comes first followed by block driver conversions,
>   filesystem conversions and then finally removal of deprecate stuff.
> 
> I think the patchset is mostly ready now although I would love to have
> acked/reviewd-by's on bio-based dm conversion patches.  Jens,
> Christoph, how do you guys wanna steer these patches?  I think it
> would be best to go through the block tree but splitting parts and
> doing cross tree pull/merges should be doable too.

Since the core changes are block churn, it would probably be
sanest to pull everything in there. Core into for-2.6.37/core,
then have for-2.6.37/fs and 2.6.37/drivers on top of that.

-- 
Jens Axboe


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]