Re: [BENCHMARK] 2.5.70-mm2 with contest

[Andrew Morton]

Nick Piggin <[e-mail]> wrote:
>
> It will be interesting to see what happens if we set the
>  ext3 journal write paths as PF_SYNCWRITE. I'll try some tests
>  a bit later today.
> 

OK.

Longer-term it would be best to lose the PF_SYNCWRITE thing and to just
mark the BIOs as synchronous prior to submitting them.  It's a matter of
transferring the info in writeback_control.sync_mode at the pagecache/BIO
boundary: mpage_bio_submit(), __block_write_full_page->submit_bh(), etc.

But we can worry about that later, once it is established that the
synchronous write detection is sufficiently useful.