[lkml]   [2009]   [May]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH 0/6] kill-the-BKL/reiserfs3: performance improvements, faster than Bkl based scheme

* Linus Torvalds <> wrote:

> On Fri, 1 May 2009, Frederic Weisbecker wrote:
> >
> > This reiserfs patchset applies against latest
> > tip:core/kill-the-BKL It adds various explicit write lock
> > releases on specific sleeping sections.
> Btw, is there any reason why it cannot just be re-based on top of
> standard -rc4?
> I'd love to pull a "reiserfs: remove bkl" branch when the next
> merge window opens, but there's no way I'll pull the kill-bkl
> thing with all the odd random tty stuff etc that is totally
> unrelated.

Btw., i can name another reason why we'd want to do reiserfs
separately: if our testing and efforts are any proof, then reiser3
turned out to be the hardest BKL nut to crack, by a wide margin.

All the other hacks in kill-the-BKL are really of relatively low
complexity and really just tried to map out the problem areas. Even
the tty ones are simple - just a few recursion assumptions.

We _suspected_ that kind of status quo before, but we never had any
conclusive proof of that. I think we now know that for sure, and we
have to fear the BKL no more.

After the reiser3 conversion we should just tackle all the other BKL
users one by one, and go for a straight subsystem mutex in every
case: it will cause little trouble, and it will be a job that can be
finished within a reasonable time frame.


 \ /
  Last update: 2009-05-01 23:35    [W:0.315 / U:1.556 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site