[prev in list] [next in list] [prev in thread] [next in thread] 

List:       swsusp-devel
Subject:    RE: [Swsusp] Nigel's Big Suspend Patch Version 2.
From:       Nigel Cunningham <ncunningham () clear ! net ! nz>
Date:       2002-09-22 19:39:07
[Download RAW message or body]

Hi!

> > Hi.
> >
> > Thanks for the feedback. I am still working on it, and have
> another patch
> in
> > the works. Actually I'm having a go at getting the 2.4.19
> patch going at
> the
> > moment. I've fixed the missing unlock in
> is_head_of_free_chunck, and am
> now
> > dealing with the next issue. It will probably be Monday
> (Sunday for you
> > Americans) before you see a patch though.
> >
> > Regards,
> >

I've posted a 2.4.19 patch this morning. The page_alloc message in the oops
was misleading. The problem was that block_flushpage call was eventually
trying to free the page we're saving to disk. This was resulting in an oops
because the page still had a mapping and was locked. I fixed this by making
the call to page_cache_release dependant upon the page not being mapped and
not being locked. This shouldn't adversely affect anything else because
anything that met these criteria would cause the same oops anyway (and we
don't actually want to free the page!).

In addition to this, there were a couple of other typos in Florent's changes
to the read/write code in suspend.c - nothing major. (Oh, and the original
lack of a spin-unlock in the is_head_of_free_chunck routine).

>
> I'm wondering where / how you fixed the page locking
> problem... could you
> tell me what you changed so that I may continue to look at the 2.4.19
> issue(s) as well?  As I mentioned in another reply, I took a
> break from it,
> but I think I'll continue working on it in the next few days
> as well.  What
> is the 'next issue' that you have found?
>

> You should NOT be suspending RAID threads.  What I did was
> set the flag so
> that swsusp would skip over RAID threads and let them continue running
> (using current->flags |= PF_IOTHREAD)... if PF_SUSPEND_TASK
> is equivalent to
> PF_IOTHREAD, then you're a-okay there.  I noticed that you
> kept the code
> that notifies md (RAID) of the suspend and resume (in
> drivers_suspend/resume), so that's good too.  I didn't
> actually apply the
> patch and try it yet, but it the code for RAID looks good.

Okee doke. I'll leave it as PF_SUSPEND_TASK then. I'm trying to keep my
patch up to date with Florent's, so anything added by her(?) should appear
shortly afterwards from me.

Regards,

Nigel



_______________________________________________
swsusp mailing list
swsusp@lister.fornax.hu
http://lister.fornax.hu/mailman/listinfo/swsusp
[prev in list] [next in list] [prev in thread] [next in thread]