Re: [PATCH 10 of 20] ipath - support for userspace apps using coredriver

[Linus Torvalds]

On Wed, 22 Mar 2006, Bryan O'Sullivan wrote:
> 
> Under 2.6.15, what seems to be actually happening is that vmops->nopage
> is being called on each page of a 32K compound page, driving the page
> count from 1 (prior to any nopage calls) to 9.  By the time I get to my
> cleanup code, the page count has gone from 9 to 8 (whereas under 2.6.16,
> the page count has gone from 9 back to 1, where it belongs).  From this,
> it seems fairly clear that the kernel isn't decrementing the use counts
> correctly on compound pages in 2.6.15.

Ahh. Isn't this the same problem that the fairly recent "mm: compound 
release fix" by Nick should fix?

Git commit ID 8519fb30e438f8088b71a94a7d5a660a814d3872, to be exact:

Author: Nick Piggin <npiggin@xxxxxxx>
Date:   Tue Feb 7 12:58:52 2006 -0800

    [PATCH] mm: compound release fix
    
    Compound pages on SMP systems can now often be freed from pagetables via
    the release_pages path.  This uses put_page_testzero which does not handle
    compound pages at all.  Releasing constituent pages from process mappings
    decrements their count to a large negative number and leaks the reference
    at the head page - net result is a memory leak.
    
    The problem was hidden because the debug check in put_page_testzero itself
    actually did take compound pages into consideration.
    
    Fix the bug and the debug check.
    
    Signed-off-by: Nick Piggin <npiggin@xxxxxxx>
    Acked-by: Hugh Dickins <hugh@xxxxxxxxxxx>
    Signed-off-by: Andrew Morton <akpm@xxxxxxxx>
    Signed-off-by: Linus Torvalds <torvalds@xxxxxxxx>

But anything based on 2.6.16-rc6 should be fine (The bug was fixed by 
2.6.16-rc3, methinks). You said:

   "First of all, it turns out that the BUG I mentioned was reported 
    against the SLES10 2.6.16-rc6 kernel.  I haven't had a chance to chase 
    it down yet, but I'm going to have to, because..."

but if that _really_ is 2.6.16-rc6-based, this problem should have been 
fixed already.

Maybe SLES is based on 2.6._15_-rc6?

		Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/