linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
* Re: [PATCH] mm: call pte_unmap() against a proper pte (Re: [PATCH 7/9] swap_info: swap count continuations)
@ 2009-10-17 22:44 hugh.dickins
  0 siblings, 0 replies; 3+ messages in thread
From: hugh.dickins @ 2009-10-17 22:44 UTC (permalink / raw)
  To: nishimura
  Cc: Andrew Morton, Nitin Gupta, KAMEZAWA Hiroyuki, hongshin,
	linux-kernel, linux-mm



>----Original Message----
>From: nishimura@mxp.nes.nec.co.jp
>Date: 16/10/2009 7:30 
>To: "Hugh Dickins"<hugh.dickins@tiscali.co.uk>
>Cc: "Andrew Morton"<akpm@linux-foundation.org>, "Nitin Gupta"<ngupta@vflare.
org>, "KAMEZAWA Hiroyuki"<kamezawa.hiroyu@jp.fujitsu.com>, <hongshin@gmail.
com>, <linux-kernel@vger.kernel.org>, <linux-mm@kvack.org>, "Daisuke Nishimura"
<nishimura@mxp.nes.nec.co.jp>
>Subj: [PATCH] mm: call pte_unmap() against a proper pte (Re: [PATCH 7/9] 
swap_info: swap count continuations)
>
>Hi.
>
>> @@ -645,6 +648,7 @@ static int copy_pte_range(struct mm_stru
>>  	spinlock_t *src_ptl, *dst_ptl;
>>  	int progress = 0;
>>  	int rss[2];
>> +	swp_entry_t entry = (swp_entry_t){0};
>>  
>>  again:
>>  	rss[1] = rss[0] = 0;
>> @@ -671,7 +675,10 @@ again:
>>  			progress++;
>>  			continue;
>>  		}
>> -		copy_one_pte(dst_mm, src_mm, dst_pte, src_pte, vma, addr, rss);
>> +		entry.val = copy_one_pte(dst_mm, src_mm, dst_pte, src_pte,
>> +							vma, addr, rss);
>> +		if (entry.val)
>> +			break;
>>  		progress += 8;
>>  	} while (dst_pte++, src_pte++, addr += PAGE_SIZE, addr != end);
>>  
>It isn't the fault of only this patch, but I think breaking the loop without 
incrementing
>dst_pte(and src_pte) would be bad behavior because we do unmap_pte(dst_pte - 
1) later.
>(current copy_pte_range() already does it though... and this is only 
problematic
>when we break the first loop, IIUC.)

Good catch, thanks a lot for finding that.  I believe this is entirely a fault 
in my 7/9, the existing code
takes care not to break before it has made some progress (in part because of 
this unmap issue).

>
>> @@ -681,6 +688,12 @@ again:
>>  	add_mm_rss(dst_mm, rss[0], rss[1]);
>>  	pte_unmap_unlock(dst_pte - 1, dst_ptl);
>>  	cond_resched();
>> +
>> +	if (entry.val) {
>> +		if (add_swap_count_continuation(entry, GFP_KERNEL) < 0)
>> +			return -ENOMEM;
>> +		progress = 0;
>> +	}
>>  	if (addr != end)
>>  		goto again;
>>  	return 0;
>
>I've searched other places where we break a similar loop and do pte_unmap(pte 
- 1).
>Current copy_pte_range() and apply_to_pte_range() has the same problem.

And thank you for taking the trouble to look further afield: yes, 
apply_to_pte_range()
was already wrong.

>
>How about a patch like this ?
>===
>From: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
>
>There are some places where we do like:
>
>	pte = pte_map();
>	do {
>		(do break in some conditions)
>	} while (pte++, ...);
>	pte_unmap(pte - 1);
>
>But if the loop breaks at the first loop, pte_unmap() unmaps invalid pte.
>
>This patch is a fix for this problem.
>
>Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>

Acked-by: Hugh Dickins <hugh.dickins@tiscali.co.uk>





Forget the rest, get the best - http://www.tiscali.co.uk/music

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

^ permalink raw reply	[flat|nested] 3+ messages in thread
* [PATCH 0/9] swap_info and swap_map patches
@ 2009-10-15  0:44 Hugh Dickins
  2009-10-15  0:56 ` [PATCH 7/9] swap_info: swap count continuations Hugh Dickins
  0 siblings, 1 reply; 3+ messages in thread
From: Hugh Dickins @ 2009-10-15  0:44 UTC (permalink / raw)
  To: Andrew Morton; +Cc: KAMEZAWA Hiroyuki, linux-kernel, linux-mm

Here's a series of nine patches around the swap_info_struct: against
2.6.32-rc4, but intended for mmotm, which is currently similar here.

They start out with some old and not very important cleanups, but get
around to solving the swap count overflow problem: our handling above
32765 has depended on hoping that it won't coincide with other races.

That problem exists in theory today (when pid_max is raised from its
default), though never reported in practice; but the motivation for
solving it now comes from the impending KSM swapping patches - it
becomes very easy for anyone to overflow the maximum that way.

But most people will never have a swap count overflow in their life:
the benefit for them is that the vmalloc'ed swap_map halves in size.

This is all internal housekeeping: no change to actual swapping and
page reclaim.

 include/linux/swap.h |   66 ++-
 mm/memory.c          |   19 
 mm/page_io.c         |   19 
 mm/rmap.c            |    6 
 mm/shmem.c           |   11 
 mm/swapfile.c        |  834 +++++++++++++++++++++++++----------------
 6 files changed, 599 insertions(+), 356 deletions(-)

Hugh

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

^ permalink raw reply	[flat|nested] 3+ messages in thread
end of thread, other threads:[~2009-10-17 22:44 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-10-17 22:44 [PATCH] mm: call pte_unmap() against a proper pte (Re: [PATCH 7/9] swap_info: swap count continuations) hugh.dickins
  -- strict thread matches above, loose matches on Subject: below --
2009-10-15  0:44 [PATCH 0/9] swap_info and swap_map patches Hugh Dickins
2009-10-15  0:56 ` [PATCH 7/9] swap_info: swap count continuations Hugh Dickins
2009-10-16  6:30   ` [PATCH] mm: call pte_unmap() against a proper pte (Re: [PATCH 7/9] swap_info: swap count continuations) Daisuke Nishimura
2009-10-16  8:01     ` KAMEZAWA Hiroyuki

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).