linux-kernel - Re: [RPC][PATCH 2.6.20-rc5] limit total vfs page cache

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <6d6a94c50701232153w7bfd28d2s62aaec7ed48a9b0c@mail.gmail.com>
Date:	Wed, 24 Jan 2007 13:53:40 +0800
From:	"Aubrey Li" <aubreylee@...il.com>
To:	"Vaidyanathan Srinivasan" <svaidy@...ux.vnet.ibm.com>
Cc:	linux-kernel@...r.kernel.org, linux-mm@...ck.org,
	"Linus Torvalds" <torvalds@...l.org>,
	"Andrew Morton" <akpm@...l.org>,
	"Nick Piggin" <nickpiggin@...oo.com.au>,
	"linux-os (Dick Johnson)" <linux-os@...logic.com>,
	"Robin Getz" <rgetz@...ckfin.uclinux.org>
Subject: Re: [RPC][PATCH 2.6.20-rc5] limit total vfs page cache

On 1/24/07, Vaidyanathan Srinivasan <svaidy@...ux.vnet.ibm.com> wrote:
>
>
> Aubrey Li wrote:
> > On 1/19/07, Vaidyanathan Srinivasan <svaidy@...ux.vnet.ibm.com> wrote:
> >> Hi Aubrey,
> >>
> >> I used your patch on my PPC64 box and I do not get expected
> >> behavior.  As you had requested, I am attaching zoneinfo and meminfo
> >> dumps:
> >>
> >> Please let me know if you need any further data to help me out with
> >> the test/experiment.
> >>
> >
> > Although I have no PPC64 box in hand, I think the logic should be the same.
> > get_page_from_freelist() is called 5 times in __alloc_pages().
> >
> > 1) alloc_flags = ALLOC_WMARK_LOW | ALLOC_PAGECACHE;
> > 2) alloc_flags = ALLOC_WMARK_MIN | ALLOC_PAGECACHE;
> > We should have the same result on the first two times get_page_from_freelist().
> >
> > 3) if (((p->flags & PF_MEMALLOC) || unlikely(test_thread_flag(TIF_MEMDIE)))
> >                       && !in_interrupt())
> >    alloc_flags = ALLOC_NO_WATERMARKS
> > The case on my platform will never enter this branch. If the branch
> > occurs on your side,
> > The limit will be omitted. Because NO watermark, zone_watermark_ok()
> > will not be checked. memory will be allocated directly.
> >
> > 4)if (likely(did_some_progress)) {
> >    alloc_flags should include ALLOC_PAGECACHE.
> > So we should have the same result on this call.
> >
> > 5)    } else if ((gfp_mask & __GFP_FS) && !(gfp_mask & __GFP_NORETRY)) {
> >    alloc_flags = ALLOC_WMARK_HIGH, without ALLOC_PAGECACHE
> >
> > This branch will not hit on my case. You may need to check it.
> >
> > If 3) or 5) occurs on your platform, I think you can easily fix it.
> > Please confirm it and let me know the result.
>
>
> None of the above condition was the problem in my PPC64 box.  I
> added __GFP_PAGECACHE flag in pagecache_alloc_cold() and
> grab_cache_page_nowait() routines and the reclaim seemed to work.
>
> --- linux-2.6.20-rc5.orig/include/linux/pagemap.h
> +++ linux-2.6.20-rc5/include/linux/pagemap.h
> @@ -62,12 +62,12 @@ static inline struct page *__page_cache_
>
>  static inline struct page *page_cache_alloc(struct address_space *x)
>  {
> -       return __page_cache_alloc(mapping_gfp_mask(x));
> +       return __page_cache_alloc(mapping_gfp_mask(x)|__GFP_PAGECACHE);
>  }
>
>  static inline struct page *page_cache_alloc_cold(struct
> address_space *x)
>  {
> -       return __page_cache_alloc(mapping_gfp_mask(x)|__GFP_COLD);
> +       return
> __page_cache_alloc(mapping_gfp_mask(x)|__GFP_COLD|__GFP_PAGECACHE);
>  }
>
>  typedef int filler_t(void *, struct page *);
>
> [snip]
>
> --- linux-2.6.20-rc5.orig/mm/filemap.c
> +++ linux-2.6.20-rc5/mm/filemap.c
> @@ -823,7 +823,7 @@ grab_cache_page_nowait(struct address_sp
>                 page_cache_release(page);
>                 return NULL;
>         }
> -       page = __page_cache_alloc(mapping_gfp_mask(mapping) & ~__GFP_FS);
> +       page = __page_cache_alloc(mapping_gfp_mask(mapping) & ~__GFP_FS |
> __GFP_PAGECACHE);
>         if (page && add_to_page_cache_lru(page, mapping, index, GFP_KERNEL)) {
>                 page_cache_release(page);
>                 page = NULL;
>
>
> pagecache_alloc_cold() is used in the read-ahead path which was
> being called in my case of large file operations.
>
> --Vaidy
>
Thanks to point it out. There is another patch on the LKML which I
think is better.
Checking the zone->max_pagecache  in the get_page_from_freelist() is
better than checking the watermark in zone_watermark_ok(). Let me know
if it works for you.

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