Re: [RFC][PATCH v3] readahead: introduce O_RANDOM for POSIX_FADV_RANDOM

[Minchan Kim <minchan.kim@gmail.com>]

Hi, Wu.

On Mon, Jan 4, 2010 at 1:50 PM, Wu Fengguang <fengguang.wu@intel.com> wrote:
> This fixes inefficient page-by-page reads on POSIX_FADV_RANDOM.
>
> POSIX_FADV_RANDOM used to set ra_pages=0, which leads to poor
> performance: a 16K read will be carried out in 4 _sync_ 1-page reads.
>
> In other places, ra_pages==0 means
> - it's ramfs/tmpfs/hugetlbfs/sysfs/configfs
> - some IO error happened
> where multi-page read IO won't help or should be avoided.
>
> POSIX_FADV_RANDOM actually want a different semantics: to disable the
> *heuristic* readahead algorithm, and to use a dumb one which faithfully
> submit read IO for whatever application requests.
>
> So introduce a flag O_RANDOM for POSIX_FADV_RANDOM.
> It will be visible to fcntl(F_GETFL).
>
> Note that the random hint is not likely to help random reads performance
> noticeably. And it may be too permissive on huge request size (its IO
> size is not limited by read_ahead_kb).
>
> In Quentin's report (http://lkml.org/lkml/2009/12/24/145), the overall
> (NFS read) performance of the application increased by 313%!
>
> v3: use O_RANDOM to indicate both read/write access pattern as in
>    posix_fadvise(), although it only takes effect for read() now
>    (proposed by Quentin)
> v2: use O_RANDOM_READ to avoid race conditions (pointed out by Andi)
>
> CC: Nick Piggin <npiggin@suse.de>
> CC: Andi Kleen <andi@firstfloor.org>
> CC: Steven Whitehouse <swhiteho@redhat.com>
> CC: David Howells <dhowells@redhat.com>
> CC: Al Viro <viro@zeniv.linux.org.uk>
> CC: Jonathan Corbet <corbet@lwn.net>
> CC: Christoph Hellwig <hch@infradead.org>
> Tested-by: Quentin Barnes <qbarnes+nfs@yahoo-inc.com>
> Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
> ---
>  include/asm-generic/fcntl.h |    4 ++++
>  mm/fadvise.c                |   10 +++++++++-
>  mm/readahead.c              |    6 ++++++
>  3 files changed, 19 insertions(+), 1 deletion(-)
>
> --- linux.orig/include/asm-generic/fcntl.h      2010-01-04 12:39:29.000000000 +0800
> +++ linux/include/asm-generic/fcntl.h   2010-01-04 12:40:11.000000000 +0800
> @@ -80,6 +80,10 @@
>  #define O_NDELAY       O_NONBLOCK
>  #endif
>
> +#ifndef O_RANDOM
> +#define O_RANDOM       010000000       /* random access pattern hint */
> +#endif
> +
>  #define F_DUPFD                0       /* dup */
>  #define F_GETFD                1       /* get close_on_exec */
>  #define F_SETFD                2       /* set/clear close_on_exec */
> --- linux.orig/mm/fadvise.c     2010-01-04 12:39:29.000000000 +0800
> +++ linux/mm/fadvise.c  2010-01-04 12:39:30.000000000 +0800
> @@ -77,12 +77,20 @@ SYSCALL_DEFINE(fadvise64_64)(int fd, lof
>        switch (advice) {
>        case POSIX_FADV_NORMAL:
>                file->f_ra.ra_pages = bdi->ra_pages;
> +               spin_lock(&file->f_lock);
> +               file->f_flags &= ~O_RANDOM;
> +               spin_unlock(&file->f_lock);
>                break;
>        case POSIX_FADV_RANDOM:
> -               file->f_ra.ra_pages = 0;
> +               spin_lock(&file->f_lock);
> +               file->f_flags |= O_RANDOM;
> +               spin_unlock(&file->f_lock);
>                break;
>        case POSIX_FADV_SEQUENTIAL:
>                file->f_ra.ra_pages = bdi->ra_pages * 2;
> +               spin_lock(&file->f_lock);
> +               file->f_flags &= ~O_RANDOM;
> +               spin_unlock(&file->f_lock);
>                break;
>        case POSIX_FADV_WILLNEED:
>                if (!mapping->a_ops->readpage) {
> --- linux.orig/mm/readahead.c   2010-01-04 12:39:29.000000000 +0800
> +++ linux/mm/readahead.c        2010-01-04 12:39:30.000000000 +0800
> @@ -501,6 +501,12 @@ void page_cache_sync_readahead(struct ad
>        if (!ra->ra_pages)
>                return;
>
> +       /* be dumb */
> +       if (filp->f_flags & O_RANDOM) {
> +               force_page_cache_readahead(mapping, filp, offset, req_size);
> +               return;
> +       }
> +

Let me have a dumb question. :)

How about testing O_RANDOM in front of ra_pages testing?

My intention is that although we turn off ra, it would be better to read
contiguous block all at once than readpage() callback doing I/O
one page at a time.

Is it break some semantics or happen some problem in ondemand readahead?

>        /* do read-ahead */
>        ondemand_readahead(mapping, ra, filp, false, offset, req_size);
>  }
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>



-- 
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html