From: Wu Fengguang <fengguang.wu@intel.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Linux Memory Management List <linux-mm@kvack.org>,
linux-fsdevel@vger.kernel.org, Ingo Molnar <mingo@elte.hu>,
Jens Axboe <jens.axboe@oracle.com>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
Rik van Riel <riel@redhat.com>,
LKML <linux-kernel@vger.kernel.org>,
Andi Kleen <andi@firstfloor.org>
Subject: Re: [PATCH 4/8] readahead: record readahead patterns
Date: Tue, 29 Nov 2011 10:40:15 +0800 [thread overview]
Message-ID: <20111129024015.GA19506@localhost> (raw)
In-Reply-To: <20111121151919.4b76a475.akpm@linux-foundation.org>
On Mon, Nov 21, 2011 at 03:19:19PM -0800, Andrew Morton wrote:
> On Mon, 21 Nov 2011 17:18:23 +0800
> Wu Fengguang <fengguang.wu@intel.com> wrote:
>
> > Record the readahead pattern in ra_flags and extend the ra_submit()
> > parameters, to be used by the next readahead tracing/stats patches.
> >
> > 7 patterns are defined:
> >
> > pattern readahead for
> > -----------------------------------------------------------
> > RA_PATTERN_INITIAL start-of-file read
> > RA_PATTERN_SUBSEQUENT trivial sequential read
> > RA_PATTERN_CONTEXT interleaved sequential read
> > RA_PATTERN_OVERSIZE oversize read
> > RA_PATTERN_MMAP_AROUND mmap fault
> > RA_PATTERN_FADVISE posix_fadvise()
> > RA_PATTERN_RANDOM random read
>
> It would be useful to spell out in full detail what an "interleaved
> sequential read" is, and why a read is considered "oversized", etc.
> The 'enum readahead_pattern' definition site would be a good place for
> this.
Good point, here is the added comments:
/*
* Which policy makes decision to do the current read-ahead IO?
*
* RA_PATTERN_INITIAL readahead window is initially opened,
* normally when reading from start of file
* RA_PATTERN_SUBSEQUENT readahead window is pushed forward
* RA_PATTERN_CONTEXT no readahead window available, querying the
* page cache to decide readahead start/size.
* This typically happens on interleaved reads (eg.
* reading pages 0, 1000, 1, 1001, 2, 1002, ...)
* where one file_ra_state struct is not enough
* for recording 2+ interleaved sequential read
* streams.
* RA_PATTERN_MMAP_AROUND read-around on mmap page faults
* (w/o any sequential/random hints)
* RA_PATTERN_BACKWARDS reverse reading detected
* RA_PATTERN_FADVISE triggered by POSIX_FADV_WILLNEED or FMODE_RANDOM
* RA_PATTERN_OVERSIZE a random read larger than max readahead size,
* do max readahead to break down the read size
* RA_PATTERN_RANDOM a small random read
*/
> > Note that random reads will be recorded in file_ra_state now.
> > This won't deteriorate cache bouncing because the ra->prev_pos update
> > in do_generic_file_read() already pollutes the data cache, and
> > filemap_fault() will stop calling into us after MMAP_LOTSAMISS.
> >
> > --- linux-next.orig/include/linux/fs.h 2011-11-20 20:10:48.000000000 +0800
> > +++ linux-next/include/linux/fs.h 2011-11-20 20:18:29.000000000 +0800
> > @@ -951,6 +951,39 @@ struct file_ra_state {
> >
> > /* ra_flags bits */
> > #define READAHEAD_MMAP_MISS 0x000003ff /* cache misses for mmap access */
> > +#define READAHEAD_MMAP 0x00010000
>
> Why leave a gap?
Never mind, it's now converted to a bit field :)
> And what is READAHEAD_MMAP anyway?
READAHEAD_MMAP will be set for mmap page faults.
> > +#define READAHEAD_PATTERN_SHIFT 28
>
> Why 28?
Bits 28-32 are for READAHEAD_PATTERN.
Anyway it will be gone when breaking down the ra_flags fields into
individual variables.
> > +#define READAHEAD_PATTERN 0xf0000000
> > +
> > +/*
> > + * Which policy makes decision to do the current read-ahead IO?
> > + */
> > +enum readahead_pattern {
> > + RA_PATTERN_INITIAL,
> > + RA_PATTERN_SUBSEQUENT,
> > + RA_PATTERN_CONTEXT,
> > + RA_PATTERN_MMAP_AROUND,
> > + RA_PATTERN_FADVISE,
> > + RA_PATTERN_OVERSIZE,
> > + RA_PATTERN_RANDOM,
> > + RA_PATTERN_ALL, /* for summary stats */
> > + RA_PATTERN_MAX
> > +};
>
> Again, the behaviour is all undocumented. I see from the code that
> multiple flags can be set at the same time. So afacit a file can be
> marked RANDOM and SUBSEQUENT at the same time, which seems oxymoronic.
Nope, it will be classified into one "pattern" exclusively.
> This reader wants to know what the implications of this are - how the
> code chooses, prioritises and acts. But this code doesn't tell me.
Hope the comment addresses this issue. The precise logic happens
mainly inside ondemand_readahead().
> > +static inline unsigned int ra_pattern(unsigned int ra_flags)
> > +{
> > + unsigned int pattern = ra_flags >> READAHEAD_PATTERN_SHIFT;
>
> OK, no masking is needed because the code silently assumes that arg
> `ra_flags' came out of an ra_state.ra_flags and it also silently
> assumes that no higher bits are used in ra_state.ra_flags.
>
> That's a bit of a handgrenade - if someone redoes the flags
> enumeration, the code will explode.
Yeah sorry for playing with such tricks. Will get rid of this function
totally and use a plain assign to ra->pattern.
> > + return min_t(unsigned int, pattern, RA_PATTERN_ALL);
> > +}
>
> <scratches head>
>
> What the heck is that min_t() doing in there?
Just for safety... not really necessary given correct code.
Thanks,
Fengguang
next prev parent reply other threads:[~2011-11-29 2:40 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-11-21 9:18 [PATCH 0/8] readahead stats/tracing, backwards prefetching and more Wu Fengguang
2011-11-21 9:18 ` [PATCH 1/8] block: limit default readahead size for small devices Wu Fengguang
2011-11-21 10:00 ` Christoph Hellwig
2011-11-21 11:24 ` Wu Fengguang
2011-11-21 12:47 ` Andi Kleen
2011-11-21 14:46 ` Jeff Moyer
2011-11-21 22:52 ` Andrew Morton
2011-11-22 14:23 ` Jeff Moyer
2011-11-23 12:18 ` Wu Fengguang
2011-11-21 9:18 ` [PATCH 2/8] readahead: make default readahead size a kernel parameter Wu Fengguang
2011-11-21 10:01 ` Christoph Hellwig
2011-11-21 11:35 ` Wu Fengguang
2011-11-24 22:28 ` Jan Kara
2011-11-25 0:36 ` Dave Chinner
2011-11-28 2:39 ` Wu Fengguang
2011-11-30 13:04 ` Christian Ehrhardt
2011-11-30 13:29 ` Wu Fengguang
2011-11-30 16:09 ` Jan Kara
2011-11-21 9:18 ` [PATCH 3/8] readahead: replace ra->mmap_miss with ra->ra_flags Wu Fengguang
2011-11-21 11:04 ` Steven Whitehouse
2011-11-21 11:42 ` Wu Fengguang
2011-11-21 23:01 ` Andrew Morton
2011-11-23 12:47 ` Wu Fengguang
2011-11-23 20:31 ` Andrew Morton
2011-11-29 3:42 ` Wu Fengguang
2011-11-21 9:18 ` [PATCH 4/8] readahead: record readahead patterns Wu Fengguang
2011-11-21 23:19 ` Andrew Morton
2011-11-29 2:40 ` Wu Fengguang [this message]
2011-11-21 9:18 ` [PATCH 5/8] readahead: add /debug/readahead/stats Wu Fengguang
2011-11-21 14:17 ` Andi Kleen
2011-11-22 14:14 ` Wu Fengguang
2011-11-21 23:29 ` Andrew Morton
2011-11-21 23:32 ` Andi Kleen
2011-11-29 3:23 ` Wu Fengguang
2011-11-29 4:49 ` Andrew Morton
2011-11-29 6:41 ` Wu Fengguang
2011-11-29 12:29 ` Wu Fengguang
2011-11-21 9:18 ` [PATCH 6/8] readahead: add debug tracing event Wu Fengguang
2011-11-21 14:01 ` Steven Rostedt
2011-11-21 9:18 ` [PATCH 7/8] readahead: basic support for backwards prefetching Wu Fengguang
2011-11-21 23:33 ` Andrew Morton
2011-11-29 3:08 ` Wu Fengguang
2011-11-21 9:18 ` [PATCH 8/8] readahead: dont do start-of-file readahead after lseek() Wu Fengguang
2011-11-21 23:36 ` Andrew Morton
2011-11-22 14:18 ` Wu Fengguang
2011-11-21 9:56 ` [PATCH 0/8] readahead stats/tracing, backwards prefetching and more Christoph Hellwig
2011-11-21 12:00 ` Wu Fengguang
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20111129024015.GA19506@localhost \
--to=fengguang.wu@intel.com \
--cc=a.p.zijlstra@chello.nl \
--cc=akpm@linux-foundation.org \
--cc=andi@firstfloor.org \
--cc=jens.axboe@oracle.com \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mingo@elte.hu \
--cc=riel@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).