archive mirror
 help / color / mirror / Atom feed
From: Stephen Satchell <>
Cc: linux-kernel <>
Subject: Re: Swap performance statistics in 2.6 -- which /proc file has it?
Date: Tue, 09 Dec 2003 06:46:25 -0800	[thread overview]
Message-ID: <> (raw)
In-Reply-To: <Pine.LNX.4.53.0312090854080.8425@chaos>

On Tue, 2003-12-09 at 05:56, Richard B. Johnson wrote:
> On Tue, 9 Dec 2003, Stephen Satchell wrote:
> > I also grepped the proc filesystem source (linux/fs/proc) for "swap" and
> > "Swap" and didn't find anything that had to do with swap request accounting,
> > only with swap memory allocation (which I do use, but which for me is only
> > half the story).

> /proc/meminfo may give you the information you need, just not
> as directly as you propose.

I appreciate your answer.  From the sprintf statement in
fs/proc/proc_misc.c that builds the return for mem_info:


Tracing around a bit, I find /mm/swapfile.c defines the routine void
si_swapinfo(struct sysinfo *) which, in part, does this:

        for (i = 0; i < nr_swapfiles; i++) {
                if (!(swap_info[i].flags & SWP_USED) ||
                     (swap_info[i].flags & SWP_WRITEOK))
                nr_to_be_unused += swap_info[i].inuse_pages;
        val->freeswap = nr_swap_pages + nr_to_be_unused;
        val->totalswap = total_swap_pages + nr_to_be_unused;
So it collect allocation information, but doesn't collect any
performance information (assuming that performance information is saved
for each swap file in the swapfile list) to return to the meminfo
handler, or to anyone else for that matter.
Is there performance information to collect?  I'm beginning to think
not, based on looking at include/linux/swap.h:

struct swap_info_struct {
        unsigned int flags;
        spinlock_t sdev_lock;
        struct file *swap_file;
        struct block_device *bdev;
        struct list_head extent_list;
        int nr_extents;
        struct swap_extent *curr_swap_extent;
        unsigned old_block_size;
        unsigned short * swap_map;
        unsigned int lowest_bit;
        unsigned int highest_bit;
        unsigned int cluster_next;
        unsigned int cluster_nr;
        int prio;                       /* swap priority */
        int pages;
        unsigned long max;
        unsigned long inuse_pages;
        int next;                       /* next entry on swap list */
If I were performance-collecting code, where would I live?  Of course! 
In the routines that actually perform the input-output to the swap
file(s).  So I went hunting again, and my feeble brain found some
likely-looking code in mm/page_io.c that appears to start and wait for
I/O operations involving swap. But where is the accounting in that
code?  I don't see any.  Now, I did try to trace through some of the
subroutines, but shuffling the accounting to routines that clearly had a
different intent didn't make sense.

How about higher level functions?  Well, I ran out of time in this
sitting, as I have to get ready for work.

I'm not trying to get on your case, Dick, for trying to provide an
answer, and I do appreciate your response. It's just that I don't see
how I can measure the amount of swap activity given the information
available.  I can *guess*, but then the whole idea of performance
*measurement* is to avoid guessing.

Anyone:  what have I missed?

Respectfully pondered,
Stephen Satchell

  reply	other threads:[~2003-12-09 14:46 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <>
2003-12-08 19:29 ` balance interrupts Len Brown
2003-12-08 20:00   ` Julien Oster
2003-12-08 20:02     ` Julien Oster
2003-12-08 21:46     ` Len Brown
2003-12-09  3:51   ` Bob
2003-12-09  5:11     ` Stephen Satchell
2003-12-09 13:19     ` Swap performance statistics in 2.6 -- which /proc file has it? Stephen Satchell
2003-12-09 13:56       ` Richard B. Johnson
2003-12-09 14:46         ` Stephen Satchell [this message]
2003-12-09 15:25           ` Richard B. Johnson
2003-12-09 19:53             ` Dominik Kubla
2003-12-09 20:24               ` Richard B. Johnson
2003-12-10 10:18                 ` Dominik Kubla
2003-12-10  1:28               ` Stephen Satchell
2003-12-10 10:34                 ` Dominik Kubla
2003-12-10 13:06                   ` Answer to Swap performance statistics in 2.6 -- which /proc file has it Stephen Satchell

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:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \

* 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).