lkml.org 
[lkml]   [2001]   [Jul]   [13]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] VM statistics code


On Sat, 14 Jul 2001, Chris Wedgwood wrote:

> On Fri, Jul 13, 2001 at 07:08:23PM -0300, Marcelo Tosatti wrote:
>
> With this data we are able to know more about what is really going
> on in the VM.
>
> +vm_pglaunder: nr of page_launder() calls
> +vm_pglaunder_write: nr of times page_launder() started writting out data to free
> +memory.
> +vm_refill_inactive_scan: nr of refill_inactive_scan() calls
> +vm_alloc_resched: nr of reschedule's in __alloc_pages() due to a memory shortage.
> +vm_kswapd_wakeup: nr of kswapd wakeup's
> +vm_kreclaimd_wakeup: nr of kreclaimd wakeup's
> +vm_kflushd_wakeup: nr of kflushd wakeup's
> +
> +Per-zone statistics:
> +free shortage: per-zone free shortage
> +inactive shortage: per-zone inactive shortage
> +vm_launder_pgscan: number of pages scanned by page_launder
> +vm_pgclean: number of pages cleaned (moved to the inactive clean list) by page_launder
> +vm_pgskiplocked: number of locked pages skipped by page_launder
> +vm_pgskipdirty: number of dirty pages skipped by page_launder
> +vm_pglaundered: laundered pages by page_launder
> +vm_pgreact: pages reactivated in page_launder
> +vm_pgrescue: pages rescue in reclaim_page
> +vm_pgagescan: pages scanned by refill_inactive_scan()
> +vm_pgagedown: pages aged down by refill_inative_scan()
> +vm_pgageup: pages aged up by refill_inactive_scan()/try_to_swap_out()
> +vm_pgdeact: deactivated pages by deactivate_page
> +vm_pgdeactfail_age: nr of deactivation failures on refill_inactive_scan()
> +due to >0 age
> +vm_pgdeactfail_ref: nr of deactivation failures on refill_inactive_scan()
> +due to zero aged pages with more users than the pagecache
> +vm_reclaimfail: failures of reclaim_page() (no freeable clean pages in the inactive
> +clean list for this zone)
> +vm_ptescan: nr of present ptes scanned by swap_out()
> +vm_pteunmap: nr of present ptes unmapped by swap_out()
>
> Some of these seem very specialised and low-level.

Yes they are.

> Perhaps another /proc entry might be a better idea along with either a
> different vmstat,

Maybe. Personally I don't really care about the way we are doing this, as
long as I can get the information. I can add /proc/vmstat easily if
needed...

Linus, how do you want this thing done to be accepted in the stock tree?

> or changes to vmstat such that it will behave as always if not given a
> magic command line switch and/or this new /proc entry isn't present?

Well, I don't want to include this stuff on the stock vmstat code right
now. I've done an ugly hack in vmstat.c to get the thing to work and thats
it.

I will keep an updated version of the "hacked" vmstat.c on my homepage
until I find time/balls to code something which does not stink.

> Right now, much of what comes form vmstat vaguley resembles other OSs,
> with the above, it will be very different.

Indeed.

-
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/

\
 
 \ /
  Last update: 2005-03-22 12:57    [W:0.053 / U:2.840 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site