* Understanding page statistics
@ 2014-08-13 21:31 Martin Ichilevici de Oliveira
2014-08-14 11:39 ` Steffen Vogel
0 siblings, 1 reply; 2+ messages in thread
From: Martin Ichilevici de Oliveira @ 2014-08-13 21:31 UTC (permalink / raw)
To: kernelnewbies
Hello,
I'm starting to study and learn the kernel, and I'm particulary
interested in the virtual memory mechanism and it's application in NUMA
architectures.
As far as I could understand, the kernel keeps track of some
information/statistics about each page, and I'd like to read them with a
kernel module - just to see what kind of information is available
(something more detailed than /proc/<pid>/maps).
In a LWN article [1], I read the following:
"The scheduler will periodically scan through each process's address
space, revoking all access permissions to the pages that are currently
resident in RAM. The next time the affected process tries to access that
memory, a page fault will result. The scheduler will trap that fault and
restore access to the page in question; it will also increment an access
counter in a per-process array indexed by the NUMA node number."
I've built a simple kernel module that, given a PID and virtual address,
retrieves the corresponding struct page. I'm now unsure now where to look
for such statistics and counters. Any hints on relevant data structures are
appreciated.
Thank you,
Martin
[1] : http://lwn.net/Articles/568870/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
Url : http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20140813/eb62c3ac/attachment.bin
^ permalink raw reply [flat|nested] 2+ messages in thread
* Understanding page statistics
2014-08-13 21:31 Understanding page statistics Martin Ichilevici de Oliveira
@ 2014-08-14 11:39 ` Steffen Vogel
0 siblings, 0 replies; 2+ messages in thread
From: Steffen Vogel @ 2014-08-14 11:39 UTC (permalink / raw)
To: kernelnewbies
Am Mittwoch, den 13.08.2014, 18:31 -0300 schrieb Martin Ichilevici de
Oliveira:
> "The scheduler will periodically scan through each process's address
> space, revoking all access permissions to the pages that are currently
> resident in RAM. The next time the affected process tries to access that
> memory, a page fault will result.
I don't know how often the flags are reset.
But isn't this a huge overhead? I don't see any advantages for this on
UMA machines.
> I've built a simple kernel module that, given a PID and virtual address,
> retrieves the corresponding struct page. I'm now unsure now where to look
> for such statistics and counters. Any hints on relevant data structures are
> appreciated.
Oh okay, If I'm correct 'struct page' represents a physical page frame
rather than virtual mapped page (see [2]).
The first thing which comes in my mind is LRU aka the kernels page frame
reclamation which is used for swapping unused pages to disk.
I'm mot really sure wheather the LRU lists are based on physical pages.
But I assume so.
You might want to have a look at /proc/pid/pagemap for detailed
information of the VA per process [1].
Theres also a /proc/kpageflags and /proc/kpagecount which contain
information over physical pages [1].
Steffen
[1] https://www.kernel.org/doc/Documentation/vm/pagemap.txt
[2] http://www.tldp.org/LDP/tlk/ds/ds.html
[3] http://linux-mm.org/LRU
[4] https://www.kernel.org/doc/gorman/html/understand/understand013.html
--
Steffen Vogel
Robensstra?e 69
52070 Aachen
Mail: post at steffenvogel.de
Mobil: +49 1575 7180927
Web: http://www.steffenvogel.de
Jabber: steffen.vogel at jabber.rwth-aachen.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part
Url : http://lists.kernelnewbies.org/pipermail/kernelnewbies/attachments/20140814/1d065013/attachment-0001.bin
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2014-08-14 11:39 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-08-13 21:31 Understanding page statistics Martin Ichilevici de Oliveira
2014-08-14 11:39 ` Steffen Vogel
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).