From: Andi Kleen <andi@firstfloor.org> To: Ingo Molnar <mingo@elte.hu> Cc: "Wu Fengguang" <fengguang.wu@intel.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frédéric Weisbecker" <fweisbec@gmail.com>, "Larry Woodman" <lwoodman@redhat.com>, "Peter Zijlstra" <a.p.zijlstra@chello.nl>, "Pekka Enberg" <penberg@cs.helsinki.fi>, "Eduard - Gabriel Munteanu" <eduard.munteanu@linux360.ro>, "Andrew Morton" <akpm@linux-foundation.org>, LKML <linux-kernel@vger.kernel.org>, "KOSAKI Motohiro" <kosaki.motohiro@jp.fujitsu.com>, "Andi Kleen" <andi@firstfloor.org>, "Matt Mackall" <mpm@selenic.com>, "Alexey Dobriyan" <adobriyan@gmail.com>, "linux-mm@kvack.org" <linux-mm@kvack.org> Subject: Re: [PATCH 5/5] proc: export more page flags in /proc/kpageflags Date: Tue, 28 Apr 2009 09:40:31 +0200 [thread overview] Message-ID: <20090428074031.GK27382@one.firstfloor.org> (raw) In-Reply-To: <20090428065507.GA2024@elte.hu> > I think i have to NAK this kind of ad-hoc instrumentation of kernel > internals and statistics until we clear up why such instrumentation I think because it has zero fast path overhead and can be used any time without enabling anything special. > measures are being accepted into the MM while other, more dynamic While the dynamic instrumentation you're proposing has non zero fast path overhead, especially if you consider the CPU time needed for the backend computation in user space too. And it requires explicit tracing first and some backend that counts the events and maintains a shadow data structure covering all of mem_map again. So it's clear your alternative will be much more costly, plus have additional drawbacks (needs enabling first, cannot take a snapshot at arbitary time) Also dynamic tracing tends to have trouble with full memory observation. I experimented with systemtap tracing for my memory usage paper I did a couple of years ago, but ended up with integrated counters (similar to those) because it was impossible to do proper accounting for the pages set up in early boot with the standard tracers. I suspect both have their uses (that's indeed some things that can only be done with dynamic tracing), but they're clearly complementary and the static facility seems useful enough on its own. I think Fengguang is demonstrating that clearly by the great improvements he's doing for readahead which are enabled by these patches. -Andi
WARNING: multiple messages have this Message-ID (diff)
From: Andi Kleen <andi@firstfloor.org> To: Ingo Molnar <mingo@elte.hu> Cc: "Wu Fengguang" <fengguang.wu@intel.com>, "Steven Rostedt" <rostedt@goodmis.org>, "Frédéric Weisbecker" <fweisbec@gmail.com>, "Larry Woodman" <lwoodman@redhat.com>, "Peter Zijlstra" <a.p.zijlstra@chello.nl>, "Pekka Enberg" <penberg@cs.helsinki.fi>, "Eduard - Gabriel Munteanu" <eduard.munteanu@linux360.ro>, "Andrew Morton" <akpm@linux-foundation.org>, LKML <linux-kernel@vger.kernel.org>, "KOSAKI Motohiro" <kosaki.motohiro@jp.fujitsu.com>, "Andi Kleen" <andi@firstfloor.org>, "Matt Mackall" <mpm@selenic.com>, "Alexey Dobriyan" <adobriyan@gmail.com>, "linux-mm@kvack.org" <linux-mm@kvack.org> Subject: Re: [PATCH 5/5] proc: export more page flags in /proc/kpageflags Date: Tue, 28 Apr 2009 09:40:31 +0200 [thread overview] Message-ID: <20090428074031.GK27382@one.firstfloor.org> (raw) In-Reply-To: <20090428065507.GA2024@elte.hu> > I think i have to NAK this kind of ad-hoc instrumentation of kernel > internals and statistics until we clear up why such instrumentation I think because it has zero fast path overhead and can be used any time without enabling anything special. > measures are being accepted into the MM while other, more dynamic While the dynamic instrumentation you're proposing has non zero fast path overhead, especially if you consider the CPU time needed for the backend computation in user space too. And it requires explicit tracing first and some backend that counts the events and maintains a shadow data structure covering all of mem_map again. So it's clear your alternative will be much more costly, plus have additional drawbacks (needs enabling first, cannot take a snapshot at arbitary time) Also dynamic tracing tends to have trouble with full memory observation. I experimented with systemtap tracing for my memory usage paper I did a couple of years ago, but ended up with integrated counters (similar to those) because it was impossible to do proper accounting for the pages set up in early boot with the standard tracers. I suspect both have their uses (that's indeed some things that can only be done with dynamic tracing), but they're clearly complementary and the static facility seems useful enough on its own. I think Fengguang is demonstrating that clearly by the great improvements he's doing for readahead which are enabled by these patches. -Andi -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2009-04-28 7:36 UTC|newest] Thread overview: 137+ messages / expand[flat|nested] mbox.gz Atom feed top 2009-04-28 1:09 [PATCH 0/5] proc: export more page flags in /proc/kpageflags (take 4) Wu Fengguang 2009-04-28 1:09 ` Wu Fengguang 2009-04-28 1:09 ` [PATCH 1/5] pagemap: document clarifications Wu Fengguang 2009-04-28 1:09 ` Wu Fengguang 2009-04-28 7:11 ` Tommi Rantala 2009-04-28 7:11 ` Tommi Rantala 2009-04-28 1:09 ` [PATCH 2/5] pagemap: documentation 9 more exported page flags Wu Fengguang 2009-04-28 1:09 ` Wu Fengguang 2009-04-28 1:09 ` [PATCH 3/5] mm: introduce PageHuge() for testing huge/gigantic pages Wu Fengguang 2009-04-28 1:09 ` Wu Fengguang 2009-04-28 1:09 ` [PATCH 4/5] proc: kpagecount/kpageflags code cleanup Wu Fengguang 2009-04-28 1:09 ` Wu Fengguang 2009-04-28 1:09 ` [PATCH 5/5] proc: export more page flags in /proc/kpageflags Wu Fengguang 2009-04-28 1:09 ` Wu Fengguang 2009-04-28 6:55 ` Ingo Molnar 2009-04-28 6:55 ` Ingo Molnar 2009-04-28 7:40 ` Andi Kleen [this message] 2009-04-28 7:40 ` Andi Kleen 2009-04-28 9:04 ` Pekka Enberg 2009-04-28 9:04 ` Pekka Enberg 2009-04-28 9:10 ` Andi Kleen 2009-04-28 9:10 ` Andi Kleen 2009-04-28 9:15 ` Pekka Enberg 2009-04-28 9:15 ` Pekka Enberg 2009-04-28 9:15 ` Ingo Molnar 2009-04-28 9:15 ` Ingo Molnar 2009-04-28 9:19 ` Pekka Enberg 2009-04-28 9:19 ` Pekka Enberg 2009-04-28 9:25 ` Pekka Enberg 2009-04-28 9:25 ` Pekka Enberg 2009-04-28 9:36 ` Wu Fengguang 2009-04-28 9:36 ` Wu Fengguang 2009-04-28 9:36 ` Ingo Molnar 2009-04-28 9:36 ` Ingo Molnar 2009-04-28 9:57 ` Pekka Enberg 2009-04-28 9:57 ` Pekka Enberg 2009-04-28 10:10 ` KOSAKI Motohiro 2009-04-28 10:10 ` KOSAKI Motohiro 2009-04-28 10:21 ` Pekka Enberg 2009-04-28 10:21 ` Pekka Enberg 2009-04-28 10:56 ` Ingo Molnar 2009-04-28 10:56 ` Ingo Molnar 2009-04-28 11:09 ` KOSAKI Motohiro 2009-04-28 11:09 ` KOSAKI Motohiro 2009-04-28 12:42 ` Ingo Molnar 2009-04-28 12:42 ` Ingo Molnar 2009-04-28 11:03 ` Ingo Molnar 2009-04-28 11:03 ` Ingo Molnar 2009-04-28 17:42 ` Matt Mackall 2009-04-28 17:42 ` Matt Mackall 2009-04-28 9:29 ` Ingo Molnar 2009-04-28 9:29 ` Ingo Molnar 2009-04-28 9:34 ` KOSAKI Motohiro 2009-04-28 9:34 ` KOSAKI Motohiro 2009-04-28 9:38 ` Ingo Molnar 2009-04-28 9:38 ` Ingo Molnar 2009-04-28 9:55 ` Wu Fengguang 2009-04-28 9:55 ` Wu Fengguang 2009-04-28 10:11 ` KOSAKI Motohiro 2009-04-28 10:11 ` KOSAKI Motohiro 2009-04-28 11:05 ` Ingo Molnar 2009-04-28 11:05 ` Ingo Molnar 2009-04-28 11:36 ` Wu Fengguang 2009-04-28 11:36 ` Wu Fengguang 2009-04-28 12:17 ` [rfc] object collection tracing (was: [PATCH 5/5] proc: export more page flags in /proc/kpageflags) Ingo Molnar 2009-04-28 12:17 ` Ingo Molnar 2009-04-28 13:31 ` Wu Fengguang 2009-04-28 13:31 ` Wu Fengguang 2009-05-12 13:01 ` Frederic Weisbecker 2009-05-12 13:01 ` Frederic Weisbecker 2009-05-17 13:36 ` Wu Fengguang 2009-05-17 13:55 ` Frederic Weisbecker 2009-05-17 13:55 ` Frederic Weisbecker 2009-05-17 14:12 ` Wu Fengguang 2009-05-17 14:12 ` Wu Fengguang 2009-05-18 11:44 ` KOSAKI Motohiro 2009-05-18 11:44 ` KOSAKI Motohiro 2009-05-18 11:47 ` Wu Fengguang 2009-05-18 11:47 ` Wu Fengguang 2009-04-28 10:18 ` [PATCH 5/5] proc: export more page flags in /proc/kpageflags Andi Kleen 2009-04-28 10:18 ` Andi Kleen 2009-04-28 8:33 ` Wu Fengguang 2009-04-28 8:33 ` Wu Fengguang 2009-04-28 9:24 ` Ingo Molnar 2009-04-28 9:24 ` Ingo Molnar 2009-04-28 18:11 ` Tony Luck 2009-04-28 18:11 ` Tony Luck 2009-04-28 18:34 ` Matt Mackall 2009-04-28 18:34 ` Matt Mackall 2009-04-28 20:47 ` Tony Luck 2009-04-28 20:47 ` Tony Luck 2009-04-28 20:54 ` Andi Kleen 2009-04-28 20:54 ` Andi Kleen 2009-04-28 20:59 ` Matt Mackall 2009-04-28 20:59 ` Matt Mackall 2009-04-28 21:17 ` Andrew Morton 2009-04-28 21:17 ` Andrew Morton 2009-04-28 21:49 ` Matt Mackall 2009-04-28 21:49 ` Matt Mackall 2009-04-29 0:02 ` Robin Holt 2009-04-29 0:02 ` Robin Holt 2009-04-28 17:49 ` Matt Mackall 2009-04-28 17:49 ` Matt Mackall 2009-04-29 8:05 ` Wu Fengguang 2009-04-29 8:05 ` Wu Fengguang 2009-04-29 19:13 ` Matt Mackall 2009-04-29 19:13 ` Matt Mackall 2009-04-30 1:00 ` Wu Fengguang 2009-04-30 1:00 ` Wu Fengguang 2009-04-28 21:32 ` Andrew Morton 2009-04-28 21:32 ` Andrew Morton 2009-04-28 22:46 ` Matt Mackall 2009-04-28 22:46 ` Matt Mackall 2009-04-28 23:02 ` Andrew Morton 2009-04-28 23:02 ` Andrew Morton 2009-04-28 23:31 ` Matt Mackall 2009-04-28 23:31 ` Matt Mackall 2009-04-28 23:42 ` Andrew Morton 2009-04-28 23:42 ` Andrew Morton 2009-04-28 23:55 ` Matt Mackall 2009-04-28 23:55 ` Matt Mackall 2009-04-29 3:33 ` Wu Fengguang 2009-04-29 3:33 ` Wu Fengguang 2009-04-29 2:38 ` Wu Fengguang 2009-04-29 2:38 ` Wu Fengguang 2009-04-29 2:55 ` Andrew Morton 2009-04-29 2:55 ` Andrew Morton 2009-04-29 3:48 ` Wu Fengguang 2009-04-29 3:48 ` Wu Fengguang 2009-04-29 5:09 ` Wu Fengguang 2009-04-29 5:09 ` Wu Fengguang 2009-04-29 4:41 ` Nathan Lynch 2009-04-29 4:41 ` Nathan Lynch 2009-04-29 4:41 ` Nathan Lynch 2009-04-29 4:50 ` Andrew Morton 2009-04-29 4:50 ` Andrew Morton 2009-04-29 4:50 ` Andrew Morton
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=20090428074031.GK27382@one.firstfloor.org \ --to=andi@firstfloor.org \ --cc=a.p.zijlstra@chello.nl \ --cc=adobriyan@gmail.com \ --cc=akpm@linux-foundation.org \ --cc=eduard.munteanu@linux360.ro \ --cc=fengguang.wu@intel.com \ --cc=fweisbec@gmail.com \ --cc=kosaki.motohiro@jp.fujitsu.com \ --cc=linux-kernel@vger.kernel.org \ --cc=linux-mm@kvack.org \ --cc=lwoodman@redhat.com \ --cc=mingo@elte.hu \ --cc=mpm@selenic.com \ --cc=penberg@cs.helsinki.fi \ --cc=rostedt@goodmis.org \ /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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.