This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: [PATCH] Linux Kernel Markers 0.13 for 2.6.17
- From: Mathieu Desnoyers <compudj at krystal dot dyndns dot org>
- To: "Frank Ch. Eigler" <fche at redhat dot com>
- Cc: Jeremy Fitzhardinge <jeremy at goop dot org>, Martin Bligh <mbligh at google dot com>, Masami Hiramatsu <masami dot hiramatsu dot pt at hitachi dot com>, prasanna at in dot ibm dot com, Andrew Morton <akpm at osdl dot org>, Ingo Molnar <mingo at elte dot hu>, Paul Mundt <lethal at linux-sh dot org>, linux-kernel <linux-kernel at vger dot kernel dot org>, Jes Sorensen <jes at sgi dot com>, Tom Zanussi <zanussi at us dot ibm dot com>, Richard J Moore <richardj_moore at uk dot ibm dot com>, Michel Dagenais <michel dot dagenais at polymtl dot ca>, Christoph Hellwig <hch at infradead dot org>, Greg Kroah-Hartman <gregkh at suse dot de>, Thomas Gleixner <tglx at linutronix dot de>, William Cohen <wcohen at redhat dot com>, ltt-dev at shafik dot org, systemtap at sources dot redhat dot com, Alan Cox <alan at lxorguk dot ukuu dot org dot uk>, Karim Yaghmour <karim at opersys dot com>, Pavel Machek <pavel at suse dot cz>, Joe Perches <joe at perches dot com>, "Randy.Dunlap" <rdunlap at xenotime dot net>, "Jose R. Santos" <jrs at us dot ibm dot com>
- Date: Tue, 26 Sep 2006 16:05:14 -0400
- Subject: Re: [PATCH] Linux Kernel Markers 0.13 for 2.6.17
- References: <45187146.8040302@goop.org> <20060926002551.GA18276@Krystal> <20060926004535.GA2978@Krystal> <45187C0E.1080601@goop.org> <20060926025924.GA27366@Krystal> <4518B4A0.6070509@goop.org> <20060926180414.GA10497@Krystal> <4519781D.9040503@goop.org> <20060926190849.GA2280@Krystal> <y0mhcyue7ch.fsf@ton.toronto.redhat.com>
* Frank Ch. Eigler (fche@redhat.com) wrote:
> Mathieu Desnoyers <compudj@krystal.dyndns.org> writes:
>
> > [...]
> > > Yep, that looks reasonable. Though you could just directly test a
> > > per-marker enable flag, rather than using "condition"...
> > [...]
> > I am not sure I understand your suggestion correctly.. do you mean having
> > a per-marker flag that would be loaded and tested at every marker site ?
>
> I gather that one reason for working so hard with the inline assembly
> is a race condition problem with the plain STAP_MARK style of marker
> disconnection:
>
> if (pointer) (*pointer)(args ...);
>
> Granted, but this problem could almost certainly be dealt with simpler
> than that. How about a compxchg or other atomic-fetch of the static
> pointer with a local variable? That should solve the worry of an
> (*NULL) call.
>
I don't really see how cmpxchg might be needed here.
Atomic fetch of a static variable is how I will do it in my next version for the
non optimized case :
volatile static var = 0;
if(var) {
preempt disable
call
preempt_enable_no_resched
}
But, still, in the optimized case, the if(var) will depend on an immediate
value, therefore saving the memory read.
> If we then become concerned with a valid pointer become obsolete (the
> probe handler function wanting to unload), we might be able to use
> some RCU-type deferral mechanism and/or preempt controls to ensure
> that this does not happen.
>
This is exactly why the preemption is disabled around the call. However, RCU
must always _see_ a coherent version of the structure in memory.
Calling an empty function, disabling preemption around the call and calling
synchronize_sched() before deleting the removed function looks very much like
a RCU-style protection (actually, that's what it is).
Mathieu
OpenPGP public key: http://krystal.dyndns.org:8080/key/compudj.gpg
Key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68