This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
[Bug runtime/5194] IO problem on begin/end probes
- From: "fche at redhat dot com" <sourceware-bugzilla at sourceware dot org>
- To: systemtap at sources dot redhat dot com
- Date: 18 Oct 2007 15:19:00 -0000
- Subject: [Bug runtime/5194] IO problem on begin/end probes
- References: <20071018140048.5194.hunt@redhat.com>
- Reply-to: sourceware-bugzilla at sourceware dot org
------- Additional Comments From fche at redhat dot com 2007-10-18 15:18 -------
(In reply to comment #3)
> calling a sleeping function in a begin probe.
That remains invalid, even with the slight relaxing we did earlier.
> If the begin/end probe does sleep or get interrupted, it might
> not get the same cpu when it resumes,
Interrupts do not cause a reschedule or cpu reassingnment if
preemption was blocked (as it is).
> which means the output buffer will change without the
> previous buffer getting flushed and data will be lost.
If this is the limit of what can happen when a user schemes to call
a sleeping function (which can only happen in "-g" mode anyway), then
this problem is a minor one.
> IIRC, one of the reasons we enabled preemption was to handle situations where
> large arrays are dumped during end probes.
That was only a side-effect. We wanted to produce "long" reports, for "long"
meaning perhaps thousands of lines. It's not megabytes. With the status
quo, SMP machines may be able to keep up with even that.
--
http://sourceware.org/bugzilla/show_bug.cgi?id=5194
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.