[lkml]   [2004]   [Oct]   [27]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[patch] Real-Time Preemption, -RT-2.6.9-mm1-V0.4

i have released the -V0.4 Real-Time Preemption patch, which can be
downloaded from:

this is a fixes-only release, but still experimental.

this release should fix more bugs of the 'slowdown' and 'interactivity
problems' variety.

To debug the wakeup anomalies reported i've implemented a new variant of
the latency tracer, which now traces 'wakeup latencies' too - i.e. it
measures and traces maximum delays observed from the point of wakeup to
the point of the really starting to execute. Only the highest-priority
runnable task in the system is traced at a time, but this should be more
than enough to find the high latency scheduling paths.

This new tracing mode can be enabled by compiling a LATENCY_TRACE kernel
as usual and doing:

echo 4 > /proc/sys/kernel/trace_enabled

then to start tracing, reset the current max latency value via e.g.:

echo 10 > /proc/sys/kernel/preempt_max_latency

then the kernel should signal wakeup latency events in the syslog:

(sshd/3093/CPU#0): new 18 us maximum-latency wakeup.
(sshd/3093/CPU#0): new 19 us maximum-latency wakeup.
(hackbench/3818/CPU#0): new 20 us maximum-latency wakeup.
(hackbench/3762/CPU#0): new 21 us maximum-latency wakeup.
(hackbench/3814/CPU#0): new 22 us maximum-latency wakeup.
(ksoftirqd/0/3/CPU#0): new 35 us maximum-latency wakeup.

the latency trace of the last (and highest) event can always be found in
/proc/latency_trace, as usual. Note that the trace output is a bit
different in the wakeup-tracing case.

NOTE: the tracer works on SMP too, but since on SMP tasks can switch
from one CPU to another a given trace can be less useful if the delay
happened on another CPU.

using this wakeup tracer i found and fixed a couple of 'missed
preemption check' bugs - all introduced by PREEMPT_REALTIME in the -U/-V
timeframe. So if you had latency/interactivity problems please re-check

the wakeup tracer is nice in the sense of that it traces actual, realy.

Changes since -V0.3.2:

- fixed the rtc_lock related crash reported by K.R. Foley and Robert

- fixed missing preemption checks in rwsem-generic.c

- fixed missing preemption check in schedule_tail() [==new task wakeup]

- implemented wakeup-latency tracer

to create a -V0.4 tree from scratch, the patching order is:

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 14:07    [W:1.032 / U:7.048 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site