Re: [2.6.18-rt6] BUG / typo

[Guennadi Liakhovetski]

Hi again

Just a bit more detail to the way the bug occurs: you just need

CONFIG_LATENCY_TIMING=y

and in my case I had 2 consoles: tty1 and ttyS0. tty1 is a framebuffer. I 
don't know where irqs get disabled (the original BUG was

BUG: scheduling with irqs disabled: posix_cpu_timer/0x00000001/2
caller is rt_spin_lock_slowlock+0xd8/0x1c8

but is it at all ok to schedule in kmalloc(GFP_ATOMIC)? Which is exactly 
what happens here as kmalloc tries to acquire the per-cpu "spinlock" / 
mutex slab_irq_locks and is forced into the slow path.

The bug often triggers when I reset /proc/sys/kernel/preempt_max_latency 
with 0 and if I have high enough console logging level in 
/proc/sys/kernel/printk.

Thanks
Guennadi
---------------------------------
Guennadi Liakhovetski, Ph.D.
DSA Daten- und Systemtechnik GmbH
Pascalstr. 28
D-52076 Aachen
Germany
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/