Re: SMP stuck on TLB IPI (2.2.7)

[George]

On Thu, 13 May 1999, Thomas van Gulick wrote:

>My systems locks up with the message
>
>stuck on TLB IPI wait (CPU#0)
>
>after about an hour or 2 when runngin kernel 2.2.7 .. I have the same
>problem with other 2.2.x kernels and 2.1.132 seems to lock up too, but a lot
>less frequent (2.2.x can't run longer than a day, 2.1.132 can hold out as
>long as a couple of days).

2.2.0-pre6 I cannot reproduce it with my disk access test.  2.2.0-pre7 I
can easily do it with.  Do you have a way to reproduce it reliably?  It'd
be handy to pin it down further.

>It seems to depend on the network load.. The heavier I transfer files over
>the network, the quicker the system locks up.

Odd, I can do it without a network.  All it takes for me is a
multi-threaded program that thrashes into swap.  It might not even have to
be multi-threaded but GIMP v1.1.5 with MP works quite reproducibly for me.

Since I get 'hdc: lost interrupt' messages also, I suppose it's probably a
generic IRQ thing that would explain why network traffic does it for you
and hard drive access does it for me.

-George Greer