Re: __DEATH__ of SMP

Linus Torvalds (torvalds@transmeta.com)
20 Apr 1997 17:46:30 GMT

In article <Pine.LNX.3.91.970420103436.2429A-100000@chaos.analogic.com>,
Richard B. Johnson <root@analogic.com> wrote:
>Version 2.1.35 still has a problem with the system halting with no
>evidence of any problem displayed in any log or on the console.
>
>The symptoms are:
>
>o Dead. No response to any external event like Numlock on the keyboard.
>o Dead. No response to network pings.
>o Dead. No evidence that it was a computer.

I get the feeling that you think that the system has died?

Could you test out the new pre-patch-2.1.36.gz stuff currently on
ftp.kernel.org (under linux/kernel/testing)? I'm not making a real
2.1.36 release because I want to actually boot this on my own machine
first at some point, but the pre-patch might be interesting to test out
anyway because it does fix a few potential problems that I glossed over
in 2.1.35..

This one has the new and oh-so-clever read-write locks, and they are
used to protect the task list. With basic 2.1.35 you could potentially
get a corrupted task list, although it was rather unlikely. That would
certainly have killed your machine.

(pre-2.1.36 also has the threaded signal code, and that's kind of
bleeding edge. But _fun_ bleeding edge.)

Linus