Re: kmsgdump not SMP-safe ?

[=?iso-8859-1?q?willy=20tarreau?=]

> Out of 5 panic() tries, kmsgdump was able to do
> something only once.
> The four other times, kmsgdump either:
>    - Printed "disabling sumetric I/O mode. Last
> chance for SysRQ" and
>      then got stuck.
> or - Caused an oops (available).
> 
> This was on a two-CPU box.
> 
> Is kmsgdump SMP-safe ?

It should try to, because in machine_dump(), the
other CPUs are stopped, IIRC, although I'm not
really sure about this since it's been a long time
since I last looked at the code.

Indeed, there are many cases where the kernel is under
weird condition, and the oops isn't catchable.

Even with a UP kernel I often get stuck. The reason
is mainly because kmsgdump still has lots of work to
do after the oops and sometimes this is not completely
possible. I began to write a new version in which
the system would be pre-initialized so that the CPU
would just have to reset to enter kmsgdump (even a
triple fault would work). But I don't have time to
work on it, even if I really would like to finish it.

> Phil.

Willy

___________________________________________________________
Do You Yahoo!?
Achetez, vendez! À votre prix! Sur http://encheres.yahoo.fr