Bug #188
closed
Added by swildner almost 20 years ago.
Updated over 19 years ago.
Description
Hi,
similar to an issue reported by Kevin Kane on users@:
One of my boxes freezes within 1-2 minutes after booting (HEAD as of
today), when mprime is running. mprime is some tool to search for large
prime numbers, which eats all idle time when it runs. If I kill mprime
directly after booting, the system seems to run fine.
Sascha
:Hi,
:
:similar to an issue reported by Kevin Kane on users@:
:
:One of my boxes freezes within 1-2 minutes after booting (HEAD as of
:today), when mprime is running. mprime is some tool to search for large
:prime numbers, which eats all idle time when it runs. If I kill mprime
:directly after booting, the system seems to run fine.
:
:Sascha
:
:--
:http://yoyodyne.ath.cx
Ok, there is clearly still a bug in there somewhere that I
need to track down.
Are you specifying a particular priority, i.e. with idprio or
something like that?
-Matt
Matthew Dillon
<dillon@backplane.com>
Matthew Dillon wrote:
:Hi,
:
:similar to an issue reported by Kevin Kane on users@:
:
:One of my boxes freezes within 1-2 minutes after booting (HEAD as of
:today), when mprime is running. mprime is some tool to search for large
:prime numbers, which eats all idle time when it runs. If I kill mprime
:directly after booting, the system seems to run fine.
:
:Sascha
:
:--
:http://yoyodyne.ath.cx
Ok, there is clearly still a bug in there somewhere that I
need to track down.
Are you specifying a particular priority, i.e. with idprio or
something like that?
Hmm, on a different box running an older DragonFly version, the output
of ps -axl looks like this for mprime:
UID PID PPID CPU PRI NI VSZ RSS WCHAN STAT TT TIME COMMAND
0 651 650 128 254 20 7052 1332 - RNL con- 67174:50.48
./mprime -d
Sascha
Ok, I'm close to tracking this one down. It's a critical section
mismatch somewhere. A critical section gets entered and never gets
released, the user process then runs in a critical section which as
you can imagine can lead to, oh, things like interrupts not working
any more, and infinite loops locking up the box as well.
I am adding INVARIANTS code to panic if the critical section count
becomes mismatched and hopefully I'll track down the current mismatch
soon.
-Matt
Also available in: Atom
PDF