[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: os x kernel panics

Hi Phobos,

I am running tor-cvs here on the latest tiger as well as on panther... not having the issues you are experiencing..
tor was compiled with the bundled compiler and userland for xcode 1.5 on panther and xcode-2.0 on tiger...

a tor user

phobos@xxxxxxxxxx wrote:

I've been doing some digging into the loose correlation of tor and osx
kernel panics. So far, it only happens on 10.3.x; that I know of. If
anyone has panic.log's they'd like to send me, i'm trying to get a feel
for what's going on.

The bulk of the work happens at kernel panic when I can do a remote gdb
to the machine and figure out what the backtrace is telling me.

One interesting thing I noted was that on my 8 panics as recorded in
panic.log, the last 3 have the same memory addresses of the stack
pointer where the call was initiated (lr). In theory, the way apple
handles kernel modules and loading, this shouldn't happen.

Anyway, if you've experienced an osx kernel panic, could you send me:

1) your latest kernel panic info from the log file
2) which version of tor you are running
3) which install you did (fink, darwinports, package from the tor website, or source compile)

I believe this may be a dual bug, one in apple's kernel for allowing a
userland program to cause a panic, and two, something in tor or a
dependent lib creating the condition to cause the first bug.  As far as
I can loosely correlate, it appears to happen when a circuit times out
and a new one is created.  This may cause the 0x300 DAR panic, as you
try to dereference freed memory.  In theory, this should only crash
tor, not the kernel.