From: Paul Stewart <stewart@pa...> - 2005-09-18 00:18:32
On 2005-09-17 12:05, John Byrne wrote:
> I'm very curious to know how you got the stack information from the =20
Actually, my mailing was mostly about how I could avoid getting such =20
stack information from the kernel. :-) Those messages were being sent =20
gratuitously (and continuously) to /var/log/messages for every beacon =20
frame the Atheros card sent out, so my machine was constantly under =20
disk activity. I didn't invoke any special kernel back doors -- this =20
is just the a standard behavior of the Fedora 2.6.12 kernels.
It was triggered by the statement:
in the first statement in the body of local_bh_enable(). The later =20
kernels are a lot smarter about a) complaining about things that aren't =20
right, but not necessarily fatal, as in this case and b) being able to =20
convey in a human readable manner the stack trace leading up to the =20
error. I remember a time when I would have had to load a debug version =20
of my module and then through various gdb machinations line up the hex =20
values in the stack dump with /proc/ksyms and the symbol tables of the =20
various modules of interest. That form of gdb mojo has gotten easier =20
with time as well, but a stack trace (from start_kernel!) is nicer.
Of course, kgdb has existed for quite some time, but a setup of that =20
sort has never been handy when I've needed it.
Get latest updates about Open Source Projects, Conferences and News.