Re: Strange crash building "transmission"

[Tobias Nygren <tnn%NetBSD.org@localhost>]

On Mon, 09 Feb 2009 15:22:13 -0800
Scott Ellis <scotte%warped.com@localhost> wrote:

> Okay, this is weird, and I am wondering if anyone else has seen this.
> 
> I was trying to build 1.50b4 of Transmission 
> (http://www.transmissionbt.com/), using curl 7.19.3 on amd64, with 
> -current as of today using "-O3 -march=nocona" as my CFLAGS on a 
> filesystem mounted with "-o log".
> 
> Running (in the transmission directory) "./configure --enable-cli 
> --disable-nls" repeatedly yields:
> 
> fatal protection fault in supervisor mode
> trap type 4 code 0 rip ffffffff801c18da cs 8 rflags 10282 cr2 
> 7f7ffda011fc cpl 0 rsp ffff800052123ab0
> panic: trap
> Begin traceback...
> uvm_fault(0xffff800052070b88, 0x0, 1) -> e
> fatal page fault in supervisor mode
> trap type 6 code 0 rip ffffffff801562af cs 8 rflags 10246 cr2  0 cpl 0 
> rsp ffff800052123680
> panic: trap
> Faulted in mid-traceback; aborting...rebooting...
> 
> This happens just after "Appending 'F77' to libtool flags" in the 
> libevent configure script (called by transmission's configure script).
> 
> What the heck?  The system is otherwise quiescent (technically, there 
> was a Raidframe parity rebuild in the background on that particular 
> crash, but it happened when the array was fine too), and I can't get a 
> decent backtrace to see even where to look.
> 
> Does this sound like a kernel bug with -current (happens with multiple 
> kernels from random times in Feb)?  FS corruption (fsck says the 
> filesystems are clean)?  RAM issues (memcheck says all is well)?
> 
>       ScottE

Yeah, I see the exact same issue while attempting to build
net/Transmission from pkgsrc on current -i386. For a while there's
a /bin/sh spinning at 100% CPU inside the kernel (kill -9 doesn't
work). Then the kernel will usually panic within a minute or so.

-Tobias