After porting strace to 2.6.12 I hacked it some more to sync the log
file to disk after trace entry (instead of just fflush as it was doing)
and I now get a complete system call trace (albeit after a much, much
longer wait!).
Restoretextmode appears to crash in a call to mmap as shown by the last
few lines of the trace below. (The full trace is attached).
open("/proc/bus/pci/01/00.0", O_RDONLY) = 4
lseek(4, 60, SEEK_SET) = 60
read(4, "\v\1\5\1", 4) = 4
close(4) = 0
mmap2(NULL, 8388608, PROT_READ|PROT_WRITE, MAP_SHARED, 3, 0xee000) 0xb77d7000
write(2, "Using nvidia driver, 65536KB, Ty"..., 53) = 53
open("/sys/bus/pci/devices/0000:01:00.0/resource", O_RDONLY) = 4
fstat64(4, {st_mode=S_IFREG|0444, st_size=4096, ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1,
0) = 0xb77d6000
read(4, "0x00000000ee000000 0x00000000eef"..., 4096) = 399
open("/sys/bus/pci/devices/0000:01:00.0/resource1", O_RDWR) = 5
close(4) = 0
munmap(0xb77d6000, 4096) = 0
mmap2(NULL, 65536, PROT_READ|PROT_WRITE, MAP_SHARED, 3, 0xa0) 0xb77c7000
mmap2(NULL, 67108864, PROT_READ|PROT_WRITE, MAP_SHARED, 5, 0
Regards,
Dave Bonnell
Corporate Architect
BMC Software
-----Original Message-----
From: Ian Pratt [mailto:m+Ian.Pratt@cl.cam.ac.uk]
Sent: Wednesday, 14 September 2005 1:04 PM
To: Bonnell, David; xen-devel@lists.xensource.com
Cc: Keir Fraser; ian.pratt@cl.cam.ac.uk; ian.pratt@cl.cam.ac.uk
Subject: RE: [UPDATE] Xen crashes
> The complete Xen boot log and strace are attached however I''m
> not sure the strace log is complete (in fact, I''m almost
> certain it is not!) as the latest entries may not have been
> flushed to disk before the crash and hence lost. (Unless Xen
> is booted with "noreboot" the strace log file is not even
created!).
>
> The only output from restoretextmode before the crash is:
>
> Using nvidia driver, 65536KB, Type: GeForce (NV 11).
>
> I''ll pull down the source and compile restoretextmode/svgalib
> with debug symbols so I can step through it in gdb until the
> point of the "bang".
It looks like it''s going at it through /proc/pci raher than /dev/mem. I
seem to recall we fixed the memmap functions od /dev/mem a couple of
weeks back, but I wander if read/write have been attended to?
Having the fragment of source that''s causing the crash would be useful
to target the search.
Ian
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel