On Sun, Aug 24, 2008 at 11:50:52AM -0400, Robert Noland wrote: > On Sun, 2008-08-24 at 14:03 +0300, Kostik Belousov wrote: > > Unfortunately, I still get a hang with r300 in drm_lastclose(). > > I don't understand this. You are the only one that I'm hearing this > from now. Our open/close routines are an ugly maze, but the last time I > had radeon hardware it worked great. I had tests with this patch on a > 9250 and some newer, higher end radeons also. Can you get a trace or > see if it is hung on a lock? > > I don't currently have any radeon hardware, so I have to work blind on > it. > > robert.
I believe I already posted the backtrace of the X server whan this happen
(see below). The thing I cannot explain is the mouse pointer that still
follows the mouse. Everything else in X is frozen; system as a whole
works, X server eats 100% of the CPU in the loop that is easily
deducable from the trace.
I get this almost deterministically when I run either firefox or xdvi.
Probably, the hang could be reproduced with other apps too, but I
use xterm, xchat and xpdf only. My graphic card is r300 (PCI:*(1:0:0)
ATI Technologies Inc R420 JI [Radeon X800PRO] rev 0). The same problem
was observed on r200.
For some time, I blamed rev. 1.7 of drm_drv.c, but it appears that
reverting it does not make any change.
Would be happy to help get this resolved.
#4 0xc04c9bb9 in _sleep (ident=0xe7c91840, lock=0xc54a98d8, priority=84,
wmesg=0xc550ed1a "rdnrel", timo=1)
at /usr/bsd/src/sys/kern/kern_synch.c:235
#5 0xc54fecfe in radeon_do_release () from /boot/kernel/radeon.ko
#6 0xc551f0c5 in drm_lastclose () from /boot/kernel/drm.ko
#7 0xc551ff3b in drm_close () from /boot/kernel/drm.ko
#8 0xc0487fb0 in giant_close (dev=0xc5527a00, fflag=67, devtype=8192,
td=0xc5871880) at /usr/bsd/src/sys/kern/kern_conf.c:371
#9 0xc046ad2c in devfs_close (ap=0xe7c91940)
at /usr/bsd/src/sys/fs/devfs/devfs_vnops.c:372
#10 0xc06eddc2 in VOP_CLOSE_APV (vop=0xc0731380, a=0xe7c91940)
at vnode_if.c:424
pgpNfzb1geaZg.pgp
Description: PGP signature
