X.Org is stuck in a loop on resume from S3

If I use fglrx-experimental-9 instead of radeon, X gets stuck when going to sleep. All I can see is a black screen with the backlight on. Just like before, Ctrl+Alt+Sysrq+K kills X and allows the system to go back to the login screen.

I could reproduce this radeon, but when I installed the kernel 12.10 released with (-17), it resumed just fine. So sounds like a regression in the kernel between -17..-22. Would need to double check with ftws and stress testing it a bit.

I tested 3.5.0-17.28 from Quantal and 3.5.0-18.29~precise1 from Precise-updates, the -17 kernel can suspend, but the -18 cannot.

During Ubuntu tag 3.5.0-17.28 and 3.5.0-18.29, there are two upstream stable releases 3.5.7 and 3.5.8.

3.5.7 can suspend (with fglrx), 3.5.8 cannot.

3.5.7 (with fglrx) resumes with the black screen. The led on the monitor is not lightened on.

With 3.5.0-22.34 kernel (with fglrx), there is another way to reproduce this. On the menu on the right top, choose "logout", the monitor goes blank and the lightdm login screen does not show.

>From `top` can see the Xorg process occupies 100% CPU.

>From the Xorg.0.log the last lines are:

[ 35.682] (II) fglrx(0): Shutdown CMMQS
[ 35.682] (II) fglrx(0): [uki] removed 1 reserved context for kernel
[ 35.682] (II) fglrx(0): [uki] unmapping 8192 bytes of SAREA 0x2000 at 0x7f981aec6000

To correct comment #7, I re-did the fwts s3 test with the 3.5.0-17.28 kernel, the 100% CPU consumption of the Xorg process is observed today. This usually happens in a few cycles of suspending and resuming.

I couldn't reproduce the bug with an older machine with BIOS instead of UEFI, did over 40 S/R cycles and it worked, also with my old card and the stock fglrx version from 12.10.

On some AMD graphic card, I can reproduce the issue even with the xserver-xorg-lts-precise and quantal backported kernel (3.5.x).

The 100% Xorg CPU consumption can also be reproduced by logging out from the desktop system (back to lightdm) besides suspending the system.

Both scenarios happens with high possibility (though not always, but usually in dozens of times).

BTW, the virtual terminal switching leads to blank screen.

 When the Xorg occupied 100% CPU, ctrl+c in gdb, the backtrace:

(gdb) bt
#0 0x00007fb22e17b633 in __pread_nocancel () from /lib/x86_64-linux-gnu/libpthread.so.0
#1 0x00007fb22e38dc1b in ?? () from /usr/lib/x86_64-linux-gnu/libpciaccess.so.0
#2 0x00007fb22b89a29f in amd_xs113_int10_x_inl () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#3 0x00007fb22b892dfb in ?? () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#4 0x00007fb22b886515 in X86EMU_exec () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#5 0x00007fb22b89b246 in amd_xs113_int10_xf86ExecX86int10 ()
   from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#6 0x00007fb22b3e52fd in xf86ExecX86int10 () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#7 0x00007fb229bdd3ed in VBESetVBEMode () from /usr/lib/xorg/modules/libvbe.so
#8 0x00007fb22b416ce3 in ?? () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#9 0x00007fb22b416b34 in atiddxVBESetConsoleMode () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#10 0x00007fb22b53508b in xdl_xs113_atiddxLeaveVT () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/drivers/fglrx_drv.so
#11 0x00007fb22eef6282 in ?? ()
#12 0x00007fb22bf5f30e in ?? () from /usr/lib/x86_64-linux-gnu/xorg/extra-modules/modules/extensions/libglx.so
#13 0x00007fb22eedbc27 in xf86Wakeup ()
#14 0x00007fb22ee9fb1b in WakeupHandler ()
#15 0x00007fb22eff8a06 in WaitForSomething ()
#16 0x00007fb22ee9b4f2 in ?? ()
#17 0x00007fb22ee8a15a in ?? ()
#18 0x00007fb22cdf276d in __libc_start_main () from /lib/x86_64-linux-gnu/libc.so.6
#19 0x00007fb22ee8a4b1 in _start ()

13.2 beta6 fixed this for me:

http://www2.ati.com/drivers/beta/amd-driver-installer-catalyst-13.2-beta6-linux-x86.x86_64.zip

I encountered the similar issue (Kengyu's comment#12) when testing one machine with AMD graphic (using fglrx 13.2 beta3 driver). The problem goes away after updating lightdm.

The verification of this Stable Release Update has completed successfully and the package has now been released to -updates. Subsequently, the Ubuntu Stable Release Updates Team is being unsubscribed and will not receive messages about this bug report. In the event that you encounter a regression using the package from -updates please report a new bug using ubuntu-bug and tag the bug report regression-update so we can easily find any regresssions.

This should really be fixed in fglrx-experimental-12. Please test the package and let me know.