Bug 29574

Overlay hang:

0x01ff0710:      0x18800080: MI_BATCH_BUFFER_START
0x01ff0714:      0x029b8001:    dword 1
0x01ff0718:      0x02000004: MI_FLUSH
0x01ff071c:      0x00000000: MI_NOOP
0x01ff0720:      0x10800001: MI_STORE_DATA_INDEX
0x01ff0724:      0x00000080:    dword 1
0x01ff0728:      0x000007ff:    dword 2
0x01ff072c:      0x01000000: MI_USER_INTERRUPT
0x01ff0730:      0x08a00000: MI_OVERLAY_FLIP
0x01ff0734:      0x3716a001:    dword 1
0x01ff0738: HEAD 0x01810000: MI_WAIT_FOR_EVENT
0x01ff073c:      0x00000000: MI_NOOP
0x01ff0740:      0x10800001: MI_STORE_DATA_INDEX
0x01ff0744:      0x00000080:    dword 1
0x01ff0748:      0x00000800:    dword 2
0x01ff074c:      0x01000000: MI_USER_INTERRUPT

Feel brave? Care to try my overlay branch in http://cgit.freedesktop.org/~ickle/linux-2.6/log/?h=overlay ?

Created attachment 37857 [details]
Tarball with i915 stat (debugfs), dmesg, Xorg.log

Here is the result of braveness:
linus's tree with anholt/drm-intel-next and ickle/overlay applied (plus a few more things that fix bugs in other areas - all as of 2-3 hours ago)

This time no crash on mplayer -vo gl (but no video either). GPU hangs at -vo xv time with some new messages in dmesg:
[drm:intel_calculate_wm] *ERROR* Insufficient FIFO for plane, expect flickering: entries required = 70, available = 42.

The ISR reports the Overlay Plane Flip Pending, which implies that no subsequent VBLANK event happened. Are we enabling the overlay on the wrong plane? [The hang occurs on OVERLAY_ON.] Daniel is also reporting misbehaviour with the forced Pipe A quirk.

Chris' overlay branch works okey for me on ThinkPad X40.

Rebasing the overlay branch onto drm-intel.git does not work so well. Xv doesn't seem to work at all; I get X11 error: BadAlloc (insufficient resources for operation) and the kernel locked up on resume at least once.

Plain overlay branch seems more stable and gives working video.

Dan likes overlay bugs.

Same problem on laptop HP Compaq nx9020.

00:02.0 VGA compatible controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)

Arch Linux i686
kernel26 2.6.35.4-1
xf86-video-intel 2.12.0-1
libdrm 2.4.21-2
xorg-server 1.8.1.902-1
mesa 7.8.2-1

Lines from Xorg.0.log after run mplayer with xv output:

[  3227.674] (EE) intel(0): Detected a hung GPU, disabling acceleration.

After run mplayer with gl output:

[  3251.158] 
Backtrace:
[  3251.159] 0: /usr/bin/X (xorg_backtrace+0x3b) [0x80df05b]
[  3251.159] 1: /usr/bin/X (0x8048000+0x63da5) [0x80abda5]
[  3251.159] 2: (vdso) (__kernel_rt_sigreturn+0x0) [0xb77a440c]
[  3251.159] 3: /usr/lib/xorg/modules/extensions/libdri2.so (0xb7398000+0x31ef) [0xb739b1ef]
[  3251.159] 4: /usr/bin/X (0x8048000+0x2170f) [0x806970f]
[  3251.159] 5: /usr/bin/X (0x8048000+0x1a4b5) [0x80624b5]
[  3251.159] 6: /lib/libc.so.6 (__libc_start_main+0xe6) [0xb74d9c76]
[  3251.159] 7: /usr/bin/X (0x8048000+0x1a0a1) [0x80620a1]
[  3251.159] Segmentation fault at address (nil)
[  3251.159] 
Fatal server error:
[  3251.159] Caught signal 11 (Segmentation fault). Server aborting
[  3251.159]

Sorry, I can't try Chris' overlay branch because I don't have necessary skill.
I just want to confirm a bug.

Bruno, can you please give a status update for this bug? If it still happens, please report current versions of your graphics stack.

Thx, Daniel

(In reply to comment #7)
> Bruno, can you please give a status update for this bug? If it still happens,
> please report current versions of your graphics stack.

Seems either kernel update or libdrm update fixed it as it does not happen anymore (be it mplayer -vo gl or mplayer -vo xv) and I get video as expected.

Though mplayer -vo gl causes a line in Xorg.log:
(EE) DoSwapInterval: cx = (nil), GLX screen = (nil)



My current stack is:
 - kernel 2.6.36-rc7-00125-gb88d3c2
 - x11-drivers/xf86-video-intel-2.12.0
 - media-libs/mesa-7.8.2
 - x11-libs/libdrm-2.4.22
 - x11-base/xorg-server-1.7.7-r1

Please reopen in case it shows up again.