Closed Bug 862397 Opened 8 years ago Closed 8 years ago

genlock_create_lock failure causes b2g process to crash

Categories

(Firefox OS Graveyard :: General, defect)

ARM
Gonk (Firefox OS)
defect
Not set
critical

Tracking

(blocking-b2g:-)

RESOLVED FIXED
blocking-b2g -

People

(Reporter: sotaro, Assigned: jhford)

References

()

Details

(Keywords: crash, Whiteboard: [b2g-crash])

Attachments

(1 file)

+++ This bug was initially created as a clone of Bug #862097 +++
Bug 862097 is reported as abort in content process. Though it is actually caused by a crash in b2g process. When a gralloc buffer allocation in b2g process is failed by genlock_create_lock failure, b2g process crash.
logcat before the crash from attachment 737869 [details].

-----------------------------------
04-16 08:46:36.044   109   237 E libgenlock: genlock_create_lock: open genlock device failed (err=Too many open files)
04-16 08:46:36.044   109   237 E msm7627a.gralloc: alloc_impl: genlock_create_lock failed
04-16 08:46:36.044   109   237 E memalloc: getAllocator: Invalid flags passed: 0x0
Nominate to tef. The bug blocks tef+ bug(Bug 862097).
blocking-b2g: --- → tef?
Flags: needinfo?(mvines)
Stack trace looks very much like the bug was fixed by https://www.codeaurora.org/gitweb/quic/b2g/?p=build.git;a=commitdiff;h=cee98d7cbfd59c1c4b4379c4b094aebc0d601c82

I suggest retesting with this patch applied!
Flags: needinfo?(mvines)
(In reply to Michael Vines [:m1] [:evilmachines] from comment #5)
> Stack trace looks very much like the bug was fixed by
> https://www.codeaurora.org/gitweb/quic/b2g/?p=build.git;a=commitdiff;
> h=cee98d7cbfd59c1c4b4379c4b094aebc0d601c82
> 
> I suggest retesting with this patch applied!

Thanks, it works! It needs to update the related source. v1.01 and v1-train do not use recent source. need to update.
Ugh, that's unfortunate.  Moz builds really need to track the vendor tree more closely, especially for v1.0.1 and v1-train.  Happy to discuss my POV on this tomorrow if anybody cares (oh prolly not ;) )
Yeah, I also think so. I did not know about it until yesterday.
This bug itself does not affect our partners, but it can affect our testing on unagi devices. Given that, we're not treating this as a blocker.

To fix this ourselves we just need a set of patches applied to our unagi builds, and mwu will provide the list of said patches.
Assignee: nobody → mwu
blocking-b2g: tef? → -
From https://www.codeaurora.org/gitweb/quic/lf/?p=build.git;a=summary , we'll need:

patch/all/hardware/qcom/display/remap-copybit-formats.patch
patch/ics/system/vold/Remove-g-rx-from-sdcard.patch
patch/ics_strawberry/external/bluetooth/bluez/0001-Remove-OPP-L2cap-SDP-record.patch
patch/ics_strawberry/frameworks/base/OMXCodec-stop-now-waits-for-all-buffers-to-return.patch
patch/ics_strawberry/frameworks/base/surfaceflinger_remove_layer_alpha.patch
patch/ics_strawberry/hardware/libhardware/hwc_header_remove_layer_alpha.patch
patch/ics_strawberry/hardware/msm7k/OMX_Deinit-symbol-fix.patch
patch/ics_strawberry/hardware/qcom/display/display-add-fillcolor.patch
patch/ics_strawberry/hardware/qcom/display/gpu-free-when-alloc-fails.patch
patch/ics_strawberry/hardware/qcom/display/hwc_impl_remove_layer_alpha.patch
patch/ics_strawberry/hardware/qcom/display/no-rgb565-dither.patch
patch/ics_strawberry/hardware/qcom/display/preserve-framebuffer-contents-on-initialization.patch
patch/ics_strawberry/hardware/qcom/media/OMX_Deinit-symbol-fix.patch
patch/ics_strawberry/kernel/kernel-add-fillcolor.patch

For hamachi, use the v1 branch. For leo, use the v1.1 branch.

Unagi requires a separate set of patches but it probably won't be worth it to attempt fixing things there.
Can we add:

patch/ics_strawberry/system/core/Revert-msm7627a-Change-adj-level-and-min_free_kbytes.patch
gonk-patches@master: ca1113b

The bluez patch from comment 10 wasn't included in hamachi for some reason so I added it in this patch.
Patching infrastructure is up and patching the important things.
Assignee: mwu → jhford
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
mwu, are there a new kernel(boot.img) for moz build that applying the above kernel patch?
Flags: needinfo?(mwu)
After the bug was fixed, I ceated a v1.0.1 buri's ROM by enabling HwComposer. I see following problem. Is it related to kernel?
 - During playing a wideo video on Bug 850566, I rotate an orientation of the device, drawing at previous orientation appeared in black background drawing.

- During scrolling in a web content like following, drawing often flashed by white color.
https://blog.mozilla.org/sumo/2013/03/09/content-tuning/
We still can't enable HWC. There's a patch we need on the gecko side that hasn't landed yet.
Flags: needinfo?(mwu)
(In reply to Michael Wu [:mwu] from comment #16)
> We still can't enable HWC. There's a patch we need on the gecko side that
> hasn't landed yet.

mwu, is there a bug for it?
Flags: needinfo?(mwu)
Thanks for the info! By applying the patch the first problem in comment #15 is solved. But the second one is not solved.
You need to log in before you can comment on or make changes to this bug.