Skip to first unread message
hpo14
Jan 3, 2018, 1:38:24 PM1/3/18
to Android-x86
Hi android-x86 devs
i got a wired hotplug issue between HDMI and DP. ill just keep the description clearly as i can.
OK, here the story is.
# Hardware Desc:
--------------------------------------------------------------------------------
Board or PC: UP Squared (official link, with output DP x1, HDMI x1)
monitor: ViewSonic VS16261 (with input DP x1, HDMI x1)
# Software
--------------------------------------------------------------------------------
Android Version: Android-x86 7.1.2-rc1
Kernel Version: 4.9.31
# when Issue happened (also this is the steps to repoduce)
--------------------------------------------------------------------------------
1. Boot up with HDMI connected.
2. Plug-out HDMI from board after boot to home.
3. Plug-in DP to the board.
4. monitor screen show android logo again. (this was caused by zygote restart due to surfaceflinger crashed)
# when Issue *NOT* happen
--------------------------------------------------------------------------------
1. Boot up with DP connected.
2. Plug-out HDMI from board, after boot to home.
3. Plug-in HDMI to the board.
4. nothing happened.
# Crash log of surfaceflinger
--------------------------------------------------------------------------------
12-22 09:41:08.713 2246 2246 F DEBUG : * * * * * * * * * * * * * * * *
12-22 09:41:08.714 2246 2246 F DEBUG : Revision: ‘0’
12-22 09:41:08.714 2246 2246 F DEBUG : ABI: ‘x86_64’
12-22 09:41:08.714 2246 2246 F DEBUG : pid: 1220, tid: 1269, name: surfaceflinger >>> /system/bin/surfaceflinger <<<
12-22 09:41:08.714 2246 2246 F DEBUG : signal 11 (SIGSEGV), code 1 (SEGV_MAPERR), fault addr 0×10
12-22 09:41:08.714 2246 2246 F DEBUG : rax 0000000000000000 rbx 000079d06b05e1c0 rcx 8f46cce8f92b8f55 rdx 0000000000000002
12-22 09:41:08.714 2246 2246 F DEBUG : rsi 000079d06907f3e0 rdi 000079d06b05e1c0
12-22 09:41:08.714 2246 2246 F DEBUG : r8 ffffffffffffffe8 r9 0000000000000000 r10 fffffffffffff945 r11 0000000000000246
12-22 09:41:08.714 2246 2246 F DEBUG : r12 000079d06aaf68e3 r13 0000000000000000 r14 000079d06aaf63ca r15 0000000000000000
12-22 09:41:08.714 2246 2246 F DEBUG : cs 0000000000000033 ss 000000000000002b
12-22 09:41:08.714 2246 2246 F DEBUG : rip 000079d06aaf1cee rbp 000079d06907f3e0 rsp 000079d06a00e350 eflags 0000000000010206
12-22 09:41:08.860 2246 2246 F DEBUG :
12-22 09:41:08.860 2246 2246 F DEBUG : backtrace:
12-22 09:41:08.860 2246 2246 F DEBUG : #00 pc 0000000000004cee /system/lib64/libgralloc_drm.so
12-22 09:41:08.860 2246 2246 F DEBUG : #01 pc 0000000000004ec2 /system/lib64/libgralloc_drm.so
12-22 09:41:08.860 2246 2246 F DEBUG : #02 pc 0000000000089831 /system/lib64/libc.so (_ZL15__pthread_startPv+177)
12-22 09:41:08.860 2246 2246 F DEBUG : #03 pc 0000000000029aab /system/lib64/libc.so (start_thread+11)
12-22 09:41:08.860 2246 2246 F DEBUG : #04 pc 000000000001cb25 /system/lib64/libc.so (bionic_clone+53)
# backtrace with addr2line
--------------------------------------------------------------------------------
addr2line -Cfe symbols/system/lib64/libgralloc_drm.so 4cee
Routine init_external_output at external/drm_gralloc/gralloc_drm_kms.c:1127
1121: static int init_external_output(struct gralloc_drm_t *drm,
1122: drmModeConnectorPtr connector)
1123: {
1124: struct gralloc_drm_output *output;
1125: output = drm_kms_init_with_new_connector(drm, connector); // return NULL due to no outputs[] could use,
1126:
1127: ALOGD("%s, allocate private buffer for hdmi [%dx%d]", // >>>>>> then cause surfaceflinger crash
1128: __func__, output->mode.hdisplay, output->mode.vdisplay);
11xx: ......
11xx: } /* end of init_external_output */
--------------------------------------------------------------------------------
ill upload logcat later for "issue not happen" and "issue happened".
Also according my observation and analysis of logcat.
this issue seems may relative with drm_gralloc receive hotplug event more than once (here is TWICE)
which sent from i915 driver through uevent.
correct me if i was wrong, please
Thanks
Chih-Wei Huang
Jan 3, 2018, 11:49:27 PM1/3/18
to Android-x86
The code should check it before using.
I've just pushed a fix that you can try.
> 1126:
> 1127: ALOGD("%s, allocate private buffer for hdmi [%dx%d]", //
>>>>>>> then cause surfaceflinger crash
> 1128: __func__, output->mode.hdisplay, output->mode.vdisplay);
> 11xx: ......
> 11xx: } /* end of init_external_output */
> --------------------------------------------------------------------------------
>
> ill upload logcat later for "issue not happen" and "issue happened".
>
> Also according my observation and analysis of logcat.
> this issue seems may relative with drm_gralloc receive hotplug event more
> than once (here is TWICE)
> which sent from i915 driver through uevent.
Chih-Wei
Android-x86 project
http://www.android-x86.org
Message has been deleted
hpo14
Jan 4, 2018, 1:44:20 AM1/4/18
to Android-x86
Hi CW
this patch resolved my issue, thank you.
Just in i guess.
Maybe be the *real* root cause is "i915 driver send hotplug event to uevent for a connector with same status more than once".
About hotplug event relative patch:
1. [v2] drm/i915: Enable hotplug polling after registering the outputs
i have no test it yet.
2. drm/i915: Fix DisplayPort Hotplug
Tested, but not work.
Reply all
Reply to author
Forward
0 new messages