Closed Bug 1003721 Opened 11 years ago Closed 7 years ago

[Tarako][monkey test] Homescreen was killed continuously

Categories

(Firefox OS Graveyard :: General, defect)

ARM
Gonk (Firefox OS)
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: dliang, Unassigned)

Details

Attachments

(5 files)

Attached file slog, top and ps log
We met a problem that homescreen (w/ oom_adj=8) was killed continuously on monkey test. In this case, CPU usage is high, user can feel the performance is bad. Finally, the device showed black screen, adb can work but no any b2g process. kernel log: 266086:04-29 23:22:22.413 <4>0[43228.353782] lowmem_shrink send sigkill to 23461 ((Preallocated a), adj 8, size 2514 266284:04-29 23:23:43.026 <4>0[43308.969269] lowmem_shrink send sigkill to 23471 (Usage), adj 2, size 4014 266449:04-29 23:24:40.633 <4>0[43366.571146] lowmem_shrink send sigkill to 23412 (Browser), adj 10, size 3962 266491:04-29 23:24:41.503 <4>0[43367.441287] lowmem_shrink send sigkill to 23516 (Homescreen), adj 8, size 4668 266531:04-29 23:24:43.083 <4>0[43369.021130] lowmem_shrink send sigkill to 23728 (Homescreen), adj 8, size 4859 266571:04-29 23:24:43.993 <4>0[43369.935261] lowmem_shrink send sigkill to 23741 (Homescreen), adj 8, size 4874 266611:04-29 23:24:46.025 <4>0[43371.968942] lowmem_shrink send sigkill to 23754 (Homescreen), adj 8, size 5020 266651:04-29 23:24:47.133 <4>0[43373.077472] lowmem_shrink send sigkill to 23770 (Homescreen), adj 8, size 5093 266691:04-29 23:24:48.213 <4>0[43374.151598] lowmem_shrink send sigkill to 23785 (Homescreen), adj 8, size 5496 266731:04-29 23:24:49.583 <4>0[43375.523398] lowmem_shrink send sigkill to 23801 (Homescreen), adj 8, size 4897 ... 476058:04-30 02:36:32.767 <4>0[53908.188465] lowmem_shrink send sigkill to 3234 (Homescreen), adj 8, size 3704 476098:04-30 02:36:37.948 <4>0[53913.376706] lowmem_shrink send sigkill to 3256 (Homescreen), adj 8, size 4043 476138:04-30 02:36:44.207 <4>0[53919.629496] lowmem_shrink send sigkill to 3274 (Homescreen), adj 8, size 3965 476178:04-30 02:36:48.677 <4>0[53924.099493] lowmem_shrink send sigkill to 3282 (Homescreen), adj 8, size 4009 476269:04-30 02:36:54.887 <4>0[53930.309971] lowmem_shrink send sigkill to 3303 (Homescreen), adj 8, size 3790 476309:04-30 02:37:00.587 <4>0[53936.012324] lowmem_shrink send sigkill to 3311 (Homescreen), adj 8, size 3921 Top: User 20%, System 78%, IOW 0%, IRQ 0% User 22 + Nice 0 + Sys 86 + Idle 0 + IOW 0 + IRQ 0 + SIRQ 1 = 109 PID TID PR CPU% S VSS RSS PCY UID Thread Proc 2088 2088 0 68% R 55768K 7448K fg root (Preallocated a /system/b2g/plugin-container 9499 9499 0 18% R 176980K 50780K fg root b2g /system/b2g/b2g 881 881 0 5% R 1056K 280K fg root top top 14 14 0 3% S 0K 0K fg root kswapd0 63 63 0 1% D 0K 0K fg root mmcqd/0 9499 9531 0 1% S 176980K 50880K fg root Compositor /system/b2g/b2g 9499 9524 0 1% S 176980K 50780K fg root DOM Worker /system/b2g/b2g 10 10 0 0% S 0K 0K fg root kblockd 12 12 0 0% S 0K 0K fg root cfg80211 13 13 0 0% S 0K 0K fg root pm_print 15 15 0 0% S 0K 0K fg root fsnotify_mark 16 16 0 0% S 0K 0K fg root crypto 29 29 0 0% D 0K 0K fg root vaudio-fe 30 30 0 0% S 0K 0K unk root mux_send 31 31 0 0% S 0K 0K unk root mux_post_recv
upload the log "log_5-1_cpu_busy.tar.bz2" same situation as danny's log: CPU usage is high,the phone become very slow!
(In reply to yiwen.liu from comment #1) > Created attachment 8415761 [details] > log_5-1_cpu_busy.tar.bz2 > > upload the log "log_5-1_cpu_busy.tar.bz2" > > same situation as danny's log: CPU usage is high,the phone become very slow! I didn't see homescreen was killed continuously in this log, so it might be the different situation.
We met this issue by monkey test again, some findings as following: 1. memory usage of b2g is bigger, and might have some leak. 2. Homescreen(8) was killed by LMK due to low cache and free memory. 3. b2g is busy due to memory pressure event. 4. Homescreen keep restart at background w/ screen off. I also cat the b2g-info every 0.5s to monitor app status.
Attached file about-memory.tar.gz
get about memory
Attached file b2g-info.txt
b2g-info every 0.5 seconds.
Attached patch debug.patchSplinter Review
Hi Danny, try monkey test again with this debug patch. We'd like to know if homescreen window is in foreground or background when killed.
From cc-edges of attachment 8426108 [details], I can see leaks in callers and _activities found in bug 1007520.
Seems there are other leaks: ting@sweet:~/Desktop/about-memory-3$ python ~/w/fx/tools/heapgraph/cc/census.py --num-show-rc 10 cc-edges.84.log Object frequency. Showing no more than 5 classes of objects, with at least 5 objects each. 155433 JS Object 21675 XPCWrappedNative 10880 nsXPCWrappedJS 9346 JS Script 2062 nsGenericDOMDataNode Objects with highest ref counts. Showing no more than 10 objects, with ref count of at least 0 rc=1502 0x47343470 nsNodeInfo ([none]) #text rc=959 0x47b94bf0 nsNodeInfo (xhtml) div rc=834 0x4668f650 FragmentOrElement (xhtml) div class='anonymous-div' app://browser.gaiamobile.org/index.html# rc=505 0x473b1580 FragmentOrElement (xhtml) head app://system.gaiamobile.org/index.html rc=429 0x46a8c330 nsNodeInfo ([none]) #text rc=203 0x478f07e0 FragmentOrElement (xhtml) div class='value-picker-year' app://system.gaiamobile.org/index.html rc=128 0x47343600 nsNodeInfo (xhtml) script rc=126 0x46a2d0b0 nsNodeInfo (XBL) inherits rc=124 0x438ed150 FragmentOrElement (xhtml) div id='activity-window-16' class='appWindow activityWindow inline-activity active slideup opening' app://system.gaiamobile.org/index.html rc=114 0x44507000 nsDocument normal (xhtml) app://browser.gaiamobile.org/index.html#
what happens that homescreen is in process now? Will the memory leak issue be minor in this case or will b2g suffer more?
Firefox OS is not being worked on
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: