#12929 NORM 14.1.0: display hang, high load

Zarro Boogs per Child bugtracker at laptop.org
Thu Feb 4 00:21:33 EST 2016


#12929: display hang, high load
-------------------------------------+-------------------------------------
     Reporter:  Quozl                |                 Owner:
         Type:  defect               |                Status:  new
     Priority:  normal               |             Milestone:  14.1.0
    Component:  not assigned         |               Version:  Development
     Keywords:                       |  build as of this date
     Blocking:                       |            Blocked By:
Action Needed:  reproduce            |  Deployments affected:
                                     |              Verified:  0
-------------------------------------+-------------------------------------
 Context: Fedora 20, XO-4, local build 11, test of extreme load web site
 only.

 Problem: Browse hang, normal response to F6, normal response to serial
 port and SSH.

 Analysis: Browse process looping with 100% CPU usage, attached gdb,
 captured PC sample (six times, all same PC), captured core dump, captured
 backtraces;

 {{{
 (gdb) thread apply all bt

 Thread 1 (process 809):
 #0  0xb51a135c in ?? () from /lib/libpixman-1.so.0
 #1  0x00000000 in ?? ()
 (gdb) info reg
 r0             0x50     80
 r1             0x76     118
 r2             0x9dc17000       2646700032
 r3             0x78     120
 r4             0xa77a2020       2809798688
 r5             0x78     120
 r6             0x9dc16fa0       2646699936
 r7             0x78     120
 r8             0x0      0
 r9             0x760    1888
 r10            0x70     112
 r11            0xa77a1fa0       2809798560
 r12            0x9dc11000       2646675456
 sp             0xbe86eb30       0xbe86eb30
 lr             0x0      0
 pc             0xb51a135c       0xb51a135c
 cpsr           0x20030010       537067536
 (gdb) disassemble 0xb51a1350,+32
 Dump of assembler code from 0xb51a1350 to 0xb51a1370:
    0xb51a1350:  ldrbge  r8, [r11, r5, lsl #2]!
    0xb51a1354:  subs    r0, r0, #16
    0xb51a1358:  blt     0xb51a1390
 => 0xb51a135c:  vst1.32 {d0-d3}, [r2 :128]!
    0xb51a1360:  vld1.32 {d0-d3}, [r4]!
    0xb51a1364:  add     r10, r10, #8
    0xb51a1368:  tst     r9, #15
    0xb51a136c:  addne   r10, r10, #8
 End of assembler dump.
 (gdb)
 }}}

 Confirmed by /proc/809/maps:

 {{{
 b5126000-b5127000 rw-p 00022000 b3:02 61252      /usr/lib/libEGL.so.1.0.0
 b5127000-b51b6000 r-xp 00000000 b3:02 63668
 /usr/lib/libpixman-1.so.0.30.0
 b51b6000-b51bd000 ---p 0008f000 b3:02 63668
 /usr/lib/libpixman-1.so.0.30.0
 b51bd000-b51c2000 r--p 0008e000 b3:02 63668
 /usr/lib/libpixman-1.so.0.30.0
 b51c2000-b51c3000 rw-p 00093000 b3:02 63668
 /usr/lib/libpixman-1.so.0.30.0
 b51c3000-b51dc000 r-xp 00000000 b3:02 60129      /usr/lib/libxcb.so.1.1.0
 }}}

 Versions:

 {{{
 glibc-2.22.90-29.fc24.armv7hl
 glibc-common-2.22.90-29.fc24.armv7hl
 pixman-0.30.0-5.fc20.armv7hl
 xorg-x11-drv-vivante-M5.3.12.5.6-6.olpc.0.armv7hl
 xorg-x11-server-Xorg-1.11.4-3.olpc.1.armv7hl
 kernel-3.5.7-00539-gdf0031f-dirty
 }}}

 Possibly related to #12928, and if so suggests loss of NEON register
 contents.

--
Ticket URL: <http://dev.laptop.org/ticket/12929>
One Laptop per Child <http://one.laptop.org/>
One Laptop per Child bug tracking system


More information about the Bugs mailing list