[#5101] Random hang when repeating test hardware error trap

Submitted By: Graf Yang

Open Date

2009-04-29 01:54:12 Close Date

2009-05-04 23:22:07

Priority:

Medium Assignee:

Graf Yang

Status:

Closed Fixed In Release:

N/A

Found In Release:

N/A Release:

Category:

Kernel Functions Board:

N/A

Processor:

BF533 Silicon Revision:

0.4

Is this bug repeatable?:

Yes Resolution:

Rejected

Uboot version or rev.:

Toolchain version or rev.:

2009R1-rc5

App binary format:

FDPIC

Summary: Random hang when repeating test hardware error trap

Details:

system will random hang when repeating test the hardware error, may hang on the 8th, or the 19th.

It existed on 0.3 533(when doing hardware error testing), and 0.2 537(when doing dcplb miss testing). It is frequently on 0.4 533.

root:/> traps_test 18

RuEnxng test ternal Memory Addressing Error

HW Error context

CURRENT PROCESS:

COMM=traps_test1 PID=142

CPU = 0

TEXT = 0x00366000-0x003677ec DATA = 0x003627ec-0x003631e8

BSS = 0x003631e8-0x00340000 USER-STACK = 0x0035fea0

return address: [0x00366fdc]; contents of:

0x00366fb0: bbf0 b9f0 6408 bbe0 b9ea 9151 ac53 0061

0x00366fc0: e801 0000 0010 0000 e800 0005�

(here the system hang, watchdog seems can't reboot it)

Follow-ups

--- Graf Yang 2009-05-04 06:11:59

This bug is caused by watchdog fail.

When watchdog time out, the board hang there, not reset.

--- Mike Frysinger 2009-05-04 13:17:08

not a bug then ...

--- Graf Yang 2009-05-04 21:57:25

Is it a bootrom bug? I can not find it in anomaly list.

--- Mike Frysinger 2009-05-04 22:08:36

the bootrom is not executed at all on the BF533 in bypass mode

the Blackfin system is known to not be recoverable under some circumstances and

i imagine this is one of them (e.g. CFI flash is put into some intermediate mode

where reading is not possible, and then a reset occurs)

--- Sonic Zhang 2009-05-04 22:50:56

Is it better to turn of watchdog when running trap test?

--- Mike Frysinger 2009-05-04 22:58:53

no ... Graf's test case is running traps over and over in a tight loop. that

doesnt occur in the normal test case that gets run on the test bench.

it is also triggering a bug in our hardware handling code that Robin is working

on fixing, so once that gets resolved, it should be much less of an issue.

--- Graf Yang 2009-05-04 23:22:07

It is not a bug, I will close it.

Files

Changes

Commits

Dependencies

Duplicates

Associations