[Bug 1001145] New: Cache handling crashes ARM920T with MMU enabled

[bugzilla-daemon at bugs dot ecos dot sourceware dot org]

Please do not reply to this email. Use the web interface provided at:
http://bugs.ecos.sourceware.org/show_bug.cgi?id=1001145

           Summary: Cache handling crashes ARM920T with MMU enabled
           Product: eCos
           Version: CVS
          Platform: Other (please specify)
        OS/Version: ARM
            Status: UNCONFIRMED
          Severity: critical
          Priority: low
         Component: HAL
        AssignedTo: unassigned@bugs.ecos.sourceware.org
        ReportedBy: stano@meduna.org
                CC: ecos-bugs@ecos.sourceware.org
             Class: Advice Request


Created an attachment (id=1122)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=1122)
HAL and the failing application

The attached application causes random crashes/lockups on a Samsung S3C2440
ARM920T-based processor with MMU activated, RAM region actually remapped by the
MMU and the RAM region set to cacheable, bufferable, copyback. The example
either stops ticking in one of the threads after a while, or stops delaying
between the ticks. I am running it via redboot and gdb through serial. The
crashes are caused by the HAL_FLASH_CACHES_OFF/_ON macros - commenting them out
makes the crash go away.

The behaviour looks consistent to the D-cache not being synced to memory before
invalidating - various linked lists contain stale pointers, context switches
end in another thread's stack, scheduler lock has unexpected value etc. 

The HAL cache tests themselves do run, but they are probably not long enough to
be hit by the problem.

The hardware is a FriendlyARM development board
(http://www.friendlyarm.net/products/mini2440) and I am attaching the HAL for
it, but I hope it is reproducible also in some ARM920T based boards the eCos
supports natively.

I am willing to test any patches. Unfortunately I am not so good in ARM lower
level stuff and I am not yet able to write ARM assembler to test and debug such
stuff myself.

-- 
Configure bugmail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.