[#6055] Build bf561-ezkit SMP kernel failed at anomaly 05000491
Submitted By: Vivi Li
Open Date
2010-06-02 06:52:36 Close Date
2010-08-17 22:04:19
Priority:
Medium Assignee:
Sonic Zhang
Status:
Closed Fixed In Release:
N/A
Found In Release:
2010R1 Release:
Category:
N/A Board:
EZKIT Lite
Processor:
BF561 Silicon Revision:
Is this bug repeatable?:
Yes Resolution:
Fixed
Uboot version or rev.:
Toolchain version or rev.:
gcc4.3-2010_Apr_07
App binary format:
N/A
Summary: Build bf561-ezkit SMP kernel failed at anomaly 05000491
Details:
Build bf561-ezkit SMP kernel failed at anomaly 05000491.
When smp option is selected, we don't have CONFIG_CACHE_FLUSH_L1, so it fails..
--
arch/blackfin/mach-common/arch_checks.c:65:3: error: #error You need IFLUSH in L1 inst while Anomaly 05000491 applies^M
make[2]: *** [arch/blackfin/mach-common/arch_checks.o] Error 1^M
make[1]: *** [arch/blackfin/mach-common] Error 2^M
--
Follow-ups
--- Mike Frysinger 2010-06-02 20:20:25
until we get more feedback from design, this isnt going to change
--- Vivi Li 2010-06-02 22:42:01
If this anomaly exists, we can not run SMP with bf561-ezkit now.
SMP test on bf561-ezkit will be stopped.
--- Mike Frysinger 2010-06-02 23:22:48
yep, that's correct. but until we have information on how exactly to workaround
the anomaly, implementing stuff that will be removed later is a waste of time.
especially since support L1 inst isnt going to be easy on SMP systems.
--- Sonic Zhang 2010-06-22 02:14:17
Any update for anomaly 491 from the hardware team? If there is not a suitable
workaround except L1 memory, we may have to drop SMP support in 2010R1 release.
--- Robin Getz 2010-06-22 09:08:48
No updates so far. I'll poke the person again today.
Worst case is that a helper function in SDRAM is necessary - which determines
which core you are running on, and jumping to that core's L1.
-Robin
--- Sonic Zhang 2010-06-28 03:15:27
Robin,
I wonder if anomaly 491 is exactly the same IFLUSH issue we walked around by
run IFLUSH twice? Could you check with the hardware owner if this is a valid
walk around for anomaly 491?
--- Mike Frysinger 2010-06-28 12:52:04
it is the same issue. the anomaly # came out of that iflush report.
--- Robin Getz 2010-06-28 17:44:17
Sonic:
IFLUSH twice isn't a workaround - it will work sometime, but not all times. The
only reliable workaround at this time is execute from L1.
-Robin
--- Sonic Zhang 2010-08-05 03:50:30
Fixed by executing iflush in different L1 address on different core in SMP
kernel.
--- Vivi Li 2010-08-17 22:04:19
OK, close it.
Files
Changes
Commits
Dependencies
Duplicates
Associations
Tags
File Name File Type File Size Posted By
No Files Were Found