On 12 June 2013 12:10, Ryan Harkin ryan.harkin@linaro.org wrote:
On 12 June 2013 18:42, Roy Franz roy.franz@linaro.org wrote:
Hi Oliver,
Sorry for the missing information. I am using version 7.1.63 of the
A15
fast model.
I use 7.1 too, I never got around to upgrading to the 8.0.1 models... but you could try upgrading. They're listed on the FexLM page...
I may try that - I must have missed those, as I though I was getting the latest when I set this up.
I'm using the ArmVExpress-RTSM-A15_MPCore.dsc dsc file,and building against the linaro-tracking-2013.05 branch of the uefi-next repository. I have been making some changes, mostly debug related.
I had made the following changes to the DSC file: gArmPlatformTokenSpaceGuid.PcdCoreCount|1 gArmPlatformTokenSpaceGuid.PcdSendSgiToBringUpSecondaryCores|FALSE
Try doing a diff between ./ArmPlatformPkg/ArmVExpressPkg/ArmVExpress-RTSM-A15_MPCore.dsc ./ArmPlatformPkg/ArmVExpressPkg/ArmVExpress-RTSM-A15.dsc
It begs the question that if you want a single core mode, why you don't use the single core model DSC file? You can then build both and interchange more easily.
Of course, you can also run the MPCore binary on the A15x1 model too. So then you only have to build 1 UEFI binary, but change the model binary instead...
Obviously, the model shouldn't crash, but that's a different point altogether ;-)
I don't need the multi-core stuff, so I have switched to the single core.
This was running fine on a single core until I started removing some
debug
prints.
Reece had the problem where *adding* debug causes the model to seg fault, but not the other way round.
I always run my models with release builds, because I find debug to be too unreliable. Consequently, my debug is always marked as *_ERROR.
This is probably the way to go. I haven't had to do any symbol lookups with the dll loading information, so I'm really just using prints most of the time.
Roy
On 12 June 2013 02:10, Olivier Martin olivier.martin@arm.com wrote:
Hello Roy,
Which platform are you referring to? Which EDK2 DSC file are you using
for
your UEFI firmware?
Thanks,
Olivier
From: boot-architecture-bounces@lists.linaro.org [mailto:boot-architecture-bounces@lists.linaro.org] On Behalf Of Roy
Franz
Sent: 12 June 2013 03:03 To: boot-architecture@lists.linaro.org Subject: Possible race condition in multi-core EDK2
I've been working with the DEBUG build of EDK2, and thought I had
changed
the configuration to be 'proper' for a single core, and it ran fine on a single core simulation. As I started removing debug prints, I started getting simulator seg faults. I confirmed that the difference between running and crashing versions were simply debug prints.
Changing the simulation to simulator 2 cores resolved the problem. I suspect that there may be a race condition or some other multi-core detection bug that causes this crash.
I don't plan to investigate this further at this point - just wanted to report this as I was surprised to find that a debug print being removed would cause the simulator to crash.
Roy
boot-architecture mailing list boot-architecture@lists.linaro.org http://lists.linaro.org/mailman/listinfo/boot-architecture
boot-architecture mailing list boot-architecture@lists.linaro.org http://lists.linaro.org/mailman/listinfo/boot-architecture