-
Notifications
You must be signed in to change notification settings - Fork 5.3k
Fix USB/FIQ lock-ups #3336
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix USB/FIQ lock-ups #3336
Conversation
2a2944b to
0cdd7ec
Compare
|
V2 Thinko was on my side :) |
Found this by sending patch to raspberrypi/linux [1] [1] raspberrypi/linux#3336 Signed-off-by: Andreas Müller <[email protected]>
What I did - no rocket science [1]:
* Applied RT-patches 4.19.72-rt26 on top of rpi-4.19.y / 4.19.79 (most recent
version applying rt-patch properly)
* Applied a slightly rebased version of the original (4.14) fiq-patch [2]
* grepped for 'fiq_fsm_spin_lock(' and 'fiq_fsm_spin_unlock(' and added missing
rt-specific replacements
* rebased changes back to rpi-4.19.y-rt
What this patch does:
* add one missing pair of fiq_fsm_spin_lock/fiq_fsm_spin_unlock replacements
With builds of [1] Rapsi3 is running without a singe issue for two weeks now
and it was stressed by
* moving gigabytes from USB-Stick to SDCard
* several usb-midi-keyboard jam sessions
Addresses [3]
[1] https://github.com/schnitzeltony/meta-raspi-light/tree/master/recipes-kernel/linux
[2] raspberrypi@05dd5c4
[3] raspberrypi#2943
Signed-off-by: Andreas Müller <[email protected]>
0cdd7ec to
4e27fba
Compare
|
Hey, I ran a test build of RealtimePi based on this branch. Would appreciate testers, might not work on Pi4 yet, but that I can add on my end later: |
|
@guysoft In the event we get no third-party confirmation, I'll accept your gut feel that the patch is plausible. |
|
I will try find a way to test it, currently out of free SD cards. I'd
recommended at least someone boots and confirms with uname before you pull
it.
|
|
Looks like its booting on a 3B+. The pi4 is not booting for me. Unsure why. |
|
Yes - I tested with P3 only since I do not have a running pi4 yet (have no cooling solution with soundcard plugged) @guysoft Can you share some logs? |
|
Ok I have located I was building the Pi4 kernel wrong. So Its likely this would also work for the Pi4. But it might take me a day or so to get the configuration right and push another image out (this is regarding kenrel7l.img) After that I can also add kernel8.img build (aka 64bit kernel) as well. |
|
@schnitzeltony The logs are available on the Jenkins server building this. I closed public view after Jenkins had a vulnerability that compromised the jenkins user. |
|
OK - will wait for pi4 rebuild results. If it still fails I'll send you an email |
|
Looks like we have a go: You can download the image at: I am adding the 64bit build too and making a release candidate on my side. This was stalling long enough. |
|
@pelwell Ok, so it seems all is good except I can't build and test on the 64bit kernel. Can you tell me what I need to set instead of Also is |
|
Assuming you can get hold of a cross compiler (which might be tricky if building on a 32-bit userland - it's probably easier from x86-64) you'll need something like: |
|
And yes, the same DTB can be used for both. |
|
@pelwell Ah, hell. That would be a pain. At the moment CustomPiOS was mainly built so you can install packages on an SD card as if you are doing so on a Pi. That is, it uses the toolchain from Raspbian. This indeed would require the 64bit toolchain to run on the current release of Raspbian I think if this is the case for now I will release the current build, and then figure if there is call for a 64bit realtime kernel and if so how to build it. |
|
Ok, rebuilding against |
What I did - no rocket science [1]:
version applying rt-patch properly)
rt-specific replacements
What this patch does:
With builds of [1] Rapsi3 is running without a singe issue for two weeks now
and it was stressed by
Addresses [3]
[1] https://github.com/schnitzeltony/meta-raspi-light/tree/master/recipes-kernel/linux
[2] 05dd5c4
[3] #2943
Signed-off-by: Andreas Müller [email protected]