kernel BUG at kernel/sched/deadline ?

[MagnaboscoL]

Hi,

I am just starting to test a "custom distro" I have made using this kernel and Yocto.
I am facing a "kernel BUG" I cannot understand.

Using:

• kernel 5.4.20-ti-rt-r6 1fe65cd (or 5.4.38-ti-rt-r8 487bc1d)
• defconfig from ti-linux-rt-5.4.y/patches/defconfig. The only modifications I have done are unsetting both CONFIG_MODULE_COMPRESS and CONFIG_MODULE_COMPRESS_XZ
• kernel module omap5-sgx-ddk for GPU support from https://git.ti.com
• other packages from meta-ti yocto layer (tag: ti2020.00)
• BB-BONE-LCD5-01-00A1.dts

Hardware:

• BeagleBone green
• a resistive LCD Seeed-Studio-BeagleBone-Green-LCD-Cape

When I start Weston touch screen calibration I get:

[   63.286156] 000: kernel BUG at kernel/sched/deadline.c:1495!
[   63.291842] 000: Internal error: Oops - BUG: 0 [#1] PREEMPT_RT SMP ARM
[   63.298400] 000: Modules linked in:
[   63.301902] 000:  pvrsrvkm(O)
[   63.304877] 000:  joydev
[   63.307415] 000:  evdev
[   63.309866] 000:
[   63.311798] 000: CPU: 0 PID: 644 Comm: weston Tainted: G           O      5.4.20 #1
[   63.319491] 000: Hardware name: Generic AM33XX (Flattened Device Tree)
[   63.326044] 000: PC is at enqueue_task_dl+0x90/0xd8c
[   63.331054] 000: LR is at rt_mutex_setprio+0x358/0x568
[   63.336223] 000: pc : [<c018c014>]    lr : [<c01754d4>]    psr: 20000093
[   63.342950] 000: sp : da4afa58  ip : da4afb20  fp : da4afb1c
[   63.348630] 000: r10: df926140  r9 : dc32c420  r8 : 00000001
...

Pleased find attached the entire log
bug_log.txt

Note: attached log has been captured when starting the touch screen calibration, but I got the very same bug with an image without Weston at the end of ts_calibrate from tslib.

Considerations

1. if I use the kernel linux-ti-staging-rt-5.4 from TI it seems that the issue disappears. Therefore I suppose that I am not configuring this kernel properly.

2. I suppose that the issue is independent form the applications I was runnig and maybe related to some wrong compilation settings (that probably influenced the module pvrsrvkm.ko), but for the moment I am quite clueless.

3. I am not sure that this is worth mentioning but, since the DRM seems to be involved the details about libdrm I am using are:

   • https://dri.freedesktop.org/libdrm/libdrm-2.4.99.tar.bz2 with patch musl-ioctl.patch
   • I tried with and without the patches implemented in meta-arago (see TI pathces) with the same results

Any hint would be very appreciated.

Regards
Luca

[pdp7]

@jadonk @RobertCNelson do you have the Seeed LCD cape? Maybe we could ask one of the Seeed engineers to check?

[RobertCNelson]

This needs to be retested, early 5.4 versions where not being tested.

[pdp7]

@Pillar1989 are you able to test the Seeed BeagleBone Green LCD Cape mentioned in this issue?

[Pillar1989]

@pdp7 not yet. we tested it on version 4.19.

[MagnaboscoL]

Hi @pdp7,

I have a Yocto setup that seems to work* with this LCD Cape.
At the moment I am using TI kernel and I could switch kernel for testing.
Please let me know if you think I could do something to speed up the testing.

*Just to give complete info: It seems to work but I am still facing a color issue when using the GPU for QT apps as described in e2e.ti.com. Anyway that could probably be unrelated to the tests needed for this issue.

Regards