[PATCH v4 00/19] Basic clock, reset & device tree support for StarFive JH7110 RISC-V SoC

[Hal Feng]

On Tue, 7 Mar 2023 08:51:49 +0000, Conor Dooley wrote:
> On Tue, Mar 07, 2023 at 04:36:41PM +0800, Hal Feng wrote:
>> On Tue, 21 Feb 2023 10:46:26 +0800, Hal Feng wrote:
>> > This patch series adds basic clock, reset & DT support for StarFive
>> > JH7110 SoC. Patch 17 depends on series [1] which provides pinctrl
>> > dt-bindings. Patch 19 depends on series [2] which provides dt-bindings
>> > of VisionFive 2 board and JH7110 SoC.
>> > 
>> > You can simply review or test the patches at the link [3].
>> > 
>> > [1]: https://lore.kernel.org/all/20230209143702.44408-1-hal.feng@starfivetech.com/
>> > [2]: https://lore.kernel.org/all/20230216131511.3327943-1-conor.dooley@microchip.com/
>> > [3]: https://github.com/hal-feng/linux/commits/visionfive2-minimal
>> 
>> Hi Conor,
>> 
>> When I tried to rebase these patches on v6.3-rc1, I found the kernel
>> would crash on the VisionFive 2 board during startup. The logs are as
>> below. I checkout the branch to the mainline and found that the kernel
>> would also crash on the VisionFive board which is equipped with JH7100
>> SoC.
>> 
>> --------------------------------
>> Unable to handle kernel paging request at virtual address 0000004cccccccd4
>> Oops [#1]
>> Modules linked in:
>> CPU: 3 PID: 87 Comm: udevd Not tainted 6.3.0-rc1-00019-g239e7809f291 #305
>> Hardware name: StarFive VisionFive 2 v1.3B (DT)
>> epc : enqueue_timer+0x18/0x90
>>  ra : internal_add_timer+0x2c/0x38
>> epc : ffffffff8006a714 ra : ffffffff8006a7b8 sp : ffffffc80443bc80
>>  gp : ffffffff80eb5100 tp : ffffffd8c01db200 t0 : 0000000000000000
>>  t1 : 000000000000000f t2 : 0000000038b3ea28 s0 : ffffffc80443bcb0
>>  s1 : ffffffff80813940 a0 : ffffffff80813940 a1 : ffffffc80443bd48
>>  a2 : 000000000000020b a3 : cccccccd0b000000 a4 : cccccccccccccccc
>>  a5 : 000000000000020b a6 : ffffffff80814a08 a7 : 0000000000000001
>>  s2 : ffffffc80443bd48 s3 : 0000000008400040 s4 : ffffffff80813940
>>  s5 : ffffffff80eea0b8 s6 : ffffffff80eb7220 s7 : 0000000000000040
>>  s8 : ffffffff80eb61e0 s9 : 0000002ac84a2548 s10: 0000002ad53e92c0
>>  s11: 0000000000000001 t3 : 000000000000003f t4 : 0000000000000000
>>  t5 : 0000000000000004 t6 : 0000000000000003
>> status: 0000000200000100 badaddr: 0000004cccccccd4 cause: 000000000000000f
>> [<ffffffff8006a714>] enqueue_timer+0x18/0x90
>> [<ffffffff8006aa64>] add_timer_on+0xf0/0x134
>> [<ffffffff80500f18>] try_to_generate_entropy+0x1ec/0x232
>> [<ffffffff8035a636>] urandom_read_iter+0x42/0xc2
>> [<ffffffff800fff16>] vfs_read+0x17c/0x1e4
>> [<ffffffff801005b6>] ksys_read+0x78/0x98
>> [<ffffffff801005e4>] sys_read+0xe/0x16
>> [<ffffffff800035dc>] ret_from_syscall+0x0/0x2
>> Code: 9381 9713 0037 0813 0705 983a 3703 0008 e198 c311 (e70c) d713 
>> ---[ end trace 0000000000000000 ]---
>> note: udevd[87] exited with irqs disabled
>> Segmentation fault
>> FAIL
>> Saving random seed: 
>> rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
>> rcu: 	1-...0: (0 ticks this GP) idle=19c4/1/0x4000000000000000 softirq=42/42 fqs=7474
>> rcu: 	(detected by 2, t=15005 jiffies, g=-195, q=35 ncpus=4)
>> Task dump for CPU 1:
>> task:dd              state:R  running task     stack:0     pid:92    ppid:88     flags:0x00000008
>> Call Trace:
>> [<ffffffff80003764>] ret_from_fork+0x0/0xc
>> rcu: INFO: rcu_sched detected stalls on CPUs/tasks:
>> rcu: 	1-...0: (0 ticks this GP) idle=19c4/1/0x4000000000000000 softirq=42/42 fqs=29814
>> rcu: 	(detected by 2, t=60018 jiffies, g=-195, q=35 ncpus=4)
>> Task dump for CPU 1:
>> task:dd              state:R  running task     stack:0     pid:92    ppid:88     flags:0x00000008
>> Call Trace:
>> [<ffffffff80003764>] ret_from_fork+0x0/0xc
>> ...
>> --------------------------------
>> 
>> I used 'git bisect' and found out the commit 9493e6f3ce02 is the
>> cause. I tried to revert this commit on the tag v6.3-rc1, but it
>> seems there is no improvement.
> 
> Hmm, I'm not entirely sure that that is a good bisect.
> This is a fix for my stupidity in the commit you mention:
> https://lore.kernel.org/linux-riscv/20230302174154.970746-1-conor@kernel.org/
> 
> But the main backtrace there is not from that patch at all, I think it
> is Linus' fault.
> The HEAD of Linus' tree is currently 8ca09d5fa3549 ("cpumask: fix
> incorrect cpumask scanning result checks") should be a fix for the
> backtrace that you are seeing above.
> 
>> Any options I am missing? Could you please give me some suggestions
>> to adapt to the new changes between 6.2 and 6.3? Thank you in
>> advance.
> 
> LMK if the above two things don't fix it for you & I'll go digging
> tonight.

The above two methods can fix the problem. Here are my test results.
The VisionFive board can boot up successfully if and only if all above
two applied.
The VisionFive 2 board can boot up successfully if I merge Linus's new
changes.

Hope your fix will be merged in rc2. Thank you for your reply.

Best regards,
Hal