Re: [PATCH v2] tee: optee: replace might_sleep with cond_resched

[Jens Wiklander]

On Tue, Jan 5, 2021 at 11:29 AM Rouven Czerwinski
<r.czerwinski@xxxxxxxxxxxxxx> wrote:
>
> might_sleep() is a debugging aid and triggers rescheduling only for
> certain kernel configurations. Replace with an explicit check and
> reschedule to work for all kernel configurations. Fixes the following
> trace:
>
>   [  572.945146] rcu: INFO: rcu_sched self-detected stall on CPU
>   [  572.949275] rcu:     0-....: (2099 ticks this GP) idle=572/1/0x40000002 softirq=7412/7412 fqs=974
>   [  572.957964]  (t=2100 jiffies g=10393 q=21)
>   [  572.962054] NMI backtrace for cpu 0
>   [  572.965540] CPU: 0 PID: 165 Comm: xtest Not tainted 5.8.7 #1
>   [  572.971188] Hardware name: STM32 (Device Tree Support)
>   [  572.976354] [<c011163c>] (unwind_backtrace) from [<c010b7f8>] (show_stack+0x10/0x14)
>   [  572.984080] [<c010b7f8>] (show_stack) from [<c0511e4c>] (dump_stack+0xc4/0xd8)
>   [  572.991300] [<c0511e4c>] (dump_stack) from [<c0519abc>] (nmi_cpu_backtrace+0x90/0xc4)
>   [  572.999130] [<c0519abc>] (nmi_cpu_backtrace) from [<c0519bdc>] (nmi_trigger_cpumask_backtrace+0xec/0x130)
>   [  573.008706] [<c0519bdc>] (nmi_trigger_cpumask_backtrace) from [<c01a5184>] (rcu_dump_cpu_stacks+0xe8/0x110)
>   [  573.018453] [<c01a5184>] (rcu_dump_cpu_stacks) from [<c01a4234>] (rcu_sched_clock_irq+0x7fc/0xa88)
>   [  573.027416] [<c01a4234>] (rcu_sched_clock_irq) from [<c01acdd0>] (update_process_times+0x30/0x8c)
>   [  573.036291] [<c01acdd0>] (update_process_times) from [<c01bfb90>] (tick_sched_timer+0x4c/0xa8)
>   [  573.044905] [<c01bfb90>] (tick_sched_timer) from [<c01adcc8>] (__hrtimer_run_queues+0x174/0x358)
>   [  573.053696] [<c01adcc8>] (__hrtimer_run_queues) from [<c01aea2c>] (hrtimer_interrupt+0x118/0x2bc)
>   [  573.062573] [<c01aea2c>] (hrtimer_interrupt) from [<c09ad664>] (arch_timer_handler_virt+0x28/0x30)
>   [  573.071536] [<c09ad664>] (arch_timer_handler_virt) from [<c0190f50>] (handle_percpu_devid_irq+0x8c/0x240)
>   [  573.081109] [<c0190f50>] (handle_percpu_devid_irq) from [<c018ab8c>] (generic_handle_irq+0x34/0x44)
>   [  573.090156] [<c018ab8c>] (generic_handle_irq) from [<c018b194>] (__handle_domain_irq+0x5c/0xb0)
>   [  573.098857] [<c018b194>] (__handle_domain_irq) from [<c052ac50>] (gic_handle_irq+0x4c/0x90)
>   [  573.107209] [<c052ac50>] (gic_handle_irq) from [<c0100b0c>] (__irq_svc+0x6c/0x90)
>   [  573.114682] Exception stack(0xd90dfcf8 to 0xd90dfd40)
>   [  573.119732] fce0:                                                       ffff0004 00000000
>   [  573.127917] fd00: 00000000 00000000 00000000 00000000 00000000 00000000 d93493cc ffff0000
>   [  573.136098] fd20: d2bc39c0 be926998 d90dfd58 d90dfd48 c09f3384 c01151f0 400d0013 ffffffff
>   [  573.144281] [<c0100b0c>] (__irq_svc) from [<c01151f0>] (__arm_smccc_smc+0x10/0x20)
>   [  573.151854] [<c01151f0>] (__arm_smccc_smc) from [<c09f3384>] (optee_smccc_smc+0x3c/0x44)
>   [  573.159948] [<c09f3384>] (optee_smccc_smc) from [<c09f4170>] (optee_do_call_with_arg+0xb8/0x154)
>   [  573.168735] [<c09f4170>] (optee_do_call_with_arg) from [<c09f4638>] (optee_invoke_func+0x110/0x190)
>   [  573.177786] [<c09f4638>] (optee_invoke_func) from [<c09f1ebc>] (tee_ioctl+0x10b8/0x11c0)
>   [  573.185879] [<c09f1ebc>] (tee_ioctl) from [<c029f62c>] (ksys_ioctl+0xe0/0xa4c)
>   [  573.193101] [<c029f62c>] (ksys_ioctl) from [<c0100060>] (ret_fast_syscall+0x0/0x54)
>   [  573.200750] Exception stack(0xd90dffa8 to 0xd90dfff0)
>   [  573.205803] ffa0:                   be926bf4 be926a78 00000003 8010a403 be926908 004e3cf8
>   [  573.213987] ffc0: be926bf4 be926a78 00000000 00000036 be926908 be926918 be9269b0 bffdf0f8
>   [  573.222162] ffe0: b6d76fb0 be9268fc b6d66621 b6c7e0d8
>
> seen on STM32 DK2 with CONFIG_PREEMPT_NONE.
>
> Signed-off-by: Rouven Czerwinski <r.czerwinski@xxxxxxxxxxxxxx>
> Tested-by: Sumit Garg <sumit.garg@xxxxxxxxxx>
> ---
> v2:
> - Add tested-by from Sumit Garg
> - Adjust commit message as agreed upon with Lucas Stach
>
>  drivers/tee/optee/call.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)

I'm picking this up.

Thanks,
Jens