Re: [PATCH] lkdtm/bugs: add test for hung smp_call_function_single()

From: Paul E. McKenney
Date: Tue Apr 23 2024 - 22:10:45 EST


On Fri, Apr 19, 2024 at 11:34:52AM +0100, Mark Rutland wrote:
> The CONFIG_CSD_LOCK_WAIT_DEBUG option enables debugging of hung
> smp_call_function*() calls (e.g. when the target CPU gets stuck within
> the callback function). Testing this option requires triggering such
> hangs.
>
> This patch adds an lkdtm test with a hung smp_call_function_single()
> callbac, which can be used to test CONFIG_CSD_LOCK_WAIT_DEBUG and NMI
> backtraces (as CONFIG_CSD_LOCK_WAIT_DEBUG will attempt an NMI backtrace
> of the hung target CPU).
>
> On arm64 using pseudo-NMI, this looks like:
>
> | # mount -t debugfs none /sys/kernel/debug/
> | # echo CSDLOCKUP > /sys/kernel/debug/provoke-crash/DIRECT
> | lkdtm: Performing direct entry CSDLOCKUP
> | smp: csd: Detected non-responsive CSD lock (#1) on CPU#0, waiting 5000001136 ns for CPU#01 __lkdtm_CSDLOCKUP+0x0/0x8(0x0).
> | smp: csd: CSD lock (#1) handling this request.
> | Sending NMI from CPU 0 to CPUs 1:
> | NMI backtrace for cpu 1
> | CPU: 1 PID: 0 Comm: swapper/1 Not tainted 6.9.0-rc4-00001-gda84b9dede43 #7
> | Hardware name: linux,dummy-virt (DT)
> | pstate: 60401005 (nZCv daif +PAN -UAO -TCO -DIT +SSBS BTYPE=--)
> | pc : __lkdtm_CSDLOCKUP+0x0/0x8
> | lr : __flush_smp_call_function_queue+0x1b0/0x290
> | sp : ffff80008000bf30
> | pmr_save: 00000060
> | x29: ffff80008000bf30 x28: fff00000c02dc500 x27: 0000000000000000
> | x26: 0000000000000000 x25: fff00000c02dc500 x24: ffffa41b939aa140
> | x23: ffffa41b939aa140 x22: 0000000000000000 x21: ffff80008066bc40
> | x20: 0000000000000000 x19: 0000000000000000 x18: fff05be56bd37000
> | x17: fff05be56bd07000 x16: ffff800080008000 x15: 00005b132023e6fd
> | x14: 00005aeabb53d8c3 x13: 000000000000032e x12: 0000000000000001
> | x11: 0000000000000040 x10: fff00000c003d0a8 x9 : fff00000c003d0a0
> | x8 : fff00000c0400270 x7 : 0000000000000000 x6 : ffffa41b9251b810
> | x5 : 0000000000000000 x4 : fff05be56bd07000 x3 : ffff80008000bf30
> | x2 : fff05be56bd07000 x1 : ffffa41b939aa140 x0 : 0000000000000000
> | Call trace:
> | __lkdtm_CSDLOCKUP+0x0/0x8
> | generic_smp_call_function_single_interrupt+0x14/0x20
> | ipi_handler+0xb8/0x178
> | handle_percpu_devid_irq+0x84/0x130
> | generic_handle_domain_irq+0x2c/0x44
> | gic_handle_irq+0x118/0x240
> | call_on_irq_stack+0x24/0x4c
> | do_interrupt_handler+0x80/0x84
> | el1_interrupt+0x44/0xc0
> | el1h_64_irq_handler+0x18/0x24
> | el1h_64_irq+0x78/0x7c
> | default_idle_call+0x40/0x60
> | do_idle+0x23c/0x2d0
> | cpu_startup_entry+0x38/0x3c
> | secondary_start_kernel+0x148/0x180
> | __secondary_switched+0xb8/0xbc
> | CPU: 0 PID: 143 Comm: sh Not tainted 6.9.0-rc4-00001-gda84b9dede43 #7
> | Hardware name: linux,dummy-virt (DT)
> | Call trace:
> | dump_backtrace+0x90/0xe8
> | show_stack+0x18/0x24
> | dump_stack_lvl+0xac/0xe8
> | dump_stack+0x18/0x24
> | csd_lock_wait_toolong+0x268/0x338
> | smp_call_function_single+0x1dc/0x2f0
> | lkdtm_CSDLOCKUP+0xcc/0xfc
> | lkdtm_do_action+0x1c/0x38
> | direct_entry+0xbc/0x14c
> | full_proxy_write+0x60/0xb4
> | vfs_write+0xd0/0x35c
> | ksys_write+0x70/0x104
> | __arm64_sys_write+0x1c/0x28
> | invoke_syscall+0x48/0x114
> | el0_svc_common.constprop.0+0x40/0xe0
> | do_el0_svc+0x1c/0x28
> | el0_svc+0x38/0x108
> | el0t_64_sync_handler+0x120/0x12c
> | el0t_64_sync+0x1a4/0x1a8
> | smp: csd: Continued non-responsive CSD lock (#1) on CPU#0, waiting 10000001888 ns for CPU#01 __lkdtm_CSDLOCKUP+0x0/0x8(0x0).
> | smp: csd: CSD lock (#1) handling this request.
>
> Signed-off-by: Mark Rutland <mark.rutland@xxxxxxx>
> Cc: Kees Cook <keescook@xxxxxxxxxxxx>
> Cc: Paul E. McKenney <paulmck@xxxxxxxxxx>

Acked-by: Paul E. McKenney <paulmck@xxxxxxxxxx>

> ---
> drivers/misc/lkdtm/bugs.c | 30 ++++++++++++++++++++++++++++++
> 1 file changed, 30 insertions(+)
>
> I wrote this because I needed to guide someone through debugging a hung
> smp_call_function() call, and I needed examples with/without an NMI
> backtrace. It seems like it'd be useful for testing the CSD lockup
> detector and NMI backtrace code in future.
>
> I'm not sure about the CSDLOCKUP name, but everything else I tried
> didn't seem great either:
>
> * IPILOCKUP sounds like it's testing IPIs generally
> * SMPCALLLOCKUP and similar look weirdly long
> * SMP_CALL_LOCKUP and similar look different to {HARD,SOFT,SPIN}LOCKUP
>
> ... and I'm happy to defer to Kees for the naming. ;)
>
> Mark.
>
> diff --git a/drivers/misc/lkdtm/bugs.c b/drivers/misc/lkdtm/bugs.c
> index 5178c02b21eba..47cd1be09ac1f 100644
> --- a/drivers/misc/lkdtm/bugs.c
> +++ b/drivers/misc/lkdtm/bugs.c
> @@ -286,6 +286,35 @@ static void lkdtm_HARDLOCKUP(void)
> cpu_relax();
> }
>
> +static void __lkdtm_CSDLOCKUP(void *unused)
> +{
> + for (;;)
> + cpu_relax();
> +}
> +
> +static void lkdtm_CSDLOCKUP(void)
> +{
> + unsigned int cpu, target;
> +
> + cpus_read_lock();
> +
> + cpu = get_cpu();
> + target = cpumask_any_but(cpu_online_mask, cpu);
> +
> + if (target >= nr_cpu_ids) {
> + pr_err("FAIL: no other online CPUs\n");
> + goto out_put_cpus;
> + }
> +
> + smp_call_function_single(target, __lkdtm_CSDLOCKUP, NULL, 1);
> +
> + pr_err("FAIL: did not hang\n");
> +
> +out_put_cpus:
> + put_cpu();
> + cpus_read_unlock();
> +}
> +
> static void lkdtm_SPINLOCKUP(void)
> {
> /* Must be called twice to trigger. */
> @@ -680,6 +709,7 @@ static struct crashtype crashtypes[] = {
> CRASHTYPE(UNALIGNED_LOAD_STORE_WRITE),
> CRASHTYPE(SOFTLOCKUP),
> CRASHTYPE(HARDLOCKUP),
> + CRASHTYPE(CSDLOCKUP),
> CRASHTYPE(SPINLOCKUP),
> CRASHTYPE(HUNG_TASK),
> CRASHTYPE(OVERFLOW_SIGNED),
> --
> 2.30.2
>