Re: [PATCH 3/4] powerpc: Add mm_cpumask warning when context switching

2023-08-18 Thread Michael Ellerman 
Nicholas Piggin  writes:
> When context switching away from an mm, add a CONFIG_DEBUG_VM warning
> check to ensure this CPU is still set in the mask. This could catch
> bugs where the mask is improperly trimmed while the CPU is still using
> the mm.
>
> Signed-off-by: Nicholas Piggin 
> ---
>  arch/powerpc/mm/mmu_context.c | 7 +--
>  1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/arch/powerpc/mm/mmu_context.c b/arch/powerpc/mm/mmu_context.c
> index 894468975a44..b24c19078eb1 100644
> --- a/arch/powerpc/mm/mmu_context.c
> +++ b/arch/powerpc/mm/mmu_context.c
> @@ -101,6 +102,8 @@ void switch_mm_irqs_off(struct mm_struct *prev, struct 
> mm_struct *next,
>* sub architectures. Out of line for now
>*/
>   switch_mmu_context(prev, next, tsk);
> +
> + VM_WARN_ON_ONCE(!cpumask_test_cpu(cpu, mm_cpumask(prev)));

This is popping during CPU hotunplug. I guess some confusion about when
the mask is cleared.

cheers


[  145.150374][T0] [ cut here ]
[  145.150459][T0] WARNING: CPU: 5 PID: 0 at 
arch/powerpc/mm/mmu_context.c:106 switch_mm_irqs_off+0x320/0x340
[  145.150519][T0] Modules linked in: bonding pseries_rng rng_core 
binfmt_misc aes_gcm_p10_crypto zram vmx_crypto gf128mul crc32c_vpmsum papr_scm 
ip6_tables ip_tables x_tables fuse autofs4
[  145.150588][T0] CPU: 5 PID: 0 Comm: swapper/5 Not tainted 
6.5.0-rc3-00084-g01477eb5e323 #47
[  145.150592][T0] Hardware name: IBM,9080-HEX POWER10 (raw) 0x800200 
0xf06 of:IBM,FW1030.00 (NH1030_019) hv:phyp pSeries
[  145.150595][T0] NIP:  c00cbc30 LR: c00cbab0 CTR: 
c0181a40
[  145.150598][T0] REGS: c985fb10 TRAP: 0700   Not tainted  
(6.5.0-rc3-00084-g01477eb5e323)
[  145.150602][T0] MSR:  8282b033  
 CR: 24000208  XER: 011e
[  145.150625][T0] CFAR: c00cbae4 IRQMASK: 1 
[  145.150625][T0] GPR00: c00cbab0 c985fdb0 
c27aaf00 c00b02955f00 
[  145.150625][T0] GPR04: c43e0b00 c67c7000 
 00030e2d 
[  145.150625][T0] GPR08: c451af00 0001 
c451af00 c451af00 
[  145.150625][T0] GPR12: c0181a40 c0050fffa300 
 1eed51a0 
[  145.150625][T0] GPR16:   
  
[  145.150625][T0] GPR20:   
 0001 
[  145.150625][T0] GPR24: 0005 dedc 
c4468470 0001 
[  145.150625][T0] GPR28: c67c7000  
0005 c00b02956780 
[  145.150711][T0] NIP [c00cbc30] switch_mm_irqs_off+0x320/0x340
[  145.150716][T0] LR [c00cbab0] switch_mm_irqs_off+0x1a0/0x340
[  145.150721][T0] Call Trace:
[  145.150724][T0] [c985fdb0] [c0448688] 
__smp_call_single_queue+0x198/0x1f0 (unreliable)
[  145.150732][T0] [c985fdf0] [c02af958] 
idle_task_exit+0xf8/0x230
[  145.150740][T0] [c985fe40] [c0181aac] 
pseries_cpu_offline_self+0x6c/0x230
[  145.150748][T0] [c985feb0] [c0092bb4] 
arch_cpu_idle_dead+0x64/0x90
[  145.150755][T0] [c985fee0] [c02fc09c] do_idle+0x25c/0x740
[  145.150761][T0] [c985ff60] [c02fcd14] 
cpu_startup_entry+0x84/0xa0
[  145.150765][T0] [c985ff90] [c0092500] 
start_secondary+0x4e0/0x510
[  145.150772][T0] [c985ffe0] [c000e258] 
start_secondary_prolog+0x10/0x14
[  145.150788][T0] Code: 0fe0 3ce201d7 e947a198 394a0001 f947a198 
4bfffd70 6000 6042 3d4201d7 e92aa210 39290001 f92aa210 <0fe0> 
3d4201d7 e8010050 e92aa218 
[  145.150828][T0] irq event stamp: 49758
[  145.150831][T0] hardirqs last  enabled at (49757): [] 
tick_nohz_idle_enter+0x118/0x2b0
[  145.150836][T0] hardirqs last disabled at (49758): [] 
do_idle+0x1f8/0x740
[  145.150839][T0] softirqs last  enabled at (49724): [] 
__do_softirq+0x5c8/0x7f4
[  145.150847][T0] softirqs last disabled at (49703): [] 
do_softirq_own_stack+0x50/0x80
[  145.150852][T0] ---[ end trace  ]---

[PATCH 3/4] powerpc: Add mm_cpumask warning when context switching

2023-05-24 Thread Nicholas Piggin 
When context switching away from an mm, add a CONFIG_DEBUG_VM warning
check to ensure this CPU is still set in the mask. This could catch
bugs where the mask is improperly trimmed while the CPU is still using
the mm.

Signed-off-by: Nicholas Piggin 
---
 arch/powerpc/mm/mmu_context.c | 7 +--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/mm/mmu_context.c b/arch/powerpc/mm/mmu_context.c
index 894468975a44..b24c19078eb1 100644
--- a/arch/powerpc/mm/mmu_context.c
+++ b/arch/powerpc/mm/mmu_context.c
@@ -43,12 +43,13 @@ static inline void switch_mm_pgdir(struct task_struct *tsk,
 void switch_mm_irqs_off(struct mm_struct *prev, struct mm_struct *next,
struct task_struct *tsk)
 {
+   int cpu = smp_processor_id();
bool new_on_cpu = false;
 
/* Mark this context has been used on the new CPU */
-   if (!cpumask_test_cpu(smp_processor_id(), mm_cpumask(next))) {
+   if (!cpumask_test_cpu(cpu, mm_cpumask(next))) {
VM_WARN_ON_ONCE(next == _mm);
-   cpumask_set_cpu(smp_processor_id(), mm_cpumask(next));
+   cpumask_set_cpu(cpu, mm_cpumask(next));
inc_mm_active_cpus(next);
 
/*
@@ -101,6 +102,8 @@ void switch_mm_irqs_off(struct mm_struct *prev, struct 
mm_struct *next,
 * sub architectures. Out of line for now
 */
switch_mmu_context(prev, next, tsk);
+
+   VM_WARN_ON_ONCE(!cpumask_test_cpu(cpu, mm_cpumask(prev)));
 }
 
 #ifndef CONFIG_PPC_BOOK3S_64
-- 
2.40.1