Skip to contentSkip to conversation

powerpc: disable sanitizer in irq_soft_mask_set

PATCH In review Zhouyi Zhou 3 participants 8 messages Aug 21, 2022 – Aug 24, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 powerpc: disable sanitizer in irq_soft_mask_set

In ppc, compiler based sanitizer will generate instrument instructions around statement WRITE_ONCE(local_paca->irq_soft_mask, mask):

   0xc000000000295cb0 <+0>:	addis   r2,r12,774
   0xc000000000295cb4 <+4>:	addi    r2,r2,16464
   0xc000000000295cb8 <+8>:	mflr    r0
   0xc000000000295cbc <+12>:	bl      0xc00000000008bb4c <mcount>
   0xc000000000295cc0 <+16>:	mflr    r0
   0xc000000000295cc4 <+20>:	std     r31,-8(r1)
   0xc000000000295cc8 <+24>:	addi    r3,r13,2354
   0xc000000000295ccc <+28>:	mr      r31,r13
   0xc000000000295cd0 <+32>:	std     r0,16(r1)
   0xc000000000295cd4 <+36>:	stdu    r1,-48(r1)
   0xc000000000295cd8 <+40>:	bl      0xc000000000609b98 <__asan_store1+8>
   0xc000000000295cdc <+44>:	nop
   0xc000000000295ce0 <+48>:	li      r9,1
   0xc000000000295ce4 <+52>:	stb     r9,2354(r31)
   0xc000000000295ce8 <+56>:	addi    r1,r1,48
   0xc000000000295cec <+60>:	ld      r0,16(r1)
   0xc000000000295cf0 <+64>:	ld      r31,-8(r1)
   0xc000000000295cf4 <+68>:	mtlr    r0

If there is a context switch before "stb r9,2354(r31)", r31 may
not equal to r13, in such case, irq soft mask will not work.

This patch disable sanitizer in irq_soft_mask_set.

---
Dear PPC developers

I found this bug when trying to do rcutorture tests in ppc VM of Open Source Lab of Oregon State University following Paul E. McKenny's guidance.

console.log report following bug:

[  346.527467][  T100] BUG: using smp_processor_id() in preemptible [00000000] code: rcu_torture_rea/100^M
[  346.529416][  T100] caller is rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.531157][  T100] CPU: 4 PID: 100 Comm: rcu_torture_rea Tainted: G        W          5.19.0-rc5-next-20220708-dirty #253^M
[  346.533620][  T100] Call Trace:^M
[  346.534449][  T100] [c0000000094876c0] [c000000000ce2b68] dump_stack_lvl+0xbc/0x108 (unreliable)^M
[  346.536632][  T100] [c000000009487710] [c000000001712954] check_preemption_disabled+0x154/0x160^M
[  346.538665][  T100] [c0000000094877a0] [c0000000002ce2d4] rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.540830][  T100] [c0000000094878b0] [c0000000002cf3c0] __rcu_read_unlock+0x290/0x3b0^M
[  346.542746][  T100] [c000000009487910] [c0000000002bb330] rcu_torture_read_unlock+0x30/0xb0^M
[  346.544779][  T100] [c000000009487930] [c0000000002b7ff8] rcutorture_one_extend+0x198/0x810^M
[  346.546851][  T100] [c000000009487a10] [c0000000002b8bfc] rcu_torture_one_read+0x58c/0xc90^M
[  346.548844][  T100] [c000000009487ca0] [c0000000002b942c] rcu_torture_reader+0x12c/0x360^M
[  346.550784][  T100] [c000000009487db0] [c0000000001de978] kthread+0x1e8/0x220^M
[  346.552555][  T100] [c000000009487e10] [c00000000000cd54] ret_from_kernel_thread+0x5c/0x64^M

After 12 days debugging, I finally narrow the problem to irq_soft_mask_set.

I am a beginner, hope I can be of some beneficial to the community ;-)

Thanks
Zhouyi
--

 arch/powerpc/include/asm/hw_irq.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/include/asm/hw_irq.h +1 -1
index 26ede09c521d..a5ae8d82cc9d 100644
@@ -121,7 +121,7 @@ static inline notrace unsigned long irq_soft_mask_return(void)
  * for the critical section and as a clobber because
  * we changed paca->irq_soft_mask
  */
-static inline notrace void irq_soft_mask_set(unsigned long mask)
+static inline notrace __no_kcsan __no_sanitize_address void irq_soft_mask_set(unsigned long mask)
 {
 	/*
 	 * The irq mask must always include the STD bit if any are set.
Signature · 1 line
2.34.1
#1 lore raw

Le 21/08/2022 à 03:00, Zhouyi Zhou a écrit :

Quoted 22 lines

In ppc, compiler based sanitizer will generate instrument instructions around statement WRITE_ONCE(local_paca->irq_soft_mask, mask):

    0xc000000000295cb0 <+0>:	addis   r2,r12,774
    0xc000000000295cb4 <+4>:	addi    r2,r2,16464
    0xc000000000295cb8 <+8>:	mflr    r0
    0xc000000000295cbc <+12>:	bl      0xc00000000008bb4c <mcount>
    0xc000000000295cc0 <+16>:	mflr    r0
    0xc000000000295cc4 <+20>:	std     r31,-8(r1)
    0xc000000000295cc8 <+24>:	addi    r3,r13,2354
    0xc000000000295ccc <+28>:	mr      r31,r13
    0xc000000000295cd0 <+32>:	std     r0,16(r1)
    0xc000000000295cd4 <+36>:	stdu    r1,-48(r1)
    0xc000000000295cd8 <+40>:	bl      0xc000000000609b98 <__asan_store1+8>
    0xc000000000295cdc <+44>:	nop
    0xc000000000295ce0 <+48>:	li      r9,1
    0xc000000000295ce4 <+52>:	stb     r9,2354(r31)
    0xc000000000295ce8 <+56>:	addi    r1,r1,48
    0xc000000000295cec <+60>:	ld      r0,16(r1)
    0xc000000000295cf0 <+64>:	ld      r31,-8(r1)
    0xc000000000295cf4 <+68>:	mtlr    r0
Quoted 4 lines

If there is a context switch before "stb r9,2354(r31)", r31 may
not equal to r13, in such case, irq soft mask will not work.

This patch disable sanitizer in irq_soft_mask_set.

Well spotted, thanks.

You should add:

Fixes ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers")

By the way, I think the READ_ONCE() likely has the same issue so you should fix irq_soft_mask_return() at the same time.

Quoted 48 lines

---
Dear PPC developers

I found this bug when trying to do rcutorture tests in ppc VM of Open Source Lab of Oregon State University following Paul E. McKenny's guidance.

console.log report following bug:

[  346.527467][  T100] BUG: using smp_processor_id() in preemptible [00000000] code: rcu_torture_rea/100^M
[  346.529416][  T100] caller is rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.531157][  T100] CPU: 4 PID: 100 Comm: rcu_torture_rea Tainted: G        W          5.19.0-rc5-next-20220708-dirty #253^M
[  346.533620][  T100] Call Trace:^M
[  346.534449][  T100] [c0000000094876c0] [c000000000ce2b68] dump_stack_lvl+0xbc/0x108 (unreliable)^M
[  346.536632][  T100] [c000000009487710] [c000000001712954] check_preemption_disabled+0x154/0x160^M
[  346.538665][  T100] [c0000000094877a0] [c0000000002ce2d4] rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.540830][  T100] [c0000000094878b0] [c0000000002cf3c0] __rcu_read_unlock+0x290/0x3b0^M
[  346.542746][  T100] [c000000009487910] [c0000000002bb330] rcu_torture_read_unlock+0x30/0xb0^M
[  346.544779][  T100] [c000000009487930] [c0000000002b7ff8] rcutorture_one_extend+0x198/0x810^M
[  346.546851][  T100] [c000000009487a10] [c0000000002b8bfc] rcu_torture_one_read+0x58c/0xc90^M
[  346.548844][  T100] [c000000009487ca0] [c0000000002b942c] rcu_torture_reader+0x12c/0x360^M
[  346.550784][  T100] [c000000009487db0] [c0000000001de978] kthread+0x1e8/0x220^M
[  346.552555][  T100] [c000000009487e10] [c00000000000cd54] ret_from_kernel_thread+0x5c/0x64^M

After 12 days debugging, I finally narrow the problem to irq_soft_mask_set.

I am a beginner, hope I can be of some beneficial to the community ;-)

Thanks
Zhouyi
--

  arch/powerpc/include/asm/hw_irq.h | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/include/asm/hw_irq.h +1 -1
index 26ede09c521d..a5ae8d82cc9d 100644
@@ -121,7 +121,7 @@ static inline notrace unsigned long irq_soft_mask_return(void)
   * for the critical section and as a clobber because
   * we changed paca->irq_soft_mask
   */
-static inline notrace void irq_soft_mask_set(unsigned long mask)
+static inline notrace __no_kcsan __no_sanitize_address void irq_soft_mask_set(unsigned long mask)
  {
  	/*
  	 * The irq mask must always include the STD bit if any are set.
#2 lore raw
Quoted 26 lines · Christophe Leroy
Quoted 22 lines · Zhouyi Zhou

In ppc, compiler based sanitizer will generate instrument instructions around statement WRITE_ONCE(local_paca->irq_soft_mask, mask):

    0xc000000000295cb0 <+0>:  addis   r2,r12,774
    0xc000000000295cb4 <+4>:  addi    r2,r2,16464
    0xc000000000295cb8 <+8>:  mflr    r0
    0xc000000000295cbc <+12>: bl      0xc00000000008bb4c <mcount>
    0xc000000000295cc0 <+16>: mflr    r0
    0xc000000000295cc4 <+20>: std     r31,-8(r1)
    0xc000000000295cc8 <+24>: addi    r3,r13,2354
    0xc000000000295ccc <+28>: mr      r31,r13
    0xc000000000295cd0 <+32>: std     r0,16(r1)
    0xc000000000295cd4 <+36>: stdu    r1,-48(r1)
    0xc000000000295cd8 <+40>: bl      0xc000000000609b98 <__asan_store1+8>
    0xc000000000295cdc <+44>: nop
    0xc000000000295ce0 <+48>: li      r9,1
    0xc000000000295ce4 <+52>: stb     r9,2354(r31)
    0xc000000000295ce8 <+56>: addi    r1,r1,48
    0xc000000000295cec <+60>: ld      r0,16(r1)
    0xc000000000295cf0 <+64>: ld      r31,-8(r1)
    0xc000000000295cf4 <+68>: mtlr    r0
Quoted 6 lines
Quoted 4 lines

If there is a context switch before "stb r9,2354(r31)", r31 may
not equal to r13, in such case, irq soft mask will not work.

This patch disable sanitizer in irq_soft_mask_set.

Well spotted, thanks.

Thank Christophe for reviewing my patch and your guidance!

Quoted 4 lines

You should add:

Fixes ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers")

OK, I will do it!

Quoted 3 lines

By the way, I think the READ_ONCE() likely has the same issue so you should fix irq_soft_mask_return() at the same time.

Yes, after disassembling irq_soft_mask_return, she has the same issue as irq_soft_mask_set.

In addition, I read hw_irq.h by naked eye to search for removed inline assembly one by one,
I found another place that we could possible enhance (Thank Paul E. McKenny for teaching me use git blame ;-)):
077fc62b2b66a("powerpc/irq: remove inline assembly in hard_irq_disable macro")

arch/powerpc/include/asm/hw_irq.h +1 -3
@@ -282,9 +282,7 @@ static inline bool pmi_irq_pending(void)
        flags = irq_soft_mask_set_return(IRQS_ALL_DISABLED);            \
        local_paca->irq_happened |= PACA_IRQ_HARD_DIS;                  \
        if (!arch_irqs_disabled_flags(flags)) {                         \
-               asm ("stdx %%r1, 0, %1 ;"                               \
-                    : "=m" (local_paca->saved_r1)                      \
-                    : "b" (&local_paca->saved_r1));                    \
+               WRITE_ONCE(local_paca->saved_r1, current_stack_pointer);\
                trace_hardirqs_off();                                   \
        }                                                               \
 } while(0)

I wrap the macro hard_irq_disable into a test function and disassemble it, she has the above issue too:
(gdb) disassemble test002
Dump of assembler code for function test002:

   0xc000000000295db0 <+0>:    addis   r2,r12,774
   0xc000000000295db4 <+4>:    addi    r2,r2,16464
   0xc000000000295db8 <+8>:    mflr    r0
   0xc000000000295dbc <+12>:    bl      0xc00000000008bacc <mcount>
   0xc000000000295dc0 <+16>:    mflr    r0
   0xc000000000295dc4 <+20>:    std     r30,-16(r1)
   0xc000000000295dc8 <+24>:    std     r31,-8(r1)
   0xc000000000295dcc <+28>:    li      r9,2
   0xc000000000295dd0 <+32>:    std     r0,16(r1)
   0xc000000000295dd4 <+36>:    stdu    r1,-48(r1)
   0xc000000000295dd8 <+40>:    mtmsrd  r9,1
   0xc000000000295ddc <+44>:    addi    r3,r13,2354
   0xc000000000295de0 <+48>:    mr      r31,r13
   0xc000000000295de4 <+52>:    bl      0xc000000000609838 <__asan_load1+8>
   0xc000000000295de8 <+56>:    nop
   0xc000000000295dec <+60>:    li      r3,3
   0xc000000000295df0 <+64>:    lbz     r30,2354(r31)
   0xc000000000295df4 <+68>:    bl      0xc00000000028de90 <irq_soft_mask_set>
   0xc000000000295df8 <+72>:    mr      r31,r13
   0xc000000000295dfc <+76>:    addi    r3,r13,2355
   0xc000000000295e00 <+80>:    bl      0xc000000000609838 <__asan_load1+8>
   0xc000000000295e04 <+84>:    nop
   0xc000000000295e08 <+88>:    lbz     r9,2355(r31)
   0xc000000000295e0c <+92>:    andi.   r10,r30,1
   0xc000000000295e10 <+96>:    ori     r9,r9,1
   0xc000000000295e14 <+100>:    stb     r9,2355(r31)
   0xc000000000295e18 <+104>:    bne     0xc000000000295e30 <test002+128>
   0xc000000000295e1c <+108>:    mr      r30,r1
   0xc000000000295e20 <+112>:    addi    r3,r31,2328
   0xc000000000295e24 <+116>:    bl      0xc000000000609dd8 <__asan_store8+8>
   0xc000000000295e28 <+120>:    nop
   0xc000000000295e2c <+124>:    std     r30,2328(r31)
   0xc000000000295e30 <+128>:    addi    r1,r1,48
   0xc000000000295e34 <+132>:    ld      r0,16(r1)
   0xc000000000295e38 <+136>:    ld      r30,-16(r1)
   0xc000000000295e3c <+140>:    ld      r31,-8(r1)
   0xc000000000295e40 <+144>:    mtlr    r0
   0xc000000000295e44 <+148>:    blr

Could we rewrite this macro into a static inline function as irq_soft_mask_set does, and disable sanitizer for it?

Thanks again
Cheers
Zhouyi

Quoted 49 lines
Quoted 49 lines

---
Dear PPC developers

I found this bug when trying to do rcutorture tests in ppc VM of Open Source Lab of Oregon State University following Paul E. McKenny's guidance.

console.log report following bug:

[  346.527467][  T100] BUG: using smp_processor_id() in preemptible [00000000] code: rcu_torture_rea/100^M
[  346.529416][  T100] caller is rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.531157][  T100] CPU: 4 PID: 100 Comm: rcu_torture_rea Tainted: G        W          5.19.0-rc5-next-20220708-dirty #253^M
[  346.533620][  T100] Call Trace:^M
[  346.534449][  T100] [c0000000094876c0] [c000000000ce2b68] dump_stack_lvl+0xbc/0x108 (unreliable)^M
[  346.536632][  T100] [c000000009487710] [c000000001712954] check_preemption_disabled+0x154/0x160^M
[  346.538665][  T100] [c0000000094877a0] [c0000000002ce2d4] rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.540830][  T100] [c0000000094878b0] [c0000000002cf3c0] __rcu_read_unlock+0x290/0x3b0^M
[  346.542746][  T100] [c000000009487910] [c0000000002bb330] rcu_torture_read_unlock+0x30/0xb0^M
[  346.544779][  T100] [c000000009487930] [c0000000002b7ff8] rcutorture_one_extend+0x198/0x810^M
[  346.546851][  T100] [c000000009487a10] [c0000000002b8bfc] rcu_torture_one_read+0x58c/0xc90^M
[  346.548844][  T100] [c000000009487ca0] [c0000000002b942c] rcu_torture_reader+0x12c/0x360^M
[  346.550784][  T100] [c000000009487db0] [c0000000001de978] kthread+0x1e8/0x220^M
[  346.552555][  T100] [c000000009487e10] [c00000000000cd54] ret_from_kernel_thread+0x5c/0x64^M

After 12 days debugging, I finally narrow the problem to irq_soft_mask_set.

I am a beginner, hope I can be of some beneficial to the community ;-)

Thanks
Zhouyi
--

  arch/powerpc/include/asm/hw_irq.h | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
arch/powerpc/include/asm/hw_irq.h +1 -1
index 26ede09c521d..a5ae8d82cc9d 100644
@@ -121,7 +121,7 @@ static inline notrace unsigned long irq_soft_mask_return(void)
   * for the critical section and as a clobber because
   * we changed paca->irq_soft_mask
   */
-static inline notrace void irq_soft_mask_set(unsigned long mask)
+static inline notrace __no_kcsan __no_sanitize_address void irq_soft_mask_set(unsigned long mask)
  {
      /*
       * The irq mask must always include the STD bit if any are set.
#3 lore raw

Le 23/08/2022 à 03:43, Zhouyi Zhou a écrit :

Quoted 68 lines
Quoted 26 lines · Christophe Leroy
Quoted 22 lines · Zhouyi Zhou

In ppc, compiler based sanitizer will generate instrument instructions around statement WRITE_ONCE(local_paca->irq_soft_mask, mask):

     0xc000000000295cb0 <+0>:  addis   r2,r12,774
     0xc000000000295cb4 <+4>:  addi    r2,r2,16464
     0xc000000000295cb8 <+8>:  mflr    r0
     0xc000000000295cbc <+12>: bl      0xc00000000008bb4c <mcount>
     0xc000000000295cc0 <+16>: mflr    r0
     0xc000000000295cc4 <+20>: std     r31,-8(r1)
     0xc000000000295cc8 <+24>: addi    r3,r13,2354
     0xc000000000295ccc <+28>: mr      r31,r13
     0xc000000000295cd0 <+32>: std     r0,16(r1)
     0xc000000000295cd4 <+36>: stdu    r1,-48(r1)
     0xc000000000295cd8 <+40>: bl      0xc000000000609b98 <__asan_store1+8>
     0xc000000000295cdc <+44>: nop
     0xc000000000295ce0 <+48>: li      r9,1
     0xc000000000295ce4 <+52>: stb     r9,2354(r31)
     0xc000000000295ce8 <+56>: addi    r1,r1,48
     0xc000000000295cec <+60>: ld      r0,16(r1)
     0xc000000000295cf0 <+64>: ld      r31,-8(r1)
     0xc000000000295cf4 <+68>: mtlr    r0
Quoted 6 lines
Quoted 4 lines

If there is a context switch before "stb r9,2354(r31)", r31 may
not equal to r13, in such case, irq soft mask will not work.

This patch disable sanitizer in irq_soft_mask_set.

Well spotted, thanks.

Thank Christophe for reviewing my patch and your guidance!

Quoted 4 lines

You should add:

Fixes ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers")

OK, I will do it!

Quoted 3 lines

By the way, I think the READ_ONCE() likely has the same issue so you should fix irq_soft_mask_return() at the same time.

Yes, after disassembling irq_soft_mask_return, she has the same issue as irq_soft_mask_set.

In addition, I read hw_irq.h by naked eye to search for removed inline assembly one by one,
I found another place that we could possible enhance (Thank Paul E. McKenny for teaching me use git blame ;-)):
077fc62b2b66a("powerpc/irq: remove inline assembly in hard_irq_disable macro")

arch/powerpc/include/asm/hw_irq.h +1 -3
@@ -282,9 +282,7 @@ static inline bool pmi_irq_pending(void)
         flags = irq_soft_mask_set_return(IRQS_ALL_DISABLED);            \
         local_paca->irq_happened |= PACA_IRQ_HARD_DIS;                  \
         if (!arch_irqs_disabled_flags(flags)) {                         \
-               asm ("stdx %%r1, 0, %1 ;"                               \
-                    : "=m" (local_paca->saved_r1)                      \
-                    : "b" (&local_paca->saved_r1));                    \
+               WRITE_ONCE(local_paca->saved_r1, current_stack_pointer);\
                 trace_hardirqs_off();                                   \
         }                                                               \
  } while(0)

I wrap the macro hard_irq_disable into a test function and disassemble it, she has the above issue too:
(gdb) disassemble test002
Dump of assembler code for function test002:

...

Quoted 2 lines

Could we rewrite this macro into a static inline function as irq_soft_mask_set does, and disable sanitizer for it?

Yes we could try to do that, hoping it will not bring too much troubles with circular header inclusion.

Another possible approach could be to create a WRITE_ONCE_NOCHECK() more or less similar to existing READ_ONCE_NOCHECK().

We could also change READ_ONCE_NOCHECK() to accept bytes size then use it for irq_soft_mask_return() .

Christophe

#4 lore raw
In reply to Zhouyi Zhou
Quoted 51 lines · Zhouyi Zhou

In ppc, compiler based sanitizer will generate instrument instructions around statement WRITE_ONCE(local_paca->irq_soft_mask, mask):

   0xc000000000295cb0 <+0>:	addis   r2,r12,774
   0xc000000000295cb4 <+4>:	addi    r2,r2,16464
   0xc000000000295cb8 <+8>:	mflr    r0
   0xc000000000295cbc <+12>:	bl      0xc00000000008bb4c <mcount>
   0xc000000000295cc0 <+16>:	mflr    r0
   0xc000000000295cc4 <+20>:	std     r31,-8(r1)
   0xc000000000295cc8 <+24>:	addi    r3,r13,2354
   0xc000000000295ccc <+28>:	mr      r31,r13
   0xc000000000295cd0 <+32>:	std     r0,16(r1)
   0xc000000000295cd4 <+36>:	stdu    r1,-48(r1)
   0xc000000000295cd8 <+40>:	bl      0xc000000000609b98 <__asan_store1+8>
   0xc000000000295cdc <+44>:	nop
   0xc000000000295ce0 <+48>:	li      r9,1
   0xc000000000295ce4 <+52>:	stb     r9,2354(r31)
   0xc000000000295ce8 <+56>:	addi    r1,r1,48
   0xc000000000295cec <+60>:	ld      r0,16(r1)
   0xc000000000295cf0 <+64>:	ld      r31,-8(r1)
   0xc000000000295cf4 <+68>:	mtlr    r0

If there is a context switch before "stb r9,2354(r31)", r31 may
not equal to r13, in such case, irq soft mask will not work.

This patch disable sanitizer in irq_soft_mask_set.

---
Dear PPC developers

I found this bug when trying to do rcutorture tests in ppc VM of Open Source Lab of Oregon State University following Paul E. McKenny's guidance.

console.log report following bug:

[  346.527467][  T100] BUG: using smp_processor_id() in preemptible [00000000] code: rcu_torture_rea/100^M
[  346.529416][  T100] caller is rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.531157][  T100] CPU: 4 PID: 100 Comm: rcu_torture_rea Tainted: G        W          5.19.0-rc5-next-20220708-dirty #253^M
[  346.533620][  T100] Call Trace:^M
[  346.534449][  T100] [c0000000094876c0] [c000000000ce2b68] dump_stack_lvl+0xbc/0x108 (unreliable)^M
[  346.536632][  T100] [c000000009487710] [c000000001712954] check_preemption_disabled+0x154/0x160^M
[  346.538665][  T100] [c0000000094877a0] [c0000000002ce2d4] rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.540830][  T100] [c0000000094878b0] [c0000000002cf3c0] __rcu_read_unlock+0x290/0x3b0^M
[  346.542746][  T100] [c000000009487910] [c0000000002bb330] rcu_torture_read_unlock+0x30/0xb0^M
[  346.544779][  T100] [c000000009487930] [c0000000002b7ff8] rcutorture_one_extend+0x198/0x810^M
[  346.546851][  T100] [c000000009487a10] [c0000000002b8bfc] rcu_torture_one_read+0x58c/0xc90^M
[  346.548844][  T100] [c000000009487ca0] [c0000000002b942c] rcu_torture_reader+0x12c/0x360^M
[  346.550784][  T100] [c000000009487db0] [c0000000001de978] kthread+0x1e8/0x220^M
[  346.552555][  T100] [c000000009487e10] [c00000000000cd54] ret_from_kernel_thread+0x5c/0x64^M
Quoted 1 line

After 12 days debugging, I finally narrow the problem to irq_soft_mask_set.

Thanks for spending 12 days debugging it! O_o

arch/powerpc/include/asm/hw_irq.h static inline notrace unsigned long irq_soft_mask_return(void) · line 127

Quoted 13 lines
arch/powerpc/include/asm/hw_irq.h +1 -1
index 26ede09c521d..a5ae8d82cc9d 100644
@@ -121,7 +121,7 @@ static inline notrace unsigned long irq_soft_mask_return(void)
  * for the critical section and as a clobber because
  * we changed paca->irq_soft_mask
  */
-static inline notrace void irq_soft_mask_set(unsigned long mask)
+static inline notrace __no_kcsan __no_sanitize_address void irq_soft_mask_set(unsigned long mask)
 {
 	/*
 	 * The irq mask must always include the STD bit if any are set.

My worry is that this will force irq_soft_mask_set() out of line, which we would rather avoid. It's meant to be a fast path.

In fact with this applied I see nearly 300 out-of-line copies of the function when building a defconfig, and ~1700 calls to it.

Normally it is inlined at every call site.

So I think I'm inclined to revert ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers").

It was a nice looking cleanup, but those loads must not be instrumented by KASAN, but we also want them inlined, and AFAICS the only way to achieve that is to go back to inline asm.

cheers

#5 lore raw

Le 23/08/2022 à 10:33, Michael Ellerman a écrit :

Quoted 73 lines
Quoted 51 lines · Zhouyi Zhou

In ppc, compiler based sanitizer will generate instrument instructions around statement WRITE_ONCE(local_paca->irq_soft_mask, mask):

    0xc000000000295cb0 <+0>:	addis   r2,r12,774
    0xc000000000295cb4 <+4>:	addi    r2,r2,16464
    0xc000000000295cb8 <+8>:	mflr    r0
    0xc000000000295cbc <+12>:	bl      0xc00000000008bb4c <mcount>
    0xc000000000295cc0 <+16>:	mflr    r0
    0xc000000000295cc4 <+20>:	std     r31,-8(r1)
    0xc000000000295cc8 <+24>:	addi    r3,r13,2354
    0xc000000000295ccc <+28>:	mr      r31,r13
    0xc000000000295cd0 <+32>:	std     r0,16(r1)
    0xc000000000295cd4 <+36>:	stdu    r1,-48(r1)
    0xc000000000295cd8 <+40>:	bl      0xc000000000609b98 <__asan_store1+8>
    0xc000000000295cdc <+44>:	nop
    0xc000000000295ce0 <+48>:	li      r9,1
    0xc000000000295ce4 <+52>:	stb     r9,2354(r31)
    0xc000000000295ce8 <+56>:	addi    r1,r1,48
    0xc000000000295cec <+60>:	ld      r0,16(r1)
    0xc000000000295cf0 <+64>:	ld      r31,-8(r1)
    0xc000000000295cf4 <+68>:	mtlr    r0

If there is a context switch before "stb r9,2354(r31)", r31 may
not equal to r13, in such case, irq soft mask will not work.

This patch disable sanitizer in irq_soft_mask_set.

---
Dear PPC developers

I found this bug when trying to do rcutorture tests in ppc VM of Open Source Lab of Oregon State University following Paul E. McKenny's guidance.

console.log report following bug:

[  346.527467][  T100] BUG: using smp_processor_id() in preemptible [00000000] code: rcu_torture_rea/100^M
[  346.529416][  T100] caller is rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.531157][  T100] CPU: 4 PID: 100 Comm: rcu_torture_rea Tainted: G        W          5.19.0-rc5-next-20220708-dirty #253^M
[  346.533620][  T100] Call Trace:^M
[  346.534449][  T100] [c0000000094876c0] [c000000000ce2b68] dump_stack_lvl+0xbc/0x108 (unreliable)^M
[  346.536632][  T100] [c000000009487710] [c000000001712954] check_preemption_disabled+0x154/0x160^M
[  346.538665][  T100] [c0000000094877a0] [c0000000002ce2d4] rcu_preempt_deferred_qs_irqrestore+0x74/0xed0^M
[  346.540830][  T100] [c0000000094878b0] [c0000000002cf3c0] __rcu_read_unlock+0x290/0x3b0^M
[  346.542746][  T100] [c000000009487910] [c0000000002bb330] rcu_torture_read_unlock+0x30/0xb0^M
[  346.544779][  T100] [c000000009487930] [c0000000002b7ff8] rcutorture_one_extend+0x198/0x810^M
[  346.546851][  T100] [c000000009487a10] [c0000000002b8bfc] rcu_torture_one_read+0x58c/0xc90^M
[  346.548844][  T100] [c000000009487ca0] [c0000000002b942c] rcu_torture_reader+0x12c/0x360^M
[  346.550784][  T100] [c000000009487db0] [c0000000001de978] kthread+0x1e8/0x220^M
[  346.552555][  T100] [c000000009487e10] [c00000000000cd54] ret_from_kernel_thread+0x5c/0x64^M
Quoted 1 line

After 12 days debugging, I finally narrow the problem to irq_soft_mask_set.

Thanks for spending 12 days debugging it! O_o

arch/powerpc/include/asm/hw_irq.h static inline notrace unsigned long irq_soft_mask_return(void) · line 127

Quoted 13 lines
arch/powerpc/include/asm/hw_irq.h +1 -1
index 26ede09c521d..a5ae8d82cc9d 100644
@@ -121,7 +121,7 @@ static inline notrace unsigned long irq_soft_mask_return(void)
   * for the critical section and as a clobber because
   * we changed paca->irq_soft_mask
   */
-static inline notrace void irq_soft_mask_set(unsigned long mask)
+static inline notrace __no_kcsan __no_sanitize_address void irq_soft_mask_set(unsigned long mask)
  {
  	/*
  	 * The irq mask must always include the STD bit if any are set.

My worry is that this will force irq_soft_mask_set() out of line, which we would rather avoid. It's meant to be a fast path.

Quoted 8 lines

In fact with this applied I see nearly 300 out-of-line copies of the function when building a defconfig, and ~1700 calls to it.

Normally it is inlined at every call site.

So I think I'm inclined to revert ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers").

Could you revert it only partially ? In extenso, revert the READ/WRITE_ONCE and bring back the inline asm in irq_soft_mask_return()
and irq_soft_mask_set(), but keep other changes.

Quoted 5 lines

It was a nice looking cleanup, but those loads must not be instrumented by KASAN, but we also want them inlined, and AFAICS the only way to achieve that is to go back to inline asm.

It's a pitty.

Would it be acceptable to have it out of line when KASAN is selected and inline otherwise ? In that case there is __no_sanitize_or_inline

Christophe

#6 lore raw

Le 23/08/2022 à 10:47, Christophe Leroy a écrit :

Quoted 11 lines

Le 23/08/2022 à 10:33, Michael Ellerman a écrit :

Quoted 8 lines

Zhouyi Zhou <[email protected]> writes:

My worry is that this will force irq_soft_mask_set() out of line, which we would rather avoid. It's meant to be a fast path.

In fact with this applied I see nearly 300 out-of-line copies of the function when building a defconfig, and ~1700 calls to it.

Quoted 8 lines
Quoted 5 lines

Normally it is inlined at every call site.

So I think I'm inclined to revert ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers").

Could you revert it only partially ? In extenso, revert the READ/WRITE_ONCE and bring back the inline asm in irq_soft_mask_return()
 and irq_soft_mask_set(), but keep other changes.

I sent a patch doing that.

Christophe

#7 lore raw
Quoted 18 lines · Christophe Leroy
Quoted 14 lines · Christophe Leroy
Quoted 11 lines · Michael Ellerman

Zhouyi Zhou <[email protected]> writes:

My worry is that this will force irq_soft_mask_set() out of line, which we would rather avoid. It's meant to be a fast path.

In fact with this applied I see nearly 300 out-of-line copies of the function when building a defconfig, and ~1700 calls to it.

Normally it is inlined at every call site.

Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

So I think I'm inclined to revert ef5b570d3700 ("powerpc/irq: Don't open code irq_soft_mask helpers").

Could you revert it only partially ? In extenso, revert the READ/WRITE_ONCE and bring back the inline asm in irq_soft_mask_return()
and irq_soft_mask_set(), but keep other changes.

I sent a patch doing that.

Thank Christophe for the fix. I am very glad to be of benefit to the community ;-)
Also thank Michael and Paul for your constant encouragement and guidance, I learned to use objdump to count the number of failed inline function calls today ;-)

By the way, from my experiments, both gcc-11 and clang-14 behave the same as Michael has described.

Cheers
Zhouyi

Quoted 2 lines

Christophe