Skip to contentSkip to conversation

hazptrtorture: Fix inverted sleep condition in do_pending kthread

PATCH In review Bradley Morgan 4 participants 18 messages Aug 4 – Sep 7 RCU

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 hazptrtorture: Fix inverted sleep condition in do_pending kthread

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and thats false while the test runs. So the kthread never sleeps. It just loops. It drains the pending lists over and over, no pause, and burns a full CPU for the whole test.

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does one last cleanup and exits. Nothing left to do, so no sleep needed then.

Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release deferred hazard pointers")
---
 kernel/rcu/hazptrtorture.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +1 -1
index 72aeb50668e9..7c8b5899fb01 100644
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
 			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
 			hazptr_torture_do_one_pending(cpu, &rand);
 		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
 			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
 		// Omit stutter_wait() because this function needs to do cleanup.
 	} while (!torture_must_stop());
Signature · 1 line
2.47.3
maintainer
#1 lore raw
Quoted 7 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and thats false while the test runs. So the kthread never sleeps. It just loops. It drains the pending lists over and over, no pause, and burns a full CPU for the whole test.

Quoted 6 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does one last cleanup and exits. Nothing left to do, so no sleep needed then.

Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release deferred hazard pointers")

Good catch, applied, thank you!

Thanx, Paul

Quoted 20 lines
---
 kernel/rcu/hazptrtorture.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +1 -1
index 72aeb50668e9..7c8b5899fb01 100644
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
 			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
 			hazptr_torture_do_one_pending(cpu, &rand);
 		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
 			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
 		// Omit stutter_wait() because this function needs to do cleanup.
 	} while (!torture_must_stop());
Signature · 1 line
2.47.3
#2 lore raw
In reply to Bradley Morgan
Quoted 6 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and thats false while the test runs. So the kthread never sleeps. It just loops. It drains the pending lists over and over, no pause, and burns a full CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe relevant for you ?

Thanks,

Mathieu

Quoted 24 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does one last cleanup and exits. Nothing left to do, so no sleep needed then.

Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release deferred hazard pointers")
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +1 -1
index 72aeb50668e9..7c8b5899fb01 100644
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
maintainer
#3 lore raw
Quoted 10 lines · Mathieu Desnoyers
Quoted 6 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and thats false while the test runs. So the kthread never sleeps. It just loops. It drains the pending lists over and over, no pause, and burns a full CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe relevant for you ?

We got it in my -rcu tree:

15b9fca2f5ce ("hazptrtorture: Fix inverted sleep condition in do_pending kthread")

But I do appreciate your checking for missed patches, thank you!

Thanx, Paul

Quoted 34 lines

Thanks,

Mathieu

Quoted 24 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does one last cleanup and exits. Nothing left to do, so no sleep needed then.

Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release deferred hazard pointers")
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +1 -1
index 72aeb50668e9..7c8b5899fb01 100644
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
#4 lore raw
Quoted 26 lines · Paul E. McKenney
Quoted 10 lines · Mathieu Desnoyers
Quoted 6 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and thats false while the test runs. So the kthread never sleeps. It just loops. It drains the pending lists over and over, no pause, and burns a full CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe relevant for you ?

We got it in my -rcu tree:

15b9fca2f5ce ("hazptrtorture: Fix inverted sleep condition in do_pending kthread")

But I do appreciate your checking for missed patches, thank you!

Thanx, Paul

Quoted 5 lines

Thanks,

Mathieu

Quoted 1 line
Quoted 32 lines
Quoted 5 lines
Quoted 5 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does one last cleanup and exits. Nothing left to do, so no sleep needed then.

Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release deferred

hazard pointers")

Quoted 24 lines
Quoted 18 lines
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +1 -1
index 72aeb50668e9..7c8b5899fb01 100644
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com

Hi, new email.

Could I ask where the heck hazptr is? Its not in next, would you like, help???

--- Thanks!
https://lore.kernel.org/all/[email protected]/

maintainer
#5 lore raw
Quoted 60 lines · Bradley Morgan
Quoted 26 lines · Paul E. McKenney
Quoted 10 lines · Mathieu Desnoyers
Quoted 6 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and thats false while the test runs. So the kthread never sleeps. It just loops. It drains the pending lists over and over, no pause, and burns a full CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe relevant for you ?

We got it in my -rcu tree:

15b9fca2f5ce ("hazptrtorture: Fix inverted sleep condition in do_pending kthread")

But I do appreciate your checking for missed patches, thank you!

Thanx, Paul

Quoted 5 lines

Thanks,

Mathieu

Quoted 1 line
Quoted 32 lines
Quoted 5 lines
Quoted 5 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does one last cleanup and exits. Nothing left to do, so no sleep needed then.

Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release deferred

hazard pointers")

Quoted 24 lines
Quoted 18 lines
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +1 -1
index 72aeb50668e9..7c8b5899fb01 100644
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
Quoted 7 lines

Hi, new email.

Could I ask where the heck hazptr is? Its not in next, would you like, help???

--- Thanks!
https://lore.kernel.org/all/[email protected]/

It is still in my -rcu tree.

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard hazard pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

Thanx, Paul

#6 lore raw
Quoted 83 lines · Paul E. McKenney
Quoted 1 line · Bradley Morgan

On 4 September 2026 17:00:22 BST, "Paul E. McKenney"

<[email protected]>

Quoted 6 lines
Quoted 5 lines · wrote:
Quoted 4 lines · Mathieu Desnoyers
Quoted 3 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and

thats

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

false while the test runs. So the kthread never sleeps. It just

loops.

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

It drains the pending lists over and over, no pause, and burns a

full

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
Quoted 1 line

CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe relevant for you ?

We got it in my -rcu tree:

15b9fca2f5ce ("hazptrtorture: Fix inverted sleep condition in

do_pending

Quoted 6 lines
Quoted 6 lines

kthread")

But I do appreciate your checking for missed patches, thank you!

Thanx, Paul

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Thanks,

Mathieu

Quoted 3 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does

one

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

last cleanup and exits. Nothing left to do, so no sleep needed

then.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release

deferred

Quoted 7 lines
Quoted 7 lines

hazard pointers")

Quoted 6 lines

Review

Quoted 6 lines
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/kernel/rcu/hazptrtorture.c

Quoted 20 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines

index 72aeb50668e9..7c8b5899fb01 100644

kernel/rcu/hazptrtorture.c +1 -1
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
Quoted 10 lines
Quoted 8 lines
Quoted 2 lines
kernel/rcu/hazptrtorture.c +0 -0
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
Quoted 7 lines

Hi, new email.

Could I ask where the heck hazptr is? Its not in next, would you like, help???

--- Thanks!

https://lore.kernel.org/all/[email protected]/

It is still in my -rcu tree.

I would not say "no" to a fix for this issue:

Quoted 7 lines

https://lore.kernel.org/all/[email protected]/

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard hazard pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

Well, okay, tell me more and I'll develop it.

Quoted 2 lines

Thanx, Paul

--- Thanks!
https://lore.kernel.org/all/[email protected]/

maintainer
#7 lore raw
Quoted 87 lines · Bradley Morgan
Quoted 85 lines · Paul E. McKenney
Quoted 1 line · Bradley Morgan

On 4 September 2026 17:00:22 BST, "Paul E. McKenney"

<[email protected]>

Quoted 6 lines
Quoted 5 lines · wrote:
Quoted 4 lines · Mathieu Desnoyers
Quoted 3 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the docs say so too. But the sleep sits behind torture_must_stop(), and

thats

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

false while the test runs. So the kthread never sleeps. It just

loops.

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

It drains the pending lists over and over, no pause, and burns a

full

Quoted 8 lines
Quoted 8 lines
Quoted 4 lines
Quoted 1 line

CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe relevant for you ?

We got it in my -rcu tree:

15b9fca2f5ce ("hazptrtorture: Fix inverted sleep condition in

do_pending

Quoted 6 lines
Quoted 6 lines

kthread")

But I do appreciate your checking for missed patches, thank you!

Thanx, Paul

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Thanks,

Mathieu

Quoted 3 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it does

one

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

last cleanup and exits. Nothing left to do, so no sleep needed

then.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release

deferred

Quoted 7 lines
Quoted 7 lines

hazard pointers")

Quoted 6 lines

Review

Quoted 6 lines
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
(quoted diff) +0 -0

b/kernel/rcu/hazptrtorture.c

Quoted 20 lines
Quoted 10 lines
Quoted 10 lines
Quoted 10 lines

index 72aeb50668e9..7c8b5899fb01 100644

kernel/rcu/hazptrtorture.c +1 -1
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void *arg)
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
Quoted 10 lines
Quoted 8 lines
Quoted 2 lines
kernel/rcu/hazptrtorture.c +0 -0
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
Quoted 7 lines

Hi, new email.

Could I ask where the heck hazptr is? Its not in next, would you like, help???

--- Thanks!

https://lore.kernel.org/all/[email protected]/

It is still in my -rcu tree.

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 7 lines
Quoted 5 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard hazard pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

Well, okay, tell me more and I'll develop it.

I must confess that I have not looked much into that bug report, so I don't have much more to tell.

Thanx, Paul

Quoted 3 lines

--- Thanks!
https://lore.kernel.org/all/[email protected]/

#8 lore raw
Quoted 99 lines · Paul E. McKenney
Quoted 1 line · Bradley Morgan

On 5 September 2026 21:40:28 BST, "Paul E. McKenney"

<[email protected]>

Quoted 9 lines
Quoted 8 lines · wrote:
Quoted 1 line · Bradley Morgan

On 4 September 2026 17:00:22 BST, "Paul E. McKenney"

<[email protected]>

Quoted 5 lines
Quoted 4 lines · wrote:
Quoted 3 lines · Mathieu Desnoyers
Quoted 2 lines · Bradley Morgan

The do_pending kthread never sleeps. Its supposed to. kthread_do_pending_ms sets the gap between cleanups, and the

docs

Quoted 9 lines
Quoted 9 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

say so too. But the sleep sits behind torture_must_stop(), and

thats

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

false while the test runs. So the kthread never sleeps. It just

loops.

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

It drains the pending lists over and over, no pause, and burns a

full

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 1 line

CPU for the whole test.

Paul, I just found this patch in my junk email folder. Maybe

relevant

Quoted 12 lines
Quoted 12 lines
Quoted 5 lines
Quoted 5 lines
Quoted 1 line

for you ?

We got it in my -rcu tree:

15b9fca2f5ce ("hazptrtorture: Fix inverted sleep condition in

do_pending

Quoted 6 lines
Quoted 6 lines

kthread")

But I do appreciate your checking for missed patches, thank you!

Thanx, Paul

Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines

Thanks,

Mathieu

Quoted 3 lines

Flip the check. With !torture_must_stop() it sleeps for kthread_do_pending_ms between passes. When the test stops it

does

Quoted 12 lines
Quoted 12 lines

one

Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

last cleanup and exits. Nothing left to do, so no sleep needed

then.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Fixes 94d2e93c222e ("hazptrtorture: Add kthread to release

deferred

Quoted 6 lines
Quoted 6 lines

hazard pointers")

Quoted 5 lines
Quoted 5 lines
---
  kernel/rcu/hazptrtorture.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 6 lines
Quoted 6 lines
Quoted 1 line
Quoted 1 line
Quoted 1 line

Review

Quoted 1 line
(quoted diff) +0 -0

b/kernel/rcu/hazptrtorture.c

Quoted 4 lines
Quoted 4 lines
Quoted 4 lines

kernel/rcu/hazptrtorture.c static int hazptr_torture_do_pending(void

Quoted 4 lines

index 72aeb50668e9..7c8b5899fb01 100644

kernel/rcu/hazptrtorture.c +0 -0
@@ -555,7 +555,7 @@ static int hazptr_torture_do_pending(void

*arg)

Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 16 lines
Quoted 14 lines
Quoted 8 lines
kernel/rcu/hazptrtorture.c +1 -1
  			cpu = cpumask_next_wrap(cpu, cpu_possible_mask);
  			hazptr_torture_do_one_pending(cpu, &rand);
  		}
-		if (torture_must_stop())
+		if (!torture_must_stop())
  			torture_hrtimeout_ms(kthread_do_pending_ms, USEC_PER_MSEC, &rand);
  		// Omit stutter_wait() because this function needs to do cleanup.
  	} while (!torture_must_stop());
Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

Hi, new email.

Could I ask where the heck hazptr is? Its not in next, would you

like,

Quoted 5 lines
Quoted 4 lines
Quoted 4 lines

help???

--- Thanks!

Quoted 12 lines

https://lore.kernel.org/all/[email protected]/

Quoted 5 lines

It is still in my -rcu tree.

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 4 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard

Quoted 8 lines

hazard

Quoted 4 lines
Quoted 2 lines

pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

Well, okay, tell me more and I'll develop it.

I must confess that I have not looked much into that bug report, so I don't have much more to tell.

No, like the getting rid of wildcard hazptrs?

Quoted 7 lines

Thanx, Paul

Quoted 2 lines

--- Thanks!

https://lore.kernel.org/all/[email protected]/

Quoted 1 line

--- Thanks!
https://lore.kernel.org/all/[email protected]/

#9 lore raw
In reply to Paul E. McKenney
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

I'm not sure this URL actually points to a relevant issue ?

Quoted 6 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard hazard pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See: https://lore.kernel.org/all/[email protected]/

Thanks,

Mathieu

Signature · 3 lines
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
#10 lore raw
Quoted 10 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

I'm not sure this URL actually points to a relevant issue ?

No, that URL is my signature, just in case people

1: add my old email to R-Bs, or SOBs.

2: makes people aware this is me.

Quoted 17 lines
Quoted 6 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard hazard pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

Quoted 7 lines

See: https://lore.kernel.org/all/[email protected]/

Thanks,

Mathieu

--- Thanks!
https://lore.kernel.org/all/[email protected]/

#11 lore raw
Quoted 17 lines · Bradley Morgan
Quoted 10 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

I'm not sure this URL actually points to a relevant issue ?

No, that URL is my signature, just in case people

1: add my old email to R-Bs, or SOBs.

Quoted 20 lines

2: makes people aware this is me.

Quoted 4 lines
Quoted 4 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard

hazard

Quoted 13 lines
Quoted 2 lines

pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

Sure, I can help write it if you want

Quoted 11 lines
Quoted 1 line

See:

https://lore.kernel.org/all/[email protected]/

Quoted 6 lines

Thanks,

Mathieu

--- Thanks!
https://lore.kernel.org/all/[email protected]/

--- Thanks!
https://lore.kernel.org/all/[email protected]/

maintainer
#12 lore raw
In reply to Mathieu Desnoyers
Quoted 8 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

I'm not sure this URL actually points to a relevant issue ?

Indeed, it does not, apologies! Here you go:

https://lore.kernel.org/all/[email protected]/

Quoted 9 lines
Quoted 5 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard hazard pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Quoted 8 lines

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See: https://lore.kernel.org/all/[email protected]/

The main objection was over the content and style of the kernel-doc header comment, right? I am guessing that it should be possible to resolve this to roughly equal disgust of all concerned. ;-)

We did make some progress on this sort of pointer issue in C++29 this past June:

https://people.kernel.org/paulmck/c-pointer-zap-and-oota-progress

But the piece you need is this guy, which is still in process:

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3790r1.pdf

Plus it will be some time before this reaches all the compilers used to build the Linux kernel, and probably even more time to reach the C language. I do have pen-on-paper notes that will lead to a draft of the corresponding C-language working paper, but these things do not move quickly.

So, yes, we will need something like ptr_eq() for some years to come.

Back to your original question, given the fix for the above bug and given the current use case, I believe we can get the current series into mainline. Give or take Linus's thoughts on the matter. But either way, we will need a version that allows the user to avoid all wildcard use sooner rather than later.

So having a series on top of the current one for a later merge window would be a very good thing!

Thanx, Paul

#13 lore raw
Quoted 62 lines · Paul E. McKenney
Quoted 6 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 2 lines

I'm not sure this URL actually points to a relevant issue ?

Indeed, it does not, apologies! Here you go:

https://lore.kernel.org/all/[email protected]/

Quoted 3 lines
Quoted 3 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard

hazard

Quoted 6 lines
Quoted 2 lines

pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Quoted 8 lines

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See:

https://lore.kernel.org/all/[email protected]/

The main objection was over the content and style of the kernel-doc header comment, right? I am guessing that it should be possible to resolve this to roughly equal disgust of all concerned. ;-)

We did make some progress on this sort of pointer issue in C++29 this past June:

https://people.kernel.org/paulmck/c-pointer-zap-and-oota-progress

But the piece you need is this guy, which is still in process:

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3790r1.pdf

Plus it will be some time before this reaches all the compilers used to build the Linux kernel, and probably even more time to reach the C language. I do have pen-on-paper notes that will lead to a draft of the corresponding C-language working paper, but these things do not move quickly.

So, yes, we will need something like ptr_eq() for some years to come.

Back to your original question, given the fix for the above bug and given the current use case, I believe we can get the current series into mainline. Give or take Linus's thoughts on the matter. But either way, we will need a version that allows the user to avoid all wildcard use sooner rather than later.

Quoted 4 lines

So having a series on top of the current one for a later merge window would be a very good thing!

Thanx, Paul

Can I participate in this? :)
--- Thanks!
https://lore.kernel.org/all/[email protected]/

maintainer
#14 lore raw
Quoted 64 lines · Bradley Morgan
Quoted 62 lines · Paul E. McKenney
Quoted 6 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 2 lines

I'm not sure this URL actually points to a relevant issue ?

Indeed, it does not, apologies! Here you go:

https://lore.kernel.org/all/[email protected]/

Quoted 3 lines
Quoted 3 lines

Once that is in place, I would be happy to put this back into -next.

At some point, we will need to get rid of the concept of wildcard

hazard

Quoted 6 lines
Quoted 2 lines

pointers, as those end up instead emulating RCU, but I don't see that as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Quoted 8 lines

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See:

https://lore.kernel.org/all/[email protected]/

The main objection was over the content and style of the kernel-doc header comment, right? I am guessing that it should be possible to resolve this to roughly equal disgust of all concerned. ;-)

We did make some progress on this sort of pointer issue in C++29 this past June:

https://people.kernel.org/paulmck/c-pointer-zap-and-oota-progress

But the piece you need is this guy, which is still in process:

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3790r1.pdf

Plus it will be some time before this reaches all the compilers used to build the Linux kernel, and probably even more time to reach the C language. I do have pen-on-paper notes that will lead to a draft of the corresponding C-language working paper, but these things do not move quickly.

So, yes, we will need something like ptr_eq() for some years to come.

Back to your original question, given the fix for the above bug and given the current use case, I believe we can get the current series into mainline. Give or take Linus's thoughts on the matter. But either way, we will need a version that allows the user to avoid all wildcard use sooner rather than later.

Quoted 4 lines
Quoted 3 lines

So having a series on top of the current one for a later merge window would be a very good thing!

Can I participate in this? :)

If Mathieu is OK with it, feel free to look at the patch stack that Mathieu sent the URL for earlier in this thread. Either way, please feel free to look at the stack in my -rcu tree based on v7.3-rc1 and headed by this commit:

4398b7c192d ("hazptr: Implement two-phase wildcard scan")

Perhaps you can find the bug that kernel test robot located. ;-)

My -rcu tree is here:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git

Just so you know, in all cases, your taking on a task does not preclude others from also taking that same task on.

Thanx, Paul

Quoted 2 lines

--- Thanks!
https://lore.kernel.org/all/[email protected]/

#15 lore raw
Quoted 90 lines · Paul E. McKenney
Quoted 1 line · Bradley Morgan

On 6 September 2026 19:46:50 BST, "Paul E. McKenney"

<[email protected]>

Quoted 18 lines
Quoted 7 lines · wrote:
Quoted 6 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 8 lines
Quoted 2 lines

I'm not sure this URL actually points to a relevant issue ?

Indeed, it does not, apologies! Here you go:

https://lore.kernel.org/all/[email protected]/

Quoted 1 line
Quoted 1 line

Once that is in place, I would be happy to put this back into

-next.

Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines

At some point, we will need to get rid of the concept of wildcard

hazard

Quoted 1 line
Quoted 1 line

pointers, as those end up instead emulating RCU, but I don't see

that

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 1 line

as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Do you want me to resurrect it on top of the current series ? This depends on:

Quoted 31 lines
Quoted 5 lines
Quoted 5 lines

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See:

https://lore.kernel.org/all/[email protected]/

Quoted 24 lines

The main objection was over the content and style of the kernel-doc header comment, right? I am guessing that it should be possible to resolve this to roughly equal disgust of all concerned. ;-)

We did make some progress on this sort of pointer issue in C++29 this past June:

https://people.kernel.org/paulmck/c-pointer-zap-and-oota-progress

But the piece you need is this guy, which is still in process:

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3790r1.pdf

Plus it will be some time before this reaches all the compilers used to build the Linux kernel, and probably even more time to reach the C language. I do have pen-on-paper notes that will lead to a draft of the corresponding C-language working paper, but these things do not move quickly.

So, yes, we will need something like ptr_eq() for some years to come.

Back to your original question, given the fix for the above bug and given the current use case, I believe we can get the current series

into

Quoted 1 line
Quoted 1 line

mainline. Give or take Linus's thoughts on the matter. But either

way,

Quoted 7 lines
Quoted 6 lines

we will need a version that allows the user to avoid all wildcard use sooner rather than later.

So having a series on top of the current one for a later merge window would be a very good thing!

Can I participate in this? :)

If Mathieu is OK with it, feel free to look at the patch stack that Mathieu sent the URL for earlier in this thread. Either way, please feel free to look at the stack in my -rcu tree based on v7.3-rc1 and headed by this commit:

4398b7c192d ("hazptr: Implement two-phase wildcard scan")

Perhaps you can find the bug that kernel test robot located. ;-)

My -rcu tree is here:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git

Quoted 8 lines

Just so you know, in all cases, your taking on a task does not preclude others from also taking that same task on.

Thanx, Paul

Quoted 2 lines

--- Thanks!

https://lore.kernel.org/all/[email protected]/

Hey, test this fix?

From 3e92b8153c31106d6a080e1c9bbe9bf1e86e1f63 Mon Sep 17 00:00:00 2001
From: Bradley Morgan <[email protected]>
Date: Mon, 7 Sep 2026 18:00:08 +0000
Subject: [PATCH] hazptrtorture: Only detach acquired hazard pointers

hazptr_torture_acquire() detaches unconditionally, even when the readlock fails. A failed acquire leaves nothing to detach, but the detach still promotes the context to its backup slot and chains that slot into the running CPU's overflow list. The reader then retries on its own CPU, the fast path hands out a per-CPU slot and overwrites ctx->slot, and the chained backup node is orphaned, still linked, with nobody left to unchain it.

The next detach of the same context chains the same node a second time, into another CPU's list, and the node ends up reachable from both. The eventual release unchains it once, hlist_del() poisons node->next, and the first list is left pointing at the poisoned node. The writer's next hazptr_synchronize() walks that list, steps onto LIST_POISON1 (0x100 on i386, where POISON_POINTER_DELTA is 0), and reads slot.addr at offset 8 of the backup slot, address 0x108, which is the crash the robot hit.

  cpuA (IPI acquire)     cpuR (reader)          cpuD (do_pending)
  ---------------------  ---------------------  -------------------
  readlock() returns
    NULL
  detach chains the
    backup node into
    cpuA list
                         hpp_htp is NULL,
                           continue
                         reacquire, ctx->slot
                           is now a cpuR
                           per-CPU slot
                         acquire succeeds,
                           defer, detach chains
                           the SAME node into
                           cpuR list
                                                release, unchain
                                                  once, node->next
                                                  is POISON1
                                                kfree(hppp)
  synchronize walks cpuA
    list, node->next is
    0x100, reads 0x108,
    Oops

Skip the detach when the acquire failed. The slot holds NULL in that case, note_context_switch() and the synchronize scanners skip NULL slots, and the next acquire overwrites ctx->slot, so leaving the context attached is safe.

The robot's original report was against the defer path before detach existed, which 4bd7f458229a fixed. This is the same crash surviving through the IPI acquire path that 6357ec235c59 added.

Fixes 6357ec235c59 ("hazptrtorture: Fix hazptr ownership issue")
---
 kernel/rcu/hazptrtorture.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +4 -1
index 7c8b589..267f262 100644
@@ -373,8 +373,11 @@ static void hazptr_torture_acquire(void *hppp_in)
 	/*
 	 * Acquiring a hazard pointer from a remote CPU.
 	 * Detach hazptr from its task so it can be released by another task.
+	 * A failed acquire has nothing to detach, and detaching one anyway
+	 * orphans the chained backup slot on this CPU's overflow list.
 	 */
-	hazptr_detach(&hppp->hpp_hc);
+	if (hppp->hpp_htp)
+		hazptr_detach(&hppp->hpp_hc);
 	atomic_long_inc(per_cpu_ptr(&hazptr_torture_acquires_irq, raw_smp_processor_id()));
 }
 
Signature · 5 lines
2.47.3


--- Thanks!
https://lore.kernel.org/all/[email protected]/
maintainer
#16 lore raw
Quoted 95 lines · Bradley Morgan
Quoted 93 lines · Paul E. McKenney
Quoted 1 line · Bradley Morgan

On 6 September 2026 19:46:50 BST, "Paul E. McKenney"

<[email protected]>

Quoted 18 lines
Quoted 7 lines · wrote:
Quoted 6 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 8 lines
Quoted 2 lines

I'm not sure this URL actually points to a relevant issue ?

Indeed, it does not, apologies! Here you go:

https://lore.kernel.org/all/[email protected]/

Quoted 1 line
Quoted 1 line

Once that is in place, I would be happy to put this back into

-next.

Quoted 4 lines
Quoted 4 lines
Quoted 2 lines
Quoted 2 lines

At some point, we will need to get rid of the concept of wildcard

hazard

Quoted 1 line
Quoted 1 line

pointers, as those end up instead emulating RCU, but I don't see

that

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 1 line

as an immediate obstacle.

I already have the implementation which eliminates the wildcard if we care about this. It was part of a previous hazptr series version.

Do you want me to resurrect it on top of the current series ? This depends on:

Quoted 31 lines
Quoted 5 lines
Quoted 5 lines

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See:

https://lore.kernel.org/all/[email protected]/

Quoted 24 lines

The main objection was over the content and style of the kernel-doc header comment, right? I am guessing that it should be possible to resolve this to roughly equal disgust of all concerned. ;-)

We did make some progress on this sort of pointer issue in C++29 this past June:

https://people.kernel.org/paulmck/c-pointer-zap-and-oota-progress

But the piece you need is this guy, which is still in process:

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3790r1.pdf

Plus it will be some time before this reaches all the compilers used to build the Linux kernel, and probably even more time to reach the C language. I do have pen-on-paper notes that will lead to a draft of the corresponding C-language working paper, but these things do not move quickly.

So, yes, we will need something like ptr_eq() for some years to come.

Back to your original question, given the fix for the above bug and given the current use case, I believe we can get the current series

into

Quoted 1 line
Quoted 1 line

mainline. Give or take Linus's thoughts on the matter. But either

way,

Quoted 7 lines
Quoted 6 lines

we will need a version that allows the user to avoid all wildcard use sooner rather than later.

So having a series on top of the current one for a later merge window would be a very good thing!

Can I participate in this? :)

If Mathieu is OK with it, feel free to look at the patch stack that Mathieu sent the URL for earlier in this thread. Either way, please feel free to look at the stack in my -rcu tree based on v7.3-rc1 and headed by this commit:

4398b7c192d ("hazptr: Implement two-phase wildcard scan")

Perhaps you can find the bug that kernel test robot located. ;-)

My -rcu tree is here:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git

Just so you know, in all cases, your taking on a task does not preclude others from also taking that same task on.

Quoted 6 lines
Quoted 5 lines

Thanx, Paul

Quoted 2 lines

--- Thanks!

https://lore.kernel.org/all/[email protected]/

Hey, test this fix?

Very good, thank you!

Please post this patch as a reply to the report, asking them to test:

https://lore.kernel.org/all/[email protected]/

But first, have you tested it locally? Something like this:

tools/testing/selftests/rcutorture/bin/kvm.sh --torture hazptr --allcpus --duration 2h

Would run a two-hour test of each of the two scenarios, within a guest OS. If your host system has 32 or more CPUs, it will run both scenarios concurrently.

Thanx, Paul

Quoted 87 lines
Quoted 1 line

From 3e92b8153c31106d6a080e1c9bbe9bf1e86e1f63 Mon Sep 17 00:00:00 2001

From: Bradley Morgan <[email protected]>
Date: Mon, 7 Sep 2026 18:00:08 +0000
Subject: [PATCH] hazptrtorture: Only detach acquired hazard pointers

hazptr_torture_acquire() detaches unconditionally, even when the readlock fails. A failed acquire leaves nothing to detach, but the detach still promotes the context to its backup slot and chains that slot into the running CPU's overflow list. The reader then retries on its own CPU, the fast path hands out a per-CPU slot and overwrites ctx->slot, and the chained backup node is orphaned, still linked, with nobody left to unchain it.

The next detach of the same context chains the same node a second time, into another CPU's list, and the node ends up reachable from both. The eventual release unchains it once, hlist_del() poisons node->next, and the first list is left pointing at the poisoned node. The writer's next hazptr_synchronize() walks that list, steps onto LIST_POISON1 (0x100 on i386, where POISON_POINTER_DELTA is 0), and reads slot.addr at offset 8 of the backup slot, address 0x108, which is the crash the robot hit.

  cpuA (IPI acquire)     cpuR (reader)          cpuD (do_pending)
  ---------------------  ---------------------  -------------------
  readlock() returns
    NULL
  detach chains the
    backup node into
    cpuA list
                         hpp_htp is NULL,
                           continue
                         reacquire, ctx->slot
                           is now a cpuR
                           per-CPU slot
                         acquire succeeds,
                           defer, detach chains
                           the SAME node into
                           cpuR list
                                                release, unchain
                                                  once, node->next
                                                  is POISON1
                                                kfree(hppp)
  synchronize walks cpuA
    list, node->next is
    0x100, reads 0x108,
    Oops

Skip the detach when the acquire failed. The slot holds NULL in that case, note_context_switch() and the synchronize scanners skip NULL slots, and the next acquire overwrites ctx->slot, so leaving the context attached is safe.

The robot's original report was against the defer path before detach existed, which 4bd7f458229a fixed. This is the same crash surviving through the IPI acquire path that 6357ec235c59 added.

Fixes 6357ec235c59 ("hazptrtorture: Fix hazptr ownership issue")
---
 kernel/rcu/hazptrtorture.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +4 -1
index 7c8b589..267f262 100644
@@ -373,8 +373,11 @@ static void hazptr_torture_acquire(void *hppp_in)
 	/*
 	 * Acquiring a hazard pointer from a remote CPU.
 	 * Detach hazptr from its task so it can be released by another task.
+	 * A failed acquire has nothing to detach, and detaching one anyway
+	 * orphans the chained backup slot on this CPU's overflow list.
 	 */
-	hazptr_detach(&hppp->hpp_hc);
+	if (hppp->hpp_htp)
+		hazptr_detach(&hppp->hpp_hc);
 	atomic_long_inc(per_cpu_ptr(&hazptr_torture_acquires_irq, raw_smp_processor_id()));
 }
 
Signature · 5 lines
2.47.3


--- Thanks!
https://lore.kernel.org/all/[email protected]/
#17 lore raw
Quoted 121 lines · Paul E. McKenney
Quoted 1 line · Bradley Morgan

On 7 September 2026 00:16:55 BST, "Paul E. McKenney"

<[email protected]>

Quoted 18 lines
Quoted 17 lines · wrote:
Quoted 1 line · Bradley Morgan

On 6 September 2026 19:46:50 BST, "Paul E. McKenney"

<[email protected]>

Quoted 9 lines
Quoted 7 lines · wrote:
Quoted 6 lines · Mathieu Desnoyers
Quoted 1 line · Paul E. McKenney

On Fri, Sep 04, 2026 at 06:28:45PM +0100, Bradley Morgan wrote:

[...]

Quoted 3 lines

I would not say "no" to a fix for this issue:

https://lore.kernel.org/all/[email protected]/

Quoted 3 lines
Quoted 3 lines
Quoted 2 lines

I'm not sure this URL actually points to a relevant issue ?

Quoted 8 lines
Quoted 8 lines
Quoted 5 lines
Quoted 5 lines

Indeed, it does not, apologies! Here you go:

https://lore.kernel.org/all/[email protected]/

Quoted 1 line
Quoted 1 line

Once that is in place, I would be happy to put this back into

-next.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

At some point, we will need to get rid of the concept of

wildcard

Quoted 6 lines
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines

hazard

Quoted 1 line
Quoted 1 line

pointers, as those end up instead emulating RCU, but I don't see

that

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 1 line

as an immediate obstacle.

I already have the implementation which eliminates the wildcard if

we

Quoted 23 lines
Quoted 23 lines
Quoted 11 lines
Quoted 10 lines
Quoted 10 lines

care about this. It was part of a previous hazptr series version.

Do you want me to resurrect it on top of the current series ? This depends on:

- ptr_eq(),
- then use ptr_eq() to compare the loaded pointer (pre mb) with the re-loaded pointer (post-mb).

See:

https://lore.kernel.org/all/[email protected]/

Quoted 10 lines
Quoted 10 lines

The main objection was over the content and style of the kernel-doc header comment, right? I am guessing that it should be possible to resolve this to roughly equal disgust of all concerned. ;-)

We did make some progress on this sort of pointer issue in C++29 this past June:

https://people.kernel.org/paulmck/c-pointer-zap-and-oota-progress

Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

But the piece you need is this guy, which is still in process:

https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p3790r1.pdf

Plus it will be some time before this reaches all the compilers used to build the Linux kernel, and probably even more time to reach the C language. I do have pen-on-paper notes that will lead to a draft of the corresponding C-language working paper, but these things do

not

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

move quickly.

So, yes, we will need something like ptr_eq() for some years to

come.

Quoted 7 lines
Quoted 7 lines
Quoted 3 lines
Quoted 3 lines

Back to your original question, given the fix for the above bug and given the current use case, I believe we can get the current series

into

Quoted 1 line
Quoted 1 line

mainline. Give or take Linus's thoughts on the matter. But either

way,

Quoted 1 line
Quoted 1 line

we will need a version that allows the user to avoid all wildcard

use

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

sooner rather than later.

So having a series on top of the current one for a later merge

window

Quoted 20 lines
Quoted 20 lines
Quoted 3 lines
Quoted 2 lines

would be a very good thing!

Can I participate in this? :)

If Mathieu is OK with it, feel free to look at the patch stack that Mathieu sent the URL for earlier in this thread. Either way, please feel free to look at the stack in my -rcu tree based on v7.3-rc1 and headed by this commit:

4398b7c192d ("hazptr: Implement two-phase wildcard scan")

Perhaps you can find the bug that kernel test robot located. ;-)

My -rcu tree is here:

git://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git

Just so you know, in all cases, your taking on a task does not preclude others from also taking that same task on.

Quoted 7 lines
Quoted 4 lines

Thanx, Paul

Quoted 2 lines

--- Thanks!

https://lore.kernel.org/all/[email protected]/
Hey, test this fix?

Very good, thank you!

Please post this patch as a reply to the report, asking them to test:

https://lore.kernel.org/all/[email protected]/

But first, have you tested it locally? Something like this:

Quoted 8 lines

tools/testing/selftests/rcutorture/bin/kvm.sh --torture hazptr --allcpus --duration 2h

Would run a two-hour test of each of the two scenarios, within a guest OS. If your host system has 32 or more CPUs, it will run both scenarios concurrently.

Thanx, Paul

My server is pathetic, it's terrible, it takes a while to build a full kernel, in, (gulp) 4 gb of ram (oh no!) but I leave it sitting when doing full kernel builds, a torture test would kill it. Hah.

I'll post the patch soon.

Quoted 83 lines
Quoted 53 lines
Quoted 2 lines

From 3e92b8153c31106d6a080e1c9bbe9bf1e86e1f63 Mon Sep 17 00:00:00 2001

From: Bradley Morgan <[email protected]>
Date: Mon, 7 Sep 2026 18:00:08 +0000
Subject: [PATCH] hazptrtorture: Only detach acquired hazard pointers

hazptr_torture_acquire() detaches unconditionally, even when the readlock fails. A failed acquire leaves nothing to detach, but the detach still promotes the context to its backup slot and chains that slot into the running CPU's overflow list. The reader then retries on its own CPU, the fast path hands out a per-CPU slot and overwrites ctx->slot, and the chained backup node is orphaned, still linked, with nobody left to unchain it.

The next detach of the same context chains the same node a second time, into another CPU's list, and the node ends up reachable from both. The eventual release unchains it once, hlist_del() poisons node->next, and the first list is left pointing at the poisoned node. The writer's next hazptr_synchronize() walks that list, steps onto LIST_POISON1 (0x100 on i386, where POISON_POINTER_DELTA is 0), and reads slot.addr at offset 8 of the backup slot, address 0x108, which is the crash the robot hit.

  cpuA (IPI acquire)     cpuR (reader)          cpuD (do_pending)
  ---------------------  ---------------------  -------------------
  readlock() returns
    NULL
  detach chains the
    backup node into
    cpuA list
                         hpp_htp is NULL,
                           continue
                         reacquire, ctx->slot
                           is now a cpuR
                           per-CPU slot
                         acquire succeeds,
                           defer, detach chains
                           the SAME node into
                           cpuR list
                                                release, unchain
                                                  once, node->next
                                                  is POISON1
                                                kfree(hppp)
  synchronize walks cpuA
    list, node->next is
    0x100, reads 0x108,
    Oops

Skip the detach when the acquire failed. The slot holds NULL in that case, note_context_switch() and the synchronize scanners skip NULL slots, and the next acquire overwrites ctx->slot, so leaving the context attached is safe.

Quoted 7 lines

The robot's original report was against the defer path before detach existed, which 4bd7f458229a fixed. This is the same crash surviving through the IPI acquire path that 6357ec235c59 added.

Fixes 6357ec235c59 ("hazptrtorture: Fix hazptr ownership issue")

Closes:

https://lore.kernel.org/oe-lkp/[email protected]

Quoted 22 lines
---
 kernel/rcu/hazptrtorture.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
kernel/rcu/hazptrtorture.c +4 -1
index 7c8b589..267f262 100644
@@ -373,8 +373,11 @@ static void hazptr_torture_acquire(void *hppp_in)
 	/*
 	 * Acquiring a hazard pointer from a remote CPU.
 	 * Detach hazptr from its task so it can be released by another task.
+	 * A failed acquire has nothing to detach, and detaching one anyway
+	 * orphans the chained backup slot on this CPU's overflow list.
 	 */
-	hazptr_detach(&hppp->hpp_hc);
+	if (hppp->hpp_htp)
+		hazptr_detach(&hppp->hpp_hc);
 	atomic_long_inc(per_cpu_ptr(&hazptr_torture_acquires_irq, raw_smp_processor_id()));
 }
 
Quoted 7 lines
Quoted 6 lines
Signature · 4 lines
2.47.3


--- Thanks!

https://lore.kernel.org/all/[email protected]/

--- Thanks!
https://lore.kernel.org/all/[email protected]/