acpi/hmat: hmat_register_target() refactor to address lockdep warning

PATCH Superseded Dave Jiang 2 participants 8 messages Nov 5, 2025 – Nov 6, 2025

Subsystems (2)

ACPI libnvdimm: non-volatile memory device

lore

Share & tools

lore

Dave Jiang

author maintainer

2025-11-05 23:48 UTC #0 lore raw

v3 cover acpi/hmat: hmat_register_target() refactor to address lockdep warning

This series has changes that refactor the hmat_register_target() function to clean up a lockdep warning.

Dave Jiang (2):
acpi/hmat: Return when generic target is updated acpi/hmat: Fix lockdep warning for hmem_register_resource()

 drivers/acpi/numa/hmat.c | 47 ++++++++++++++++++++++------------------
 1 file changed, 26 insertions(+), 21 deletions(-)

base-commit: 3a8660878839faadb4f1a6dd72c3179c1df56787

Signature · 1 line

2.51.0

Dave Jiang

author maintainer

2025-11-05 23:48 UTC #1 lore raw

PATCH v3 1/2 acpi/hmat: Return when generic target is updated

With the current code flow, once the generic target is updated target->registered is set and the remaining code is skipped. So return immediately instead of going through the checks and then skip.

Signed-off-by Dave Jiang <[email protected]>

---
 drivers/acpi/numa/hmat.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

drivers/acpi/numa/hmat.c +6 -5

index 5a36d57289b4..1dc73d20d989 100644
@@ -888,12 +888,13 @@ static void hmat_register_target(struct memory_target *target)
 	 * Register generic port perf numbers. The nid may not be
 	 * initialized and is still NUMA_NO_NODE.
 	 */
-	mutex_lock(&target_lock);
-	if (*(u16 *)target->gen_port_device_handle) {
-		hmat_update_generic_target(target);
-		target->registered = true;
+	scoped_guard(mutex, &target_lock) {
+		if (*(u16 *)target->gen_port_device_handle) {
+			hmat_update_generic_target(target);
+			target->registered = true;
+			return;
+		}
 	}
-	mutex_unlock(&target_lock);
 
 	/*
 	 * Skip offline nodes. This can happen when memory

Signature · 1 line

2.51.0

Dave Jiang

author maintainer

2025-11-05 23:48 UTC #2 lore raw

PATCH v3 2/2 acpi/hmat: Fix lockdep warning for hmem_register_resource()

In reply to Dave Jiang

The following lockdep splat was observed while kernel auto-online a CXL memory region:

======================================================
WARNING: possible circular locking dependency detected
6.17.0djtest+ #53 Tainted: G W
------------------------------------------------------
systemd-udevd/3334 is trying to acquire lock:
ffffffff90346188 (hmem_resource_lock){+.+.}-{4:4}, at: hmem_register_resource+0x31/0x50

but task is already holding lock:
ffffffff90338890 ((node_chain).rwsem){++++}-{4:4}, at: blocking_notifier_call_chain+0x2e/0x70

which lock already depends on the new lock.
[..]

Chain exists of:
  hmem_resource_lock --> mem_hotplug_lock --> (node_chain).rwsem

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  rlock((node_chain).rwsem);
                               lock(mem_hotplug_lock);
                               lock((node_chain).rwsem);

lock(hmem_resource_lock);

The lock ordering can cause potential deadlock. There are instances where hmem_resource_lock is taken after (node_chain).rwsem, and vice versa.

Split out the target update section of hmat_register_target() so that hmat_callback() only envokes that section instead of attempt to register hmem devices that it does not need to.

Fixes cf8741ac57ed ("ACPI: NUMA: HMAT: Register "soft reserved" memory as a

n "hmem" device")
notmuch/

Signed-off-by Dave Jiang <[email protected]>

---
v3:
- Refactor to split out target device setup vs target update (Dan)

---
 drivers/acpi/numa/hmat.c | 48 ++++++++++++++++++++++------------------
 1 file changed, 26 insertions(+), 22 deletions(-)

drivers/acpi/numa/hmat.c +26 ~~-22~~

index 1dc73d20d989..ddbdd32e79a8 100644
@@ -874,28 +874,10 @@ static void hmat_register_target_devices(struct memory_target *target)
 	}
 }
 
-static void hmat_register_target(struct memory_target *target)
+static void hmat_hotplug_target(struct memory_target *target)
 {
 	int nid = pxm_to_node(target->memory_pxm);
 
-	/*
-	 * Devices may belong to either an offline or online
-	 * node, so unconditionally add them.
-	 */
-	hmat_register_target_devices(target);
-
-	/*
-	 * Register generic port perf numbers. The nid may not be
-	 * initialized and is still NUMA_NO_NODE.
-	 */
-	scoped_guard(mutex, &target_lock) {
-		if (*(u16 *)target->gen_port_device_handle) {
-			hmat_update_generic_target(target);
-			target->registered = true;
-			return;
-		}
-	}
-
 	/*
 	 * Skip offline nodes. This can happen when memory
 	 * marked EFI_MEMORY_SP, "specific purpose", is applied

@@ -906,7 +888,7 @@ static void hmat_register_target(struct memory_target *target)
 	if (nid == NUMA_NO_NODE || !node_online(nid))
 		return;
 
-	mutex_lock(&target_lock);
+	guard(mutex)(&target_lock);
 	if (!target->registered) {
 		hmat_register_target_initiators(target);
 		hmat_register_target_cache(target);

@@ -914,7 +896,29 @@ static void hmat_register_target(struct memory_target *target)
 		hmat_register_target_perf(target, ACCESS_COORDINATE_CPU);
 		target->registered = true;
 	}
-	mutex_unlock(&target_lock);
+}
+
+static void hmat_register_target(struct memory_target *target)
+{
+	/*
+	 * Devices may belong to either an offline or online
+	 * node, so unconditionally add them.
+	 */
+	hmat_register_target_devices(target);
+
+	/*
+	 * Register generic port perf numbers. The nid may not be
+	 * initialized and is still NUMA_NO_NODE.
+	 */
+	scoped_guard(mutex, &target_lock) {
+		if (*(u16 *)target->gen_port_device_handle) {
+			hmat_update_generic_target(target);
+			target->registered = true;
+			return;
+		}
+	}
+
+	hmat_hotplug_target(target);
 }
 
 static void hmat_register_targets(void)

@@ -940,7 +944,7 @@ static int hmat_callback(struct notifier_block *self,
 	if (!target)
 		return NOTIFY_OK;
 
-	hmat_register_target(target);
+	hmat_hotplug_target(target);
 	return NOTIFY_OK;
 }

Signature · 1 line

2.51.0

Dave Jiang

author maintainer

2025-11-05 23:48 UTC #3 lore raw

v4 cover acpi/hmat: hmat_register_target() refactor to address lockdep warning

In reply to Dave Jiang

v4:
- Refactor hmat_hotplug_target() (Jonathan)
- Fix fixes tag. (Jonathan)

This series has changes that refactor the hmat_register_target() function to clean up a lockdep warning.

Dave Jiang (2):
acpi/hmat: Return when generic target is updated acpi/hmat: Fix lockdep warning for hmem_register_resource()

 drivers/acpi/numa/hmat.c | 60 ++++++++++++++++++++++------------------
 1 file changed, 33 insertions(+), 27 deletions(-)

base-commit: 6146a0f1dfae5d37442a9ddcba012add260bceb0

Signature · 1 line

2.51.1

Dave Jiang

author maintainer

2025-11-05 23:49 UTC #4 lore raw

PATCH v4 1/2 acpi/hmat: Return when generic target is updated

In reply to Dave Jiang

With the current code flow, once the generic target is updated target->registered is set and the remaining code is skipped. So return immediately instead of going through the checks and then skip.

Reviewed-by Jonathan Cameron <[email protected]>

Signed-off-by Dave Jiang <[email protected]>

---
no changes

---
 drivers/acpi/numa/hmat.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

drivers/acpi/numa/hmat.c +6 -5

index 5a36d57289b4..1dc73d20d989 100644
@@ -888,12 +888,13 @@ static void hmat_register_target(struct memory_target *target)
 	 * Register generic port perf numbers. The nid may not be
 	 * initialized and is still NUMA_NO_NODE.
 	 */
-	mutex_lock(&target_lock);
-	if (*(u16 *)target->gen_port_device_handle) {
-		hmat_update_generic_target(target);
-		target->registered = true;
+	scoped_guard(mutex, &target_lock) {
+		if (*(u16 *)target->gen_port_device_handle) {
+			hmat_update_generic_target(target);
+			target->registered = true;
+			return;
+		}
 	}
-	mutex_unlock(&target_lock);
 
 	/*
 	 * Skip offline nodes. This can happen when memory

Signature · 1 line

2.51.1

Dave Jiang

author maintainer

2025-11-05 23:49 UTC #5 lore raw

PATCH v4 2/2 acpi/hmat: Fix lockdep warning for hmem_register_resource()

In reply to Dave Jiang

The following lockdep splat was observed while kernel auto-online a CXL memory region:

but task is already holding lock:
ffffffff90338890 ((node_chain).rwsem){++++}-{4:4}, at: blocking_notifier_call_chain+0x2e/0x70

which lock already depends on the new lock.
[..]

Chain exists of:
  hmem_resource_lock --> mem_hotplug_lock --> (node_chain).rwsem

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  rlock((node_chain).rwsem);
                               lock(mem_hotplug_lock);
                               lock((node_chain).rwsem);

lock(hmem_resource_lock);

The lock ordering can cause potential deadlock. There are instances where hmem_resource_lock is taken after (node_chain).rwsem, and vice versa.

Split out the target update section of hmat_register_target() so that hmat_callback() only envokes that section instead of attempt to register hmem devices that it does not need to.

Fixes cf8741ac57ed ("ACPI: NUMA: HMAT: Register "soft reserved" memory as an "hmem" device")

Signed-off-by Dave Jiang <[email protected]>

---
v4:
- Fix fixes tag. (Jonathan)
- Refactor hmat_hotplug_target(). (Jonathan)

---
 drivers/acpi/numa/hmat.c | 47 ++++++++++++++++++++++------------------
 1 file changed, 26 insertions(+), 21 deletions(-)

drivers/acpi/numa/hmat.c +26 ~~-21~~

index 1dc73d20d989..d10cbe93c3a7 100644
@@ -874,10 +874,33 @@ static void hmat_register_target_devices(struct memory_target *target)
 	}
 }
 
-static void hmat_register_target(struct memory_target *target)
+static void hmat_hotplug_target(struct memory_target *target)
 {
 	int nid = pxm_to_node(target->memory_pxm);
 
+	/*
+	 * Skip offline nodes. This can happen when memory
+	 * marked EFI_MEMORY_SP, "specific purpose", is applied
+	 * to all the memory in a proximity domain leading to
+	 * the node being marked offline / unplugged, or if
+	 * memory-only "hotplug" node is offline.
+	 */
+	if (nid == NUMA_NO_NODE || !node_online(nid))
+		return;
+
+	guard(mutex)(&target_lock);
+	if (target->registered)
+		return;
+
+	hmat_register_target_initiators(target);
+	hmat_register_target_cache(target);
+	hmat_register_target_perf(target, ACCESS_COORDINATE_LOCAL);
+	hmat_register_target_perf(target, ACCESS_COORDINATE_CPU);
+	target->registered = true;
+}
+
+static void hmat_register_target(struct memory_target *target)
+{
 	/*
 	 * Devices may belong to either an offline or online
 	 * node, so unconditionally add them.

@@ -896,25 +919,7 @@ static void hmat_register_target(struct memory_target *target)
 		}
 	}
 
-	/*
-	 * Skip offline nodes. This can happen when memory
-	 * marked EFI_MEMORY_SP, "specific purpose", is applied
-	 * to all the memory in a proximity domain leading to
-	 * the node being marked offline / unplugged, or if
-	 * memory-only "hotplug" node is offline.
-	 */
-	if (nid == NUMA_NO_NODE || !node_online(nid))
-		return;
-
-	mutex_lock(&target_lock);
-	if (!target->registered) {
-		hmat_register_target_initiators(target);
-		hmat_register_target_cache(target);
-		hmat_register_target_perf(target, ACCESS_COORDINATE_LOCAL);
-		hmat_register_target_perf(target, ACCESS_COORDINATE_CPU);
-		target->registered = true;
-	}
-	mutex_unlock(&target_lock);
+	hmat_hotplug_target(target);
 }
 
 static void hmat_register_targets(void)

@@ -940,7 +945,7 @@ static int hmat_callback(struct notifier_block *self,
 	if (!target)
 		return NOTIFY_OK;
 
-	hmat_register_target(target);
+	hmat_hotplug_target(target);
 	return NOTIFY_OK;
 }

Signature · 1 line

2.51.1

Koralahalli Channabasappa, Smita

Tested-by Smita Koralahalli Reviewed-by Smita Koralahalli on v4 2/2

2025-11-06 20:28 UTC #6 lore raw

Re: [PATCH v4 2/2] acpi/hmat: Fix lockdep warning for hmem_register_resource()

Hi Dave,

Quoted 31 lines · Dave Jiang

The following lockdep splat was observed while kernel auto-online a CXL memory region:
======================================================
WARNING: possible circular locking dependency detected
6.17.0djtest+ #53 Tainted: G W
------------------------------------------------------
systemd-udevd/3334 is trying to acquire lock:
ffffffff90346188 (hmem_resource_lock){+.+.}-{4:4}, at: hmem_register_resource+0x31/0x50
but task is already holding lock:
ffffffff90338890 ((node_chain).rwsem){++++}-{4:4}, at: blocking_notifier_call_chain+0x2e/0x70
which lock already depends on the new lock.
[..]
Chain exists of:
   hmem_resource_lock --> mem_hotplug_lock --> (node_chain).rwsem

  Possible unsafe locking scenario:
        CPU0                    CPU1
        ----                    ----
   rlock((node_chain).rwsem);
                                lock(mem_hotplug_lock);
                                lock((node_chain).rwsem);
lock(hmem_resource_lock);
The lock ordering can cause potential deadlock. There are instances where hmem_resource_lock is taken after (node_chain).rwsem, and vice versa.

Quoted 7 lines

Split out the target update section of hmat_register_target() so that hmat_callback() only envokes that section instead of attempt to register hmem devices that it does not need to.
Fixes cf8741ac57ed ("ACPI: NUMA: HMAT: Register "soft reserved" memory as an "hmem" device")
Signed-off-by Dave Jiang <[email protected]>
---

Thanks for this fix.

This resolves a deadlock I was hitting while working on the Soft Reserved series.

Before this change, the HMEM deferred work path could hold hmem_resource_lock while eventually triggering memory onlining where udev holds mem_hotplug_lock.

Tested-by Smita Koralahalli <[email protected]>

Reviewed-by Smita Koralahalli <[email protected]>

Quoted 82 lines

v4:
- Fix fixes tag. (Jonathan)
- Refactor hmat_hotplug_target(). (Jonathan)

---
  drivers/acpi/numa/hmat.c | 47 ++++++++++++++++++++++------------------
  1 file changed, 26 insertions(+), 21 deletions(-)

drivers/acpi/numa/hmat.c +26 ~~-21~~

index 1dc73d20d989..d10cbe93c3a7 100644
@@ -874,10 +874,33 @@ static void hmat_register_target_devices(struct memory_target *target)
  	}
  }
 
-static void hmat_register_target(struct memory_target *target)
+static void hmat_hotplug_target(struct memory_target *target)
  {
  	int nid = pxm_to_node(target->memory_pxm);
 
+	/*
+	 * Skip offline nodes. This can happen when memory
+	 * marked EFI_MEMORY_SP, "specific purpose", is applied
+	 * to all the memory in a proximity domain leading to
+	 * the node being marked offline / unplugged, or if
+	 * memory-only "hotplug" node is offline.
+	 */
+	if (nid == NUMA_NO_NODE || !node_online(nid))
+		return;
+
+	guard(mutex)(&target_lock);
+	if (target->registered)
+		return;
+
+	hmat_register_target_initiators(target);
+	hmat_register_target_cache(target);
+	hmat_register_target_perf(target, ACCESS_COORDINATE_LOCAL);
+	hmat_register_target_perf(target, ACCESS_COORDINATE_CPU);
+	target->registered = true;
+}
+
+static void hmat_register_target(struct memory_target *target)
+{
  	/*
  	 * Devices may belong to either an offline or online
  	 * node, so unconditionally add them.

@@ -896,25 +919,7 @@ static void hmat_register_target(struct memory_target *target)
  		}
  	}
 
-	/*
-	 * Skip offline nodes. This can happen when memory
-	 * marked EFI_MEMORY_SP, "specific purpose", is applied
-	 * to all the memory in a proximity domain leading to
-	 * the node being marked offline / unplugged, or if
-	 * memory-only "hotplug" node is offline.
-	 */
-	if (nid == NUMA_NO_NODE || !node_online(nid))
-		return;
-
-	mutex_lock(&target_lock);
-	if (!target->registered) {
-		hmat_register_target_initiators(target);
-		hmat_register_target_cache(target);
-		hmat_register_target_perf(target, ACCESS_COORDINATE_LOCAL);
-		hmat_register_target_perf(target, ACCESS_COORDINATE_CPU);
-		target->registered = true;
-	}
-	mutex_unlock(&target_lock);
+	hmat_hotplug_target(target);
  }
 
  static void hmat_register_targets(void)

@@ -940,7 +945,7 @@ static int hmat_callback(struct notifier_block *self,
  	if (!target)
  		return NOTIFY_OK;
 
-	hmat_register_target(target);
+	hmat_hotplug_target(target);
  	return NOTIFY_OK;
  }

Dave Jiang

author maintainer

2025-11-05 23:52 UTC #7 lore raw

In reply to Dave Jiang

Quoted 7 lines · Dave Jiang

This series has changes that refactor the hmat_register_target() function to clean up a lockdep warning.
Dave Jiang (2):
acpi/hmat: Return when generic target is updated acpi/hmat: Fix lockdep warning for hmem_register_resource()

Quoted 5 lines

 drivers/acpi/numa/hmat.c | 47 ++++++++++++++++++++++------------------
 1 file changed, 26 insertions(+), 21 deletions(-)

base-commit: 3a8660878839faadb4f1a6dd72c3179c1df56787

Ignore please. See v4 resend series.