Skip to contentSkip to conversation

sched/isolation: Defer freeing of cpumask memblock memory to initcall

PATCH Applied Waiman Long 2 participants 5 messages Jul 1 – Aug 13 Scheduler

lore
Share & tools
lore
#0 lore raw

PATCH v4 1/1 sched/isolation: Defer freeing of cpumask memblock memory to initcall

When testing a linux-next kernel with commit 59bd1d914bb5 ("memblock:
warn when freeing reserved memory before memory map is initialized"), the following warning was hit when there was a "nohz_full" kernel boot parameter.

Cannot free reserved memory because of deferred initialization of the memory map

  WARNING: mm/memblock.c:904 at __free_reserved_area+0xde/0xf0, CPU#0: swapper/0/0
    :
  Call Trace:
   <TASK>
   memblock_phys_free+0xcb/0x100
   housekeeping_init+0x14c/0x170
   start_kernel+0x207/0x450
   x86_64_start_reservations+0x24/0x30
   x86_64_start_kernel+0xda/0xe0
   common_startup_64+0x13e/0x141
   </TASK>

IOW, we shouldn't free memblock allocated memory so early in the boot process when memory map isn't fully initialized in deferred_init_memmap().

Fix it by saving the housekeeping cpumask memblock memory to be freed into a llist free list in housekeeping_init() and add a new housekeeping_late_init() helper to defer the actual freeing of memblock memory to when initcall's are being processed. The cpumask memblock memory is treated as a llist_node with the size of a "long" type which is also smallest cpumask size that can be allocated.

The non-atomic version of the llist APIs are used as there is no contention.

This commit depends on the presence of commit 7c2eee9c1367 ("memblock:
don't touch memblock arrays when memblock_free() is called late") to prevent a KASAN UAF bug report [1].

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 27c3a5967f05 ("sched/isolation: Convert housekeeping cpumasks to rcu pointers")
---
 kernel/sched/isolation.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)

[v4] Collect tags and update commit log

kernel/sched/isolation.c +15 -1
index ef152d401fe2..156025ef81b7 100644
@@ -8,6 +8,7 @@
  *
  */
 #include <linux/sched/isolation.h>
+#include <linux/llist.h>
 #include <linux/pci.h>
 #include "sched.h"
 
@@ -27,6 +28,7 @@ struct housekeeping {
 };
 
 static struct housekeeping housekeeping;
+static __initdata LLIST_HEAD(memblock_freelist);
 
 bool housekeeping_enabled(enum hk_type type)
 {
@@ -189,10 +191,22 @@ void __init housekeeping_init(void)
 		WARN_ON_ONCE(cpumask_empty(omask));
 		cpumask_copy(nmask, omask);
 		RCU_INIT_POINTER(housekeeping.cpumasks[type], nmask);
-		memblock_free(omask, cpumask_size());
+		__llist_add((struct llist_node *)omask, &memblock_freelist);
 	}
 }
 
+static int __init housekeeping_late_init(void)
+{
+	struct llist_node *llnode, *pos, *t;
+
+	/* Free allocated memblock memory, if any */
+	llnode = __llist_del_all(&memblock_freelist);
+	llist_for_each_safe(pos, t, llnode)
+		memblock_free(pos, cpumask_size());
+	return 0;
+}
+pure_initcall(housekeeping_late_init);
+
 static void __init housekeeping_setup_type(enum hk_type type,
 					   cpumask_var_t housekeeping_staging)
 {
Signature · 1 line
2.54.0
#1 lore raw
Quoted 36 lines · Waiman Long

When testing a linux-next kernel with commit 59bd1d914bb5 ("memblock:
warn when freeing reserved memory before memory map is initialized"), the following warning was hit when there was a "nohz_full" kernel boot parameter.

   Cannot free reserved memory because of deferred initialization of the memory map
   WARNING: mm/memblock.c:904 at __free_reserved_area+0xde/0xf0, CPU#0: swapper/0/0
     :
   Call Trace:
    <TASK>
    memblock_phys_free+0xcb/0x100
    housekeeping_init+0x14c/0x170
    start_kernel+0x207/0x450
    x86_64_start_reservations+0x24/0x30
    x86_64_start_kernel+0xda/0xe0
    common_startup_64+0x13e/0x141
    </TASK>

IOW, we shouldn't free memblock allocated memory so early in the boot process when memory map isn't fully initialized in deferred_init_memmap().

Fix it by saving the housekeeping cpumask memblock memory to be freed into a llist free list in housekeeping_init() and add a new housekeeping_late_init() helper to defer the actual freeing of memblock memory to when initcall's are being processed. The cpumask memblock memory is treated as a llist_node with the size of a "long" type which is also smallest cpumask size that can be allocated.

The non-atomic version of the llist APIs are used as there is no contention.

This commit depends on the presence of commit 7c2eee9c1367 ("memblock:
don't touch memblock arrays when memblock_free() is called late") to prevent a KASAN UAF bug report [1].

Quoted 6 lines

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 27c3a5967f05 ("sched/isolation: Convert housekeeping cpumasks to rcu pointers")

Peter,

Are you OK with this patch as I would like to get it merged into v7.3 if possible?

Cheers,
Longman

Quoted 50 lines
---
  kernel/sched/isolation.c | 16 +++++++++++++++-
  1 file changed, 15 insertions(+), 1 deletion(-)

[v4] Collect tags and update commit log

kernel/sched/isolation.c +15 -1
index ef152d401fe2..156025ef81b7 100644
@@ -8,6 +8,7 @@
   *
   */
  #include <linux/sched/isolation.h>
+#include <linux/llist.h>
  #include <linux/pci.h>
  #include "sched.h"
 
@@ -27,6 +28,7 @@ struct housekeeping {
  };
 
  static struct housekeeping housekeeping;
+static __initdata LLIST_HEAD(memblock_freelist);
 
  bool housekeeping_enabled(enum hk_type type)
  {
@@ -189,10 +191,22 @@ void __init housekeeping_init(void)
  		WARN_ON_ONCE(cpumask_empty(omask));
  		cpumask_copy(nmask, omask);
  		RCU_INIT_POINTER(housekeeping.cpumasks[type], nmask);
-		memblock_free(omask, cpumask_size());
+		__llist_add((struct llist_node *)omask, &memblock_freelist);
  	}
  }
 
+static int __init housekeeping_late_init(void)
+{
+	struct llist_node *llnode, *pos, *t;
+
+	/* Free allocated memblock memory, if any */
+	llnode = __llist_del_all(&memblock_freelist);
+	llist_for_each_safe(pos, t, llnode)
+		memblock_free(pos, cpumask_size());
+	return 0;
+}
+pure_initcall(housekeeping_late_init);
+
  static void __init housekeeping_setup_type(enum hk_type type,
  					   cpumask_var_t housekeeping_staging)
  {
#2 lore raw
In reply to Waiman Long
Quoted 82 lines · Waiman Long

When testing a linux-next kernel with commit 59bd1d914bb5 ("memblock:
warn when freeing reserved memory before memory map is initialized"), the following warning was hit when there was a "nohz_full" kernel boot parameter.

   Cannot free reserved memory because of deferred initialization of the memory map
   WARNING: mm/memblock.c:904 at __free_reserved_area+0xde/0xf0, CPU#0: swapper/0/0
     :
   Call Trace:
    <TASK>
    memblock_phys_free+0xcb/0x100
    housekeeping_init+0x14c/0x170
    start_kernel+0x207/0x450
    x86_64_start_reservations+0x24/0x30
    x86_64_start_kernel+0xda/0xe0
    common_startup_64+0x13e/0x141
    </TASK>

IOW, we shouldn't free memblock allocated memory so early in the boot process when memory map isn't fully initialized in deferred_init_memmap().

Fix it by saving the housekeeping cpumask memblock memory to be freed into a llist free list in housekeeping_init() and add a new housekeeping_late_init() helper to defer the actual freeing of memblock memory to when initcall's are being processed. The cpumask memblock memory is treated as a llist_node with the size of a "long" type which is also smallest cpumask size that can be allocated.

The non-atomic version of the llist APIs are used as there is no contention.

This commit depends on the presence of commit 7c2eee9c1367 ("memblock:
don't touch memblock arrays when memblock_free() is called late") to prevent a KASAN UAF bug report [1].

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 27c3a5967f05 ("sched/isolation: Convert housekeeping cpumasks to rcu pointers")
---
  kernel/sched/isolation.c | 16 +++++++++++++++-
  1 file changed, 15 insertions(+), 1 deletion(-)

[v4] Collect tags and update commit log

kernel/sched/isolation.c +8 -1
index ef152d401fe2..156025ef81b7 100644
@@ -8,6 +8,7 @@
   *
   */
  #include <linux/sched/isolation.h>
+#include <linux/llist.h>
  #include <linux/pci.h>
  #include "sched.h"
 
@@ -27,6 +28,7 @@ struct housekeeping {
  };
 
  static struct housekeeping housekeeping;
+static __initdata LLIST_HEAD(memblock_freelist);
 
  bool housekeeping_enabled(enum hk_type type)
  {
@@ -189,10 +191,22 @@ void __init housekeeping_init(void)
  		WARN_ON_ONCE(cpumask_empty(omask));
  		cpumask_copy(nmask, omask);
  		RCU_INIT_POINTER(housekeeping.cpumasks[type], nmask);
-		memblock_free(omask, cpumask_size());
+		__llist_add((struct llist_node *)omask, &memblock_freelist);
  	}
  }
 
+static int __init housekeeping_late_init(void)
+{
+	struct llist_node *llnode, *pos, *t;
+
+	/* Free allocated memblock memory, if any */

kernel/sched/isolation.c void __init housekeeping_init(void) · line 212

Quoted 10 lines
kernel/sched/isolation.c +7 -0
+	llnode = __llist_del_all(&memblock_freelist);
+	llist_for_each_safe(pos, t, llnode)
+		memblock_free(pos, cpumask_size());
+	return 0;
+}
+pure_initcall(housekeeping_late_init);
+
  static void __init housekeeping_setup_type(enum hk_type type,
  					   cpumask_var_t housekeeping_staging)
  {

Ping!

Is this patch in a state that is merge-able?

Cheers,
Longman

#3 lore raw
Quoted 92 lines · Waiman Long
Quoted 78 lines · Waiman Long

When testing a linux-next kernel with commit 59bd1d914bb5 ("memblock:
warn when freeing reserved memory before memory map is initialized"), the following warning was hit when there was a "nohz_full" kernel boot parameter.

   Cannot free reserved memory because of deferred initialization of the memory map

   WARNING: mm/memblock.c:904 at __free_reserved_area+0xde/0xf0, CPU#0: swapper/0/0
     :
   Call Trace:
    <TASK>
    memblock_phys_free+0xcb/0x100
    housekeeping_init+0x14c/0x170
    start_kernel+0x207/0x450
    x86_64_start_reservations+0x24/0x30
    x86_64_start_kernel+0xda/0xe0
    common_startup_64+0x13e/0x141
    </TASK>

IOW, we shouldn't free memblock allocated memory so early in the boot process when memory map isn't fully initialized in deferred_init_memmap().

Fix it by saving the housekeeping cpumask memblock memory to be freed into a llist free list in housekeeping_init() and add a new housekeeping_late_init() helper to defer the actual freeing of memblock memory to when initcall's are being processed. The cpumask memblock memory is treated as a llist_node with the size of a "long" type which is also smallest cpumask size that can be allocated.

The non-atomic version of the llist APIs are used as there is no contention.

This commit depends on the presence of commit 7c2eee9c1367 ("memblock:
don't touch memblock arrays when memblock_free() is called late") to prevent a KASAN UAF bug report [1].

  [1] https://lore.kernel.org/lkml/[email protected]/

Fixes 27c3a5967f05 ("sched/isolation: Convert housekeeping cpumasks to rcu pointers")

---
  kernel/sched/isolation.c | 16 +++++++++++++++-

  1 file changed, 15 insertions(+), 1 deletion(-)

  [v4] Collect tags and update commit log

kernel/sched/isolation.c +1 -0
index ef152d401fe2..156025ef81b7 100644
@@ -8,6 +8,7 @@
    *
    */
   #include <linux/sched/isolation.h>
+#include <linux/llist.h>
   #include <linux/pci.h>
   #include "sched.h"
   @@ -27,6 +28,7 @@ struct housekeeping {

  };
    static struct housekeeping housekeeping;
+static __initdata LLIST_HEAD(memblock_freelist);
    bool housekeeping_enabled(enum hk_type type)
  {

kernel/sched/isolation.c +5 -1
@@ -189,10 +191,22 @@ void __init housekeeping_init(void)
           WARN_ON_ONCE(cpumask_empty(omask));
           cpumask_copy(nmask, omask);
           RCU_INIT_POINTER(housekeeping.cpumasks[type], nmask);
-        memblock_free(omask, cpumask_size());
+        __llist_add((struct llist_node *)omask, &memblock_freelist);
       }
   }
   +static int __init housekeeping_late_init(void)
+{
+    struct llist_node *llnode, *pos, *t;
+
+    /* Free allocated memblock memory, if any */

kernel/sched/isolation.c void __init housekeeping_init(void) · line 211

Quoted 10 lines
kernel/sched/isolation.c +7 -0
+    llnode = __llist_del_all(&memblock_freelist);
+    llist_for_each_safe(pos, t, llnode)
+        memblock_free(pos, cpumask_size());
+    return 0;
+}
+pure_initcall(housekeeping_late_init);
+
   static void __init housekeeping_setup_type(enum hk_type type,
                          cpumask_var_t housekeeping_staging)
   {

Ping!

Quoted 1 line

Is this patch in a state that is merge-able?

Any feedback?

Thanks,
Longman

merge bot Applied
#4 lore raw
In reply to Waiman Long
The following commit has been merged into the sched/core branch of tip:

The following commit has been merged into the sched/core branch of tip:

Commit-ID:     2b58c749b8c5244e259a0230bc57b10b010dc545
Gitweb:        https://git.kernel.org/tip/2b58c749b8c5244e259a0230bc57b10b010dc545
Author:        Waiman Long <[email protected]>
AuthorDate:    Wed, 01 Jul 2026 15:58:10 -04:00
Committer:     Peter Zijlstra <[email protected]>
CommitterDate: Thu, 13 Aug 2026 12:37:03 +02:00

sched/isolation: Defer freeing of cpumask memblock memory to initcall

When testing a linux-next kernel with commit 59bd1d914bb5 ("memblock:
warn when freeing reserved memory before memory map is initialized"), the following warning was hit when there was a "nohz_full" kernel boot parameter.

Cannot free reserved memory because of deferred initialization of the memory map

  WARNING: mm/memblock.c:904 at __free_reserved_area+0xde/0xf0, CPU#0: swapper/0/0
    :
  Call Trace:
   <TASK>
   memblock_phys_free+0xcb/0x100
   housekeeping_init+0x14c/0x170
   start_kernel+0x207/0x450
   x86_64_start_reservations+0x24/0x30
   x86_64_start_kernel+0xda/0xe0
   common_startup_64+0x13e/0x141
   </TASK>

IOW, we shouldn't free memblock allocated memory so early in the boot process when memory map isn't fully initialized in deferred_init_memmap().

Fix it by saving the housekeeping cpumask memblock memory to be freed into a llist free list in housekeeping_init() and add a new housekeeping_late_init() helper to defer the actual freeing of memblock memory to when initcall's are being processed. The cpumask memblock memory is treated as a llist_node with the size of a "long" type which is also smallest cpumask size that can be allocated.

The non-atomic version of the llist APIs are used as there is no contention.

This commit depends on the presence of commit 7c2eee9c1367 ("memblock:
don't touch memblock arrays when memblock_free() is called late") to prevent a KASAN UAF bug report [1].

[1] https://lore.kernel.org/lkml/[email protected]/

Fixes 27c3a5967f05 ("sched/isolation: Convert housekeeping cpumasks to rcu pointers")
---
 kernel/sched/isolation.c | 16 +++++++++++++++-
 1 file changed, 15 insertions(+), 1 deletion(-)
kernel/sched/isolation.c +15 -1
index ef152d4..156025e 100644
@@ -8,6 +8,7 @@
  *
  */
 #include <linux/sched/isolation.h>
+#include <linux/llist.h>
 #include <linux/pci.h>
 #include "sched.h"
 
@@ -27,6 +28,7 @@ struct housekeeping {
 };
 
 static struct housekeeping housekeeping;
+static __initdata LLIST_HEAD(memblock_freelist);
 
 bool housekeeping_enabled(enum hk_type type)
 {
@@ -189,10 +191,22 @@ void __init housekeeping_init(void)
 		WARN_ON_ONCE(cpumask_empty(omask));
 		cpumask_copy(nmask, omask);
 		RCU_INIT_POINTER(housekeeping.cpumasks[type], nmask);
-		memblock_free(omask, cpumask_size());
+		__llist_add((struct llist_node *)omask, &memblock_freelist);
 	}
 }
 
+static int __init housekeeping_late_init(void)
+{
+	struct llist_node *llnode, *pos, *t;
+
+	/* Free allocated memblock memory, if any */
+	llnode = __llist_del_all(&memblock_freelist);
+	llist_for_each_safe(pos, t, llnode)
+		memblock_free(pos, cpumask_size());
+	return 0;
+}
+pure_initcall(housekeeping_late_init);
+
 static void __init housekeeping_setup_type(enum hk_type type,
 					   cpumask_var_t housekeeping_staging)
 {