Skip to contentSkip to conversation

mm/page_alloc: couple of followups for recent cleanups

PATCH Applied Brendan Jackman 6 participants 16 messages Jul 15 – Jul 17

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

v3 cover mm/page_alloc: couple of followups for recent cleanups

---
Changes in v3:
- Fixed Sashiko Links: (Zi Yan)
- Fixed kerneldoc syntax (Sashiko)
- Removed extra VM_BUG_ON() (Sashiko)
- Link to v2: https://patch.msgid.link/[email protected]

Changes in v2:
- Separated from functional fixes
- Added cpuset cleanup, comment fixup, VM_BUG_ON() removal from Zi Yan's
review of [0].
- Link to v1: https://patch.msgid.link/[email protected]

Based on mm-new, these are followups to [0]

The alloc_pages_nolock_noprof() comment fixup could be squashed into "mm/page_alloc: relax GFP WARN in nolock allocs" - currently 11770f8836f44 in mm-new.

The VM_BUG_ON() removal could be squashed into "mm: remove __alloc_pages_node()", currently fba100a6cdfc5.
[0]: https://lore.kernel.org/all/[email protected]/

To: Andrew Morton <[email protected]>
To: Vlastimil Babka <[email protected]>
To: Suren Baghdasaryan <[email protected]>
To: Michal Hocko <[email protected]>
To: Brendan Jackman <[email protected]>
To: Johannes Weiner <[email protected]>
To: Zi Yan <[email protected]>
To: Sebastian Andrzej Siewior <[email protected]>
To: Clark Williams <[email protected]>
To: Steven Rostedt <[email protected]>
To: Waiman Long <[email protected]>
To: Ridong Chen <[email protected]>
To: Tejun Heo <[email protected]>
To: Michal Koutný <[email protected]>
To: David Hildenbrand <[email protected]>
To: Lorenzo Stoakes <[email protected]>
To: "Liam R. Howlett" <[email protected]>
To: Mike Rapoport <[email protected]>
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]
Cc: [email protected]

---
Brendan Jackman (4):
mm/page_alloc: rename FPI_TRYLOCK -> FPI_NOLOCK cgroup/cpuset: update some comments about the page allocator mm/page_alloc: fixup alloc_pages_nolock_noprof() comment mm/page_alloc: remove a couple of VM_BUG_ON()st

 include/linux/gfp.h    |  1 -
 kernel/cgroup/cpuset.c | 13 +++++--------
 mm/page_alloc.c        | 22 +++++++++++-----------
 3 files changed, 16 insertions(+), 20 deletions(-)

---
base-commit: 59c684a9908d2e6f7a791f7f033eae57ec2b3a61
change-id: 20260710-spin-trylock-followup-332c636e0d99

Best regards,

Signature · 1 line
Brendan Jackman <[email protected]>
#1 lore raw

PATCH v3 1/4 mm/page_alloc: rename FPI_TRYLOCK -> FPI_NOLOCK

As discussed in the linked patch, the there is some inconsistency between "trylock" and "nolock" nomenclature, let's align it. Since "nolock" is used in the public API it seems to have more mindshare so do that.

The linked patch did this for the ALLOC_ flag but forgot about FPI_.

---
 mm/page_alloc.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)
mm/page_alloc.c +9 -9
index af63558391345..c2da85e69a0f8 100644
@@ -90,7 +90,7 @@ typedef int __bitwise fpi_t;
 #define FPI_TO_TAIL		((__force fpi_t)BIT(1))
 
 /* Free the page without taking locks. Rely on trylock only. */
-#define FPI_TRYLOCK		((__force fpi_t)BIT(2))
+#define FPI_NOLOCK		((__force fpi_t)BIT(2))
 
 /* free_pages_prepare() has already been called for page(s) being freed. */
 #define FPI_PREPARED		((__force fpi_t)BIT(3))
@@ -1419,7 +1419,7 @@ static __always_inline bool __free_pages_prepare(struct page *page,
 	page_table_check_free(page, order);
 	pgalloc_tag_sub(page, 1 << order);
 
-	if (!PageHighMem(page) && !(fpi_flags & FPI_TRYLOCK)) {
+	if (!PageHighMem(page) && !(fpi_flags & FPI_NOLOCK)) {
 		debug_check_no_locks_freed(page_address(page),
 					   PAGE_SIZE << order);
 		debug_check_no_obj_freed(page_address(page),
@@ -1558,7 +1558,7 @@ static void free_one_page(struct zone *zone, struct page *page,
 	struct llist_head *llhead;
 	unsigned long flags;
 
-	if (unlikely(fpi_flags & FPI_TRYLOCK)) {
+	if (unlikely(fpi_flags & FPI_NOLOCK)) {
 		if (!spin_trylock_irqsave(&zone->lock, flags)) {
 			add_page_to_zone_llist(zone, page, order);
 			return;
@@ -1569,7 +1569,7 @@ static void free_one_page(struct zone *zone, struct page *page,
 
 	/* The lock succeeded. Process deferred pages. */
 	llhead = &zone->trylock_free_pages;
-	if (unlikely(!llist_empty(llhead) && !(fpi_flags & FPI_TRYLOCK))) {
+	if (unlikely(!llist_empty(llhead) && !(fpi_flags & FPI_NOLOCK))) {
 		struct llist_node *llnode;
 		struct page *p, *tmp;
 
@@ -2882,7 +2882,7 @@ static bool free_frozen_page_commit(struct zone *zone,
 	if (pcp->free_count < (batch << CONFIG_PCP_BATCH_SCALE_MAX))
 		pcp->free_count += (1 << order);
 
-	if (unlikely(fpi_flags & FPI_TRYLOCK)) {
+	if (unlikely(fpi_flags & FPI_NOLOCK)) {
 		/*
 		 * Do not attempt to take a zone lock. Let pcp->count get
 		 * over high mark temporarily.
@@ -2979,7 +2979,7 @@ static void __free_frozen_pages(struct page *page, unsigned int order,
 		migratetype = MIGRATE_MOVABLE;
 	}
 
-	if (unlikely((fpi_flags & FPI_TRYLOCK) && IS_ENABLED(CONFIG_PREEMPT_RT)
+	if (unlikely((fpi_flags & FPI_NOLOCK) && IS_ENABLED(CONFIG_PREEMPT_RT)
 		     && (in_nmi() || in_hardirq()))) {
 		add_page_to_zone_llist(zone, page, order);
 		return;
@@ -3002,7 +3002,7 @@ void free_frozen_pages(struct page *page, unsigned int order)
 
 void free_frozen_pages_nolock(struct page *page, unsigned int order)
 {
-	__free_frozen_pages(page, order, FPI_TRYLOCK);
+	__free_frozen_pages(page, order, FPI_NOLOCK);
 }
 
 /*
@@ -5399,7 +5399,7 @@ struct page *__alloc_frozen_pages_noprof(gfp_t gfp, unsigned int order,
 	if (memcg_kmem_online() && (gfp & __GFP_ACCOUNT) && page &&
 	    unlikely(__memcg_kmem_charge_page(page, gfp, order) != 0)) {
 		__free_frozen_pages(page, order,
-				    alloc_flags & ALLOC_NOLOCK ? FPI_TRYLOCK : 0);
+				    alloc_flags & ALLOC_NOLOCK ? FPI_NOLOCK : 0);
 		page = NULL;
 	}
 
@@ -5522,7 +5522,7 @@ EXPORT_SYMBOL(__free_pages);
  */
 void free_pages_nolock(struct page *page, unsigned int order)
 {
-	___free_pages(page, order, FPI_TRYLOCK);
+	___free_pages(page, order, FPI_NOLOCK);
 }
 
 /**
Signature · 1 line
2.54.0
maintainer Reviewed-by on v3 1/4
#2 lore raw

Re: [PATCH v3 1/4] mm/page_alloc: rename FPI_TRYLOCK -> FPI_NOLOCK

Quoted 9 lines · Brendan Jackman

As discussed in the linked patch, the there is some inconsistency between "trylock" and "nolock" nomenclature, let's align it. Since "nolock" is used in the public API it seems to have more mindshare so do that.

The linked patch did this for the ALLOC_ flag but forgot about FPI_.

Quoted 90 lines
---
 mm/page_alloc.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)
mm/page_alloc.c +9 -9
index af63558391345..c2da85e69a0f8 100644
@@ -90,7 +90,7 @@ typedef int __bitwise fpi_t;
 #define FPI_TO_TAIL		((__force fpi_t)BIT(1))
 
 /* Free the page without taking locks. Rely on trylock only. */
-#define FPI_TRYLOCK		((__force fpi_t)BIT(2))
+#define FPI_NOLOCK		((__force fpi_t)BIT(2))
 
 /* free_pages_prepare() has already been called for page(s) being freed. */
 #define FPI_PREPARED		((__force fpi_t)BIT(3))
@@ -1419,7 +1419,7 @@ static __always_inline bool __free_pages_prepare(struct page *page,
 	page_table_check_free(page, order);
 	pgalloc_tag_sub(page, 1 << order);
 
-	if (!PageHighMem(page) && !(fpi_flags & FPI_TRYLOCK)) {
+	if (!PageHighMem(page) && !(fpi_flags & FPI_NOLOCK)) {
 		debug_check_no_locks_freed(page_address(page),
 					   PAGE_SIZE << order);
 		debug_check_no_obj_freed(page_address(page),
@@ -1558,7 +1558,7 @@ static void free_one_page(struct zone *zone, struct page *page,
 	struct llist_head *llhead;
 	unsigned long flags;
 
-	if (unlikely(fpi_flags & FPI_TRYLOCK)) {
+	if (unlikely(fpi_flags & FPI_NOLOCK)) {
 		if (!spin_trylock_irqsave(&zone->lock, flags)) {
 			add_page_to_zone_llist(zone, page, order);
 			return;
@@ -1569,7 +1569,7 @@ static void free_one_page(struct zone *zone, struct page *page,
 
 	/* The lock succeeded. Process deferred pages. */
 	llhead = &zone->trylock_free_pages;
-	if (unlikely(!llist_empty(llhead) && !(fpi_flags & FPI_TRYLOCK))) {
+	if (unlikely(!llist_empty(llhead) && !(fpi_flags & FPI_NOLOCK))) {
 		struct llist_node *llnode;
 		struct page *p, *tmp;
 
@@ -2882,7 +2882,7 @@ static bool free_frozen_page_commit(struct zone *zone,
 	if (pcp->free_count < (batch << CONFIG_PCP_BATCH_SCALE_MAX))
 		pcp->free_count += (1 << order);
 
-	if (unlikely(fpi_flags & FPI_TRYLOCK)) {
+	if (unlikely(fpi_flags & FPI_NOLOCK)) {
 		/*
 		 * Do not attempt to take a zone lock. Let pcp->count get
 		 * over high mark temporarily.
@@ -2979,7 +2979,7 @@ static void __free_frozen_pages(struct page *page, unsigned int order,
 		migratetype = MIGRATE_MOVABLE;
 	}
 
-	if (unlikely((fpi_flags & FPI_TRYLOCK) && IS_ENABLED(CONFIG_PREEMPT_RT)
+	if (unlikely((fpi_flags & FPI_NOLOCK) && IS_ENABLED(CONFIG_PREEMPT_RT)
 		     && (in_nmi() || in_hardirq()))) {
 		add_page_to_zone_llist(zone, page, order);
 		return;
@@ -3002,7 +3002,7 @@ void free_frozen_pages(struct page *page, unsigned int order)
 
 void free_frozen_pages_nolock(struct page *page, unsigned int order)
 {
-	__free_frozen_pages(page, order, FPI_TRYLOCK);
+	__free_frozen_pages(page, order, FPI_NOLOCK);
 }
 
 /*
@@ -5399,7 +5399,7 @@ struct page *__alloc_frozen_pages_noprof(gfp_t gfp, unsigned int order,
 	if (memcg_kmem_online() && (gfp & __GFP_ACCOUNT) && page &&
 	    unlikely(__memcg_kmem_charge_page(page, gfp, order) != 0)) {
 		__free_frozen_pages(page, order,
-				    alloc_flags & ALLOC_NOLOCK ? FPI_TRYLOCK : 0);
+				    alloc_flags & ALLOC_NOLOCK ? FPI_NOLOCK : 0);
 		page = NULL;
 	}
 
@@ -5522,7 +5522,7 @@ EXPORT_SYMBOL(__free_pages);
  */
 void free_pages_nolock(struct page *page, unsigned int order)
 {
-	___free_pages(page, order, FPI_TRYLOCK);
+	___free_pages(page, order, FPI_NOLOCK);
 }
 
 /**
#3 lore raw

PATCH v3 2/4 cgroup/cpuset: update some comments about the page allocator

In reply to Brendan Jackman

These comments describing the page allocator are out of date:

- __alloc_pages() is no longer a public API and has no business being described outside of mm/.

- The wait variable is gone.

It may be out of date for other reasons too but this patch is just fixing the issues that stood out.

To fix it:

- Instead of referring to a specific function, instead to "the page allocator"

- Completely drop out-of-date details of that function's internal behaviour, since they were irrelevant anyway.

---
 kernel/cgroup/cpuset.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)
kernel/cgroup/cpuset.c +5 -8
index 24ea2d09cdbdb..dfd0f827e3b92 100644
@@ -4193,7 +4193,7 @@ static struct cpuset *nearest_hardwall_ancestor(struct cpuset *cs)
  * nearest enclosing hardwalled ancestor cpuset.
  *
  * Scanning up parent cpusets requires callback_lock.  The
- * __alloc_pages() routine only calls here with __GFP_HARDWALL bit
+ * page allocator only calls here with __GFP_HARDWALL bit
  * _not_ set if it's a GFP_KERNEL allocation, and all nodes in the
  * current tasks mems_allowed came up empty on the first pass over
  * the zonelist.  So only GFP_KERNEL allocations, if all nodes in the
@@ -4206,11 +4206,8 @@ static struct cpuset *nearest_hardwall_ancestor(struct cpuset *cs)
  * come before the __GFP_HARDWALL check, otherwise a dying task
  * would be blocked on the fast path.
  *
- * The second pass through get_page_from_freelist() doesn't even call
- * here for GFP_ATOMIC calls.  For those calls, the __alloc_pages()
- * variable 'wait' is not set, and the bit ALLOC_CPUSET is not set
- * in alloc_flags.  That logic and the checks below have the combined
- * affect that:
+ * The second pass through get_page_from_freelist() doesn't even call here for
+ * GFP_ATOMIC calls.  That, and the checks below have the combined affect that:
  *	in_interrupt - any node ok (current task context irrelevant)
  *	GFP_ATOMIC   - any node ok
  *	tsk_is_oom_victim   - any node ok
@@ -4327,8 +4324,8 @@ void cpuset_nodes_allowed(struct cgroup *cgroup, nodemask_t *mask)
  * should not be possible for the following code to return an
  * offline node.  But if it did, that would be ok, as this routine
  * is not returning the node where the allocation must be, only
- * the node where the search should start.  The zonelist passed to
- * __alloc_pages() will include all nodes.  If the slab allocator
+ * the node where the search should start.  The zonelist used by
+ * the allocator will include all nodes.  If the slab allocator
  * is passed an offline node, it will fall back to the local node.
  * See kmem_cache_alloc_node().
  */
Signature · 1 line
2.54.0
maintainer Reviewed-by on v3 2/4
#4 lore raw

Re: [PATCH v3 2/4] cgroup/cpuset: update some comments about the page allocator

Quoted 15 lines · Brendan Jackman

These comments describing the page allocator are out of date:

- __alloc_pages() is no longer a public API and has no business being described outside of mm/.

- The wait variable is gone.

It may be out of date for other reasons too but this patch is just fixing the issues that stood out.

To fix it:

- Instead of referring to a specific function, instead to "the page allocator"

Quoted 7 lines

- Completely drop out-of-date details of that function's internal behaviour, since they were irrelevant anyway.

Quoted 43 lines
---
 kernel/cgroup/cpuset.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)
kernel/cgroup/cpuset.c +5 -8
index 24ea2d09cdbdb..dfd0f827e3b92 100644
@@ -4193,7 +4193,7 @@ static struct cpuset *nearest_hardwall_ancestor(struct cpuset *cs)
  * nearest enclosing hardwalled ancestor cpuset.
  *
  * Scanning up parent cpusets requires callback_lock.  The
- * __alloc_pages() routine only calls here with __GFP_HARDWALL bit
+ * page allocator only calls here with __GFP_HARDWALL bit
  * _not_ set if it's a GFP_KERNEL allocation, and all nodes in the
  * current tasks mems_allowed came up empty on the first pass over
  * the zonelist.  So only GFP_KERNEL allocations, if all nodes in the
@@ -4206,11 +4206,8 @@ static struct cpuset *nearest_hardwall_ancestor(struct cpuset *cs)
  * come before the __GFP_HARDWALL check, otherwise a dying task
  * would be blocked on the fast path.
  *
- * The second pass through get_page_from_freelist() doesn't even call
- * here for GFP_ATOMIC calls.  For those calls, the __alloc_pages()
- * variable 'wait' is not set, and the bit ALLOC_CPUSET is not set
- * in alloc_flags.  That logic and the checks below have the combined
- * affect that:
+ * The second pass through get_page_from_freelist() doesn't even call here for
+ * GFP_ATOMIC calls.  That, and the checks below have the combined affect that:
  *	in_interrupt - any node ok (current task context irrelevant)
  *	GFP_ATOMIC   - any node ok
  *	tsk_is_oom_victim   - any node ok
@@ -4327,8 +4324,8 @@ void cpuset_nodes_allowed(struct cgroup *cgroup, nodemask_t *mask)
  * should not be possible for the following code to return an
  * offline node.  But if it did, that would be ok, as this routine
  * is not returning the node where the allocation must be, only
- * the node where the search should start.  The zonelist passed to
- * __alloc_pages() will include all nodes.  If the slab allocator
+ * the node where the search should start.  The zonelist used by
+ * the allocator will include all nodes.  If the slab allocator
  * is passed an offline node, it will fall back to the local node.
  * See kmem_cache_alloc_node().
  */
maintainer Acked-by on v3 2/4
#5 lore raw

Re: [PATCH v3 2/4] cgroup/cpuset: update some comments about the page allocator

In reply to Brendan Jackman
Quoted 54 lines · Brendan Jackman

These comments describing the page allocator are out of date:

- __alloc_pages() is no longer a public API and has no business being described outside of mm/.

- The wait variable is gone.

It may be out of date for other reasons too but this patch is just fixing the issues that stood out.

To fix it:

- Instead of referring to a specific function, instead to "the page allocator"

- Completely drop out-of-date details of that function's internal behaviour, since they were irrelevant anyway.

---
  kernel/cgroup/cpuset.c | 13 +++++--------
  1 file changed, 5 insertions(+), 8 deletions(-)
kernel/cgroup/cpuset.c +3 -6
index 24ea2d09cdbdb..dfd0f827e3b92 100644
@@ -4193,7 +4193,7 @@ static struct cpuset *nearest_hardwall_ancestor(struct cpuset *cs)
   * nearest enclosing hardwalled ancestor cpuset.
   *
   * Scanning up parent cpusets requires callback_lock.  The
- * __alloc_pages() routine only calls here with __GFP_HARDWALL bit
+ * page allocator only calls here with __GFP_HARDWALL bit
   * _not_ set if it's a GFP_KERNEL allocation, and all nodes in the
   * current tasks mems_allowed came up empty on the first pass over
   * the zonelist.  So only GFP_KERNEL allocations, if all nodes in the
@@ -4206,11 +4206,8 @@ static struct cpuset *nearest_hardwall_ancestor(struct cpuset *cs)
   * come before the __GFP_HARDWALL check, otherwise a dying task
   * would be blocked on the fast path.
   *
- * The second pass through get_page_from_freelist() doesn't even call
- * here for GFP_ATOMIC calls.  For those calls, the __alloc_pages()
- * variable 'wait' is not set, and the bit ALLOC_CPUSET is not set
- * in alloc_flags.  That logic and the checks below have the combined
- * affect that:
+ * The second pass through get_page_from_freelist() doesn't even call here for
+ * GFP_ATOMIC calls.  That, and the checks below have the combined affect that:
   *	in_interrupt - any node ok (current task context irrelevant)
   *	GFP_ATOMIC   - any node ok
   *	tsk_is_oom_victim   - any node ok

kernel/cgroup/cpuset.c void cpuset_nodes_allowed(struct cgroup *cgroup, nodemask_t *mask) · line 4331

Quoted 10 lines
kernel/cgroup/cpuset.c +2 -2
@@ -4327,8 +4324,8 @@ void cpuset_nodes_allowed(struct cgroup *cgroup, nodemask_t *mask)
   * should not be possible for the following code to return an
   * offline node.  But if it did, that would be ok, as this routine
   * is not returning the node where the allocation must be, only
- * the node where the search should start.  The zonelist passed to
- * __alloc_pages() will include all nodes.  If the slab allocator
+ * the node where the search should start.  The zonelist used by
+ * the allocator will include all nodes.  If the slab allocator
   * is passed an offline node, it will fall back to the local node.
   * See kmem_cache_alloc_node().
   */
maintainer Applied
#6 lore raw

Re: [PATCH v3 2/4] cgroup/cpuset: update some comments about the page allocator

In reply to Brendan Jackman

Hello,

Applied this one to cgroup/for-7.3 with Vlastimil's reviewed-by and Waiman's ack added. Please drop it from the mm series.

Thanks.

Signature · 1 line
tejun
maintainer
#7 lore raw

Re: [PATCH v3 2/4] cgroup/cpuset: update some comments about the page allocator

Quoted 4 lines · Tejun Heo

Hello,

Applied this one to cgroup/for-7.3 with Vlastimil's reviewed-by and Waiman's ack added. Please drop it from the mm series.

Never mind. I'll drop it from mine. Sorry about the noise.

Signature · 1 line
tejun
maintainer
#8 lore raw

Re: [PATCH v3 2/4] cgroup/cpuset: update some comments about the page allocator

Quoted 7 lines · On Fri, 17 Jul 2026 07:26:28 -1000 Tejun Heo <[email protected]> wrote:
Quoted 4 lines · Tejun Heo

Hello,

Applied this one to cgroup/for-7.3 with Vlastimil's reviewed-by and Waiman's ack added. Please drop it from the mm series.

Never mind. I'll drop it from mine. Sorry about the noise.

Thanks. Plucking one patch from a 4-patch series is awkward. Better to retain this in mm.git with your ack(?).

maintainer Acked-by
#9 lore raw

Re: [PATCH v3 2/4] cgroup/cpuset: update some comments about the page allocator

Quoted 5 lines · Andrew Morton
Quoted 3 lines · On Fri, 17 Jul 2026 07:26:28 -1000 Tejun Heo <[email protected]> wrote:
Quoted 2 lines · Tejun Heo

Hello,

Quoted 7 lines
Quoted 4 lines
Quoted 2 lines

Applied this one to cgroup/for-7.3 with Vlastimil's reviewed-by and Waiman's ack added. Please drop it from the mm series.

Never mind. I'll drop it from mine. Sorry about the noise.

Thanks. Plucking one patch from a 4-patch series is awkward. Better to retain this in mm.git with your ack(?).

Yes, that was a tooling problem. Shouldn't happen again.

Thanks.

Signature · 1 line
tejun
#10 lore raw

PATCH v3 3/4 mm/page_alloc: fixup alloc_pages_nolock_noprof() comment

In reply to Brendan Jackman

Update the comment to reflect the recent change to allow flags in gfp_nolock.

---
 mm/page_alloc.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
mm/page_alloc.c +2 -1
index c2da85e69a0f8..25a83a57aab66 100644
@@ -7983,7 +7983,8 @@ struct page *alloc_frozen_pages_nolock_noprof(gfp_t gfp_flags, int nid, unsigned
 }
 /**
  * alloc_pages_nolock - opportunistic reentrant allocation from any context
- * @gfp_flags: GFP flags. Only __GFP_ACCOUNT allowed.
+ * @gfp_flags: GFP flags. Only __GFP_ACCOUNT, plus some flags that get set
+ *             internally regardless (see %gfp_nolock) are allowed.
  * @nid: node to allocate from
  * @order: allocation order size
  *
Signature · 1 line
2.54.0
maintainer Reviewed-by on v3 3/4
#11 lore raw

Re: [PATCH v3 3/4] mm/page_alloc: fixup alloc_pages_nolock_noprof() comment

Quoted 7 lines · Brendan Jackman

Update the comment to reflect the recent change to allow flags in gfp_nolock.

Quoted 19 lines
---
 mm/page_alloc.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
mm/page_alloc.c +2 -1
index c2da85e69a0f8..25a83a57aab66 100644
@@ -7983,7 +7983,8 @@ struct page *alloc_frozen_pages_nolock_noprof(gfp_t gfp_flags, int nid, unsigned
 }
 /**
  * alloc_pages_nolock - opportunistic reentrant allocation from any context
- * @gfp_flags: GFP flags. Only __GFP_ACCOUNT allowed.
+ * @gfp_flags: GFP flags. Only __GFP_ACCOUNT, plus some flags that get set
+ *             internally regardless (see %gfp_nolock) are allowed.
  * @nid: node to allocate from
  * @order: allocation order size
  *
#12 lore raw

PATCH v3 4/4 mm/page_alloc: remove a couple of VM_BUG_ON()st

In reply to Brendan Jackman

VM_BUG_ON() is out of favour and on the way to removal, since I recently touched alloc_pages_node_noprof() I am removing that invocation, and also removing the __folio_alloc_node_noprof() one for consistency. If this precondition is violated, the system will soon crash anyway.

---
 include/linux/gfp.h | 1 -
 mm/page_alloc.c     | 1 -
 2 files changed, 2 deletions(-)
include/linux/gfp.h +0 -1
index 4d57e9c0bf204..872bc53f32ec8 100644
@@ -255,7 +255,6 @@ static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask)
 static inline
 struct folio *__folio_alloc_node_noprof(gfp_t gfp, unsigned int order, int nid)
 {
-	VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
 	warn_if_node_offline(nid, gfp);
 
 	return __folio_alloc_noprof(gfp, order, nid, NULL);
mm/page_alloc.c +0 -1
index 25a83a57aab66..4c6815f84adc6 100644
@@ -5427,7 +5427,6 @@ struct page *alloc_pages_node_noprof(int nid, gfp_t gfp_mask, unsigned int order
 	if (nid == NUMA_NO_NODE)
 		nid = numa_mem_id();
 
-	VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
 	warn_if_node_offline(nid, gfp_mask);
 
 	return __alloc_pages_noprof(gfp_mask, order, nid, NULL, ALLOC_DEFAULT);
Signature · 1 line
2.54.0
maintainer Reviewed-by on v3 4/4
#13 lore raw

Re: [PATCH v3 4/4] mm/page_alloc: remove a couple of VM_BUG_ON()st

Subject has stray 't' at the end?

Quoted 9 lines · Brendan Jackman

VM_BUG_ON() is out of favour and on the way to removal, since I recently touched alloc_pages_node_noprof() I am removing that invocation, and also removing the __folio_alloc_node_noprof() one for consistency. If this precondition is violated, the system will soon crash anyway.

Quoted 10 lines
---
 include/linux/gfp.h | 1 -
 mm/page_alloc.c     | 1 -
 2 files changed, 2 deletions(-)
include/linux/gfp.h +0 -0
index 4d57e9c0bf204..872bc53f32ec8 100644

include/linux/gfp.h static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask) · line 260

Quoted 7 lines
include/linux/gfp.h +0 -1
@@ -255,7 +255,6 @@ static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask)
 static inline
 struct folio *__folio_alloc_node_noprof(gfp_t gfp, unsigned int order, int nid)
 {
-	VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
 	warn_if_node_offline(nid, gfp);
 
 	return __folio_alloc_noprof(gfp, order, nid, NULL);

Well if you want more cleanups, I can see in iommu_alloc_pages_node_sz():

        /*
         * __folio_alloc_node() does not handle NUMA_NO_NODE like
         * alloc_pages_node() did.
         */
        if (nid == NUMA_NO_NODE)
                nid = numa_mem_id();

        folio = __folio_alloc_node(gfp | __GFP_ZERO, order, nid);

Should we introduce folio_alloc_node() and make __folio_alloc_node() mm-internal, for consistency?

Quoted 13 lines
mm/page_alloc.c +0 -1
index 25a83a57aab66..4c6815f84adc6 100644
@@ -5427,7 +5427,6 @@ struct page *alloc_pages_node_noprof(int nid, gfp_t gfp_mask, unsigned int order
 	if (nid == NUMA_NO_NODE)
 		nid = numa_mem_id();
 
-	VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
 	warn_if_node_offline(nid, gfp_mask);
 
 	return __alloc_pages_noprof(gfp_mask, order, nid, NULL, ALLOC_DEFAULT);
#14 lore raw

Re: [PATCH v3 4/4] mm/page_alloc: remove a couple of VM_BUG_ON()st

Quoted 1 line · Vlastimil Babka (SUSE)

Subject has stray 't' at the end?

Thanks - will fix if we do a v3 (otherwise Andrew, please can you amend when you apply it?)

Quoted 43 lines · Brendan Jackman
Quoted 9 lines

VM_BUG_ON() is out of favour and on the way to removal, since I recently touched alloc_pages_node_noprof() I am removing that invocation, and also removing the __folio_alloc_node_noprof() one for consistency. If this precondition is violated, the system will soon crash anyway.

Quoted 10 lines
---
 include/linux/gfp.h | 1 -
 mm/page_alloc.c     | 1 -
 2 files changed, 2 deletions(-)
include/linux/gfp.h +0 -0
index 4d57e9c0bf204..872bc53f32ec8 100644

include/linux/gfp.h static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask) · line 260

Quoted 7 lines
include/linux/gfp.h +0 -1
@@ -255,7 +255,6 @@ static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask)
 static inline
 struct folio *__folio_alloc_node_noprof(gfp_t gfp, unsigned int order, int nid)
 {
-	VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
 	warn_if_node_offline(nid, gfp);
 
 	return __folio_alloc_noprof(gfp, order, nid, NULL);

Well if you want more cleanups, I can see in iommu_alloc_pages_node_sz():

        /*
         * __folio_alloc_node() does not handle NUMA_NO_NODE like
         * alloc_pages_node() did.
         */
        if (nid == NUMA_NO_NODE)
                nid = numa_mem_id();

        folio = __folio_alloc_node(gfp | __GFP_ZERO, order, nid);
Quoted 2 lines

Should we introduce folio_alloc_node() and make __folio_alloc_node() mm-internal, for consistency?

Ha, I literally just wrote that patch. I'm planning to do it as yet another series as there's a little dance needed to get it all into shape. But, also happy to just expand this one if you prefer.

Then I'm gonna add alloc_flags to the __ variant so filemap.c can set ALLOC_UNMAPPED for AS_NO_DIRECT_MAP.

#15 lore raw

Re: [PATCH v3 4/4] mm/page_alloc: remove a couple of VM_BUG_ON()st

Quoted 47 lines · Brendan Jackman
Quoted 1 line · Vlastimil Babka (SUSE)

Subject has stray 't' at the end?

Thanks - will fix if we do a v3 (otherwise Andrew, please can you amend when you apply it?)

Quoted 41 lines · Brendan Jackman
Quoted 9 lines

VM_BUG_ON() is out of favour and on the way to removal, since I recently touched alloc_pages_node_noprof() I am removing that invocation, and also removing the __folio_alloc_node_noprof() one for consistency. If this precondition is violated, the system will soon crash anyway.

Quoted 10 lines
---
 include/linux/gfp.h | 1 -
 mm/page_alloc.c     | 1 -
 2 files changed, 2 deletions(-)
include/linux/gfp.h +0 -0
index 4d57e9c0bf204..872bc53f32ec8 100644

include/linux/gfp.h static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask) · line 260

Quoted 7 lines
include/linux/gfp.h +0 -1
@@ -255,7 +255,6 @@ static inline void warn_if_node_offline(int this_node, gfp_t gfp_mask)
 static inline
 struct folio *__folio_alloc_node_noprof(gfp_t gfp, unsigned int order, int nid)
 {
-	VM_BUG_ON(nid < 0 || nid >= MAX_NUMNODES);
 	warn_if_node_offline(nid, gfp);
 
 	return __folio_alloc_noprof(gfp, order, nid, NULL);

Well if you want more cleanups, I can see in iommu_alloc_pages_node_sz():

        /*
         * __folio_alloc_node() does not handle NUMA_NO_NODE like
         * alloc_pages_node() did.
         */
        if (nid == NUMA_NO_NODE)
                nid = numa_mem_id();
Quoted 7 lines
Quoted 4 lines
        folio = __folio_alloc_node(gfp | __GFP_ZERO, order, nid);

Should we introduce folio_alloc_node() and make __folio_alloc_node() mm-internal, for consistency?

Ha, I literally just wrote that patch. I'm planning to do it as yet another series as there's a little dance needed to get it all into

Cool!

Quoted 1 line

shape. But, also happy to just expand this one if you prefer.

Not necessary, another series is fine.

Quoted 2 lines

Then I'm gonna add alloc_flags to the __ variant so filemap.c can set ALLOC_UNMAPPED for AS_NO_DIRECT_MAP.

Great!