Skip to contentSkip to conversation

mm/mmu_notifier: fix mmu_notifier_range_init warning

PATCH In review Arnd Bergmann 3 participants 8 messages Dec 11, 2018

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm/mmu_notifier: fix mmu_notifier_range_init warning

The macro version of mmu_notifier_range_init() for CONFIG_MMU_NOTIFIER=n does not evaluate all its arguments, leading to a warning in one case:

mm/migrate.c: In function 'migrate_vma_pages':
mm/migrate.c:2711:20: error: unused variable 'mm' [-Werror=unused-variable]
struct mm_struct *mm = vma->vm_mm;

Pass down the 'mm' as into the inline function as well so gcc can see why the variable exists.

Fixes 137d92bd73b1 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")
---
 include/linux/mmu_notifier.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
include/linux/mmu_notifier.h +2 -1
index 29f7b9670ba3..b13ea00ded5d 100644
@@ -476,6 +476,7 @@ struct mmu_notifier_range {
 };
 
 static inline void _mmu_notifier_range_init(struct mmu_notifier_range *range,
+					    struct mm_struct *mm,
 					    unsigned long start,
 					    unsigned long end)
 {
@@ -484,7 +485,7 @@ static inline void _mmu_notifier_range_init(struct mmu_notifier_range *range,
 }
 
 #define mmu_notifier_range_init(range, mm, start, end, event) \
-	_mmu_notifier_range_init(range, start, end)
+	_mmu_notifier_range_init(range, mm, start, end)
 
 
 static inline int mm_has_notifiers(struct mm_struct *mm)
Signature · 1 line
2.20.0
#1 lore raw
Quoted 11 lines · Arnd Bergmann

The macro version of mmu_notifier_range_init() for CONFIG_MMU_NOTIFIER=n does not evaluate all its arguments, leading to a warning in one case:

mm/migrate.c: In function 'migrate_vma_pages':
mm/migrate.c:2711:20: error: unused variable 'mm' [-Werror=unused-variable]
struct mm_struct *mm = vma->vm_mm;

Pass down the 'mm' as into the inline function as well so gcc can see why the variable exists.

Fixes 137d92bd73b1 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")

What about changing migrate.c (i actualy tried to do that everywhere in the patchset but i missed that spot) So we avoid one useless variable on such configuration:

mm/migrate.c +2 -3
index f02bb4b22c1a..883fce631f47 100644
@@ -2701,7 +2701,6 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 	const unsigned long npages = migrate->npages;
 	const unsigned long start = migrate->start;
 	struct vm_area_struct *vma = migrate->vma;
-	struct mm_struct *mm = vma->vm_mm;
 	struct mmu_notifier_range range;
 	unsigned long addr, i;
 	bool notified = false;
@@ -2724,8 +2723,8 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 			if (!notified) {
 				notified = true;
 
-				mmu_notifier_range_init(&range, mm, addr,
-							migrate->end,
+				mmu_notifier_range_init(&range, vma->vm_mm,
+							addr, migrate->end,
 							MMU_NOTIFY_CLEAR);
 				mmu_notifier_invalidate_range_start(&range);
 			}
Quoted 29 lines
---
 include/linux/mmu_notifier.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
include/linux/mmu_notifier.h +2 -1
index 29f7b9670ba3..b13ea00ded5d 100644
@@ -476,6 +476,7 @@ struct mmu_notifier_range {
 };
 
 static inline void _mmu_notifier_range_init(struct mmu_notifier_range *range,
+					    struct mm_struct *mm,
 					    unsigned long start,
 					    unsigned long end)
 {
@@ -484,7 +485,7 @@ static inline void _mmu_notifier_range_init(struct mmu_notifier_range *range,
 }
 
 #define mmu_notifier_range_init(range, mm, start, end, event) \
-	_mmu_notifier_range_init(range, start, end)
+	_mmu_notifier_range_init(range, mm, start, end)
 
 
 static inline int mm_has_notifiers(struct mm_struct *mm)
Signature · 1 line
2.20.0
#2 lore raw
Quoted 30 lines · Jerome Glisse
Quoted 11 lines · Arnd Bergmann

The macro version of mmu_notifier_range_init() for CONFIG_MMU_NOTIFIER=n does not evaluate all its arguments, leading to a warning in one case:

mm/migrate.c: In function 'migrate_vma_pages':
mm/migrate.c:2711:20: error: unused variable 'mm' [-Werror=unused-variable]
struct mm_struct *mm = vma->vm_mm;

Pass down the 'mm' as into the inline function as well so gcc can see why the variable exists.

Fixes 137d92bd73b1 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")

What about changing migrate.c (i actualy tried to do that everywhere in the patchset but i missed that spot) So we avoid one useless variable on such configuration:

mm/migrate.c +0 -1
index f02bb4b22c1a..883fce631f47 100644
@@ -2701,7 +2701,6 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 	const unsigned long npages = migrate->npages;
 	const unsigned long start = migrate->start;
 	struct vm_area_struct *vma = migrate->vma;
-	struct mm_struct *mm = vma->vm_mm;
 	struct mmu_notifier_range range;
 	unsigned long addr, i;
 	bool notified = false;

mm/migrate.c static void migrate_vma_pages(struct migrate_vma *migrate) · line 2730

Quoted 10 lines
mm/migrate.c +2 -2
@@ -2724,8 +2723,8 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 			if (!notified) {
 				notified = true;
 
-				mmu_notifier_range_init(&range, mm, addr,
-							migrate->end,
+				mmu_notifier_range_init(&range, vma->vm_mm,
+							addr, migrate->end,
 							MMU_NOTIFY_CLEAR);
 				mmu_notifier_invalidate_range_start(&range);
 			}

Wouldn't it be better to just declare mmu_notifier_range_init() as a static inline function rather than a #define to avoid this warning?

#3 lore raw
Quoted 35 lines · David Rientjes
Quoted 33 lines · Jerome Glisse
Quoted 11 lines · Arnd Bergmann

The macro version of mmu_notifier_range_init() for CONFIG_MMU_NOTIFIER=n does not evaluate all its arguments, leading to a warning in one case:

mm/migrate.c: In function 'migrate_vma_pages':
mm/migrate.c:2711:20: error: unused variable 'mm' [-Werror=unused-variable]
struct mm_struct *mm = vma->vm_mm;

Pass down the 'mm' as into the inline function as well so gcc can see why the variable exists.

Fixes 137d92bd73b1 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")

What about changing migrate.c (i actualy tried to do that everywhere in the patchset but i missed that spot) So we avoid one useless variable on such configuration:

mm/migrate.c +0 -1
index f02bb4b22c1a..883fce631f47 100644
@@ -2701,7 +2701,6 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 	const unsigned long npages = migrate->npages;
 	const unsigned long start = migrate->start;
 	struct vm_area_struct *vma = migrate->vma;
-	struct mm_struct *mm = vma->vm_mm;
 	struct mmu_notifier_range range;
 	unsigned long addr, i;
 	bool notified = false;
@@ -2724,8 +2723,8 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 			if (!notified) {
 				notified = true;
 
Quoted 10 lines

mm/migrate.c static void migrate_vma_pages(struct migrate_vma *migrate) · line 2730

Quoted 7 lines
mm/migrate.c +2 -2
-				mmu_notifier_range_init(&range, mm, addr,
-							migrate->end,
+				mmu_notifier_range_init(&range, vma->vm_mm,
+							addr, migrate->end,
 							MMU_NOTIFY_CLEAR);
 				mmu_notifier_invalidate_range_start(&range);
 			}

Wouldn't it be better to just declare mmu_notifier_range_init() as a static inline function rather than a #define to avoid this warning?

The define trick is use to drop the event parameter so that we do not need to define the event value for CONFIG_MMU_NOTIFIER=n case.

Cheers,
Jérôme

#4 lore raw
Quoted 43 lines · Jerome Glisse
Quoted 32 lines
Quoted 32 lines
Quoted 11 lines

The macro version of mmu_notifier_range_init() for CONFIG_MMU_NOTIFIER=n does not evaluate all its arguments, leading to a warning in one case:

mm/migrate.c: In function 'migrate_vma_pages':
mm/migrate.c:2711:20: error: unused variable 'mm' [-Werror=unused-variable]
struct mm_struct *mm = vma->vm_mm;

Pass down the 'mm' as into the inline function as well so gcc can see why the variable exists.

Fixes 137d92bd73b1 ("mm/mmu_notifier: use structure for invalidate_range_start/end calls v2")

What about changing migrate.c (i actualy tried to do that everywhere in the patchset but i missed that spot) So we avoid one useless variable on such configuration:

mm/migrate.c +0 -1
index f02bb4b22c1a..883fce631f47 100644
@@ -2701,7 +2701,6 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 	const unsigned long npages = migrate->npages;
 	const unsigned long start = migrate->start;
 	struct vm_area_struct *vma = migrate->vma;
-	struct mm_struct *mm = vma->vm_mm;
 	struct mmu_notifier_range range;
 	unsigned long addr, i;
 	bool notified = false;
@@ -2724,8 +2723,8 @@ static void migrate_vma_pages(struct migrate_vma *migrate)
 			if (!notified) {
 				notified = true;
 
Quoted 10 lines

mm/migrate.c static void migrate_vma_pages(struct migrate_vma *migrate) · line 2730

Quoted 7 lines
mm/migrate.c +2 -2
-				mmu_notifier_range_init(&range, mm, addr,
-							migrate->end,
+				mmu_notifier_range_init(&range, vma->vm_mm,
+							addr, migrate->end,
 							MMU_NOTIFY_CLEAR);
 				mmu_notifier_invalidate_range_start(&range);
 			}

Wouldn't it be better to just declare mmu_notifier_range_init() as a static inline function rather than a #define to avoid this warning?

Quoted 3 lines

The define trick is use to drop the event parameter so that we do not need to define the event value for CONFIG_MMU_NOTIFIER=n case.

Hmm, strange that Arnd's build failure is only reporting about an unused variable instead of MMU_NOTIFY_CLEAR being undefined :/

I think this should be done so that anybody using mmu_notifier_range_init() doesn't need to worry about the implications of *any* unused formal parameter as a result of how the #define is formed:

include/linux/mmu_notifier.h +23 -18
@@ -10,21 +10,6 @@
 struct mmu_notifier;
 struct mmu_notifier_ops;
 
-#ifdef CONFIG_MMU_NOTIFIER
-
-/*
- * The mmu notifier_mm structure is allocated and installed in
- * mm->mmu_notifier_mm inside the mm_take_all_locks() protected
- * critical section and it's released only when mm_count reaches zero
- * in mmdrop().
- */
-struct mmu_notifier_mm {
-	/* all mmu notifiers registerd in this mm are queued in this list */
-	struct hlist_head list;
-	/* to serialize the list modifications and hlist_unhashed */
-	spinlock_t lock;
-};
-
 /**
  * enum mmu_notifier_event - reason for the mmu notifier callback
  * @MMU_NOTIFY_UNMAP: either munmap() that unmap the range or a mremap() that
@@ -53,6 +38,21 @@ enum mmu_notifier_event {
 	MMU_NOTIFY_SOFT_DIRTY,
 };
 
+#ifdef CONFIG_MMU_NOTIFIER
+
+/*
+ * The mmu notifier_mm structure is allocated and installed in
+ * mm->mmu_notifier_mm inside the mm_take_all_locks() protected
+ * critical section and it's released only when mm_count reaches zero
+ * in mmdrop().
+ */
+struct mmu_notifier_mm {
+	/* all mmu notifiers registerd in this mm are queued in this list */
+	struct hlist_head list;
+	/* to serialize the list modifications and hlist_unhashed */
+	spinlock_t lock;
+};
+
 struct mmu_notifier_range {
 	struct mm_struct *mm;
 	unsigned long start;
@@ -483,9 +483,14 @@ static inline void _mmu_notifier_range_init(struct mmu_notifier_range *range,
 	range->end = end;
 }
 
-#define mmu_notifier_range_init(range, mm, start, end, event) \
-	_mmu_notifier_range_init(range, start, end)
-
+static inline void mmu_notifier_range_init(struct mmu_notifier_range *range,
+					   struct mm_struct *mm,
+					   unsigned long start,
+					   unsigned long end,
+					   enum mmu_notifier_event event)
+{
+	_mmu_notifier_range_init(range, start, end);
+}
 
 static inline int mm_has_notifiers(struct mm_struct *mm)
 {
#5 lore raw
Quoted 10 lines · David Rientjes

On Tue, 11 Dec 2018, Jerome Glisse wrote:

Hmm, strange that Arnd's build failure is only reporting about an unused variable instead of MMU_NOTIFY_CLEAR being undefined :/

I think this should be done so that anybody using mmu_notifier_range_init() doesn't need to worry about the implications of *any* unused formal parameter as a result of how the #define is formed:

Your patch below is more or less what I tried at first, and that resulted in another build failure for

mm/hugetlb.c: mmu_notifier_range_init(&range, mm, start, end,
MMU_NOTIFY_CLEAR);
mm/hugetlb.c- adjust_range_if_pmd_sharing_possible(vma,
&range.start, &range.end);

where range.end refers to a nonexisting member of range.

Arnd

#6 lore raw
Quoted 15 lines · Arnd Bergmann
Quoted 10 lines · David Rientjes

On Tue, 11 Dec 2018, Jerome Glisse wrote:

Hmm, strange that Arnd's build failure is only reporting about an unused variable instead of MMU_NOTIFY_CLEAR being undefined :/

I think this should be done so that anybody using mmu_notifier_range_init() doesn't need to worry about the implications of *any* unused formal parameter as a result of how the #define is formed:

Your patch below is more or less what I tried at first, and that resulted in another build failure for

Quoted 8 lines

mm/hugetlb.c: mmu_notifier_range_init(&range, mm, start, end,
MMU_NOTIFY_CLEAR);
mm/hugetlb.c- adjust_range_if_pmd_sharing_possible(vma,
&range.start, &range.end);

where range.end refers to a nonexisting member of range.

Arnd

I will post a v3 with htmldoc fix and build warning fix incorporated.

Cheers,
Jérôme

#7 lore raw
In reply to Arnd Bergmann
Quoted 10 lines · Arnd Bergmann
Quoted 6 lines

Hmm, strange that Arnd's build failure is only reporting about an unused variable instead of MMU_NOTIFY_CLEAR being undefined :/

I think this should be done so that anybody using mmu_notifier_range_init() doesn't need to worry about the implications of *any* unused formal parameter as a result of how the #define is formed:

Your patch below is more or less what I tried at first, and that resulted in another build failure for

Quoted 7 lines

mm/hugetlb.c: mmu_notifier_range_init(&range, mm, start, end,
MMU_NOTIFY_CLEAR);
mm/hugetlb.c- adjust_range_if_pmd_sharing_possible(vma,
&range.start, &range.end);

where range.end refers to a nonexisting member of range.

Isn't that separate? It seems like there is a strict dependency that CONFIG_HUGETLBFS has on CONFIG_MMU_NOTIFIER.