[PATCH 0/3] userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request

[PATCH 0/3] userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request

[Mike Rapoport]

Hi,

These patches add notification of madvise(MADV_REMOVE) event to
non-cooperative userfaultfd monitor.

The first pacth renames EVENT_MADVDONTNEED to EVENT_REMOVE along with
relevant functions and structures. Using _REMOVE instead of _MADVDONTNEED
describes the event semantics more clearly and I hope it's not too late for
such change in the ABI.

The patches are against current -mm tree.

Mike Rapoport (3):
  userfaultfd: non-cooperative: rename *EVENT_MADVDONTNEED to *EVENT_REMOVE
  userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request
  userfaultfd: non-cooperative: selftest: enable REMOVE event test for shmem

 fs/userfaultfd.c                         | 14 +++++++-------
 include/linux/userfaultfd_k.h            | 16 ++++++++--------
 include/uapi/linux/userfaultfd.h         |  8 ++++----
 mm/madvise.c                             |  3 ++-
 tools/testing/selftests/vm/userfaultfd.c | 22 ++++++++++------------
 5 files changed, 31 insertions(+), 32 deletions(-)

-- 
1.9.1

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 1/3] userfaultfd: non-cooperative: rename *EVENT_MADVDONTNEED to *EVENT_REMOVE

[Mike Rapoport]

The UFFD_EVENT_MADVDONTNEED purpose is to notify uffd monitor about removal
of certain range from address space tracked by userfaultfd.
Hence, UFFD_EVENT_REMOVE seems to better reflect the operation semantics.
Respectively, 'madv_dn' field of uffd_msg is renamed to 'remove' and the
madvise_userfault_dontneed callback is renamed to userfaultfd_remove.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 fs/userfaultfd.c                         | 14 +++++++-------
 include/linux/userfaultfd_k.h            | 16 ++++++++--------
 include/uapi/linux/userfaultfd.h         |  8 ++++----
 mm/madvise.c                             |  2 +-
 tools/testing/selftests/vm/userfaultfd.c | 16 ++++++++--------
 5 files changed, 28 insertions(+), 28 deletions(-)

diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c
index a817588..e9b4a50 100644
--- a/fs/userfaultfd.c
+++ b/fs/userfaultfd.c
@@ -681,16 +681,16 @@ void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *vm_ctx,
 	userfaultfd_event_wait_completion(ctx, &ewq);
 }
 
-void madvise_userfault_dontneed(struct vm_area_struct *vma,
-				struct vm_area_struct **prev,
-				unsigned long start, unsigned long end)
+void userfaultfd_remove(struct vm_area_struct *vma,
+			struct vm_area_struct **prev,
+			unsigned long start, unsigned long end)
 {
 	struct mm_struct *mm = vma->vm_mm;
 	struct userfaultfd_ctx *ctx;
 	struct userfaultfd_wait_queue ewq;
 
 	ctx = vma->vm_userfaultfd_ctx.ctx;
-	if (!ctx || !(ctx->features & UFFD_FEATURE_EVENT_MADVDONTNEED))
+	if (!ctx || !(ctx->features & UFFD_FEATURE_EVENT_REMOVE))
 		return;
 
 	userfaultfd_ctx_get(ctx);
@@ -700,9 +700,9 @@ void madvise_userfault_dontneed(struct vm_area_struct *vma,
 
 	msg_init(&ewq.msg);
 
-	ewq.msg.event = UFFD_EVENT_MADVDONTNEED;
-	ewq.msg.arg.madv_dn.start = start;
-	ewq.msg.arg.madv_dn.end = end;
+	ewq.msg.event = UFFD_EVENT_REMOVE;
+	ewq.msg.arg.remove.start = start;
+	ewq.msg.arg.remove.end = end;
 
 	userfaultfd_event_wait_completion(ctx, &ewq);
 
diff --git a/include/linux/userfaultfd_k.h b/include/linux/userfaultfd_k.h
index f431861..2521542 100644
--- a/include/linux/userfaultfd_k.h
+++ b/include/linux/userfaultfd_k.h
@@ -61,10 +61,10 @@ extern void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *,
 					unsigned long from, unsigned long to,
 					unsigned long len);
 
-extern void madvise_userfault_dontneed(struct vm_area_struct *vma,
-				       struct vm_area_struct **prev,
-				       unsigned long start,
-				       unsigned long end);
+extern void userfaultfd_remove(struct vm_area_struct *vma,
+			       struct vm_area_struct **prev,
+			       unsigned long start,
+			       unsigned long end);
 
 #else /* CONFIG_USERFAULTFD */
 
@@ -112,10 +112,10 @@ static inline void mremap_userfaultfd_complete(struct vm_userfaultfd_ctx *ctx,
 {
 }
 
-static inline void madvise_userfault_dontneed(struct vm_area_struct *vma,
-					      struct vm_area_struct **prev,
-					      unsigned long start,
-					      unsigned long end)
+static inline void userfaultfd_remove(struct vm_area_struct *vma,
+				      struct vm_area_struct **prev,
+				      unsigned long start,
+				      unsigned long end)
 {
 }
 #endif /* CONFIG_USERFAULTFD */
diff --git a/include/uapi/linux/userfaultfd.h b/include/uapi/linux/userfaultfd.h
index 9ac4b68..b742c40 100644
--- a/include/uapi/linux/userfaultfd.h
+++ b/include/uapi/linux/userfaultfd.h
@@ -20,7 +20,7 @@
 #define UFFD_API ((__u64)0xAA)
 #define UFFD_API_FEATURES (UFFD_FEATURE_EVENT_FORK |		\
 			   UFFD_FEATURE_EVENT_REMAP |		\
-			   UFFD_FEATURE_EVENT_MADVDONTNEED |	\
+			   UFFD_FEATURE_EVENT_REMOVE |	\
 			   UFFD_FEATURE_MISSING_HUGETLBFS |	\
 			   UFFD_FEATURE_MISSING_SHMEM)
 #define UFFD_API_IOCTLS				\
@@ -92,7 +92,7 @@ struct uffd_msg {
 		struct {
 			__u64	start;
 			__u64	end;
-		} madv_dn;
+		} remove;
 
 		struct {
 			/* unused reserved fields */
@@ -109,7 +109,7 @@ struct uffd_msg {
 #define UFFD_EVENT_PAGEFAULT	0x12
 #define UFFD_EVENT_FORK		0x13
 #define UFFD_EVENT_REMAP	0x14
-#define UFFD_EVENT_MADVDONTNEED	0x15
+#define UFFD_EVENT_REMOVE	0x15
 
 /* flags for UFFD_EVENT_PAGEFAULT */
 #define UFFD_PAGEFAULT_FLAG_WRITE	(1<<0)	/* If this was a write fault */
@@ -155,7 +155,7 @@ struct uffdio_api {
 #define UFFD_FEATURE_PAGEFAULT_FLAG_WP		(1<<0)
 #define UFFD_FEATURE_EVENT_FORK			(1<<1)
 #define UFFD_FEATURE_EVENT_REMAP		(1<<2)
-#define UFFD_FEATURE_EVENT_MADVDONTNEED		(1<<3)
+#define UFFD_FEATURE_EVENT_REMOVE		(1<<3)
 #define UFFD_FEATURE_MISSING_HUGETLBFS		(1<<4)
 #define UFFD_FEATURE_MISSING_SHMEM		(1<<5)
 	__u64 features;
diff --git a/mm/madvise.c b/mm/madvise.c
index b530a49..ab5ef14 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -479,7 +479,7 @@ static long madvise_dontneed(struct vm_area_struct *vma,
 	if (!can_madv_dontneed_vma(vma))
 		return -EINVAL;
 
-	madvise_userfault_dontneed(vma, prev, start, end);
+	userfaultfd_remove(vma, prev, start, end);
 	zap_page_range(vma, start, end - start);
 	return 0;
 }
diff --git a/tools/testing/selftests/vm/userfaultfd.c b/tools/testing/selftests/vm/userfaultfd.c
index 5a840a6..9eb77df 100644
--- a/tools/testing/selftests/vm/userfaultfd.c
+++ b/tools/testing/selftests/vm/userfaultfd.c
@@ -398,12 +398,12 @@ static void *uffd_poll_thread(void *arg)
 			uffd = msg.arg.fork.ufd;
 			pollfd[0].fd = uffd;
 			break;
-		case UFFD_EVENT_MADVDONTNEED:
-			uffd_reg.range.start = msg.arg.madv_dn.start;
-			uffd_reg.range.len = msg.arg.madv_dn.end -
-				msg.arg.madv_dn.start;
+		case UFFD_EVENT_REMOVE:
+			uffd_reg.range.start = msg.arg.remove.start;
+			uffd_reg.range.len = msg.arg.remove.end -
+				msg.arg.remove.start;
 			if (ioctl(uffd, UFFDIO_UNREGISTER, &uffd_reg.range))
-				fprintf(stderr, "madv_dn failure\n"), exit(1);
+				fprintf(stderr, "remove failure\n"), exit(1);
 			break;
 		case UFFD_EVENT_REMAP:
 			area_dst = (char *)(unsigned long)msg.arg.remap.to;
@@ -570,7 +570,7 @@ static int userfaultfd_open(int features)
  * mremap, the entire monitored area is accessed in a single pass for
  * HUGETLB_TEST.
  * The release of the pages currently generates event only for
- * anonymous memory (UFFD_EVENT_MADVDONTNEED), hence it is not checked
+ * anonymous memory (UFFD_EVENT_REMOVE), hence it is not checked
  * for hugetlb and shmem.
  */
 static int faulting_process(void)
@@ -715,14 +715,14 @@ static int userfaultfd_events_test(void)
 	pid_t pid;
 	char c;
 
-	printf("testing events (fork, remap, madv_dn): ");
+	printf("testing events (fork, remap, remove): ");
 	fflush(stdout);
 
 	if (release_pages(area_dst))
 		return 1;
 
 	features = UFFD_FEATURE_EVENT_FORK | UFFD_FEATURE_EVENT_REMAP |
-		UFFD_FEATURE_EVENT_MADVDONTNEED;
+		UFFD_FEATURE_EVENT_REMOVE;
 	if (userfaultfd_open(features) < 0)
 		return 1;
 	fcntl(uffd, F_SETFL, uffd_flags | O_NONBLOCK);
-- 
1.9.1

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 1/3] userfaultfd: non-cooperative: rename *EVENT_MADVDONTNEED to *EVENT_REMOVE

[Hillf Danton]

On Thursday, January 19, 2017 4:23 PM Mike Rapoport wrote: 
> 
> The UFFD_EVENT_MADVDONTNEED purpose is to notify uffd monitor about removal
> of certain range from address space tracked by userfaultfd.
> Hence, UFFD_EVENT_REMOVE seems to better reflect the operation semantics.
> Respectively, 'madv_dn' field of uffd_msg is renamed to 'remove' and the
> madvise_userfault_dontneed callback is renamed to userfaultfd_remove.
> 
Looks that "no function change" is missing.

> Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
> ---
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 2/3] userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request

[Mike Rapoport]

When a page is removed from a shared mapping, the uffd reader should be
notified, so that it won't attempt to handle #PF events for the removed
pages.
We can reuse the UFFD_EVENT_REMOVE because from the uffd monitor point
of view, the semantices of madvise(MADV_DONTNEED) and madvise(MADV_REMOVE)
is exactly the same.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 mm/madvise.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/mm/madvise.c b/mm/madvise.c
index ab5ef14..0012071 100644
--- a/mm/madvise.c
+++ b/mm/madvise.c
@@ -520,6 +520,7 @@ static long madvise_remove(struct vm_area_struct *vma,
 	 * mmap_sem.
 	 */
 	get_file(f);
+	userfaultfd_remove(vma, prev, start, end);
 	up_read(&current->mm->mmap_sem);
 	error = vfs_fallocate(f,
 				FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
-- 
1.9.1

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 2/3] userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request

[Hillf Danton]

On Thursday, January 19, 2017 4:23 PM Mike Rapoport wrote: 
> 
> When a page is removed from a shared mapping, the uffd reader should be
> notified, so that it won't attempt to handle #PF events for the removed
> pages.
> We can reuse the UFFD_EVENT_REMOVE because from the uffd monitor point
> of view, the semantices of madvise(MADV_DONTNEED) and madvise(MADV_REMOVE)
> is exactly the same.
> 
> Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
> ---
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>

>  mm/madvise.c | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/mm/madvise.c b/mm/madvise.c
> index ab5ef14..0012071 100644
> --- a/mm/madvise.c
> +++ b/mm/madvise.c
> @@ -520,6 +520,7 @@ static long madvise_remove(struct vm_area_struct *vma,
>  	 * mmap_sem.
>  	 */
>  	get_file(f);
> +	userfaultfd_remove(vma, prev, start, end);
>  	up_read(&current->mm->mmap_sem);
>  	error = vfs_fallocate(f,
>  				FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
> --
> 1.9.1

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

[PATCH 3/3] userfaultfd: non-cooperative: selftest: enable REMOVE event test for shmem

[Mike Rapoport]

Now when madvise(MADV_REMOVE) notifies uffd reader, we should verify that
appliciation actually sees zeros at the removed range.

Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
---
 tools/testing/selftests/vm/userfaultfd.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/tools/testing/selftests/vm/userfaultfd.c b/tools/testing/selftests/vm/userfaultfd.c
index 9eb77df..e9449c8 100644
--- a/tools/testing/selftests/vm/userfaultfd.c
+++ b/tools/testing/selftests/vm/userfaultfd.c
@@ -569,9 +569,9 @@ static int userfaultfd_open(int features)
  * part is accessed after mremap. Since hugetlbfs does not support
  * mremap, the entire monitored area is accessed in a single pass for
  * HUGETLB_TEST.
- * The release of the pages currently generates event only for
+ * The release of the pages currently generates event for shmem and
  * anonymous memory (UFFD_EVENT_REMOVE), hence it is not checked
- * for hugetlb and shmem.
+ * for hugetlb.
  */
 static int faulting_process(void)
 {
@@ -610,7 +610,6 @@ static int faulting_process(void)
 		}
 	}
 
-#ifndef SHMEM_TEST
 	if (release_pages(area_dst))
 		return 1;
 
@@ -618,7 +617,6 @@ static int faulting_process(void)
 		if (my_bcmp(area_dst + nr * page_size, zeropage, page_size))
 			fprintf(stderr, "nr %lu is not zero\n", nr), exit(1);
 	}
-#endif /* SHMEM_TEST */
 
 #endif /* HUGETLB_TEST */
 
-- 
1.9.1

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 0/3] userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request

[Andrea Arcangeli]

On Thu, Jan 19, 2017 at 10:22:31AM +0200, Mike Rapoport wrote:
> Hi,
> 
> These patches add notification of madvise(MADV_REMOVE) event to
> non-cooperative userfaultfd monitor.
> 
> The first pacth renames EVENT_MADVDONTNEED to EVENT_REMOVE along with
> relevant functions and structures. Using _REMOVE instead of _MADVDONTNEED
> describes the event semantics more clearly and I hope it's not too late for
> such change in the ABI.
> 
> The patches are against current -mm tree.

Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>

Not sure if they should be folded inside -mm before they go upstream,
but if they're sent all along it probably doesn't make a difference,
it just adds one more handler.

In theory we could have differentiated MADV_REMOVE and MADV_DONTNEED
in two different events, but I couldn't see any downside in sharing
the same event for both and this looks more compact.

Thanks,
Andrea

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH 0/3] userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request

[Pavel Emelyanov]

On 01/19/2017 11:22 AM, Mike Rapoport wrote:
> Hi,
> 
> These patches add notification of madvise(MADV_REMOVE) event to
> non-cooperative userfaultfd monitor.
> 
> The first pacth renames EVENT_MADVDONTNEED to EVENT_REMOVE along with
> relevant functions and structures. Using _REMOVE instead of _MADVDONTNEED
> describes the event semantics more clearly and I hope it's not too late for
> such change in the ABI.
> 
> The patches are against current -mm tree.
> 
> Mike Rapoport (3):
>   userfaultfd: non-cooperative: rename *EVENT_MADVDONTNEED to *EVENT_REMOVE
>   userfaultfd: non-cooperative: add madvise() event for MADV_REMOVE request
>   userfaultfd: non-cooperative: selftest: enable REMOVE event test for shmem

Acked-by: Pavel Emelyanov <xemul@virtuozzo.com>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>