Skip to contentSkip to conversation

vfs: renumber FMODE_NONOTIFY and add to uniqueness check

PATCH In review David Drysdale 4 participants 6 messages Nov 24, 2014 – Jan 8, 2015

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

v2 cover vfs: renumber FMODE_NONOTIFY and add to uniqueness check

Changes since v1:
- reword commit to indicate origin of clash [Heinrich Schuchardt]
- remove comment fix [Heinrich Schuchardt]
- acked-by [David S. Miller]

Against v3.18-rc5.

Tested by running fatrace on x86_64, and by cross-compiling against sparc, arm, arm64, m68k, mips, powerpc, alpha, parisc, blackfin.

David Drysdale (1):
vfs: renumber FMODE_NONOTIFY and add to uniqueness check

 fs/fcntl.c                       | 5 +++--
 include/linux/fs.h               | 2 +-
 include/uapi/asm-generic/fcntl.h | 2 +-
 3 files changed, 5 insertions(+), 4 deletions(-)
Signature · 1 line
2.1.0.rc2.206.gedb03e5
#1 lore raw

PATCH v2 1/1 vfs: renumber FMODE_NONOTIFY and add to uniqueness check

Fix clashing values for O_PATH and FMODE_NONOTIFY on sparc. The clashing O_PATH value was added in 5229645bdc35f1cc43eb ("vfs: add nonconflicting values for O_PATH") but this can't be changed as it is user-visible.

FMODE_NONOTIFY is only used internally in the kernel, but it is in the same numbering space as the other O_* flags, as indicated by the comment at the top of include/uapi/asm-generic/fcntl.h (and its use in fs/notify/fanotify/fanotify_user.c). So renumber it to avoid the clash.

All of this has happened before (12ed2e36c98aec6c4155, "fanotify:
FMODE_NONOTIFY and __O_SYNC in sparc conflict"), and all of this will happen again -- so update the uniqueness check in fcntl_init() to include __FMODE_NONOTIFY.

---
 fs/fcntl.c                       | 5 +++--
 include/linux/fs.h               | 2 +-
 include/uapi/asm-generic/fcntl.h | 2 +-
 3 files changed, 5 insertions(+), 4 deletions(-)
fs/fcntl.c +3 -2
index 99d440a4a6ba..ee85cd4e136a 100644
@@ -740,14 +740,15 @@ static int __init fcntl_init(void)
 	 * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY
 	 * is defined as O_NONBLOCK on some platforms and not on others.
 	 */
-	BUILD_BUG_ON(20 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
+	BUILD_BUG_ON(21 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
 		O_RDONLY	| O_WRONLY	| O_RDWR	|
 		O_CREAT		| O_EXCL	| O_NOCTTY	|
 		O_TRUNC		| O_APPEND	| /* O_NONBLOCK	| */
 		__O_SYNC	| O_DSYNC	| FASYNC	|
 		O_DIRECT	| O_LARGEFILE	| O_DIRECTORY	|
 		O_NOFOLLOW	| O_NOATIME	| O_CLOEXEC	|
-		__FMODE_EXEC	| O_PATH	| __O_TMPFILE
+		__FMODE_EXEC	| O_PATH	| __O_TMPFILE	|
+		__FMODE_NONOTIFY
 		));
 
 	fasync_cache = kmem_cache_create("fasync_cache",
include/linux/fs.h +1 -1
index 9ab779e8a63c..fc1c2488adbc 100644
@@ -134,7 +134,7 @@ typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
 #define FMODE_CAN_WRITE         ((__force fmode_t)0x40000)
 
 /* File was opened by fanotify and shouldn't generate fanotify events */
-#define FMODE_NONOTIFY		((__force fmode_t)0x1000000)
+#define FMODE_NONOTIFY		((__force fmode_t)0x4000000)
 
 /*
  * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
include/uapi/asm-generic/fcntl.h +1 -1
index 7543b3e51331..e063effe0cc1 100644
@@ -5,7 +5,7 @@
 
 /*
  * FMODE_EXEC is 0x20
- * FMODE_NONOTIFY is 0x1000000
+ * FMODE_NONOTIFY is 0x4000000
  * These cannot be used by userspace O_* until internal and external open
  * flags are split.
  * -Eric Paris
Signature · 1 line
2.1.0.rc2.206.gedb03e5
reviewer Acked-by on v2 1/1
#2 lore raw
Quoted 11 lines · David Drysdale

Fix clashing values for O_PATH and FMODE_NONOTIFY on sparc. The clashing O_PATH value was added in 5229645bdc35f1cc43eb ("vfs: add nonconflicting values for O_PATH") but this can't be changed as it is user-visible.

FMODE_NONOTIFY is only used internally in the kernel, but it is in the same numbering space as the other O_* flags, as indicated by the comment at the top of include/uapi/asm-generic/fcntl.h (and its use in fs/notify/fanotify/fanotify_user.c). So renumber it to avoid the clash.

Quoted 7 lines

All of this has happened before (12ed2e36c98aec6c4155, "fanotify:
FMODE_NONOTIFY and __O_SYNC in sparc conflict"), and all of this will happen again -- so update the uniqueness check in fcntl_init() to include __FMODE_NONOTIFY.

Looks good. You can add:

Honza

Quoted 57 lines
---
 fs/fcntl.c                       | 5 +++--
 include/linux/fs.h               | 2 +-
 include/uapi/asm-generic/fcntl.h | 2 +-
 3 files changed, 5 insertions(+), 4 deletions(-)
fs/fcntl.c +3 -2
index 99d440a4a6ba..ee85cd4e136a 100644
@@ -740,14 +740,15 @@ static int __init fcntl_init(void)
 	 * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY
 	 * is defined as O_NONBLOCK on some platforms and not on others.
 	 */
-	BUILD_BUG_ON(20 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
+	BUILD_BUG_ON(21 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
 		O_RDONLY	| O_WRONLY	| O_RDWR	|
 		O_CREAT		| O_EXCL	| O_NOCTTY	|
 		O_TRUNC		| O_APPEND	| /* O_NONBLOCK	| */
 		__O_SYNC	| O_DSYNC	| FASYNC	|
 		O_DIRECT	| O_LARGEFILE	| O_DIRECTORY	|
 		O_NOFOLLOW	| O_NOATIME	| O_CLOEXEC	|
-		__FMODE_EXEC	| O_PATH	| __O_TMPFILE
+		__FMODE_EXEC	| O_PATH	| __O_TMPFILE	|
+		__FMODE_NONOTIFY
 		));
 
 	fasync_cache = kmem_cache_create("fasync_cache",
include/linux/fs.h +1 -1
index 9ab779e8a63c..fc1c2488adbc 100644
@@ -134,7 +134,7 @@ typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
 #define FMODE_CAN_WRITE         ((__force fmode_t)0x40000)
 
 /* File was opened by fanotify and shouldn't generate fanotify events */
-#define FMODE_NONOTIFY		((__force fmode_t)0x1000000)
+#define FMODE_NONOTIFY		((__force fmode_t)0x4000000)
 
 /*
  * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
include/uapi/asm-generic/fcntl.h +1 -1
index 7543b3e51331..e063effe0cc1 100644
@@ -5,7 +5,7 @@
 
 /*
  * FMODE_EXEC is 0x20
- * FMODE_NONOTIFY is 0x1000000
+ * FMODE_NONOTIFY is 0x4000000
  * These cannot be used by userspace O_* until internal and external open
  * flags are split.
  * -Eric Paris
Signature · 1 line
2.1.0.rc2.206.gedb03e5
Signature · 2 lines
Jan Kara <[email protected]>
SUSE Labs, CR
#3 lore raw

Hello Andrew,

could you, please, consider the patch proposed in
https://lkml.org/lkml/2014/11/24/196
for inclusion into the mm-tree.

This 2nd version of the patch considers the review comments in the thread starting at https://lkml.org/lkml/2014/11/21/141.

Best regards

Heinrich Schuchardt

Quoted 81 lines · Jan Kara
Quoted 11 lines · David Drysdale

Fix clashing values for O_PATH and FMODE_NONOTIFY on sparc. The clashing O_PATH value was added in 5229645bdc35f1cc43eb ("vfs: add nonconflicting values for O_PATH") but this can't be changed as it is user-visible.

FMODE_NONOTIFY is only used internally in the kernel, but it is in the same numbering space as the other O_* flags, as indicated by the comment at the top of include/uapi/asm-generic/fcntl.h (and its use in fs/notify/fanotify/fanotify_user.c). So renumber it to avoid the clash.

Quoted 7 lines

All of this has happened before (12ed2e36c98aec6c4155, "fanotify:
FMODE_NONOTIFY and __O_SYNC in sparc conflict"), and all of this will happen again -- so update the uniqueness check in fcntl_init() to include __FMODE_NONOTIFY.

Looks good. You can add:

Honza

Quoted 57 lines
---
  fs/fcntl.c                       | 5 +++--
  include/linux/fs.h               | 2 +-
  include/uapi/asm-generic/fcntl.h | 2 +-
  3 files changed, 5 insertions(+), 4 deletions(-)
fs/fcntl.c +3 -2
index 99d440a4a6ba..ee85cd4e136a 100644
@@ -740,14 +740,15 @@ static int __init fcntl_init(void)
  	 * Exceptions: O_NONBLOCK is a two bit define on parisc; O_NDELAY
  	 * is defined as O_NONBLOCK on some platforms and not on others.
  	 */
-	BUILD_BUG_ON(20 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
+	BUILD_BUG_ON(21 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32(
  		O_RDONLY	| O_WRONLY	| O_RDWR	|
  		O_CREAT		| O_EXCL	| O_NOCTTY	|
  		O_TRUNC		| O_APPEND	| /* O_NONBLOCK	| */
  		__O_SYNC	| O_DSYNC	| FASYNC	|
  		O_DIRECT	| O_LARGEFILE	| O_DIRECTORY	|
  		O_NOFOLLOW	| O_NOATIME	| O_CLOEXEC	|
-		__FMODE_EXEC	| O_PATH	| __O_TMPFILE
+		__FMODE_EXEC	| O_PATH	| __O_TMPFILE	|
+		__FMODE_NONOTIFY
  		));
 
  	fasync_cache = kmem_cache_create("fasync_cache",
include/linux/fs.h +1 -1
index 9ab779e8a63c..fc1c2488adbc 100644
@@ -134,7 +134,7 @@ typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
  #define FMODE_CAN_WRITE         ((__force fmode_t)0x40000)
 
  /* File was opened by fanotify and shouldn't generate fanotify events */
-#define FMODE_NONOTIFY		((__force fmode_t)0x1000000)
+#define FMODE_NONOTIFY		((__force fmode_t)0x4000000)
 
  /*
   * Flag for rw_copy_check_uvector and compat_rw_copy_check_uvector
include/uapi/asm-generic/fcntl.h +1 -2
index 7543b3e51331..e063effe0cc1 100644
@@ -5,7 +5,7 @@
 
  /*
   * FMODE_EXEC is 0x20
- * FMODE_NONOTIFY is 0x1000000
+ * FMODE_NONOTIFY is 0x4000000
   * These cannot be used by userspace O_* until internal and external open
   * flags are split.
   * -Eric Paris
--

2.1.0.rc2.206.gedb03e5

#4 lore raw
In reply to David Drysdale
Quoted 11 lines · On Mon, 24 Nov 2014 11:23:19 +0000 David Drysdale <[email protected]> wrote:

Fix clashing values for O_PATH and FMODE_NONOTIFY on sparc. The clashing O_PATH value was added in 5229645bdc35f1cc43eb ("vfs: add nonconflicting values for O_PATH") but this can't be changed as it is user-visible.

FMODE_NONOTIFY is only used internally in the kernel, but it is in the same numbering space as the other O_* flags, as indicated by the comment at the top of include/uapi/asm-generic/fcntl.h (and its use in fs/notify/fanotify/fanotify_user.c). So renumber it to avoid the clash.

Quoted 4 lines

All of this has happened before (12ed2e36c98aec6c4155, "fanotify:
FMODE_NONOTIFY and __O_SYNC in sparc conflict"), and all of this will happen again -- so update the uniqueness check in fcntl_init() to include __FMODE_NONOTIFY.

What are the user-visible runtime effects of the bug?

Please always include this info when fixing bugs, so people can work out which kernel version(s) need the fix.

#5 lore raw
Quoted 18 lines · Andrew Morton
Quoted 11 lines · On Mon, 24 Nov 2014 11:23:19 +0000 David Drysdale <[email protected]> wrote:

Fix clashing values for O_PATH and FMODE_NONOTIFY on sparc. The clashing O_PATH value was added in 5229645bdc35f1cc43eb ("vfs: add nonconflicting values for O_PATH") but this can't be changed as it is user-visible.

FMODE_NONOTIFY is only used internally in the kernel, but it is in the same numbering space as the other O_* flags, as indicated by the comment at the top of include/uapi/asm-generic/fcntl.h (and its use in fs/notify/fanotify/fanotify_user.c). So renumber it to avoid the clash.

Quoted 4 lines

All of this has happened before (12ed2e36c98aec6c4155, "fanotify:
FMODE_NONOTIFY and __O_SYNC in sparc conflict"), and all of this will happen again -- so update the uniqueness check in fcntl_init() to include __FMODE_NONOTIFY.

Quoted 4 lines

What are the user-visible runtime effects of the bug?

Please always include this info when fixing bugs, so people can work out which kernel version(s) need the fix.

I don't know for sure what the user-visible effects are -- I noticed the problem from the source and I don't have a sparc system (the only affected architecture) to test with. But I'd guess that...

The likely user-visible effect (on sparc) of the clash would be that:
- Filesystem activity on files opened with O_PATH does not get reported via the fanotify(7) API.
- The file descriptors for accessed objects reported via the fanotify(7) API will effectively have O_PATH set, and so will have the limitations described for O_PATH in the open(2) manpage.

However, we'd need some intersection of sparc/fanotify folk to verify those guesses -- Eric / David, any thoughts?

Thanks,
David