vfs: O_* bit numbers uniqueness check
PATCH 2/5 vfs: O_* bit numbers uniqueness check
From: Wu Fengguang <[email protected]>
The O_* bit numbers are defined in 20+ arch/*, and can silently overlap. Add a compile time check to ensure the uniqueness as suggested by David Miller.
---
fs/fcntl.c | 15 +++++++++++++-- include/asm-generic/fcntl.h | 4 ++++ 2 files changed, 17 insertions(+), 2 deletions(-)
fs/fcntl.c renamed from fs/fcntl.c~vfs-o_-bit-numbers-uniqueness-check +13 -2
@@ -767,11 +767,22 @@ void kill_fasync(struct fasync_struct ** }EXPORT_SYMBOL (kill_fasync); -static int __init fasync_init (void ) +static int __init fcntl_init (void ) { +/* please add new bits here to ensure allocation uniqueness */ +BUILD_BUG_ON (19 -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 + )); + fasync_cache =kmem_cache_create ("fasync_cache" ,sizeof (struct fasync_struct ),0 ,SLAB_PANIC ,NULL );return 0 ; } -module_init (fasync_init) +module_init (fcntl_init)
include/asm-generic/fcntl.h renamed from include/asm-generic/fcntl.h~vfs-o_-bit-numbers-uniqueness-check +4 -0
@@ -11,6 +11,10 @@* -Eric Paris */ +/* +* When introducing new O_* bits, please check its uniqueness in fcntl_init(). +*/ +#define O_ACCMODE 00000003#define O_RDONLY 00000000#define O_WRONLY 00000001
_
Quoted 5 lines · On Tue, 2010-07-20 at 15:29 -0700, [email protected] wrote:
From: Wu Fengguang <[email protected]>
The O_* bit numbers are defined in 20+
arch/*, and can silently overlap. Add a compile time check to ensure the uniqueness as suggested by David Miller.
Can we get this reverted or fixed? It's causing the parisc compiles to fail. The reason is O_NONBLOCK on parisc has a dual value:
#define O_NONBLOCK 000200004 /* HPUX has separate NDELAY & NONBLOCK */
The fix would be to take O_NONBLOCK out.
James
---
fs/fcntl.c +3 -2
@@ -770,10 +770,11 @@ EXPORT_SYMBOL(kill_fasync); static int __init fcntl_init(void) { /* please add new bits here to ensure allocation uniqueness */ - BUILD_BUG_ON(19 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32( + BUILD_BUG_ON(18 - 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 | + /* remove O_NONBLOCK it's a two bit define on parisc */ + O_TRUNC | O_APPEND | /* O_NONBLOCK | */ __O_SYNC | O_DSYNC | FASYNC | O_DIRECT | O_LARGEFILE | O_DIRECTORY | O_NOFOLLOW | O_NOATIME | O_CLOEXEC |
Quoted 7 lines · James Bottomley
Quoted 5 lines · On Tue, 2010-07-20 at 15:29 -0700, [email protected] wrote:
From: Wu Fengguang <[email protected]>
The O_* bit numbers are defined in 20+
arch/*, and can silently overlap. Add a compile time check to ensure the uniqueness as suggested by David Miller.
Quoted 6 lines
Can we get this reverted or fixed? It's causing the parisc compiles to fail. The reason is
O_NONBLOCKon parisc has a dual value:#define
O_NONBLOCK000200004 /* HPUX has separate NDELAY & NONBLOCK */The fix would be to take
O_NONBLOCKout.
A more thoroughly checking fix would be
BUILD_BUG_ON(18 - 1 /* For O_RDONLY being 0 */
+ HWEIGHT32(O_NONBLOCK) /* Because it's 2 bits on parisc */
!= HWEIGHT32( .... all the bits .... ));
Am I allowed to Sign-off handwavy pseudocode? ;-)
-- Jamie
Quoted 13 lines · Jamie Lokier
Quoted 12 lines · James Bottomley
Quoted 5 lines · On Tue, 2010-07-20 at 15:29 -0700, [email protected] wrote:
From: Wu Fengguang <[email protected]>
The O_* bit numbers are defined in 20+
arch/*, and can silently overlap. Add a compile time check to ensure the uniqueness as suggested by David Miller.Can we get this reverted or fixed? It's causing the parisc compiles to fail. The reason is
O_NONBLOCKon parisc has a dual value:#define
O_NONBLOCK000200004 /* HPUX has separate NDELAY & NONBLOCK */
Quoted 7 lines
Quoted 1 line
The fix would be to take
O_NONBLOCKout.A more thoroughly checking fix would be
BUILD_BUG_ON(18 - 1 /* For
O_RDONLYbeing 0 */
+HWEIGHT32(O_NONBLOCK)/* Because it's 2 bits on parisc */
!= HWEIGHT32( .... all the bits .... ));
Well, it works, but is it wise? There are several missing flags in this
expression: O_NDELAY being the most noticeable one. It's defined to be the sames as O_NONBLOCK on some platforms and not on others. Our O_NONBLOCK problem is essentially the same class.
Quoted 1 line
Am I allowed to Sign-off handwavy pseudocode? ;-)
Nope ... you'd have to code it as a real patch ...
James
PATCH 1/1 vfs: take O_NONBLOCK out of the O_* uniqueness test
I tend to agree with James' approach. How about this patch? Compile tested on x86.
===
vfs: take O_NONBLOCK out of the O_* uniqueness test
From: James Bottomley <[email protected]>
O_NONBLOCK on parisc has a dual value:
#define O_NONBLOCK 000200004 /* HPUX has separate NDELAY & NONBLOCK */
It is caught by the O_* bits uniqueness check and leads to a parisc compile error. The fix would be to take O_NONBLOCK out.
--- fs/fcntl.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-)
fs/fcntl.c +7 -3
@@ -769,11 +769,15 @@ EXPORT_SYMBOL(kill_fasync); static int __init fcntl_init(void) { - /* please add new bits here to ensure allocation uniqueness */ - BUILD_BUG_ON(19 - 1 /* for O_RDONLY being 0 */ != HWEIGHT32( + /* + * Please add new bits here to ensure allocation uniqueness. + * 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(18 - 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_TRUNC | O_APPEND | /* O_NONBLOCK | */ __O_SYNC | O_DSYNC | FASYNC | O_DIRECT | O_LARGEFILE | O_DIRECTORY | O_NOFOLLOW | O_NOATIME | O_CLOEXEC |