Skip to contentSkip to conversation

fs: O_* bit numbers uniqueness check

PATCH Superseded Wu Fengguang 6 participants 13 messages Jan 6, 2010 – Jan 9, 2010

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 fs: O_* bit numbers uniqueness check

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>
---
 fs/fcntl.c                  |   14 ++++++++++++--
 include/asm-generic/fcntl.h |    2 ++
 2 files changed, 14 insertions(+), 2 deletions(-)
fs/fcntl.c +12 -2
@@ -709,11 +709,21 @@ 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 */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+
 	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 +2 -0
@@ -4,6 +4,8 @@
 #include <linux/types.h>
 
 /*
+ * When introducing new O_* bits, please check its uniqueness in fcntl_init().
+ *
  * FMODE_EXEC is 0x20
  * FMODE_NONOTIFY is 0x1000000
  * These cannot be used by userspace O_* until internal and external open
#1 lore raw

fs/fcntl.c void kill_fasync(struct fasync_struct ** · line 716

Quoted 4 lines
fs/fcntl.c +3 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|

I wonder if there's a way to make this BUILD_BUG_ON(), so the problem is caught at compile time (a change like adding an O_ flag is often compile tested only for obscure archs). One could create a compile-time macro-ized version of hweight32() and use that, I guess, although I'm not sure it's worth the ugliness.

Failing that maybe this should be WARN_ON()? I'd be annoyed if my arch wouldn't boot because some strange new O_ flag happened to collide.

- R.

#7 lore raw
In reply to Wu Fengguang
Quoted 12 lines · Wu Fengguang

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>

---
{

fs/fcntl.c void kill_fasync(struct fasync_struct ** · line 723

Quoted 10 lines
fs/fcntl.c +10 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+

I cannot test it, but given O_RDONLY is 0, are you sure 20 bits are actually set ? --
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw

PATCH v2 1/1 fs: O_* bit numbers uniqueness check

In reply to Roland Dreier
Quoted 10 lines · Roland Dreier

Review

Quoted 4 lines
(quoted diff) +3 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|

I wonder if there's a way to make this BUILD_BUG_ON(), so the problem is caught at compile time (a change like adding an O_ flag is often compile tested only for obscure archs). One could create a compile-time macro-ized version of hweight32() and use that, I guess, although I'm not sure it's worth the ugliness.

Yeah, I chose the boot time check because of hweight32()..

Quoted 2 lines

Failing that maybe this should be WARN_ON()? I'd be annoyed if my arch wouldn't boot because some strange new O_ flag happened to collide.

Good suggestion! Here is the updated patch.

Andrew, this patch relies on the newly introduced O_RANDOM and Eric's FMODE_NONOTIFY.

Thanks,
Fengguang
---
fs: O_* bit numbers uniqueness check

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

v2: change to WARN_ON() as suggested by Roland

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>
CC Roland Dreier <[email protected]>
---
 fs/fcntl.c                  |   14 ++++++++++++--
 include/asm-generic/fcntl.h |    2 ++
 2 files changed, 14 insertions(+), 2 deletions(-)
fs/fcntl.c +12 -2
@@ -709,11 +709,21 @@ 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 */
+	WARN_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+
 	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 +2 -0
@@ -4,6 +4,8 @@
 #include <linux/types.h>
 
 /*
+ * When introducing new O_* bits, please check its uniqueness in fcntl_init().
+ *
  * FMODE_EXEC is 0x20
  * FMODE_NONOTIFY is 0x1000000
  * These cannot be used by userspace O_* until internal and external open
#8 lore raw
In reply to Eric Dumazet
Quoted 17 lines · Eric Dumazet
Quoted 16 lines · Wu Fengguang

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>

---
{

fs/fcntl.c +4 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
Quoted 8 lines

fs/fcntl.c void kill_fasync(struct fasync_struct ** · line 723

Quoted 6 lines
fs/fcntl.c +6 -0
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+

I cannot test it, but given O_RDONLY is 0, are you sure 20 bits are actually set ?

Yes, I tested it. The tricky one is O_SYNC, which actually has two bits..

Thanks,
Fengguang

#3 lore raw
In reply to Wu Fengguang
Quoted 2 lines

Yeah, I chose the boot time check because of hweight32()..

One could do something like

#define HWEIGHT32(x) (!!((x) & (1u <<  0)) +
		      !!((x) & (1u <<  1)) +
//...
		      !!((x) & (1u << 31)));

that would probably work with BUILD_BUG_ON().

But as I said maybe it's too ugly.

- R.

#4 lore raw
Quoted 5 lines · Roland Dreier
Quoted 2 lines

Yeah, I chose the boot time check because of hweight32()..

One could do something like

Quoted 8 lines
#define HWEIGHT32(x) (!!((x) & (1u <<  0)) +
		      !!((x) & (1u <<  1)) +
//...
		      !!((x) & (1u << 31)));

that would probably work with BUILD_BUG_ON().

But as I said maybe it's too ugly.

Yes, it'll feel like this. Then you'll have to fix the possible error in order to compile ;)

---
fs: O_* bit numbers uniqueness check

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

v3: change to BUILD_BUG_ON() as suggested by Roland

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>
CC Roland Dreier <[email protected]>
---
 fs/fcntl.c                  |   14 ++++++++++++--
 include/asm-generic/fcntl.h |    2 ++
 include/linux/bitops.h      |   33 +++++++++++++++++++++++++++++++++
 3 files changed, 47 insertions(+), 2 deletions(-)
fs/fcntl.c +12 -2
@@ -709,11 +709,21 @@ 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(20 != HWEIGHT32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+
 	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 +2 -0
@@ -4,6 +4,8 @@
 #include <linux/types.h>
 
 /*
+ * When introducing new O_* bits, please check its uniqueness in fcntl_init().
+ *
  * FMODE_EXEC is 0x20
  * FMODE_NONOTIFY is 0x1000000
  * These cannot be used by userspace O_* until internal and external open
include/linux/bitops.h +33 -0
@@ -8,6 +8,39 @@
 #define BIT_WORD(nr)		((nr) / BITS_PER_LONG)
 #define BITS_PER_BYTE		8
 #define BITS_TO_LONGS(nr)	DIV_ROUND_UP(nr, BITS_PER_BYTE * sizeof(long))
+
+#define HWEIGHT32(x) (!!((x) & (1u <<  0)) +	\
+                      !!((x) & (1u <<  1)) +	\
+                      !!((x) & (1u <<  2)) +	\
+                      !!((x) & (1u <<  3)) +	\
+                      !!((x) & (1u <<  4)) +	\
+                      !!((x) & (1u <<  5)) +	\
+                      !!((x) & (1u <<  6)) +	\
+                      !!((x) & (1u <<  7)) +	\
+                      !!((x) & (1u <<  8)) +	\
+                      !!((x) & (1u <<  9)) +	\
+                      !!((x) & (1u << 10)) +	\
+                      !!((x) & (1u << 11)) +	\
+                      !!((x) & (1u << 12)) +	\
+                      !!((x) & (1u << 13)) +	\
+                      !!((x) & (1u << 14)) +	\
+                      !!((x) & (1u << 15)) +	\
+                      !!((x) & (1u << 16)) +	\
+                      !!((x) & (1u << 17)) +	\
+                      !!((x) & (1u << 18)) +	\
+                      !!((x) & (1u << 19)) +	\
+                      !!((x) & (1u << 20)) +	\
+                      !!((x) & (1u << 21)) +	\
+                      !!((x) & (1u << 22)) +	\
+                      !!((x) & (1u << 23)) +	\
+                      !!((x) & (1u << 24)) +	\
+                      !!((x) & (1u << 25)) +	\
+                      !!((x) & (1u << 26)) +	\
+                      !!((x) & (1u << 27)) +	\
+                      !!((x) & (1u << 28)) +	\
+                      !!((x) & (1u << 29)) +	\
+                      !!((x) & (1u << 30)) +	\
+                      !!((x) & (1u << 31)))
 #endif
 
 /*
#9 lore raw
In reply to Wu Fengguang
Quoted 27 lines · Wu Fengguang
Quoted 17 lines · Eric Dumazet
Quoted 16 lines · Wu Fengguang

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>

---
{

fs/fcntl.c +4 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
Quoted 8 lines

fs/fcntl.c void kill_fasync(struct fasync_struct ** · line 723

Quoted 6 lines
fs/fcntl.c +6 -0
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+

I cannot test it, but given O_RDONLY is 0, are you sure 20 bits are actually set ?

Quoted 1 line

Yes, I tested it. The tricky one is O_SYNC, which actually has two bits..

What if a new architecture wants to use a single bit value (since it does not need backwards compatibility)?

Andreas.

Signature · 3 lines
Andreas Schwab, [email protected]
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."
#11 lore raw
In reply to Wu Fengguang
Quoted 3 lines · Wu Fengguang

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

The right fix is to stop overlyaing the FMODE_ flags into the O_ namespace. Al has been working towards this.

#5 lore raw
In reply to Roland Dreier
Quoted 10 lines · Roland Dreier

fs/fcntl.c void kill_fasync(struct fasync_struct ** · line 716

Quoted 4 lines
fs/fcntl.c +3 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|

I wonder if there's a way to make this BUILD_BUG_ON(), so the problem is caught at compile time (a change like adding an O_ flag is often compile tested only for obscure archs). One could create a compile-time macro-ized version of hweight32() and use that, I guess, although I'm not sure it's worth the ugliness.

Not ugly at all:

#define hweight32(x) __builtin_popcount(x)

Checked GCC 3.4.3 / 4.1 / 4.4: It expands as a compile-time constant if the argument is a compile-time constant, so can be used in BUILD_BUG_ON() and even for array sizes etc.

If GCC's __builtin_popcount() isn't good enough for non-constant values, the macro would be more involved:

#define hweight32(x) (__builtin_constant_p(x) ? __builtin_popcount(x) \
                      : hweight32_nonconstant(x))

-- Jamie

#10 lore raw
In reply to Andreas Schwab
Quoted 29 lines · Andreas Schwab
Quoted 27 lines · Wu Fengguang
Quoted 25 lines · Eric Dumazet
Quoted 12 lines · Wu Fengguang

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>

---
{

fs/fcntl.c void kill_fasync(struct fasync_struct ** · line 723

Quoted 10 lines
fs/fcntl.c +10 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|
+		O_CREAT		| O_EXCL	| O_NOCTTY	|
+		O_TRUNC		| O_APPEND	| O_NONBLOCK	|
+		O_SYNC		| FASYNC	| O_DIRECT	|
+		O_LARGEFILE	| O_DIRECTORY	| O_NOFOLLOW	|
+		O_NOATIME	| O_CLOEXEC	| O_RANDOM	|
+		FMODE_EXEC	| FMODE_NONOTIFY));
+

I cannot test it, but given O_RDONLY is 0, are you sure 20 bits are actually set ?

Quoted 4 lines
Quoted 1 line

Yes, I tested it. The tricky one is O_SYNC, which actually has two bits..

What if a new architecture wants to use a single bit value (since it does not need backwards compatibility)?

You mean to test __O_SYNC | O_DSYNC instead of O_SYNC?
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#6 lore raw
In reply to Jamie Lokier
Quoted 20 lines · Jamie Lokier
Quoted 10 lines · Roland Dreier

Review

Quoted 4 lines
(quoted diff) +3 -0
+	/* please add new bits here to ensure allocation uniqueness */
+	BUG_ON(20 != hweight32(
+		O_RDONLY	| O_WRONLY	| O_RDWR	|

I wonder if there's a way to make this BUILD_BUG_ON(), so the problem is caught at compile time (a change like adding an O_ flag is often compile tested only for obscure archs). One could create a compile-time macro-ized version of hweight32() and use that, I guess, although I'm not sure it's worth the ugliness.

Not ugly at all:

#define hweight32(x) __builtin_popcount(x)

Checked GCC 3.4.3 / 4.1 / 4.4: It expands as a compile-time constant if the argument is a compile-time constant, so can be used in BUILD_BUG_ON() and even for array sizes etc.

Quoted 5 lines

If GCC's __builtin_popcount() isn't good enough for non-constant values, the macro would be more involved:

#define hweight32(x) (__builtin_constant_p(x) ? __builtin_popcount(x) \
                      : hweight32_nonconstant(x))

Jamie, Thanks for the hint! Would you sign this patch?

---
bitops: compile time optimization for hweight_long(CONSTANT)

This allows use of hweight_long() in BUILD_BUG_ON(). Suggested by Jamie.

CC Jamie Lokier <[email protected]>
CC Roland Dreier <[email protected]>
---
 include/linux/bitops.h |   12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
include/linux/bitops.h +8 -4
@@ -40,10 +40,14 @@ static __inline__ int get_count_order(un
 	return order;
 }
 
-static inline unsigned long hweight_long(unsigned long w)
-{
-	return sizeof(w) == 4 ? hweight32(w) : hweight64(w);
-}
+#define hweight_long(x)				\
+(						\
+	__builtin_constant_p(x) ?		\
+		__builtin_popcountl(x) :	\
+		(sizeof(x) <= 4 ?		\
+			 hweight32(x) :		\
+			 hweight64(x))		\
+)
 
 /**
  * rol32 - rotate a 32-bit value left
#12 lore raw
In reply to Christoph Hellwig
Quoted 7 lines · Christoph Hellwig
Quoted 3 lines · Wu Fengguang

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

The right fix is to stop overlyaing the FMODE_ flags into the O_ namespace. Al has been working towards this.

I thinks they are orthogonal problems -- this patch watches for the possible conflicting O_* in so many archs.

Anyway here is the new version without the FMODE_ and O_RANDOM.

Thanks,
Fengguang
---
fs: O_* bit numbers uniqueness check

The O_* bit numbers are defined in 20+ arch/*, and hence can silently overlap. Add a boot time check to ensure the uniqueness as suggested by David Miller.

v4: use the nice hweight_long() as suggested by Jamie
split O_SYNC to (__O_SYNC | O_DSYNC) as suggested by Andreas take away FMODE_* and O_RANDOM
v3: change to BUILD_BUG_ON() as suggested by Roland

CC David Miller <[email protected]>
CC Stephen Rothwell <[email protected]>
CC Al Viro <[email protected]>
CC Christoph Hellwig <[email protected]>
CC Eric Paris <[email protected]>
CC Roland Dreier <[email protected]>
CC Jamie Lokier <[email protected]>
CC Andreas Schwab <[email protected]>
---
 fs/fcntl.c                  |   14 ++++++++++++--
 include/asm-generic/fcntl.h |    2 ++
 2 files changed, 14 insertions(+), 2 deletions(-)
fs/fcntl.c +12 -2
@@ -709,11 +709,21 @@ 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(17 != hweight_long(
+		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
+		));
+
 	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 +2 -0
@@ -4,6 +4,8 @@
 #include <linux/types.h>
 
 /*
+ * When introducing new O_* bits, please check its uniqueness in fcntl_init().
+ *
  * FMODE_EXEC is 0x20
  * FMODE_NONOTIFY is 0x1000000
  * These cannot be used by userspace O_* until internal and external open