1. Patchew
  2. linux
  3. View series

[PATCH] uapi/fcntl: conditionally define AT_RENAME* macros

Randy Dunlap posted 1 patch 1 month, 1 week ago 
include/uapi/linux/fcntl.h |    2 ++
1 file changed, 2 insertions(+)
[PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Randy Dunlap 1 month, 1 week ago 
Don't define the AT_RENAME_* macros when __USE_GNU is defined since
/usr/include/stdio.h defines them in that case (i.e. when _GNU_SOURCE
is defined, which causes __USE_GNU to be defined).

Having them defined in 2 places causes build warnings (duplicate
definitions) in both samples/watch_queue/watch_test.c and
samples/vfs/test-statx.c.

Fixes: b4fef22c2fb9 ("uapi: explain how per-syscall AT_* flags should be allocated")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
---
Cc: Amir Goldstein <amir73il@gmail.com>
Cc: Jeff Layton <jlayton@kernel.org>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Alexander Aring <alex.aring@gmail.com>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: Aleksa Sarai <cyphar@cyphar.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Christian Brauner <brauner@kernel.org>
Cc: linux-fsdevel@vger.kernel.org
CC: linux-api@vger.kernel.org

 include/uapi/linux/fcntl.h |    2 ++
 1 file changed, 2 insertions(+)

--- linux-next-20250819.orig/include/uapi/linux/fcntl.h
+++ linux-next-20250819/include/uapi/linux/fcntl.h
@@ -155,10 +155,12 @@
  * as possible, so we can use them for generic bits in the future if necessary.
  */
 
+#ifndef __USE_GNU
 /* Flags for renameat2(2) (must match legacy RENAME_* flags). */
 #define AT_RENAME_NOREPLACE	0x0001
 #define AT_RENAME_EXCHANGE	0x0002
 #define AT_RENAME_WHITEOUT	0x0004
+#endif
 
 /* Flag for faccessat(2). */
 #define AT_EACCESS		0x200	/* Test access permitted for
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Matthew Wilcox 1 month, 1 week ago 
On Sun, Aug 24, 2025 at 03:10:55PM -0700, Randy Dunlap wrote:
> Don't define the AT_RENAME_* macros when __USE_GNU is defined since
> /usr/include/stdio.h defines them in that case (i.e. when _GNU_SOURCE
> is defined, which causes __USE_GNU to be defined).
> 
> Having them defined in 2 places causes build warnings (duplicate
> definitions) in both samples/watch_queue/watch_test.c and
> samples/vfs/test-statx.c.

It does?  What flags?

#define AT_RENAME_NOREPLACE     0x0001
#define AT_RENAME_NOREPLACE     0x0001

int main(void)
{
	return AT_RENAME_NOREPLACE;
}

gcc -W -Wall testA.c -o testA

(no warnings)

I'm pretty sure C says that duplicate definitions are fine as long
as they're identical.
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Randy Dunlap 1 month, 1 week ago 

On 8/24/25 4:21 PM, Matthew Wilcox wrote:
> On Sun, Aug 24, 2025 at 03:10:55PM -0700, Randy Dunlap wrote:
>> Don't define the AT_RENAME_* macros when __USE_GNU is defined since
>> /usr/include/stdio.h defines them in that case (i.e. when _GNU_SOURCE
>> is defined, which causes __USE_GNU to be defined).
>>
>> Having them defined in 2 places causes build warnings (duplicate
>> definitions) in both samples/watch_queue/watch_test.c and
>> samples/vfs/test-statx.c.
> 
> It does?  What flags?
> 

for samples/vfs/test-statx.c:

In file included from ../samples/vfs/test-statx.c:23:
usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
  159 | #define AT_RENAME_NOREPLACE     0x0001
In file included from ../samples/vfs/test-statx.c:13:
/usr/include/stdio.h:171:10: note: this is the location of the previous definition
  171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
usr/include/linux/fcntl.h:160:9: warning: ‘AT_RENAME_EXCHANGE’ redefined
  160 | #define AT_RENAME_EXCHANGE      0x0002
/usr/include/stdio.h:173:10: note: this is the location of the previous definition
  173 | # define AT_RENAME_EXCHANGE RENAME_EXCHANGE
usr/include/linux/fcntl.h:161:9: warning: ‘AT_RENAME_WHITEOUT’ redefined
  161 | #define AT_RENAME_WHITEOUT      0x0004
/usr/include/stdio.h:175:10: note: this is the location of the previous definition
  175 | # define AT_RENAME_WHITEOUT RENAME_WHITEOUT

for samples/watch_queue/watch_test.c:

In file included from usr/include/linux/watch_queue.h:6,
                 from ../samples/watch_queue/watch_test.c:19:
usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
  159 | #define AT_RENAME_NOREPLACE     0x0001
In file included from ../samples/watch_queue/watch_test.c:11:
/usr/include/stdio.h:171:10: note: this is the location of the previous definition
  171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
usr/include/linux/fcntl.h:160:9: warning: ‘AT_RENAME_EXCHANGE’ redefined
  160 | #define AT_RENAME_EXCHANGE      0x0002
/usr/include/stdio.h:173:10: note: this is the location of the previous definition
  173 | # define AT_RENAME_EXCHANGE RENAME_EXCHANGE
usr/include/linux/fcntl.h:161:9: warning: ‘AT_RENAME_WHITEOUT’ redefined
  161 | #define AT_RENAME_WHITEOUT      0x0004
/usr/include/stdio.h:175:10: note: this is the location of the previous definition
  175 | # define AT_RENAME_WHITEOUT RENAME_WHITEOUT


> #define AT_RENAME_NOREPLACE     0x0001
> #define AT_RENAME_NOREPLACE     0x0001
> 
> int main(void)
> {
> 	return AT_RENAME_NOREPLACE;
> }
> 
> gcc -W -Wall testA.c -o testA
> 
> (no warnings)
> 
> I'm pretty sure C says that duplicate definitions are fine as long
> as they're identical.
The vales are identical but the strings are not identical.

We can't fix stdio.h, but we could just change uapi/linux/fcntl.h
to match stdio.h. I suppose.

-- 
~Randy
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Matthew Wilcox 1 month, 1 week ago 
On Sun, Aug 24, 2025 at 04:54:50PM -0700, Randy Dunlap wrote:
> In file included from ../samples/vfs/test-statx.c:23:
> usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
>   159 | #define AT_RENAME_NOREPLACE     0x0001
> In file included from ../samples/vfs/test-statx.c:13:
> /usr/include/stdio.h:171:10: note: this is the location of the previous definition
>   171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE

Oh dear.  This is going to be libc-version-dependent.

$ grep -r AT_RENAME_NOREPLACE /usr/include
/usr/include/linux/fcntl.h:#define AT_RENAME_NOREPLACE	0x0001

It's not in stdio.h at all.  This is with libc6 2.41-10
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Randy Dunlap 1 month, 1 week ago 

On 8/25/25 6:03 AM, Matthew Wilcox wrote:
> On Sun, Aug 24, 2025 at 04:54:50PM -0700, Randy Dunlap wrote:
>> In file included from ../samples/vfs/test-statx.c:23:
>> usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
>>   159 | #define AT_RENAME_NOREPLACE     0x0001
>> In file included from ../samples/vfs/test-statx.c:13:
>> /usr/include/stdio.h:171:10: note: this is the location of the previous definition
>>   171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
> 
> Oh dear.  This is going to be libc-version-dependent.
> 

I am not surprised at that.

> $ grep -r AT_RENAME_NOREPLACE /usr/include
> /usr/include/linux/fcntl.h:#define AT_RENAME_NOREPLACE	0x0001
> 
> It's not in stdio.h at all.  This is with libc6 2.41-10

It was added 2025-04-22:
2025-04-22  Joseph Myers  <josmyers@redhat.com>

	COMMIT: cf9241107d12e79073ddb03bab9de115e5e0e856
	Add AT_* constants from Linux 6.12

$ grep -r AT_RENAME_NOREPLACE /usr/include
/usr/include/stdio.h:# define AT_RENAME_NOREPLACE RENAME_NOREPLACE
/usr/include/linux/fcntl.h:#define AT_RENAME_NOREPLACE	0x0001

I have libc 2.42-1.1 (openSUSE).

thanks.
-- 
~Randy
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Matthew Wilcox 1 month, 1 week ago 
On Mon, Aug 25, 2025 at 10:52:31AM -0700, Randy Dunlap wrote:
> $ grep -r AT_RENAME_NOREPLACE /usr/include
> /usr/include/stdio.h:# define AT_RENAME_NOREPLACE RENAME_NOREPLACE
> /usr/include/linux/fcntl.h:#define AT_RENAME_NOREPLACE	0x0001
> 
> I have libc 2.42-1.1 (openSUSE).

I wonder if we can fix it by changing include/uapi/linux/fcntl.h
from being an explicit 0x0001 to RENAME_NOREPLACE?  There's probably
a horrendous include problem between linux/fcntl.h and linux/fs.h
though?
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Randy Dunlap 1 month, 1 week ago 

On 8/25/25 11:45 AM, Matthew Wilcox wrote:
> On Mon, Aug 25, 2025 at 10:52:31AM -0700, Randy Dunlap wrote:
>> $ grep -r AT_RENAME_NOREPLACE /usr/include
>> /usr/include/stdio.h:# define AT_RENAME_NOREPLACE RENAME_NOREPLACE
>> /usr/include/linux/fcntl.h:#define AT_RENAME_NOREPLACE	0x0001
>>
>> I have libc 2.42-1.1 (openSUSE).
> 
> I wonder if we can fix it by changing include/uapi/linux/fcntl.h
> from being an explicit 0x0001 to RENAME_NOREPLACE?  There's probably
> a horrendous include problem between linux/fcntl.h and linux/fs.h
> though?

I'm working on something like that now (suggested by Amir),
but it might depend on whether stdio.h has been #included first.

-- 
~Randy
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Amir Goldstein 1 month, 1 week ago 
On Mon, Aug 25, 2025 at 1:54 AM Randy Dunlap <rdunlap@infradead.org> wrote:
>
>
>
> On 8/24/25 4:21 PM, Matthew Wilcox wrote:
> > On Sun, Aug 24, 2025 at 03:10:55PM -0700, Randy Dunlap wrote:
> >> Don't define the AT_RENAME_* macros when __USE_GNU is defined since
> >> /usr/include/stdio.h defines them in that case (i.e. when _GNU_SOURCE
> >> is defined, which causes __USE_GNU to be defined).
> >>
> >> Having them defined in 2 places causes build warnings (duplicate
> >> definitions) in both samples/watch_queue/watch_test.c and
> >> samples/vfs/test-statx.c.
> >
> > It does?  What flags?
> >
>
> for samples/vfs/test-statx.c:
>
> In file included from ../samples/vfs/test-statx.c:23:
> usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
>   159 | #define AT_RENAME_NOREPLACE     0x0001
> In file included from ../samples/vfs/test-statx.c:13:
> /usr/include/stdio.h:171:10: note: this is the location of the previous definition
>   171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
> usr/include/linux/fcntl.h:160:9: warning: ‘AT_RENAME_EXCHANGE’ redefined
>   160 | #define AT_RENAME_EXCHANGE      0x0002
> /usr/include/stdio.h:173:10: note: this is the location of the previous definition
>   173 | # define AT_RENAME_EXCHANGE RENAME_EXCHANGE
> usr/include/linux/fcntl.h:161:9: warning: ‘AT_RENAME_WHITEOUT’ redefined
>   161 | #define AT_RENAME_WHITEOUT      0x0004
> /usr/include/stdio.h:175:10: note: this is the location of the previous definition
>   175 | # define AT_RENAME_WHITEOUT RENAME_WHITEOUT
>
> for samples/watch_queue/watch_test.c:
>
> In file included from usr/include/linux/watch_queue.h:6,
>                  from ../samples/watch_queue/watch_test.c:19:
> usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
>   159 | #define AT_RENAME_NOREPLACE     0x0001
> In file included from ../samples/watch_queue/watch_test.c:11:
> /usr/include/stdio.h:171:10: note: this is the location of the previous definition
>   171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
> usr/include/linux/fcntl.h:160:9: warning: ‘AT_RENAME_EXCHANGE’ redefined
>   160 | #define AT_RENAME_EXCHANGE      0x0002
> /usr/include/stdio.h:173:10: note: this is the location of the previous definition
>   173 | # define AT_RENAME_EXCHANGE RENAME_EXCHANGE
> usr/include/linux/fcntl.h:161:9: warning: ‘AT_RENAME_WHITEOUT’ redefined
>   161 | #define AT_RENAME_WHITEOUT      0x0004
> /usr/include/stdio.h:175:10: note: this is the location of the previous definition
>   175 | # define AT_RENAME_WHITEOUT RENAME_WHITEOUT
>
>
> > #define AT_RENAME_NOREPLACE     0x0001
> > #define AT_RENAME_NOREPLACE     0x0001
> >
> > int main(void)
> > {
> >       return AT_RENAME_NOREPLACE;
> > }
> >
> > gcc -W -Wall testA.c -o testA
> >
> > (no warnings)
> >
> > I'm pretty sure C says that duplicate definitions are fine as long
> > as they're identical.
> The vales are identical but the strings are not identical.
>
> We can't fix stdio.h, but we could just change uapi/linux/fcntl.h
> to match stdio.h. I suppose.

I do not specifically object to a patch like this (assuming that is works?):

--- a/include/uapi/linux/fcntl.h
+++ b/include/uapi/linux/fcntl.h
@@ -156,9 +156,9 @@
  */

 /* Flags for renameat2(2) (must match legacy RENAME_* flags). */
-#define AT_RENAME_NOREPLACE    0x0001
-#define AT_RENAME_EXCHANGE     0x0002
-#define AT_RENAME_WHITEOUT     0x0004
+#define AT_RENAME_NOREPLACE    RENAME_NOREPLACE
+#define AT_RENAME_EXCHANGE     RENAME_EXCHANGE
+#define AT_RENAME_WHITEOUT     RENAME_WHITEOUT


But to be clear, this is a regression introduced by glibc that is likely
to break many other builds, not only the kernel samples
and even if we fix linux uapi to conform to its downstream
copy of definitions, it won't help those users whose programs
build was broken until they install kernel headers, so feels like you
should report this regression to glibc and they'd better not "fix" the
regression by copying the current definition string as that may change as per
the patch above.

Why would a library copy definitions from kernel uapi without
wrapping them with #ifndef or #undef?

Thanks,
Amir.
Re: [PATCH] uapi/fcntl: conditionally define AT_RENAME* macros
Posted by Randy Dunlap 1 month, 1 week ago 
Hi Amir,


On 8/24/25 10:58 PM, Amir Goldstein wrote:
> On Mon, Aug 25, 2025 at 1:54 AM Randy Dunlap <rdunlap@infradead.org> wrote:
>>
>>
>>
>> On 8/24/25 4:21 PM, Matthew Wilcox wrote:
>>> On Sun, Aug 24, 2025 at 03:10:55PM -0700, Randy Dunlap wrote:
>>>> Don't define the AT_RENAME_* macros when __USE_GNU is defined since
>>>> /usr/include/stdio.h defines them in that case (i.e. when _GNU_SOURCE
>>>> is defined, which causes __USE_GNU to be defined).
>>>>
>>>> Having them defined in 2 places causes build warnings (duplicate
>>>> definitions) in both samples/watch_queue/watch_test.c and
>>>> samples/vfs/test-statx.c.
>>>
>>> It does?  What flags?
>>>
>>
>> for samples/vfs/test-statx.c:
>>
>> In file included from ../samples/vfs/test-statx.c:23:
>> usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
>>   159 | #define AT_RENAME_NOREPLACE     0x0001
>> In file included from ../samples/vfs/test-statx.c:13:
>> /usr/include/stdio.h:171:10: note: this is the location of the previous definition
>>   171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
>> usr/include/linux/fcntl.h:160:9: warning: ‘AT_RENAME_EXCHANGE’ redefined
>>   160 | #define AT_RENAME_EXCHANGE      0x0002
>> /usr/include/stdio.h:173:10: note: this is the location of the previous definition
>>   173 | # define AT_RENAME_EXCHANGE RENAME_EXCHANGE
>> usr/include/linux/fcntl.h:161:9: warning: ‘AT_RENAME_WHITEOUT’ redefined
>>   161 | #define AT_RENAME_WHITEOUT      0x0004
>> /usr/include/stdio.h:175:10: note: this is the location of the previous definition
>>   175 | # define AT_RENAME_WHITEOUT RENAME_WHITEOUT
>>
>> for samples/watch_queue/watch_test.c:
>>
>> In file included from usr/include/linux/watch_queue.h:6,
>>                  from ../samples/watch_queue/watch_test.c:19:
>> usr/include/linux/fcntl.h:159:9: warning: ‘AT_RENAME_NOREPLACE’ redefined
>>   159 | #define AT_RENAME_NOREPLACE     0x0001
>> In file included from ../samples/watch_queue/watch_test.c:11:
>> /usr/include/stdio.h:171:10: note: this is the location of the previous definition
>>   171 | # define AT_RENAME_NOREPLACE RENAME_NOREPLACE
>> usr/include/linux/fcntl.h:160:9: warning: ‘AT_RENAME_EXCHANGE’ redefined
>>   160 | #define AT_RENAME_EXCHANGE      0x0002
>> /usr/include/stdio.h:173:10: note: this is the location of the previous definition
>>   173 | # define AT_RENAME_EXCHANGE RENAME_EXCHANGE
>> usr/include/linux/fcntl.h:161:9: warning: ‘AT_RENAME_WHITEOUT’ redefined
>>   161 | #define AT_RENAME_WHITEOUT      0x0004
>> /usr/include/stdio.h:175:10: note: this is the location of the previous definition
>>   175 | # define AT_RENAME_WHITEOUT RENAME_WHITEOUT
>>

>>>
>>> I'm pretty sure C says that duplicate definitions are fine as long
>>> as they're identical.
>> The vales are identical but the strings are not identical.
>>
>> We can't fix stdio.h, but we could just change uapi/linux/fcntl.h
>> to match stdio.h. I suppose.
> 
> I do not specifically object to a patch like this (assuming that is works?):
> 
> --- a/include/uapi/linux/fcntl.h
> +++ b/include/uapi/linux/fcntl.h
> @@ -156,9 +156,9 @@
>   */
> 
>  /* Flags for renameat2(2) (must match legacy RENAME_* flags). */
> -#define AT_RENAME_NOREPLACE    0x0001
> -#define AT_RENAME_EXCHANGE     0x0002
> -#define AT_RENAME_WHITEOUT     0x0004
> +#define AT_RENAME_NOREPLACE    RENAME_NOREPLACE
> +#define AT_RENAME_EXCHANGE     RENAME_EXCHANGE
> +#define AT_RENAME_WHITEOUT     RENAME_WHITEOUT
> 

I'll test that.

> 
> But to be clear, this is a regression introduced by glibc that is likely
> to break many other builds, not only the kernel samples
> and even if we fix linux uapi to conform to its downstream
> copy of definitions, it won't help those users whose programs
> build was broken until they install kernel headers, so feels like you
> should report this regression to glibc and they'd better not "fix" the
> regression by copying the current definition string as that may change as per
> the patch above.
> 

I'll look into that also.

> Why would a library copy definitions from kernel uapi without
> wrapping them with #ifndef or #undef?

To me it looks like they stuck them into the wrong file - stdio.h
instead of fcntl.h.

thanks.
-- 
~Randy

Patchew 2.1-devel-b67e4c2

© 2016 - 2025 Red Hat, Inc.