Re: [PATCH v2 05/16] fs: Add RWF_ATOMIC and IOCB_ATOMIC flags for atomic write support

[John Garry <john.g.garry@oracle.com>]

On 13/12/2023 13:31, Al Viro wrote:
>> Add file mode flag FMODE_CAN_ATOMIC_WRITE, so files which do not have the
>> flag set will have RWF_ATOMIC rejected and not just ignored.
>>
>> Signed-off-by: Prasad Singamsetty<prasad.singamsetty@oracle.com>
>> Signed-off-by: John Garry<john.g.garry@oracle.com>
>> ---
>>   include/linux/fs.h      | 8 ++++++++
>>   include/uapi/linux/fs.h | 5 ++++-
>>   2 files changed, 12 insertions(+), 1 deletion(-)
>>
>> diff --git a/include/linux/fs.h b/include/linux/fs.h
>> index 70329c81be31..d725c194243c 100644
>> --- a/include/linux/fs.h
>> +++ b/include/linux/fs.h
>> @@ -185,6 +185,9 @@ typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
>>   /* File supports async nowait buffered writes */
>>   #define FMODE_BUF_WASYNC	((__force fmode_t)0x80000000)
>>   
>> +/* File supports atomic writes */
>> +#define FMODE_CAN_ATOMIC_WRITE	((__force fmode_t)0x100000000)
> Have you even tried to compile that on e.g. arm?

i386 and now arm32, and no grumblings.

I think that the issue is that we only ever do a bitwise OR or test that 
bit 33 for a 32b value, and it is a void operation and ignored.

However if I have file.f_mode = FMODE_CAN_ATOMIC_WRITE and compile for 
arm32, then it complains.