Re: [PATCH util-linux v2] fallocate: add FALLOC_FL_WRITE_ZEROES support

[Zhang Yi <yi.zhang@huaweicloud.com>]

Thank you for your review comments!

On 2025/8/15 0:52, Darrick J. Wong wrote:
> On Wed, Aug 13, 2025 at 10:40:15AM +0800, Zhang Yi wrote:
>> From: Zhang Yi <yi.zhang@huawei.com>
>>
>> The Linux kernel (since version 6.17) supports FALLOC_FL_WRITE_ZEROES in
>> fallocate(2). Add support for FALLOC_FL_WRITE_ZEROES to the fallocate
>> utility by introducing a new option -w|--write-zeroes.
>>
>> Link: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=278c7d9b5e0c
>> Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
>> ---
>> v1->v2:
>>  - Minor description modification to align with the kernel.
>>
>>  sys-utils/fallocate.1.adoc | 11 +++++++++--
>>  sys-utils/fallocate.c      | 20 ++++++++++++++++----
>>  2 files changed, 25 insertions(+), 6 deletions(-)
>>
>> diff --git a/sys-utils/fallocate.1.adoc b/sys-utils/fallocate.1.adoc
>> index 44ee0ef4c..0ec9ff9a9 100644
>> --- a/sys-utils/fallocate.1.adoc
>> +++ b/sys-utils/fallocate.1.adoc
>> @@ -12,7 +12,7 @@ fallocate - preallocate or deallocate space to a file
> 
> <snip all the long lines>
> 
>> +*-w*, *--write-zeroes*::
>> +Zeroes space in the byte range starting at _offset_ and continuing
>> for _length_ bytes. Within the specified range, blocks are
>> preallocated for the regions that span the holes in the file. After a
>> successful call, subsequent reads from this range will return zeroes,
>> subsequent writes to that range do not require further changes to the
>> file mapping metadata.
> 
> "...will return zeroes and subsequent writes to that range..." ?
> 

Yeah.

>> ++
>> +Zeroing is done within the filesystem by preferably submitting write
> 
> I think we should say less about what the filesystem actually does to
> preserve some flexibility:
> 
> "Zeroing is done within the filesystem. The filesystem may use a
> hardware accelerated zeroing command, or it may submit regular writes.
> The behavior depends on the filesystem design and available hardware."
> 

Sure.

>> zeores commands, the alternative way is submitting actual zeroed data,
>> the specified range will be converted into written extents. The write
>> zeroes command is typically faster than write actual data if the
>> device supports unmap write zeroes, the specified range will not be
>> physically zeroed out on the device.
>> ++
>> +Options *--keep-size* can not be specified for the write-zeroes
>> operation.
>> +
>>  include::man-common/help-version.adoc[]
>>  
>>  == AUTHORS
[..]
>> @@ -429,6 +438,9 @@ int main(int argc, char **argv)
>>  			else if (mode & FALLOC_FL_ZERO_RANGE)
>>  				fprintf(stdout, _("%s: %s (%ju bytes) zeroed.\n"),
>>  								filename, str, length);
>> +			else if (mode & FALLOC_FL_WRITE_ZEROES)
>> +				fprintf(stdout, _("%s: %s (%ju bytes) write zeroed.\n"),
> 
> "write zeroed" is a little strange, but I don't have a better
> suggestion. :)
> 

Hmm... What about simply using "zeroed", the same to FALLOC_FL_ZERO_RANGE?
Users should be aware of the parameters they have passed to fallocate(),
so they should not use this print for further differentiation.

Thanks,
Yi.