Re: [Qemu-devel] [PATCH] block: workaround for unaligned byte range in fallocate()

[Andrey Shinkevich]

On 23/08/2019 00:09, Eric Blake wrote:
> On 8/22/19 1:31 PM, Andrey Shinkevich wrote:
>> Revert the commit 118f99442d 'block/io.c: fix for the allocation failure'
>> and make better error handling for the file systems that do not support
> 
> s/make/use/
> 
>> fallocate() for the unaligned byte range. Allow falling back to pwrite
> 
> s/the/an/
> 
>> in case fallocate() returns EINVAL.
>>
>> Suggested-by: Kevin Wolf <address@hidden>
>> Suggested-by: Eric Blake <address@hidden>
>> Signed-off-by: Andrey Shinkevich <address@hidden>
>> ---
>> Discussed in email thread with the message ID
>> <address@hidden>
>>
>>   block/file-posix.c | 7 +++++++
>>   block/io.c         | 2 +-
>>   2 files changed, 8 insertions(+), 1 deletion(-)
>>
>> diff --git a/block/file-posix.c b/block/file-posix.c
>> index fbeb006..2c254ff 100644
>> --- a/block/file-posix.c
>> +++ b/block/file-posix.c
>> @@ -1588,6 +1588,13 @@ static int handle_aiocb_write_zeroes(void *opaque)
>>       if (s->has_write_zeroes) {
>>           int ret = do_fallocate(s->fd, FALLOC_FL_ZERO_RANGE,
>>                                  aiocb->aio_offset, aiocb->aio_nbytes);
>> +        if (ret == -EINVAL) {
>> +            /*
>> +             * Allow falling back to pwrite for file systems that
>> +             * do not support fallocate() for unaligned byte range.
> 
> s/for/for an/
> 
>> +             */
>> +            return -ENOTSUP;
>> +        }
>>           if (ret == 0 || ret != -ENOTSUP) {
>>               return ret;
>>           }
>> diff --git a/block/io.c b/block/io.c
>> index 56bbf19..58f08cd 100644
>> --- a/block/io.c
>> +++ b/block/io.c
>> @@ -1558,7 +1558,7 @@ static int coroutine_fn 
>> bdrv_co_do_pwrite_zeroes(BlockDriverState *bs,
>>               assert(!bs->supported_zero_flags);
>>           }
>>   
>> -        if (ret < 0 && !(flags & BDRV_REQ_NO_FALLBACK)) {
>> +        if (ret == -ENOTSUP && !(flags & BDRV_REQ_NO_FALLBACK)) {
>>               /* Fall back to bounce buffer if write zeroes is unsupported */
>>               BdrvRequestFlags write_flags = flags & ~BDRV_REQ_ZERO_WRITE;
>>   
>>
> 
> Reviewed-by: Eric Blake <address@hidden>
> 

Thank you Eric!
Also, I would prefer "make the finer error handling" because the word 
'better' has a wide sense.
I have prepared the v2 with your corrections and am waiting for Denis 
Lunev's response. He is coming back to work from his vacation on Tuesday.

Andrey
-- 
With the best regards,
Andrey Shinkevich