Re: [PATCH -next 1/2] nfs: nfs{,4}_file_flush should consume writeback error

["chenxiaosong (A)" <chenxiaosong2@huawei.com>]

在 2022/4/20 16:50, chenxiaosong (A) 写道:
> 在 2022/4/12 22:27, Trond Myklebust 写道:
> 
>>
>> It will clear ENOSPC, EDQUOT and EFBIG. It should not clear other
>> errors that are not supposed to be reported by write().
>>
>> As I keep repeating, that is _documented behaviour_!
>>
> 
> Hi Trond:
> 
> You may mean that write(2) manpage described:
> 
>> Since Linux 4.13, errors from write-back come with a promise that
>> they may be reported by subsequent.  write(2) requests, and will be
>> reported by a subsequent fsync(2) (whether or not they were also
>> reported by write(2)).
> 
> The manpage mentioned that "reported by a subsequent fsync(2)", your 
> patch[1] clear the wb err on _async_ write(), and wb err will _not_ be 
> reported by subsequent fsync(2), is it documented behaviour?
> 
> All other filesystems will _not_ clear any wb err on _async_ write().
> 
> [1] 
> https://patchwork.kernel.org/project/linux-nfs/patch/20220411213346.762302-4-trondmy@kernel.org/ 
> 


Hi Trond:

write(2) manpage described:

> On some filesystems, including NFS, it does not even guarantee that
> space has successfully been reserved for the data.  In this case, some
> errors might be delayed until a future write(2), fsync(2), or even
> close(2).  The only way to be sure is to call fsync(2) after you are
> done writing all your data.

Maybe it mean that: writeback errors of NFS is delayed until future sync 
write() and fsync(), because sync write() will call fsync(). We all 
agreed that close()->flush() should not clear writeback errors, should 
async write() do the same thing like other filesystems?