Re: fixed up version of testcfr.c

Trond Myklebust wrote:
>Hi Rick
>
>On Tue, 2019-10-15 at 23:33 +0000, Rick Macklem wrote:
>> Attached is a version of testcfr.c that seems to work for a Linux
>> NFSv4.2 client when compiled with "-DDOSYNC". DOSYNC adds
>> some fsync(2) calls plus close()/open() after doing the
>> copy_file_range().
>> { I'm not sure that a client app should need to do these things, but
>> that's
>>   up to the Linux community. }
>>
>> At least for Fedora30, a copy_file_range() where the in_fd and out_fd
>> are
>> the same file and the byte range overlaps returns success, but the
>> data
>> seems bogus. However, this occurs when run on a local FS as well, so
>> I don't think it's a Linux NFSv4.2 client bug.
>>
>
>
>The above would be a server bug. According to RFC7862 section 15.2.3:
>
>   SAVED_FH and CURRENT_FH must be different files.  If SAVED_FH and
>   CURRENT_FH refer to the same file, the operation MUST fail with
>   NFS4ERR_INVAL.
If the client attempted a Copy for this case, then, yes.
But if you look at a packet trace, the client doesn't attempt a Copy, but tries
to do this with a Write.

rick


--
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]



_______________________________________________
nfsv4 mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/nfsv4