> Steven N. Hirsch wrote:
>
> > The latest patch you sent against 2.1.63 has made a huge improvement. I'm
> > still seeing a write error the very first time a file create is attempted
> > after mount, though.
> >
> > Here's some log output:
>
> > Nov 15 14:33:09 cy kernel: smb_proc_open: BENCH/iozone.tmp open failed, error=-26, retrying R/O
> > Nov 15 14:33:09 cy kernel: smb_open: BENCH/iozone.tmp access denied, access=0, wish=1
> >
> > ^^^^^^^^^^^^^- Here's the first attempt.
>
> Hi Steve,
> Glad to hear we're making progress again. From the error log it appears
> that on your first attempt the tmp file was still busy when smbfs tried
> to reopen it. (Code -26 is ETXTBUSY). Perhaps it would be worthwhile to
> try a small delay again?
Bill,
The first access to the tmp file was when iozone first tried to create it
after mounting the share. It failed immediately with a write error, and
I'm quite sure that the file did _not_ exist on the server prior to
running the benchmark. The second time, it ran to completion.
>
> > Nov 15 14:35:51 cy kernel: smb_request_ok: rcls=1, err=18 mapped to 0
> > Nov 15 14:35:51 cy kernel: smb_request_ok: verify failed, result=-5
> > Nov 15 14:36:00 cy kernel: smb_errno: ERRDOS code 50, returning EIO
> > Nov 15 14:36:01 cy kernel: smb_errno: ERRDOS code 50, returning EIO
>
> The "ERRDOS code 50" errors are what others have reported occur when
> creating a file, except that in some cases it seems to prevent correct
> operation. Is this repeatable for you? I really want to track this down,
> but I need to find what procedure is generating the messages.
I'm not sure what is triggering the ERRDOS messages, they just seem to be
sprinkled throughout the logs.
I'll try to quantify what I'm doing when they emerge.
Steve