[PATCH] Fix bug 9412 - SMB2 server doesn't support recvfile.

[Volker Lendecke]

On Tue, Apr 01, 2014 at 07:42:50PM +0800, Jones wrote:
> 2014-03-31 18:02 GMT+08:00 Volker Lendecke <Volker.Lendecke at sernet.de>:
> 
> > Just last week I did a comparison of 4.0.something against
> > master with async I/O, and the user-space CPU had dropped by
> > roughly half.
> >
> 
> 
> Hello Volker,
> 
> Thanks for kindly response!
> 
> After testing the same workload on different release,
> including samba-4.0 and master (git on 2014/03/31),
> and the iometer shows an inspiring result:
> 
> (unit: IOps)
> samba-4.0.13  read:46712 write:28038
> samba-4.0.14  read:84230 write:28627
> samba-master  read:95363 write:50472
> 
> Guess the improvement between 4.0.{13,14} might be this one,
> that's really amazing:
> https://bugzilla.samba.org/show_bug.cgi?id=10298

Yep, that's the one. That and some more patches in master. I
think metze has some more in the pipeline.

> With 'strace' on samba-master,
> there are 1 RECVFILE with 4 fcntl too,
> hence i would like to try the part as Jeremy mentioned earlier,
> thanks.
> 
> fcntl64(31, F_GETFL)                    = 0x802 (flags O_RDWR|O_NONBLOCK)
> fcntl64(31, F_GETFL)                    = 0x802 (flags O_RDWR|O_NONBLOCK)
> fcntl64(31, F_SETFL, O_RDWR)            = 0
> _llseek(37, 290012672, [290012672], SEEK_SET) = 0
> read(31, "g\1\307_\233\304\203\243\344\212\277VJ\310;\325P-\235\272"...,
> 512) = 512
> write(37, "g\1\307_\233\304\203\243\344\212\277VJ\310;\325P-\235\272"...,
> 512) = 512
> fcntl64(31, F_SETFL, O_RDWR|O_NONBLOCK) = 0

Can you send me a more complete strace? I haven't followed,
but have you tried "strict locking = no" and "posix locking
= no"?

Volker

-- 
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de