[linux-cifs-client] Re: sendmsg blocking with sendtimeout vs. non-blocking

[Evgeniy Polyakov]

Hi.

On Thu, Oct 23, 2008 at 12:40:05PM -0500, Steve French (smfrench at gmail.com) wrote:
> On Thu, Oct 23, 2008 at 11:42 AM, Steve French <smfrench at gmail.com> wrote:
> > On Thu, Oct 23, 2008 at 11:26 AM, Shirish Pargaonkar
> > As I look into this sndbuf and rcvbuf size setting ... what concerns
> > me is why nfs sets these sizes for snd and rcvbuf sizes still if they
> > don't need to be set?  We (cifs) have larger write sizes (56K) than
> > nfs's default.   See svc_set_sockbufsize in net/sunrpc/svcsock.c
> 
> What is the valid range for the sndbuf and rcvbuf size so I can sanity
> check this if the user overrides it on mount?

>From zero to infinity. Actual size will be automatically ajusted by the
kernel, but will not exceed specified one, so there is no need to tune
this parameter. Very likely you do not want to change socket queue sizes,
since it may hurt performance, when previously autotuning could rise the
maximum socket buffer size.

-- 
	Evgeniy Polyakov