Terrible NFS performance under 9.2-RELEASE?

[John-Mark Gurney]

J David wrote this message on Tue, Jan 28, 2014 at 18:28 -0500:
> On Mon, Jan 27, 2014 at 9:14 PM, John-Mark Gurney <jmg at funkthat.com> wrote:
> > P.S. if someone wants to submit a patch to mbuf.9 to update the docs
> > that would be helpful... I'll review and commit...  and m_append is
> > also undocumented...
> 
> Would that look something like this be a start in that direction?
> 
> http://pastebin.com/UVir1BET

It might be better to move most of m_getm's docs under m_getm2, and
document that m_getm is just m_getm2 w/ M_PKTHDR flag set.

Could you also document that only M_PKTHDR and M_EOR are valid
flags for m_getm2?

> This is all very new to me, so I apologize if that's completely wrong.

Nope, good first start...

> m_append does appear to be documented, it's between m_adj and
> m_prepend.  (At least it is on 9.2, which is the latest tree I have
> access to.)

You are correct.. the problem is that the MLINK isn't setup in the
Makefile, so:
$ man m_append
No manual entry for m_append

I've fixed that, r261254...

> It does also look like m_getm is just a macro in mbuf.h that calls
> m_getm2 with flags set to M_PKTHDR, not a function as described in the
> man page.  It was not immediately obvious if that was intentional or
> something that should be fixed or if it's intentionally meant to be
> treated as a function from an API standpoint.

It's common to use a macro when the change isn't complicated, i.e.
just adding a flag...

Thanks.

-- 
  John-Mark Gurney				Voice: +1 415 225 5579

     "All that I will do, has been done, All that I have, has not."