Em Domingo 29 Julho 2007 11:07, David Laight escreveu:
> On Sun, Jul 29, 2007 at 01:43:50PM +0900, Izumi Tsutsui wrote:
> > I've upgrade a kernel on my server (i386) from 4.99.20 (around June 20)
> > to yesterday's 4.99.25, but after that ypserv(8) (both old 4.99.20
> > and new 4.99.25 binaries) seems to stop working.
> >
> > ypserv(8) is started via rc.d(8) normally, but it doesn't
> > respond any requests from ypbind(8) even from localhost.
> > ypserv -l (logging requests) option just says
> > "domain_nonack_2: request from localhost, domain foo, served TRUE"
> >
> > Does anyone see the similar problem?

I upgraded on 2007/08/08 (4.99.27), and:

    - ntpd causes the machine to hang (DDB works to some extent, but if I kill 
ntpd, in an attempt to resume operation, things get really worse: I start 
getting kernel messages telling me that viaide (SATA) is missing interrupts 
(or something like that -- I was too tired when I tried it, and didn't write 
it down, nor could possibly get a crash dump once it started, since the disks 
became inaccessible).
    - ypserv still fails to be contacted via broadcast or even direct 
specification, both for remote clients and localhost.

Before that, I had kernels built about 2007/07/25 and 2007/08/02, with which I 
don't remember having ntpd related hangs.  I also cannott remember problems 
with NIS with (then-)new kernels before I updated userland.  But after I 
began having problems, I did some investigation, and learned that if I 
started ypbind with "-ypsetme" option, than ypbind would come up, but not 
work until I used ypset to force it to contact the troublesome server.  Even 
then, however, data retrieved from NIS maps was truncated or corrupted.

I have not yet repeated the tests with the latest kernel/userland.

> There is a PR against it, I suspect it got broken somewhere
> in my changes to the socket code.
> Unfortunately I don't run ypbind ...

What PR?

-- 
	Pappires

... Qui habet aurem audiat quid Spiritus dicat ecclesiis.