Subject: Re: [netatalk-admins] Adrian's afpd server not letting go?
From: Michael M Han (han@windy.ckm.ucsf.edu)
Date: Tue Mar 10 1998 - 12:41:57 EST
There's a patch against umich 1.4b2 for Solaris. Its description seems
to match your symptoms. I had to be a little intuitive about applying
it to netatalk+asun, but I *think* what you need to do is this:
*** libatalk/asp/asp_getsess.c-
--- libatalk/asp/asp_getsess.c
***************
*** 135,141 ****
atpb.atp_rreqdata = asp->cmdbuf;
atpb.atp_rreqdlen = sizeof( asp->cmdbuf );
while ( atp_rreq( asp->asp_atp, &atpb ) < 0 ) {
! if ( errno == EINTR ) {
continue;
}
return( NULL );
--- 135,141 ----
atpb.atp_rreqdata = asp->cmdbuf;
atpb.atp_rreqdlen = sizeof( asp->cmdbuf );
while ( atp_rreq( asp->asp_atp, &atpb ) < 0 ) {
! if ( errno == EINTR || errno == EAGAIN ) {
continue;
}
return( NULL );
Now while I *did* do this to my own source and got it to compile and
run correctly, my knowledge of C is more or less limited to things
like 10-liners to do "Hello, world!\n" so I really couldn't say if
what I've done is right. It seems to be to my rather naive eye. The
original patch I found that I adapted to +asun was found at:
http://www.umich.edu/~rsug/netatalk/patches/1.4b2/solariseagain.diff
If someone more knowledgable about C and Solaris error conditions
would care to confirm or deny, I'd appreciate it.
Of course, a simpler approach would be to bump the -c flag handed to
afpd in /etc/init.d/atalk.
Previously...
>
>I'm running netatalk-1.4b2+asun18.2 on the new Solaris kernel module.
>
>The other day when I tried to login, I got the "too many users logged in"
>message. As I have fewer Mac users than users allowed, this came as a
>surprise to me.
>
>Further investigation showed old afpd processes which had not died,
>despite the client being no longer in touch with the server. In fact
>there was one person with 3 afpd processes against her name.
>
>Anybody seen this? I think it happens more when the user's Mac has
>crashed, rather than the connection being terminated the normal way.
>
>Either way I get messages like this on my console:
>Mar 10 13:19:18 lomax afpd[2896]: atp_rresp: Connection timed out
>Mar 10 13:19:28 lomax afpd[2896]: afp_die: asp_shutdown: Connection timed out
>
>Before "asun", I would only get the last error...
>
>
>While I'm here, more problems.... On one occasion the master "afpd"
>daemon died completely, requiring a manual restart. No apparent
>reason for this.
>
>
>A new class of error message for me:
>Mar 4 09:15:27 lomax afpd[12219]: afp_addicon: Permission denied
>Mar 4 09:15:36 lomax last message repeated 258 times
>Mar 9 15:24:06 lomax afpd[2060]: afp_addicon: No such file or directory
>
>As you can see, the "Permission denied" error caused a lot of fun for
>my syslogd.
>
>
>Anyone, perhaps Adrian, got any thoughts on these matters?
>
>Of the three the dying afpd is the most serious, but has only happened
>once or twice.
>
>Thanks a bunch anyway, I've just got my hands on MacOS 8 and it's nice to
>be able to access my files at a more respectable speed now.
>
>Cheers,
>
>
>Duncan Sinclair.
_________
mike (han@library.ucsf.edu)
I will not squeak chalk
- The collected wisdom of Bart Simpson
This archive was generated by hypermail 2b28 : Sat Dec 18 1999 - 16:31:25 EST