Subject: Re: Massive lossage with -current as of tonight?
To: Scott Ellis <email@example.com>
From: Scott Ellis <firstname.lastname@example.org>
Date: 11/04/2006 07:24:49
Scott Ellis wrote:
> Scott Ellis wrote:
>> I'm seeing significant problems with -current as of this evening.
> Well, after cvs updating and doing a complete rebuild (so using -current
> as of ~10pm PST Nov 3rd), I get the same behavior as before: Various
> programs appear to hang when booting multi-user.
The sshd problem turns out to be a missing openlog_r():
"/usr/lib/libssh.so.5: Undefined PLT symbol \"openlog_r\" (symnum = 321\
Makes perfect sense since I had to revert to libc.so.12.147, which is
prior to the whole _r/_ss thrash.
Looking into why pam_start suddenly started failing, a ktrace of "login
foobar" under single user shows:
218 1 login CALL __socket30(1,2,0)
218 1 login RET __socket30 4
218 1 login CALL fcntl(4,2,1)
218 1 login RET fcntl 0
218 1 login CALL connect(4,0x7f7ffd7ed7e0,0xe)
218 1 login NAMI "/var/run/log"
218 1 login RET connect -1 errno 2 No such file or directory
218 1 login CALL close(4)
218 1 login RET close 0
218 1 login CALL sendto(0xffffffff,0x7f7fffffd860,0x4e,0,0,0)
218 1 login RET sendto -1 errno 9 Bad file descriptor
218 1 login CALL close(0xffffffff)
218 1 login RET close -1 errno 9 Bad file descriptor
This doesn't make much sense since /var/run/log is a socket that looks
to have been newly created. Perhaps this is fallout of using an old
libc as well, w/o all the syslog changes.
I've nuked my source, and re-checked it out via CVS, to see if a totally
fresh build will change things.
Is really nobody else seeing this with a kernel and userland from Nov 2nd+?