Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 28 Feb 1996 13:26:51 -0600 (CST)
From:      Joe Greco <jgreco@brasil.moneng.mei.com>
To:        hackers@freebsd.org
Subject:   Latest 2.1R panic. Hmm.
Message-ID:  <199602281926.NAA03151@brasil.moneng.mei.com>

next in thread | raw e-mail | index | archive | help
Hi,

This is not so much a panic report as it is a growing pains report :-)

My news server freaked an hour ago, when I got back from lunch it was
happily running.  I did "dmesg" to see the crash and got...

hummin# dmesg
ode segment             = base 0x0, limit 0xfffff, type 0x1b
                        = DPL 0, pres 1, de\M^?\^Bran 1
processor eflags        = interrupt enabled, resume, IOPL = 0
current process         = 8891 (innd)
interrupt mask          = net tty bio 
panic: page fault

syncing disks... 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 29 giving up
Automatic reboot in 15 seconds - press a key on the console to abort
Rebooting...
followed by a reasonably normal boot.

Well that "DPL" line looks weird BUT - _MY_ complaint is that there was not
enough room to save the entire message.

A typical "dmesg" output from my news server, freshly booted, is 3693 bytes
from banner to changing root device.  Why?  Because I have
lllllllllllllllllllllllllooooots of stuff connected to my news box.

390 bytes of old message was saved.  That's 4083, very close to 4K.  So I
went looking and found

#define MSG_BSIZE       (4096 - 3 * sizeof(unsigned int))

in sys/msgbuf.h.  There's my 4083.

Ok.  Well.  Can I safely bump this up, recompile dmesg and friends, and be
happy?  (yes syslog too, I know)

This also brings a larger issue to light.  I've spent a lot of time tracking
down these Magick Constants(tm) in BSD code.  Is there a way we can at least
document (or better yet make into kernel compile time options) all of these
magic little critters that big machines might want to tweak?  Things like
MSG_BSIZE and DK_NDRIVE and... and...  oh.  And all the damn pieces of code
that assume that one will never have more than {10,32,etc} network
interfaces.  That doesn't even seem to HAVE a well known constant (I really
blew my stack the other day when I found #define MAXIF 10 in
/usr/src/usr.bin/netstat/if.c).

I would be VERY HAPPY to track down and report things like this that I find
if somebody assures me that efforts will be made to correct them -
preferably with some dynamic methods.  I can of course submit patches that
raise the constants, which is generally what I do to fix the problems..

Still it is sorta cool to run into limits  ;-)

... Joe

-------------------------------------------------------------------------------
Joe Greco - Systems Administrator			      jgreco@ns.sol.net
Solaria Public Access UNIX - Milwaukee, WI			   414/546-7968



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199602281926.NAA03151>