Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cornell!uw-beaver!rice!sun-spots-request
From: phil@Rice.edu (William LeFebvre)
Newsgroups: comp.sys.sun
Subject: Re: ldtermrsrv: out of blocks
Keywords: SunOS
Message-ID: <2993@kalliope.rice.edu>
Date: 30 Mar 89 23:05:34 GMT
Sender: usenet@rice.edu
Organization: Sun-Spots
Lines: 44
Approved: Sun-Spots@rice.edu
Original-Date: Tue, 28 Mar 89 12:28:42 CST
X-Sun-Spots-Digest: Volume 7, Issue 217, message 2 of 11

After seeing so many people complain about this problem, I have decided to
post what we are currenty using for NBLK settings in "param.c".  Before
increasing these values, we would routinely get error messages from the
terminal driver.  Unfortuantely, there is no easy way to set these values
in the config file.  So here's what you do.  After "config TYPE" (where
"TYPE" is whatever hthe kernel id name is), cd to "../TYPE".  Then edit
the file called "param.c" in that directory.  Most of the changes are
going to occur around line 144, just after the comment "Stream data
structures".  After making the appropriate changes, type "make".

Here is what we use for a 3/280 that has 32 ALM lines, 2 console lines,
and 32 pseudo tty lines.  I have noted the original values, if different,
in comments off to the right.
__________

/*
 * Stream data structures.
 * XXX - should be dynamically allocated.
 */
#define NBLK4096          1		/* was   0 */
#define NBLK2048         32
#define NBLK1024         12
#define NBLK512          10		/* was   8 */
#define NBLK256          48		/* was  32 */
#define NBLK128         192		/* was  96 */
#define NBLK64          240
#define NBLK16          400
#define NBLK4           256		/* was 128 */
#define NBLK (NBLK4096 + NBLK2048 + NBLK1024 + NBLK512 + NBLK256 + NBLK128 + \
        NBLK64 + NBLK16 + NBLK4)

#define NSTREAM         64		/* was  32 */
__________

I based these figures on the output of "netstat -m", which tells you the
number of allocation failures for each block size.  We had a few failures
for 4096 blocks so I added one, but I am still getting failures so it
doesn't appear to have done any good.  We are also still getting failures
for 4 bytes blocks.  But all the other failues are 0, and we haven't had
any terminal problems since.  If you want to go to 64 pseudo ttys, then
blocks of size 512, 256, and 128 are going to have to be increased more,
especially 128 byte blocks.

	William LeFebvre