Subject: kern/697: no buffer cache?
To: None <gnats-admin@NetBSD.ORG>
From: None <muir@idiom.com>
List: netbsd-bugs
Date: 01/04/1995 13:50:05
>Number:         697
>Category:       kern
>Synopsis:       there doesn't appear to be any disk buffering
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people (Kernel Bug People)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Jan  4 13:50:03 1995
>Originator:     David Muir Sharnoff
>Organization:
Idiom Consulting
>Release:        1.0
>Environment:

System: NetBSD again.idiom.com 1.0 NetBSD 1.0 (AGAIN) #5: Wed Jan 4 12:34:19 PST 1995 root@again.idiom.com:/home/src/sys/arch/i386/compile/AGAIN i386

>Description:
	
	I noticed that when I copy a new kernel to /, my disk seeks madly.

	I got to thinking...  why would it be seeking?  You only hear seeking
	if the head is moving some distance.
	
		1.  I just linked it, it should still be in the buffer cache 
		    and thus it shouldn't need to read it.

		2.  For writing, it should buffer the write until a sync 
		    occurs.  Thus it shouldn't need to write anything.

		3.  The kernel copy operation should be almost instant.

	Instead, it takes more than a second and the disk is seeking
	wildly throughout.

	I noticed other oddities too: the mailer I use writes a tmp file
	when you quit.   It takes longer to do this on my 90Mhz pentium
	with 7200rpm disk than it does on my 25Mhz sparc (SunOS) with
	3600rpm disk.

	This isn't because I'm out of buffers.  I thought the number the
	system gave me by default seemed a little low, so I added some:
	"using 1024 buffers containing 6291456 bytes of memory."
	Didn't help.

>How-To-Repeat:
	
>Fix:
>Audit-Trail:
>Unformatted: