Out of inodes

[Paul Campbell]

In article <11317 at dasys1.UUCP> jsb at dasys1.UUCP (Lothar of the Hill People) writes:
>Remember the inode bug discussed on the net a while ago?  A file system
>bug interacted with news to cause inodes to get lost?  Well, my A/UX
>system seems to be so aflicted.  Is there a fix?  Other than running

I worked out a bugfix a while ago and gave it a quick test, it seemed
to work. I then went off on a tangent completely rewritting the system
V inode code to keep a FS inode bitmap table instead which I never fininshed
all of which goes to say that I lost the patch and don't use it from day
to day. A quick check of that part of the kernel with adb makes me think
that the following is probably it (but I haven't tested it so try it on a
copy of your kernel and maybe on a disk you don't want to trash). The
main difference between this code and the current is that it slows down
the inode scavenger so that it searches the whole disk from the beginning
(ie back to system III before AT&T put the bug in in the first place). This
does not solve the other inode hole that sometimes causes just 1 inode to be
lost - also an AT&T bug). 

For A/UX 1.0 (1.1 may be different check it out!) use adb -k /unix
and change:

	ialloc+0x142:	addq.l	#1, d0

		to	moveq	#2, d0

>fsck -s regularly to reclaim the inodes?  I'm also running out of
>file table space.  Is this a sign of the same bug?

No - it means just that, you're out of file table space, use kconfig
to bump it .... 

I hope this helps some


	Paul
-- 
Paul Campbell    UUCP: ..!mtxinu!taniwha!paul     AppleLink: CAMPBELL.P
"We got a thousand points of light for the homeless man,
Got a kinder, gentler, machine gun hand ..." - Neil Young 'Freedom'