summaryrefslogtreecommitdiff
path: root/recipes/lilo-sh/files/lilo-0.21-second.patch
blob: 12a9af04c7fc09228ce552b3e0552ebc96a04b47 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
I have a problem to report with lilo, and the fix for it. I am not sure
whether this is the right place for it though, I am also CCing the lilo
maintainer Werner Almesberger.

When lilo gets control from the bios (whether lilo is on disk, or on the
floppy), it is checking the wrong location when it tries to parse the
parameter line construction area to determine whether it needs to clean
up extra blank spaces. This is causing lilo to eat up the last character(s)
of the command line that it passes to the kernel.

Specifically, the code

cpcodn:	cmp	byte ptr (di-1),#32 ! last was space ?

is using the memory contents of ds:di-1 for the compare. It should
really be using es:di-1. I checked, ds=0x0b00 around here, so the
code ended up checking 0x0b00:(di-1) instead of 0x9000:(di-1). It
is quite possible that the memory that lilo checks has garbage, 
causing it to make wrong decision. A "seg es" just before the cmp 
fixes the problem. I am attaching the patch below.

Kanoj

--- second.0	Mon Sep 20 13:52:01 1999
+++ second.S	Tue Sep 28 10:20:54 1999
@@ -816,7 +816,8 @@
 	je	cpcodsp		! yes -> discard next
 	lodsb			! get next byte
 	jmp	cpcolp
-cpcodn:	cmp	byte ptr (di-1),#32 ! last was space ?
+cpcodn:	seg	es
+	cmp	byte ptr (di-1),#32 ! last was space ?
 	jne	nocopt		! no -> go on
 	dec	di		! discard it
 nocopt:	mov	si,options	! append variable options