> It might be better to use one kbyte from PARAM+1024 as it's guaranteed
> to be free (see Documentation/i386/zero-page.txt).
Huh ?
>From Documentation/i386/zero-page.txt :
0x224 unsigned short setup.S heap end pointer
0x226 - 0x7ff setup.S code.
0x800 string, 256bytes COMMAND_LINE, the kernel commandline as
copied using CL_OFFSET.
Note: this will be copied once more by setup.c
into a local buffer which is only 256 bytes long.
( #define COMMAND_LINE_SIZE 256 )
---------------
(my patched version, of course, but that applies only to the length of
whatever is at 0x800)
I actually tried the place at 0x400 -- it did froze at boot (wonder why
:-))
(actually, I tried first, at random, 0x400. First I retrieved only one
entry, was happy. Then tried more. Froze everything. Was puzzled until I
found out about Doc/zero-page ;-) )
I also tried in a subsequent version of my page to use memory at 0xc00,
but stepped back, as strange problems struck me (it *might* be that egcs
compiles :
#define TOTO ((PARAM + 2048 + 1024 + 4))
...
stuff[TOTO] = foo;
differently (correct) than when TOTO is defined to be (PARAM + 3076).
Or perhaps (more probable) that there's something else, undocumented, in
the last bits of the zero page.
I'll investigate that a bit later, if no one does before me. That QSAM
patch is not really necessary for 2.2 (unless it is felt that systems with
a RAM hole at 14-16 Mb or whatever else need something to fix -- in which
case my patch, with its zero-page mysteries solved, fits perfectly the
bill).
-- Cyrille
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/