[lkml]   [2000]   [Nov]   [19]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH] Large "clipped" IDE disk support for 2.4 when using old BIOS
On Sun, Nov 19, 2000 at 04:30:34AM +0900, Taisuke Yamada wrote:

> Earlier this month, I had sent in a patch to 2.2.18pre17 (with
> IDE-patch from applied) to add support
> for IDE disk larger than 32GB, even if the disk required "clipping"
> to reduce apparent disk size due to BIOS limitation.
> BIOS known to have this limitation is Award 4.51 (and before) and
> it seems many mainboards with not-so-great vendor support still use it.
> Now I'm moving to 2.4-based system, and so ported the patch to
> 2.4-test10. It also applies cleanly to 2.4-test11.
> With this patch, you will be able to use disk capacity above
> 32GB (or 2GB/8GB depending on how clipping take effect), and
> still be able to boot off from the disk because you can leave
> the "clipping" turned on.

Hi Taisuke,

I suppose you know that no kernel patch is required
(since setmax.c does the same from user space).
Did you try setmax?
I would like to see the results - am still in the information
gathering stage - I have two largish Maxtor disks myself, one
40 GB and one 60 GB, and their behaviour is different, so
to me it seems a bit too early to come with kernel patches.

I think I already sent you setmax.c, but in case my memory
is confused let me include it here again. This is for 2.4.


/* setmax.c - aeb, 000326 */
#include <stdio.h>
#include <fcntl.h>
#include <getopt.h>
#include <linux/hdreg.h>

#define HDIO_DRIVE_CMD_AEB 0x031e

#define CHECK_POWER_MODE 0xe5
#define SET_MAX 0xf9

#define LBA 0x40
#define VV 1 /* if set in sectorct then NOT volatile */

struct idecmdin {
unsigned char cmd;
unsigned char feature;
unsigned char nsect;
unsigned char sect, lcyl, hcyl;
unsigned char select;

struct idecmdout {
unsigned char status;
unsigned char error;
unsigned char nsect;
unsigned char sect, lcyl, hcyl;
unsigned char select;

unsigned int
tolba(unsigned char *args) {
return ((args[6] & 0xf) << 24) + (args[5] << 16) + (args[4] << 8) + args[3];

fromlba(unsigned char *args, unsigned int lba) {
args[3] = (lba & 0xff);
lba >>= 8;
args[4] = (lba & 0xff);
lba >>= 8;
args[5] = (lba & 0xff);
lba >>= 8;
args[6] = (args[6] & 0xf0) | (lba & 0xf);

get_identity(int fd) {
unsigned char args[4+512] = {WIN_IDENTIFY,0,0,1,};
struct hd_driveid *id = (struct hd_driveid *)&args[4];

if (ioctl(fd, HDIO_DRIVE_CMD, &args)) {
"WIN_IDENTIFY failed - trying WIN_PIDENTIFY\n");
args[0] = WIN_PIDENTIFY;
if (ioctl(fd, HDIO_DRIVE_CMD, &args)) {
"WIN_PIDENTIFY also failed - giving up\n");

printf("lba capacity: %d sectors (%lld bytes)\n",
(long long) id->lba_capacity * 512);

* result: in LBA mode precisely what is expected
* in CHS mode the correct H and S, and C mod 65536.
unsigned int
get_native_max(int fd, int slave) {
unsigned char args[7];
int i, max;

for (i=0; i<7; i++)
args[i] = 0;
args[6] = (slave ? 0x10 : 0) | LBA;
if (ioctl(fd, HDIO_DRIVE_CMD_AEB, &args)) {
for (i=0; i<7; i++)
printf("%d = 0x%x\n", args[i], args[i]);
return tolba(args);

* SET_MAX_ADDRESS requires immediately preceding READ_NATIVE_MAX_ADDRESS
* Results: this fails for delta <= 254, succeeds for delta >= 255.
* So, in order to get the last 255*512=130560 bytes back one has to reboot.
* Side effect: reset to CurCHS=16383/16/63, CurSects=16514064.
set_max_address(int fd, int slave, int delta) {
unsigned char args[7];
int i, nativemax;

nativemax = get_native_max(fd, slave);
printf("nativemax=%d (0x%x)\n", nativemax, nativemax);

for (i=0; i<7; i++)
args[i] = 0;
args[0] = SET_MAX;
args[1] = 0;
fromlba(args, nativemax-delta);
args[6] |= LBA;

if (ioctl(fd, HDIO_DRIVE_CMD_AEB, &args)) {
perror("HDIO_DRIVE_CMD_AEB failed SET_MAX");
for (i=0; i<7; i++)
printf("%d = 0x%x\n", args[i], args[i]);

static char short_opts[] = "d:";
static const struct option long_opts[] = {
{ "delta", required_argument, NULL, 'd' },
{ NULL, 0, NULL, 0 }

static char *usage_txt =
"Call: setmax [-d D] DEVICE\n"
"The call \"setmax --delta D DEVICE\" will do a SET_MAX command\n"
"to set the maximum accessible sector number D sectors\n"
"below end-of-disk.\n"
"The call \"setmax DEVICE\" will do a READ_NATIVE_MAX_ADDRESS\n"
"command, and report the maximum accessible sector number.\n"
"This is IDE-only. Probably DEVICE is /dev/hdx for some x.\n\n";

main(int argc, char **argv){
int fd, c;
int delta;

/* If you modify device, also update slave, if necessary. */
/* master: hda, hdc, hde; slave: hdb, hdd, hdf */
char *device = NULL; /* e.g. "/dev/hda" */
int slave = 0;

delta = -1;
while ((c = getopt_long (argc, argv, short_opts, long_opts, NULL)) != -1) {
switch(c) {
case 'd':
delta = atoi(optarg);
case '?':
fprintf(stderr, "unknown option\n");
fprintf(stderr, usage_txt);

if (optind < argc)
device = argv[optind];
if (!device) {
fprintf(stderr, "no device specified - "
"use e.g. \"setmax /dev/hdb\"\n");
fprintf(stderr, usage_txt);
printf("Using device %s\n", device);

fd = open(device, O_RDONLY);
if (fd == -1) {

if (delta != -1) {
printf("setting delta=%d\n", delta);
set_max_address(fd, slave, delta);
} else {
int mad = get_native_max(fd, slave);
long long bytes = (long long) (mad+1) * 512;
int hMB = (bytes+50000000)/100000000;

printf("native max address: %d\n", mad);
printf("that is %lld bytes, %d.%d GB\n",
bytes, hMB/10, hMB%10);

return 0;

features, sectorct, sectornr, cyllow, cylhi: NA
device/head: 8 bits: obs LBA obs DEV na na na na
command: F8
error: 8 bits: na na na na na ABRT na na
sectorct: NA
sectornr, cyllo, cylhi: native max address
device/head: 8 bits: obs na obs DEV; 4bits native max address
status: BSY DRDY DF na DRQ na na ERR [010x0xx0]

SET_MAX_ADDRESS (obsolete?):
Error: ABRT=4: command not supported, max requested exceeds
device capacity, CYL > 16383, command not preceded by
Also error if the device is in Set_Max_Locked or Set_Max_Frozen state.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
Please read the FAQ at

 \ /
  Last update: 2005-03-22 12:47    [W:0.124 / U:2.240 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site