How to debug 'out of disk' error.

How to debug 'out of disk' error.

[Lennart Sorensen]

I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
of disk' error when trying to boot from the HD after doing an install.
It ends up at a grub rescue prompt, given it can't even read the disk
well enough to load the full grub.  Doing ls shows the disks, and I can
show the root directory of a partition, but anything further seems to
hit the same 'out of disk' error.

The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
(the new one) with a phoenix bios.  So far Compulab's response when told
it won't boot grub2 was "We know, but syslinux works fine.", which to
me is not a solution at all.  The same system also hangs the Linux 2.6.32
kernel unless 'edd=off' is added to the kernel command line.

So any suggestions on how to go about debuagging this?  Or even any
ideas why grub would give such an error.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

On 16.03.2011 16:13, Lennart Sorensen wrote:
> I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
> of disk' error when trying to boot from the HD after doing an install.
> It ends up at a grub rescue prompt, given it can't even read the disk
> well enough to load the full grub.  Doing ls shows the disks, and I can
> show the root directory of a partition, but anything further seems to
> hit the same 'out of disk' error.
>
>   
ls -l should show how big it thinks the disk is. Compare it with its
real size
> The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
> (the new one) with a phoenix bios.  So far Compulab's response when told
> it won't boot grub2 was "We know, but syslinux works fine.", which to
> me is not a solution at all.  The same system also hangs the Linux 2.6.32
> kernel unless 'edd=off' is added to the kernel command line.
>
>   
Looks like int13 is borked. Perhaps it's borked in a way which prevents
GRUB from detecting 13/42 function. You can try to force int13/42 by:
=== modified file 'grub-core/disk/i386/pc/biosdisk.c'
--- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
+++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 15:28:26 +0000
@@ -145,6 +145,8 @@
 {
   struct grub_bios_int_registers regs;
 
+  return 1;
+
   regs.edx = drive & 0xff;
   regs.eax = 0x4100;
   regs.ebx = 0x55aa;


> So any suggestions on how to go about debuagging this?  Or even any
> ideas why grub would give such an error.
>
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Mar 16, 2011 at 04:29:32PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> On 16.03.2011 16:13, Lennart Sorensen wrote:
> > I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
> > of disk' error when trying to boot from the HD after doing an install.
> > It ends up at a grub rescue prompt, given it can't even read the disk
> > well enough to load the full grub.  Doing ls shows the disks, and I can
> > show the root directory of a partition, but anything further seems to
> > hit the same 'out of disk' error.
> >
> >   
> ls -l should show how big it thinks the disk is. Compare it with its
> real size
> > The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
> > (the new one) with a phoenix bios.  So far Compulab's response when told
> > it won't boot grub2 was "We know, but syslinux works fine.", which to
> > me is not a solution at all.  The same system also hangs the Linux 2.6.32
> > kernel unless 'edd=off' is added to the kernel command line.
> >
> >   
> Looks like int13 is borked. Perhaps it's borked in a way which prevents
> GRUB from detecting 13/42 function. You can try to force int13/42 by:
> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 15:28:26 +0000
> @@ -145,6 +145,8 @@
>  {
>    struct grub_bios_int_registers regs;
>  
> +  return 1;
> +
>    regs.edx = drive & 0xff;
>    regs.eax = 0x4100;
>    regs.ebx = 0x55aa;

OK, the code in 1.98(plus some git stuff) that is in debian uses assembly
code for that part.  I changed the place calling it instead and added
some debug info.

When running from linux, the diskinfo bios call returns a total_sectors of
4352 sectors, for an 80G drive.  That's just messed up.  I haven't tried
booting it yet to see if it returns the same value then, but I suspect
it will.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 2512 bytes --]

On 16.03.2011 17:58, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 04:29:32PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>   
>> On 16.03.2011 16:13, Lennart Sorensen wrote:
>>     
>>> I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
>>> of disk' error when trying to boot from the HD after doing an install.
>>> It ends up at a grub rescue prompt, given it can't even read the disk
>>> well enough to load the full grub.  Doing ls shows the disks, and I can
>>> show the root directory of a partition, but anything further seems to
>>> hit the same 'out of disk' error.
>>>
>>>   
>>>       
>> ls -l should show how big it thinks the disk is. Compare it with its
>> real size
>>     
>>> The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
>>> (the new one) with a phoenix bios.  So far Compulab's response when told
>>> it won't boot grub2 was "We know, but syslinux works fine.", which to
>>> me is not a solution at all.  The same system also hangs the Linux 2.6.32
>>> kernel unless 'edd=off' is added to the kernel command line.
>>>
>>>   
>>>       
>> Looks like int13 is borked. Perhaps it's borked in a way which prevents
>> GRUB from detecting 13/42 function. You can try to force int13/42 by:
>> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
>> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
>> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 15:28:26 +0000
>> @@ -145,6 +145,8 @@
>>  {
>>    struct grub_bios_int_registers regs;
>>  
>> +  return 1;
>> +
>>    regs.edx = drive & 0xff;
>>    regs.eax = 0x4100;
>>    regs.ebx = 0x55aa;
>>     
> OK, the code in 1.98(plus some git stuff) that is in debian uses assembly
> code for that part.  I changed the place calling it instead and added
> some debug info.
>
> When running from linux, the diskinfo bios call returns a total_sectors of
> 4352 sectors, for an 80G drive.  That's just messed up.  I haven't tried
> booting it yet to see if it returns the same value then, but I suspect
> it will.
>
>   
Following change would make GRUB ignore the total_sectors (as an
investigation measure):
@@ -398,7 +400,7 @@
         total_sectors = data->cylinders * data->heads * data->sectors;
     }
 
-  disk->total_sectors = total_sectors;
+  disk->total_sectors = GRUB_DISK_SIZE_UNKNOWN;
   disk->data = data;
 
   return GRUB_ERR_NONE;




-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Mar 16, 2011 at 04:29:32PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> On 16.03.2011 16:13, Lennart Sorensen wrote:
> > I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
> > of disk' error when trying to boot from the HD after doing an install.
> > It ends up at a grub rescue prompt, given it can't even read the disk
> > well enough to load the full grub.  Doing ls shows the disks, and I can
> > show the root directory of a partition, but anything further seems to
> > hit the same 'out of disk' error.
> >
> >   
> ls -l should show how big it thinks the disk is. Compare it with its
> real size
> > The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
> > (the new one) with a phoenix bios.  So far Compulab's response when told
> > it won't boot grub2 was "We know, but syslinux works fine.", which to
> > me is not a solution at all.  The same system also hangs the Linux 2.6.32
> > kernel unless 'edd=off' is added to the kernel command line.
> >
> >   
> Looks like int13 is borked. Perhaps it's borked in a way which prevents
> GRUB from detecting 13/42 function. You can try to force int13/42 by:
> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 15:28:26 +0000
> @@ -145,6 +145,8 @@
>  {
>    struct grub_bios_int_registers regs;
>  
> +  return 1;
> +
>    regs.edx = drive & 0xff;
>    regs.eax = 0x4100;
>    regs.ebx = 0x55aa;

So I now have some debug info:

version=48
total_sectors=156312576 C=16383 H=16 S=63
error: hd0,msdos1 out of disk.
Entering rescue mode...
grub rescue>

So version is returning something, and it is actually getting the disk
size properly.  I wonder what else could cause the out of disk message
then.

-- 
len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 2131 bytes --]

On 16.03.2011 18:32, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 04:29:32PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>   
>> On 16.03.2011 16:13, Lennart Sorensen wrote:
>>     
>>> I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
>>> of disk' error when trying to boot from the HD after doing an install.
>>> It ends up at a grub rescue prompt, given it can't even read the disk
>>> well enough to load the full grub.  Doing ls shows the disks, and I can
>>> show the root directory of a partition, but anything further seems to
>>> hit the same 'out of disk' error.
>>>
>>>   
>>>       
>> ls -l should show how big it thinks the disk is. Compare it with its
>> real size
>>     
>>> The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
>>> (the new one) with a phoenix bios.  So far Compulab's response when told
>>> it won't boot grub2 was "We know, but syslinux works fine.", which to
>>> me is not a solution at all.  The same system also hangs the Linux 2.6.32
>>> kernel unless 'edd=off' is added to the kernel command line.
>>>
>>>   
>>>       
>> Looks like int13 is borked. Perhaps it's borked in a way which prevents
>> GRUB from detecting 13/42 function. You can try to force int13/42 by:
>> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
>> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
>> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 15:28:26 +0000
>> @@ -145,6 +145,8 @@
>>  {
>>    struct grub_bios_int_registers regs;
>>  
>> +  return 1;
>> +
>>    regs.edx = drive & 0xff;
>>    regs.eax = 0x4100;
>>    regs.ebx = 0x55aa;
>>     
> So I now have some debug info:
>
> version=48
> total_sectors=156312576 C=16383 H=16 S=63
> error: hd0,msdos1 out of disk.
> Entering rescue mode...
> grub rescue>
>
> So version is returning something, and it is actually getting the disk
> size properly.  I wonder what else could cause the out of disk message
> then.
>
>   
Could you post the debug code you added?


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

[-- Attachment #1: Type: text/plain, Size: 773 bytes --]

On Wed, Mar 16, 2011 at 06:58:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> On 16.03.2011 18:32, Lennart Sorensen wrote:
> > So I now have some debug info:
> >
> > version=48
> > total_sectors=156312576 C=16383 H=16 S=63
> > error: hd0,msdos1 out of disk.
> > Entering rescue mode...
> > grub rescue>
> >
> > So version is returning something, and it is actually getting the disk
> > size properly.  I wonder what else could cause the out of disk message
> > then.
> >
> >   
> Could you post the debug code you added?

I will try.  I am currently having an argument with the quilt patch
system debian's grub package is using.

Here is the modified biosdisk.c I am using (slightly changed since
the above was done).

-- 
Len Sorensen

[-- Attachment #2: biosdisk.c --]
[-- Type: text/x-csrc, Size: 11233 bytes --]

/*
 *  GRUB  --  GRand Unified Bootloader
 *  Copyright (C) 1999,2000,2001,2002,2003,2004,2005,2006,2007,2008,2009,2010  Free Software Foundation, Inc.
 *
 *  GRUB is free software: you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation, either version 3 of the License, or
 *  (at your option) any later version.
 *
 *  GRUB is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with GRUB.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <grub/machine/biosdisk.h>
#include <grub/machine/memory.h>
#include <grub/machine/kernel.h>
#include <grub/disk.h>
#include <grub/dl.h>
#include <grub/mm.h>
#include <grub/types.h>
#include <grub/misc.h>
#include <grub/err.h>
#include <grub/term.h>

static int cd_drive = 0;

static int
grub_biosdisk_get_drive (const char *name)
{
  unsigned long drive;

  if ((name[0] != 'f' && name[0] != 'h') || name[1] != 'd')
    goto fail;

  drive = grub_strtoul (name + 2, 0, 10);
  if (grub_errno != GRUB_ERR_NONE)
    goto fail;

  if (name[0] == 'h')
    drive += 0x80;

  return (int) drive ;

 fail:
  grub_error (GRUB_ERR_UNKNOWN_DEVICE, "not a biosdisk");
  return -1;
}

static int
grub_biosdisk_call_hook (int (*hook) (const char *name), int drive)
{
  char name[10];

  grub_snprintf (name, sizeof (name),
		 (drive & 0x80) ? "hd%d" : "fd%d", drive & (~0x80));
  return hook (name);
}

static int
grub_biosdisk_iterate (int (*hook) (const char *name))
{
  int drive;
  int num_floppies;

  /* For hard disks, attempt to read the MBR.  */
  for (drive = 0x80; drive < 0x90; drive++)
    {
      if (grub_biosdisk_rw_standard (0x02, drive, 0, 0, 1, 1,
				     GRUB_MEMORY_MACHINE_SCRATCH_SEG) != 0)
	{
	  grub_dprintf ("disk", "Read error when probing drive 0x%2x\n", drive);
	  break;
	}

      if (grub_biosdisk_call_hook (hook, drive))
	return 1;
    }

  if (cd_drive)
    {
      if (grub_biosdisk_call_hook (hook, cd_drive))
      return 1;
    }

  /* For floppy disks, we can get the number safely.  */
  num_floppies = grub_biosdisk_get_num_floppies ();
  for (drive = 0; drive < num_floppies; drive++)
    if (grub_biosdisk_call_hook (hook, drive))
      return 1;

  return 0;
}

static grub_err_t
grub_biosdisk_open (const char *name, grub_disk_t disk)
{
  grub_uint64_t total_sectors = 0;
  int drive;
  struct grub_biosdisk_data *data;

  drive = grub_biosdisk_get_drive (name);
  if (drive < 0)
    return grub_errno;

  disk->has_partitions = cd_drive ? (drive != cd_drive) : 1;
  disk->id = drive;

  data = (struct grub_biosdisk_data *) grub_zalloc (sizeof (*data));
  if (! data)
    return grub_errno;

  data->drive = drive;

  if ((cd_drive) && (drive == cd_drive))
    {
      data->flags = GRUB_BIOSDISK_FLAG_LBA | GRUB_BIOSDISK_FLAG_CDROM;
      data->sectors = 32;
      /* TODO: get the correct size.  */
      total_sectors = GRUB_DISK_SIZE_UNKNOWN;
    }
  else if (drive & 0x80)
    {
      /* HDD */
      int version;

      grub_printf ("Checking int13_extensions.\n");
      version = grub_biosdisk_check_int13_extensions (drive);
      grub_printf ("...version=%d.\n",version);
      if (!version)
	{
	  version = 1; 
	}
      if (version)
	{
	  struct grub_biosdisk_drp *drp
	    = (struct grub_biosdisk_drp *) GRUB_MEMORY_MACHINE_SCRATCH_ADDR;

	  /* Clear out the DRP.  */
	  grub_memset (drp, 0, sizeof (*drp));
	  drp->size = sizeof (*drp);
	  if (! grub_biosdisk_get_diskinfo_int13_extensions (drive, drp))
	    {
	      data->flags = GRUB_BIOSDISK_FLAG_LBA;

	      if (drp->total_sectors)
		{
		  total_sectors = drp->total_sectors;
		  grub_printf ("total_sectors (C=%u H=%u S=%u)=%llu.\n",drp->cylinders,drp->heads,drp->sectors,total_sectors);
		}
	      else
		{
                /* Some buggy BIOSes doesn't return the total sectors
                   correctly but returns zero. So if it is zero, compute
                   it by C/H/S returned by the LBA BIOS call.  */
                  total_sectors = drp->cylinders * drp->heads * drp->sectors;
		  grub_printf ("total_sectors (buggy bios) (C=%u H=%u S=%u)=%llu.\n",drp->cylinders,drp->heads,drp->sectors,total_sectors);
		}
	    }
	}
    }

  if (! (data->flags & GRUB_BIOSDISK_FLAG_CDROM))
    {
      if (grub_biosdisk_get_diskinfo_standard (drive,
					       &data->cylinders,
					       &data->heads,
					       &data->sectors) != 0)
        {
	  if (total_sectors && (data->flags & GRUB_BIOSDISK_FLAG_LBA))
	    {
	      data->sectors = 63;
	      data->heads = 255;
	      data->cylinders
		= grub_divmod64 (total_sectors
				 + data->heads * data->sectors - 1,
				 data->heads * data->sectors, 0);
	    }
	  else
	    {
	      grub_free (data);
	      return grub_error (GRUB_ERR_BAD_DEVICE, "%s cannot get C/H/S values", disk->name);
	    }
        }

      if (! total_sectors)
        total_sectors = data->cylinders * data->heads * data->sectors;
    }

  disk->total_sectors = total_sectors;
  disk->data = data;

  return GRUB_ERR_NONE;
}

static void
grub_biosdisk_close (grub_disk_t disk)
{
  grub_free (disk->data);
}

/* For readability.  */
#define GRUB_BIOSDISK_READ	0
#define GRUB_BIOSDISK_WRITE	1

#define GRUB_BIOSDISK_CDROM_RETRY_COUNT 3

static grub_err_t
grub_biosdisk_rw (int cmd, grub_disk_t disk,
		  grub_disk_addr_t sector, grub_size_t size,
		  unsigned segment)
{
  struct grub_biosdisk_data *data = disk->data;

  if (data->flags & GRUB_BIOSDISK_FLAG_LBA)
    {
      struct grub_biosdisk_dap *dap;

      dap = (struct grub_biosdisk_dap *) (GRUB_MEMORY_MACHINE_SCRATCH_ADDR
					  + (data->sectors
					     << GRUB_DISK_SECTOR_BITS));
      dap->length = sizeof (*dap);
      dap->reserved = 0;
      dap->blocks = size;
      dap->buffer = segment << 16;	/* The format SEGMENT:ADDRESS.  */
      dap->block = sector;

      if (data->flags & GRUB_BIOSDISK_FLAG_CDROM)
        {
	  int i;

	  if (cmd)
	    return grub_error (GRUB_ERR_WRITE_ERROR, "can\'t write to cdrom");

	  dap->blocks = ALIGN_UP (dap->blocks, 4) >> 2;
	  dap->block >>= 2;

	  for (i = 0; i < GRUB_BIOSDISK_CDROM_RETRY_COUNT; i++)
            if (! grub_biosdisk_rw_int13_extensions (0x42, data->drive, dap))
	      break;

	  if (i == GRUB_BIOSDISK_CDROM_RETRY_COUNT)
	    return grub_error (GRUB_ERR_READ_ERROR, "cdrom read error");
	}
      else
        if (grub_biosdisk_rw_int13_extensions (cmd + 0x42, data->drive, dap))
	  {
	    /* Fall back to the CHS mode.  */
	    data->flags &= ~GRUB_BIOSDISK_FLAG_LBA;
	    grub_printf ("Falling back to CHS\n");
	    disk->total_sectors = data->cylinders * data->heads * data->sectors;
	    return grub_biosdisk_rw (cmd, disk, sector, size, segment);
	  }
    }
  else
    {
      unsigned coff, hoff, soff;
      unsigned head;

      grub_printf ("CHS=%d,%d,%d\n",data->cylinders,data->heads,data->sectors);
      /* It is impossible to reach over 8064 MiB (a bit less than LBA24) with
	 the traditional CHS access.  */
      if (sector >
	  1024 /* cylinders */ *
	  256 /* heads */ *
	  63 /* spt */)
	return grub_error (GRUB_ERR_OUT_OF_RANGE, "%s out of disk(1)", disk->name);

      soff = ((grub_uint32_t) sector) % data->sectors + 1;
      head = ((grub_uint32_t) sector) / data->sectors;
      hoff = head % data->heads;
      coff = head / data->heads;

      if (coff >= data->cylinders)
	return grub_error (GRUB_ERR_OUT_OF_RANGE, "%s out of disk(2)", disk->name);

      if (grub_biosdisk_rw_standard (cmd + 0x02, data->drive,
				     coff, hoff, soff, size, segment))
	{
	  switch (cmd)
	    {
	    case GRUB_BIOSDISK_READ:
	      return grub_error (GRUB_ERR_READ_ERROR, "%s read error", disk->name);
	    case GRUB_BIOSDISK_WRITE:
	      return grub_error (GRUB_ERR_WRITE_ERROR, "%s write error", disk->name);
	    }
	}
    }

  return GRUB_ERR_NONE;
}

/* Return the number of sectors which can be read safely at a time.  */
static grub_size_t
get_safe_sectors (grub_disk_addr_t sector, grub_uint32_t sectors)
{
  grub_size_t size;
  grub_uint32_t offset;

  /* OFFSET = SECTOR % SECTORS */
  grub_divmod64 (sector, sectors, &offset);

  size = sectors - offset;

  /* Limit the max to 0x7f because of Phoenix EDD.  */
  if (size > 0x7f)
    size = 0x7f;

  return size;
}

static grub_err_t
grub_biosdisk_read (grub_disk_t disk, grub_disk_addr_t sector,
		    grub_size_t size, char *buf)
{
  struct grub_biosdisk_data *data = disk->data;

  while (size)
    {
      grub_size_t len;
      grub_size_t cdoff = 0;

      len = get_safe_sectors (sector, data->sectors);

      if (data->flags & GRUB_BIOSDISK_FLAG_CDROM)
	{
	  cdoff = (sector & 3) << GRUB_DISK_SECTOR_BITS;
	  len = ALIGN_UP (sector + len, 4) - (sector & ~3);
	  sector &= ~3;
	}

      if (len > size)
	len = size;

      if (grub_biosdisk_rw (GRUB_BIOSDISK_READ, disk, sector, len,
			    GRUB_MEMORY_MACHINE_SCRATCH_SEG))
	return grub_errno;

      grub_memcpy (buf, (void *) (GRUB_MEMORY_MACHINE_SCRATCH_ADDR + cdoff),
		   len << GRUB_DISK_SECTOR_BITS);
      buf += len << GRUB_DISK_SECTOR_BITS;
      sector += len;
      size -= len;
    }

  return grub_errno;
}

static grub_err_t
grub_biosdisk_write (grub_disk_t disk, grub_disk_addr_t sector,
		     grub_size_t size, const char *buf)
{
  struct grub_biosdisk_data *data = disk->data;

  if (data->flags & GRUB_BIOSDISK_FLAG_CDROM)
    return grub_error (GRUB_ERR_IO, "can't write to CDROM");

  while (size)
    {
      grub_size_t len;

      len = get_safe_sectors (sector, data->sectors);
      if (len > size)
	len = size;

      grub_memcpy ((void *) GRUB_MEMORY_MACHINE_SCRATCH_ADDR, buf,
		   len << GRUB_DISK_SECTOR_BITS);

      if (grub_biosdisk_rw (GRUB_BIOSDISK_WRITE, disk, sector, len,
			    GRUB_MEMORY_MACHINE_SCRATCH_SEG))
	return grub_errno;

      buf += len << GRUB_DISK_SECTOR_BITS;
      sector += len;
      size -= len;
    }

  return grub_errno;
}

static struct grub_disk_dev grub_biosdisk_dev =
  {
    .name = "biosdisk",
    .id = GRUB_DISK_DEVICE_BIOSDISK_ID,
    .iterate = grub_biosdisk_iterate,
    .open = grub_biosdisk_open,
    .close = grub_biosdisk_close,
    .read = grub_biosdisk_read,
    .write = grub_biosdisk_write,
    .next = 0
  };

static void
grub_disk_biosdisk_fini (void)
{
  grub_disk_dev_unregister (&grub_biosdisk_dev);
}

GRUB_MOD_INIT(biosdisk)
{
  struct grub_biosdisk_cdrp *cdrp
        = (struct grub_biosdisk_cdrp *) GRUB_MEMORY_MACHINE_SCRATCH_ADDR;

  if (grub_disk_firmware_is_tainted)
    {
      grub_printf ("Firmware is marked as tainted, refusing to initialize.\n");
      return;
    }
  grub_disk_firmware_fini = grub_disk_biosdisk_fini;

  grub_memset (cdrp, 0, sizeof (*cdrp));
  cdrp->size = sizeof (*cdrp);
  cdrp->media_type = 0xFF;
  if ((! grub_biosdisk_get_cdinfo_int13_extensions (grub_boot_drive, cdrp)) &&
      ((cdrp->media_type & GRUB_BIOSDISK_CDTYPE_MASK)
       == GRUB_BIOSDISK_CDTYPE_NO_EMUL))
    cd_drive = cdrp->drive_no;

  grub_disk_dev_register (&grub_biosdisk_dev);
}

GRUB_MOD_FINI(biosdisk)
{
  grub_disk_biosdisk_fini ();
}

Re: How to debug 'out of disk' error.

[Phillip Susi]

On 3/16/2011 1:32 PM, Lennart Sorensen wrote:
> So I now have some debug info:
> 
> version=48
> total_sectors=156312576 C=16383 H=16 S=63
> error: hd0,msdos1 out of disk.
> Entering rescue mode...
> grub rescue>
> 
> So version is returning something, and it is actually getting the disk
> size properly.  I wonder what else could cause the out of disk message
> then.

What does the partition table look like?  It doesn't extend beyond
sector 156312575 does it?

Also I wonder about why H=16.  It should be 255.

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Mar 16, 2011 at 03:19:48PM -0400, Phillip Susi wrote:
> On 3/16/2011 1:32 PM, Lennart Sorensen wrote:
> > So I now have some debug info:
> > 
> > version=48
> > total_sectors=156312576 C=16383 H=16 S=63
> > error: hd0,msdos1 out of disk.
> > Entering rescue mode...
> > grub rescue>
> > 
> > So version is returning something, and it is actually getting the disk
> > size properly.  I wonder what else could cause the out of disk message
> > then.
> 
> What does the partition table look like?  It doesn't extend beyond
> sector 156312575 does it?
> 
> Also I wonder about why H=16.  It should be 255.

Well usually it should, but not always.

Here is the partition table:

Disk /dev/sda: 80.0 GB, 80032038912 bytes
255 heads, 63 sectors/track, 9730 cylinders, total 156312576 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0004a073

   Device Boot      Start         End      Blocks   Id  System
/dev/sda1   *        2048   152350719    76174336   83  Linux
/dev/sda2       152352766   156311551     1979393    5  Extended
/dev/sda5       152352768   156311551     1979392   82  Linux swap / Solaris

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Mar 16, 2011 at 03:23:58PM -0400, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 03:19:48PM -0400, Phillip Susi wrote:
> > On 3/16/2011 1:32 PM, Lennart Sorensen wrote:
> > > So I now have some debug info:
> > > 
> > > version=48
> > > total_sectors=156312576 C=16383 H=16 S=63
> > > error: hd0,msdos1 out of disk.
> > > Entering rescue mode...
> > > grub rescue>
> > > 
> > > So version is returning something, and it is actually getting the disk
> > > size properly.  I wonder what else could cause the out of disk message
> > > then.
> > 
> > What does the partition table look like?  It doesn't extend beyond
> > sector 156312575 does it?
> > 
> > Also I wonder about why H=16.  It should be 255.
> 
> Well usually it should, but not always.
> 
> Here is the partition table:
> 
> Disk /dev/sda: 80.0 GB, 80032038912 bytes
> 255 heads, 63 sectors/track, 9730 cylinders, total 156312576 sectors
> Units = sectors of 1 * 512 = 512 bytes
> Sector size (logical/physical): 512 bytes / 512 bytes
> I/O size (minimum/optimal): 512 bytes / 512 bytes
> Disk identifier: 0x0004a073
> 
>    Device Boot      Start         End      Blocks   Id  System
> /dev/sda1   *        2048   152350719    76174336   83  Linux
> /dev/sda2       152352766   156311551     1979393    5  Extended
> /dev/sda5       152352768   156311551     1979392   82  Linux swap / Solaris

I have now determined that the call to grub_biosdisk_rw_int13_extensions
fails here:

        if (grub_biosdisk_rw_int13_extensions (cmd + 0x42, data->drive, dap))
          {
            /* Fall back to the CHS mode.  */
            data->flags &= ~GRUB_BIOSDISK_FLAG_LBA;
            disk->total_sectors = data->cylinders * data->heads * data->sectors;
            return grub_biosdisk_rw (cmd, disk, sector, size, segment);
          }

Then LBA gets disabled, and access beyond CHS range is gone, and the
error about 'out of disk' happens.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 2582 bytes --]

On 16.03.2011 21:09, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 03:23:58PM -0400, Lennart Sorensen wrote:
>   
>> On Wed, Mar 16, 2011 at 03:19:48PM -0400, Phillip Susi wrote:
>>     
>>> On 3/16/2011 1:32 PM, Lennart Sorensen wrote:
>>>       
>>>> So I now have some debug info:
>>>>
>>>> version=48
>>>> total_sectors=156312576 C=16383 H=16 S=63
>>>> error: hd0,msdos1 out of disk.
>>>> Entering rescue mode...
>>>> grub rescue>
>>>>
>>>> So version is returning something, and it is actually getting the disk
>>>> size properly.  I wonder what else could cause the out of disk message
>>>> then.
>>>>         
>>> What does the partition table look like?  It doesn't extend beyond
>>> sector 156312575 does it?
>>>
>>> Also I wonder about why H=16.  It should be 255.
>>>       
>> Well usually it should, but not always.
>>
>> Here is the partition table:
>>
>> Disk /dev/sda: 80.0 GB, 80032038912 bytes
>> 255 heads, 63 sectors/track, 9730 cylinders, total 156312576 sectors
>> Units = sectors of 1 * 512 = 512 bytes
>> Sector size (logical/physical): 512 bytes / 512 bytes
>> I/O size (minimum/optimal): 512 bytes / 512 bytes
>> Disk identifier: 0x0004a073
>>
>>    Device Boot      Start         End      Blocks   Id  System
>> /dev/sda1   *        2048   152350719    76174336   83  Linux
>> /dev/sda2       152352766   156311551     1979393    5  Extended
>> /dev/sda5       152352768   156311551     1979392   82  Linux swap / Solaris
>>     
> I have now determined that the call to grub_biosdisk_rw_int13_extensions
> fails here:
>
>         if (grub_biosdisk_rw_int13_extensions (cmd + 0x42, data->drive, dap))
>           {
>             /* Fall back to the CHS mode.  */
>             data->flags &= ~GRUB_BIOSDISK_FLAG_LBA;
>             disk->total_sectors = data->cylinders * data->heads * data->sectors;
>             return grub_biosdisk_rw (cmd, disk, sector, size, segment);
>           }
>
> Then LBA gets disabled, and access beyond CHS range is gone, and the
> error about 'out of disk' happens.
>
>   
Following may help:
=== modified file 'grub-core/disk/i386/pc/biosdisk.c'
--- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
+++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 20:17:57 +0000
@@ -80,6 +80,9 @@
   regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
 
   grub_bios_interrupt (0x13, &regs);
+  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
+    return 0;
+
   return (regs.eax >> 8) & 0xff;
 }
 


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

[-- Attachment #1: Type: text/plain, Size: 1138 bytes --]

On Wed, Mar 16, 2011 at 09:19:12PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> Following may help:
> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 20:17:57 +0000
> @@ -80,6 +80,9 @@
>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
>  
>    grub_bios_interrupt (0x13, &regs);
> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> +    return 0;
> +
>    return (regs.eax >> 8) & 0xff;
>  }

I now got this:

GRUB loading.
Welcome to GRUB!

int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
int13_ext fixup due to CARRY flag
Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
error: hd0 out of disk.
Entering rescue mode...
grub rescue>

I am now using 1.99-rc1.  Here is the biosdisk.c with my changes.

-- 
Len Sorensen

[-- Attachment #2: biosdisk.c --]
[-- Type: text/x-csrc, Size: 17070 bytes --]

/*
 *  GRUB  --  GRand Unified Bootloader
 *  Copyright (C) 1999,2000,2001,2002,2003,2004,2005,2006,2007,2008,2009,2010  Free Software Foundation, Inc.
 *
 *  GRUB is free software: you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation, either version 3 of the License, or
 *  (at your option) any later version.
 *
 *  GRUB is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with GRUB.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <grub/machine/biosdisk.h>
#include <grub/machine/kernel.h>
#include <grub/machine/memory.h>
#include <grub/machine/int.h>
#include <grub/disk.h>
#include <grub/dl.h>
#include <grub/mm.h>
#include <grub/types.h>
#include <grub/misc.h>
#include <grub/err.h>
#include <grub/term.h>

static int cd_drive = 0;
static int grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap);

static int grub_biosdisk_get_num_floppies (void)
{
  struct grub_bios_int_registers regs;
  int drive;

  /* reset the disk system first */
  regs.eax = 0;
  regs.edx = 0;
  regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;

  grub_bios_interrupt (0x13, &regs);

  for (drive = 0; drive < 2; drive++)
    {
      regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT | GRUB_CPU_INT_FLAGS_CARRY;
      regs.edx = drive;

      /* call GET DISK TYPE */
      regs.eax = 0x1500;
      grub_bios_interrupt (0x13, &regs);
      if (regs.flags & GRUB_CPU_INT_FLAGS_CARRY)
	break;

      /* check if this drive exists */
      if (!(regs.eax & 0x300))
	break;
    }

  return drive;
}

/*
 *   Call IBM/MS INT13 Extensions (int 13 %ah=AH) for DRIVE. DAP
 *   is passed for disk address packet. If an error occurs, return
 *   non-zero, otherwise zero.
 */

static int 
grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
{
  int ret;
  struct grub_bios_int_registers regs;
  regs.eax = ah << 8;
  /* compute the address of disk_address_packet */
  regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
  regs.esi = (((grub_addr_t) dap) & 0xffff);
  regs.edx = drive;
  regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;

  grub_bios_interrupt (0x13, &regs);
  ret = (regs.eax >> 8) & 0xff;
  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
    {
      grub_printf ("int13_ext fixup due to CARRY flag\n");
      ret = 0;
    }
  if (ret) 
    {
      grub_printf ("Failed int13_ext call: ah=%x, drive=%x, dap=%8x returned eax=%x\n", ah, drive, (grub_addr_t) dap, regs.eax);
    }
  //else
    //{
      //grub_printf ("Successful int13_ext call: ah=%x, drive=%x, dap=%8x returned eax=%x\n", ah, drive, (grub_addr_t) dap, regs.eax);
    //}

  return ret;
}

/*
 *   Call standard and old INT13 (int 13 %ah=AH) for DRIVE. Read/write
 *   NSEC sectors from COFF/HOFF/SOFF into SEGMENT. If an error occurs,
 *   return non-zero, otherwise zero.
 */
static int 
grub_biosdisk_rw_standard (int ah, int drive, int coff, int hoff,
			   int soff, int nsec, int segment)
{
  int ret, i;

  /* Try 3 times.  */
  for (i = 0; i < 3; i++)
    {
      struct grub_bios_int_registers regs;

      /* set up CHS information */
      /* set %ch to low eight bits of cylinder */
      regs.ecx = (coff << 8) & 0xff00;
      /* set bits 6-7 of %cl to high two bits of cylinder */
      regs.ecx |= (coff >> 2) & 0xc0;
      /* set bits 0-5 of %cl to sector */
      regs.ecx |= soff & 0x3f;

      /* set %dh to head and %dl to drive */  
      regs.edx = (drive & 0xff) | ((hoff << 8) & 0xff00);
      /* set %ah to AH */
      regs.eax = (ah << 8) & 0xff00;
      /* set %al to NSEC */
      regs.eax |= nsec & 0xff;

      regs.ebx = 0;
      regs.es = segment;

      regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;

      grub_bios_interrupt (0x13, &regs);
      /* check if successful */
      if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
	return 0;

      /* save return value */
      ret = regs.eax >> 8;

      /* if fail, reset the disk system */
      regs.eax = 0;
      regs.edx = (drive & 0xff);
      regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
      grub_bios_interrupt (0x13, &regs);
    }
  return ret;
}

/*
 *   Check if LBA is supported for DRIVE. If it is supported, then return
 *   the major version of extensions, otherwise zero.
 */
static int
grub_biosdisk_check_int13_extensions (int drive)
{
  struct grub_bios_int_registers regs;

  regs.edx = drive & 0xff;
  regs.eax = 0x4100;
  regs.ebx = 0x55aa;
  regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
  grub_bios_interrupt (0x13, &regs);
  
  if (regs.flags & GRUB_CPU_INT_FLAGS_CARRY)
    return 0;

  if ((regs.ebx & 0xffff) != 0xaa55)
    return 0;

  /* check if AH=0x42 is supported */
  if (!(regs.ecx & 1))
    return 0;

  return (regs.eax >> 8) & 0xff;
}

/*
 *   Return the geometry of DRIVE in CYLINDERS, HEADS and SECTORS. If an
 *   error occurs, then return non-zero, otherwise zero.
 */
static int 
grub_biosdisk_get_diskinfo_standard (int drive,
				     unsigned long *cylinders,
				     unsigned long *heads,
				     unsigned long *sectors)
{
  struct grub_bios_int_registers regs;

  regs.eax = 0x0800;
  regs.edx = drive & 0xff;

  regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
  grub_bios_interrupt (0x13, &regs);

  /* Check if unsuccessful. Ignore return value if carry isn't set to 
     workaround some buggy BIOSes. */
  if ((regs.flags & GRUB_CPU_INT_FLAGS_CARRY) && ((regs.eax & 0xff00) != 0))
    return (regs.eax & 0xff00) >> 8;

  /* bogus BIOSes may not return an error number */  
  /* 0 sectors means no disk */
  if (!(regs.ecx & 0x3f))
    /* XXX 0x60 is one of the unused error numbers */
    return 0x60;

  /* the number of heads is counted from zero */
  *heads = ((regs.edx >> 8) & 0xff) + 1;
  *cylinders = (((regs.ecx >> 8) & 0xff) | ((regs.ecx << 2) & 0x0300)) + 1;
  *sectors = regs.ecx & 0x3f;
  return 0;
}

static int
grub_biosdisk_get_diskinfo_real (int drive, void *drp, grub_uint16_t ax)
{
  struct grub_bios_int_registers regs;

  regs.eax = ax;

  /* compute the address of drive parameters */
  regs.esi = ((grub_addr_t) drp) & 0xf;
  regs.ds = ((grub_addr_t) drp) >> 4;
  regs.edx = drive & 0xff;

  regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
  grub_bios_interrupt (0x13, &regs);

  /* Check if unsuccessful. Ignore return value if carry isn't set to 
     workaround some buggy BIOSes. */
  if ((regs.flags & GRUB_CPU_INT_FLAGS_CARRY) && ((regs.eax & 0xff00) != 0))
    return (regs.eax & 0xff00) >> 8;

  return 0;
}

/*
 *   Return the cdrom information of DRIVE in CDRP. If an error occurs,
 *   then return non-zero, otherwise zero.
 */
static int
grub_biosdisk_get_cdinfo_int13_extensions (int drive, void *cdrp)
{
  return grub_biosdisk_get_diskinfo_real (drive, cdrp, 0x4b01);
}

/*
 *   Return the geometry of DRIVE in a drive parameters, DRP. If an error
 *   occurs, then return non-zero, otherwise zero.
 */
static int
grub_biosdisk_get_diskinfo_int13_extensions (int drive, void *drp)
{
  return grub_biosdisk_get_diskinfo_real (drive, drp, 0x4800);
}

static int
grub_biosdisk_get_drive (const char *name)
{
  unsigned long drive;

  if (name[0] == 'c' && name[1] == 'd' && name[2] == 0 && cd_drive)
    return cd_drive;

  if ((name[0] != 'f' && name[0] != 'h') || name[1] != 'd')
    goto fail;

  drive = grub_strtoul (name + 2, 0, 10);
  if (grub_errno != GRUB_ERR_NONE)
    goto fail;

  if (name[0] == 'h')
    drive += 0x80;

  return (int) drive ;

 fail:
  grub_error (GRUB_ERR_UNKNOWN_DEVICE, "not a biosdisk");
  return -1;
}

static int
grub_biosdisk_call_hook (int (*hook) (const char *name), int drive)
{
  char name[10];

  if (cd_drive && drive == cd_drive)
    return hook ("cd");

  grub_snprintf (name, sizeof (name),
		 (drive & 0x80) ? "hd%d" : "fd%d", drive & (~0x80));
  return hook (name);
}

static int
grub_biosdisk_iterate (int (*hook) (const char *name))
{
  int drive;
  int num_floppies;

  /* For hard disks, attempt to read the MBR.  */
  for (drive = 0x80; drive < 0x90; drive++)
    {
      if (grub_biosdisk_rw_standard (0x02, drive, 0, 0, 1, 1,
				     GRUB_MEMORY_MACHINE_SCRATCH_SEG) != 0)
	{
	  grub_dprintf ("disk", "Read error when probing drive 0x%2x\n", drive);
	  break;
	}

      if (grub_biosdisk_call_hook (hook, drive))
	return 1;
    }

  if (cd_drive)
    {
      if (grub_biosdisk_call_hook (hook, cd_drive))
      return 1;
    }

  /* For floppy disks, we can get the number safely.  */
  num_floppies = grub_biosdisk_get_num_floppies ();
  for (drive = 0; drive < num_floppies; drive++)
    if (grub_biosdisk_call_hook (hook, drive))
      return 1;

  return 0;
}

static grub_err_t
grub_biosdisk_open (const char *name, grub_disk_t disk)
{
  grub_uint64_t total_sectors = 0;
  int drive;
  struct grub_biosdisk_data *data;

  drive = grub_biosdisk_get_drive (name);
  if (drive < 0)
    return grub_errno;

  disk->id = drive;

  data = (struct grub_biosdisk_data *) grub_zalloc (sizeof (*data));
  if (! data)
    return grub_errno;

  data->drive = drive;

  if ((cd_drive) && (drive == cd_drive))
    {
      data->flags = GRUB_BIOSDISK_FLAG_LBA | GRUB_BIOSDISK_FLAG_CDROM;
      data->sectors = 32;
      /* TODO: get the correct size.  */
      total_sectors = GRUB_DISK_SIZE_UNKNOWN;
    }
  else
    {
      /* HDD */
      int version;

      version = grub_biosdisk_check_int13_extensions (drive);
      if (version)
	{
	  struct grub_biosdisk_drp *drp
	    = (struct grub_biosdisk_drp *) GRUB_MEMORY_MACHINE_SCRATCH_ADDR;

	  /* Clear out the DRP.  */
	  grub_memset (drp, 0, sizeof (*drp));
	  drp->size = sizeof (*drp);
	  if (! grub_biosdisk_get_diskinfo_int13_extensions (drive, drp))
	    {
	      data->flags = GRUB_BIOSDISK_FLAG_LBA;

	      if (drp->total_sectors)
		total_sectors = drp->total_sectors;
	      else
                /* Some buggy BIOSes doesn't return the total sectors
                   correctly but returns zero. So if it is zero, compute
                   it by C/H/S returned by the LBA BIOS call.  */
                total_sectors = drp->cylinders * drp->heads * drp->sectors;
	    }
	}
    }

  if (! (data->flags & GRUB_BIOSDISK_FLAG_CDROM))
    {
      if (grub_biosdisk_get_diskinfo_standard (drive,
					       &data->cylinders,
					       &data->heads,
					       &data->sectors) != 0)
        {
	  if (total_sectors && (data->flags & GRUB_BIOSDISK_FLAG_LBA))
	    {
	      data->sectors = 63;
	      data->heads = 255;
	      data->cylinders
		= grub_divmod64 (total_sectors
				 + data->heads * data->sectors - 1,
				 data->heads * data->sectors, 0);
	    }
	  else
	    {
	      grub_free (data);
	      return grub_error (GRUB_ERR_BAD_DEVICE, "%s cannot get C/H/S values", disk->name);
	    }
        }

      if (! total_sectors)
        total_sectors = data->cylinders * data->heads * data->sectors;
    }

  disk->total_sectors = total_sectors;
  disk->data = data;

  return GRUB_ERR_NONE;
}

static void
grub_biosdisk_close (grub_disk_t disk)
{
  grub_free (disk->data);
}

/* For readability.  */
#define GRUB_BIOSDISK_READ	0
#define GRUB_BIOSDISK_WRITE	1

#define GRUB_BIOSDISK_CDROM_RETRY_COUNT 3

static grub_err_t
grub_biosdisk_rw (int cmd, grub_disk_t disk,
		  grub_disk_addr_t sector, grub_size_t size,
		  unsigned segment)
{
  struct grub_biosdisk_data *data = disk->data;

  if (data->flags & GRUB_BIOSDISK_FLAG_LBA)
    {
      struct grub_biosdisk_dap *dap;

      dap = (struct grub_biosdisk_dap *) (GRUB_MEMORY_MACHINE_SCRATCH_ADDR
					  + (data->sectors
					     << GRUB_DISK_SECTOR_BITS));
      dap->length = sizeof (*dap);
      dap->reserved = 0;
      dap->blocks = size;
      dap->buffer = segment << 16;	/* The format SEGMENT:ADDRESS.  */
      dap->block = sector;

      if (data->flags & GRUB_BIOSDISK_FLAG_CDROM)
        {
	  int i;

	  if (cmd)
	    return grub_error (GRUB_ERR_WRITE_ERROR, "can\'t write to cdrom");

	  dap->blocks = ALIGN_UP (dap->blocks, 4) >> 2;
	  dap->block >>= 2;

	  for (i = 0; i < GRUB_BIOSDISK_CDROM_RETRY_COUNT; i++)
            if (! grub_biosdisk_rw_int13_extensions (0x42, data->drive, dap))
	      break;

	  if (i == GRUB_BIOSDISK_CDROM_RETRY_COUNT)
	    return grub_error (GRUB_ERR_READ_ERROR, "cdrom read error");
	}
      else
        if (grub_biosdisk_rw_int13_extensions (cmd + 0x42, data->drive, dap))
	  {
	    /* Fall back to the CHS mode.  */
	    data->flags &= ~GRUB_BIOSDISK_FLAG_LBA;
	    disk->total_sectors = data->cylinders * data->heads * data->sectors;
	    return grub_biosdisk_rw (cmd, disk, sector, size, segment);
	  }
    }
  else
    {
      unsigned coff, hoff, soff;
      unsigned head;

      /* It is impossible to reach over 8064 MiB (a bit less than LBA24) with
	 the traditional CHS access.  */
      if (sector >
	  1024 /* cylinders */ *
	  256 /* heads */ *
	  63 /* spt */)
	return grub_error (GRUB_ERR_OUT_OF_RANGE, "%s out of disk", disk->name);

      soff = ((grub_uint32_t) sector) % data->sectors + 1;
      head = ((grub_uint32_t) sector) / data->sectors;
      hoff = head % data->heads;
      coff = head / data->heads;

      if (coff >= data->cylinders)
	return grub_error (GRUB_ERR_OUT_OF_RANGE, "%s out of disk", disk->name);

      if (grub_biosdisk_rw_standard (cmd + 0x02, data->drive,
				     coff, hoff, soff, size, segment))
	{
	  switch (cmd)
	    {
	    case GRUB_BIOSDISK_READ:
	      return grub_error (GRUB_ERR_READ_ERROR, "%s read error", disk->name);
	    case GRUB_BIOSDISK_WRITE:
	      return grub_error (GRUB_ERR_WRITE_ERROR, "%s write error", disk->name);
	    }
	}
    }

  return GRUB_ERR_NONE;
}

/* Return the number of sectors which can be read safely at a time.  */
static grub_size_t
get_safe_sectors (grub_disk_addr_t sector, grub_uint32_t sectors)
{
  grub_size_t size;
  grub_uint32_t offset;

  /* OFFSET = SECTOR % SECTORS */
  grub_divmod64 (sector, sectors, &offset);

  size = sectors - offset;

  /* Limit the max to 0x7f because of Phoenix EDD.  */
  if (size > 0x7f)
    size = 0x7f;

  return size;
}

static grub_err_t
grub_biosdisk_read (grub_disk_t disk, grub_disk_addr_t sector,
		    grub_size_t size, char *buf)
{
  struct grub_biosdisk_data *data = disk->data;

  while (size)
    {
      grub_size_t len;
      grub_size_t cdoff = 0;

      len = get_safe_sectors (sector, data->sectors);

      if (data->flags & GRUB_BIOSDISK_FLAG_CDROM)
	{
	  cdoff = (sector & 3) << GRUB_DISK_SECTOR_BITS;
	  len = ALIGN_UP (sector + len, 4) - (sector & ~3);
	  sector &= ~3;
	}

      if (len > size)
	len = size;

      if (grub_biosdisk_rw (GRUB_BIOSDISK_READ, disk, sector, len,
			    GRUB_MEMORY_MACHINE_SCRATCH_SEG))
	return grub_errno;

      grub_memcpy (buf, (void *) (GRUB_MEMORY_MACHINE_SCRATCH_ADDR + cdoff),
		   len << GRUB_DISK_SECTOR_BITS);
      buf += len << GRUB_DISK_SECTOR_BITS;
      sector += len;
      size -= len;
    }

  return grub_errno;
}

static grub_err_t
grub_biosdisk_write (grub_disk_t disk, grub_disk_addr_t sector,
		     grub_size_t size, const char *buf)
{
  struct grub_biosdisk_data *data = disk->data;

  if (data->flags & GRUB_BIOSDISK_FLAG_CDROM)
    return grub_error (GRUB_ERR_IO, "can't write to CDROM");

  while (size)
    {
      grub_size_t len;

      len = get_safe_sectors (sector, data->sectors);
      if (len > size)
	len = size;

      grub_memcpy ((void *) GRUB_MEMORY_MACHINE_SCRATCH_ADDR, buf,
		   len << GRUB_DISK_SECTOR_BITS);

      if (grub_biosdisk_rw (GRUB_BIOSDISK_WRITE, disk, sector, len,
			    GRUB_MEMORY_MACHINE_SCRATCH_SEG))
	return grub_errno;

      buf += len << GRUB_DISK_SECTOR_BITS;
      sector += len;
      size -= len;
    }

  return grub_errno;
}

static struct grub_disk_dev grub_biosdisk_dev =
  {
    .name = "biosdisk",
    .id = GRUB_DISK_DEVICE_BIOSDISK_ID,
    .iterate = grub_biosdisk_iterate,
    .open = grub_biosdisk_open,
    .close = grub_biosdisk_close,
    .read = grub_biosdisk_read,
    .write = grub_biosdisk_write,
    .next = 0
  };

static void
grub_disk_biosdisk_fini (void)
{
  grub_disk_dev_unregister (&grub_biosdisk_dev);
}

GRUB_MOD_INIT(biosdisk)
{
  struct grub_biosdisk_cdrp *cdrp
        = (struct grub_biosdisk_cdrp *) GRUB_MEMORY_MACHINE_SCRATCH_ADDR;

  if (grub_disk_firmware_is_tainted)
    {
      grub_printf ("Firmware is marked as tainted, refusing to initialize.\n");
      return;
    }
  grub_disk_firmware_fini = grub_disk_biosdisk_fini;

  grub_memset (cdrp, 0, sizeof (*cdrp));
  cdrp->size = sizeof (*cdrp);
  cdrp->media_type = 0xFF;
  if ((! grub_biosdisk_get_cdinfo_int13_extensions (grub_boot_drive, cdrp)) &&
      ((cdrp->media_type & GRUB_BIOSDISK_CDTYPE_MASK)
       == GRUB_BIOSDISK_CDTYPE_NO_EMUL))
    cd_drive = cdrp->drive_no;
  /* Since diskboot.S rejects devices over 0x90 it must be a CD booted with
     cdboot.S
   */
  if (grub_boot_drive >= 0x90)
    cd_drive = grub_boot_drive;

  grub_disk_dev_register (&grub_biosdisk_dev);
}

GRUB_MOD_FINI(biosdisk)
{
  grub_disk_biosdisk_fini ();
}

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

On 16.03.2011 23:00, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 09:19:12PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>   
>> Following may help:
>> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
>> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
>> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 20:17:57 +0000
>> @@ -80,6 +80,9 @@
>>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
>>  
>>    grub_bios_interrupt (0x13, &regs);
>> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
>> +    return 0;
>> +
>>    return (regs.eax >> 8) & 0xff;
>>  }
>>     
> I now got this:
>
> GRUB loading.
> Welcome to GRUB!
>
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
> int13_ext fixup due to CARRY flag
>   
It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
was a wild guess.
> Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
>   
Error 0x80 according to
http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
Following code would retry on timeout:
=== modified file 'grub-core/disk/i386/pc/biosdisk.c'
--- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
+++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
@@ -72,6 +72,8 @@
 grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
 {
   struct grub_bios_int_registers regs;
+  int tries = 4;
+ retry:
   regs.eax = ah << 8;
   /* compute the address of disk_address_packet */
   regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
@@ -80,6 +82,16 @@
   regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
 
   grub_bios_interrupt (0x13, &regs);
+  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
+    return 0;
+
+  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
+    {
+      grub_millisleep (10);
+      tries--;
+      goto retry;
+    }
+
   return (regs.eax >> 8) & 0xff;
 }
 

> error: hd0 out of disk.
> Entering rescue mode...
> grub rescue>
>
> I am now using 1.99-rc1.  Here is the biosdisk.c with my changes.
>
>   
>
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/grub-devel
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Mar 16, 2011 at 11:16:05PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> On 16.03.2011 23:00, Lennart Sorensen wrote:
> > On Wed, Mar 16, 2011 at 09:19:12PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> >   
> >> Following may help:
> >> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> >> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> >> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 20:17:57 +0000
> >> @@ -80,6 +80,9 @@
> >>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
> >>  
> >>    grub_bios_interrupt (0x13, &regs);
> >> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> >> +    return 0;
> >> +
> >>    return (regs.eax >> 8) & 0xff;
> >>  }
> >>     
> > I now got this:
> >
> > GRUB loading.
> > Welcome to GRUB!
> >
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> > int13_ext fixup due to CARRY flag
> >   
> It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
> BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
> was a wild guess.
> > Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
> >   
> Error 0x80 according to
> http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
> Following code would retry on timeout:
> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
> @@ -72,6 +72,8 @@
>  grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
>  {
>    struct grub_bios_int_registers regs;
> +  int tries = 4;
> + retry:
>    regs.eax = ah << 8;
>    /* compute the address of disk_address_packet */
>    regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
> @@ -80,6 +82,16 @@
>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
>  
>    grub_bios_interrupt (0x13, &regs);
> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> +    return 0;
> +
> +  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
> +    {
> +      grub_millisleep (10);
> +      tries--;
> +      goto retry;
> +    }
> +
>    return (regs.eax >> 8) & 0xff;
>  }
>  
> 
> > error: hd0 out of disk.
> > Entering rescue mode...
> > grub rescue>
> >
> > I am now using 1.99-rc1.  Here is the biosdisk.c with my changes.
> >
> >   
> >
> >
> > _______________________________________________
> > Grub-devel mailing list
> > Grub-devel@gnu.org
> > http://lists.gnu.org/mailman/listinfo/grub-devel
> >   

I am trying that now.

By the way I think I spotted a bug:

      else
        if (grub_biosdisk_rw_int13_extensions (cmd + 0x42, data->drive, dap))
          {
            /* Fall back to the CHS mode.  */
            data->flags &= ~GRUB_BIOSDISK_FLAG_LBA;
            disk->total_sectors = data->cylinders * data->heads * data->sectors;
            return grub_biosdisk_rw (cmd, disk, sector, size, segment);
          }

I think that + 0x42 is wrong.  The extended commands are 0x40 above the
normal ones.  + 0x40 makes more sense to me.

0x42 = extended read.
0x02 = normal read.

0x48 = extended parameter read.
0x08 = normal parameter read.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Thu, Mar 17, 2011 at 10:06:06AM -0400, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 11:16:05PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> > On 16.03.2011 23:00, Lennart Sorensen wrote:
> > > On Wed, Mar 16, 2011 at 09:19:12PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> > >   
> > >> Following may help:
> > >> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> > >> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> > >> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 20:17:57 +0000
> > >> @@ -80,6 +80,9 @@
> > >>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
> > >>  
> > >>    grub_bios_interrupt (0x13, &regs);
> > >> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> > >> +    return 0;
> > >> +
> > >>    return (regs.eax >> 8) & 0xff;
> > >>  }
> > >>     
> > > I now got this:
> > >
> > > GRUB loading.
> > > Welcome to GRUB!
> > >
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > > int13_ext fixup due to CARRY flag
> > >   
> > It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
> > BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
> > was a wild guess.
> > > Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
> > >   
> > Error 0x80 according to
> > http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
> > Following code would retry on timeout:
> > === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> > --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> > +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
> > @@ -72,6 +72,8 @@
> >  grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
> >  {
> >    struct grub_bios_int_registers regs;
> > +  int tries = 4;
> > + retry:
> >    regs.eax = ah << 8;
> >    /* compute the address of disk_address_packet */
> >    regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
> > @@ -80,6 +82,16 @@
> >    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
> >  
> >    grub_bios_interrupt (0x13, &regs);
> > +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> > +    return 0;
> > +
> > +  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
> > +    {
> > +      grub_millisleep (10);
> > +      tries--;
> > +      goto retry;
> > +    }
> > +
> >    return (regs.eax >> 8) & 0xff;
> >  }
> >  
> > 
> > > error: hd0 out of disk.
> > > Entering rescue mode...
> > > grub rescue>
> > >
> > > I am now using 1.99-rc1.  Here is the biosdisk.c with my changes.
> > >
> > >   
> > >
> > >
> > > _______________________________________________
> > > Grub-devel mailing list
> > > Grub-devel@gnu.org
> > > http://lists.gnu.org/mailman/listinfo/grub-devel
> > >   
> 
> I am trying that now.
> 
> By the way I think I spotted a bug:
> 
>       else
>         if (grub_biosdisk_rw_int13_extensions (cmd + 0x42, data->drive, dap))
>           {
>             /* Fall back to the CHS mode.  */
>             data->flags &= ~GRUB_BIOSDISK_FLAG_LBA;
>             disk->total_sectors = data->cylinders * data->heads * data->sectors;
>             return grub_biosdisk_rw (cmd, disk, sector, size, segment);
>           }
> 
> I think that + 0x42 is wrong.  The extended commands are 0x40 above the
> normal ones.  + 0x40 makes more sense to me.
> 
> 0x42 = extended read.
> 0x02 = normal read.
> 
> 0x48 = extended parameter read.
> 0x08 = normal parameter read.

Never mind.  cmd = 0 for read, not 0x2.  Oh well.  I found it adds 0x02
in _standard, so it works the same.

Still testing the other stuff.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Mar 16, 2011 at 11:16:05PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
> BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
> was a wild guess.
> > Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
> >   
> Error 0x80 according to
> http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
> Following code would retry on timeout:
> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
> @@ -72,6 +72,8 @@
>  grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
>  {
>    struct grub_bios_int_registers regs;
> +  int tries = 4;
> + retry:
>    regs.eax = ah << 8;
>    /* compute the address of disk_address_packet */
>    regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
> @@ -80,6 +82,16 @@
>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
>  
>    grub_bios_interrupt (0x13, &regs);
> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> +    return 0;
> +
> +  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
> +    {
> +      grub_millisleep (10);
> +      tries--;
> +      goto retry;
> +    }
> +
>    return (regs.eax >> 8) & 0xff;
>  }

So that didn't make any difference to it.

It just keeps returning 0x8000 in eax on every retry.

I really want to find out what this BIOS is doing wrong so I can tell
the system maker what to fix.

I am going to add some debuging to find out what is in the dap being
passed to the BIOS.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 2119 bytes --]

On 17.03.2011 17:09, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 11:16:05PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>   
>> It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
>> BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
>> was a wild guess.
>>     
>>> Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
>>>   
>>>       
>> Error 0x80 according to
>> http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
>> Following code would retry on timeout:
>> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
>> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
>> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
>> @@ -72,6 +72,8 @@
>>  grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
>>  {
>>    struct grub_bios_int_registers regs;
>> +  int tries = 4;
>> + retry:
>>    regs.eax = ah << 8;
>>    /* compute the address of disk_address_packet */
>>    regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
>> @@ -80,6 +82,16 @@
>>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
>>  
>>    grub_bios_interrupt (0x13, &regs);
>> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
>> +    return 0;
>> +
>> +  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
>> +    {
>> +      grub_millisleep (10);
>> +      tries--;
>> +      goto retry;
>> +    }
>> +
>>    return (regs.eax >> 8) & 0xff;
>>  }
>>     
> So that didn't make any difference to it.
>
> It just keeps returning 0x8000 in eax on every retry.
>
>   
Another wild guess:
@@ -506,6 +520,8 @@ get_safe_sectors (grub_disk_addr_t secto
   grub_size_t size;
   grub_uint32_t offset;
 
+  return 1;
+
   /* OFFSET = SECTOR % SECTORS */
   grub_divmod64 (sector, sectors, &offset);
 

> I really want to find out what this BIOS is doing wrong so I can tell
> the system maker what to fix.
>
> I am going to add some debuging to find out what is in the dap being
> passed to the BIOS.
>
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Thu, Mar 17, 2011 at 05:29:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> On 17.03.2011 17:09, Lennart Sorensen wrote:
> > On Wed, Mar 16, 2011 at 11:16:05PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> >   
> >> It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
> >> BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
> >> was a wild guess.
> >>     
> >>> Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
> >>>   
> >>>       
> >> Error 0x80 according to
> >> http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
> >> Following code would retry on timeout:
> >> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
> >> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
> >> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
> >> @@ -72,6 +72,8 @@
> >>  grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
> >>  {
> >>    struct grub_bios_int_registers regs;
> >> +  int tries = 4;
> >> + retry:
> >>    regs.eax = ah << 8;
> >>    /* compute the address of disk_address_packet */
> >>    regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
> >> @@ -80,6 +82,16 @@
> >>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
> >>  
> >>    grub_bios_interrupt (0x13, &regs);
> >> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
> >> +    return 0;
> >> +
> >> +  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
> >> +    {
> >> +      grub_millisleep (10);
> >> +      tries--;
> >> +      goto retry;
> >> +    }
> >> +
> >>    return (regs.eax >> 8) & 0xff;
> >>  }
> >>     
> > So that didn't make any difference to it.
> >
> > It just keeps returning 0x8000 in eax on every retry.
> >
> >   
> Another wild guess:
> @@ -506,6 +520,8 @@ get_safe_sectors (grub_disk_addr_t secto
>    grub_size_t size;
>    grub_uint32_t offset;
>  
> +  return 1;
> +
>    /* OFFSET = SECTOR % SECTORS */
>    grub_divmod64 (sector, sectors, &offset);

No go either.

Here is what I see with some more debugging:

[snip]
int13_ext call: ah=42, drive=80, dap=   6fe00 (len=16 rsvd=0 blks=1 buf=1744830464 sec=2078)
int13_ext call: ah=42, drive=80, dap=   6fe00 (len=16 rsvd=0 blks=1 buf=1744830464 sec=2079)
int13_ext call: ah=42, drive=80, dap=   6fe00 (len=16 rsvd=0 blks=1 buf=1744830464 sec=72091664)
failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
error: hd0 out of disk.
Entering rescue mode...
grub rescue>

Now it has no excuse for failing a call for sector 72091664 when the
drive has 156000000 or so sectors.

I imagine if I made a small boot partition (less than 8G), the stupid
thing would boot fine, but that's not acceptable on a brand new design.
I am going to try resizing the partition as a test.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 3532 bytes --]

On 17.03.2011 19:40, Lennart Sorensen wrote:
> On Thu, Mar 17, 2011 at 05:29:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>   
>> On 17.03.2011 17:09, Lennart Sorensen wrote:
>>     
>>> On Wed, Mar 16, 2011 at 11:16:05PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>>>   
>>>       
>>>> It's not really a fixup. Correct return on no error is CF=0 AH=0. Some
>>>> BIOSes in some other functions returned CF=0, AH!= 0 on no error so it
>>>> was a wild guess.
>>>>     
>>>>         
>>>>> Failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
>>>>>   
>>>>>       
>>>>>           
>>>> Error 0x80 according to
>>>> http://www.delorie.com/djgpp/doc/rbinter/it/34/2.html means timeout.
>>>> Following code would retry on timeout:
>>>> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
>>>> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
>>>> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 22:13:12 +0000
>>>> @@ -72,6 +72,8 @@
>>>>  grub_biosdisk_rw_int13_extensions (int ah, int drive, void *dap)
>>>>  {
>>>>    struct grub_bios_int_registers regs;
>>>> +  int tries = 4;
>>>> + retry:
>>>>    regs.eax = ah << 8;
>>>>    /* compute the address of disk_address_packet */
>>>>    regs.ds = (((grub_addr_t) dap) & 0xffff0000) >> 4;
>>>> @@ -80,6 +82,16 @@
>>>>    regs.flags = GRUB_CPU_INT_FLAGS_DEFAULT;
>>>>  
>>>>    grub_bios_interrupt (0x13, &regs);
>>>> +  if (!(regs.flags & GRUB_CPU_INT_FLAGS_CARRY))
>>>> +    return 0;
>>>> +
>>>> +  if (((regs.eax >> 8) & 0xff) == 0x80 && tries)
>>>> +    {
>>>> +      grub_millisleep (10);
>>>> +      tries--;
>>>> +      goto retry;
>>>> +    }
>>>> +
>>>>    return (regs.eax >> 8) & 0xff;
>>>>  }
>>>>     
>>>>         
>>> So that didn't make any difference to it.
>>>
>>> It just keeps returning 0x8000 in eax on every retry.
>>>
>>>   
>>>       
>> Another wild guess:
>> @@ -506,6 +520,8 @@ get_safe_sectors (grub_disk_addr_t secto
>>    grub_size_t size;
>>    grub_uint32_t offset;
>>  
>> +  return 1;
>> +
>>    /* OFFSET = SECTOR % SECTORS */
>>    grub_divmod64 (sector, sectors, &offset);
>>     
> No go either.
>
>   
Can you dump complete drp before and after every call?

(19:52:44) phcoder: cjwatson: may the Lennart's last message be caused
by 4K sectors?(idk if he has 512B or 4K sectors)
(19:56:14) cjwatson: possible; he can find out by printing
drp->bytes_per_sector after the call to
grub_biosdisk_get_diskinfo_int13_extensions in
grub-core/disk/i386/pc/biosdisk.c
(19:56:25) cjwatson: at least find out what the BIOS thinks
> Here is what I see with some more debugging:
>
> [snip]
> int13_ext call: ah=42, drive=80, dap=   6fe00 (len=16 rsvd=0 blks=1 buf=1744830464 sec=2078)
> int13_ext call: ah=42, drive=80, dap=   6fe00 (len=16 rsvd=0 blks=1 buf=1744830464 sec=2079)
> int13_ext call: ah=42, drive=80, dap=   6fe00 (len=16 rsvd=0 blks=1 buf=1744830464 sec=72091664)
> failed int13_ext call: ah=42, drive=80, dap=   6fe00 returned eax=8000
> error: hd0 out of disk.
> Entering rescue mode...
> grub rescue>
>
> Now it has no excuse for failing a call for sector 72091664 when the
> drive has 156000000 or so sectors.
>
> I imagine if I made a small boot partition (less than 8G), the stupid
> thing would boot fine, but that's not acceptable on a brand new design.
> I am going to try resizing the partition as a test.
>
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Thu, Mar 17, 2011 at 08:01:09PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> Can you dump complete drp before and after every call?

I can try.  Of course transscribing them by hand is a pain.  I wonder
if the serial port works on this thing, so I can make grub use that
(assuming I can get grub to load itself first, before hitting the
BIOS bug.  I might be able to.)

> (19:52:44) phcoder: cjwatson: may the Lennart's last message be caused
> by 4K sectors?(idk if he has 512B or 4K sectors)
> (19:56:14) cjwatson: possible; he can find out by printing
> drp->bytes_per_sector after the call to
> grub_biosdisk_get_diskinfo_int13_extensions in
> grub-core/disk/i386/pc/biosdisk.c
> (19:56:25) cjwatson: at least find out what the BIOS thinks

It's a VERY old 80G WD SATA drive.  No 4k sectors here.

By the way, making the partition smaller than 8G made it boot just
fine, so it seems the BIOS is failing with higher sector numbers only.
That's simply pathetic.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Grégoire Sutre]

On 03/16/2011 20:19, Phillip Susi wrote:
> On 3/16/2011 1:32 PM, Lennart Sorensen wrote:
>> So I now have some debug info:
>>
>> version=48
>> total_sectors=156312576 C=16383 H=16 S=63
>> error: hd0,msdos1 out of disk.
>> Entering rescue mode...
>> grub rescue>
>>
>> So version is returning something, and it is actually getting the disk
>> size properly.  I wonder what else could cause the out of disk message
>> then.
>
> What does the partition table look like?  It doesn't extend beyond
> sector 156312575 does it?
>
> Also I wonder about why H=16.  It should be 255.

I didn't look at the code, but this is probably the ATA geometry of
the disk, which is returned by the INT-13h (Extended) function 48h
(Read Drive Parameters).  If I remember correctly, this geometry is
always 16383/16/63 for disks larger than 8.4GB.

Grégoire

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Fri, Mar 18, 2011 at 10:04:42AM +0100, Grégoire Sutre wrote:
> I didn't look at the code, but this is probably the ATA geometry of
> the disk, which is returned by the INT-13h (Extended) function 48h
> (Read Drive Parameters).  If I remember correctly, this geometry is
> always 16383/16/63 for disks larger than 8.4GB.

That does match what is returned.

Grub later assumes H=255 if it falls back to CHS as far as I can tell.

H=255 S=63 would be really bad for alignment on SSDs and such (not that
H=16 is a whole lot better, but it is better).

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Grégoire Sutre]

On 03/18/2011 15:22, Lennart Sorensen wrote:
> On Fri, Mar 18, 2011 at 10:04:42AM +0100, Grégoire Sutre wrote:
>> I didn't look at the code, but this is probably the ATA geometry of
>> the disk, which is returned by the INT-13h (Extended) function 48h
>> (Read Drive Parameters).  If I remember correctly, this geometry is
>> always 16383/16/63 for disks larger than 8.4GB.
>
> That does match what is returned.
>
> Grub later assumes H=255 if it falls back to CHS as far as I can tell.

If the standard (non-extended) INT-13h interface is used, then addressing
is performed via CHS, and w.r.t. to the BIOS geometry (which need not be
the same as the ATA geometry reported by the disk itself, due to potential
address translation by the BIOS).  I guess that H=255 is the from the BIOS
geometry of the disk (as returned by the INT-13h function 08h).

> H=255 S=63 would be really bad for alignment on SSDs and such (not that
> H=16 is a whole lot better, but it is better).

The geometry just says how to interpret CHS addresses, it does not imply
any partition alignment.  You can have 1MiB-aligned partitions even with
a n/255/63 BIOS geometry.

Grégoire

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Fri, Mar 18, 2011 at 03:56:27PM +0100, Grégoire Sutre wrote:
> If the standard (non-extended) INT-13h interface is used, then addressing
> is performed via CHS, and w.r.t. to the BIOS geometry (which need not be
> the same as the ATA geometry reported by the disk itself, due to potential
> address translation by the BIOS).  I guess that H=255 is the from the BIOS
> geometry of the disk (as returned by the INT-13h function 08h).

Very likely.

> The geometry just says how to interpret CHS addresses, it does not imply
> any partition alignment.  You can have 1MiB-aligned partitions even with
> a n/255/63 BIOS geometry.

Certainly true.

-- 
Len Sorensen

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

On 16.03.2011 18:32, Lennart Sorensen wrote:
> On Wed, Mar 16, 2011 at 04:29:32PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>   
>> On 16.03.2011 16:13, Lennart Sorensen wrote:
>>     
>>> I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
>>> of disk' error when trying to boot from the HD after doing an install.
>>> It ends up at a grub rescue prompt, given it can't even read the disk
>>> well enough to load the full grub.  Doing ls shows the disks, and I can
>>> show the root directory of a partition, but anything further seems to
>>> hit the same 'out of disk' error.
>>>
>>>   
>>>       
>> ls -l should show how big it thinks the disk is. Compare it with its
>> real size
>>     
>>> The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
>>> (the new one) with a phoenix bios.  So far Compulab's response when told
>>> it won't boot grub2 was "We know, but syslinux works fine.", which to
>>> me is not a solution at all.  The same system also hangs the Linux 2.6.32
>>> kernel unless 'edd=off' is added to the kernel command line.
>>>
>>>   
>>>       
>> Looks like int13 is borked. Perhaps it's borked in a way which prevents
>> GRUB from detecting 13/42 function. You can try to force int13/42 by:
>> === modified file 'grub-core/disk/i386/pc/biosdisk.c'
>> --- grub-core/disk/i386/pc/biosdisk.c    2011-01-04 14:42:47 +0000
>> +++ grub-core/disk/i386/pc/biosdisk.c    2011-03-16 15:28:26 +0000
>> @@ -145,6 +145,8 @@
>>  {
>>    struct grub_bios_int_registers regs;
>>  
>> +  return 1;
>> +
>>    regs.edx = drive & 0xff;
>>    regs.eax = 0x4100;
>>    regs.ebx = 0x55aa;
>>     
> So I now have some debug info:
>
> version=48
> total_sectors=156312576 C=16383 H=16 S=63
> error: hd0,msdos1 out of disk.
> Entering rescue mode...
> grub rescue>
>
>   
These are correct values. However it may be that the correct value is
returned only on the first call. Could you try forcing

version=48
total_sectors=156312576


> So version is returning something, and it is actually getting the disk
> size properly.  I wonder what else could cause the out of disk message
> then.
>
>   


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko

Re: How to debug 'out of disk' error.

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 1306 bytes --]

On 16.03.2011 16:13, Lennart Sorensen wrote:
> I am trying to figure out why grub2 (from Debian Squeeze) gives an 'out
> of disk' error when trying to boot from the HD after doing an install.
> It ends up at a grub rescue prompt, given it can't even read the disk
> well enough to load the full grub.  Doing ls shows the disks, and I can
> show the root directory of a partition, but anything further seems to
> hit the same 'out of disk' error.
>
I have another idea. Could you try GRUB 1.97 and GRUB Legacy? A similar
report on our BTS seems to indicate that GRUB Legacy works. Major
difference between Legacy and latest GRUB2 is that GRUB2 uses C-wrappers
around BIOS calls whereas in GRUB Legacy they are in asm. The change
happened after 1.97
> The system is a Compulab CM-iTC which uses an intel tunnelcreek atom
> (the new one) with a phoenix bios.  So far Compulab's response when told
> it won't boot grub2 was "We know, but syslinux works fine.", which to
> me is not a solution at all.  The same system also hangs the Linux 2.6.32
> kernel unless 'edd=off' is added to the kernel command line.
>
> So any suggestions on how to go about debuagging this?  Or even any
> ideas why grub would give such an error.
>


-- 
Regards
Vladimir 'φ-coder/phcoder' Serbinenko



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: How to debug 'out of disk' error.

[Lennart Sorensen]

On Wed, Apr 06, 2011 at 02:50:01PM +0200, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> I have another idea. Could you try GRUB 1.97 and GRUB Legacy? A similar
> report on our BTS seems to indicate that GRUB Legacy works. Major
> difference between Legacy and latest GRUB2 is that GRUB2 uses C-wrappers
> around BIOS calls whereas in GRUB Legacy they are in asm. The change
> happened after 1.97

I managed to get the maker of the board to confirm the BIOS has a bug.
They are going to fix it.  It incorrectly returns error when you try to
do LBA access beyond 8GB, which makes grub fall back to CHS and report
'out of disk'.  They have been trying to add SD card support to the BIOS
and I think they broke this in the process.

So grub is fine as far as I can tell.

-- 
Len Sorensen