does disk geometry depend on partition table?

does disk geometry depend on partition table?

[Zhixu Liu]

[-- Attachment #1: Type: text/plain, Size: 1695 bytes --]

Hi, all,

We encounter a strange problem recently. In short, following is the output
of geometry of
hd0 in situation with correct partition in disk and without partition in
disk.

When with correct partition table in disk hd0:
    grub> geometry (hd0)
    drive 0x80: C/H/S = 500/16/32, The number of sectors = 256000, LBA

When without partition table in disk hd0:
    grub> geometry (hd0)
    drive 0x80: C/H/S = 126/32/63, The number of sectors = 256000, LBA

The former is correct! However, it's strange that the disk geometry depend
on the partition table?!
Maybe this is due to a buggy BIOS implementation. We also check the code in
disk/i386/pc/biosdisk.c
or (stage2/bios.c) for grub-0.97, and for code in function
grub_biosdisk_open (or get_diskinfo in grub-0.97),
there is a standard int13 check of disk after check with ext int13,

  if(drive & 0x80) {
     ...
    version = grub_biosdisk_check_int13_extension();
      if(version) {
         grub_biosdisk_get_diskinfo_int13_extensions();
      }
   }

   grub_biosdisk_get_diskinfo_standard()

We're wondering does it really need to call
grub_biosdisk_get_diskinfo_standard() again if we have get the necessary
information
from ext int13? I think we call the later only if ext int13 failed for some
reason. Especially in our case the code should be .

  if(drive & 0x80) {
     ...
    version = grub_biosdisk_check_int13_extension();
      if(version) {
         grub_biosdisk_get_diskinfo_int13_extensions();
      } else {
         grub_biosdisk_get_diskinfo_standard()
      }
   }

and it do works!

Of course, maybe there are some details I'm not clear. Does anyone encounter
this problem before?

Best regards!
-- 
Zhixu Liu

[-- Attachment #2: Type: text/html, Size: 2181 bytes --]

Re: does disk geometry depend on partition table?

[Hollis Blanchard]

On Mon, 2006-11-06 at 17:07 +0800, Zhixu Liu wrote:
> Hi, all, 
> 
> We encounter a strange problem recently. In short, following is the
> output of geometry of
> hd0 in situation with correct partition in disk and without partition
> in disk.
> When with correct partition table in disk hd0:
> 
>     grub> geometry (hd0)
>     drive 0x80: C/H/S = 500/16/32, The number of sectors = 256000, LBA
>
> When without partition table in disk hd0:
>     grub> geometry (hd0)
>     drive 0x80: C/H/S = 126/32/63, The number of sectors = 256000, LBA
> 
> The former is correct! However, it's strange that the disk geometry
> depend on the partition table?!
> Maybe this is due to a buggy BIOS implementation. We also check the
> code in disk/i386/pc/biosdisk.c
> or (stage2/bios.c) for grub-0.97, and for code in function
> grub_biosdisk_open (or get_diskinfo in grub-0.97), 
> there is a standard int13 check of disk after check with ext int13, 
> 
>   if(drive & 0x80) {
>      ...
>     version = grub_biosdisk_check_int13_extension(); 
>       if(version) {
>          grub_biosdisk_get_diskinfo_int13_extensions();
>       }
>    }
> 
>    grub_biosdisk_get_diskinfo_standard()
> 
> We're wondering does it really need to call
> grub_biosdisk_get_diskinfo_standard() again if we have get the
> necessary information 
> from ext int13? I think we call the later only if ext int13 failed for
> some reason. Especially in our case the code should be . 
> 
>   if(drive & 0x80) {
>      ...
>     version = grub_biosdisk_check_int13_extension();
>       if(version) {
>          grub_biosdisk_get_diskinfo_int13_extensions();
>       } else {
>          grub_biosdisk_get_diskinfo_standard()
>       }
>    }
> 
> and it do works!

Thanks for your report, Zhixu. Does this patch work?

Index: disk/i386/pc/biosdisk.c
===================================================================
RCS file: /cvsroot/grub/grub2/disk/i386/pc/biosdisk.c,v
retrieving revision 1.9
diff -u -p -r1.9 biosdisk.c
--- disk/i386/pc/biosdisk.c     29 Jul 2006 10:11:01 -0000      1.9
+++ disk/i386/pc/biosdisk.c     14 Nov 2006 03:39:47 -0000
@@ -133,21 +133,24 @@ grub_biosdisk_open (const char *name, gr
        }
     }

-  if (grub_biosdisk_get_diskinfo_standard (drive,
-                                          &data->cylinders,
-                                          &data->heads,
-                                          &data->sectors) != 0)
+  if (! total_sectors)
     {
-      grub_free (data);
-      return grub_error (GRUB_ERR_BAD_DEVICE, "cannot get C/H/S values");
-    }
+      /* FDD, or HDD without int13 extensions.  */
+      if (grub_biosdisk_get_diskinfo_standard (drive,
+                                              &data->cylinders,
+                                              &data->heads,
+                                              &data->sectors) != 0)
+       {
+         grub_free (data);
+         return grub_error (GRUB_ERR_BAD_DEVICE, "cannot get C/H/S values");
+       }

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

   disk->total_sectors = total_sectors;
   disk->data = data;
-
+
   return GRUB_ERR_NONE;
 }


-Hollis

Re: does disk geometry depend on partition table?

[Zhixu Liu]

[-- Attachment #1: Type: text/plain, Size: 4129 bytes --]

Hi, Hollis,

We've tried something like this, and found it works for some platform, but
still failed for one platform.
Indeed, in this weird platform, we get both wrong CHS value with and without
partition table. But the OS always
get the correct value. The problem is due to either BIOS or Disk for this
particular platform, we think. But we're not clear.

Anyway, the patch works for some cases, and should for most cases.

Best regards!

Zhixu

On 11/14/06, Hollis Blanchard <hollis@penguinppc.org> wrote:
>
> On Mon, 2006-11-06 at 17:07 +0800, Zhixu Liu wrote:
> > Hi, all,
> >
> > We encounter a strange problem recently. In short, following is the
> > output of geometry of
> > hd0 in situation with correct partition in disk and without partition
> > in disk.
> > When with correct partition table in disk hd0:
> >
> >     grub> geometry (hd0)
> >     drive 0x80: C/H/S = 500/16/32, The number of sectors = 256000, LBA
> >
> > When without partition table in disk hd0:
> >     grub> geometry (hd0)
> >     drive 0x80: C/H/S = 126/32/63, The number of sectors = 256000, LBA
> >
> > The former is correct! However, it's strange that the disk geometry
> > depend on the partition table?!
> > Maybe this is due to a buggy BIOS implementation. We also check the
> > code in disk/i386/pc/biosdisk.c
> > or (stage2/bios.c) for grub-0.97, and for code in function
> > grub_biosdisk_open (or get_diskinfo in grub-0.97),
> > there is a standard int13 check of disk after check with ext int13,
> >
> >   if(drive & 0x80) {
> >      ...
> >     version = grub_biosdisk_check_int13_extension();
> >       if(version) {
> >          grub_biosdisk_get_diskinfo_int13_extensions();
> >       }
> >    }
> >
> >    grub_biosdisk_get_diskinfo_standard()
> >
> > We're wondering does it really need to call
> > grub_biosdisk_get_diskinfo_standard() again if we have get the
> > necessary information
> > from ext int13? I think we call the later only if ext int13 failed for
> > some reason. Especially in our case the code should be .
> >
> >   if(drive & 0x80) {
> >      ...
> >     version = grub_biosdisk_check_int13_extension();
> >       if(version) {
> >          grub_biosdisk_get_diskinfo_int13_extensions();
> >       } else {
> >          grub_biosdisk_get_diskinfo_standard()
> >       }
> >    }
> >
> > and it do works!
>
> Thanks for your report, Zhixu. Does this patch work?
>
> Index: disk/i386/pc/biosdisk.c
> ===================================================================
> RCS file: /cvsroot/grub/grub2/disk/i386/pc/biosdisk.c,v
> retrieving revision 1.9
> diff -u -p -r1.9 biosdisk.c
> --- disk/i386/pc/biosdisk.c     29 Jul 2006 10:11:01 -0000      1.9
> +++ disk/i386/pc/biosdisk.c     14 Nov 2006 03:39:47 -0000
> @@ -133,21 +133,24 @@ grub_biosdisk_open (const char *name, gr
>         }
>      }
>
> -  if (grub_biosdisk_get_diskinfo_standard (drive,
> -                                          &data->cylinders,
> -                                          &data->heads,
> -                                          &data->sectors) != 0)
> +  if (! total_sectors)
>      {
> -      grub_free (data);
> -      return grub_error (GRUB_ERR_BAD_DEVICE, "cannot get C/H/S values");
> -    }
> +      /* FDD, or HDD without int13 extensions.  */
> +      if (grub_biosdisk_get_diskinfo_standard (drive,
> +                                              &data->cylinders,
> +                                              &data->heads,
> +                                              &data->sectors) != 0)
> +       {
> +         grub_free (data);
> +         return grub_error (GRUB_ERR_BAD_DEVICE, "cannot get C/H/S
> values");
> +       }
>
> -  if (! total_sectors)
> -    total_sectors = data->cylinders * data->heads * data->sectors;
> +      total_sectors = data->cylinders * data->heads * data->sectors;
> +    }
>
>    disk->total_sectors = total_sectors;
>    disk->data = data;
> -
> +
>    return GRUB_ERR_NONE;
> }
>
>
> -Hollis
>
>
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> http://lists.gnu.org/mailman/listinfo/grub-devel
>



-- 
Zhixu Liu

[-- Attachment #2: Type: text/html, Size: 7387 bytes --]

Re: does disk geometry depend on partition table?

[Hollis Blanchard]

On Tue, 2006-11-14 at 13:03 +0800, Zhixu Liu wrote:
> Hi, Hollis,
> 
> We've tried something like this, and found it works for some platform,
> but still failed for one platform.
> Indeed, in this weird platform, we get both wrong CHS value with and
> without partition table. But the OS always 
> get the correct value. The problem is due to either BIOS or Disk for
> this particular platform, we think. But we're not clear.

That platform must have a BIOS bug and we have no known workaround. :/

> Anyway, the patch works for some cases, and should for most cases.

OK. I will commit this soon unless I hear an objection.

-Hollis