Re: [PATCH] Ensure boot CD can be accessed

[Bean]

On Tue, Mar 4, 2008 at 4:46 AM, Christian Franke
<address@hidden> wrote:
> Bean wrote:
>  > On Sat, Feb 16, 2008 at 5:13 AM, Christian Franke
>  > <...> wrote:
>  >
>  >> ...
>
> >>  Further testing outdated this patch. The boot CD appeared under the
>  >>  following drive numbers:
>  >>
>  >>  0x82 (old Toshiba Laptop with unknown BIOS)
>  >>  0x9f (PC with Phoenix-Award BIOS, VMware)
>  >>  0xe0 (VirtualBox)
>  >>  0xef (PC with AMI BIOS, VirtualPC)
>  >>
>  >>  Current CVS would only work for 0xe0, with the previous patch it works
>  >>  for all >= 0x9f.
>  >>
>  >>  This new patch also handles the case that the CD appears under the
>  >>  legacy controller drive number (0x82 = secondary master). It is accessed
>  >>  as "hd2" in this case.
>  >>
>  >>  Christian
>  >>
>  >>  ...
>
> >>
>  >
>  > I'm thinking, is there a more generic way to test for cd drive ? also,
>  > if the cd drive can appear as (hd2), maybe we can use (hdx) for other
>  > cases as well, such as (hd96), etc. The name might be a little
>  > strange, but we can get the bios drive number quite easily, which can
>  > be useful sometimes (some boot loader need to know the boot drive
>  > number).
>  >
>  >
>
>  Good point.
>
>  Besides the heuristics "(boot drive >= 0x9f) => (likely a boot CD)",
>  there is apparently no clean method to distinguish CD from HDD on the
>  legacy BIOS call level.
>
>  Is the special handling for CD in grub_biosdisk_rw() necessary, or can
>  it be generalized to work with both CD and HDD? In the latter case, no
>  actual CD detection is necessary.
>
>  So is would be probably better to remove the (cdN) device names and
>  accept funny names like e.g. (hd111) for boot CD 0xef. It also would be
>  easier to track down problems when the boot device number is no longer
>  hidden behind (cd0).
>
>  BTW: None of the BIOSes I tested provided access to a second (non-boot)
>  CD drive.

I take a look at grub legacy, it uses INT 13 FUNC 4B01 to check for
cdrom, maybe i can write a patch later.

-- 
Bean