Re: [PATCH] Fix LVM/RAID probing without device.map

["Vladimir 'φ-coder/phcoder' Serbinenko" <phcoder@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 1531 bytes --]


util/deviceiter.c is missing from your patch.
> @@ -136,6 +138,22 @@ probe (const char *path, char *device_na
>        goto end;
>      }
>  
> +  if (stat (dev_map, &dev_map_stat) == -1 &&
> +      grub_util_get_dev_abstraction (device_name) != GRUB_DEV_ABSTRACTION_NONE)
>   
Looks like it may call stat with NULL argument. I would rather avoid
doing this than rely on stat to detect it cleanly and not just segfault
> +    {
> +      /* If we don't have a device map, then we won't yet know about the
> +         physical volumes underlying this device, so probe all devices.  */
> +      grub_util_iterate_devices (grub_util_biosdisk_probe_device, 0);
> +
> +      /* Now reinitialise the higher layers.  */
> +      grub_lvm_fini ();
> +      grub_mdraid_fini ();
> +      grub_raid_fini ();
> +      grub_raid_init ();
> +      grub_mdraid_init ();
> +      grub_lvm_init ();
>   
Can we not to initialise those levels before they should be inited
rather than reinit? If it requires unclean workarounds I would rather
prefer this since it's relatively clean (approx rmmod+insmod)
> +int
> +grub_util_biosdisk_probe_device (const char *name, int is_floppy)
> +{
> +  char *grub_disk;
> +
> +  if (is_floppy)
> +    grub_disk = xasprintf ("fd%d", num_fd++);
> +  else
> +    grub_disk = xasprintf ("hd%d", num_hd++);
> +
>   
You can also choose something more straightforward for name generation.
Like just use 'name'.




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



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 293 bytes --]