Skip to contentSkip to conversation

Skip Apple ghosts

PATCH In review Vladimir 'φ-coder/phcoder' Serbinenko 2 participants 5 messages Dec 24, 2013 – Jan 7, 2014

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Skip Apple ghosts

Hello, all. It was discovered that on macs sometimes firmware defines ghost disks with vendor suffix. E.g. /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/EndEntire
is a normal disks with partitions presented with HD(...) but then /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/MediaVendor(Apple)[0:
]/EndEntire
is a ghost disk. It has as suffix a vendor path with empty vendor data. This is a problem because when chainloading on such disks GRUB can't find handle of partition as none is defined. I propose to dkip ghosts completely. This should be safe as the skip happens only for empty Apple vendor suffix and if the parent is already a known disk.
Did anyone see anything similar?

grub-core/disk/efi/efidisk.c +15 -0
index e04203f..3b12c34 100644
@@ -333,6 +333,21 @@ name_devices (struct grub_efidisk_data *devices)
       if (! dp)
 	continue;
 
+      /* Ghosts proudly presented by Apple.  */
+      if (GRUB_EFI_DEVICE_PATH_TYPE (dp) == GRUB_EFI_MEDIA_DEVICE_PATH_TYPE
+	  && GRUB_EFI_DEVICE_PATH_SUBTYPE (dp)
+	  == GRUB_EFI_VENDOR_MEDIA_DEVICE_PATH_SUBTYPE)
+	{
+	  grub_efi_vendor_device_path_t *vendor = (grub_efi_vendor_device_path_t *) dp;
+	  const struct grub_efi_guid apple = GRUB_EFI_VENDOR_APPLE_GUID;
+
+	  if (vendor->header.length == sizeof (*vendor)
+	      && grub_memcmp (&vendor->vendor_guid, &apple,
+			      sizeof (vendor->vendor_guid)) == 0
+	      && find_parent_device (devices, d))
+	    continue;
+	}
+
       m = d->block_io->media;
       if (GRUB_EFI_DEVICE_PATH_TYPE (dp) == GRUB_EFI_ACPI_DEVICE_PATH_TYPE
 	  && GRUB_EFI_DEVICE_PATH_SUBTYPE (dp)
#1 lore raw

В Вт, 24/12/2013 в 14:26 +0100, Vladimir 'φ-coder/phcoder' Serbinenko пишет:

Quoted 7 lines

Hello, all. It was discovered that on macs sometimes firmware defines ghost disks with vendor suffix. E.g. /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/EndEntire
is a normal disks with partitions presented with HD(...) but then /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/MediaVendor(Apple)[0:
]/EndEntire
is a ghost disk. It has as suffix a vendor path with empty vendor data.

Messaging subtype 12 is IPv4 Device Path. I suppose it is related to NetBoot.

Quoted 4 lines

This is a problem because when chainloading on such disks GRUB can't find handle of partition as none is defined. I propose to dkip ghosts completely. This should be safe as the skip happens only for empty Apple vendor suffix and if the parent is already a known disk.

Hmm ... should not we simply ignore all unknown media types? They can be added on case by case basis I guess.

grub-core/disk/efi/efidisk.c +1 -0
index e04203f..6dfdb0d 100644
@@ -310,6 +310,7 @@ name_devices (struct grub_efidisk_data *devices)
              grub_efi_print_device_path (d->device_path);
 #endif
              /* For now, ignore the others.  */
+             d->last_device_path = 0;
              break;
            }
        }
Quoted 5 lines

Did anyone see anything similar?
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel

#2 lore raw
Quoted 10 lines · Andrey Borzenkov

В Вт, 24/12/2013 в 14:26 +0100, Vladimir 'φ-coder/phcoder' Serbinenko пишет:

Quoted 7 lines

Hello, all. It was discovered that on macs sometimes firmware defines ghost disks with vendor suffix. E.g. /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/EndEntire
is a normal disks with partitions presented with HD(...) but then /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/MediaVendor(Apple)[0:
]/EndEntire
is a ghost disk. It has as suffix a vendor path with empty vendor data.

Quoted 2 lines

Messaging subtype 12 is IPv4 Device Path. I suppose it is related to NetBoot.

It's hex. 0x12 is SATA.

Quoted 9 lines
Quoted 5 lines

This is a problem because when chainloading on such disks GRUB can't find handle of partition as none is defined. I propose to dkip ghosts completely. This should be safe as the skip happens only for empty Apple vendor suffix and if the parent is already a known disk.

Hmm ... should not we simply ignore all unknown media types? They can be added on case by case basis I guess.

Too much breakage potential. And I don't have accessto check it on loads of different buggy EFI systems and even if I did, it would be a waste of time.

Quoted 28 lines
grub-core/disk/efi/efidisk.c +1 -0
index e04203f..6dfdb0d 100644
@@ -310,6 +310,7 @@ name_devices (struct grub_efidisk_data *devices)
              grub_efi_print_device_path (d->device_path);
 #endif
              /* For now, ignore the others.  */
+             d->last_device_path = 0;
              break;
            }
        }
Quoted 7 lines

Did anyone see anything similar?
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel

_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel

#3 lore raw
In reply to Vladimir 'φ-coder/phcoder' Serbinenko

В Вт, 24/12/2013 в 14:26 +0100, Vladimir 'φ-coder/phcoder' Serbinenko пишет:

Quoted 4 lines

Hello, all. It was discovered that on macs sometimes firmware defines ghost disks with vendor suffix. E.g. /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/EndEntire is a normal disks with partitions presented with HD(...) but then

Quoted 8 lines

/ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/MediaVendor(Apple)[0:
]/EndEntire
is a ghost disk. It has as suffix a vendor path with empty vendor data. This is a problem because when chainloading on such disks GRUB can't find handle of partition as none is defined. I propose to dkip ghosts completely. This should be safe as the skip happens only for empty Apple vendor suffix and if the parent is already a known disk.
Did anyone see anything similar?

Yes (on MacBook Pro);

Here is what I get when EFI-booting from GRUB rescue CD made with single x86_64-efi target:

/ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/EndEntire
disk
block /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/MediaVendor(2b0585eb-d8b8-49a9-8b8c-e21b01aef2b7)[0: ]/EndEntire block
simple FS /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(2,10,5c1,0000000000000000,20,0)/EndEntire disk
block /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(3,5d1,ca3,0000000000000000,20,0)/EndEntire simple FS
disk
block /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(4,1274,a2,0000000000000000,20,0)/EndEntire disk
block

The three HD media paths are for three Apple partitions (and the third one has simple FS because it does contains HFS+). And I /think/ that the very first Vendor media path refers to CD-ROM (based on the fact that it got simple FS protocol). I played a bit with it; when I try to list filesystem for this media path in EFI shell, it audibly tries to access CD-ROM but fails. May be it requires Apple ISO9660 extensions, but I do not know how to create EFI bootable image containing Apple extensions (mkisofs should support it, but I was lost in arguments. xorriso apparently does not support Apple extensions).

In your case you have SATA DVD drive, right? And path shown refers to it?

#4 lore raw
In reply to Vladimir 'φ-coder/phcoder' Serbinenko

В Вт, 24/12/2013 в 14:26 +0100, Vladimir 'φ-coder/phcoder' Serbinenko пишет:

Quoted 4 lines

Hello, all. It was discovered that on macs sometimes firmware defines ghost disks with vendor suffix. E.g. /ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/EndEntire is a normal disks with partitions presented with HD(...) but then

Quoted 8 lines

/ACPI(a0341d0,0)/PCI(2,1f)/UnknownMessaging(12)/MediaVendor(Apple)[0:
]/EndEntire
is a ghost disk. It has as suffix a vendor path with empty vendor data. This is a problem because when chainloading on such disks GRUB can't find handle of partition as none is defined. I propose to dkip ghosts completely. This should be safe as the skip happens only for empty Apple vendor suffix and if the parent is already a known disk.
Did anyone see anything similar?

Yes (on MacBook Pro);

Here is what I get when EFI-booting from GRUB rescue CD made with single x86_64-efi target:

/ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/EndEntire
disk
block /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/MediaVendor(2b0585eb-d8b8-49a9-8b8c-e21b01aef2b7)[0: ]/EndEntire block
simple FS /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(2,10,5c1,0000000000000000,20,0)/EndEntire disk
block /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(3,5d1,ca3,0000000000000000,20,0)/EndEntire simple FS
disk
block /ACPI(a0341d0,0)/PCI(1,1f)/ATAPI(0,0,0)/HD(4,1274,a2,0000000000000000,20,0)/EndEntire disk
block

The three HD media paths are for three Apple partitions (and the third one has simple FS because it does contains HFS+). And I /think/ that the very first Vendor media path refers to CD-ROM (based on the fact that it got simple FS protocol). I played a bit with it; when I try to list filesystem for this media path in EFI shell, it audibly tries to access CD-ROM but fails. May be it requires Apple ISO9660 extensions, but I do not know how to create EFI bootable image containing Apple extensions (mkisofs should support it, but I was lost in arguments. xorriso apparently does not support Apple extensions).

In your case you have SATA DVD drive, right? And path shown refers to it?