running pygrub on split-partition disk layouts.

running pygrub on split-partition disk layouts.

[Ian Campbell]

An absolute age ago Michael posted a patch to allow pygrub to work with
split-partition layouts (i.e. those where the guests disk cfg refers to
xvda1, xvda2, etc rather than an xvda with a partition table). See
http://lists.xen.org/archives/html/xen-devel/2011-01/msg02076.html
(patch itself is below).

Apparently at the time I was of the opinion that this was a reasonable
workaround for the underlying issue (although I seem to have thought I
had a better solution, not so sure right now...).

I'm not sure why it didn't go in (in the likely event I just dropped the
ball -- sorry!).

Anyway, this resurfaced recently in the context of Debian bug #745419:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=745419#27

#745419 seems to be unrelated to Michael's issue, it is a full disk
image case with an MBR etc.

Michael, are you still carrying this patch to the Fedora packages? If so
presumably it is without issue in the field so to speak?

Should the patch be resurrected for either 4.5 or 4.6? Or does anyone
have any other insights into the various issues?

Ian.

--- xen-4.1.0/tools/pygrub/src/pygrub.orig	2010-12-31 15:24:11.000000000 +0000
+++ xen-4.1.0/tools/pygrub/src/pygrub	2011-01-30 18:58:17.000000000 +0000
@@ -96,6 +96,7 @@
 
     fd = os.open(file, os.O_RDONLY)
     buf = os.read(fd, 512)
+    offzerocount = 0
     for poff in (446, 462, 478, 494): # partition offsets
 
         # MBR contains a 16 byte descriptor per partition
@@ -105,6 +106,7 @@
         
         # offset == 0 implies this partition is not enabled
         if offset == 0:
+            offzerocount += 1
             continue
 
         if type == FDISK_PART_SOLARIS or type == FDISK_PART_SOLARIS_OLD:
@@ -123,6 +125,9 @@
         else:
             part_offs.append(offset)
 
+    if offzerocount == 4:
+        # Might be a grub boot sector pretending to be an MBR
+        part_offs.append(0)
     return part_offs
 
 class GrubLineEditor(curses.textpad.Textbox):

Re: running pygrub on split-partition disk layouts.

[M A Young]

On Fri, 21 Nov 2014, Ian Campbell wrote:

> An absolute age ago Michael posted a patch to allow pygrub to work with
> split-partition layouts (i.e. those where the guests disk cfg refers to
> xvda1, xvda2, etc rather than an xvda with a partition table). See
> http://lists.xen.org/archives/html/xen-devel/2011-01/msg02076.html
> (patch itself is below).
>
> Apparently at the time I was of the opinion that this was a reasonable
> workaround for the underlying issue (although I seem to have thought I
> had a better solution, not so sure right now...).
>
> I'm not sure why it didn't go in (in the likely event I just dropped the
> ball -- sorry!).
>
> Anyway, this resurfaced recently in the context of Debian bug #745419:
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=745419#27
>
> #745419 seems to be unrelated to Michael's issue, it is a full disk
> image case with an MBR etc.
>
> Michael, are you still carrying this patch to the Fedora packages? If so
> presumably it is without issue in the field so to speak?

Yes, it is still in Fedora (with a minor change to get it to apply as a 
line was added in the surrounding code). I haven't seen any bug reports 
relating to this patch.

 	Michael Young

Re: running pygrub on split-partition disk layouts.

[Konrad Rzeszutek Wilk]

On Fri, Nov 21, 2014 at 02:18:26PM +0000, Ian Campbell wrote:
> An absolute age ago Michael posted a patch to allow pygrub to work with
> split-partition layouts (i.e. those where the guests disk cfg refers to
> xvda1, xvda2, etc rather than an xvda with a partition table). See
> http://lists.xen.org/archives/html/xen-devel/2011-01/msg02076.html
> (patch itself is below).
> 
> Apparently at the time I was of the opinion that this was a reasonable
> workaround for the underlying issue (although I seem to have thought I
> had a better solution, not so sure right now...).
> 
> I'm not sure why it didn't go in (in the likely event I just dropped the
> ball -- sorry!).
> 
> Anyway, this resurfaced recently in the context of Debian bug #745419:
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=745419#27
> 
> #745419 seems to be unrelated to Michael's issue, it is a full disk
> image case with an MBR etc.
> 
> Michael, are you still carrying this patch to the Fedora packages? If so
> presumably it is without issue in the field so to speak?

ping?
> 
> Should the patch be resurrected for either 4.5 or 4.6? Or does anyone
> have any other insights into the various issues?
> 
> Ian.
> 
> --- xen-4.1.0/tools/pygrub/src/pygrub.orig	2010-12-31 15:24:11.000000000 +0000
> +++ xen-4.1.0/tools/pygrub/src/pygrub	2011-01-30 18:58:17.000000000 +0000
> @@ -96,6 +96,7 @@
>  
>      fd = os.open(file, os.O_RDONLY)
>      buf = os.read(fd, 512)
> +    offzerocount = 0
>      for poff in (446, 462, 478, 494): # partition offsets
>  
>          # MBR contains a 16 byte descriptor per partition
> @@ -105,6 +106,7 @@
>          
>          # offset == 0 implies this partition is not enabled
>          if offset == 0:
> +            offzerocount += 1
>              continue
>  
>          if type == FDISK_PART_SOLARIS or type == FDISK_PART_SOLARIS_OLD:
> @@ -123,6 +125,9 @@
>          else:
>              part_offs.append(offset)
>  
> +    if offzerocount == 4:
> +        # Might be a grub boot sector pretending to be an MBR
> +        part_offs.append(0)
>      return part_offs
>  
>  class GrubLineEditor(curses.textpad.Textbox):
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel