RE: GPT probe: signature vs checksum

[Matt_Domsch]

> > but I want to follow the spec (or work w/ Intel to change it like we
> > did the pmbr test).
> 
> Why?  I guess "following the spec" is always going to be a religious
> issue, but I always prefer "what works best".

Sure, no argument here.  I just like documentation when multiple parties
will be implementing something that's supposed to interoperate. :-)
 
> Can I contribute somehow to the spec?

Probably.  http://developer.intel.com/technology/efi/efi.htm has the spec.
http://developer.intel.com/technology/efi/feedback.htm is to submit
feedback, but I've never gotten a response out of that.

address@hidden and address@hidden are two people at
Intel working on IA-64 Linux, and it was Asit that first contacted me about
the PMBR change.  That's your best bet.

-         if (ped_device_read(dev, legacyMbr, 0, 1)) {
+         if (ped_device_read(dev, &legacyMbr, 0, 1)) {
I think.

> Nice and simple, and should deal with all cases cleanly (eg: bad first
> sector, pmbr changed).  Why can't that go in the spec?
> 
> i.e. define GPT present (possibly corrupt) vs GPT valid.

That looks good to me.  Then the test for valid GPT really moves into
gpt_open() calling gpt_read(), where if one is valid and the other invalid,
it gets fixed up automatically.  I don't think the EFI Spec would need to be
changed here then, it's an implementation detail that we probe before
testing for valid.  OK.

Looking at 1.4.20 code right now...
ped_disk_open() calls ped_disk_probe() calls gpt_probe(), then calls
gpt_open().  gpt_open() calls gpt_probe() again (that shouldn't be
necessary, right?) and then gpt_read().  In all cases, simplifying
gpt_probe() looks like the right thing to do.

FYI, I'm testing a kernel patch from Red Hat that will allow us to drop the
#if linux read/write last odd sector of a disk kludge from parted.  The
kludge gets moved into the kernel. :-)

Thanks,
Matt

--
Matt Domsch
Sr. Software Engineer
Dell Linux Solutions
www.dell.com/linux
#1 US Linux Server provider with 24% (IDC Sept 2001)
#2 Worldwide Linux Server provider with 17% (IDC Sept 2001)
#3 Unix provider with 18% in the US (Dataquest)!