Re: [PATCH v1 3/6] driver: Google EFI SMI

From: Matt Domsch <Matt_Domsch@dell.com>
To: Mike Waychison <mikew@google.com>
Cc: Greg KH <greg@kroah.com>,
	torvalds@linux-foundation.org, San Mehat <san@google.com>,
	Aaron Durbin <adurbin@google.com>,
	Duncan Laurie <dlaurie@google.com>,
	linux-kernel@vger.kernel.org, Tim Hockin <thockin@google.com>
Subject: Re: [PATCH v1 3/6] driver: Google EFI SMI
Date: Sat, 19 Feb 2011 22:44:40 -0600	[thread overview]
Message-ID: <20110220044440.GA28167@auslistsprd01.us.dell.com> (raw)
In-Reply-To: <AANLkTim27Y+TKa54jhC1VuqXzjEzX150ysGoUKNJrBmk@mail.gmail.com>

On Thu, Jan 27, 2011 at 03:41:06PM -0800, Mike Waychison wrote:
> On Wed, Jan 26, 2011 at 5:22 PM, Mike Waychison <mikew@google.com> wrote:
> 
> I've spent the last few hours looking at efivars.c and working out how
> I can refactor it to reuse all the kobject bits it uses.  Does anybody
> use this thing though?
> 
> I can't believe I was just lectured for crappy ABI when this thing
> takes a binary packed struct on write() and process it:
>    - without regard to write length, and
>    - in a way that isn't compatible across compat (both DataSize and
> Status are unsigned long!).
> 
> struct efi_variable {
>         efi_char16_t  VariableName[1024/sizeof(efi_char16_t)];
>         efi_guid_t    VendorGuid;
>         unsigned long DataSize;
>         __u8          Data[1024];
>         efi_status_t  Status;
>         __u32         Attributes;
> } __attribute__((packed));
> 
> :(

I wrote that long before anyone believed there could be a 32-bit EFI
or 32-bit kernel running on a 64-bit EFI.  Remember, this originated
on Itanium.  I knew it was ugly, but the EFI spec itself defined the
DataSize and Status as they did.

-- 
Matt Domsch
Technology Strategist
Dell | Office of the CTO