On Mon, 10 Feb, at 03:23:33PM, Madper Xie wrote:
> Howdy,
>
> With old kernel (from 3.10 to 3.14-rc1), my hp box shows following
> outputs:
> ~~~
> [ 0.009166] Freeing SMP alternatives memory: 20K (ffffffff82234000 - ffffffff82239000)
> [ 0.010302] ioremap: invalid physical address 1376e01800000001
> [ 0.010303] ------------[ cut here ]------------
> [ 0.010308] WARNING: CPU: 0 PID: 0 at arch/x86/mm/ioremap.c:83 __ioremap_caller+0x357/0x380()
OK, this looks like a bogus BGRT address, that's a 64-bit physical
address. I suspect the BGRT address is actually 0x1376e018, but that's
just a guess. This looks like a BIOS bug.
> ~~~
> However the box can boot up and works well.
> After I update kernel to 3.14-rc2, it panic when booting. Here is
> console output:
> ~~~
> [ 0.010366] BUG: unable to handle kernel paging request at ffffffffff340003
> [ 0.017376] IP: [<ffffffff81d85ba4>] efi_bgrt_init+0x9d/0x133
> [ 0.023153] PGD 159eb067 PUD 159ed067 PMD 16240067 PTE 9376e01800000163
> [ 0.029842] Oops: 0009 [#1] SMP
Well, nuts. I think this is my bad. Does reverting commit 081cd62a010f
("x86/efi: Allow mapping BGRT on x86-32") make things work? Or at least,
work as well as they did prior to -rc2?
I wonder if we're calling early_memremap() when the fixmap slots are no
longer valid. Could you try building the 'urgent' branch at,
git://git.kernel.org/pub/scm/linux/kernel/git/mfleming/efi.git
--
Matt Fleming, Intel Open Source Technology Center
--
To unsubscribe from this list: send the line "unsubscribe linux-efi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
