Re: ohci1394_dma=early crash since 2.6.32 (was Re: [Bug #14487] PANIC:early exception 08 rip 246:10 error ffffffff810251b5 cr2 0)
From: Justin P. Mattock
Date: Mon Feb 01 2010 - 18:50:31 EST
On 02/01/10 14:27, Stefan Richter wrote:
Justin P. Mattock wrote:
(as for yesterdays 0xffffffffffffffff(just experimenting)Google gives me
no info on the differences between 8f's to 16f's, I was under the
impression that it's x86_32 and x86_64 for the pci address).
As Dan noted,
(class == 0xffffffff || 0xffffffffffffffff)
is always true because it is logically the same as
(class == whatever) || true
If you really meant
class == 0xffffffff || class == 0xffffffffffffffff
yeah that's what I was going for(just to see).
then the latter half will never become true because class is declared as
u32 and got its value from read_pci_config() which also returns u32.
That's what I was afraid of. I'm guessing there probably would be a lot
of things to change for(if this correct) u64.
BTW, whether a PCI device is capable of accessing 32 bit bus addresses
or also 64 bit bus addresses depends on the device, not on the CPU.
Originally, PCI only had a 32 bit addressing model. OHCI 1394 1.0/1.1
implementations only deal with 32 bit local bus addresses.
I haven't even looked at what the device was capable of doing.
The 'class' however is not an address but merely a register value with
24 bits width. (Defined in the PCI Local Bus spec which is not freely
available, cited in OHCI 1394 annex A.3.) This register is read as a 32
bits wide register from which the excess byte is later discarded. If
all bits read are 1, the bus:slot:function is not actually populated.
So(correct me if I'm wrong), I'm generating a 64 bit register
and the kernel is looking for a 32 bit register causing the crash.
Justin P. Mattock
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/