[HD64461_PCMCIA_driver] - Shared IRQ lines / Segfaults

[Manuel Lauss]

On Wed, Jan 30, 2008 at 01:20:36PM +0100, Kristoffer Ericson wrote:
> Greetings,
> 
> Driver detects card insertion / ejection and sets power properly. The mapping looks resonable also compairing to a working 2.6.17 driver.
> modprobe hostap_cs
> hostap_cs: Registered netdevice wifi0
> Unable to handle kernel NULL pointer dereference at virtual address 00000028
> pc = 8d0031b8
> *pde = 00000000
> Oops: 0000 [#1]
> Modules linked in: hostap_cs hostap orinoco_cs orinoco hermes atmel_cs atmel
> 
> Pid : 283, Comm:             modprobe
> PC is at generic_inw+0x8/0x20
> PC  : 8d0031b8 SP  : 8ecaba74 SR  : 400000f0 TEA : c0045040    Not tainted
> R0  : 00000028 R1  : 8d2d9030 R2  : 8d2ac050 R3  : 00000007
> R4  : 00000028 R5  : 00000002 R6  : 00000080 R7  : 8ecf06a0
> R8  : 8ee62000 R9  : 00008a32 R10 : c002a1b0 R11 : 8ee62000
> R12 : 00000080 R13 : 8ee62000 R14 : 8ecf06a0
> MACH: 0000458f MACL: 00001110 GBR : 296ce450 PR  : 8d0031b8
> 
> Call trace: 
> [<c002a1fa>] prism2_interrupt+0x4a/0xc50 [hostap_cs]

The oops output suggests it crashes in 
drivers/net/wireless/hostap/hostap_hw.c::prism2_check_magic(),
specifically when it tries to read the HF384X_SWSUPPORT0_OFF register
(which ist at offset 0x28, just like the oopsing address).
So, looking at the HFA384X_INW macro, it seems dev->base_addr is NULL.
I suspect this is because of set_io_map() callback in the hd64461.c
socket driver.  The comment there says IO area is statically mapped,
yet it returns null; I suggest you let it return the base address of
the IO area instead.

-- 
 Manuel Lauss