[RFC v3] PCMCIA locking updates for 2.6.34

[Dominik Brodowski]

Hey,

On Thu, Feb 11, 2010 at 06:39:37PM +0900, Komuro wrote:
> Here is the output of dmesg.

This is most interesting.

> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x51, 5)
> pcmcia_socket pcmcia_socket0:   0x00 0x91 0x00 0x00 ...
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x91, 5)
> pcmcia_socket pcmcia_socket0:   0x13 0x03 0x43 0x49 ...

This is how it should look like. But then, all of a sudden:

> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x8d, 2)
> pcmcia_socket pcmcia_socket0:   0xff 0x9c 0x02 0x40 ...
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x51, 5)
> pcmcia_socket pcmcia_socket0:   0xff 0x9c 0x02 0x40 ...
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x51, 5)
> pcmcia_socket pcmcia_socket0:   0x00 0x91 0x00 0xff ...
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0xffff0091, 5)
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x7fff8048, 5)
> pcmcia_socket pcmcia_socket0:   0x00 0x91 0x00 0x00 ...
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x91, 5)
> pcmcia_socket pcmcia_socket0: pcmcia_read_cis_mem(1, 0x2, 2)

Things become confused. addr should never become this large, AFAICS
(well, surely not above >0xffff). However, we also get some concurrent
accesses, which might be the cause of this. 

Finally, it doesn't seem to be v4 (the following line is no longer in the
code), so could you re-run the test (with the debug patch applied), please,
with the code found (now) at
git://git.kernel.org/pub/scm/linux/kernel/git/brodo/pcmcia-2.6.git ? If the
problem persists, please send the dmesg -s 200000 output to me again.

> pcmcia_socket pcmcia_socket0: re-scanning socket 0

Thanks for all your effort in improving the Linux PCMCIA subsystem!

Best,
	Dominik