On 2023-05-11 14:35, Geert Uytterhoeven wrote:

Hi Arnd,

CC Artur, who's working on HP Jornada 680.

Thanks for CC'ing me - I faced this exact issue while working on my
(still not upstreamed) hd6446x PCMCIA controller driver. The PCMCIA
subsystem uses `inb/outb`, which expect the `sh_io_port_base` to be set
to something else than the default `-1`. At first I tried to set it to
`0xa0000000`, so that all I/O goes through the fixed, non-cacheable P2
area. That however broke some other driver code (I had no time to debug
which one). Eventually I ended up taking a suggestion from a MIPS PCMCIA
driver [1] and simply substract the broken `sh_io_port_base` address
from `HD64461_IOBASE`, as the base for `socket.io_offset`. This way all
the PCMCIA `inb/outb` accesses are absolute, no matter what the
`sh_io_port_base` is set to. This of course is a very ugly solution and
we should instead fix the root cause of this mess. I will have a better
look at this patch set and the problem at hand at a later date.

Cheers,
Artur

[1] 
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/pcmcia/db1xxx_ss.c?h=v6.4-rc1#n527

On Wed, May 10, 2023 at 5:55 PM Arnd Bergmann [off-list ref] wrote:

quoted

On Wed, May 10, 2023, at 16:27, Thomas Zimmermann wrote:

quoted

Am 10.05.23 um 16:15 schrieb Arnd Bergmann:

quoted

On Wed, May 10, 2023, at 16:03, kernel test robot wrote:

quoted

quoted

I think that's a preexisting bug and I have no idea what the
correct solution is. Looking for HD64461 shows it being used
both with inw/outw and readw/writew, so there is no way to have
the correct type. The sh __raw_readw() definition hides this bug,
but that is a problem with arch/sh and it probably hides others
as well.

The constant HD64461_IOBASE is defined as integer at


https://elixir.bootlin.com/linux/latest/source/arch/sh/include/asm/hd64461.h#L17

but fb_readw() expects a volatile-void pointer. I guess we could add a
cast somewhere to silence the problem. In the current upstream code,
that appears to be done by sh's __raw_readw() internally:


https://elixir.bootlin.com/linux/latest/source/arch/sh/include/asm/io.h#L35

Sure, that would make it build again, but that still doesn't make the
code correct, since it's completely unclear what base address the
HD64461_IOBASE is relative to. The hp6xx platform code only passes it
through inw()/outw(), which take an offset relative to 
sh_io_port_base,
but that is not initialized on hp6xx. I tried to find in the history
when it broke, apparently that was in 2007 commit 34a780a0afeb ("sh:
hp6xx pata_platform support."), which removed the custom inw/outw
implementations.

See also commit 4aafae27d0ce73f8 ("sh: hd64461 tidying."), which
claims they are no longer needed.

Don't the I/O port macros just treat the port as an absolute base 
address
when sh_io_port_base isn't set?

Gr{oetje,eeting}s,

                        Geert