Re: [linuxsh-dev] PCMCIA patch & question

[Greg B. <gb...@po...>]

NIIBE Yutaka wrote:
> 
> My network connection is quite bad these days, because of the
> restructuring of my company.  I think I can be "in sync" this weekend.

  I wondered where you'd been ;-)

> 
> For the patch by Greg, here's my comment.

  Thanks.

> 
> +++ kernel/arch/sh/config.in    Fri Sep 29 03:13:26 2000
> +++ kernel/arch/sh/kernel/Makefile      Wed Sep 27 03:28:54 2000
> +++ kernel/arch/sh/kernel/sh_ksyms.c    Tue Oct 10 19:56:39 2000
> +++ kernel/include/asm-sh/io.h  Wed Sep 27 10:11:34 2000
> +++ kernel/include/asm-sh/mmu_context.h Sun Oct  1 13:18:06 2000
> +++ kernel/include/asm-sh/machvec.h     Sat Sep 23 05:34:00 2000
> +++ kernel/include/asm-sh/io_hd64461.h  Sat Sep 23 05:21:26 2000
> +++ kernel/arch/sh/mm/fault.c   Sun Oct  1 13:56:02 2000
> 
> OK.
> 
> +++ kernel/arch/sh/kernel/io_hd64465.c  Sun Oct  1 04:57:59 2000
> +++ kernel/arch/sh/kernel/mach_foobar.c Sat Sep 30 10:59:41 2000
> +++ kernel/arch/sh/kernel/setup_hd64465.c       Tue Oct 10 20:02:53 2000
> +++ kernel/include/asm-sh/hd64465.h     Sat Sep 30 11:00:17 2000
> +++ kernel/include/asm-sh/io_hd64465.h  Tue Oct 10 20:18:29 2000
> 
> OK.
> 
> +++ kernel/drivers/pcmcia/Config.in     Mon Sep 25 23:41:41 2000
> +++ kernel/drivers/pcmcia/Makefile      Mon Sep 25 23:54:05 2000
> +++ kernel/drivers/pcmcia/hd64465_ss.c  Tue Oct 10 20:14:28 2000
> 
> OK for me.  (Actually I don't care. :-)  Could you please send this part
> to PCMCIA driver maintainer directly?

  I'll do that once we resolve the dispatch_virtual_irq() issue.

> 
> Please check them in.
> ----------------------------------------------
> 
> +++ kernel/arch/sh/kernel/io.c  Fri Sep 29 09:44:43 2000
> +++ kernel/arch/sh/kernel/io_generic.c  Tue Sep 26 01:59:37 2000
> 
> For those, I'm not sure now.  EXPORT_SYMBOL shoule be here or in sh_ksyms.c.

  I wasn't sure either, and it doesn't particularly bother me where
they go, as long as they do go somewhere.  Your call.

> 
> +++ kernel/include/asm-sh/irq.h Tue Oct 10 22:27:45 2000
> 
> Basically OK.  But I'm not sure about the implementation of
> dispatch_virtual_irq.

  Neither am I   ;-)  I'm pretty unhappy about needing to use it
at all, but that need arises from the braindead nature of the HD64465
PCMCIA features.  Note that the current code is my *third* attempt at
delivering IRQs properly.  If you can suggest a better way I'd be
happy to follow it.

  As for implementation issues, I know it doesn't handle IRQ_DISABLED
at all.  I can add that easily enough.

> +++ kernel/arch/sh/kernel/irq.c Tue Oct 10 19:56:23 2000
> +int dispatch_virtual_irq(unsigned int irq, struct pt_regs * regs)
> 
> Please let me think for a while for this.

  Sure.

> +++ kernel/include/asm-sh/ide.h Fri Sep 29 06:10:52 2000
> +#define NO_ONBOARD_IDE 1
> 
> Umm... it looks not good.  Pending.

  I agree, ugly.  At the time it seemed necessary because the IDE
code was probing non-existant ISA ports looking for the standard
ide0 and ide1 ports, and this was oopsing.  I've since fixed the
ISA port mapping code so that probing non-existant ports is
harmless;  perhaps NO_ONBOARD_IDE code can go away now.  All it
does now is prevent wastage of those ports in systems which
support IDE on PCMCIA but have no native IDE of their own (a
system like this is sitting on my desk now).

> 
> +++ kernel/include/asm-sh/pgtable.h     Sun Oct  1 12:01:12 2000
> 
> Currently I don't understand this.  Please give me more time.

  I should have provided more comments ;-)  Basically, I grab 4 of
the unused (either by hw or sw) bits of the PTE and use them to
store bits which get shifted into the TC and SA bits of the PTEA
by the TLB miss handler.  The rest is just fancy footwork to
provide an interface for encoding those bits.


Greg.
-- 
These are my opinions not PPIs.