Re: [PATCH] advansys: ISA support depends on ISA_DMA_API

[Ben Hutchings <ben@decadent.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 1457 bytes --]

On Fri, 2010-02-19 at 01:38 +0000, Ben Hutchings wrote:
> The advansys driver supports ISA cards using DMA but not PIO-only.
> Fix the dependency and preprocessor conditions.

If I recall correctly, we subsequently discussed this on IRC and you
said this issue had been raised before and that ISA should imply
ISA_DMA_API.

So, where is it documented that ISA implies ISA_DMA_API?  There are many
dependencies in drivers/scsi/Kconfig on ISA && ISA_DMA_API which are be
redundant if this is true.

You previously wrote (archived as
<http://thread.gmane.org/gmane.linux.scsi/44923>):
> On Thu, Oct 02, 2008 at 09:49:15AM +0200, Martin Michlmayr wrote:
> > * Matthew Wilcox <matthew@wil.cx> [2008-09-30 10:13]:
> > > Huh?  I've looked into it.  ARM and MIPS provide half of an API but not
> > > the other half.  These architectures are broken.
> > 
> > Maybe, but the point is that advansys was marked as BROKEN on !X86_32
> > before and now it's not, so the ARM/MIPS compilation errors show up
> > and that's a regression.  What you say is a good argument for my patch
> > to disable advansys for ARM/MIPS in Kconfig.
> 
> No, ARM and MIPS need to just add stub functions that do nothing.

You mean that request_dma() should always fail?  Won't that just allow
us to build a bunch of drivers for ARM and MIPS that will never work?

Ben.

-- 
Ben Hutchings
Once a job is fouled up, anything done to improve it makes it worse.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 828 bytes --]