ISA DMA

ISA DMA

[Pierre Ossman]

I'm trying to figure out how to do ISA DMA transfers. I can't figure out 
  how to satisfy all the requirements the ISA DMA controller sets. I've 
set a DMA mask of 0x00ffffff but mappings end up above the 16MB limit 
nonetheless. And I have no idea how to keep transfers within the same 
64k boundary.

I've been trying to figure out how other drivers do it but I can't see 
what I'm missing. And the documentation doesn't cover ISA DMA.

I'm basically stuck now so I'd appreciate any tips you can give me.

Rgds
Pierre Ossman

Re: ISA DMA

[Alan Cox]

On Mer, 2004-09-08 at 22:28, Pierre Ossman wrote:
> I'm trying to figure out how to do ISA DMA transfers. I can't figure out 
>   how to satisfy all the requirements the ISA DMA controller sets. I've 
> set a DMA mask of 0x00ffffff but mappings end up above the 16MB limit 
> nonetheless. And I have no idea how to keep transfers within the same 
> 64k boundary.

drivers/net/lance.c is a pretty good worked example of driving the old
ISA bus grunge. 

Re: ISA DMA

[Pierre Ossman]

Alan Cox wrote:

>drivers/net/lance.c is a pretty good worked example of driving the old
>ISA bus grunge. 
>
>  
>
Thanks. That helped solve some of the problems at least. The transfers 
still only work some of the time. Everything seems ok but only junk in 
the buffers. Not sure which part is causing it so it'll require some 
more research.

Rgds
Pierre

Re: ISA DMA

[Tommy Reynolds]

[-- Attachment #1: Type: text/plain, Size: 447 bytes --]

Uttered Pierre Ossman <drzeus-list@drzeus.cx>, spake thus:

> I've been trying to figure out how other drivers do it but I can't see 
> what I'm missing. And the documentation doesn't cover ISA DMA.

Allocate your DMA buffer area by OR'ing in the "GFP_DMA" flag.  This
keeps your DMA buffer below the magical 16MB limit.  Also be sure to
use "virt_to_bus()" to convert the kernel buffer address into one
that you can give to the DMA engine.

HTH.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: ISA DMA

[Michael Sokolov]

Adrian Cox <adrian@humboldt.co.uk> wrote:

> Via686a. As I understand it, 32-bit DMA uses the DMA_HI_PAGE_x registers
> from 0x487 to 0x48a, and the Via manual doesn't claim to have any
> registers in that range.

It is also my understanding that VT82C686/A/B offers 24-bit DMA addressing
only.

On the Adirondack I plan to use the floppy disk controller (the only ISA DMA
user) in non-DMA mode only. I hope Linux can do that. The memory map we've
adopted does not have memory at PCI address 0 at all, so 24-bit DMA means no
DMA at all rather than bottom 16 MB only. It is really too late to change the
memory map now, it would be so much pain that I would much rather settle for
using non-DMA mode only for floppies.

> The only 32-bit DMA southbridge I've
> encountered is the Winbond chip on the Sandpoint, and that chip doesn't
> have UDMA66, USB, and audio.

Acer Labs 1543 has it according to its manual (see the latter for where the
registers are, I don't remember).

We have that bridge on the K2. Ironically, there we don't have a floppy disk
port, so we have good ISA DMA on the board where we don't need it but don't
have it on a board where we do...

MS

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

ISA DMA

[Adrian Cox]

[-- Attachment #1: Type: text/plain, Size: 452 bytes --]

The current 2.4-devel tree never assigns any memory to the ISA DMA zone,
irrespective of the setting of ISA_DMA_THRESHOLD. As I'm currently
trying to get a floppy working on a platform with 24-bit ISA DMA in its
Southbridge, this looks like a problem.

How do people feel about the attached patch? This should only change
behavior for platforms which set ISA_DMA_THRESHOLD to a value other than
0 or (-1)L.


--
Adrian Cox   http://www.humboldt.co.uk/

[-- Attachment #2: isadma.patch --]
[-- Type: text/plain, Size: 819 bytes --]

===== arch/ppc/mm/init.c 1.41 vs edited =====
--- 1.41/arch/ppc/mm/init.c	Fri Nov 30 15:07:14 2001
+++ edited/arch/ppc/mm/init.c	Wed Dec 19 10:15:44 2001
@@ -440,11 +440,18 @@
 	kmap_prot = PAGE_KERNEL;
 #endif /* CONFIG_HIGHMEM */

-	/*
-	 * All pages are DMA-able so we put them all in the DMA zone.
-	 */
-	zones_size[ZONE_DMA] = total_lowmem >> PAGE_SHIFT;
-	for (i = 1; i < MAX_NR_ZONES; i++)
+    	if (ISA_DMA_THRESHOLD && ISA_DMA_THRESHOLD < total_lowmem - 1)
+	{
+		zones_size[ZONE_DMA] = (ISA_DMA_THRESHOLD + 1) >> PAGE_SHIFT;
+		zones_size[ZONE_NORMAL] =
+			(total_lowmem - ISA_DMA_THRESHOLD - 1) >> PAGE_SHIFT;
+	}
+	else
+	{
+	    	zones_size[ZONE_DMA] = total_lowmem >> PAGE_SHIFT;
+	    	zones_size[ZONE_NORMAL] = 0;
+	}
+	for (i = 2; i < MAX_NR_ZONES; i++)
 		zones_size[i] = 0;

 #ifdef CONFIG_HIGHMEM

Re: ISA DMA

[David Müller (ELSOFT AG)]

Hi

Seems like you are facing the same problem as i do. But IIRC i think you
have to adjust the code in include/asm/dma.h (and possibly other places)
as well (at least in 2.4.14, not sure about the later kernels).

BTW:

What south bridge do you use? Some of the newer ones offer 32bit
addressing capability for DMA.

Adrian Cox wrote:

> The current 2.4-devel tree never assigns any memory to the ISA DMA zone,
> irrespective of the setting of ISA_DMA_THRESHOLD. As I'm currently
> trying to get a floppy working on a platform with 24-bit ISA DMA in its
> Southbridge, this looks like a problem.
>
> How do people feel about the attached patch? This should only change
> behavior for platforms which set ISA_DMA_THRESHOLD to a value other than
> 0 or (-1)L.
>


[stuff deleted]


Dave


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ISA DMA

[Adrian Cox]

David Müller (ELSOFT AG) wrote:


> Seems like you are facing the same problem as i do. But IIRC i think you
> have to adjust the code in include/asm/dma.h (and possibly other places)
> as well (at least in 2.4.14, not sure about the later kernels).


The main problem I can see is that the code always writes to the
DMA_HI_PAGE_x registers, irrespective of whether they actually exist.
This should be harmless, as nothing else is mapped to those locations.


> What south bridge do you use? Some of the newer ones offer 32bit
> addressing capability for DMA.


Via686a. As I understand it, 32-bit DMA uses the DMA_HI_PAGE_x registers
from 0x487 to 0x48a, and the Via manual doesn't claim to have any
registers in that range. The only 32-bit DMA southbridge I've
encountered is the Winbond chip on the Sandpoint, and that chip doesn't
have UDMA66, USB, and audio.


--
Adrian Cox   http://www.humboldt.co.uk/


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: ISA DMA

[Geert Uytterhoeven]

On Thu, 20 Dec 2001, Adrian Cox wrote:
> David Müller (ELSOFT AG) wrote:
> > What south bridge do you use? Some of the newer ones offer 32bit
> > addressing capability for DMA.
>
> Via686a. As I understand it, 32-bit DMA uses the DMA_HI_PAGE_x registers
> from 0x487 to 0x48a, and the Via manual doesn't claim to have any
> registers in that range. The only 32-bit DMA southbridge I've
> encountered is the Winbond chip on the Sandpoint, and that chip doesn't
> have UDMA66, USB, and audio.

And the Winbond South Bridge on the LongTrail (most probably the same chip).

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/