@AMiGR

Then this is the only thing that people care about? i.e. that cache coherency is not done by the hardware but done by software (in the driver)? that's all?
Damn, I never understood that. The most important is the result, not how it is done.
This is what the end user (like me) wants to know and the result is that OS4 does full DMA.

Last edited by elwood on 03-Jun-2005 at 10:41 AM.
Last edited by elwood on 03-Jun-2005 at 10:40 AM.

_________________
Philippe 'Elwood' Ferrucci
Sam460 1.10 Ghz
AmigaOS 4 betatester
Amiga Translator Organisation

Oops, double post.

Last edited by elwood on 03-Jun-2005 at 10:40 AM.

_________________
Philippe 'Elwood' Ferrucci
Sam460 1.10 Ghz
AmigaOS 4 betatester
Amiga Translator Organisation

@AMiGR

Quote:

AMiGR wrote: @Varthall The cache coherency can be "fixed" with an external hardware fix, that's how it's done on the Peg1 and ?A1.

You're right, I haven't thought about it.

Varthall

_________________
AmigaOne XE - AmigaOS 4.1 - Freescale 7457 1GHz - 1GB ram

Is that how its done on the micro-A1? Can someone point me to it on the schematics. Ty.

@elwood

Quote:

elwood wrote: @AMiGR Then this is the only thing that people care about? i.e. that cache coherency is not done by the hardware but done by software (in the driver)? that's all? Damn, I never understood that. The most important is the result, not how it is done. This is what the end user (like me) whants to know.

It's not about this. I don't know if the software patches added to the Linux kernel actually do cache coherency, or if they are just workarounds to this problem, the fact is that even with these patches DMA isn't reliable on unfixed XE boards.

Varthall

_________________
AmigaOne XE - AmigaOS 4.1 - Freescale 7457 1GHz - 1GB ram

@Varthall

On unfixed? Yeah sure.
If you meant on "fixed". This is only valid in Debian, not for OS4.

Last edited by elwood on 03-Jun-2005 at 11:12 AM.

_________________
Philippe 'Elwood' Ferrucci
Sam460 1.10 Ghz
AmigaOS 4 betatester
Amiga Translator Organisation

@Helgis

Quote:

I could assign you as my personal secretary to correct my "bugs"

You could check your facts, before posting them.

_________________

@Varthall

Quote:

I don't know if the software patches added to the Linux kernel actually do cache coherency, or if they are just workarounds to this problem, the fact is that even with these patches DMA isn't reliable on unfixed XE boards.

I use the Sil-thingy with my harddrives (dma=on).
3 atapi connecting to the unfixed VIA-thingy. (dma=off).
Using kernel 2.4.26 that Ross has made, I have not noticed any problems with file-corruption this far.
just my 2 cents...

_________________
Site admins are people too..pooff!

@elwood


Quote:

Well in this case, the OS4 drivers does it so the result is that it works so it's ok for me. The Debian driver doesn't do it? Fine, but personally I don't care.

Please do stop trolling against Linux.
Like it or not, the A1 needs working Linux too.

_________________
Site admins are people too..pooff!

@Wiffy
The uA1-C’s are still broken for under Linux with DMA switched on.

As I understand it this is an Artica-S problem and is the cache problem people have been speaking about. I do not personally understand why this means that DMA work with an external card but not the onboard IDE with the micros but I am sure one of the Linux guys can explain it.

_________________
Sam440ep 667mhz, 512MB, 120GB 2.5" HD, OS4.1FE
WinUae 3.0.0, OS 3.9, BB3, Catweasel MkIV
Amiga 1200, Blizzard 040/40 (BlizzardPPC 060/200 with SCSI removed at present), mediatorSX pci, Voodoo3, PCI network card os 3.9BB2
4MB Minimig with ARM addon boar

@AMiGR

lol!!

@elwood

I meant on the "not fixed" boards. Try the md5 check on the copied files, if the DMA works now I would make me very happy, and it would be a reason more to do the fix

Varthall

_________________
AmigaOne XE - AmigaOS 4.1 - Freescale 7457 1GHz - 1GB ram

@tomazkid

Quote:

I use the Sil-thingy with my harddrives (dma=on). 3 atapi connecting to the unfixed VIA-thingy. (dma=off). Using kernel 2.4.26 that Ross has made, I have not noticed any problems with file-corruption this far. just my 2 cents...

Yes, I've heard of DMA working on a SIL controller... but I'm one of those who prefer to use a PCI slot for other cards

Varthall

_________________
AmigaOne XE - AmigaOS 4.1 - Freescale 7457 1GHz - 1GB ram

@Dwyloc
Quote:

The uA1-C’s are still broken for under Linux with DMA switched on.

Last time I checked, Linux cannot run correctly on the MicroA1-C due to the GX CPU not being supported and different VIA settings required (sorry, I don't remember the precise details). So I'm curious, how exactly did you run your test?

_________________
ExecSG Team Lead

@tomazkid

you use your HD at the same time as ethernet and got no corruption, is that right ?
Lio

@AMiGR

Quote:

AMiGR wrote: @Helgis You should SERIOUSLY become a marketing person. Lot's of excitement and big and nice words about stuff you know nothing about, you qualify 100%!

_________________
Are we not done with the same silly arguments and flames yet??!

@lionstorm

Quote:

you use your HD at the same time as ethernet and got no corruption, is that right ?

Yes. I've burnt dvd's while surfing the net.
Dvd's pass the verification-check OK.

Last edited by tomazkid on 03-Jun-2005 at 06:31 PM.

_________________
Site admins are people too..pooff!

@Dwyloc

Quote:

As I understand it this is an Artica-S problem and is the cache problem people have been speaking about. I do not personally understand why this means that DMA work with an external card but not the onboard IDE with the micros but I am sure one of the Linux guys can explain it.

Simple: your understanding is incorrect. The onboard IDE on an AmigaOne board is done by the VIA686B southbridge, which is connected to the PCI bus (as if it were on a PCI card). In other words, you have one PCI card (the one with the sil controller) that lets you do full DMA, and another "card" (the 686B) that has broken DMA. Simple logic should be enough to prove there's something wrong with either the way the 686 is "plugged in" (ie wired on the board) or the drivers for it. The Articia is just not part of the equation; if it were, we'd be seeing problems with both IDE controllers.

From the description of the IDE DMA fix that's been floating around (cut some traces, bridge some lines, replace a few capacitors), I'd say that the problem is that someone at MAI simply screwed up the "articia to 686" part of the PCI bus on the board. If it were more complex than that, the fix would probably involve custom logic. Since the Pegasos I needed a similar fix (someone who knows about chip design told me that he recognized the kind of chip the April boards used, and they have a speed grade too low to do anything complicated on a 33MHz bus), I suspect this is yet another case of VIA not documenting things properly.

@all

For Linux DMA to work 100% with the A1 you must use a kernel with the Coherency
patch (as well as the hardware fix,or Sil card) in it such as Ross`s latest
kernel.

The original 2.4.19 will still cause problems.

OS4 works differently so only the hardware fix or Sil card are required.


Just

EDIT: Ooops. This wanted to write this in private.

FYI: There are no April like hardware 'patches' on uA1. April chip is needed for OLD Articia chips (what can be found on A1-SE and Peg1 boards). The XE and the uA1 has got newer Articia. I thought it was already clear for everybody...

Last edited by Chip on 03-Jun-2005 at 11:00 PM.