sun4c driver/dma problems

sun4c driver/dma problems

[Uli Luckas]

Hi, 
there seem to be all sorts of driver problems with the 2.6 Kernels and my 
Sparc Clone "GoldStar SPARC Workstation GWS-20".
The box is a sun4c architecture with integrated SunLANCE networking (external 
tranciver), cg3 framebuffer and 1 ESP SCSI controller.
It is operated as a diskless workstation and works with 2.2 series kernels 
(2.2.26).

As the idprom's battery is empty I boot it with the following commands:

n			 	# new command mode
f idprom@ 1 xor f mkp		# reset idprom
8 0 20 76 4c 78 c0ffee mkpl	# set mac addr/host id
boot net - ip=dhcp		# boot linux

Framebuffer
=====While 2.2 works nicely with the framebuffer console, the first twist was a 
lockup at boot time, worked around by using the prom console.

Sun LANCE
====Now I am stuck at IP auto configuration with "Sending DHCP requests ...... 
timed out!"
While DHCP requests are being sent, neither does my switch show any activity, 
nor does tcpdump on the server report any packets being received.
The card is detected properly though "sunlance.c:v2.02.... eth0: LANCE 
08:00:20:76:4c.78"

When I instead configure a fixed IP:
boot net - ip\x10.0.0.5:10.0.0.3:10.0.0.3:255.255.255.0::: 
nfsroot\x10.0.0.3:/data/export/euler/root
The machine get's stuck mounting the nfs-root:
Looking up port or RPC 100003/2 on 10.0.0.3
portmap: server 10.0.0.3 not responding, timed out
Root-NFS: Unable to get nfsd port number from server, using default
Looking up port or RPC 100005/1 on 10.0.0.3
NETDEV WATCHDOG: eth0: transmit timed out
eth0 transmit timed out, status ff3c, reset

Again, there is no activity, what so ever, on the wire (switch/tcpdump)

DMA
=All problems (see below) could of course be linked to this boot message?
dma0 unknown dma version 70000000

ESP SCSI
====
Cannot find dvma for ESP0's SCSI
ESP: Total of 1 ESP hosts found, 0 actually in use.

This all happens either with a native, 2.95 compiled kernel, as with a 3.3.2 
cross x86 compiled kernel.

Any help would be appreciated.

Thanks
Uli

Re: sun4c driver/dma problems

[Uli Luckas]

> DMA
> => All problems (see below) could of course be linked to this boot message?
> dma0 unknown dma version 70000000
By the way, 2.2.26 reports:
dma0: Revision 1

Uli

Re: sun4c driver/dma problems

[Uli Luckas]

Am Mittwoch, 2. Juni 2004 08:18 schrieb Uli Luckas:
> Hi,
> there seem to be all sorts of driver problems with the 2.6 Kernels and my
> Sparc Clone "GoldStar SPARC Workstation GWS-20".
> The box is a sun4c architecture with integrated SunLANCE networking
> (external tranciver), cg3 framebuffer and 1 ESP SCSI controller.
> It is operated as a diskless workstation and works with 2.2 series kernels
> (2.2.26).

> DMA
> => All problems (see below) could of course be linked to this boot message?
> dma0 unknown dma version 70000000
>
Hi,
the above cited DMA message seems to be the root of the driver problems. From 
the 2.2.26 kernel I know it should read:
dma0: Revision 1

So what can I do to help, getting this fixed? Who works on this code?
I already know, it's not only the version detection. Defining a new version 
70000000 "Revision 1 broken" and initialising it just as a "Revision 1" works 
around the above message, but all the device drivers report all kind of DMA 
related problems later on and of course don't work.

Please, if someone would look into this, I could arange for access to this 
machine, running 2.2.26. And would be more than pleased to assist in 
isolating the problem.

Regards
Uli

Re: sun4c driver/dma problems

[Keith M Wesolowski]

On Tue, Jun 08, 2004 at 09:22:44AM +0200, Uli Luckas wrote:

> the above cited DMA message seems to be the root of the driver problems. From 
> the 2.2.26 kernel I know it should read:
> dma0: Revision 1
> 
> So what can I do to help, getting this fixed? Who works on this code?
> I already know, it's not only the version detection. Defining a new version 
> 70000000 "Revision 1 broken" and initialising it just as a "Revision 1" works 
> around the above message, but all the device drivers report all kind of DMA 
> related problems later on and of course don't work.

Reading your messages, I'm uncertain exactly how well 2.2.26 works.
Can you please clarify?  If it's perfect, then I may invest a few
hours trying to fix this for you in 2.6.  If there's no known-working
kernel, it's unlikely that it ever will be fixed.

-- 
Keith M Wesolowski

Re: sun4c driver/dma problems

[Uli Luckas]

Am Dienstag, 8. Juni 2004 17:53 schrieb Keith M Wesolowski:
> On Tue, Jun 08, 2004 at 09:22:44AM +0200, Uli Luckas wrote:
> > the above cited DMA message seems to be the root of the driver problems.
> > From the 2.2.26 kernel I know it should read:
> > dma0: Revision 1
> >
> > So what can I do to help, getting this fixed? Who works on this code?
> > I already know, it's not only the version detection. Defining a new
> > version 70000000 "Revision 1 broken" and initialising it just as a
> > "Revision 1" works around the above message, but all the device drivers
> > report all kind of DMA related problems later on and of course don't
> > work.
>
> Reading your messages, I'm uncertain exactly how well 2.2.26 works.
> Can you please clarify?  If it's perfect, then I may invest a few
> hours trying to fix this for you in 2.6.  If there's no known-working
> kernel, it's unlikely that it ever will be fixed.
Hi Keith,
thanks for your reply. 
The machine works nicely and stable with 2.2 kernels, nfs root and nbd swap.
With 2.2 kernels I can also access the local SCSI cdrom and hard drives which 
host an old SunOS 4.1.3 installation.

Please let me know, if you need more information.

Uli

Re: sun4c driver/dma problems

[Uli Luckas]

Am Dienstag, 8. Juni 2004 18:12 schrieb Uli Luckas:
> Am Dienstag, 8. Juni 2004 17:53 schrieb Keith M Wesolowski:
> > On Tue, Jun 08, 2004 at 09:22:44AM +0200, Uli Luckas wrote:
> > > the above cited DMA message seems to be the root of the driver
> > > problems. From the 2.2.26 kernel I know it should read:
> > > dma0: Revision 1
> > >
> > > So what can I do to help, getting this fixed? Who works on this code?
> > > I already know, it's not only the version detection. Defining a new
> > > version 70000000 "Revision 1 broken" and initialising it just as a
> > > "Revision 1" works around the above message, but all the device drivers
> > > report all kind of DMA related problems later on and of course don't
> > > work.
> >
> > Reading your messages, I'm uncertain exactly how well 2.2.26 works.
> > Can you please clarify?  If it's perfect, then I may invest a few
> > hours trying to fix this for you in 2.6.  If there's no known-working
> > kernel, it's unlikely that it ever will be fixed.
>
> Hi Keith,
> thanks for your reply.
> The machine works nicely and stable with 2.2 kernels, nfs root and nbd
> swap. With 2.2 kernels I can also access the local SCSI cdrom and hard
> drives which host an old SunOS 4.1.3 installation.
>
> Please let me know, if you need more information.

Hi Keith,
did you get any chance to look into the problem?

Uli

-- 
"[...]Yes, Mr. Gates, recently you have helped open source succeed -- in
much the same way Osama bin Laden has helped beef up airport security 
lately."
                                -- Eric S. Raymond, 9.11.2001

Re: sun4c driver/dma problems

[Uli Luckas]

Am Dienstag, 8. Juni 2004 17:53 schrieb Keith M Wesolowski:
> On Tue, Jun 08, 2004 at 09:22:44AM +0200, Uli Luckas wrote:
> > the above cited DMA message seems to be the root of the driver problems.
> > From the 2.2.26 kernel I know it should read:
> > dma0: Revision 1
> >
> > So what can I do to help, getting this fixed? Who works on this code?
> > I already know, it's not only the version detection. Defining a new
> > version 70000000 "Revision 1 broken" and initialising it just as a
> > "Revision 1" works around the above message, but all the device drivers
> > report all kind of DMA related problems later on and of course don't
> > work.
>
> Reading your messages, I'm uncertain exactly how well 2.2.26 works.
> Can you please clarify?  If it's perfect, then I may invest a few
> hours trying to fix this for you in 2.6.  If there's no known-working
> kernel, it's unlikely that it ever will be fixed.

Dear Keith,
thanks for your work. 
As I have read, you are retiering from sparc32 kernel development. Is there 
any hint that could get me going on fixing the above problems myself? Do you 
have asumptions on where the problem can be found?

Regards
Uli

Re: sun4c driver/dma problems

[Keith M Wesolowski]

On Tue, Jul 13, 2004 at 11:39:00PM +0200, Uli Luckas wrote:

> > > So what can I do to help, getting this fixed? Who works on this code?
> > > I already know, it's not only the version detection. Defining a new
> > > version 70000000 "Revision 1 broken" and initialising it just as a
> > > "Revision 1" works around the above message, but all the device drivers
> > > report all kind of DMA related problems later on and of course don't
> > > work.

> As I have read, you are retiering from sparc32 kernel development. Is there 
> any hint that could get me going on fixing the above problems myself? Do you 
> have asumptions on where the problem can be found?

I do not know the cause of this problem.  However, it's entirely
possible it's just a detection issue.  You might consider grepping for
these messages in the source, then diffing between a working and
non-working kernel to see what's changed.  If it's not simple
detection, it may be related to the hypersparc DMA issue I've noted,
which has not been diagnosed.

-- 
Keith M Wesolowski

Re: sun4c driver/dma problems

[Uli Luckas]

Am Donnerstag, 15. Juli 2004 00:23 schrieb Keith M Wesolowski:
> On Tue, Jul 13, 2004 at 11:39:00PM +0200, Uli Luckas wrote:
> > > > So what can I do to help, getting this fixed? Who works on this code?
> > > > I already know, it's not only the version detection. Defining a new
> > > > version 70000000 "Revision 1 broken" and initialising it just as a
> > > > "Revision 1" works around the above message, but all the device
> > > > drivers report all kind of DMA related problems later on and of
> > > > course don't work.
> >
> > As I have read, you are retiering from sparc32 kernel development. Is
> > there any hint that could get me going on fixing the above problems
> > myself? Do you have asumptions on where the problem can be found?
>
> I do not know the cause of this problem.  However, it's entirely
> possible it's just a detection issue.  You might consider grepping for
> these messages in the source, then diffing between a working and
> non-working kernel to see what's changed.  If it's not simple
> detection, it may be related to the hypersparc DMA issue I've noted,
> which has not been diagnosed.
Hi Keith,
thanks for your reply. I'll see what I can do. Even though I might not be 
clued in to sparc internals enough.

Regards
Uli

-- 
Im Kapitalismus beutet der Mensch den Menschen aus.
Im Sozialismus ist es genau umgekehrt.
- Ben Tucker