vmlinuz-2.6.29-1-11May on IIfx

vmlinuz-2.6.29-1-11May on IIfx

[Riccardo]

Hi,

since I was so happy of the reults on the Q950, I tried my luck with 
the current kernel on the IIfx .

The kernel completes boot. The keyboard works; I'm able to login.
But any activity on the ethernet (remlote login or just a Ping) gives 
me:

trigger_send() called  with the transmitter busy

and after a bit


eth0: mismatched read page pointers 45 vs  c.

booting into 2.6.26 I see the driver used is the mac8390. From dmesg of 
2.6.26

[42949395.010000] mac8390.c: v0.4 2001-05-15 David Huggins-Daines 
<dhd@debian.org> and others
[42949395.030000] eth0: Memory length resource for slot A not found, 
probing
[42949395.040000] eth0: EtherNet card in slot A (type apple)
[42949395.050000] MAC 00:80:ad:13:7c:d0 IRQ 57, 32 KB shared memory at 
0xfaad0000,  32-bit access.


I hope that helps a bit,

Riccardo

Re: vmlinuz-2.6.29-1-11May on IIfx

[Brad Boyer]

On Thu, May 14, 2009 at 07:35:03PM +0200, Riccardo wrote:
> But any activity on the ethernet (remlote login or just a Ping) gives 
> me:
> 
> trigger_send() called  with the transmitter busy
> 
> and after a bit
> 
> eth0: mismatched read page pointers 45 vs  c.
> 
> booting into 2.6.26 I see the driver used is the mac8390. From dmesg of 
> 2.6.26
> 
> [42949395.010000] mac8390.c: v0.4 2001-05-15 David Huggins-Daines 
> <dhd@debian.org> and others
> [42949395.030000] eth0: Memory length resource for slot A not found, 
> probing
> [42949395.040000] eth0: EtherNet card in slot A (type apple)
> [42949395.050000] MAC 00:80:ad:13:7c:d0 IRQ 57, 32 KB shared memory at 
> 0xfaad0000,  32-bit access.

Did this ethernet card work in an older kernel? The card type detection
in mac8390.c is partially based on guesses and just trying different
cards, as I remember. It's possible that it's detecting the wrong type,
particularly since it isn't finding all the NuBus ROM resources it is
expecting to be present on this card type. I suspect that it's not
accessing the card correctly. Both the error messages are that the
card isn't in a sane state based on what the driver tried to do.

Incidentally, "apple" is the default type for any card the driver
thinks it can use but can't identify with 100% certainty.

	Brad Boyer
	flar@allandria.com

Re: vmlinuz-2.6.29-1-11May on IIfx

[Michael Schmitz]

Hi,

> > But any activity on the ethernet (remlote login or just a Ping) gives 
> > me:
> > 
> > trigger_send() called  with the transmitter busy
> > 
> > and after a bit
> > 
> > eth0: mismatched read page pointers 45 vs  c.
> > 
> > booting into 2.6.26 I see the driver used is the mac8390. From dmesg of 
> > 2.6.26
> > 
> > [42949395.010000] mac8390.c: v0.4 2001-05-15 David Huggins-Daines 
> > <dhd@debian.org> and others
> > [42949395.030000] eth0: Memory length resource for slot A not found, 
> > probing
> > [42949395.040000] eth0: EtherNet card in slot A (type apple)
> > [42949395.050000] MAC 00:80:ad:13:7c:d0 IRQ 57, 32 KB shared memory at 
> > 0xfaad0000,  32-bit access.
> 
> Did this ethernet card work in an older kernel? The card type detection
> in mac8390.c is partially based on guesses and just trying different
> cards, as I remember. It's possible that it's detecting the wrong type,
> particularly since it isn't finding all the NuBus ROM resources it is
> expecting to be present on this card type. I suspect that it's not
> accessing the card correctly. Both the error messages are that the
> card isn't in a sane state based on what the driver tried to do.

The 'mismatched read page  pointers' I've seen on the EtherNEC 8390 based card 
as well, it seems to indicate that the receive buffer overruns. I'm using a 
timer interrupt there, so that is expected to happen. 

Might 'trigger send called with tx busy' also indicate a failure to receive the 
txdone interrupt? Maybe you're not getting any interrupts from the card? What 
does /proc/interrupts register regarding the 8390 driver?

	Michael

Re: vmlinuz-2.6.29-1-11May on IIfx

[Riccardo]

Hi,

>> Did this ethernet card work in an older kernel? The card type 
>> detection
>> in mac8390.c is partially based on guesses and just trying different
>> cards, as I remember. It's possible that it's detecting the wrong 
>> type,
>> particularly since it isn't finding all the NuBus ROM resources it is
>> expecting to be present on this card type. I suspect that it's not
>> accessing the card correctly. Both the error messages are that the
>> card isn't in a sane state based on what the driver tried to do.
>
> The 'mismatched read page  pointers' I've seen on the EtherNEC 8390 
> based card
> as well, it seems to indicate that the receive buffer overruns. I'm 
> using a
> timer interrupt there, so that is expected to happen.
>
> Might 'trigger send called with tx busy' also indicate a failure to 
> receive the
> txdone interrupt? Maybe you're not getting any interrupts from the 
> card? What
> does /proc/interrupts register regarding the 8390 driver?
>

under linux, 2.6.29 I get:

mac 57: 74 8390 Ethernet

I attempt to ping a host, which fails, then it says:

mac 57: 86 8390 Ethernet


Riccardo

Re: vmlinuz-2.6.29-1-11May on IIfx

[Riccardo]

Hi Brad,

On Thursday, May 14, 2009, at 11:09  PM, Brad Boyer wrote:

>> booting into 2.6.26 I see the driver used is the mac8390. From dmesg 
>> of
>> 2.6.26
>>
>> [42949395.010000] mac8390.c: v0.4 2001-05-15 David Huggins-Daines
>> <dhd@debian.org> and others
>> [42949395.030000] eth0: Memory length resource for slot A not found,
>> probing
>> [42949395.040000] eth0: EtherNet card in slot A (type apple)
>> [42949395.050000] MAC 00:80:ad:13:7c:d0 IRQ 57, 32 KB shared memory at
>> 0xfaad0000,  32-bit access.
>
> Did this ethernet card work in an older kernel? The card type detection
> in mac8390.c is partially based on guesses and just trying different
> cards, as I remember. It's possible that it's detecting the wrong type,
> particularly since it isn't finding all the NuBus ROM resources it is
> expecting to be present on this card type. I suspect that it's not
> accessing the card correctly. Both the error messages are that the
> card isn't in a sane state based on what the driver tried to do.

maybe I did not express myself clearly: in 2.6.29 the card doesn't work 
and I typed the error message manually,

I then rebooted into 2.6.26, where the card still worked, and pasted 
its dmesg.

Riccardo

Re: vmlinuz-2.6.29-1-11May on IIfx

[Finn Thain]

On Fri, 15 May 2009, Riccardo wrote:
> 
> ...in 2.6.29 the card doesn't work and I typed the error message 
> manually,
> 
> I then rebooted into 2.6.26, where the card still worked, and pasted its 
> dmesg.

This commit looks a bit suspect -- it was not CC'd to this list -- in 
particular the Makefile change:

http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=ca17584bf2ad1b1e37a5c0e4386728cc5fc9dabc

So now mac8390.c and 8390.c both #include "lib8390.c"... might this 
perhaps lead to the wrong ei_inb/outb etc in the final link?

Finn

Re: vmlinuz-2.6.29-1-11May on IIfx

[Brad Boyer]

On Fri, May 15, 2009 at 08:28:51PM +1000, Finn Thain wrote:
> This commit looks a bit suspect -- it was not CC'd to this list -- in 
> particular the Makefile change:
> 
> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=ca17584bf2ad1b1e37a5c0e4386728cc5fc9dabc
> 
> So now mac8390.c and 8390.c both #include "lib8390.c"... might this 
> perhaps lead to the wrong ei_inb/outb etc in the final link?

It does seem like an odd change. It would be nice if this list got
copied on changes to any nubus drivers, but I suppose it's not anywhere
on the maintainers list. It might be something to bring up with David
for any future changes to nubus network cards.

I'll try to find some time to build a kernel and try it out. I have
several cards that use this driver.

	Brad Boyer
	flar@allandria.com

Re: vmlinuz-2.6.29-1-11May on IIfx

[Geert Uytterhoeven]

On Sun, May 17, 2009 at 08:10, Brad Boyer <flar@allandria.com> wrote:
> On Fri, May 15, 2009 at 08:28:51PM +1000, Finn Thain wrote:
>> This commit looks a bit suspect -- it was not CC'd to this list -- in
>> particular the Makefile change:
>>
>> http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=ca17584bf2ad1b1e37a5c0e4386728cc5fc9dabc
>>
>> So now mac8390.c and 8390.c both #include "lib8390.c"... might this
>> perhaps lead to the wrong ei_inb/outb etc in the final link?
>
> It does seem like an odd change. It would be nice if this list got
> copied on changes to any nubus drivers, but I suppose it's not anywhere
> on the maintainers list. It might be something to bring up with David
> for any future changes to nubus network cards.

Probably the change looked sufficient trivial/innocent. Trivial updates for core
changes are usually posted on the subsystem mailing list only.

> I'll try to find some time to build a kernel and try it out. I have
> several cards that use this driver.

Good, thanks!

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

Re: vmlinuz-2.6.29-1-11May on IIfx

[Brad Boyer]

On Fri, May 15, 2009 at 10:32:46AM +0200, Riccardo wrote:
> maybe I did not express myself clearly: in 2.6.29 the card doesn't work 
> and I typed the error message manually,
> 
> I then rebooted into 2.6.26, where the card still worked, and pasted 
> its dmesg.

My apologies. I did misunderstand your original message.

	Brad Boyer
	flar@allandria.com