[9fans] pcmcia/cardbus rtl8139 support

[9fans] pcmcia/cardbus rtl8139 support

[Gabriel Diaz]

Hello

This card used to work, but i updated with a new plan9.iso and now
doesn't work.

I can see this message on boot:

rtl8139: port 0x400 in use

how can i see what is using that port or force the card to use
otherone (i tryed ether0=type=rtl839 port=0x400 to "reserve" the port
:-? but doesn't work)

thanks

gabi

Re: [9fans] pcmcia/cardbus rtl8139 support

[Steve Simon]

> otherone (i tryed ether0=type=rtl839 port=0x400 to "reserve" the port

is this a typo or cut & paste? surely you meant

	ether0=type=rtl8139 port=0x400 

Alternatively look in /dev/ioalloc and see what is using
0x400, and try forcing your ether card to use a different
address (I thiunk you can do this), eg.

	ether0=type=rtl8139 port=0x380

-Steve

Re: [9fans] pcmcia/cardbus rtl8139 support

[Gabriel Diaz]

hi

yes that was a typo.

ioalloc tells that the port 0x400 is used by cardbus
i have:

#Y0: Ricoh 476 PCI/Cardbus bridge, B0000000 intl 11

i tried with that plan9.ini but still doesn't work.

i recompiled with DEBUG 1 in devpccard and i got the following:

#Y0: Ricoh 476 PCI/Cardbus bridge, B0000000 intl 11
engine(0): SlotEmpty(CardDetected)
#Y0: card powered at 3 volt
engine(0): SlotFull(CardPowered)
configuring slot 0 (SlotPowered)
engine(0): SlotPowered(CardConfigured)
PCI.21.0.0 mem[1] 80000000 512
rtl8139: reset device failed
#U/usb0: uhci port 0x1820 irq 11
... three more usb ports...
1015 memory: 256M kernel data, 758M user, 1383 swap

then the computer is hanged, i cannot run  ^t^t+p or r, ctrl-alt-del
prints cpu0: exiting but does not reboot.

the kernel from the cd doesn't notice the card.

i added the print Richard Miller suggested and he is right, the reset
device command failed.

I suppose i messed up the kernel source or something, but i don't get
why that print is hanging the computer.

Also, note that the pci.21.0.0 is the network card on the cardbus.

windows tells the card io is on 0xff00-0xffff

thanks all

gabi

On 3/27/07, Steve Simon <steve@quintile.net> wrote:
> > otherone (i tryed ether0=type=rtl839 port=0x400 to "reserve" the port
>
> is this a typo or cut & paste? surely you meant
>
>         ether0=type=rtl8139 port=0x400
>
> Alternatively look in /dev/ioalloc and see what is using
> 0x400, and try forcing your ether card to use a different
> address (I thiunk you can do this), eg.
>
>         ether0=type=rtl8139 port=0x380
>
> -Steve
>

Re: [9fans] pcmcia/cardbus rtl8139 support

[Richard Miller]

> I can see this message on boot:
> 
> rtl8139: port 0x400 in use

Sometimes this message is misleading.  In rtl8139match, if the call to
rtl8139reset fails, the adapter isn't marked active but the port has
already been allocated.  That means rtl8139match can be called
again for the same adapter, and this time the ioalloc will fail.

It might be helpful to print an error message in rtl8139reset
so you can see that it failed.

-- Richard

Re: [9fans] pcmcia/cardbus rtl8139 support

[geoff]

I fixed the bug in rtl8139reset that wasn't freeing the port when
rtl8139reset failed.  I've just pushed the fixed driver to sources.

Re: [9fans] pcmcia/cardbus rtl8139 support

[Richard Miller]

> I fixed the bug in rtl8139reset that wasn't freeing the port when
> rtl8139reset failed.

ethervt6102.c needs a similar correction:

952a953
> 			iofree(port);
958a960
> 			iofree(port);

Re: [9fans] pcmcia/cardbus rtl8139 support

[Gabriel Diaz]

hello

i'm still trying to get this notebook work. (ibm z60t)

Any ideas why the 9pcf kernel provided with the install cdrom is able
to boot the system while a local compiled version (with the sources
from the cdrom and the configuration untouched) hangs after the memory
line is printed (1015M memory: 256 kernel data, 758 user, 1383 swap) ?

i added debug codes in sdata.c but i've to much output to see
anything, i saw the last command returns a timeout (i suppose was a
probe command and is not an error).

I added also the iofree(port) to the rtl, now the port is free (the
complain message doesn't appear), but the card is still not
recognized.

i added to pci.c the id 8086/2641 of the intel southbridge ich6 mobile
 (with the original kernels, the one that boots, and the one that
doesn't boot, both, prints the line about not finding southbridge).
And added too the sdata disk id to the dma switch as commented on a
recent thread.

The modified kernel also does not boot and hangs in the same point.

this notebook has no serial and no network (broadcom 57x, rtl8139
card), should i give up? (it was working(tm) with an old plan9
installation from oct 2006 or so, but i lost the cd somewhere and i
wanted to add venti. . .).

thanks

gabi

On 3/28/07, Richard Miller <9fans@hamnavoe.com> wrote:
> > I fixed the bug in rtl8139reset that wasn't freeing the port when
> > rtl8139reset failed.
>
> ethervt6102.c needs a similar correction:
>
> 952a953
> >                       iofree(port);
> 958a960
> >                       iofree(port);
>
>

Re: [9fans] pcmcia/cardbus rtl8139 support

[erik quanstrom]

On Wed Mar 28 04:54:47 EDT 2007, gabidiaz@gmail.com wrote:
> hello
> 
> i'm still trying to get this notebook work. (ibm z60t)
> 
> Any ideas why the 9pcf kernel provided with the install cdrom is able
> to boot the system while a local compiled version (with the sources
> from the cdrom and the configuration untouched) hangs after the memory
> line is printed (1015M memory: 256 kernel data, 758 user, 1383 swap) ?

/386/9pc* are not regenerated each time a change is made to
the kernel sources.  with the date of 9pc from the cd in hand
and sources, you should be able to work out what changed.

one possibility is that it's an interrupt coming when a device
is partially configured and the old kernel also had the bug,
but it was hidden.  sdata.c is somewhat notorious for this.

> i added to pci.c the id 8086/2641 of the intel southbridge ich6 mobile
>  (with the original kernels, the one that boots, and the one that
> doesn't boot, both, prints the line about not finding southbridge).
> And added too the sdata disk id to the dma switch as commented on a
> recent thread.
> 
> The modified kernel also does not boot and hangs in the same point.

i think you'll have better luck first isolating exactly the 
function/functions and line numbers where you're stuck.  are you
sure you're getting ata interrupts?

the standard kernel prints are a very inexact way of determining 
what has happeend and what hasn't.  sometimes the best
way is to do a binary search with print statements.
 
> i added debug codes in sdata.c but i've to much output to see
> anything, i saw the last command returns a timeout (i suppose was a
> probe command and is not an error).

unfortunately, i don't know of a usb console.

> this notebook has no serial and no network (broadcom 57x, rtl8139
> card), should i give up?

heck, no.  you're doing the right things and it seems like you are
making quite a bit of progress.  

since you can boot from the cd, you have a lot of options.  you
could try rebuilding from sources history on the day the kernel
on the cd was built and adding newer bits until it breaks.
(kernel work is so unglamorous.)  also, if you have a fileserver, 
you can try compiling a kernel without sd in the configuration. 

- erik

Re: [9fans] pcmcia/cardbus rtl8139 support

[Gabriel Diaz]

hello

thanks for the encouragement.

I compiled a new kernel without pccard support, and now it boots, even
the sdata wiht the new ids and the hd dma. I'm pretty sure i will get
used to the dma speed quickly :)


I added the DEBUG param in devpccard and got:
#Y0: Ricoh 476 PCI/Cardbus bridge, B0000000 intl 11
engine(0): SlotFull(CardPowered)
configuring slot 0 (SLotPowered)
engine(0): SloptPowered(CardConfigured)

that is with the card plugged and the kernel hanging, without the
card, the kernel has started to work (before wasn't, the *nopcirouting
seems to be unrelated as i tested with and without and get the same
results, i suppose i'm the guilty of that previous failure)

so i suppose the problem is between devpccard and rtl8139 driver.

that was the only pccard i have, i will look into rtl8139 to see if i
can get a conclusion.

thanks all,

gabi

btw: the card is "Conceptronic CSP100TCL 10/100Mbps LAN SnapPort Card
for notebooks"


On 3/28/07, erik quanstrom <quanstro@coraid.com> wrote:
> On Wed Mar 28 04:54:47 EDT 2007, gabidiaz@gmail.com wrote:
> > hello
> >
> > i'm still trying to get this notebook work. (ibm z60t)
> >
> > Any ideas why the 9pcf kernel provided with the install cdrom is able
> > to boot the system while a local compiled version (with the sources
> > from the cdrom and the configuration untouched) hangs after the memory
> > line is printed (1015M memory: 256 kernel data, 758 user, 1383 swap) ?
>
> /386/9pc* are not regenerated each time a change is made to
> the kernel sources.  with the date of 9pc from the cd in hand
> and sources, you should be able to work out what changed.
>
> one possibility is that it's an interrupt coming when a device
> is partially configured and the old kernel also had the bug,
> but it was hidden.  sdata.c is somewhat notorious for this.
>
> > i added to pci.c the id 8086/2641 of the intel southbridge ich6 mobile
> >  (with the original kernels, the one that boots, and the one that
> > doesn't boot, both, prints the line about not finding southbridge).
> > And added too the sdata disk id to the dma switch as commented on a
> > recent thread.
> >
> > The modified kernel also does not boot and hangs in the same point.
>
> i think you'll have better luck first isolating exactly the
> function/functions and line numbers where you're stuck.  are you
> sure you're getting ata interrupts?
>
> the standard kernel prints are a very inexact way of determining
> what has happeend and what hasn't.  sometimes the best
> way is to do a binary search with print statements.
>
> > i added debug codes in sdata.c but i've to much output to see
> > anything, i saw the last command returns a timeout (i suppose was a
> > probe command and is not an error).
>
> unfortunately, i don't know of a usb console.
>
> > this notebook has no serial and no network (broadcom 57x, rtl8139
> > card), should i give up?
>
> heck, no.  you're doing the right things and it seems like you are
> making quite a bit of progress.
>
> since you can boot from the cd, you have a lot of options.  you
> could try rebuilding from sources history on the day the kernel
> on the cd was built and adding newer bits until it breaks.
> (kernel work is so unglamorous.)  also, if you have a fileserver,
> you can try compiling a kernel without sd in the configuration.
>
> - erik
>

Re: [9fans] pcmcia/cardbus rtl8139 support

[erik quanstrom]

that's great progress for one day. 

*nopcirouting is not typically helpful and if your machine worked
before without pcirouting, it is unlikely that it would help now.

one silly quesiton.  is the card plugged into a working network?
is this different than before when it was working?

at this point it would make a lot of sense to try the older driver.
if the older driver works, there don't appear to be many changes
so if this is the problem, you can easily figure out which one
broke your 8139.

- erik

On Thu Mar 29 04:58:55 EDT 2007, gabidiaz@gmail.com wrote:
> hello
> 
> thanks for the encouragement.
> 
> I compiled a new kernel without pccard support, and now it boots, even
> the sdata wiht the new ids and the hd dma. I'm pretty sure i will get
> used to the dma speed quickly :)
> 
> 
> I added the DEBUG param in devpccard and got:
> #Y0: Ricoh 476 PCI/Cardbus bridge, B0000000 intl 11
> engine(0): SlotFull(CardPowered)
> configuring slot 0 (SLotPowered)
> engine(0): SloptPowered(CardConfigured)
> 
> that is with the card plugged and the kernel hanging, without the
> card, the kernel has started to work (before wasn't, the *nopcirouting
> seems to be unrelated as i tested with and without and get the same
> results, i suppose i'm the guilty of that previous failure)
> 
> so i suppose the problem is between devpccard and rtl8139 driver.
> 
> that was the only pccard i have, i will look into rtl8139 to see if i
> can get a conclusion.
> 
> thanks all,
> 
> gabi

Re: [9fans] pcmcia/cardbus rtl8139 support

[Gabriel Diaz]

hello

Seems that devpccard is the only one that hangs the computer.

If i plug a pcmcia card (network card or tv card those are the one i
have been able to get) the pc hangs, and more funny, sometimes, when i
poweroff-poweron the computer and it hanged due to pccard, the kernel
hangs without a card in the slop.

i mean, boot with a card -> kernel hangs -> poweroff -> poweron ->boot
without a card -> kernel hangs (crappy hardware :-? too much smart
bios with weird settings? )

Also if i insert a card on an already booted system, it get's hanged.

i will try to see any devpccard changes in the last months, it will be
a great success to recover plan9-bility of my work notebook (while the
broadcom drivers gets done the pccard is the only way to have
network).

thanks

gabi


On 3/29/07, erik quanstrom <quanstro@coraid.com> wrote:
> that's great progress for one day.
>
> *nopcirouting is not typically helpful and if your machine worked
> before without pcirouting, it is unlikely that it would help now.
>
> one silly quesiton.  is the card plugged into a working network?
> is this different than before when it was working?
>
> at this point it would make a lot of sense to try the older driver.
> if the older driver works, there don't appear to be many changes
> so if this is the problem, you can easily figure out which one
> broke your 8139.
>
> - erik
>
> On Thu Mar 29 04:58:55 EDT 2007, gabidiaz@gmail.com wrote:
> > hello
> >
> > thanks for the encouragement.
> >
> > I compiled a new kernel without pccard support, and now it boots, even
> > the sdata wiht the new ids and the hd dma. I'm pretty sure i will get
> > used to the dma speed quickly :)
> >
> >
> > I added the DEBUG param in devpccard and got:
> > #Y0: Ricoh 476 PCI/Cardbus bridge, B0000000 intl 11
> > engine(0): SlotFull(CardPowered)
> > configuring slot 0 (SLotPowered)
> > engine(0): SloptPowered(CardConfigured)
> >
> > that is with the card plugged and the kernel hanging, without the
> > card, the kernel has started to work (before wasn't, the *nopcirouting
> > seems to be unrelated as i tested with and without and get the same
> > results, i suppose i'm the guilty of that previous failure)
> >
> > so i suppose the problem is between devpccard and rtl8139 driver.
> >
> > that was the only pccard i have, i will look into rtl8139 to see if i
> > can get a conclusion.
> >
> > thanks all,
> >
> > gabi
>

Re: [9fans] pcmcia/cardbus rtl8139 support

[geoff]

[-- Attachment #1: Type: text/plain, Size: 29 bytes --]

Done and pushed to sources.

[-- Attachment #2: Type: message/rfc822, Size: 2566 bytes --]

From: Richard Miller <9fans@hamnavoe.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] pcmcia/cardbus rtl8139 support
Date: Wed, 28 Mar 2007 09:26:07 +0100
Message-ID: <fdd5c49c35baea31bfb583125151d560@hamnavoe.com>

> I fixed the bug in rtl8139reset that wasn't freeing the port when
> rtl8139reset failed.

ethervt6102.c needs a similar correction:

952a953
> 			iofree(port);
958a960
> 			iofree(port);