ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[Michael Frank]

Hello.

Have some trouble with loading modules (see earlier message).  
Tried to compile a driver in.

dmesg:
-------
isapnp: Scanning for PnP cards...
isapnp: Card Plug & Play Ethernet card
isapnp: 1 Plug and Play card detected total
------

- no further references do isapnp in logs

- Same card works (with pnp disabled (jumper) and driver compiled 
as a module) by modprobing it with io=0x300


- Same card works with 2.4.21-pre5 driver as module both with pnp 
and modual probing

	Regards
	Michael

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[M.H.VanLeeuwen]

[-- Attachment #1: Type: text/plain, Size: 1162 bytes --]

> Hello.
> 
> Have some trouble with loading modules (see earlier message).
> Tried to compile a driver in.
> 
> dmesg:
> -------
> isapnp: Scanning for PnP cards...
> isapnp: Card Plug & Play Ethernet card
> isapnp: 1 Plug and Play card detected total
> ------
> 
> - no further references do isapnp in logs
> 
> - Same card works (with pnp disabled (jumper) and driver compiled
> as a module) by modprobing it with io=0x300
> 
> - Same card works with 2.4.21-pre5 driver as module both with pnp
> and modual probing
> 
>         Regards
>         Michael
> -

Michael,

NE2k ISAPNP broke around 2.5.64, again.  There are 2 parts to the attached
patch, one to move the NIC initialization earlier in the boot sequence
and the second is a HACK to get ne2k to work when compiled into the
kernel, I've never tried NE2k as a module...

1. The level of isapnp_init was moved to after apci.  Since it is now
   after net_dev_init, ISA PNP NICs fail to initialized at boot.

   This fix allows ISA PNP NIC cards to work during net_dev_init, and still
   leaves isapnp_init after apci_init.

2. The second piece kills off a now ?unnecessary? probe.

Works for me,
Martin

[-- Attachment #2: 65.2k.diff --]
[-- Type: text/plain, Size: 670 bytes --]

--- ./drivers/net/ne.c	Tue Mar  4 22:44:58 2003
+++ ../linux-2.5.65/./drivers/net/ne.c	Tue Mar  4 22:41:36 2003
@@ -176,8 +176,10 @@
 		return -ENXIO;
 
 	/* Then look for any installed ISAPnP clones */
+/*  HACK to kill this probe, also done now by new isapnp code
 	if (isapnp_present() && (ne_probe_isapnp(dev) == 0))
 		return 0;
+*/
 
 #ifndef MODULE
 	/* Last resort. The semi-risky ISA auto-probe. */
--- ./drivers/pnp/isapnp/core.c	Tue Mar 18 05:43:40 2003
+++ ../linux-2.5.65/./drivers/pnp/isapnp/core.c	Tue Mar 18 05:43:14 2003
@@ -1173,7 +1173,7 @@
 	return 0;
 }
 
-device_initcall(isapnp_init);
+subsys_initcall(isapnp_init);
 
 /* format is: noisapnp */
 

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[Michael Frank]

Martin,

Thank you for your message and patch which I tested. Firstlye some more background info.

I just started testing 2.5 on all kinds of hw.  This is being done on RH beta 8.094 (phoebe)

This machine is an old 586 w/o ACPI and bootet with acpi=off. I sent messages  and dmesg to your email only.

1) 2.5.65 + acpi20030321, isapnp fails on the _first_ attempt and works on the second and further(rmmod ne and 8390) attempts. There is a kernel message during ifup which may give you a hint

1st attempt:
 kernel: pnp: res: The PnP device '01:01.00' is already active.
 kernel: ne.c: You must supply "io=0xNNN" value(s) for ISA cards.
 ifup: ne device eth0 does not seem to be present, delaying initialization.
 network: Bringing up interface eth0:  failed

2nd (and further after rmmod) attempt:
 kernel: ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
 kernel: Last modified Nov 1, 2000 by Paul Gortmaker
 kernel: NE*000 ethercard probe at 0x2a0: 00 00 ff ff 27 ef
 kernel: eth0: NE2000 found at 0x2a0, using IRQ 15.
 network: Bringing up interface eth0:  succeeded

2) 2.5.65 + acpi20030321 + your patch, pnp does not work at all, but the module can be inserted with modprobe ne io=0x2a0 irq=15. It seems it does not talk to the ne code as Beckers msg missing?

All attempts of ifup eth0 or service network start:

ifup: Cannot find device "eth0"
network: Bringing up loopback interface:  succeeded
kernel: ne.c: You must supply "io=0xNNN" value(s) for ISA cards.
ifup: ne device eth0 does not seem to be present, delaying initialization.
network: Bringing up interface eth0:  failed

modprobe ne io=0x2a0 irq=15:

kernel: ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
kernel: Last modified Nov 1, 2000 by Paul Gortmaker
kernel: NE*000 ethercard probe at 0x2a0: 00 00 ff ff 27 ef
kernel: eth0: NE2000 found at 0x2a0, using IRQ 15.
 

for me not yet ... :-)
Michael


On Monday 24 March 2003 00:26, you wrote:
> > Hello.
> >
> > Have some trouble with loading modules (see earlier
> > message). Tried to compile a driver in.
> >
> > dmesg:
> > -------
> > isapnp: Scanning for PnP cards...
> > isapnp: Card Plug & Play Ethernet card
> > isapnp: 1 Plug and Play card detected total
> > ------
> >
> > - no further references do isapnp in logs
> >
> > - Same card works (with pnp disabled (jumper) and driver
> > compiled as a module) by modprobing it with io=0x300
> >
> > - Same card works with 2.4.21-pre5 driver as module both
> > with pnp and modual probing
> >
> >         Regards
> >         Michael
> > -
>
> Michael,
>
> NE2k ISAPNP broke around 2.5.64, again.  There are 2 parts to
> the attached patch, one to move the NIC initialization earlier
> in the boot sequence and the second is a HACK to get ne2k to
> work when compiled into the kernel, I've never tried NE2k as a
> module...
>
> 1. The level of isapnp_init was moved to after apci.  Since it
> is now after net_dev_init, ISA PNP NICs fail to initialized at
> boot.
>
>    This fix allows ISA PNP NIC cards to work during
> net_dev_init, and still leaves isapnp_init after apci_init.
>
> 2. The second piece kills off a now ?unnecessary? probe.
>
> Works for me,
> Martin

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[Alan Cox]

On Sun, 2003-03-23 at 16:26, M.H.VanLeeuwen wrote:
> NE2k ISAPNP broke around 2.5.64, again.  There are 2 parts to the attached
> patch, one to move the NIC initialization earlier in the boot sequence
> and the second is a HACK to get ne2k to work when compiled into the
> kernel, I've never tried NE2k as a module...
> 
> 1. The level of isapnp_init was moved to after apci.  Since it is now
>    after net_dev_init, ISA PNP NICs fail to initialized at boot.
> 
>    This fix allows ISA PNP NIC cards to work during net_dev_init, and still
>    leaves isapnp_init after apci_init.

We must initialise ACPI before ISAPnP because we need PCI and ACPI to
know what system resources we must not hit. How about moving the
net_dev_init to later ?

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[M.H.VanLeeuwen]

Michael,

Thanks for the info.  Like I said in previous e-mail I've not tested
modules nor do I use them unless absolutely necessary.

The question is who is the maintainer of the NE2K ISAPNP interface?
Is this an NE2k maintainer issue or an ISAPNP issue or a "you have
the hdwr" you fix it thingy?

I can hack on ne2k but don't want to spend time on it if someone else
already is or has a better solution or possibly colliding with ISAPNP
development again as did my last patches sent to Linus.

Martin


Michael Frank wrote:
> 
> Martin,
> 
> Thank you for your message and patch which I tested. Firstlye some more background info.
> 
> I just started testing 2.5 on all kinds of hw.  This is being done on RH beta 8.094 (phoebe)
> 
> This machine is an old 586 w/o ACPI and bootet with acpi=off. I sent messages  and dmesg to your email only.
> 
> 1) 2.5.65 + acpi20030321, isapnp fails on the _first_ attempt and works on the second and further(rmmod ne and 8390) attempts. There is a kernel message during ifup which may give you a hint
> 
> 1st attempt:
>  kernel: pnp: res: The PnP device '01:01.00' is already active.
>  kernel: ne.c: You must supply "io=0xNNN" value(s) for ISA cards.
>  ifup: ne device eth0 does not seem to be present, delaying initialization.
>  network: Bringing up interface eth0:  failed
> 
> 2nd (and further after rmmod) attempt:
>  kernel: ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
>  kernel: Last modified Nov 1, 2000 by Paul Gortmaker
>  kernel: NE*000 ethercard probe at 0x2a0: 00 00 ff ff 27 ef
>  kernel: eth0: NE2000 found at 0x2a0, using IRQ 15.
>  network: Bringing up interface eth0:  succeeded
> 
> 2) 2.5.65 + acpi20030321 + your patch, pnp does not work at all, but the module can be inserted with modprobe ne io=0x2a0 irq=15. It seems it does not talk to the ne code as Beckers msg missing?
> 
> All attempts of ifup eth0 or service network start:
> 
> ifup: Cannot find device "eth0"
> network: Bringing up loopback interface:  succeeded
> kernel: ne.c: You must supply "io=0xNNN" value(s) for ISA cards.
> ifup: ne device eth0 does not seem to be present, delaying initialization.
> network: Bringing up interface eth0:  failed
> 
> modprobe ne io=0x2a0 irq=15:
> 
> kernel: ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
> kernel: Last modified Nov 1, 2000 by Paul Gortmaker
> kernel: NE*000 ethercard probe at 0x2a0: 00 00 ff ff 27 ef
> kernel: eth0: NE2000 found at 0x2a0, using IRQ 15.
> 
> 
> for me not yet ... :-)
> Michael
> 
> On Monday 24 March 2003 00:26, you wrote:
> > > Hello.
> > >
> > > Have some trouble with loading modules (see earlier
> > > message). Tried to compile a driver in.
> > >
> > > dmesg:
> > > -------
> > > isapnp: Scanning for PnP cards...
> > > isapnp: Card Plug & Play Ethernet card
> > > isapnp: 1 Plug and Play card detected total
> > > ------
> > >
> > > - no further references do isapnp in logs
> > >
> > > - Same card works (with pnp disabled (jumper) and driver
> > > compiled as a module) by modprobing it with io=0x300
> > >
> > > - Same card works with 2.4.21-pre5 driver as module both
> > > with pnp and modual probing
> > >
> > >         Regards
> > >         Michael
> > > -
> >
> > Michael,
> >
> > NE2k ISAPNP broke around 2.5.64, again.  There are 2 parts to
> > the attached patch, one to move the NIC initialization earlier
> > in the boot sequence and the second is a HACK to get ne2k to
> > work when compiled into the kernel, I've never tried NE2k as a
> > module...
> >
> > 1. The level of isapnp_init was moved to after apci.  Since it
> > is now after net_dev_init, ISA PNP NICs fail to initialized at
> > boot.
> >
> >    This fix allows ISA PNP NIC cards to work during
> > net_dev_init, and still leaves isapnp_init after apci_init.
> >
> > 2. The second piece kills off a now ?unnecessary? probe.
> >
> > Works for me,
> > Martin

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[M.H.VanLeeuwen]

Alan Cox wrote:
> 
> On Sun, 2003-03-23 at 16:26, M.H.VanLeeuwen wrote:
> > NE2k ISAPNP broke around 2.5.64, again.  There are 2 parts to the attached
> > patch, one to move the NIC initialization earlier in the boot sequence
> > and the second is a HACK to get ne2k to work when compiled into the
> > kernel, I've never tried NE2k as a module...
> >
> > 1. The level of isapnp_init was moved to after apci.  Since it is now
> >    after net_dev_init, ISA PNP NICs fail to initialized at boot.
> >
> >    This fix allows ISA PNP NIC cards to work during net_dev_init, and still
> >    leaves isapnp_init after apci_init.
> 
> We must initialise ACPI before ISAPnP because we need PCI and ACPI to
> know what system resources we must not hit. How about moving the
> net_dev_init to later ?

Here is the ordering of initcall from System.map file w/ my change.
I take it that you want isapnp_init after pci*_init also, or is it sufficient
like it is, after the acpi*_init?

c0410f30 t __initcall_init_bio
c0410f34 t __initcall_acpi_init
c0410f38 t __initcall_acpi_ec_init
c0410f3c t __initcall_acpi_pci_root_init
c0410f40 t __initcall_acpi_pci_link_init
c0410f44 t __initcall_acpi_power_init
c0410f48 t __initcall_acpi_system_init
c0410f4c t __initcall_acpi_event_init
c0410f50 t __initcall_acpi_scan_init
c0410f54 t __initcall_pnp_init
c0410f58 t __initcall_pnp_system_init
c0410f5c t __initcall_isapnp_init   <<<<<< patch moved it here <<<
c0410f60 t __initcall_device_init
c0410f64 t __initcall_deadline_slab_setup
c0410f68 t __initcall_pci_acpi_init
c0410f6c t __initcall_pci_legacy_init
c0410f70 t __initcall_pcibios_irq_init
c0410f74 t __initcall_pcibios_init
c0410f78 t __initcall_net_dev_init
c0410f7c t __initcall_init_8259A_devicefs

Or is this not the place to see initcall ordering?

Looking into moving net_dev_init to later...
Martin

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[Alan Cox]

On Sun, 2003-03-23 at 19:26, M.H.VanLeeuwen wrote:
> Here is the ordering of initcall from System.map file w/ my change.
> I take it that you want isapnp_init after pci*_init also, or is it sufficient
> like it is, after the acpi*_init?

pci and pnpbios must come before isapnp as well. Otherwise IGP based systems
will MCE on boot

Re: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[Michael Frank]

Original 2.5.65. Put some printks in.

Two tries with service network start. Success on the second try.

Here are the messages

First try
Mar 24 08:01:16 mhfl1 network: Setting network parameters:  succeeded
Mar 24 08:01:17 mhfl1 ifup: Cannot find device "eth0"
Mar 24 08:01:18 mhfl1 network: Bringing up loopback interface:  succeeded
Mar 24 08:01:19 mhfl1 kernel: ne: ne_probe
Mar 24 08:01:19 mhfl1 kernel: ne: ne_probe_isapnp

It seems that pnp does not work the first try:

Mar 24 08:01:19 mhfl1 kernel: pnp: res: The PnP device '01:01.00' is already active.

Mar 24 08:01:19 mhfl1 kernel: ne.c: You must supply "io=0xNNN" value(s) for ISA cards.
Mar 24 08:01:19 mhfl1 ifup: ne device eth0 does not seem to be present, delaying initialization.
Mar 24 08:01:19 mhfl1 network: Bringing up interface eth0:  failed

Second try
Mar 24 08:01:20 mhfl1 network: Setting network parameters:  succeeded
Mar 24 08:01:21 mhfl1 ifup: Cannot find device "eth0"
Mar 24 08:01:22 mhfl1 network: Bringing up loopback interface:  succeeded
Mar 24 08:01:23 mhfl1 kernel: ne: ne_probe
Mar 24 08:01:23 mhfl1 kernel: ne: ne_probe_isapnp
Mar 24 08:01:23 mhfl1 kernel: ne.c: ISAPnP reports Generic PNP at i/o 0x2a0, irq 15.
Mar 24 08:01:23 mhfl1 kernel: ne: ne_probe1
Mar 24 08:01:23 mhfl1 kernel: ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
Mar 24 08:01:23 mhfl1 kernel: Last modified Nov 1, 2000 by Paul Gortmaker
Mar 24 08:01:23 mhfl1 kernel: NE*000 ethercard probe at 0x2a0: 00 00 ff ff 27 ef
Mar 24 08:01:23 mhfl1 kernel: eth0: NE2000 found at 0x2a0, using IRQ 15.
Mar 24 08:01:23 mhfl1 kernel: ne: ne_probe_isapnp == 0
Mar 24 08:01:23 mhfl1 kernel: ne: ne_probe
Mar 24 08:01:23 mhfl1 kernel: ne: ne_probe_isapnp
Mar 24 08:01:26 mhfl1 network: Bringing up interface eth0:  succeeded

---------

ne_probe1 and ne_probe_isapnp have only a printk at their top

---------

int __init ne_probe(struct net_device *dev)
{
  unsigned int base_addr = dev->base_addr;

  printk("ne: ne_probe\n");
  SET_MODULE_OWNER(dev);

/* First check any supplied i/o locations. User knows best. <cough> */
  if (base_addr > 0x1ff)	/* Check a single specified location. */
    return ne_probe1(dev, base_addr);
  else if (base_addr != 0)	/* Don't probe at all. */
    return -ENXIO;

/* Then look for any installed ISAPnP clones */
  if (isapnp_present() == 0)
    printk("ne: no isapnp found\n");
  else if (ne_probe_isapnp(dev) == 0) {
    printk("ne: ne_probe_isapnp == 0\n");
    return 0;
  }

#ifndef MODULE
/* Last resort. The semi-risky ISA auto-probe. */
  for (base_addr = 0; netcard_portlist[base_addr] != 0; base_addr++) {
    int ioaddr = netcard_portlist[base_addr];
    if (ne_probe1(dev, ioaddr) == 0)
      return 0;
  }
#endif
  
  return -ENODEV;
}

--------


On Monday 24 March 2003 04:40, Alan Cox wrote:
> On Sun, 2003-03-23 at 19:26, M.H.VanLeeuwen wrote:
> > Here is the ordering of initcall from System.map file w/ my
> > change. I take it that you want isapnp_init after pci*_init
> > also, or is it sufficient like it is, after the acpi*_init?
>
> pci and pnpbios must come before isapnp as well. Otherwise IGP
> based systems will MCE on boot

Resolved: ISAPNP BUG: 2.4.65 ne2000 driver w. isapnp not working

[Michael Frank]

Martin,

Thank you for getting the fix going and sending me this patch which I applid on 2.5.66.

The problem is resolved:

  .config: CONFIG_NE2000=m

  Tested:
    network start on boot: 		OK
    several network restart:		OK
    rmmod ne 8390; network start:	OK
    several network restart:		OK

I like also to thank Adam Belay for getting this fixed.


Best Regards
Michael

 

On Wednesday 26 March 2003 21:01, you wrote:
> Michael,
>
> Adam Belay has posted some PNP patches you may want to try for 2.5.66.
>
> One of them looks like it may help...I haven't tried.
>
> Martin
>
>
>
> diff -Nru a/drivers/pnp/manager.c b/drivers/pnp/manager.c
> --- a/drivers/pnp/manager.c Tue Mar 25 21:44:41 2003
> +++ b/drivers/pnp/manager.c Tue Mar 25 21:44:41 2003
> @@ -632,8 +632,7 @@
>          if (!dev)
>                  return -EINVAL;
>          if (dev->active) {
> - pnp_info("res: The PnP device '%s' is already active.", dev->dev.bus_id);
> - return -EBUSY;
> + return 0; /* the device is already active */
>          }
>          /* If this condition is true, advanced configuration failed, we
> need to get this device up and running * so we use the simple config engine
> which ignores cold conflicts, this of course may lead to new failures */ @@
> -698,8 +697,7 @@
>          if (!dev)
>                  return -EINVAL;
>          if (!dev->active) {
> - pnp_info("res: The PnP device '%s' is already disabled.",
> dev->dev.bus_id); - return -EINVAL;
> + return 0; /* the device is already disabled */
>          }
>          if (dev->status != PNP_READY){
>                  pnp_info("res: Disable failed becuase the PnP device '%s'
> is busy.", dev->dev.bus_id);