Re: 2.6.7: sk98lin unload oops

Re: 2.6.7: sk98lin unload oops

[Christoph Hellwig]

On Sun, Jul 04, 2004 at 04:15:09PM +0100, Christoph Hellwig wrote:
> > Fortunality everything still works fine (I'm running the script over the 
> > network of the syskonnect cards).
> > 
> > This machine has two of those syskonnect cards, on another machine which has 
> > only one syskonnect card this oops doesn't occur.
> 
> As a colleteral damage the following huge patch should fix it, and I need
> testers for it anyway ;-)

Actually the problem sits deeper.  sk98line tries to register a procfile with
the interfacename of the struct net_device.  The patch below (ontop of
the previous one) makes it work unless you change the interface name manually,
but as Linux explicitly allows that the interface is fundamentally broken and
probably should just go away.


--- drivers/net/sk98lin/skge.c~	2004-07-04 19:15:43.219326648 +0200
+++ drivers/net/sk98lin/skge.c	2004-07-04 19:18:21.562254864 +0200
@@ -5119,9 +5119,12 @@
 	if ((pAC->GIni.GIMacsFound == 2) && pAC->RlmtNets == 2)
 		have_second_mac = 1;
 
+	remove_proc_entry(dev->name, pSkRootDir);
 	unregister_netdev(dev);
-	if (have_second_mac)
+	if (have_second_mac) {
+		remove_proc_entry(pAC->dev[1]->name, pSkRootDir);
 		unregister_netdev(pAC->dev[1]);
+	}
 
 	SkGeYellowLED(pAC, pAC->IoBase, 0);
 

Re: 2.6.7: sk98lin unload oops

[Denis Vlasenko]

On Sunday 04 July 2004 18:23, Christoph Hellwig wrote:
> On Sun, Jul 04, 2004 at 04:15:09PM +0100, Christoph Hellwig wrote:
> > > Fortunality everything still works fine (I'm running the script over
> > > the network of the syskonnect cards).
> > >
> > > This machine has two of those syskonnect cards, on another machine
> > > which has only one syskonnect card this oops doesn't occur.
> >
> > As a colleteral damage the following huge patch should fix it, and I need
> > testers for it anyway ;-)
>
> Actually the problem sits deeper.  sk98line tries to register a procfile
> with the interfacename of the struct net_device.  The patch below (ontop of
> the previous one) makes it work unless you change the interface name
> manually, but as Linux explicitly allows that the interface is
> fundamentally broken and probably should just go away.

hostap driver does this too
--
vda

Re: 2.6.7: sk98lin unload oops

[Christoph Hellwig]

[-- Attachment #1: Type: text/plain, Size: 270 bytes --]

> Is your patch for vanilla 2.6.7?

Sorry, it's for Linus' current BK tree.  I've attached three files that
you should be able to just copy over your regular 2.6.7 tree:

include/linux/pci_ids.h
drivers/net/sk98lin/skge.c
drivers/net/sk98lin/h/skdrv2nd.h

(all gzipped)

[-- Attachment #2: pci_ids.h.gz --]
[-- Type: application/x-gzip, Size: 16580 bytes --]

[-- Attachment #3: skge.c.gz --]
[-- Type: application/x-gzip, Size: 34211 bytes --]

[-- Attachment #4: skdrv2nd.h.gz --]
[-- Type: application/x-gzip, Size: 5091 bytes --]

Re: 2.6.7: sk98lin unload oops

[Bernd Schubert]

> Sorry, it's for Linus' current BK tree.  I've attached three files that
> you should be able to just copy over your regular 2.6.7 tree:

Thanks, this driver compiles fine.


[from your other mail]
> the previous one) makes it work unless you change the interface name
> manually, but as Linux explicitly allows that the interface is
> fundamentally broken and probably should just go away.

Unfortunality we rename all interfaces using ifrename to make sure that the 
interface names won't change with different kernel versions (we have this 
problem when we switch between 2.4. and 2.6.). So it is normal that the oops 
occurs on unloading the modules?


Btw, on 22th June I got another skge.c patch from Herbert Xu to fix another 
oops:

http://lkml.org/lkml/2004/6/22/44

This patch applies fine on top of your new versions (with 400 lines offset), 
maybe this patch should also be included into the current BK tree?

Thanks a lot,
	Bernd

Re: 2.6.7: sk98lin unload oops

[Christoph Hellwig]

On Mon, Jul 05, 2004 at 12:01:35AM +0200, Bernd Schubert wrote:
> [from your other mail]
> > the previous one) makes it work unless you change the interface name
> > manually, but as Linux explicitly allows that the interface is
> > fundamentally broken and probably should just go away.
> 
> Unfortunality we rename all interfaces using ifrename to make sure that the 
> interface names won't change with different kernel versions (we have this 
> problem when we switch between 2.4. and 2.6.). So it is normal that the oops 
> occurs on unloading the modules?

Well, the problem is that someone smoked bad crack when designing the sk98lin
procfs interface ;-)  We should probably just kill it and find a better way
to export the information if nessecary.  I'll take a look at that.

> Btw, on 22th June I got another skge.c patch from Herbert Xu to fix another 
> oops:
> 
> http://lkml.org/lkml/2004/6/22/44
> 
> This patch applies fine on top of your new versions (with 400 lines offset), 
> maybe this patch should also be included into the current BK tree?

Jeff already merged that patch.