PROBLEM: sk98lin misbehaves with D-Link DGE-530T which doesn't have readable VPD

PROBLEM: sk98lin misbehaves with D-Link DGE-530T which doesn't have readable VPD

[Josip Loncaric]

Driver sk98lin makes repeated attempts to read VPD even after the first 
VpdInit() fails.  This is wrong.

Lots of people seem to be getting repeated "Vpd: Cannot read VPD keys" 
errors.  When nifd is active, this can happen every second, causing 
kernel stalls that disrupt time-critical operations (e.g. DVD use).

Inexpensive cards like D-Link DGE-530T may lack a readable VPD, so there 
is no point in trying to access this missing feature and causing brief 
system stalls (see 
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=136158).

Can sk98lin be enhanced to give up on VPD after the first VpdInit() 
fails?  The NIC appears to operate stably even without a readable VPD.

I haven't tested the experimental driver skge, but the same comment may 
apply.  Obviously it makes no sense to repeatedly look for VPD keys when 
the VPD is missing or incorrect.

Sincerely,
Josip

Re: PROBLEM: sk98lin misbehaves with D-Link DGE-530T which doesn't have readable VPD

[Stephen Hemminger]

On Fri, 09 Sep 2005 11:49:45 -0600
Josip Loncaric <josip@lanl.gov> wrote:

> Driver sk98lin makes repeated attempts to read VPD even after the first 
> VpdInit() fails.  This is wrong.
> 
> Lots of people seem to be getting repeated "Vpd: Cannot read VPD keys" 
> errors.  When nifd is active, this can happen every second, causing 
> kernel stalls that disrupt time-critical operations (e.g. DVD use).

The sk98lin driver is wrapped around VPD for it's silly proprietary
network management interface.  


> Inexpensive cards like D-Link DGE-530T may lack a readable VPD, so there 
> is no point in trying to access this missing feature and causing brief 
> system stalls (see 
> https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=136158).
> 
> Can sk98lin be enhanced to give up on VPD after the first VpdInit() 
> fails?  The NIC appears to operate stably even without a readable VPD.
> 
> I haven't tested the experimental driver skge, but the same comment may 
> apply.  Obviously it makes no sense to repeatedly look for VPD keys when 
> the VPD is missing or incorrect.

Skge doesn't use VPD for anything, it works fine on my D-Link card.
The same data is available for applications that have a need
(like linux-diag lsvpd) via the standard pci interface in sysfs.

Now that skge is in 2.6.13, perhaps the proper thing to do is to take
DGE-530T out of the PCI table for sk98lin?

Re: PROBLEM: sk98lin misbehaves with D-Link DGE-530T which doesn't have readable VPD

[Josip Loncaric]

Stephen Hemminger wrote:
> On Fri, 09 Sep 2005 11:49:45 -0600
> Josip Loncaric <josip@lanl.gov> wrote:
> 
> 
>>Driver sk98lin makes repeated attempts to read VPD even after the first 
>>VpdInit() fails.  This is wrong.
> 
> Now that skge is in 2.6.13, perhaps the proper thing to do is to take
> DGE-530T out of the PCI table for sk98lin?

Sounds good to me -- since your skge works fine -- but there are lots of 
people with Marvell GBE on the motherboard that are in the same 
predicament.  The PCI table for sk98lin could be pruned further (in 
favor of skge).

Sincerely,
Josip