[driverloader] WMP45G PCI on Kernel 2.6.0 temporary freezups (driverloader: message 4 of 20)

[Linuxant support]

Regis,

The problem is that the affected objects can sometimes be statically 
used without prior "allocation", so it is not possible in this context 
to just store a pointer to dynamically allocated storage, since there 
is no explicit way of finding out when the object should be freed later 
on.

We have thought of a few possible solutions but they generally add 
significant overhead when spinlocks should remain extremely light and 
efficient. Therefore it was decided to simply exclude support of the 
CONFIG_DEBUG_SPINLOCK option for now since it isn't meant to be used in 
production systems anyway.

Marc

On Jan 23, 2004, at 1:30 AM, Regis "HPReg" Duchesne wrote:

> Marc,
>
> Thanks for your explanation. It makes a lot of sense now.
>
>> I don't think CONFIG_DEBUG_SPINLOCK is at cause here. Recent versions
>> of DriverLoader will refuse to compile if this option is set. Why?
>> because it adds too much overhead to some kernel structures  causing
>> them to no longer fit in their equivalent fixed-size NDIS objects.
>
> Sounds like one way to work around that would be to only store a 
> pointer
> to the kernel structure inside the NDIS object when
> CONFIG_DEBUG_SPINLOCK is set, instead of the whole kernel structure
> itself.
>
> Then DriverLoader would not have to require that CONFIG_DEBUG_SPINLOCK
> be disabled to function properly.
>
> -- 
> Regis
>