Re: Vortex 3c900 passing driver parameters

Re: Vortex 3c900 passing driver parameters

[Randy.Dunlap]

Andrew Morton wrote:
| Karel Kulhavý <clock@twibright.com> wrote:
| >
| > Hello
| > 
| > How do I do a ether=... (kernel boot-time) equivalent of
| > insmod 3c59x.o options=0x201 full_duplex=1 ?
| 
| Unfortunately you cannot.  `ether=' is broken for all drivers which use the
| new(ish) alloc_etherdev() API.
| 
| It is due to ordering problems: the name of the interface is not known at
| the time of parsing the setup info and nobody has got down and worked out
| how to fix it.

Does this ordering problem apply to both 2.4.current and 2.6.0-test?

--
~Randy

Re: Vortex 3c900 passing driver parameters

[Andrew Morton]

"Randy.Dunlap" <rddunlap@osdl.org> wrote:
>
> Andrew Morton wrote:
> | Karel Kulhavý <clock@twibright.com> wrote:
> | >
> | > Hello
> | > 
> | > How do I do a ether=... (kernel boot-time) equivalent of
> | > insmod 3c59x.o options=0x201 full_duplex=1 ?
> | 
> | Unfortunately you cannot.  `ether=' is broken for all drivers which use the
> | new(ish) alloc_etherdev() API.
> | 
> | It is due to ordering problems: the name of the interface is not known at
> | the time of parsing the setup info and nobody has got down and worked out
> | how to fix it.
> 
> Does this ordering problem apply to both 2.4.current and 2.6.0-test?

Well it was a problem a year or so ago.

But init_netdev()'s call to netdev_boot_setup_check() looks like it
should fix things up, so I'm not sure what's going on...

Re: Vortex 3c900 passing driver parameters

[Randy.Dunlap]

On Tue, 14 Oct 2003 21:13:16 -0700 Andrew Morton <akpm@osdl.org> wrote:

| "Randy.Dunlap" <rddunlap@osdl.org> wrote:
| >
| > Andrew Morton wrote:
| > | Karel Kulhavý <clock@twibright.com> wrote:
| > | >
| > | > Hello
| > | > 
| > | > How do I do a ether=... (kernel boot-time) equivalent of
| > | > insmod 3c59x.o options=0x201 full_duplex=1 ?
| > | 
| > | Unfortunately you cannot.  `ether=' is broken for all drivers which use the
| > | new(ish) alloc_etherdev() API.
| > | 
| > | It is due to ordering problems: the name of the interface is not known at
| > | the time of parsing the setup info and nobody has got down and worked out
| > | how to fix it.
| > 
| > Does this ordering problem apply to both 2.4.current and 2.6.0-test?
| 
| Well it was a problem a year or so ago.
| 
| But init_netdev()'s call to netdev_boot_setup_check() looks like it
| should fix things up, so I'm not sure what's going on...

[I'm looking at 2.4.22 since that is where the problem was reported.]

I don't see anything in 3c59x.c that calls init_netdev().
vortex_probe1() calls alloc_etherdev().
It doesn't call init_etherdev() -> init_netdev(),
AFAICT. ??

But I agree with you, I think there is still an ordering problem.

--
~Randy

Vortex 3c900 passing driver parameters

[Karel Kulhavý]

Hello

How do I do a ether=... (kernel boot-time) equivalent of
insmod 3c59x.o options=0x201 full_duplex=1 ?

I have been looking in Documentation/networking/vortex.txt
and make htmldocs, also on the driver homepage at Scyld and found
it on none of these places.

Cl<

Re: Vortex 3c900 passing driver parameters

[Andrew Morton]

Karel Kulhavý <clock@twibright.com> wrote:
>
> Hello
> 
> How do I do a ether=... (kernel boot-time) equivalent of
> insmod 3c59x.o options=0x201 full_duplex=1 ?
> 

Unfortunately you cannot.  `ether=' is broken for all drivers which use the
new(ish) alloc_etherdev() API.

It is due to ordering problems: the name of the interface is not known at
the time of parsing the setup info and nobody has got down and worked out
how to fix it.