2003-11-17 20:16:30

by Andrew Pimlott

[permalink] [raw]
Subject: CONFIG_CRC32 in 2.4.22 breaks PCMCIA

CONFIG_CRC32 was introduced in 2.4.22. I found that if I didn't
explicitly set it, the pcnet_cs driver from stand-alone PCMCIA
distribution didn't work. PCMCIA relies on the crc functions, and
since they were always available before 2.4.22, it doesn't check for
them.

This seems to be significant breakage, and it took me a good while
to figure out what was going on. Is this change reasonable in the
stable kernel series?

Andrew


2003-11-17 20:23:51

by Maciej Żenczykowski

[permalink] [raw]
Subject: Re: CONFIG_CRC32 in 2.4.22 breaks PCMCIA

> CONFIG_CRC32 was introduced in 2.4.22. I found that if I didn't
> explicitly set it, the pcnet_cs driver from stand-alone PCMCIA
> distribution didn't work. PCMCIA relies on the crc functions, and
> since they were always available before 2.4.22, it doesn't check for
> them.

Something wrong with the in-kernel pcnet_cs?

> This seems to be significant breakage, and it took me a good while
> to figure out what was going on. Is this change reasonable in the
> stable kernel series?

Well, it's in the help for the CRC32 option that it's available to enable
external-kernel tree drivers to access these functions. If you are
running make oldconfig you'll hit the question and if you don't know what
it's about you should consult help...
Seems reasonable to me. Same aplies to deflate, etc. support routines.

Cheers,
MaZe.


2003-11-17 20:52:22

by Andrew Pimlott

[permalink] [raw]
Subject: Re: CONFIG_CRC32 in 2.4.22 breaks PCMCIA

On Mon, Nov 17, 2003 at 09:23:31PM +0100, Maciej Zenczykowski wrote:
> > CONFIG_CRC32 was introduced in 2.4.22. I found that if I didn't
> > explicitly set it, the pcnet_cs driver from stand-alone PCMCIA
> > distribution didn't work. PCMCIA relies on the crc functions, and
> > since they were always available before 2.4.22, it doesn't check for
> > them.
>
> Something wrong with the in-kernel pcnet_cs?

No, something is wrong with in-kernel ray_cs. :-( Oops posted a
few weeks ago.

http://groups.google.com/groups?selm=fa.gevojmd.732qbh%40ifi.uio.no

> > This seems to be significant breakage, and it took me a good while
> > to figure out what was going on. Is this change reasonable in the
> > stable kernel series?
>
> Well, it's in the help for the CRC32 option that it's available to enable
> external-kernel tree drivers to access these functions. If you are
> running make oldconfig you'll hit the question and if you don't know what
> it's about you should consult help...

I think I used xconfig the first time I configured this kernel
(because I coincidentally wanted to change something). It was a
while ago, and I only tried pcnet_cs today, so my memory isn't
perfect. Maybe I should have used oldconfig first, but I doubt
everyone else does that for stable kernels.

It still seems unwise to change the default in a stable kernel. Let
the people who want it set CONFIG_OMIT_CRC32 or something.

Andrew

2003-11-18 00:45:26

by Ed Tomlinson

[permalink] [raw]
Subject: Re: CONFIG_CRC32 in 2.4.22 breaks PCMCIA

On November 17, 2003 03:40 pm, Andrew Pimlott wrote:
> > Well, it's in the help for the CRC32 option that it's available to enable
> > external-kernel tree drivers to access these functions. ?If you are
> > running make oldconfig you'll hit the question and if you don't know what
> > it's about you should consult help...
>
> I think I used xconfig the first time I configured this kernel
> (because I coincidentally wanted to change something). ?It was a
> while ago, and I only tried pcnet_cs today, so my memory isn't
> perfect. ?Maybe I should have used oldconfig first, but I doubt
> everyone else does that for stable kernels.
>
> It still seems unwise to change the default in a stable kernel. ?Let
> the people who want it set CONFIG_OMIT_CRC32 or something.

Andrew,

I think its reasonable to have to do a make oldconfig for stable kernels.
Stable does not mean new drivers and/or filesystems do not get added...

Ed Tomlinson