2003-06-16 13:11:25

by Mikael Pettersson

[permalink] [raw]
Subject: 2.5.71 cardbus problem + possible solution

2.5.71 changed the name of the Yenta module from yenta_socket
to yenta. In my case (Latitude with RH9 user-space), this
prevented cardmgr from starting properly.

Quick fix: add 'alias yenta_socket yenta' to /etc/modprobe.conf,
or s/yenta_socket/yenta/ in the appropriate config file (but
then you make multi-booting 2.4/2.5 more difficult).

/Mikael


2003-06-16 14:19:06

by Russell King

[permalink] [raw]
Subject: Re: 2.5.71 cardbus problem + possible solution

On Mon, Jun 16, 2003 at 03:25:16PM +0200, [email protected] wrote:
> 2.5.71 changed the name of the Yenta module from yenta_socket
> to yenta. In my case (Latitude with RH9 user-space), this
> prevented cardmgr from starting properly.
>
> Quick fix: add 'alias yenta_socket yenta' to /etc/modprobe.conf,
> or s/yenta_socket/yenta/ in the appropriate config file (but
> then you make multi-booting 2.4/2.5 more difficult).

What do people want to do about this? I have no particular desire to
answer all those emails asking about this, so unless Dominik objects,
I think we should just rename "yenta.c" to "yenta_socket.c" so we have
back-compatibility.

(This issue has appeared because yenta_socket.ko used to be created
by combining yenta.o with pci_socket.o. Since pci_socket.c no longer
exists, we create the module from yenta.c directly, so its now called
yenta.ko.)

--
Russell King ([email protected]) The developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html

2003-06-16 14:25:04

by Mikael Pettersson

[permalink] [raw]
Subject: Re: 2.5.71 cardbus problem + possible solution

Russell King writes:
> On Mon, Jun 16, 2003 at 03:25:16PM +0200, [email protected] wrote:
> > 2.5.71 changed the name of the Yenta module from yenta_socket
> > to yenta. In my case (Latitude with RH9 user-space), this
> > prevented cardmgr from starting properly.
> >
> > Quick fix: add 'alias yenta_socket yenta' to /etc/modprobe.conf,
> > or s/yenta_socket/yenta/ in the appropriate config file (but
> > then you make multi-booting 2.4/2.5 more difficult).
>
> What do people want to do about this? I have no particular desire to
> answer all those emails asking about this, so unless Dominik objects,
> I think we should just rename "yenta.c" to "yenta_socket.c" so we have
> back-compatibility.
>
> (This issue has appeared because yenta_socket.ko used to be created
> by combining yenta.o with pci_socket.o. Since pci_socket.c no longer
> exists, we create the module from yenta.c directly, so its now called
> yenta.ko.)

I'd prefer compatibility, if there are no technical reasons for
breaking it. Reverting to the old name in 2.5.72 sounds like a
good idea.

2003-06-16 14:39:18

by Felipe Alfaro Solana

[permalink] [raw]
Subject: Re: 2.5.71 cardbus problem + possible solution

On Mon, 2003-06-16 at 16:38, [email protected] wrote:
> Russell King writes:
> > On Mon, Jun 16, 2003 at 03:25:16PM +0200, [email protected] wrote:
> > > 2.5.71 changed the name of the Yenta module from yenta_socket
> > > to yenta. In my case (Latitude with RH9 user-space), this
> > > prevented cardmgr from starting properly.
> > >
> > > Quick fix: add 'alias yenta_socket yenta' to /etc/modprobe.conf,
> > > or s/yenta_socket/yenta/ in the appropriate config file (but
> > > then you make multi-booting 2.4/2.5 more difficult).
> >
> > What do people want to do about this? I have no particular desire to
> > answer all those emails asking about this, so unless Dominik objects,
> > I think we should just rename "yenta.c" to "yenta_socket.c" so we have
> > back-compatibility.
> >
> > (This issue has appeared because yenta_socket.ko used to be created
> > by combining yenta.o with pci_socket.o. Since pci_socket.c no longer
> > exists, we create the module from yenta.c directly, so its now called
> > yenta.ko.)
>
> I'd prefer compatibility, if there are no technical reasons for
> breaking it. Reverting to the old name in 2.5.72 sounds like a
> good idea.

I must agree with. I think backwards compatibility is important if we
want widespread adoption of 2.6 from the beginning. But there's a
question I had in mind for long time: is cardmgr really needed? Isn't
hotplug more than enough to handle CardBus devices?

2003-06-16 14:44:28

by Dominik Brodowski

[permalink] [raw]
Subject: Re: 2.5.71 cardbus problem + possible solution

On Mon, Jun 16, 2003 at 04:38:52PM +0200, [email protected] wrote:
> > What do people want to do about this? I have no particular desire to
> > answer all those emails asking about this, so unless Dominik objects,
> > I think we should just rename "yenta.c" to "yenta_socket.c" so we have
> > back-compatibility.

Agreed.

Dominik

2003-06-16 14:45:37

by Russell King

[permalink] [raw]
Subject: Re: 2.5.71 cardbus problem + possible solution

On Mon, Jun 16, 2003 at 04:53:07PM +0200, Felipe Alfaro Solana wrote:
> I must agree with. I think backwards compatibility is important if we
> want widespread adoption of 2.6 from the beginning. But there's a
> question I had in mind for long time: is cardmgr really needed? Isn't
> hotplug more than enough to handle CardBus devices?

We're slowly working towards the point where our reliance on cardmgr
getting less and less.

--
Russell King ([email protected]) The developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html

2003-06-16 14:47:33

by Dominik Brodowski

[permalink] [raw]
Subject: Re: 2.5.71 cardbus problem + possible solution

On Mon, Jun 16, 2003 at 04:53:07PM +0200, Felipe Alfaro Solana wrote:
>
> I must agree with. I think backwards compatibility is important if we
> want widespread adoption of 2.6 from the beginning. But there's a
> question I had in mind for long time: is cardmgr really needed? Isn't
> hotplug more than enough to handle CardBus devices?

One aim of my work is to deprecate cardmgr for both CardBus (32-bit) and
PCMCIA (16-bit) devices. For CardBus, most pieces should be there already,
for PCMCIA the road is much longer.

Dominik