Need new modutils aliases for 2.4.0? (Was) Re: Cannot hotplug cardbus cards.
Quoted 7 lines · Andrew Morton
Quoted 3 lines · Miles Lane
...
Jan 22 12:55:30 agate cardmgr[444]: module
/lib/modules/2.4.1-pre9/pcmcia/3c575_cb.onot available
Quoted 5 lines
The hotplug command is doing the right thing - it's trying to load 3c575_cb.o. But you're using 3c59x.o under 2.4.
I think you need an `alias eth0 3c59x' in modules.conf.
I'll try that. Is this something that modutils should be doing? I kinda think it is. Basically, we'd need aliasing of any pcmcia-cs drivers whose function is now served by a new driver in the 2.4.0 kernel tree.
BTW: I would *really* like to see someone championing creating a smooth transition path for pcmcia-cs users. This means getting all the usermode stuff in place for removing drivers (thanks, David) and also building a way to prepare a PCMCIA/Cardbus card for removal.
I guess the tough part, here, is the PCMCIA support. Which, I gather, isn't really there in the 2.4.0 kernel code, yet. Is that correct?
Thanks,
Miles
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
Quoted 24 lines · Miles Lane
Quoted 10 lines · Andrew Morton
Quoted 2 lines · Miles Lane
Jan 22 12:55:30 agate cardmgr[444]: module
/lib/modules/2.4.1-pre9/pcmcia/3c575_cb.onot availableThe hotplug command is doing the right thing - it's trying to load 3c575_cb.o. But you're using 3c59x.o under 2.4.
I think you need an `alias eth0 3c59x' in modules.conf.
I'll try that. Is this something that modutils should be doing? I kinda think it is. Basically, we'd need aliasing of any pcmcia-cs drivers whose function is now served by a new driver in the 2.4.0 kernel tree.
BTW: I would *really* like to see someone championing creating a smooth transition path for pcmcia-cs users. This means getting all the usermode stuff in place for removing drivers (thanks, David) and also building a way to prepare a PCMCIA/Cardbus card for removal.
Quoted 3 lines
I guess the tough part, here, is the PCMCIA support. Which, I gather, isn't really there in the 2.4.0 kernel code, yet. Is that correct?
On further exploration:
Adding "alias eth0 3c59x" or "alias /dev/eth0 3c59x" or "alias 3c575_cb 3c59x" and then running "depmod -ae" before inserting the 3c575 card did not help.
Also, I am questioning whether /sbin/hotplug uses cardmgr to get the modules loaded. That means that the messages I sent earlier:
Jan 22 16:43:26 agate /sbin/hotplug: arguments (pci) env
(PCI_CLASS 000 PCI_IDB7:5157 ACTIONd TERM=dumb HOSTTYPE=i386 PATH=/bin:/sbin:/usr/sbin:/usr/bin HOME=/ SHELL=/bin/bash DEBUG=yes PCI_SUBSYS_IDB7:5B57 OSTYPE=Linux PCI_SLOT_NAME:00.0 SHLVL=1 _=/usr/bin/env)
Jan 22 16:43:26 agate cardmgr[444]: module /lib/modules/2.4.1-pre9/pcmcia/cb_enabler.o not available Jan 22 16:43:26 agate cardmgr[444]: module
/lib/modules/2.4.1-pre9/pcmcia/3c575_cb.o not available
may simply show that Card Services is trying to handle the driver loading, not /sbin/hotplug.
I will have to check and see whether David Brownell is correct and I am missing the recent hotplug patch sent by Dan Zink.
Anyhow, if /sbin/hotplug is going to handle the driver loading using modprobe, than we will probably need a way to get cardmgr out of the picture.
Miles
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
Quoted 10 lines · Miles Lane
Quoted 5 lines
The hotplug command is doing the right thing - it's trying to load 3c575_cb.o. But you're using 3c59x.o under 2.4.
I think you need an `alias eth0 3c59x' in modules.conf.
I'll try that. Is this something that modutils should be doing? I kinda think it is. Basically, we'd need aliasing of any pcmcia-cs drivers whose function is now served by a new driver in the 2.4.0 kernel tree.
I don't think it's something that should be solved in modutils; there's no reason it needs to know in modutils itself about what modules do what with what kernel version; alias.h is big enough as is. ;)
I would think that if you had working modules.<foo>map support, it shouldn't be a problem.
Bill
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
Quoted 4 lines · Miles Lane
Adding "alias eth0 3c59x" or "alias
/dev/eth03c59x" or "alias 3c575_cb 3c59x" and then running "depmod -ae" before inserting the 3c575 card did not help.
The hotplug script should not need this alias. That's the whole point of having the PCI ID matching stuff, right? (the driver gets loaded based on the device ID, not based on a user request for eth0)
Quoted 3 lines
Also, I am questioning whether
/sbin/hotpluguses cardmgr to get the modules loaded. That means that the messages I sent earlier:
/sbin/hotplug and cardmgr do not interact. cardmgr decided to try loading 3c575_cb based on /etc/pcmcia/config; no hotplug event was involved.
-- Dave
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
Re: Need new modutils aliases for 2.4.0? (Was)
In reply to Miles LaneQuoted 6 lines
From: David Hinds <[email protected]>
Sent: Thursday, January 25, 2001 8:54 PMActually I think all the evidence suggests that at this point the unbundled pcmcia drivers still get considerably more attention than the kernel drivers, at least at the end user level.
You're likely right about that, but I guess I thought that was something that folk wanted to change. Maybe not all at once ... and focussing on Cardbus at first.
Quoted 8 lines
Quoted 4 lines
I would like to explore the "mutual-noninterference pact" between Cardmgr and
/sbin/hotplug. It seems important to get right. Obviously, David Hinds will play a big role in this.I have added code to cardmgr to parse modules.pcimap to figure out which cards should be handled by the hotplug stuff. I still have a few things to clean up but I don't think it will be hard.
That sounds good. Can that also understand the network hotplugging? Or is that going to need work yet?
- Dave
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
http://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel