[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Pcmcia and nfs



Fabio Massimo Di Nitto <f.dinitto@seabone.net> wrote:

> I don't use /etc/pcmcia/* to configure my pcmcia cards since i have
> some problems using netenv + pcmcia with 2 eth cards (both ethernet
> get same address etc..) so i have to setup them by hand (i used to few
> scripts to do it) but the problem come out when I forget to umount
> nfs partition (i know that a good admin never forget and so on) but
> sometimes happen also to me.

If you customize your system so that it does not use the method provided 
by the package, then it is your responsibility that you replace that 
method with something that will do an equivalent job.  I cannot be 
responsible for what happens when you override my scripts.

> So this is the reason why i think that stop pcmcia right after
> umounting all nfs partition can be a good solution or to evaluate the
> possibility to perform a check in /etc/init.d/pcmcia stop code about
> nfs (and maybe also samba or other network shared filesystems ...why
> not) mountend partition over pcmcia... umount them and then stop the
> pcmcia service itself.

Shutting down the pcmcia services involves two distinct and separate steps:

(1) turning off and "ejecting" the cards; and

(2) stopping the cardmgr daemon.

These two steps must be completely independent to allow for the first step 
to be performed without performing the second.  (Otherwise, "hot-swapping" 
of cards is not possible.)  Unmounting network filesystems is part of step 
1.  Performing this task during step 2 is unnecessarily redundant, since 
step 2 is never performed unless step 1 has been completed.  Since the 
/etc/init.d/pcmcia script is part of step 2, the NFS umount code should 
not be moved there.

> If you think this can be done I'll work on it asap otherwise I'll
> simply change run levels of my box.

You should set up your system to umount the NFS filesystems when the card 
is shut down (during the "eject" stage).  You can borrow the code in the 
/etc/pcmcia/network script if you want (and actually, I recommend that you 
do), but you should modify your scripts to do this important task.

> Might be also interesting to take in mind the possibility to improve
> netenv to handle 2 eths pcmcia cards at the same time.

I have nothing to do with the netenv package.  If you want to work on this 
or offer suggestions, I suggest that you contact the netenv author.  I'm 
sure he would welcome any prototype code or suggestions that you are 
willing to give him.

Sincerely,

- Brian




Reply to: