Skip to contentSkip to conversation

backfstype parameter supported.

DISCUSSION Hans Deragon 2 participants 4 messages Jun 29, 2004 Kernel automounter

lore
Share & tools
lore
#0 lore raw

Greetings.

Newbie with automount. Its been a week I tried to communicate with H. Peter
Anvin at hpa|at|zytor.com but never received any answer. I then found this mailing list.

cachefs is not supported by Linux. In a mix Solaris/Linux environment,
where NIS is provided by Solaris, many yp maps specify cachefs as the filesystem, but backfstype is set to nfs. Automount does not support backfstype.

Here is a exert of an entry:

.. fstype=cachefs,cachedir=<dir>,backfstype=nfs ...

I changed the code and now I have a modified 4.1.3 version which tries for
the filesystem defined with "fstype" and if this fails, "backfstype" is then tried. It works.

My patch is not perfect though. While trying "cachefs", modprobe errors
shows up in /var/log/messages. Until backfstype also fails, ideally no error message should be emitted. Maybe there is a way to test for the existance of the kernel module before actually try to load it, thus avoiding error messages.

Altough I am a programmer, I do not program in C and do know pratically
nothing about system calls. I would like my patch to be accepted in the main branch and that someone experienced with autofs clean it up.

Please comment. I would be glade to send you the patch, but first I have to
read "man patch" to figure out how to generate one. :) Its a pretty small change to the code.

Best regards,
Hans Deragon
--
Consultant en informatique/Software Consultant

Deragon Informatique inc.     Open source:
http://www.deragon.biz        http://facil.qc.ca (Promotion du libre)
mailto://[email protected]     http://autopoweroff.sourceforge.net (Logiciel)
#1 lore raw
Quoted 6 lines · Hans Deragon

Greetings.

Newbie with automount. Its been a week I tried to communicate with H. Peter
Anvin at hpa|at|zytor.com but never received any answer. I then found this mailing list.

I think Peter is a very busy man.

In any case he's happy for me to handle the autofs maintenance.

Quoted 5 lines

cachefs is not supported by Linux. In a mix Solaris/Linux environment,
where NIS is provided by Solaris, many yp maps specify cachefs as the filesystem, but backfstype is set to nfs. Automount does not support backfstype.

Yes. Handling unknown options is a headache and needs attention.

I haven't been able to work on this at all and I haven't really thought a great deal about how we can handle it in a clean way.

Quoted 4 lines

Here is a exert of an entry:

.. fstype=cachefs,cachedir=<dir>,backfstype=nfs ...

Quoted 8 lines

I changed the code and now I have a modified 4.1.3 version which tries for
the filesystem defined with "fstype" and if this fails, "backfstype" is then tried. It works.

My patch is not perfect though. While trying "cachefs", modprobe errors
shows up in /var/log/messages. Until backfstype also fails, ideally no error message should be emitted. Maybe there is a way to test for the existance of the kernel module before actually try to load it, thus avoiding error messages.

Those messages are difficult to eliminate as they often come from mount(8) itself.

Quoted 8 lines

Altough I am a programmer, I do not program in C and do know pratically
nothing about system calls. I would like my patch to be accepted in the main branch and that someone experienced with autofs clean it up.

Please comment. I would be glade to send you the patch, but first I have to
read "man patch" to figure out how to generate one. :) Its a pretty small change to the code.

Certainly, send the patch over.

But please be aware that I can't say when I will merge it. There's a backlog. I've had some patches for ages and when I start work on them I seem to get side tracked onto bug fixes and the like.

Fact is if it looks like it won't produce side effects, has no hard coded paths and works straight out then there's a good chance.

Bottom line is if it doesn't get merged then it's not a reflection on you and I would encourage you to continue to maintain your patch(es) until I can get to it. Unfortunately most people become to busy to keep their patches up to date and when I eventually ask it becomes a bit of a big job and the vicious cycle continues.

To make a patch I usually create a clean copy of the distribution tar and call it, say, autofs-4.1.3.orig. Then make another, say autofs-4.1.3 and make the changes to it. Then do something like

diff -Nur autofs-4.1.3.orig autofs-4.1.3 >autofs-4.1.3-my_patch.patch

Unified diffs are best please (the -u).

Ian

#2 lore raw

Inline.

Quoted 19 lines · [email protected] wrote:
Quoted 3 lines · Hans Deragon

cachefs is not supported by Linux. In a mix Solaris/Linux environment,
where NIS is provided by Solaris, many yp maps specify cachefs as the filesystem, but backfstype is set to nfs. Automount does not support backfstype.

Yes. Handling unknown options is a headache and needs attention.

I haven't been able to work on this at all and I haven't really thought a great deal about how we can handle it in a clean way.

Quoted 8 lines

Here is a exert of an entry:

.. fstype=cachefs,cachedir=<dir>,backfstype=nfs ...

I changed the code and now I have a modified 4.1.3 version which tries for
the filesystem defined with "fstype" and if this fails, "backfstype" is then tried. It works.

Quoted 6 lines
Quoted 4 lines

My patch is not perfect though. While trying "cachefs", modprobe errors
shows up in /var/log/messages. Until backfstype also fails, ideally no error message should be emitted. Maybe there is a way to test for the existance of the kernel module before actually try to load it, thus avoiding error messages.

Those messages are difficult to eliminate as they often come from mount(8) itself.

In cachefs case, we know that Linux does not support it. Should we simply catch it and immediately try the backfstype? This would avoid the errors.

Is automount used on any OS where cachefs is supported? Anyhow, I could implement a #ifdef LINUX statement which will catch cachefs only for Linux OS. Anybody knows of an actual macro I could use to figure out if the kernel is Linux?

Quoted 9 lines
Quoted 7 lines

Altough I am a programmer, I do not program in C and do know pratically
nothing about system calls. I would like my patch to be accepted in the main branch and that someone experienced with autofs clean it up.

Please comment. I would be glade to send you the patch, but first I have to
read "man patch" to figure out how to generate one. :) Its a pretty small change to the code.

Quoted 4 lines

Certainly, send the patch over.

But please be aware that I can't say when I will merge it.
There's a backlog. I've had some patches for ages and when I start work on them I seem to get side tracked onto bug fixes and the like.

Rgr. I will submit it once I have refined it a bit more.

Quoted 1 line

Fact is if it looks like it won't produce side effects, has no hard coded paths and works straight out then there's a good chance.

For the moment, it is the case. Pretty simple patch.

Thanks,
Hans Deragon
--
Consultant en informatique/Software Consultant

Deragon Informatique inc.     Open source:
http://www.deragon.biz        http://facil.qc.ca (Promotion du libre)
mailto://[email protected]     http://autopoweroff.sourceforge.net (Logiciel)
maintainer
#3 lore raw
Quoted 9 lines · Hans Deragon
Quoted 2 lines

Those messages are difficult to eliminate as they often come from mount(8) itself.

In cachefs case, we know that Linux does not support it. Should we simply catch it and immediately try the backfstype? This would avoid the errors.

Is automount used on any OS where cachefs is supported? Anyhow, I could implement a #ifdef LINUX statement which will catch cachefs only for Linux OS. Anybody knows of an actual macro I could use to figure out if the kernel is Linux?

autofs is a Linux automounter so we need to worry about clashes, like this, with centralised map stores shared by other OSs.

Ian