Skip to contentSkip to conversation

Bug in parse_sun.c:parse_mount() single mount code.

DISCUSSION Hans Deragon 1 participant 1 message Dec 22, 2005 Kernel automounter

lore
Share & tools
lore
#0 lore raw

Greetings.

In parse_sun.c, I notice that parse_mount() does not handle cases the same way if there are multi mount points or not. Why is that? Why isn't the single case treated as a single pass in the multi loop?

I ask this because I have a problem to report. But first, here is the output of /etc/auto.net when querying a host with multiple mount points and one with a single one.

[root@asterix root] /etc/auto.net server1

-fstype=nfs,hard,intr,nodev,nosuid \
        /export/archives server1:/export/archives \
        /export/devel server1:/export/devel \
        /export/users server1:/export/users

[root@asterix root] /etc/auto.net obelix
-fstype=nfs,hard,intr,nodev,nosuid \
/pathxyz obelix:/pathxyz

As you can see, there is not much diff between the two. I am running:

/usr/sbin/automount --debug --ghost --timeout 60 /net program
/etc/auto.net

The following happens in the single mount point case (/var/log/messages; I removed the leading headers on each line to protect the innocent):

: attempting to mount entry /net/obelix
: lookup(program): looking up obelix : lookup(program): obelix -> -fstype=nfs,hard,intr,nodev,nosuid,nonstrict /pathxyz obelix:/pathxyz
: parse(sun): expanded entry: -fstype=nfs,hard,intr,nodev,nosuid,nonstrict /pathxyz obelix:/pathxyz
: parse(sun): dequote("fstype=nfs,hard,intr,nodev,nosuid,nonstrict") -> fstype=nfs,hard,intr,nodev,nosuid,nonstrict
: parse(sun): gathered options: fstype=nfs,hard,intr,nodev,nosuid,nonstrict
: parse(sun): dequote("/pathxyz") -> /pathxyz : parse(sun): dequote("obelix:/pathxyz") -> obelix:/pathxyz
: parse(sun): core of entry: options=fstype=nfs,hard,intr,nodev,nosuid,nonstrict, loc=/pathxyz obelix:/pathxyz : parse(sun): mounting root >>/net<<, mountpoint >>obelix<<, what >>/pathxyz obelix:/pathxyz<<, fstype >>nfs<<, options >>hard,intr,nodev,nosuid<< : mount(nfs): root=/net name=obelix what=/pathxyz obelix:/pathxyz, fstype=nfs, options=hard,intr,nodev,nosuid : mount(nfs): nfs options="hard,intr,nodev,nosuid", nosymlink=0, ro=0
: mount(nfs): host /pathxyz: lookup failure
: mount(nfs): winner = obelix local = 0
: mount(nfs): winner = obelix local = 0
: mount(nfs): from /pathxyz obelix:/pathxyz elected obelix:/pathxyz
: mount(nfs): calling mkdir_path /net/obelix
: mount(nfs): calling mount -t nfs -s -o hard,intr,nodev,nosuid obelix:/pathxyz /net/obelix
: >> nfs server reported tcp not available, falling back to udp
: mount(nfs): mounted obelix:/pathxyz on /net/obelix

Notice that host "/pathxyz" lookup failed. Obviously, since /pathxyz is a path, not a hostname. The NFS share is being mounted, but as /net/obelix, instead of the expected /net/obelix/pathxyz . There is a bug somewhere, but I have not put my finger on it. But I come back to my initial question: why is the single mount code different then the multi mount code? Why not make the single case just a single loop execution of the multi mount code? The multi code works fine, but not the single code.

Any comments are welcomed.

BTW, Ian Kent, I cannot write to you directly. Somewhere my email is
being blocked ([email protected]). Can you check this out?

Best regards,
Hans Deragon
--
Consultant en informatique/Software Consultant
Deragon Informatique inc.
http://www.deragon.biz Open source (contribution):
mailto://[email protected] http://autopoweroff.deragon.biz