Re: Here's a new really odd auto.net bug

Re: Here's a new really odd auto.net bug

[Mike Marion]

On  4 Aug, Michael Marion wrote:

> Trying some more things...

Turns out it was my fault again.. I seem to have built the daemon for
sles9-sp1 and sp2 and missed one of the patches.  I rebuilt it again to make
damn sure I hadn't, and the new build works properly. 

Don't know how I missed a patch when I built it for the suse9.2 host at the
same time and didn't miss it.

I'm thinking those 3 patches need to be rolled into the main codebase.

-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
"This software comes with ABSOLUTELY NO WARRANTY. Even if it erases your hard
drive, too bad. Although we did fix that bug from the last release." 
==> README from a long-ago release of DJGPP 

Re: Here's a new really odd auto.net bug

[Ian Kent]

On Fri, 5 Aug 2005, Mike Marion wrote:

> On  4 Aug, Michael Marion wrote:
> 
> > Trying some more things...
> 
> Turns out it was my fault again.. I seem to have built the daemon for
> sles9-sp1 and sp2 and missed one of the patches.  I rebuilt it again to make
> damn sure I hadn't, and the new build works properly. 
> 
> Don't know how I missed a patch when I built it for the suse9.2 host at the
> same time and didn't miss it.
> 
> I'm thinking those 3 patches need to be rolled into the main codebase.

Yep. I need to get back to working on 4.1.

Which three?

I've got four waiting to go into a new revision but the three you are 
refering to may be older still.

Ian

Re: Here's a new really odd auto.net bug

[Mike Marion]

Ian Kent wrote:

>Yep. I need to get back to working on 4.1.
>
>Which three?
>
>I've got four waiting to go into a new revision but the three you are 
>refering to may be older still.
>

The following three that have been in the path for awhile now...

 autofs-4.1.4-misc-fixes.patch
 autofs-4.1.4-multi-parse-fix.patch  
 autofs-4.1.4-non-replicated-ping.patch

-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
If Microsoft built cars instead of software, the airbag system would say
"Are you sure?" before going off.

Re: Here's a new really odd auto.net bug

[Ian Kent]

On Mon, 8 Aug 2005, Mike Marion wrote:

> Ian Kent wrote:
> 
> >Yep. I need to get back to working on 4.1.
> >
> >Which three?
> >
> >I've got four waiting to go into a new revision but the three you are 
> >refering to may be older still.
> >
> 
> The following three that have been in the path for awhile now...
> 
>  autofs-4.1.4-misc-fixes.patch
>  autofs-4.1.4-multi-parse-fix.patch  
>  autofs-4.1.4-non-replicated-ping.patch

OK. I thought you were talking about kernel patches. Sorry.

But that remionds me there was a very important patch from Chris Feist.
I must apply that and post it.

Ian

Re: Here's a new really odd auto.net bug

[Mike Marion]

On 10 Aug, Ian Kent wrote:

> OK. I thought you were talking about kernel patches. Sorry.

No, sorry.. daemon patches.

BTW, so far I've only seen one bug in the updated daemon with those patches,
with ghosting and direct maps.  If a mount gets hung, then HUPing the daemons,
even ones not related to the hung path (/net path hung and direct maps on
paths /prj and /iceng) doesn't seem to work.

e.g. had a host where some new paths under /prj added to map weren't taking 
effect.  Numerous hups to daemons didn't help.  Finally noticed an error about 
stating a filesystem in /net.  Was able to force umount that path, and next 
HUPs made new paths now accessible.   

Can recreate and strace daemon(s) sometime this week for more data.

-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
"...Microsoft has taken a perfectly good standard, broken it, and then told us 
that we have to buy expensive programs that support the broken interface rather
than use the free ones that come with all operating systems in the world except
Microsoft operating systems." 
==> Allen Holub (http://i-want-a-website.com/about-microsoft/quotes.html)

Re: Here's a new really odd auto.net bug

[Ian Kent]

On Sun, 14 Aug 2005, Mike Marion wrote:

> On 10 Aug, Ian Kent wrote:
> 
> > OK. I thought you were talking about kernel patches. Sorry.
> 
> No, sorry.. daemon patches.
> 
> BTW, so far I've only seen one bug in the updated daemon with those patches,
> with ghosting and direct maps.  If a mount gets hung, then HUPing the daemons,
> even ones not related to the hung path (/net path hung and direct maps on
> paths /prj and /iceng) doesn't seem to work.
> 
> e.g. had a host where some new paths under /prj added to map weren't taking 
> effect.  Numerous hups to daemons didn't help.  Finally noticed an error about 
> stating a filesystem in /net.  Was able to force umount that path, and next 
> HUPs made new paths now accessible.   
> 
> Can recreate and strace daemon(s) sometime this week for more data.

That would be useful, thanks.

I fear that the resolution may not be simple as the daemon works as a 
finite state machine. If something blocks state transitions then 
everything stops.

Bummer.

Ian

Re: Here's a new really odd auto.net bug

[Michael Marion]

On  3 Aug, Ian Kent wrote:

>> Aug  2 13:43:54 kydland automount[631]: lookup(program): lookup for
>> cornholio/tftpboot failed
>   ^^^^^^^^^^^^^^^^^^
> This looks a bit like the bad lookup bug.
> 
> The patch below is for 2.6 so you may need to apply it by had for a 2.4 
> kernel.
> 
> Give it a try.

Didn't seem to fix at least the one problem:
rs-workstation mmarion {527}$ ls /net/tau
licenses  lost+found  mlb  usr2  workspace
rs-workstation mmarion {528}$ /etc/auto.net tau
-fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
\
        /local/mnt tau:/local/mnt
rs-workstation mmarion {530}$ mount | grep tau
tau:/local/mnt on /net/tau type nfs
(rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.39)

Logs are same:
Aug  4 17:31:58 rs-workstation automount[9487]: mount(nfs): host /local/mnt:
lookup failure
Aug  4 17:31:58 rs-workstation automount[9487]: >> nfs server reported tcp not
available, falling back to udp
Aug  4 17:31:58 rs-workstation automount[9487]: >> Using NFS over UDP can
cause data corruption.
Aug  4 17:31:58 rs-workstation automount[9487]: >> Please refer to the
WARNINGS section of the nfs(5) manual page.

2.6.8-24.16-smp amd64 sles9-sp2 system btw.. but I've seen the problem on 
sles9 and -sp1 hosts too.  

This is the _really_ confusing part:
cornholio ~ {605}$ /etc/auto.net tau
-fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
\
	/local/mnt tau:/local/mnt
cornholio ~ {606}$ ls /net/tau
local
cornholio ~ {607}$ mount | grep tau
tau:/local/mnt on /net/tau/local/mnt type nfs
(rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.39)

Suse 9.2 host, but exact same kernel as the sles9 and sp1 hosts (used a
src.rpm from suse 9.2 for autofs4 patches that sles9 didn't have) and exact
same automount daemon.. only other difference is this is a 32bit host vs the
rest are 64bit.

Trying some more things...
-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
Frasier: "Well I had to, what with you underneath the covers with a flashlight 
looking at the National Geographics." 
Niles: "I was looking at the maps!"
Frasier: "THAT'S what makes it so scary!" -- Frasier

Re: Here's a new really odd auto.net bug

[Michael Marion]

On  3 Aug, Jeff Moyer wrote:

> Then trigger the bug and send the logs.  I think the logs you sent were
> incomplete.

Didn't seem to add much more... same output with attempt to mount /net/tau
(which should do tau:/local/mnt on /net/tau/local/mnt but puts it on /net/tau
directly instead):

Aug  4 15:42:16 rs-workstation automount[20058]: mount(nfs): host /local/mnt:
lookup failure
Aug  4 15:42:16 rs-workstation automount[20058]: >> nfs server reported tcp
not available, falling back to udp
Aug  4 15:42:16 rs-workstation automount[20058]: >> Using NFS over UDP can
cause data corruption.
Aug  4 15:42:16 rs-workstation automount[20058]: >> Please refer to the
WARNINGS section of the nfs(5) manual page.

Building kernel with Ian's suggested patch now...

-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
"There are no significant bugs in our released software that any
significant number of users want fixed." - Bill Gates in an interview with
Focus magazine, Oct 23, 1995.

Here's a new really odd auto.net bug

[Michael Marion]

Had a similar problem in the past that turned out to be my fault, wrong
/etc/auto.net script in place.  Today, I'm seeing similar mount issue, but
I've made sure the right script is in place.

If a host has >1 export, it works sort of as expected:
kydland mmarion {728}$ /etc/auto.net cornholio
-fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
\
        /local/mnt cornholio:/local/mnt \
        /tftpboot cornholio:/tftpboot
kydland mmarion {729}$ ls /net/cornholio
local  tftpboot
kydland mmarion {730}$ mount | grep cornholio
cornholio:/local/mnt on /net/cornholio/local/mnt type nfs
(rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.38)

but /tftpboot isn't there and debug/verbose output in syslog shows:
Aug  2 13:43:01 kydland automount[32588]: attempting to mount entry
/net/cornholio
Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
Aug  2 13:43:01 kydland automount[581]: >> mount: wrong fs type, bad option,
bad superblock on cornholio:/tftpboot,
Aug  2 13:43:01 kydland automount[581]: >>        or too many mounted file
systems
Aug  2 13:43:01 kydland automount[581]: mount(nfs): nfs: mount failure
cornholio:/tftpboot on /net/cornholio/tftpboot

trying to ls that tftpboot path gives more logs:
Aug  2 13:43:54 kydland automount[32588]: attempting to mount entry
/net/cornholio/tftpboot
Aug  2 13:43:54 kydland automount[631]: >> showmount: can't get address for
cornholio/tftpboot
Aug  2 13:43:54 kydland automount[631]: lookup(program): lookup for
cornholio/tftpboot failed
Aug  2 13:43:54 kydland automount[631]: failed to mount
/net/cornholio/tftpboot
Aug  2 13:43:54 kydland automount[631]: umount_multi: no mounts found under
/net/cornholio/tftpboot
Aug  2 13:43:54 kydland automount[631]: rm_unwanted: /net/cornholio/tftpboot
Aug  2 13:43:54 kydland automount[32588]: attempting to mount entry
/net/cornholio/tftpboot
Aug  2 13:43:54 kydland automount[643]: >> showmount: can't get address for
cornholio/tftpboot
Aug  2 13:43:54 kydland automount[643]: lookup(program): lookup for
cornholio/tftpboot failed
Aug  2 13:43:54 kydland automount[643]: failed to mount
/net/cornholio/tftpboot
Aug  2 13:43:54 kydland automount[643]: umount_multi: no mounts found under
/net/cornholio/tftpboot


Even worse, trying to mount a host with a single export mounts at the wrong
place (this is similar to what I saw before):
kydland mmarion {733}$ /etc/auto.net tau
-fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
\
        /local/mnt tau:/local/mnt
kydland mmarion {734}$ ls /net/tau
licenses  lost+found  mlb  usr2  workspace
kydland mmarion {735}$ mount | grep tau
tau:/local/mnt on /net/tau type nfs
(rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.39)

It should have mounted as /net/tau/local/mnt not /net/tau

and syslog debug output shows:
Aug  2 13:45:34 kydland automount[32588]: attempting to mount entry /net/tau
Aug  2 13:45:34 kydland automount[977]: mount(nfs): host /local/mnt: lookup
failure
Aug  2 13:45:34 kydland automount[977]: >> nfs server reported tcp not
available, falling back to udp

Host has been up for 61 days, and I could swear it was ok when first booted 
(though I could be wrong).

Module loaded is autofs4, not autofs... patched up to 61 days ago.. :)
automount daemon is 4.1.4 with patches from same timeframe.. added patches:
autofs-4.1.4-misc-fixes.patch
autofs-4.1.4-multi-parse-fix.patch
autofs-4.1.4-non-replicated-ping.patch 

Seeing same effect on multiple 64bit sles9 hosts, but same level of
daemon/patches and exact same kernel (built from exact same src.rpm anyway) on 
a 32bit host (with same /etc/auto.net script) is working fine.

-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
Marge: "Homer, I couldn't help overhearing you warp Bart's mind."
Homer: "And?"
==> The Simpsons

Re: Here's a new really odd auto.net bug

[Jeff Moyer]

==> Regarding [autofs] Here's a new really odd auto.net bug ; Michael Marion <mmarion@qualcomm.com> adds:

mmarion> Had a similar problem in the past that turned out to be my fault,
mmarion> wrong /etc/auto.net script in place.  Today, I'm seeing similar
mmarion> mount issue, but I've made sure the right script is in place.

Are you certain you are capturing all debug output?  Add an entry to
syslog.conf like so:

*.*		 /var/log/debug

Then trigger the bug and send the logs.  I think the logs you sent were
incomplete.


-Jeff

mmarion> If a host has >1 export, it works sort of as expected:
mmarion> kydland mmarion {728}$ /etc/auto.net cornholio
mmarion> -fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
mmarion> \
mmarion>         /local/mnt cornholio:/local/mnt \
mmarion>         /tftpboot cornholio:/tftpboot
mmarion> kydland mmarion {729}$ ls /net/cornholio
mmarion> local  tftpboot
mmarion> kydland mmarion {730}$ mount | grep cornholio
mmarion> cornholio:/local/mnt on /net/cornholio/local/mnt type nfs
mmarion> (rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.38)

mmarion> but /tftpboot isn't there and debug/verbose output in syslog shows:
mmarion> Aug  2 13:43:01 kydland automount[32588]: attempting to mount entry
mmarion> /net/cornholio
mmarion> Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
mmarion> Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
mmarion> Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
mmarion> Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
mmarion> Aug  2 13:43:01 kydland automount[581]: >> mount: wrong fs type, bad option,
mmarion> bad superblock on cornholio:/tftpboot,
mmarion> Aug  2 13:43:01 kydland automount[581]: >>        or too many mounted file
mmarion> systems
mmarion> Aug  2 13:43:01 kydland automount[581]: mount(nfs): nfs: mount failure
mmarion> cornholio:/tftpboot on /net/cornholio/tftpboot

mmarion> trying to ls that tftpboot path gives more logs:
mmarion> Aug  2 13:43:54 kydland automount[32588]: attempting to mount entry
mmarion> /net/cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[631]: >> showmount: can't get address for
mmarion> cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[631]: lookup(program): lookup for
mmarion> cornholio/tftpboot failed
mmarion> Aug  2 13:43:54 kydland automount[631]: failed to mount
mmarion> /net/cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[631]: umount_multi: no mounts found under
mmarion> /net/cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[631]: rm_unwanted: /net/cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[32588]: attempting to mount entry
mmarion> /net/cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[643]: >> showmount: can't get address for
mmarion> cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[643]: lookup(program): lookup for
mmarion> cornholio/tftpboot failed
mmarion> Aug  2 13:43:54 kydland automount[643]: failed to mount
mmarion> /net/cornholio/tftpboot
mmarion> Aug  2 13:43:54 kydland automount[643]: umount_multi: no mounts found under
mmarion> /net/cornholio/tftpboot


mmarion> Even worse, trying to mount a host with a single export mounts at the wrong
mmarion> place (this is similar to what I saw before):
mmarion> kydland mmarion {733}$ /etc/auto.net tau
mmarion> -fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
mmarion> \
mmarion>         /local/mnt tau:/local/mnt
mmarion> kydland mmarion {734}$ ls /net/tau
mmarion> licenses  lost+found  mlb  usr2  workspace
mmarion> kydland mmarion {735}$ mount | grep tau
mmarion> tau:/local/mnt on /net/tau type nfs
mmarion> (rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.39)

mmarion> It should have mounted as /net/tau/local/mnt not /net/tau

mmarion> and syslog debug output shows:
mmarion> Aug  2 13:45:34 kydland automount[32588]: attempting to mount entry /net/tau
mmarion> Aug  2 13:45:34 kydland automount[977]: mount(nfs): host /local/mnt: lookup
mmarion> failure
mmarion> Aug  2 13:45:34 kydland automount[977]: >> nfs server reported tcp not
mmarion> available, falling back to udp

mmarion> Host has been up for 61 days, and I could swear it was ok when first booted 
mmarion> (though I could be wrong).

mmarion> Module loaded is autofs4, not autofs... patched up to 61 days ago.. :)
mmarion> automount daemon is 4.1.4 with patches from same timeframe.. added patches:
mmarion> autofs-4.1.4-misc-fixes.patch
mmarion> autofs-4.1.4-multi-parse-fix.patch
mmarion> autofs-4.1.4-non-replicated-ping.patch 

mmarion> Seeing same effect on multiple 64bit sles9 hosts, but same level of
mmarion> daemon/patches and exact same kernel (built from exact same src.rpm anyway) on 
mmarion> a 32bit host (with same /etc/auto.net script) is working fine.

mmarion> -- 
mmarion> Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
mmarion> Marge: "Homer, I couldn't help overhearing you warp Bart's mind."
mmarion> Homer: "And?"
==> The Simpsons

mmarion> _______________________________________________
mmarion> autofs mailing list
mmarion> autofs@linux.kernel.org
mmarion> http://linux.kernel.org/mailman/listinfo/autofs

Re: Here's a new really odd auto.net bug

[Ian Kent]

On Tue, 2 Aug 2005, Michael Marion wrote:

> Had a similar problem in the past that turned out to be my fault, wrong
> /etc/auto.net script in place.  Today, I'm seeing similar mount issue, but
> I've made sure the right script is in place.
> 
> If a host has >1 export, it works sort of as expected:
> kydland mmarion {728}$ /etc/auto.net cornholio
> -fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
> \
>         /local/mnt cornholio:/local/mnt \
>         /tftpboot cornholio:/tftpboot
> kydland mmarion {729}$ ls /net/cornholio
> local  tftpboot
> kydland mmarion {730}$ mount | grep cornholio
> cornholio:/local/mnt on /net/cornholio/local/mnt type nfs
> (rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.38)
> 
> but /tftpboot isn't there and debug/verbose output in syslog shows:
> Aug  2 13:43:01 kydland automount[32588]: attempting to mount entry
> /net/cornholio
> Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
> Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
> Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
> Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
> Aug  2 13:43:01 kydland automount[581]: >> mount: wrong fs type, bad option,
> bad superblock on cornholio:/tftpboot,
> Aug  2 13:43:01 kydland automount[581]: >>        or too many mounted file
> systems
> Aug  2 13:43:01 kydland automount[581]: mount(nfs): nfs: mount failure
> cornholio:/tftpboot on /net/cornholio/tftpboot
> 
> trying to ls that tftpboot path gives more logs:
> Aug  2 13:43:54 kydland automount[32588]: attempting to mount entry
> /net/cornholio/tftpboot
> Aug  2 13:43:54 kydland automount[631]: >> showmount: can't get address for
> cornholio/tftpboot
> Aug  2 13:43:54 kydland automount[631]: lookup(program): lookup for
> cornholio/tftpboot failed
  ^^^^^^^^^^^^^^^^^^
This looks a bit like the bad lookup bug.

The patch below is for 2.6 so you may need to apply it by had for a 2.4 
kernel.

Give it a try.

Ian

diff -Nurp linux-2.6.12-rc5-mm1.orig/fs/autofs4/root.c linux-2.6.12-rc5-mm1/fs/autofs4/root.c
--- linux-2.6.12-rc5-mm1.orig/fs/autofs4/root.c	2005-05-29 14:46:30.000000000 +0800
+++ linux-2.6.12-rc5-mm1/fs/autofs4/root.c	2005-05-29 14:47:04.000000000 +0800
@@ -306,7 +306,14 @@ static int try_to_fill_dentry(struct den
 		
 		DPRINTK("expire done status=%d", status);
 		
-		return 0;
+		/*
+		 * If the directory still exists the mount request must
+		 * continue otherwise it can't be followed at the right
+		 * time during the walk.
+		 */
+		status = d_invalidate(dentry);
+		if (status != -EBUSY)
+			return 0;
 	}
 
 	DPRINTK("dentry=%p %.*s ino=%p",

Re: Here's a new really odd auto.net bug

[Jeff Moyer]

==> Regarding Re: [autofs] Here's a new really odd auto.net bug ; Ian Kent <raven@themaw.net> adds:

raven> On Tue, 2 Aug 2005, Michael Marion wrote:
>> Had a similar problem in the past that turned out to be my fault, wrong
>> /etc/auto.net script in place.  Today, I'm seeing similar mount issue, but
>> I've made sure the right script is in place.
>> 
>> If a host has >1 export, it works sort of as expected:
>> kydland mmarion {728}$ /etc/auto.net cornholio
>> -fstype=nfs,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,nodev,nosuid
>> \
>> /local/mnt cornholio:/local/mnt \
>> /tftpboot cornholio:/tftpboot
>> kydland mmarion {729}$ ls /net/cornholio
>> local  tftpboot
>> kydland mmarion {730}$ mount | grep cornholio
>> cornholio:/local/mnt on /net/cornholio/local/mnt type nfs
>> (rw,nosuid,nodev,rsize=8192,wsize=8192,hard,intr,acdirmin=1,acdirmax=5,acregmin=1,acregmax=5,addr=129.46.234.38)
>> 
>> but /tftpboot isn't there and debug/verbose output in syslog shows:
>> Aug  2 13:43:01 kydland automount[32588]: attempting to mount entry
>> /net/cornholio
>> Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
>> Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
>> Aug  2 13:43:01 kydland kernel: nfs_get_root: getattr error = 116
>> Aug  2 13:43:01 kydland kernel: nfs_read_super: get root inode failed
>> Aug  2 13:43:01 kydland automount[581]: >> mount: wrong fs type, bad option,
>> bad superblock on cornholio:/tftpboot,
>> Aug  2 13:43:01 kydland automount[581]: >>        or too many mounted file
>> systems
>> Aug  2 13:43:01 kydland automount[581]: mount(nfs): nfs: mount failure
>> cornholio:/tftpboot on /net/cornholio/tftpboot
>> 
>> trying to ls that tftpboot path gives more logs:
>> Aug  2 13:43:54 kydland automount[32588]: attempting to mount entry
>> /net/cornholio/tftpboot
>> Aug  2 13:43:54 kydland automount[631]: >> showmount: can't get address for
>> cornholio/tftpboot
>> Aug  2 13:43:54 kydland automount[631]: lookup(program): lookup for
>> cornholio/tftpboot failed
raven>   ^^^^^^^^^^^^^^^^^^
raven> This looks a bit like the bad lookup bug.

raven> The patch below is for 2.6 so you may need to apply it by had for a 2.4 
raven> kernel.

raven> Give it a try.

I thought we would only hit that bug if ghosting was enabled?

-Jeff