[Bug 297626] nfs mount of a NFSv3 share will hang if "Permission denied"

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=297626

--- Comment #3 from Dennis Clarke <[email protected]> ---

For the sake of being pedantic I made a small change. Now the machine
is running 15.1 : 

Aug 18 11:08:32 hydra kernel: ---<<BOOT>>---
Aug 18 11:08:32 hydra kernel: Copyright (c) 1992-2025 The FreeBSD Project.
Aug 18 11:08:32 hydra kernel: Copyright (c) 1979, 1980, 1983, 1986, 1988, 1989,
1991, 1992, 1993, 1994
Aug 18 11:08:32 hydra kernel:   The Regents of the University of California.
All rights reserved.
Aug 18 11:08:32 hydra kernel: FreeBSD is a registered trademark of The FreeBSD
Foundation.
Aug 18 11:08:32 hydra kernel: FreeBSD 15.1-RELEASE-p2
releng/15.1-n283596-aadd58dddcbc GENERIC amd64
Aug 18 11:08:32 hydra kernel: FreeBSD clang version 19.1.7
(https://github.com/llvm/llvm-project.git llvmorg-19.1.7-0-gcd708029e0b2)


Open two XTerms and ssh into the machine and su - root.

Then in one XTerm :

hydra# tcpdump -s 0 -w out.pcap host pluto
tcpdump: listening on re0, link-type EN10MB (Ethernet), snapshot length 262144
bytes


Just leave that running I guess?

In the other XTerm :

hydra# 
hydra# uptime
11:42AM  up 34 mins, 3 users, load averages: 0.35, 0.40, 0.43
hydra# uname -a
FreeBSD hydra 15.1-RELEASE-p2 FreeBSD 15.1-RELEASE-p2
releng/15.1-n283596-aadd58dddcbc GENERIC amd64
hydra# mkdir /mnt/pluto
mkdir: /mnt/pluto: File exists
hydra# mount -o ro pluto:/opt/bulk/passport /mnt/pluto
[tcp] pluto:/opt/bulk/passport: Permission denied
[tcp] pluto:/opt/bulk/passport: Permission denied
[tcp] pluto:/opt/bulk/passport: Permission denied
[tcp] pluto:/opt/bulk/passport: Permission denied
[tcp] pluto:/opt/bulk/passport: Permission denied
[tcp] pluto:/opt/bulk/passport: Permission denied
[tcp] pluto:/opt/bulk/passport: Permission denied
^C
hydra# echo $?
130
hydra# 

Those messages seem to pop up every 60 seconds like clock work. I am
willing to bet I could go for a walk around the neighbourhood and when
I return there would be a stack of those messages. Never ending?

Meanwhile the tcpdump XTerm shows me :

hydra# 
hydra# tcpdump -s 0 -w out.pcap host pluto
tcpdump: listening on re0, link-type EN10MB (Ethernet), snapshot length 262144
bytes
^C266 packets captured
1015 packets received by filter
0 packets dropped by kernel
hydra# 

WireShark/termshark has nothing interesting to reveal at all. Nothing.

There are half a dozen packets of DNS traffic to get the host pluto
resolved and then a trivial NFSv3 mount. The reply comes back in about
1ms as ERR_ACCESS. No surprise. A few packets of FIN, ACK and SYN with
60 seconds of nothing and then again GETPORT Call NFS(100003) Version:3 TCP
and there is nothing of interest to see here. We get the status 13 ERR_ACCESS
and the whole show goes into what looks like an infinite loop with the
same stuff every 60 secs. Over and over. The last group of this noise is
at 360 secs which would be when I hit CTRL-C. The only interesting packet
in the whole show is one ARP packet asking Who has 192.168.2.42? Big deal.

So, once again, the mount hangs and stays that way. Possibly forever.

On the other hand we have this :

hydra# 
hydra# mount -o ro pluto:/opt/bulk/users /mnt/users
hydra# echo $?
0
hydra# 

Happens pretty quick. tcpdump claims 270 packets received by filter.
They are totally boring. Some DNS and then NFS GETPORT then a call
for mount and a reply on port 876. Very fast. We get a V3 status reply
of "Status: OK" and Flavor: AUTH_UNIX and some GETATTR with eventual
V3 Procedure PATHCONF and Status NFS3_OK. There is nothing to see here.

However Permission denied results in what looks to be a hung in
an infinite loop. I can leave it all day just to be really pedantic
but this is wrong behavior.

-- 
You are receiving this mail because:
You are the assignee for the bug.