Created attachment 756160 [details] autofs 5.1.7 debug output I'm running an NFSv4-only server with autofs on the client: after upgrading from 5.1.7-r1 to autofs-5.1.8 it can no longer mount shares. Attached below are log files of the output of `automount --foreground --debug` for both versions. 5.1.8 seems to be trying NFSv3 despite everything in my config files telling it to use NFSv4; the changelog mentions "refactor get_nfs_info()" which may have caused all this, and that's the part where it times out in the log. If I try to placate it by running rpcbind on the server side, the "get_portmap_client: error 0x3 getting portmap client" lines go away but it otherwise fails the same. I couldn't find an upstream to report bugs to, sorry - the ebuild HOMEPAGE points to web.archive.org and search engines only give me a dead mailing list and a webserver directory of tarballs.
Created attachment 756163 [details] autofs 5.1.8 debug output
sounds a bug which introduced at 5.1.8.. I did a basic test here, which kinds of working fine.. could you paste your /etc/autofs/nfs configuration file? as I see you are using "noatime,proto=tcp6,timeo=20,vers=4.2"? can you try 'proto=tcp'? not sure if it fail to recognize this option
it ssem to be trying to contact the host on port 111, if the ports are blocked (as in in our case, some servers do bloque these are there are unsed for nfs4) it fails miserably
the commit that introdueced it https://git.kernel.org/pub/scm/linux/storage/autofs/autofs.git/commit/CHANGELOG?id=606795ecfaa1ba45ab58fa20376213f6089e42e9 does look like a "feature"
Created attachment 757040 [details, diff] patch to revert the commit that introduced the problem drop in /etc/portage/patches if you want
(In reply to Yixun Lan from comment #2) > sounds a bug which introduced at 5.1.8.. > > I did a basic test here, which kinds of working fine.. > > could you paste your /etc/autofs/nfs configuration file? > as I see you are using "noatime,proto=tcp6,timeo=20,vers=4.2"? can you try > 'proto=tcp'? not sure if it fail to recognize this option sorry for late response. Here's everything that may be relevant, I've toned down the mount options here (proto=tcp,vers=4) and still get the hang with any combination however: ======== # rg -Nv '^($|#)' /etc/autofs/{autofs.conf,auto.master,nfs} /etc/autofs/autofs.conf [ autofs ] timeout = 300 negative_timeout = 10 browse_mode = no mount_nfs_default_protocol = 4 [ amd ] dismount_interval = 0 /etc/autofs/auto.master /mnt/nfs /etc/autofs/nfs browse,noatime,proto=tcp,timeo=20,vers=4 /etc/autofs/nfs firmware -ro rynex:/lib/firmware/ home rynex:/home/ local -ro rynex:/usr/local/ portage rynex:/srv/portage/ src -ro rynex:/usr/src/ www rynex:/srv/www/ ======== Out of curiosity I went browsing through the source code to see how this stuff works and found something in there about setting fstype=nfs4 overriding the faulty autodetection, but trying it everywhere in the config files had no effect. (In reply to foufou33 from comment #5) > Created attachment 757040 [details, diff] [details, diff] > patch to revert the commit that introduced the problem > > drop in /etc/portage/patches if you want That works for me, and looks arguably more correct. Thanks!
The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f7586f437afc004fee03a1fdf586d6dd9b7db480 commit f7586f437afc004fee03a1fdf586d6dd9b7db480 Author: Yixun Lan <dlan@gentoo.org> AuthorDate: 2021-12-22 13:15:58 +0000 Commit: Yixun Lan <dlan@gentoo.org> CommitDate: 2021-12-22 13:17:52 +0000 net-fs/autofs: fix nfsv4 mount issue Closes: https://bugs.gentoo.org/827239 Package-Manager: Portage-3.0.28, Repoman-3.0.3 Signed-off-by: Yixun Lan <dlan@gentoo.org> ...{autofs-5.1.8.ebuild => autofs-5.1.8-r1.ebuild} | 1 + net-fs/autofs/files/autofs-5.1.8-nfsv4-mount.patch | 88 ++++++++++++++++++++++ 2 files changed, 89 insertions(+)