[Bug 235582] rpc_svc_gss / nfsd kernel panic

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=235582

--- Comment #8 from Peter Eriksson <peter.x.eriksson at liu.se> ---
I'll try out the proposed fixes. Sound like it could be the culprit. A pity I
haven't really found a way to force the issue yet (or even pinpointed which of
the 100-200 clients caused it). But perhaps it's just a random thing caused by
network retransmissions and load balancing over multiple LACP trunks...


A question - I enabled a lot of debugging output in order to try to see what's
happening and I noticed one strange thing in the "dmesg" output:

> rpcsec_gss: accepted context for \^D\^A (41 bytes) with <mech { 1 2 840 113554 1 2 2 }, qop 0, svc 1>

(I had modified the rpc_gss_log_debug() call to also print the length of the
client_principal member)

Ie, this call around line 1100 or so:

                      rpc_gss_log_debug("accepted context for %s (%d bytes)
with "
                            "<mech %.*s, qop %d, svc %d>",
                            client->cl_rawcred.client_principal->name,
                            client->cl_rawcred.client_principal->len,
                            mechname.length, (char *)mechname.value,
                            client->cl_qop, client->cl_rawcred.service);

The "client_principal->name" output looks garbled.. (In /var/log/messages those
characters get's filtered out so you won't find it there). Perhaps
cl_rawcred.client_principal isn't supposed to be printable at all.

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