[Bug 235582] rpc_svc_gss / nfsd kernel panic

[bugzilla-noreply at freebsd.org]

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

--- Comment #18 from Rick Macklem <rmacklem at FreeBSD.org> ---
Yes, I suspect that "-n 1" on the nfsd startup would probably
mask the problem, although I would not recommend that.
(This assumes that I am correct to assume this race between threads
 is the culprit.)

Increasing the size of CLIENT_MAX (which recently became a tunable
in head) might also mask it and will help w.r.t. performance if
you have more than 128 users authenticating from the clients.
(In this case, CLIENT refers to a user on an NFS mount and not an
 NFS mount from a client.)

Interesting that the HASH table is set to 256, which is twice the
number of elements being kept in the hash lists. (Not sure why
anyone would make it that way, but I'm not the author of this stuff.;-)

I am really hoping that the 2nd patch "fix race..." will fix the
problem. (The first patch would just be a "safety belt" in case
the GSSAPI did return GSS_S_COMPLETE for gss_accept_sec_context(),
but somehow leave the cname argument NULL.

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