After the update to FC5 (rawhide was fine) rpc.svcgssd dies at random times with the following error (taken with OPTIONS="-v -r" in the init script): rpc.svcgssd[13785]: leaving poll rpc.svcgssd[13785]: handling null request rpc.svcgssd[13785]: rpc.svcgssd[13785]: sname = nfs/foo.domain@DOMAIN kernel: rpc.svcgssd[13785]: segfault at 0000000000000000 rip 00002b55a8a76160 rsp 00007fffffcfd328 error 4
It looks like nscd is causing it, here is the stack trace (no debuginfo yet though, sorry). #0 0x00002aaaab865160 in strlen () from /lib64/libc.so.6 #1 0x00002aaaab8ec988 in __nscd_getgrouplist () from /lib64/libc.so.6 #2 0x00002aaaab8814e4 in internal_getgrouplist () from /lib64/libc.so.6 #3 0x00002aaaab881703 in getgrouplist () from /lib64/libc.so.6 #4 0x00002aaaaaddd19b in nss_gss_princ_to_grouplist () from /usr/lib64/libnfsidmap.so.0 #5 0x0000555555557461 in main () from /usr/sbin/rpc.svcgssd Disabling nscd cured the problem in both my servers.
Here is the output with glibc-debuginfo installed. Should I reassign the bug to glibc? (gdb) where #0 0x00002aaaab865160 in ?? () from /lib64/libc.so.6 #1 0x00002aaaab8ec988 in __nscd_getgrouplist (user=0x0, group=0, size=0x7fffffa0a7c0, groupsp=0x7fffffa0a7c8, limit=Variable "limit" is not available. ) at nscd_initgroups.c:40 #2 0x00002aaaab8814e4 in internal_getgrouplist (user=0x0, group=0, size=0x7fffffa0a7c0, groupsp=0x7fffffa0a7c8, limit=-1) at initgroups.c:60 #3 0x00002aaaab881703 in getgrouplist (user=0x0, group=0, groups=0x7fffffa0a89c, ngroups=0x7fffffa0a898) at initgroups.c:158 #4 0x00002aaaaaddd19b in nss_gss_princ_to_grouplist () from /usr/lib64/libnfsidmap.so.0 #5 0x0000555555557461 in main () from /usr/sbin/rpc.svcgssd
Aha http://linux-nfs.org/pipermail/nfsv4/2006-March/003775.html Reassigning to nfs-utils-lib.