Description of problem: rpc.mountd dies with SIGFPE Version-Release number of selected component (if applicable): nfs-utils-1.1.0-6.fc8 How reproducible: always Steps to Reproduce: 1. rpc.mountd dies with SIGFPE on a lvm2 400 Gb volume Actual results: strace of rpc.mountd: stat64("/dev/hdf1", {st_mode=S_IFBLK|0640, st_rdev=makedev(33, 65), ...}) = 0 time(NULL) = 1194952068 open("/dev/hdf1", O_RDONLY) = 12 fstat64(12, {st_mode=S_IFBLK|0640, st_rdev=makedev(33, 65), ...}) = 0 uname({sys="Linux", node="backup", ...}) = 0 ioctl(12, BLKGETSIZE64, 0xbfac1dd8) = 0 _llseek(12, 80023650304, [80023650304], SEEK_SET) = 0 read(12, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 4096) = 4096 brk(0x8006c000) = 0x8006c000 lseek(12, 0, SEEK_SET) = 0 read(12, "\353<\0\0\0\0\0\0\0\0\0\0\2\0\0\0\0\0\0\0\0\0\0\0\22\0\2\0\0\0\0\0"..., 69632) = 69632 --- SIGFPE (Floating point exception) @ 0 (0) --- Expected results: Additional info: I use lvm2, and hdf1 is a pv, part of an exported filesystem. The filesystem is 400 Gb, if that matters.
Is SELinux in the picture?
No, selinux is disabled. It seems the problem is inside of libblkid: #0 0xb7fb8ca5 in ?? () from /lib/libblkid.so.1 #1 0xb7fb96c0 in blkid_verify () from /lib/libblkid.so.1 #2 0xb7fb6808 in blkid_get_dev () from /lib/libblkid.so.1 #3 0xb7fb6a3f in ?? () from /lib/libblkid.so.1 #4 0xb7fb75f5 in ?? () from /lib/libblkid.so.1 #5 0x80005c4e in ?? () from /usr/sbin/rpc.mountd #6 0x80006022 in ?? () from /usr/sbin/rpc.mountd #7 0x800060c0 in ?? () from /usr/sbin/rpc.mountd #8 0x8000638b in ?? () from /usr/sbin/rpc.mountd #9 0x80002f51 in ?? () from /usr/sbin/rpc.mountd #10 0x8000315f in ?? () from /usr/sbin/rpc.mountd #11 0x8000d802 in ?? () from /usr/sbin/rpc.mountd #12 0x800044cc in mount_dispatch () from /usr/sbin/rpc.mountd #13 0xb7f56642 in svc_getreq_common_internal () from /lib/libc.so.6 #14 0xb7f55f31 in svc_getreqset_internal () from /lib/libc.so.6 #15 0x8000747d in ?? () from /usr/sbin/rpc.mountd #16 0x80004044 in main () from /usr/sbin/rpc.mountd
Created attachment 257441 [details] patch for libblkid probe.c This patch fixes the problem, which was division by zero cluster_size.
Looks like probably a dup of bug #398281 Can you please test: https://admin.fedoraproject.org/updates/F8/FEDORA-2007-3853 Thanks, -Eric
btw what is actually on /dev/hdf1? Thanks, -Eric
e2fsprogs-libs-1.40.2-11.fc8 fixes the problem for me. /dev/hdf1 is a part of logical volume, it does not have a filesystem superblock in the beginning. I wonder why rpc.mountd tries to determine filesystem type of it.
Thanks for testing the latest release. If you want to dd off the first 1k or so of /dev/hdf1 and attach, I can try to find some time to look at it. It may have an old signature on it...
Created attachment 282831 [details] first 1k of hdf1