Hi,
I found that running xfs_fsr introduces mmap problems (as detected by
mapcheck)
I ran the following in single user mode:
(mapcheck -v /; xfs_fsr -v; mapcheck -v /) 2>&1 | tee /tmp/xfs.out
The kernel was a CVS version from last night built with gcc3.2
/proc/version:
Linux version 2.4.19-xfs (root@localhost) (gcc version 3.2 (Red Hat Linux
Rawhide 3.2-1)) #1 Tue Aug 20 20:42:28 MDT 2002
abbr. output from 1st mapcheck -v /:
=============================
/var/ftp/lib/ld-2.2.90.so: Text file busy
/usr/bin/tee: Text file busy
/bin/bash: Text file busy
/home/wns/bin/mapcheck: Text file busy
/lib/ld-2.2.90.so: Text file busy
/sbin/init: Text file busy
/sbin/minilogd: Text file busy
323569 files scanned 0 files fixed 7 errors
summary from xfs_fsr -v:
========================
grep DONE /tmp/xfs.out | wc -l
57
abbr. output from 2nd mapcheck -v /:
==============================
323569 files scanned 52 files fixed 7 errors
I grabbed the filenames from the mapcheck -v output and looked up their inode
numbers and compared these to the inodes processed by xfs_fsr:
ino=29413088 extents before:7 after:1 DONE ino=29413088
29413088 /var/log/sa/sa20
ino=29413118 extents before:4 after:1 DONE ino=29413118
29413118 /var/log/sa/sa18
ino=29413309 extents before:3 after:1 DONE ino=29413309
29413309 /var/log/sa/sa17
ino=29439771 extents before:4 after:1 DONE ino=29439771
29439771 /var/log/sa/sa16
I have been running mapcheck regularly in the last 2 weeks on this same system
with current CVS kernels and had not detected any mmap issues.
--
Warren Stockton
mailto: wns@xxxxxxxxxxxxxx
|