fsck can *never* fix I/O errors. I/O errors are hardware failures, and fsck is a tool to fix logical and structural errors in the directories. The syslog error are very important to be checked and acted on ASAP. Consider writing a cron job that looks for problems every hour or two each day and mails the results to your sysadmin team. Something like this:
grep -i -e crit -e warn -e err -e fail -e problem /var/adm/syslog/syslog.log
Now, LVM is smart enough to fix this *IF* bad block relocation has been turned on. Use lvchange -r y to turn on the feature for each logical volume (note: some big diskarrays require this to be off and use diskarray tools to fix the error).
Once BadBlockRelocation is on, the next attepmt to write to the bad spot will cause the relocation to occur and all will be well again. Since you don't have any control over whether directory blocks and metadata is updated, newfs will handle this for you. Don't try to just restore the data without newfs. The directory metadata is corrupted and even though it may eventually get rewritten, some of it must be intact to know where to put the new data.
Note that a bad spot will always stay bad (can't be read) and will get worse with age. If this disk is not mirrored, get the Mirror/UX product, mirror your disks and plan for a replacement. If these are old (and likely very small) disks, look at a complete upgrade with mirroring.
Bill Hassell, sysadmin
