Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!constellation!osuunx.ucc.okstate.edu!moe.ksu.ksu.edu!zaphod.mps.ohio-state.edu!darwin.sura.net!udel!sbcs.sunysb.edu!stark.UUCP!gene
From: gene@stark.uucp (Gene Stark)
Newsgroups: comp.os.386bsd.bugs
Subject: Bug in /sbin/restore (PATCH)
Date: 23 Mar 93 22:35:38
Organization: Gene Stark's home system
Lines: 46
Distribution: world
Message-ID: <GENE.93Mar23223538@stark.stark.uucp>
NNTP-Posting-Host: stark.uucp
Cc: gene
After nearly trashing my filesystem with a hastily installed kernel
"experiment", I reached for my level 0 dump so that I could reinstall
my garbled INN installation. Oops, restore didn't seem to understand
that /usr/libexec/inn once was a directory on my system. In fact,
only about half the directories seemed to have made it to my dump
unadulterated. But wait! The interactive restore was terminating
the directory extraction early, immediately after beginning to read
volume 2. After several hours of head bashing, I found the answer:
restore does not work properly if the records for directories span
more than one volume.
Here's the patch that allowed me to restore my stuff. I imagine it
should somehow be directed to an official BSD patch collector (if this
hasn't been spotted already), but I don't know who that might be.
Please, no flames about the fact that I had made and was trying to restore
from a 96 floppy dump!
- Gene Stark
*** /usr/src/sbin/restore/tape.c.orig Tue Apr 16 20:30:40 1991
--- /usr/src/sbin/restore/tape.c Tue Mar 23 22:23:36 1993
***************
*** 358,363 ****
--- 358,375 ----
if (curfile.action == USING) {
if (volno == 1)
panic("active file into volume 1\n");
+ /*
+ * Skip over the dump tape header
+ */
+ if (tmpbuf.c_type == TS_TAPE && (tmpbuf.c_flags & DR_NEWHEADER))
+ for(i = tmpbuf.c_count+1; i > 0; i--)
+ readtape(&tmpbuf);
+ /*
+ * Skip over the inode map
+ */
+ if(tmpbuf.c_type == TS_BITS)
+ for(i = tmpbuf.c_count; i > 0; i--)
+ readtape(&tmpbuf);
return;
}
/*
--
stark@cs.sunysb.edu