Arla 0.19 on Linux 2.1.131 core dump

[Brandon S. Allbery]

It's abort()ing in find_first_fs(), looking for the volume ID of
root.sup.readonly in the volume structure for root.cell.  This makes me
think that it either inappropriately followed the mountpoint for root.sup
(/afs/ece.cmu.edu/support) or failed to follow it to select the correct
volume's cache structure.

This also happened with the 0.17.1 release; it's only just now that I've
had an opportunity to try to track down the arlad cores we've been having
on 2.1 kernels.

BTW, here's a minor diff to xfs/linux/getcwd.c so things don't break if you
reboot to a 2.0 kernel.  The folks who are experimenting with 2.1 around
here jump around a lot, and it's moderately annoying to see the system
break if you forget to zap the LD_PRELOAD.

--- /var/tmp/arla-0.19/xfs/linux/getcwd.c       Tue Dec 22 10:22:35 1998
+++ getcwd.c    Mon Dec 21 09:32:42 1998
@@ -34,6 +34,9 @@
  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  * SUCH DAMAGE.
+ *
+ * 1998-12-21 allbery at ece.cmu.edu
+ * If we fail with ENOSYS, fall back on the old getcwd.
  */
 
 #ifdef HAVE_CONFIG_H
@@ -58,6 +61,9 @@
     retval = sys_getcwd(buf, size);
     if (retval >= 0)
        return buf;
-    errno = -retval;
-    return NULL;
+    if (retval != -ENOSYS) {
+       errno = -retval;
+       return NULL;
+    }
+    return __getcwd(buf, size);
 }


-- 
brandon s. allbery	[os/2][linux][solaris][japh]	 allbery at kf8nh.apk.net
system administrator	     [WAY too many hats]	   allbery at ece.cmu.edu
electrical and computer engineering					 KF8NH
carnegie mellon university	      ["God, root, what is difference?" -Pitr]