[Bug libc/6966] New: hsearch almost always probes the same index when using the secondary hash

[Bug libc/6966] New: hsearch almost always probes the same index when using the secondary hash

[bugzilla at bsb dot me dot uk]

hsearch references Kunth vol. 3 6.4 on "open addressing" but the code does
implement the algorithm from Knuth.  Instead of generating the second hash from
the first, the code generates the second hash from the first index.  Because
this hash is then used to step the index, the result is that the first index
generated is (almost) always the same.

This partly defeats the purpose of using a secondary hash.

I say "almost" because it happens for all but 2 of the possible index values in
any given table.  For example, in a table of size 11, all secondary probes will
start at index 10 unless the inital hash produced 9 or 10.

This was reported on news:comp.lang.c by James Dow Allen and I decided to check
on it.  I have a simple patch but I am not sure of the procedure for posting
patches to glibc.

-- 
           Summary: hsearch almost always probes the same index when using
                    the secondary hash
           Product: glibc
           Version: 2.8
            Status: NEW
          Severity: minor
          Priority: P3
         Component: libc
        AssignedTo: drepper at redhat dot com
        ReportedBy: bugzilla at bsb dot me dot uk
                CC: glibc-bugs at sources dot redhat dot com


http://sourceware.org/bugzilla/show_bug.cgi?id=6966

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/6966] hsearch almost always probes the same index when using the secondary hash

[bugzilla at bsb dot me dot uk]

------- Additional Comments From bugzilla at bsb dot me dot uk  2008-10-19 01:33 -------
Created an attachment (id=3006)
 --> (http://sourceware.org/bugzilla/attachment.cgi?id=3006&action=view)
Suggested fix.

The patch suggests two related changes.  First, that idx be set from hval
without reducing havl % htab->size.  hval2 can now be derived from hval without
generating the same probe sequence all the time.  Secondly, a new variable must
be used to detect the wrap-around, since hval no longer contains the initial
value of idx.

-- 


http://sourceware.org/bugzilla/show_bug.cgi?id=6966

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/6966] hsearch almost always probes the same index when using the secondary hash

[drepper at redhat dot com]

------- Additional Comments From drepper at redhat dot com  2008-11-01 15:37 -------
Applied to cvs.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://sourceware.org/bugzilla/show_bug.cgi?id=6966

------- You are receiving this mail because: -------
You are on the CC list for the bug, or are watching someone who is.

[Bug libc/6966] hsearch almost always probes the same index when using the secondary hash

[fweimer at redhat dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=6966

Florian Weimer <fweimer at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
              Flags|                            |security-

-- 
You are receiving this mail because:
You are on the CC list for the bug.