67662 - Passwords don't work on FreeBSD 4.1

Status: VERIFIED DUPLICATE of bug 85472

(Bugzilla :: Bugzilla-General, defect)

Description

[Nick Barnes]

The passwords system assumes that the result of MySQL's encrypt(x) function
contains the salt used for encryption as the first two characters, and so that
if c = encrypt(p), encrypt(p,left(c,2)) == c

This is not necessarily true under FreeBSD 4.1.  MySQL's encrypt() function uses
the operating system's crypt() function, and under FreeBSD 4.1 (without the
unexportable DES encryption software), crypt() uses a modular encryption
function, and stores the encryption algorithm and the salt in the result like
this: "$<algorithm number>$<salt>$<ciphertext>", where <algorithm number> == 1
for MD5.

For instance:

$ uname -sr
FreeBSD 4.1-RELEASE
$ mysql test
Welcome to the MySQL monitor.  Commands end with ; or \g.
Your MySQL connection id is 21 to server version: 3.22.32

Type 'help' for help.

mysql> select encrypt('foo');
+------------------------------+
| encrypt('foo')               |
+------------------------------+
| $1$sH$1hNbT97TWtLDu2bHGu1el0 |
+------------------------------+
1 row in set (0.00 sec)

mysql> select encrypt('foo','sH');
+------------------------------+
| encrypt('foo','sH')          |
+------------------------------+
| $1$sH$1hNbT97TWtLDu2bHGu1el0 |
+------------------------------+
1 row in set (0.01 sec)

Possible fixes include always specifying the salt when calling encrypt.  For
instance, one could use a hash of the user_id.  I am using the simplest
workaround, which is to always use the salt '$1'.  This is the first two
characters of the result of encrypt(), so the assumptions in the code continue
to work.

Comment 1

[Dave Miller [:justdave]]

ok, so how about if it's "smart" and if the cryptpasswd starts with a $ it uses 
the text between the 2nd and 3rd $ as the salt, otherwise use the first two 
chars?

Comment 2

[Dave Miller [:justdave]]

This is already fixed, btw


*** This bug has been marked as a duplicate of 85472 ***

Comment 3

[Matthew Tuck [:CodeMachine]]

V.

Comment 4

[Dave Miller [:justdave]]

clearing milestone on closed non-FIXED bugs in case they ever get reopened.

Comment 5

[Dave Miller [:justdave]]

moving to Bugzilla product
reassign to default owner/qa for INVALID/WONTFIX/WORKSFORME/DUPLICATE