On Sun, May 03, 2009 at 11:21:04AM +0200, [email protected] wrote:
>
> As far as I can see, when looking for the next available user id, useradd
> does:
> - take max user id + 1
> - if id > 60000 then start from 1000 and search for gaps
> - repeat
>
> So if all user ids from 1000 to 60000 are taken, useradd is in an infinite
> loop.
> Admittedly not a common case, but I guess it needs to return an error in that
> case.
I don't think there is any infinite loop.
I had some reports about being very slow with LDAP systems under some
conditions (a bug in the ldap nss was mentioned at that time).
Here is the code with the loops in useradd (in libmisc/find_new_ids.c):
setpwent ();
pw_rewind ();
while ( ((pwd = getpwent ()) != NULL)
|| ((pwd = pw_next ()) != NULL)) {
if ((pwd->pw_uid >= user_id) && (pwd->pw_uid <= uid_max)) {
user_id = pwd->pw_uid + 1;
}
}
At some time, getpwent will return NULL. Unfortunately, the above
mentioned LDAP bug caused getpwent to loop back to the beginning.
So there could be (LDAP users) * (local users) at most loops here.
Without this bug, it is only (LDAP users) + (local users).
Then, there is another loop, which starts from uid_min up to at most
uid_max.
for (user_id = uid_min; user_id < uid_max; user_id++) {
/* local, no need for xgetpwuid */
if ( (getpwuid (user_id) == NULL)
&& (pw_locate_uid (user_id) == NULL)) {
break;
}
}
So the two loops are finite.
What were the symptoms you experienced?
Best Regards,
--
Nekral
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
