Hey guys, does anyone know off the top of their heads why named pipes don't appear to work across null_fs mounted partitions? i.e. if I have a named pipe in a file system,

# ls -ld /mysql/mysql.sock srwxrwxrwx 1 mysql wheel 0 Feb 3 19:01 /mysql/mysql.sock

# mysql --socket=/mysql/mysql.sock Welcome to the MySQL monitor. Commands end with ; or \g. Your MySQL connection id is 6 Server version: 5.0.33-log FreeBSD port: mysql-server-5.0.33

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

mysql>

However if I make this available elsewhere via a null_fs mount:

# mkdir /foo # mount_nullfs /mysql /foo # ls -ld /foo/mysql.sock srwxrwxrwx 1 mysql wheel 0 Feb 3 19:01 /foo/mysql.sock

# mysql --socket=/foo/mysql.sock -p Enter password: ERROR 2002 (HY000): Can't connect to local MySQL server through socket
'/foo/mysql.sock' (61)

the socket stops working.

However a hardlink to the socket works:

# umount /foo # ln /mysql/mysql.sock /foo/mysql.sock

# mysql --socket=/foo/mysql.sock Welcome to the MySQL monitor. Commands end with ; or \g. Your MySQL connection id is 10 Server version: 5.0.33-log FreeBSD port: mysql-server-5.0.33

Type 'help;' or '\h' for help. Type '\c' to clear the buffer.

Is this a fundamental design issue with null_fs or a bug?

There appears to be a lot of confusion on the lists about this point as many people are trying to do this so as to make a single mysql server available from within a number of jails, for instance. However people appear to think that this is a limitation of the jail code, not a limitation of the null_fs code. Having named pipes work in null_fs filesystems would be a very handy thing indeed.

I'd appreciate any insights into this.

Many thanks, Joe