On Mon, Nov 29, 2010 at 03:51:02PM +0300, Evgeniy Polyakov wrote:

On Mon, Nov 29, 2010 at 12:39:09PM +0000, Jarek Poplawski (jarkao2@gmail.com) wrote:

quoted

quoted

quoted

inet sockets corresponding to passive connections are added to the bind hash
using ___inet_inherit_port(). These sockets are later removed from the bind 
hash using __inet_put_port(). These two functions are not exactly symmetrical. 
__inet_put_port() decrements hashinfo->bsockets and tb->num_owners, whereas 
___inet_inherit_port() does not increment them. This results in both of these 
going to -ve values.

This patch fixes this by calling inet_bind_hash() from ___inet_inherit_port(),
which does the right thing.

'bsockets' and 'num_owners' were introduced by commit a9d8f9110d7e953c 
(inet: Allowing more than 64k connections and heavily optimize bind(0))

Yup, things changed from that simple patch a lot.
Thanks for fixing it up.
Ack.

Probably I miss something, but since bsockets is increased by each
passive connection now, it seems it will trigger "hash table is full"
too early?

Why would it? bsockets and num_owners are supposed to be increased for each
new socket added into the table, and are used as a hint to find a bucket with
the smallest number of sockets in it.

Hash table insertion did not change, only bucket selection algorithm got
a hint.

Evgeniy & Eric,

But it's compared to the numer of available port numbers in
inet_csk_get_port():

     "if (atomic_read(&hashinfo->bsockets) > (high - low) + 1)"

Can't you have bsockets higher than this with only one port used?

Jarek P.