Re: [PATCH] removed socket buffer in unix domain socket

Go Taniguchi (go@turbolinux.co.jp)
Fri, 11 Jan 2002 22:23:56 +0900


This is a multi-part message in MIME format.
--------------010005040202000502080300
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

What size of actually used hash table --unix_socket_table--?
If it is 256, probably forall_unix_sockets is dangerous.

forall_unix_sockets use 257 table size.
And If I apply this fix, test program can work.

Alan Cox wrote:

>> */
>>- if(UNIXCB(skb).fp)
>>+ if(s->dead && UNIXCB(skb).fp)
>> {
>>
>
> The bug may be real but the fix would prevent garbage collection working
> at all - which I grant would fix the problem.
>
> You don't need a socket to be dead to want to garbage collect it. If a
> socket is getting disposed of while in use then there is a
> maybe_unmark_and.. call missing, or a lock on the unix socket table missing
> somewhere.

-- GO!

--------------010005040202000502080300
Content-Type: text/plain;
name="af_net.h.patch"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline;
filename="af_net.h.patch"

--- linux/include/net/af_unix.h.orig Tue Apr 25 05:43:04 2000
+++ linux/include/net/af_unix.h Fri Jan 11 21:49:57 2002
@@ -14,7 +14,7 @@
extern atomic_t unix_tot_inflight;


-#define forall_unix_sockets(i, s) for (i=0; i<=UNIX_HASH_SIZE; i++) \
+#define forall_unix_sockets(i, s) for (i=0; i<UNIX_HASH_SIZE; i++) \
for (s=unix_socket_table[i]; s; s=s->next)

struct unix_address

--------------010005040202000502080300--

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/