Hello Jamal,
jamal wrote:
> On Mon, 2005-04-04 at 18:25, David S. Miller wrote:
>
>
>>If you only take write_lock() from process context, only the write_lock()'s
>>need BH disabling. read_lock() takers can then nest arbitrarily, BH or not.
>
>
> Ok, never mind - Ive made the change.
> As soon as Masahide tests i will post the final patch.
I've tested normal cases below with the latest patch and it works fine.
I think you can go ahead.
tested cases:
o netlink (using iproute2 "ip xfrm monitor" to confirm it)
- add/del/flush/expire for SA/SP
- acquire,allocspi,update for SA
- update for SP
o pfkey
- running racoon
o both sockets
- running racoon with using "ip xfrm monitor".
Regards,
--
Masahide NAKAMURA
|