Why use volatile for socket state?

pradeep singh 2500.pradeep at gmail.com
Tue May 15 01:06:04 CDT 2007


Hi All,

Why
do we need the socket state (for the nework layer representation of
the sockets) to be
 volatile?
It is implemented as volatile unsigned  char skc_state in struct sock_common
.

The tcp state change handler does not give me any clues also.

Is this simply an overkill here or to supress optimization from the GCC?

Can we just use memory barriers here?[refer: volatile considered harmful on
LKML]

Thank you
~psr

-- 
play the game
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://selenic.com/pipermail/kernel-mentors/attachments/20070515/56d2e2ae/attachment.htm 


More information about the Kernel-mentors mailing list