[PATCH 1/4] Set socket flags on restore using sock_setsockopt() where possible

Dan Smith danms at us.ibm.com
Thu Aug 13 12:55:46 PDT 2009

OL> When you add support to new socket due to connect() that were not
OL> yet accept()ed from the listening socket - there will be a case of
OL> a sock without sock->sk_socket.

OL> This probably means that we want the test for sock->sk_socket to
OL> remain as is (and a similar one in sock_rst_flags above)

Okay, but we've got a lot of other places above that where we assume
sock->sk_socket is valid, so the thinking here was that when we have a
real reason to handle that case, I'd fix all the other cases as well.

However, I shouldn't have done that in the middle of this patch anyway.

OL> Super nit: perhaps s/rst/restore/ ?  Besides agreeing with current
OL> practice, it may dodge a related rant from Linux :o

I'll re-send with the rename and without the sock->sk_socket and then
follow up with a fix for the other cases in a later patch.

Dan Smith
IBM Linux Technology Center
email: danms at us.ibm.com

More information about the Containers mailing list