Date: Mon, 3 May 2010 22:25:37 +0000 (UTC) From: "Bjoern A. Zeeb" <bz@FreeBSD.org> To: Marko Zec <zec@icir.org> Cc: Ana Kukec <anchie@freebsd.org>, Perforce Change Reviews <perforce@freebsd.org> Subject: Re: PERFORCE change 177677 for review Message-ID: <20100503222206.K23815@maildrop.int.zabbadoz.net> In-Reply-To: <201005040006.11204.zec@icir.org> References: <201005032103.o43L3nah081080@repoman.freebsd.org> <201005040006.11204.zec@icir.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 4 May 2010, Marko Zec wrote: > On Monday 03 May 2010 23:03:49 Ana Kukec wrote: >> http://p4web.freebsd.org/@@177677?ac=10 >> >> Change 177677 by anchie@anchie_malimis on 2010/05/03 21:03:44 >> >> Getting rid of the global variable V_send_so from files other then >> send.[ch]. > > Just wondering - isn't this change actually increasing the possibility for a > race between packet datapath and send.ko kldloading / kldunloading? I.e. had Yes, but it simplifies locking which is to come once the rest is correct; as you have figured, you cannot do it lock-free anyway in the module case ("pfil problem"). > Also, if I'm not mistaking, even if send_sendso_input() returns an error, it > will have the mbuf consumed / freed, so any further attempts to do anything > with the mbuf will crash the system. Do we really want to do m_freem(m) at > the bottom of send_sendso_input()? No we don't. Well not in the case there is no socket. Ana knows already but having this checkpointed in p4 helped me to see where we were;) /bz -- Bjoern A. Zeeb See you when I see you.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20100503222206.K23815>