Date: Mon, 5 Sep 2005 09:43:41 -0700 From: Luigi Rizzo <rizzo@icir.org> To: Gleb Smirnoff <glebius@FreeBSD.org> Cc: cvs-src@FreeBSD.org, src-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/sys/kern kern_poll.c Message-ID: <20050905094341.A23343@xorpc.icir.org> In-Reply-To: <200509051602.j85G2Bpo090258@repoman.freebsd.org>; from glebius@FreeBSD.org on Mon, Sep 05, 2005 at 04:02:11PM %2B0000 References: <200509051602.j85G2Bpo090258@repoman.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
gleb, could you document how you are using the PRF_RUNNING and PRF_LEAVING flags that you added ? also i am unclear on whether there is a race condition involving unregistering poll (second part of netisr_poll()) while running the section of ether_poll() where it drops the mutex before calling the handler. ether_poll signals its activity with PRF_RUNNING, but netisr_poll() does not check it, so it might happen that 1. on one CPU, ether_poll() drops the lock before calling pr[i].handler(), perhaps even getting descheduled 2. on another CPU, netisr_poll() gets the lock and goes all the way to the end, setting pr[i].handler = NULL 3. on the first CPU, without further checks, the code tries to dereference the field. Also, what is the overall design - do you want to support multiple polling loops running concurrently (netisr_poll, one poll_idle per CPU, and possibly the poll_in_trap) ? cheers luigi
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050905094341.A23343>