From sacadmin Sat Jun 4 12:58:22 2005 Received: from phorcys.East.Sun.COM (phorcys.East.Sun.COM [129.148.174.143]) by sac.sfbay.sun.com (8.12.9+Sun/8.12.9) with ESMTP id j54JwMs0017150 for ; Sat, 4 Jun 2005 12:58:22 -0700 (PDT) Received: from phorcys.East.Sun.COM (localhost [127.0.0.1]) by phorcys.East.Sun.COM (8.13.4+Sun/8.13.4) with ESMTP id j54JvwZU016631; Sat, 4 Jun 2005 15:57:58 -0400 (EDT) Received: (from carlsonj@localhost) by phorcys.East.Sun.COM (8.13.4+Sun/8.13.4/Submit) id j54JvwV6016628; Sat, 4 Jun 2005 15:57:58 -0400 (EDT) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <17058.1990.523888.34901@gargle.gargle.HOWL> Date: Sat, 4 Jun 2005 15:57:58 -0400 From: James Carlson To: psarc@sac.sfbay.sun.com cc: Peter.Memishian@Sun.COM, clearview-iteam@Sun.COM Subject: 2005/358 Nemo Transmit Pointers X-Mailer: VM 7.01 under Emacs 21.3.1 Status: RO Content-Length: 3576 I'm sponsoring this fast-track request for Peter Memishian and the Clearview I-team. The timer is set to 06/11/2005. Overview ======== This case proposes changes to the Nemo mac_tx_get() interface, in order to fix several key problems. The interface is currently Project Private, but given that it was included in "Nemo - a.k.a. GLD v3" (PSARC 2004/571), we felt it appropriate to notify PSARC of the change. Patch binding is requested. Problem Details =============== The current mac_tx_get() interface is declared as: void mac_tx_get(mac_handle_t mh, mac_tx_t *txp, void **argp); ... and is defined to return the underlying transmit function and a pointer to an opaque argument to pass to the Nemo driver associated with the provided mac_handle_t. This is problematic for two reasons: * The consumers of mac_tx_get() are heavily multithreaded and make use of the information provided by mac_tx_get() in performance-critical codepaths, such as packet transmit. However, because the transmit function and its argument are set independently by mac_tx_get(), locks are currently required to ensure that they appear to be set atomically. * Since mac_tx_get() does not internally check whether the MAC is in promiscuous-mode, each caller must do this -- leading to repeated code blocks like: if (mac_promisc_get(dip->di_mh, MAC_PROMISC)) { dip->di_tx = mac_txloop; dip->di_tx_arg = dip->di_mh; } else { mac_tx_get(dip->di_mh, &dip->di_tx, &dip->di_tx_arg); } In addition, the lack of centralized handling of promiscuous-mode in mac_tx_get() prevents additional debugging logic from being added, such as asserting that there are indeed promiscuous listeners when promiscuous mode is enabled. To fix these problems, we propose to introduce a new structure that contains both the transmit function and its associated argument: typedef struct mac_txinfo_s { mac_tx_t mt_fn; void *mt_arg; } mac_txinfo_t; When the mac_handle_t is created, two of these structures will be allocated and initialized: one for normal transmit, and one for loopback transmit. These structures will remain constant over the lifetime of the mac_handle_t, and will only be destroyed when the mac_handle_t is destroyed. With the above in-place, mac_tx_get() is simply changed to return the address of the appropriate transmit information structure, depending on whether or not the mac_handle_t is in promiscuous mode: const mac_txinfo_t *mac_tx_get(mac_handle_t mh); Thus, the earlier block of code is reduced to simply: dip->di_txinfo = mac_tx_get(dip->di_mh) Since we assume throughout the kernel that assignment of pointer-sized (or smaller) values is atomic, this allows us to keep our transmit functions lockless, and also preserve the tail-call optimization where applicable -- e.g., here's the new dls_tx(): mblk_t * dls_tx(dls_channel_t dc, mblk_t *mp) { mac_txinfo_t *mtp = ((dls_impl_t *)dc)->di_txinfo; return (mtp->mt_fn(mtp->mt_arg, mp)); } Interface Changes ================= mac_txinfo_t: added: typedef struct mac_txinfo_s { mac_tx_t mt_fn; void *mt_arg; } mac_txinfo_t; mac_tx_get: changed from: void mac_tx_get(mac_handle_t mh, mac_tx_t *txp, void **argp); ... to: const mac_txinfo_t *mac_tx_get(mac_handle_t mh); Again, all changes are to Project Private interfaces. From sacadmin Tue Jun 14 16:21:13 2005 Received: from phorcys.East.Sun.COM (phorcys.East.Sun.COM [129.148.174.143]) by sac.sfbay.sun.com (8.12.9+Sun/8.12.9) with ESMTP id j5ENLDs0024796 for ; Tue, 14 Jun 2005 16:21:13 -0700 (PDT) Received: from phorcys.East.Sun.COM (localhost [127.0.0.1]) by phorcys.East.Sun.COM (8.13.4+Sun/8.13.4) with ESMTP id j5ENKhgu020886; Tue, 14 Jun 2005 19:20:43 -0400 (EDT) Received: (from carlsonj@localhost) by phorcys.East.Sun.COM (8.13.4+Sun/8.13.4/Submit) id j5ENKhZk020883; Tue, 14 Jun 2005 19:20:43 -0400 (EDT) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <17071.26187.838947.518012@gargle.gargle.HOWL> Date: Tue, 14 Jun 2005 19:20:43 -0400 From: James Carlson To: psarc@sac.sfbay.sun.com cc: Peter.Memishian@Sun.COM, clearview-iteam@Sun.COM Subject: 2005/358 Nemo Transmit Pointers X-Mailer: VM 7.01 under Emacs 21.3.1 Status: RO Content-Length: 307 This case timed out three days ago without comment. I'm marking it as "approved." -- James Carlson, KISS Network Sun Microsystems / 1 Network Drive 71.234W Vox +1 781 442 2084 MS UBUR02-212 / Burlington MA 01803-2757 42.497N Fax +1 781 442 1677 From sacadmin Tue Jun 14 19:09:37 2005 Received: from triplex.East.Sun.COM (triplex.East.Sun.COM [129.148.174.104]) by sac.sfbay.sun.com (8.12.9+Sun/8.12.9) with ESMTP id j5F29as0007961 for ; Tue, 14 Jun 2005 19:09:36 -0700 (PDT) Received: from triplex.East.Sun.COM (localhost [127.0.0.1]) by triplex.East.Sun.COM (8.13.4+Sun/8.13.4) with ESMTP id j5F2BaCZ134897; Tue, 14 Jun 2005 22:11:36 -0400 (EDT) Received: (from meem@localhost) by triplex.East.Sun.COM (8.13.4+Sun/8.13.4/Submit) id j5F2BaWm134894; Tue, 14 Jun 2005 22:11:36 -0400 (EDT) From: Peter Memishian MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <17071.36439.920934.54301@gargle.gargle.HOWL> Date: Tue, 14 Jun 2005 22:11:35 -0400 To: James Carlson Cc: psarc@sac.sfbay.sun.com, Peter.Memishian@Sun.COM, clearview-iteam@Sun.COM Subject: re: 2005/358 Nemo Transmit Pointers In-Reply-To: <17071.26187.838947.518012@gargle.gargle.HOWL> References: <17071.26187.838947.518012@gargle.gargle.HOWL> X-Mailer: VM 7.17 under 21.4 (patch 15) "Security Through Obscurity" XEmacs Lucid Status: RO Content-Length: 116 > This case timed out three days ago without comment. I'm marking it as > "approved." great, thanks! -- meem