From owner-svn-src-all@freebsd.org Fri Jul 10 18:18:23 2015 Return-Path: Delivered-To: svn-src-all@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id B43A43E8F; Fri, 10 Jul 2015 18:18:23 +0000 (UTC) (envelope-from pkelsey@gmail.com) Received: from mail-yk0-x233.google.com (mail-yk0-x233.google.com [IPv6:2607:f8b0:4002:c07::233]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (Client CN "smtp.gmail.com", Issuer "Google Internet Authority G2" (verified OK)) by mx1.freebsd.org (Postfix) with ESMTPS id 5DE1BFA8; Fri, 10 Jul 2015 18:18:23 +0000 (UTC) (envelope-from pkelsey@gmail.com) Received: by ykeo3 with SMTP id o3so149149150yke.0; Fri, 10 Jul 2015 11:18:22 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20120113; h=mime-version:sender:in-reply-to:references:date:message-id:subject :from:to:cc:content-type; bh=sKsClogOMocmSskQFmrvH+K4vSnaduUH0q3Y05vDAVE=; b=BHec2xZNxX3M7wwlvCDNxpibzXARzcSXfAGJ1zFjoUS4oGB8yfBcAe/mATsklNbyh7 TKsFoEYA9Rwvtbhbaqg9T4nU1Mrp0pLYftJvwV/GgDuJFT/E1mx4yP15W3hAOZeg7F2d Gyk7I7KyXnBivuj1BtWqQSbyzdfMquwN8mZq10trY9N+IwCHoSVpWULYI4IVGCRiu+vF p+P56etk5/YlsfOfoftch6hCyfREVLjf2y6QZ7wAJl6j16ANDyWfV2S5lfj8ecnXgx6M TYFbDRN4+xdoY3Z2zxWzaKvQprZVzDtUsI98FwnS1ew6/9MhuIxBA6sDdBFjf9m0mSza iRBw== MIME-Version: 1.0 X-Received: by 10.129.36.14 with SMTP id k14mr24358657ywk.64.1436552302421; Fri, 10 Jul 2015 11:18:22 -0700 (PDT) Sender: pkelsey@gmail.com Received: by 10.129.138.68 with HTTP; Fri, 10 Jul 2015 11:18:22 -0700 (PDT) In-Reply-To: <201507100551.t6A5paZH050451@repo.freebsd.org> References: <201507100551.t6A5paZH050451@repo.freebsd.org> Date: Fri, 10 Jul 2015 14:18:22 -0400 X-Google-Sender-Auth: VHree1V7k-ZwFoPn3Nsqrpa9nys Message-ID: Subject: Re: svn commit: r285349 - in head/sys: dev/cxgbe dev/e1000 dev/ixgbe dev/netmap dev/re net From: Patrick Kelsey To: Luigi Rizzo Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Content-Type: text/plain; charset=UTF-8 X-Content-Filtered-By: Mailman/MimeDel 2.1.20 X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 10 Jul 2015 18:18:23 -0000 Thanks for the update, Luigi! For the record, the exclusive open was Juli Mallett's idea for the solutions you mention - I was just the first +1. -Patrick On Fri, Jul 10, 2015 at 1:51 AM, Luigi Rizzo wrote: > Author: luigi > Date: Fri Jul 10 05:51:36 2015 > New Revision: 285349 > URL: https://svnweb.freebsd.org/changeset/base/285349 > > Log: > Sync netmap sources with the version in our private tree. > This commit contains large contributions from Giuseppe Lettieri and > Stefano Garzarella, is partly supported by grants from Verisign and > Cisco, > and brings in the following: > > - fix zerocopy monitor ports and introduce copying monitor ports > (the latter are lower performance but give access to all traffic > in parallel with the application) > > - exclusive open mode, useful to implement solutions that recover > from crashes of the main netmap client (suggested by Patrick Kelsey) > > - revised memory allocator in preparation for the 'passthrough mode' > (ptnetmap) recently presented at bsdcan. ptnetmap is described in > S. Garzarella, G. Lettieri, L. Rizzo; > Virtual device passthrough for high speed VM networking, > ACM/IEEE ANCS 2015, Oakland (CA) May 2015 > http://info.iet.unipi.it/~luigi/research.html > > - fix rx CRC handing on ixl > > - add module dependencies for netmap when building drivers as modules > > - minor simplifications to device-specific routines (*txsync, *rxsync) > > - general code cleanup (remove unused variables, introduce macros > to access rings and remove duplicate code, > > Applications do not need to be recompiled, unless of course > they want to use the new features (monitors and exclusive open). > > Those willing to try this code on stable/10 can just update the > sys/dev/netmap/*, sys/net/netmap* with the version in HEAD > and apply the small patches to individual device drivers. > > MFC after: 1 month > Sponsored by: (partly) Verisign, Cisco > > Modified: > head/sys/dev/cxgbe/t4_main.c > head/sys/dev/cxgbe/t4_netmap.c > head/sys/dev/e1000/if_em.c > head/sys/dev/e1000/if_igb.c > head/sys/dev/e1000/if_lem.c > head/sys/dev/ixgbe/if_ix.c > head/sys/dev/netmap/if_em_netmap.h > head/sys/dev/netmap/if_igb_netmap.h > head/sys/dev/netmap/if_ixl_netmap.h > head/sys/dev/netmap/if_lem_netmap.h > head/sys/dev/netmap/if_re_netmap.h > head/sys/dev/netmap/if_vtnet_netmap.h > head/sys/dev/netmap/ixgbe_netmap.h > head/sys/dev/netmap/netmap.c > head/sys/dev/netmap/netmap_freebsd.c > head/sys/dev/netmap/netmap_generic.c > head/sys/dev/netmap/netmap_kern.h > head/sys/dev/netmap/netmap_mem2.c > head/sys/dev/netmap/netmap_mem2.h > head/sys/dev/netmap/netmap_monitor.c > head/sys/dev/netmap/netmap_pipe.c > head/sys/dev/netmap/netmap_vale.c > head/sys/dev/re/if_re.c > head/sys/net/netmap.h > head/sys/net/netmap_user.h > > Modified: head/sys/dev/cxgbe/t4_main.c > > ============================================================================== > --- head/sys/dev/cxgbe/t4_main.c Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/cxgbe/t4_main.c Fri Jul 10 05:51:36 2015 > (r285349) > @@ -8533,10 +8533,17 @@ static devclass_t cxgbe_devclass, cxl_de > DRIVER_MODULE(t4nex, pci, t4_driver, t4_devclass, mod_event, 0); > MODULE_VERSION(t4nex, 1); > MODULE_DEPEND(t4nex, firmware, 1, 1, 1); > +#ifdef DEV_NETMAP > +MODULE_DEPEND(t4nex, netmap, 1, 1, 1); > +#endif /* DEV_NETMAP */ > + > > DRIVER_MODULE(t5nex, pci, t5_driver, t5_devclass, mod_event, 0); > MODULE_VERSION(t5nex, 1); > MODULE_DEPEND(t5nex, firmware, 1, 1, 1); > +#ifdef DEV_NETMAP > +MODULE_DEPEND(t5nex, netmap, 1, 1, 1); > +#endif /* DEV_NETMAP */ > > DRIVER_MODULE(cxgbe, t4nex, cxgbe_driver, cxgbe_devclass, 0, 0); > MODULE_VERSION(cxgbe, 1); > > Modified: head/sys/dev/cxgbe/t4_netmap.c > > ============================================================================== > --- head/sys/dev/cxgbe/t4_netmap.c Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/cxgbe/t4_netmap.c Fri Jul 10 05:51:36 2015 > (r285349) > @@ -917,8 +917,6 @@ cxgbe_netmap_txsync(struct netmap_kring > kring->nr_hwtail -= kring->nkr_num_slots; > } > > - nm_txsync_finalize(kring); > - > return (0); > } > > @@ -931,7 +929,7 @@ cxgbe_netmap_rxsync(struct netmap_kring > struct port_info *pi = ifp->if_softc; > struct adapter *sc = pi->adapter; > struct sge_nm_rxq *nm_rxq = &sc->sge.nm_rxq[pi->first_nm_rxq + > kring->ring_id]; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > u_int n; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > @@ -993,8 +991,6 @@ cxgbe_netmap_rxsync(struct netmap_kring > } > } > > - nm_rxsync_finalize(kring); > - > return (0); > } > > > Modified: head/sys/dev/e1000/if_em.c > > ============================================================================== > --- head/sys/dev/e1000/if_em.c Fri Jul 10 05:07:18 2015 (r285348) > +++ head/sys/dev/e1000/if_em.c Fri Jul 10 05:51:36 2015 (r285349) > @@ -344,6 +344,9 @@ devclass_t em_devclass; > DRIVER_MODULE(em, pci, em_driver, em_devclass, 0, 0); > MODULE_DEPEND(em, pci, 1, 1, 1); > MODULE_DEPEND(em, ether, 1, 1, 1); > +#ifdef DEV_NETMAP > +MODULE_DEPEND(em, netmap, 1, 1, 1); > +#endif /* DEV_NETMAP */ > > /********************************************************************* > * Tunable default values. > > Modified: head/sys/dev/e1000/if_igb.c > > ============================================================================== > --- head/sys/dev/e1000/if_igb.c Fri Jul 10 05:07:18 2015 (r285348) > +++ head/sys/dev/e1000/if_igb.c Fri Jul 10 05:51:36 2015 (r285349) > @@ -322,6 +322,9 @@ static devclass_t igb_devclass; > DRIVER_MODULE(igb, pci, igb_driver, igb_devclass, 0, 0); > MODULE_DEPEND(igb, pci, 1, 1, 1); > MODULE_DEPEND(igb, ether, 1, 1, 1); > +#ifdef DEV_NETMAP > +MODULE_DEPEND(igb, netmap, 1, 1, 1); > +#endif /* DEV_NETMAP */ > > /********************************************************************* > * Tunable default values. > > Modified: head/sys/dev/e1000/if_lem.c > > ============================================================================== > --- head/sys/dev/e1000/if_lem.c Fri Jul 10 05:07:18 2015 (r285348) > +++ head/sys/dev/e1000/if_lem.c Fri Jul 10 05:51:36 2015 (r285349) > @@ -286,6 +286,9 @@ extern devclass_t em_devclass; > DRIVER_MODULE(lem, pci, lem_driver, em_devclass, 0, 0); > MODULE_DEPEND(lem, pci, 1, 1, 1); > MODULE_DEPEND(lem, ether, 1, 1, 1); > +#ifdef DEV_NETMAP > +MODULE_DEPEND(lem, netmap, 1, 1, 1); > +#endif /* DEV_NETMAP */ > > /********************************************************************* > * Tunable default values. > > Modified: head/sys/dev/ixgbe/if_ix.c > > ============================================================================== > --- head/sys/dev/ixgbe/if_ix.c Fri Jul 10 05:07:18 2015 (r285348) > +++ head/sys/dev/ixgbe/if_ix.c Fri Jul 10 05:51:36 2015 (r285349) > @@ -246,6 +246,9 @@ DRIVER_MODULE(ix, pci, ix_driver, ix_dev > > MODULE_DEPEND(ix, pci, 1, 1, 1); > MODULE_DEPEND(ix, ether, 1, 1, 1); > +#ifdef DEV_NETMAP > +MODULE_DEPEND(ix, netmap, 1, 1, 1); > +#endif /* DEV_NETMAP */ > > /* > ** TUNEABLE PARAMETERS: > > Modified: head/sys/dev/netmap/if_em_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/if_em_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/if_em_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -198,8 +198,6 @@ em_netmap_txsync(struct netmap_kring *kr > } > } > > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -217,7 +215,7 @@ em_netmap_rxsync(struct netmap_kring *kr > u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -303,9 +301,6 @@ em_netmap_rxsync(struct netmap_kring *kr > E1000_WRITE_REG(&adapter->hw, E1000_RDT(rxr->me), nic_i); > } > > - /* tell userspace that there might be new packets */ > - nm_rxsync_finalize(kring); > - > return 0; > > ring_reset: > > Modified: head/sys/dev/netmap/if_igb_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/if_igb_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/if_igb_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -180,8 +180,6 @@ igb_netmap_txsync(struct netmap_kring *k > kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), > lim); > } > > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -199,7 +197,7 @@ igb_netmap_rxsync(struct netmap_kring *k > u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -283,9 +281,6 @@ igb_netmap_rxsync(struct netmap_kring *k > E1000_WRITE_REG(&adapter->hw, E1000_RDT(rxr->me), nic_i); > } > > - /* tell userspace that there might be new packets */ > - nm_rxsync_finalize(kring); > - > return 0; > > ring_reset: > > Modified: head/sys/dev/netmap/if_ixl_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/if_ixl_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/if_ixl_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -68,9 +68,14 @@ extern int ixl_rx_miss, ixl_rx_miss_bufs > * count packets that might be missed due to lost interrupts. > */ > SYSCTL_DECL(_dev_netmap); > -int ixl_rx_miss, ixl_rx_miss_bufs, ixl_crcstrip; > +/* > + * The xl driver by default strips CRCs and we do not override it. > + */ > +int ixl_rx_miss, ixl_rx_miss_bufs, ixl_crcstrip = 1; > +#if 0 > SYSCTL_INT(_dev_netmap, OID_AUTO, ixl_crcstrip, > - CTLFLAG_RW, &ixl_crcstrip, 0, "strip CRC on rx frames"); > + CTLFLAG_RW, &ixl_crcstrip, 1, "strip CRC on rx frames"); > +#endif > SYSCTL_INT(_dev_netmap, OID_AUTO, ixl_rx_miss, > CTLFLAG_RW, &ixl_rx_miss, 0, "potentially missed rx intr"); > SYSCTL_INT(_dev_netmap, OID_AUTO, ixl_rx_miss_bufs, > @@ -268,8 +273,6 @@ ixl_netmap_txsync(struct netmap_kring *k > kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), > lim); > } > > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -297,7 +300,7 @@ ixl_netmap_rxsync(struct netmap_kring *k > u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -408,9 +411,6 @@ ixl_netmap_rxsync(struct netmap_kring *k > wr32(vsi->hw, rxr->tail, nic_i); > } > > - /* tell userspace that there might be new packets */ > - nm_rxsync_finalize(kring); > - > return 0; > > ring_reset: > > Modified: head/sys/dev/netmap/if_lem_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/if_lem_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/if_lem_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -302,8 +302,6 @@ lem_netmap_txsync(struct netmap_kring *k > kring->nr_hwtail = nm_prev(netmap_idx_n2k(kring, nic_i), > lim); > } > > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -321,7 +319,7 @@ lem_netmap_rxsync(struct netmap_kring *k > u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -466,9 +464,6 @@ lem_netmap_rxsync(struct netmap_kring *k > E1000_WRITE_REG(&adapter->hw, E1000_RDT(0), nic_i); > } > > - /* tell userspace that there might be new packets */ > - nm_rxsync_finalize(kring); > - > return 0; > > ring_reset: > > Modified: head/sys/dev/netmap/if_re_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/if_re_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/if_re_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -159,8 +159,6 @@ re_netmap_txsync(struct netmap_kring *kr > } > } > > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -178,7 +176,7 @@ re_netmap_rxsync(struct netmap_kring *kr > u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -273,9 +271,6 @@ re_netmap_rxsync(struct netmap_kring *kr > BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); > } > > - /* tell userspace that there might be new packets */ > - nm_rxsync_finalize(kring); > - > return 0; > > ring_reset: > > Modified: head/sys/dev/netmap/if_vtnet_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/if_vtnet_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/if_vtnet_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -214,9 +214,6 @@ vtnet_netmap_txsync(struct netmap_kring > virtqueue_postpone_intr(vq, VQ_POSTPONE_SHORT); > } > > -//out: > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -278,7 +275,7 @@ vtnet_netmap_rxsync(struct netmap_kring > // u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -340,9 +337,6 @@ vtnet_netmap_rxsync(struct netmap_kring > vtnet_rxq_enable_intr(rxq); > } > > - /* tell userspace that there might be new packets. */ > - nm_rxsync_finalize(kring); > - > ND("[C] h %d c %d t %d hwcur %d hwtail %d", > ring->head, ring->cur, ring->tail, > kring->nr_hwcur, kring->nr_hwtail); > > Modified: head/sys/dev/netmap/ixgbe_netmap.h > > ============================================================================== > --- head/sys/dev/netmap/ixgbe_netmap.h Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/ixgbe_netmap.h Fri Jul 10 05:51:36 2015 > (r285349) > @@ -322,8 +322,6 @@ ixgbe_netmap_txsync(struct netmap_kring > } > } > > - nm_txsync_finalize(kring); > - > return 0; > } > > @@ -351,7 +349,7 @@ ixgbe_netmap_rxsync(struct netmap_kring > u_int nic_i; /* index into the NIC ring */ > u_int n; > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = nm_rxsync_prologue(kring); > + u_int const head = kring->rhead; > int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & > NKR_PENDINTR; > > /* device-specific */ > @@ -458,9 +456,6 @@ ixgbe_netmap_rxsync(struct netmap_kring > IXGBE_WRITE_REG(&adapter->hw, IXGBE_RDT(rxr->me), nic_i); > } > > - /* tell userspace that there might be new packets */ > - nm_rxsync_finalize(kring); > - > return 0; > > ring_reset: > > Modified: head/sys/dev/netmap/netmap.c > > ============================================================================== > --- head/sys/dev/netmap/netmap.c Fri Jul 10 05:07:18 2015 > (r285348) > +++ head/sys/dev/netmap/netmap.c Fri Jul 10 05:51:36 2015 > (r285349) > @@ -293,7 +293,7 @@ ports attached to the switch) > * kring->nm_sync() == DEVICE_netmap_rxsync() > * 2) device interrupt handler > * na->nm_notify() == netmap_notify() > - * - tx from host stack > + * - rx from host stack > * concurrently: > * 1) host stack > * netmap_transmit() > @@ -313,31 +313,113 @@ ports attached to the switch) > * > * -= SYSTEM DEVICE WITH GENERIC SUPPORT =- > * > + * na == NA(ifp) == generic_netmap_adapter created in > generic_netmap_attach() > * > - * > - * -= VALE PORT =- > - * > - * > - * > - * -= NETMAP PIPE =- > - * > - * > - * > - * -= SYSTEM DEVICE WITH NATIVE SUPPORT, CONNECTED TO VALE, NO HOST > RINGS =- > - * > - * > - * > - * -= SYSTEM DEVICE WITH NATIVE SUPPORT, CONNECTED TO VALE, WITH HOST > RINGS =- > - * > - * > - * > - * -= SYSTEM DEVICE WITH GENERIC SUPPORT, CONNECTED TO VALE, NO HOST > RINGS =- > - * > + * - tx from netmap userspace: > + * concurrently: > + * 1) ioctl(NIOCTXSYNC)/netmap_poll() in process context > + * kring->nm_sync() == generic_netmap_txsync() > + * linux: dev_queue_xmit() with NM_MAGIC_PRIORITY_TX > + * generic_ndo_start_xmit() > + * orig. dev. start_xmit > + * FreeBSD: na->if_transmit() == orig. dev if_transmit > + * 2) generic_mbuf_destructor() > + * na->nm_notify() == netmap_notify() > + * - rx from netmap userspace: > + * 1) ioctl(NIOCRXSYNC)/netmap_poll() in process context > + * kring->nm_sync() == generic_netmap_rxsync() > + * mbq_safe_dequeue() > + * 2) device driver > + * generic_rx_handler() > + * mbq_safe_enqueue() > + * na->nm_notify() == netmap_notify() > + * - rx from host stack: > + * concurrently: > + * 1) host stack > + * linux: generic_ndo_start_xmit() > + * netmap_transmit() > + * FreeBSD: ifp->if_input() == netmap_transmit > + * both: > + * na->nm_notify() == netmap_notify() > + * 2) ioctl(NIOCRXSYNC)/netmap_poll() in process context > + * kring->nm_sync() == netmap_rxsync_from_host_compat > + * netmap_rxsync_from_host(na, NULL, NULL) > + * - tx to host stack: > + * ioctl(NIOCTXSYNC)/netmap_poll() in process context > + * kring->nm_sync() == netmap_txsync_to_host_compat > + * netmap_txsync_to_host(na) > + * NM_SEND_UP() > + * FreeBSD: na->if_input() == ??? XXX > + * linux: netif_rx() with NM_MAGIC_PRIORITY_RX > * > * > - * -= SYSTEM DEVICE WITH GENERIC SUPPORT, CONNECTED TO VALE, WITH HOST > RINGS =- > + * -= VALE =- > * > + * INCOMING: > * > + * - VALE ports: > + * ioctl(NIOCTXSYNC)/netmap_poll() in process context > + * kring->nm_sync() == netmap_vp_txsync() > + * > + * - system device with native support: > + * from cable: > + * interrupt > + * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr != > host ring) > + * kring->nm_sync() == DEVICE_netmap_rxsync() > + * netmap_vp_txsync() > + * kring->nm_sync() == DEVICE_netmap_rxsync() > + * from host stack: > + * netmap_transmit() > + * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr == > host ring) > + * kring->nm_sync() == > netmap_rxsync_from_host_compat() > + * netmap_vp_txsync() > + * > + * - system device with generic support: > + * from device driver: > + * generic_rx_handler() > + * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr != > host ring) > + * kring->nm_sync() == generic_netmap_rxsync() > + * netmap_vp_txsync() > + * kring->nm_sync() == generic_netmap_rxsync() > + * from host stack: > + * netmap_transmit() > + * na->nm_notify() == netmap_bwrap_intr_notify(ring_nr == > host ring) > + * kring->nm_sync() == > netmap_rxsync_from_host_compat() > + * netmap_vp_txsync() > + * > + * (all cases) --> nm_bdg_flush() > + * dest_na->nm_notify() == (see below) > + * > + * OUTGOING: > + * > + * - VALE ports: > + * concurrently: > + * 1) ioctlNIOCRXSYNC)/netmap_poll() in process context > + * kring->nm_sync() == netmap_vp_rxsync() > + * 2) from nm_bdg_flush() > + * na->nm_notify() == netmap_notify() > + * > + * - system device with native support: > + * to cable: > + * na->nm_notify() == netmap_bwrap_notify() > + * netmap_vp_rxsync() > + * kring->nm_sync() == DEVICE_netmap_txsync() > + * netmap_vp_rxsync() > + * to host stack: > + * netmap_vp_rxsync() > + * kring->nm_sync() == netmap_txsync_to_host_compat > + * netmap_vp_rxsync_locked() > + * > + * - system device with generic adapter: > + * to device driver: > + * na->nm_notify() == netmap_bwrap_notify() > + * netmap_vp_rxsync() > + * kring->nm_sync() == generic_netmap_txsync() > + * netmap_vp_rxsync() > + * to host stack: > + * netmap_vp_rxsync() > + * kring->nm_sync() == netmap_txsync_to_host_compat > + * netmap_vp_rxsync() > * > */ > > @@ -412,15 +494,6 @@ ports attached to the switch) > > MALLOC_DEFINE(M_NETMAP, "netmap", "Network memory map"); > > -/* > - * The following variables are used by the drivers and replicate > - * fields in the global memory pool. They only refer to buffers > - * used by physical interfaces. > - */ > -u_int netmap_total_buffers; > -u_int netmap_buf_size; > -char *netmap_buffer_base; /* also address of an invalid buffer */ > - > /* user-controlled variables */ > int netmap_verbose; > > @@ -446,7 +519,6 @@ SYSCTL_INT(_dev_netmap, OID_AUTO, adapti > > int netmap_flags = 0; /* debug flags */ > int netmap_fwd = 0; /* force transparent mode */ > -int netmap_mmap_unreg = 0; /* allow mmap of unregistered fds */ > > /* > * netmap_admode selects the netmap mode to use. > @@ -464,7 +536,6 @@ int netmap_generic_rings = 1; /* numbe > > SYSCTL_INT(_dev_netmap, OID_AUTO, flags, CTLFLAG_RW, &netmap_flags, 0 , > ""); > SYSCTL_INT(_dev_netmap, OID_AUTO, fwd, CTLFLAG_RW, &netmap_fwd, 0 , ""); > -SYSCTL_INT(_dev_netmap, OID_AUTO, mmap_unreg, CTLFLAG_RW, > &netmap_mmap_unreg, 0, ""); > SYSCTL_INT(_dev_netmap, OID_AUTO, admode, CTLFLAG_RW, &netmap_admode, 0 , > ""); > SYSCTL_INT(_dev_netmap, OID_AUTO, generic_mit, CTLFLAG_RW, > &netmap_generic_mit, 0 , ""); > SYSCTL_INT(_dev_netmap, OID_AUTO, generic_ringsize, CTLFLAG_RW, > &netmap_generic_ringsize, 0 , ""); > @@ -472,15 +543,6 @@ SYSCTL_INT(_dev_netmap, OID_AUTO, generi > > NMG_LOCK_T netmap_global_lock; > > - > -static void > -nm_kr_get(struct netmap_kring *kr) > -{ > - while (NM_ATOMIC_TEST_AND_SET(&kr->nr_busy)) > - tsleep(kr, 0, "NM_KR_GET", 4); > -} > - > - > /* > * mark the ring as stopped, and run through the locks > * to make sure other users get to see it. > @@ -495,34 +557,14 @@ netmap_disable_ring(struct netmap_kring > nm_kr_put(kr); > } > > -/* stop or enable a single tx ring */ > -void > -netmap_set_txring(struct netmap_adapter *na, u_int ring_id, int stopped) > -{ > - if (stopped) > - netmap_disable_ring(na->tx_rings + ring_id); > - else > - na->tx_rings[ring_id].nkr_stopped = 0; > - /* nofify that the stopped state has changed. This is currently > - *only used by bwrap to propagate the state to its own krings. > - * (see netmap_bwrap_intr_notify). > - */ > - na->nm_notify(na, ring_id, NR_TX, NAF_DISABLE_NOTIFY); > -} > - > -/* stop or enable a single rx ring */ > +/* stop or enable a single ring */ > void > -netmap_set_rxring(struct netmap_adapter *na, u_int ring_id, int stopped) > +netmap_set_ring(struct netmap_adapter *na, u_int ring_id, enum txrx t, > int stopped) > { > if (stopped) > - netmap_disable_ring(na->rx_rings + ring_id); > + netmap_disable_ring(NMR(na, t) + ring_id); > else > - na->rx_rings[ring_id].nkr_stopped = 0; > - /* nofify that the stopped state has changed. This is currently > - *only used by bwrap to propagate the state to its own krings. > - * (see netmap_bwrap_intr_notify). > - */ > - na->nm_notify(na, ring_id, NR_RX, NAF_DISABLE_NOTIFY); > + NMR(na, t)[ring_id].nkr_stopped = 0; > } > > > @@ -531,20 +573,15 @@ void > netmap_set_all_rings(struct netmap_adapter *na, int stopped) > { > int i; > - u_int ntx, nrx; > + enum txrx t; > > if (!nm_netmap_on(na)) > return; > > - ntx = netmap_real_tx_rings(na); > - nrx = netmap_real_rx_rings(na); > - > - for (i = 0; i < ntx; i++) { > - netmap_set_txring(na, i, stopped); > - } > - > - for (i = 0; i < nrx; i++) { > - netmap_set_rxring(na, i, stopped); > + for_rx_tx(t) { > + for (i = 0; i < netmap_real_rings(na, t); i++) { > + netmap_set_ring(na, i, t, stopped); > + } > } > } > > @@ -657,7 +694,8 @@ netmap_update_config(struct netmap_adapt > > txr = txd = rxr = rxd = 0; > if (na->nm_config == NULL || > - na->nm_config(na, &txr, &txd, &rxr, &rxd)) { > + na->nm_config(na, &txr, &txd, &rxr, &rxd)) > + { > /* take whatever we had at init time */ > txr = na->num_tx_rings; > txd = na->num_tx_desc; > @@ -738,73 +776,59 @@ netmap_krings_create(struct netmap_adapt > { > u_int i, len, ndesc; > struct netmap_kring *kring; > - u_int ntx, nrx; > + u_int n[NR_TXRX]; > + enum txrx t; > > /* account for the (possibly fake) host rings */ > - ntx = na->num_tx_rings + 1; > - nrx = na->num_rx_rings + 1; > + n[NR_TX] = na->num_tx_rings + 1; > + n[NR_RX] = na->num_rx_rings + 1; > > - len = (ntx + nrx) * sizeof(struct netmap_kring) + tailroom; > + len = (n[NR_TX] + n[NR_RX]) * sizeof(struct netmap_kring) + > tailroom; > > na->tx_rings = malloc((size_t)len, M_DEVBUF, M_NOWAIT | M_ZERO); > if (na->tx_rings == NULL) { > D("Cannot allocate krings"); > return ENOMEM; > } > - na->rx_rings = na->tx_rings + ntx; > + na->rx_rings = na->tx_rings + n[NR_TX]; > > /* > * All fields in krings are 0 except the one initialized below. > * but better be explicit on important kring fields. > */ > - ndesc = na->num_tx_desc; > - for (i = 0; i < ntx; i++) { /* Transmit rings */ > - kring = &na->tx_rings[i]; > - bzero(kring, sizeof(*kring)); > - kring->na = na; > - kring->ring_id = i; > - kring->nkr_num_slots = ndesc; > - if (i < na->num_tx_rings) { > - kring->nm_sync = na->nm_txsync; > - } else if (i == na->num_tx_rings) { > - kring->nm_sync = netmap_txsync_to_host_compat; > + for_rx_tx(t) { > + ndesc = nma_get_ndesc(na, t); > + for (i = 0; i < n[t]; i++) { > + kring = &NMR(na, t)[i]; > + bzero(kring, sizeof(*kring)); > + kring->na = na; > + kring->ring_id = i; > + kring->tx = t; > + kring->nkr_num_slots = ndesc; > + if (i < nma_get_nrings(na, t)) { > + kring->nm_sync = (t == NR_TX ? > na->nm_txsync : na->nm_rxsync); > + } else if (i == na->num_tx_rings) { > + kring->nm_sync = (t == NR_TX ? > + > netmap_txsync_to_host_compat : > + > netmap_rxsync_from_host_compat); > + } > + kring->nm_notify = na->nm_notify; > + kring->rhead = kring->rcur = kring->nr_hwcur = 0; > + /* > + * IMPORTANT: Always keep one slot empty. > + */ > + kring->rtail = kring->nr_hwtail = (t == NR_TX ? > ndesc - 1 : 0); > + snprintf(kring->name, sizeof(kring->name) - 1, "%s > %s%d", na->name, > + nm_txrx2str(t), i); > + ND("ktx %s h %d c %d t %d", > + kring->name, kring->rhead, kring->rcur, > kring->rtail); > + mtx_init(&kring->q_lock, (t == NR_TX ? > "nm_txq_lock" : "nm_rxq_lock"), NULL, MTX_DEF); > + init_waitqueue_head(&kring->si); > } > - /* > - * IMPORTANT: Always keep one slot empty. > - */ > - kring->rhead = kring->rcur = kring->nr_hwcur = 0; > - kring->rtail = kring->nr_hwtail = ndesc - 1; > - snprintf(kring->name, sizeof(kring->name) - 1, "%s TX%d", > na->name, i); > - ND("ktx %s h %d c %d t %d", > - kring->name, kring->rhead, kring->rcur, > kring->rtail); > - mtx_init(&kring->q_lock, "nm_txq_lock", NULL, MTX_DEF); > - init_waitqueue_head(&kring->si); > - } > - > - ndesc = na->num_rx_desc; > - for (i = 0; i < nrx; i++) { /* Receive rings */ > - kring = &na->rx_rings[i]; > - bzero(kring, sizeof(*kring)); > - kring->na = na; > - kring->ring_id = i; > - kring->nkr_num_slots = ndesc; > - if (i < na->num_rx_rings) { > - kring->nm_sync = na->nm_rxsync; > - } else if (i == na->num_rx_rings) { > - kring->nm_sync = netmap_rxsync_from_host_compat; > - } > - kring->rhead = kring->rcur = kring->nr_hwcur = 0; > - kring->rtail = kring->nr_hwtail = 0; > - snprintf(kring->name, sizeof(kring->name) - 1, "%s RX%d", > na->name, i); > - ND("krx %s h %d c %d t %d", > - kring->name, kring->rhead, kring->rcur, > kring->rtail); > - mtx_init(&kring->q_lock, "nm_rxq_lock", NULL, MTX_DEF); > - init_waitqueue_head(&kring->si); > + init_waitqueue_head(&na->si[t]); > } > - init_waitqueue_head(&na->tx_si); > - init_waitqueue_head(&na->rx_si); > > - na->tailroom = na->rx_rings + nrx; > + na->tailroom = na->rx_rings + n[NR_RX]; > > return 0; > } > @@ -829,6 +853,10 @@ void > netmap_krings_delete(struct netmap_adapter *na) > { > struct netmap_kring *kring = na->tx_rings; > + enum txrx t; > + > + for_rx_tx(t) > + netmap_knlist_destroy(&na->si[t]); > > /* we rely on the krings layout described above */ > for ( ; kring != na->tailroom; kring++) { > @@ -858,142 +886,35 @@ netmap_hw_krings_delete(struct netmap_ad > } > > > -/* create a new netmap_if for a newly registered fd. > - * If this is the first registration of the adapter, > - * also create the netmap rings and their in-kernel view, > - * the netmap krings. > - */ > -/* call with NMG_LOCK held */ > -static struct netmap_if* > -netmap_if_new(struct netmap_adapter *na) > -{ > - struct netmap_if *nifp; > - > - if (netmap_update_config(na)) { > - /* configuration mismatch, report and fail */ > - return NULL; > - } > - > - if (na->active_fds) /* already registered */ > - goto final; > - > - /* create and init the krings arrays. > - * Depending on the adapter, this may also create > - * the netmap rings themselves > - */ > - if (na->nm_krings_create(na)) > - return NULL; > - > - /* create all missing netmap rings */ > - if (netmap_mem_rings_create(na)) > - goto cleanup; > - > -final: > - > - /* in all cases, create a new netmap if */ > - nifp = netmap_mem_if_new(na); > - if (nifp == NULL) > - goto cleanup; > - > - return (nifp); > - > -cleanup: > - > - if (na->active_fds == 0) { > - netmap_mem_rings_delete(na); > - na->nm_krings_delete(na); > - } > - > - return NULL; > -} > - > - > -/* grab a reference to the memory allocator, if we don't have one > already. The > - * reference is taken from the netmap_adapter registered with the priv. > - */ > -/* call with NMG_LOCK held */ > -static int > -netmap_get_memory_locked(struct netmap_priv_d* p) > -{ > - struct netmap_mem_d *nmd; > - int error = 0; > - > - if (p->np_na == NULL) { > - if (!netmap_mmap_unreg) > - return ENODEV; > - /* for compatibility with older versions of the API > - * we use the global allocator when no interface has been > - * registered > - */ > - nmd = &nm_mem; > - } else { > - nmd = p->np_na->nm_mem; > - } > - if (p->np_mref == NULL) { > - error = netmap_mem_finalize(nmd, p->np_na); > - if (!error) > - p->np_mref = nmd; > - } else if (p->np_mref != nmd) { > - /* a virtual port has been registered, but previous > - * syscalls already used the global allocator. > - * We cannot continue > - */ > - error = ENODEV; > - } > - return error; > -} > - > - > -/* call with NMG_LOCK *not* held */ > -int > -netmap_get_memory(struct netmap_priv_d* p) > -{ > - int error; > - NMG_LOCK(); > - error = netmap_get_memory_locked(p); > - NMG_UNLOCK(); > - return error; > -} > - > - > -/* call with NMG_LOCK held */ > -static int > -netmap_have_memory_locked(struct netmap_priv_d* p) > -{ > - return p->np_mref != NULL; > -} > - > - > -/* call with NMG_LOCK held */ > -static void > -netmap_drop_memory_locked(struct netmap_priv_d* p) > -{ > - if (p->np_mref) { > - netmap_mem_deref(p->np_mref, p->np_na); > - p->np_mref = NULL; > - } > -} > - > > /* > - * Call nm_register(ifp,0) to stop netmap mode on the interface and > + * Undo everything that was done in netmap_do_regif(). In particular, > + * call nm_register(ifp,0) to stop netmap mode on the interface and > * revert to normal operation. > - * The second argument is the nifp to work on. In some cases it is > - * not attached yet to the netmap_priv_d so we need to pass it as > - * a separate argument. > */ > /* call with NMG_LOCK held */ > +static void netmap_unset_ringid(struct netmap_priv_d *); > +static void netmap_rel_exclusive(struct netmap_priv_d *); > static void > -netmap_do_unregif(struct netmap_priv_d *priv, struct netmap_if *nifp) > +netmap_do_unregif(struct netmap_priv_d *priv) > { > struct netmap_adapter *na = priv->np_na; > > NMG_LOCK_ASSERT(); > na->active_fds--; > + /* release exclusive use if it was requested on regif */ > + netmap_rel_exclusive(priv); > if (na->active_fds <= 0) { /* last instance */ > > if (netmap_verbose) > D("deleting last instance for %s", na->name); > + > +#ifdef WITH_MONITOR > + /* walk through all the rings and tell any monitor > + * that the port is going to exit netmap mode > + */ > + netmap_monitor_stop(na); > +#endif > /* > * (TO CHECK) This function is only called > * when the last reference to this file descriptor goes > @@ -1014,37 +935,33 @@ netmap_do_unregif(struct netmap_priv_d * > * XXX The wake up now must happen during *_down(), when > * we order all activities to stop. -gl > */ > - netmap_knlist_destroy(&na->tx_si); > - netmap_knlist_destroy(&na->rx_si); > - > /* delete rings and buffers */ > netmap_mem_rings_delete(na); > na->nm_krings_delete(na); > } > + /* possibily decrement counter of tx_si/rx_si users */ > + netmap_unset_ringid(priv); > /* delete the nifp */ > - netmap_mem_if_delete(na, nifp); > -} > - > -/* call with NMG_LOCK held */ > -static __inline int > -nm_tx_si_user(struct netmap_priv_d *priv) > -{ > - return (priv->np_na != NULL && > - (priv->np_txqlast - priv->np_txqfirst > 1)); > + netmap_mem_if_delete(na, priv->np_nifp); > + /* drop the allocator */ > + netmap_mem_deref(na->nm_mem, na); > + /* mark the priv as unregistered */ > + priv->np_na = NULL; > + priv->np_nifp = NULL; > } > > /* call with NMG_LOCK held */ > static __inline int > -nm_rx_si_user(struct netmap_priv_d *priv) > +nm_si_user(struct netmap_priv_d *priv, enum txrx t) > { > return (priv->np_na != NULL && > - (priv->np_rxqlast - priv->np_rxqfirst > 1)); > + (priv->np_qlast[t] - priv->np_qfirst[t] > 1)); > } > > - > /* > * Destructor of the netmap_priv_d, called when the fd has > - * no active open() and mmap(). Also called in error paths. > + * no active open() and mmap(). > + * Undo all the things done by NIOCREGIF. > * > * returns 1 if this is the last instance and we can free priv > */ > @@ -1066,17 +983,8 @@ netmap_dtor_locked(struct netmap_priv_d > if (!na) { > return 1; //XXX is it correct? > } > - netmap_do_unregif(priv, priv->np_nifp); > - priv->np_nifp = NULL; > - netmap_drop_memory_locked(priv); > - if (priv->np_na) { > - if (nm_tx_si_user(priv)) > - na->tx_si_users--; > - if (nm_rx_si_user(priv)) > - na->rx_si_users--; > - netmap_adapter_put(na); > - priv->np_na = NULL; > - } > + netmap_do_unregif(priv); > + netmap_adapter_put(na); > return 1; > } > > @@ -1148,7 +1056,7 @@ static void > netmap_grab_packets(struct netmap_kring *kring, struct mbq *q, int force) > { > u_int const lim = kring->nkr_num_slots - 1; > - u_int const head = kring->ring->head; > + u_int const head = kring->rhead; > u_int n; > struct netmap_adapter *na = kring->na; > > @@ -1235,7 +1143,6 @@ void > netmap_txsync_to_host(struct netmap_adapter *na) > { > struct netmap_kring *kring = &na->tx_rings[na->num_tx_rings]; > - struct netmap_ring *ring = kring->ring; > u_int const lim = kring->nkr_num_slots - 1; > u_int const head = kring->rhead; > struct mbq q; > @@ -1246,14 +1153,12 @@ netmap_txsync_to_host(struct netmap_adap > * the queue is drained in all cases. > */ > mbq_init(&q); > - ring->cur = head; > netmap_grab_packets(kring, &q, 1 /* force */); > ND("have %d pkts in queue", mbq_len(&q)); > kring->nr_hwcur = head; > kring->nr_hwtail = head + lim; > if (kring->nr_hwtail > lim) > kring->nr_hwtail -= lim + 1; > - nm_txsync_finalize(kring); > > netmap_send_up(na->ifp, &q); > } > @@ -1281,11 +1186,13 @@ netmap_rxsync_from_host(struct netmap_ad > u_int const lim = kring->nkr_num_slots - 1; > u_int const head = kring->rhead; > int ret = 0; > - struct mbq *q = &kring->rx_queue; > + struct mbq *q = &kring->rx_queue, fq; > > (void)pwait; /* disable unused warnings */ > (void)td; > > + mbq_init(&fq); /* fq holds packets to be freed */ > + > mbq_lock(q); > > /* First part: import newly received packets */ > @@ -1308,7 +1215,7 @@ netmap_rxsync_from_host(struct netmap_ad > slot->len = len; > slot->flags = kring->nkr_slot_flags; > nm_i = nm_next(nm_i, lim); > - m_freem(m); > + mbq_enqueue(&fq, m); > } > kring->nr_hwtail = nm_i; > } > @@ -1323,13 +1230,15 @@ netmap_rxsync_from_host(struct netmap_ad > kring->nr_hwcur = head; > } > > *** DIFF OUTPUT TRUNCATED AT 1000 LINES *** > _______________________________________________ > svn-src-head@freebsd.org mailing list > http://lists.freebsd.org/mailman/listinfo/svn-src-head > To unsubscribe, send any mail to "svn-src-head-unsubscribe@freebsd.org" >