From owner-svn-src-head@FreeBSD.ORG Sun Dec 15 08:37:25 2013 Return-Path: Delivered-To: svn-src-head@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTPS id 5C10BA39; Sun, 15 Dec 2013 08:37:25 +0000 (UTC) Received: from svn.freebsd.org (svn.freebsd.org [IPv6:2001:1900:2254:2068::e6a:0]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mx1.freebsd.org (Postfix) with ESMTPS id 46E8813D6; Sun, 15 Dec 2013 08:37:25 +0000 (UTC) Received: from svn.freebsd.org ([127.0.1.70]) by svn.freebsd.org (8.14.7/8.14.7) with ESMTP id rBF8bPKw011584; Sun, 15 Dec 2013 08:37:25 GMT (envelope-from luigi@svn.freebsd.org) Received: (from luigi@localhost) by svn.freebsd.org (8.14.7/8.14.7/Submit) id rBF8bON1011577; Sun, 15 Dec 2013 08:37:24 GMT (envelope-from luigi@svn.freebsd.org) Message-Id: <201312150837.rBF8bON1011577@svn.freebsd.org> From: Luigi Rizzo Date: Sun, 15 Dec 2013 08:37:24 +0000 (UTC) To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r259412 - in head/sys: conf dev/netmap modules/netmap net X-SVN-Group: head MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: svn-src-head@freebsd.org X-Mailman-Version: 2.1.17 Precedence: list List-Id: SVN commit messages for the src tree for head/-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 15 Dec 2013 08:37:25 -0000 Author: luigi Date: Sun Dec 15 08:37:24 2013 New Revision: 259412 URL: http://svnweb.freebsd.org/changeset/base/259412 Log: split netmap code according to functions: - netmap.c base code - netmap_freebsd.c FreeBSD-specific code - netmap_generic.c emulate netmap over standard drivers - netmap_mbq.c simple mbuf tailq - netmap_mem2.c memory management - netmap_vale.c VALE switch simplify devce-specific code Added: head/sys/dev/netmap/netmap_freebsd.c (contents, props changed) head/sys/dev/netmap/netmap_generic.c (contents, props changed) head/sys/dev/netmap/netmap_mbq.c (contents, props changed) head/sys/dev/netmap/netmap_mbq.h (contents, props changed) head/sys/dev/netmap/netmap_vale.c (contents, props changed) Modified: head/sys/conf/files head/sys/dev/netmap/if_em_netmap.h head/sys/dev/netmap/if_igb_netmap.h head/sys/dev/netmap/if_lem_netmap.h head/sys/dev/netmap/if_re_netmap.h head/sys/dev/netmap/ixgbe_netmap.h head/sys/dev/netmap/netmap.c head/sys/dev/netmap/netmap_kern.h head/sys/dev/netmap/netmap_mem2.c head/sys/dev/netmap/netmap_mem2.h head/sys/modules/netmap/Makefile head/sys/net/netmap.h head/sys/net/netmap_user.h Modified: head/sys/conf/files ============================================================================== --- head/sys/conf/files Sun Dec 15 07:07:13 2013 (r259411) +++ head/sys/conf/files Sun Dec 15 08:37:24 2013 (r259412) @@ -1880,7 +1880,11 @@ dev/nand/nfc_if.m optional nand dev/ncv/ncr53c500.c optional ncv dev/ncv/ncr53c500_pccard.c optional ncv pccard dev/netmap/netmap.c optional netmap +dev/netmap/netmap_freebsd.c optional netmap +dev/netmap/netmap_generic.c optional netmap +dev/netmap/netmap_mbq.c optional netmap dev/netmap/netmap_mem2.c optional netmap +dev/netmap/netmap_vale.c optional netmap # compile-with "${NORMAL_C} -Wconversion -Wextra" dev/nge/if_nge.c optional nge dev/nxge/if_nxge.c optional nxge \ Modified: head/sys/dev/netmap/if_em_netmap.h ============================================================================== --- head/sys/dev/netmap/if_em_netmap.h Sun Dec 15 07:07:13 2013 (r259411) +++ head/sys/dev/netmap/if_em_netmap.h Sun Dec 15 08:37:24 2013 (r259412) @@ -26,7 +26,7 @@ /* * $FreeBSD$ * - * netmap support for em. + * netmap support for: em. * * For more details on netmap support please see ixgbe_netmap.h */ @@ -39,10 +39,6 @@ #include -static void em_netmap_block_tasks(struct adapter *); -static void em_netmap_unblock_tasks(struct adapter *); - - // XXX do we need to block/unblock the tasks ? static void em_netmap_block_tasks(struct adapter *adapter) @@ -85,45 +81,31 @@ em_netmap_unblock_tasks(struct adapter * /* - * Register/unregister routine + * Register/unregister. We are already under netmap lock. */ static int -em_netmap_reg(struct ifnet *ifp, int onoff) +em_netmap_reg(struct netmap_adapter *na, int onoff) { + struct ifnet *ifp = na->ifp; struct adapter *adapter = ifp->if_softc; - struct netmap_adapter *na = NA(ifp); - int error = 0; - - if (na == NULL) - return EINVAL; /* no netmap support here */ + EM_CORE_LOCK(adapter); em_disable_intr(adapter); /* Tell the stack that the interface is no longer active */ ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); em_netmap_block_tasks(adapter); - + /* enable or disable flags and callbacks in na and ifp */ if (onoff) { - ifp->if_capenable |= IFCAP_NETMAP; - - na->if_transmit = ifp->if_transmit; - ifp->if_transmit = netmap_transmit; - - em_init_locked(adapter); - if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) == 0) { - error = ENOMEM; - goto fail; - } + nm_set_native_flags(na); } else { -fail: - /* return to non-netmap mode */ - ifp->if_transmit = na->if_transmit; - ifp->if_capenable &= ~IFCAP_NETMAP; - em_init_locked(adapter); /* also enable intr */ + nm_clear_native_flags(na); } + em_init_locked(adapter); /* also enable intr */ em_netmap_unblock_tasks(adapter); - return (error); + EM_CORE_UNLOCK(adapter); + return (ifp->if_drv_flags & IFF_DRV_RUNNING ? 0 : 1); } @@ -131,93 +113,103 @@ fail: * Reconcile kernel and user view of the transmit ring. */ static int -em_netmap_txsync(struct ifnet *ifp, u_int ring_nr, int flags) +em_netmap_txsync(struct netmap_adapter *na, u_int ring_nr, int flags) { - struct adapter *adapter = ifp->if_softc; - struct tx_ring *txr = &adapter->tx_rings[ring_nr]; - struct netmap_adapter *na = NA(ifp); + struct ifnet *ifp = na->ifp; struct netmap_kring *kring = &na->tx_rings[ring_nr]; struct netmap_ring *ring = kring->ring; - u_int j, k, l, n = 0, lim = kring->nkr_num_slots - 1; - + u_int nm_i; /* index into the netmap ring */ + u_int nic_i; /* index into the NIC ring */ + u_int n, new_slots; + u_int const lim = kring->nkr_num_slots - 1; + u_int const cur = nm_txsync_prologue(kring, &new_slots); /* generate an interrupt approximately every half ring */ u_int report_frequency = kring->nkr_num_slots >> 1; - k = ring->cur; - if (k > lim) + /* device-specific */ + struct adapter *adapter = ifp->if_softc; + struct tx_ring *txr = &adapter->tx_rings[ring_nr]; + + if (cur > lim) /* error checking in nm_txsync_prologue() */ return netmap_ring_reinit(kring); bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, BUS_DMASYNC_POSTREAD); /* - * Process new packets to send. j is the current index in the - * netmap ring, l is the corresponding index in the NIC ring. + * First part: process new packets to send. */ - j = kring->nr_hwcur; - if (j != k) { /* we have new packets to send */ - l = netmap_idx_k2n(kring, j); - for (n = 0; j != k; n++) { - /* slot is the current slot in the netmap ring */ - struct netmap_slot *slot = &ring->slot[j]; - /* curr is the current slot in the nic ring */ - struct e1000_tx_desc *curr = &txr->tx_base[l]; - struct em_buffer *txbuf = &txr->tx_buffers[l]; - int flags = ((slot->flags & NS_REPORT) || - j == 0 || j == report_frequency) ? - E1000_TXD_CMD_RS : 0; + + nm_i = kring->nr_hwcur; + if (nm_i != cur) { /* we have new packets to send */ + nic_i = netmap_idx_k2n(kring, nm_i); + for (n = 0; nm_i != cur; n++) { + struct netmap_slot *slot = &ring->slot[nm_i]; + u_int len = slot->len; uint64_t paddr; void *addr = PNMB(slot, &paddr); - u_int len = slot->len; - if (addr == netmap_buffer_base || len > NETMAP_BUF_SIZE) { - return netmap_ring_reinit(kring); - } + /* device-specific */ + struct e1000_tx_desc *curr = &txr->tx_base[nic_i]; + struct em_buffer *txbuf = &txr->tx_buffers[nic_i]; + int flags = (slot->flags & NS_REPORT || + nic_i == 0 || nic_i == report_frequency) ? + E1000_TXD_CMD_RS : 0; + + NM_CHECK_ADDR_LEN(addr, len); - slot->flags &= ~NS_REPORT; if (slot->flags & NS_BUF_CHANGED) { curr->buffer_addr = htole64(paddr); /* buffer has changed, reload map */ netmap_reload_map(txr->txtag, txbuf->map, addr); - slot->flags &= ~NS_BUF_CHANGED; } + slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED); + + /* Fill the slot in the NIC ring. */ curr->upper.data = 0; curr->lower.data = htole32(adapter->txd_cmd | len | (E1000_TXD_CMD_EOP | flags) ); bus_dmamap_sync(txr->txtag, txbuf->map, BUS_DMASYNC_PREWRITE); - j = (j == lim) ? 0 : j + 1; - l = (l == lim) ? 0 : l + 1; + + nm_i = nm_next(nm_i, lim); + nic_i = nm_next(nic_i, lim); } - kring->nr_hwcur = k; /* the saved ring->cur */ - kring->nr_hwavail -= n; + kring->nr_hwcur = cur; /* the saved ring->cur */ + /* decrease avail by # of packets sent minus previous ones */ + kring->nr_hwavail -= new_slots; + /* synchronize the NIC ring */ bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); + BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - E1000_WRITE_REG(&adapter->hw, E1000_TDT(txr->me), l); + /* (re)start the tx unit up to slot nic_i (excluded) */ + E1000_WRITE_REG(&adapter->hw, E1000_TDT(txr->me), nic_i); } - if (n == 0 || kring->nr_hwavail < 1) { + /* + * Second part: reclaim buffers for completed transmissions. + */ + if (flags & NAF_FORCE_RECLAIM || kring->nr_hwavail < 1) { int delta; /* record completed transmissions using TDH */ - l = E1000_READ_REG(&adapter->hw, E1000_TDH(ring_nr)); - if (l >= kring->nkr_num_slots) { /* XXX can it happen ? */ - D("TDH wrap %d", l); - l -= kring->nkr_num_slots; + nic_i = E1000_READ_REG(&adapter->hw, E1000_TDH(ring_nr)); + if (nic_i >= kring->nkr_num_slots) { /* XXX can it happen ? */ + D("TDH wrap %d", nic_i); + nic_i -= kring->nkr_num_slots; } - delta = l - txr->next_to_clean; + delta = nic_i - txr->next_to_clean; if (delta) { /* some completed, increment hwavail. */ if (delta < 0) delta += kring->nkr_num_slots; - txr->next_to_clean = l; + txr->next_to_clean = nic_i; kring->nr_hwavail += delta; } } - /* update avail to what the kernel knows */ - ring->avail = kring->nr_hwavail; + + nm_txsync_finalize(kring, cur); return 0; } @@ -227,19 +219,23 @@ em_netmap_txsync(struct ifnet *ifp, u_in * Reconcile kernel and user view of the receive ring. */ static int -em_netmap_rxsync(struct ifnet *ifp, u_int ring_nr, int flags) +em_netmap_rxsync(struct netmap_adapter *na, u_int ring_nr, int flags) { - struct adapter *adapter = ifp->if_softc; - struct rx_ring *rxr = &adapter->rx_rings[ring_nr]; - struct netmap_adapter *na = NA(ifp); + struct ifnet *ifp = na->ifp; struct netmap_kring *kring = &na->rx_rings[ring_nr]; struct netmap_ring *ring = kring->ring; - u_int j, l, n, lim = kring->nkr_num_slots - 1; + u_int nm_i; /* index into the netmap ring */ + u_int nic_i; /* index into the NIC ring */ + u_int n, resvd; + u_int const lim = kring->nkr_num_slots - 1; + u_int const cur = nm_rxsync_prologue(kring, &resvd); /* cur + res */ int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & NKR_PENDINTR; - u_int k = ring->cur, resvd = ring->reserved; - k = ring->cur; - if (k > lim) + /* device-specific */ + struct adapter *adapter = ifp->if_softc; + struct rx_ring *rxr = &adapter->rx_rings[ring_nr]; + + if (cur > lim) return netmap_ring_reinit(kring); /* XXX check sync modes */ @@ -247,84 +243,85 @@ em_netmap_rxsync(struct ifnet *ifp, u_in BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); /* - * Import newly received packets into the netmap ring. - * j is an index in the netmap ring, l in the NIC ring. + * First part: import newly received packets. */ - l = rxr->next_to_check; - j = netmap_idx_n2k(kring, l); if (netmap_no_pendintr || force_update) { uint16_t slot_flags = kring->nkr_slot_flags; + nic_i = rxr->next_to_check; + nm_i = netmap_idx_n2k(kring, nic_i); + for (n = 0; ; n++) { - struct e1000_rx_desc *curr = &rxr->rx_base[l]; + struct e1000_rx_desc *curr = &rxr->rx_base[nic_i]; uint32_t staterr = le32toh(curr->status); if ((staterr & E1000_RXD_STAT_DD) == 0) break; - ring->slot[j].len = le16toh(curr->length); - ring->slot[j].flags = slot_flags; - bus_dmamap_sync(rxr->rxtag, rxr->rx_buffers[l].map, + ring->slot[nm_i].len = le16toh(curr->length); + ring->slot[nm_i].flags = slot_flags; + bus_dmamap_sync(rxr->rxtag, rxr->rx_buffers[nic_i].map, BUS_DMASYNC_POSTREAD); - j = (j == lim) ? 0 : j + 1; + nm_i = nm_next(nm_i, lim); /* make sure next_to_refresh follows next_to_check */ - rxr->next_to_refresh = l; // XXX - l = (l == lim) ? 0 : l + 1; + rxr->next_to_refresh = nic_i; // XXX + nic_i = nm_next(nic_i, lim); } if (n) { /* update the state variables */ - rxr->next_to_check = l; + rxr->next_to_check = nic_i; kring->nr_hwavail += n; } kring->nr_kflags &= ~NKR_PENDINTR; } - /* skip past packets that userspace has released */ - j = kring->nr_hwcur; /* netmap ring index */ - if (resvd > 0) { - if (resvd + ring->avail >= lim + 1) { - D("XXX invalid reserve/avail %d %d", resvd, ring->avail); - ring->reserved = resvd = 0; // XXX panic... - } - k = (k >= resvd) ? k - resvd : k + lim + 1 - resvd; - } - if (j != k) { /* userspace has released some packets. */ - l = netmap_idx_k2n(kring, j); /* NIC ring index */ - for (n = 0; j != k; n++) { - struct netmap_slot *slot = &ring->slot[j]; - struct e1000_rx_desc *curr = &rxr->rx_base[l]; - struct em_buffer *rxbuf = &rxr->rx_buffers[l]; + /* + * Second part: skip past packets that userspace has released. + */ + nm_i = kring->nr_hwcur; + if (nm_i != cur) { + nic_i = netmap_idx_k2n(kring, nm_i); + for (n = 0; nm_i != cur; n++) { + struct netmap_slot *slot = &ring->slot[nm_i]; uint64_t paddr; void *addr = PNMB(slot, &paddr); - if (addr == netmap_buffer_base) { /* bad buf */ - return netmap_ring_reinit(kring); - } + struct e1000_rx_desc *curr = &rxr->rx_base[nic_i]; + struct em_buffer *rxbuf = &rxr->rx_buffers[nic_i]; + + if (addr == netmap_buffer_base) /* bad buf */ + goto ring_reset; if (slot->flags & NS_BUF_CHANGED) { - curr->buffer_addr = htole64(paddr); /* buffer has changed, reload map */ + curr->buffer_addr = htole64(paddr); netmap_reload_map(rxr->rxtag, rxbuf->map, addr); slot->flags &= ~NS_BUF_CHANGED; } curr->status = 0; bus_dmamap_sync(rxr->rxtag, rxbuf->map, BUS_DMASYNC_PREREAD); - j = (j == lim) ? 0 : j + 1; - l = (l == lim) ? 0 : l + 1; + nm_i = nm_next(nm_i, lim); + nic_i = nm_next(nic_i, lim); } kring->nr_hwavail -= n; - kring->nr_hwcur = k; + kring->nr_hwcur = cur; + bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); /* * IMPORTANT: we must leave one free slot in the ring, - * so move l back by one unit + * so move nic_i back by one unit */ - l = (l == 0) ? lim : l - 1; - E1000_WRITE_REG(&adapter->hw, E1000_RDT(rxr->me), l); + nic_i = (nic_i == 0) ? lim : nic_i - 1; + E1000_WRITE_REG(&adapter->hw, E1000_RDT(rxr->me), nic_i); } - /* tell userspace that there are new packets */ + + /* tell userspace that there might be new packets */ ring->avail = kring->nr_hwavail - resvd; + return 0; + +ring_reset: + return netmap_ring_reinit(kring); } @@ -342,7 +339,8 @@ em_netmap_attach(struct adapter *adapter na.nm_txsync = em_netmap_txsync; na.nm_rxsync = em_netmap_rxsync; na.nm_register = em_netmap_reg; - netmap_attach(&na, adapter->num_queues); + na.num_tx_rings = na.num_rx_rings = adapter->num_queues; + netmap_attach(&na); } /* end of file */ Modified: head/sys/dev/netmap/if_igb_netmap.h ============================================================================== --- head/sys/dev/netmap/if_igb_netmap.h Sun Dec 15 07:07:13 2013 (r259411) +++ head/sys/dev/netmap/if_igb_netmap.h Sun Dec 15 08:37:24 2013 (r259412) @@ -37,44 +37,43 @@ #include /* vtophys ? */ #include +/* + * Adaptation to different versions of the driver. + */ + +#ifndef IGB_MEDIA_RESET +/* at the same time as IGB_MEDIA_RESET was defined, the + * tx buffer descriptor was renamed, so use this to revert + * back to the old name. + */ +#define igb_tx_buf igb_tx_buffer +#endif + /* - * register-unregister routine + * Register/unregister. We are already under netmap lock. */ static int -igb_netmap_reg(struct ifnet *ifp, int onoff) +igb_netmap_reg(struct netmap_adapter *na, int onoff) { + struct ifnet *ifp = na->ifp; struct adapter *adapter = ifp->if_softc; - struct netmap_adapter *na = NA(ifp); - int error = 0; - - if (na == NULL) - return EINVAL; /* no netmap support here */ + IGB_CORE_LOCK(adapter); igb_disable_intr(adapter); /* Tell the stack that the interface is no longer active */ ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); + /* enable or disable flags and callbacks in na and ifp */ if (onoff) { - ifp->if_capenable |= IFCAP_NETMAP; - - na->if_transmit = ifp->if_transmit; - ifp->if_transmit = netmap_transmit; - - igb_init_locked(adapter); - if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) == 0) { - error = ENOMEM; - goto fail; - } + nm_set_native_flags(na); } else { -fail: - /* restore if_transmit */ - ifp->if_transmit = na->if_transmit; - ifp->if_capenable &= ~IFCAP_NETMAP; - igb_init_locked(adapter); /* also enable intr */ + nm_clear_native_flags(na); } - return (error); + igb_init_locked(adapter); /* also enable intr */ + IGB_CORE_UNLOCK(adapter); + return (ifp->if_drv_flags & IFF_DRV_RUNNING ? 0 : 1); } @@ -82,68 +81,62 @@ fail: * Reconcile kernel and user view of the transmit ring. */ static int -igb_netmap_txsync(struct ifnet *ifp, u_int ring_nr, int flags) +igb_netmap_txsync(struct netmap_adapter *na, u_int ring_nr, int flags) { - struct adapter *adapter = ifp->if_softc; - struct tx_ring *txr = &adapter->tx_rings[ring_nr]; - struct netmap_adapter *na = NA(ifp); + struct ifnet *ifp = na->ifp; struct netmap_kring *kring = &na->tx_rings[ring_nr]; struct netmap_ring *ring = kring->ring; - u_int j, k, l, n = 0, lim = kring->nkr_num_slots - 1; - + u_int nm_i; /* index into the netmap ring */ + u_int nic_i; /* index into the NIC ring */ + u_int n, new_slots; + u_int const lim = kring->nkr_num_slots - 1; + u_int const cur = nm_txsync_prologue(kring, &new_slots); /* generate an interrupt approximately every half ring */ u_int report_frequency = kring->nkr_num_slots >> 1; - k = ring->cur; - if (k > lim) + /* device-specific */ + struct adapter *adapter = ifp->if_softc; + struct tx_ring *txr = &adapter->tx_rings[ring_nr]; + /* 82575 needs the queue index added */ + u32 olinfo_status = + (adapter->hw.mac.type == e1000_82575) ? (txr->me << 4) : 0; + + if (cur > lim) /* error checking in nm_txsync_prologue() */ return netmap_ring_reinit(kring); bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_POSTREAD); + BUS_DMASYNC_POSTREAD); - /* check for new packets to send. - * j indexes the netmap ring, l indexes the nic ring, and - * j = kring->nr_hwcur, l = E1000_TDT (not tracked), - * j == (l + kring->nkr_hwofs) % ring_size + /* + * First part: process new packets to send. */ - j = kring->nr_hwcur; - if (j != k) { /* we have new packets to send */ - /* 82575 needs the queue index added */ - u32 olinfo_status = - (adapter->hw.mac.type == e1000_82575) ? (txr->me << 4) : 0; - - l = netmap_idx_k2n(kring, j); - for (n = 0; j != k; n++) { - /* slot is the current slot in the netmap ring */ - struct netmap_slot *slot = &ring->slot[j]; - /* curr is the current slot in the nic ring */ - union e1000_adv_tx_desc *curr = - (union e1000_adv_tx_desc *)&txr->tx_base[l]; -#ifndef IGB_MEDIA_RESET -/* at the same time as IGB_MEDIA_RESET was defined, the - * tx buffer descriptor was renamed, so use this to revert - * back to the old name. - */ -#define igb_tx_buf igb_tx_buffer -#endif - struct igb_tx_buf *txbuf = &txr->tx_buffers[l]; - int flags = ((slot->flags & NS_REPORT) || - j == 0 || j == report_frequency) ? - E1000_ADVTXD_DCMD_RS : 0; + + nm_i = kring->nr_hwcur; + if (nm_i != cur) { /* we have new packets to send */ + nic_i = netmap_idx_k2n(kring, nm_i); + for (n = 0; nm_i != cur; n++) { + struct netmap_slot *slot = &ring->slot[nm_i]; + u_int len = slot->len; uint64_t paddr; void *addr = PNMB(slot, &paddr); - u_int len = slot->len; - if (addr == netmap_buffer_base || len > NETMAP_BUF_SIZE) { - return netmap_ring_reinit(kring); - } + /* device-specific */ + union e1000_adv_tx_desc *curr = + (union e1000_adv_tx_desc *)&txr->tx_base[nic_i]; + struct igb_tx_buf *txbuf = &txr->tx_buffers[nic_i]; + int flags = (slot->flags & NS_REPORT || + nic_i == 0 || nic_i == report_frequency) ? + E1000_ADVTXD_DCMD_RS : 0; + + NM_CHECK_ADDR_LEN(addr, len); - slot->flags &= ~NS_REPORT; if (slot->flags & NS_BUF_CHANGED) { /* buffer has changed, reload map */ netmap_reload_map(txr->txtag, txbuf->map, addr); - slot->flags &= ~NS_BUF_CHANGED; } + slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED); + + /* Fill the slot in the NIC ring. */ curr->read.buffer_addr = htole64(paddr); // XXX check olinfo and cmd_type_len curr->read.olinfo_status = @@ -151,48 +144,56 @@ igb_netmap_txsync(struct ifnet *ifp, u_i (len<< E1000_ADVTXD_PAYLEN_SHIFT)); curr->read.cmd_type_len = htole32(len | E1000_ADVTXD_DTYP_DATA | - E1000_ADVTXD_DCMD_IFCS | - E1000_ADVTXD_DCMD_DEXT | - E1000_ADVTXD_DCMD_EOP | flags); + E1000_ADVTXD_DCMD_IFCS | + E1000_ADVTXD_DCMD_DEXT | + E1000_ADVTXD_DCMD_EOP | flags); + /* make sure changes to the buffer are synced */ bus_dmamap_sync(txr->txtag, txbuf->map, BUS_DMASYNC_PREWRITE); - j = (j == lim) ? 0 : j + 1; - l = (l == lim) ? 0 : l + 1; + + nm_i = nm_next(nm_i, lim); + nic_i = nm_next(nic_i, lim); } - kring->nr_hwcur = k; /* the saved ring->cur */ - kring->nr_hwavail -= n; + kring->nr_hwcur = cur; /* the saved ring->cur */ + /* decrease avail by # of packets sent minus previous ones */ + kring->nr_hwavail -= new_slots; /* Set the watchdog XXX ? */ txr->queue_status = IGB_QUEUE_WORKING; txr->watchdog_time = ticks; + /* synchronize the NIC ring */ bus_dmamap_sync(txr->txdma.dma_tag, txr->txdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); + BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); - E1000_WRITE_REG(&adapter->hw, E1000_TDT(txr->me), l); + /* (re)start the tx unit up to slot nic_i (excluded) */ + E1000_WRITE_REG(&adapter->hw, E1000_TDT(txr->me), nic_i); } - if (n == 0 || kring->nr_hwavail < 1) { + /* + * Second part: reclaim buffers for completed transmissions. + */ + if (flags & NAF_FORCE_RECLAIM || kring->nr_hwavail < 1) { int delta; /* record completed transmissions using TDH */ - l = E1000_READ_REG(&adapter->hw, E1000_TDH(ring_nr)); - if (l >= kring->nkr_num_slots) { /* XXX can it happen ? */ - D("TDH wrap %d", l); - l -= kring->nkr_num_slots; + nic_i = E1000_READ_REG(&adapter->hw, E1000_TDH(ring_nr)); + if (nic_i >= kring->nkr_num_slots) { /* XXX can it happen ? */ + D("TDH wrap %d", nic_i); + nic_i -= kring->nkr_num_slots; } - delta = l - txr->next_to_clean; + delta = nic_i - txr->next_to_clean; if (delta) { /* some completed, increment hwavail. */ if (delta < 0) delta += kring->nkr_num_slots; - txr->next_to_clean = l; + txr->next_to_clean = nic_i; kring->nr_hwavail += delta; } } - /* update avail to what the kernel knows */ - ring->avail = kring->nr_hwavail; + + nm_txsync_finalize(kring, cur); return 0; } @@ -202,101 +203,107 @@ igb_netmap_txsync(struct ifnet *ifp, u_i * Reconcile kernel and user view of the receive ring. */ static int -igb_netmap_rxsync(struct ifnet *ifp, u_int ring_nr, int flags) +igb_netmap_rxsync(struct netmap_adapter *na, u_int ring_nr, int flags) { - struct adapter *adapter = ifp->if_softc; - struct rx_ring *rxr = &adapter->rx_rings[ring_nr]; - struct netmap_adapter *na = NA(ifp); + struct ifnet *ifp = na->ifp; struct netmap_kring *kring = &na->rx_rings[ring_nr]; struct netmap_ring *ring = kring->ring; - u_int j, l, n, lim = kring->nkr_num_slots - 1; + u_int nm_i; /* index into the netmap ring */ + u_int nic_i; /* index into the NIC ring */ + u_int n, resvd; + u_int const lim = kring->nkr_num_slots - 1; + u_int const cur = nm_rxsync_prologue(kring, &resvd); /* cur + res */ int force_update = (flags & NAF_FORCE_READ) || kring->nr_kflags & NKR_PENDINTR; - u_int k = ring->cur, resvd = ring->reserved; - k = ring->cur; - if (k > lim) + /* device-specific */ + struct adapter *adapter = ifp->if_softc; + struct rx_ring *rxr = &adapter->rx_rings[ring_nr]; + + if (cur > lim) return netmap_ring_reinit(kring); /* XXX check sync modes */ bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, - BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); + BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); /* - * import newly received packets into the netmap ring. - * j is an index in the netmap ring, l in the NIC ring. + * First part: import newly received packets. */ - l = rxr->next_to_check; - j = netmap_idx_n2k(kring, l); if (netmap_no_pendintr || force_update) { uint16_t slot_flags = kring->nkr_slot_flags; + nic_i = rxr->next_to_check; + nm_i = netmap_idx_n2k(kring, nic_i); + for (n = 0; ; n++) { - union e1000_adv_rx_desc *curr = &rxr->rx_base[l]; + union e1000_adv_rx_desc *curr = &rxr->rx_base[nic_i]; uint32_t staterr = le32toh(curr->wb.upper.status_error); if ((staterr & E1000_RXD_STAT_DD) == 0) break; - ring->slot[j].len = le16toh(curr->wb.upper.length); - ring->slot[j].flags = slot_flags; + ring->slot[nm_i].len = le16toh(curr->wb.upper.length); + ring->slot[nm_i].flags = slot_flags; bus_dmamap_sync(rxr->ptag, - rxr->rx_buffers[l].pmap, BUS_DMASYNC_POSTREAD); - j = (j == lim) ? 0 : j + 1; - l = (l == lim) ? 0 : l + 1; + rxr->rx_buffers[nic_i].pmap, BUS_DMASYNC_POSTREAD); + nm_i = nm_next(nm_i, lim); + nic_i = nm_next(nic_i, lim); } if (n) { /* update the state variables */ - rxr->next_to_check = l; + rxr->next_to_check = nic_i; kring->nr_hwavail += n; } kring->nr_kflags &= ~NKR_PENDINTR; } - /* skip past packets that userspace has released */ - j = kring->nr_hwcur; /* netmap ring index */ - if (resvd > 0) { - if (resvd + ring->avail >= lim + 1) { - D("XXX invalid reserve/avail %d %d", resvd, ring->avail); - ring->reserved = resvd = 0; // XXX panic... - } - k = (k >= resvd) ? k - resvd : k + lim + 1 - resvd; - } - if (j != k) { /* userspace has released some packets. */ - l = netmap_idx_k2n(kring, j); - for (n = 0; j != k; n++) { - struct netmap_slot *slot = ring->slot + j; - union e1000_adv_rx_desc *curr = &rxr->rx_base[l]; - struct igb_rx_buf *rxbuf = rxr->rx_buffers + l; + /* + * Second part: skip past packets that userspace has released. + */ + nm_i = kring->nr_hwcur; + if (nm_i != cur) { + nic_i = netmap_idx_k2n(kring, nm_i); + for (n = 0; nm_i != cur; n++) { + struct netmap_slot *slot = &ring->slot[nm_i]; uint64_t paddr; void *addr = PNMB(slot, &paddr); - if (addr == netmap_buffer_base) { /* bad buf */ - return netmap_ring_reinit(kring); - } + union e1000_adv_rx_desc *curr = &rxr->rx_base[nic_i]; + struct igb_rx_buf *rxbuf = &rxr->rx_buffers[nic_i]; + + if (addr == netmap_buffer_base) /* bad buf */ + goto ring_reset; if (slot->flags & NS_BUF_CHANGED) { + /* buffer has changed, reload map */ netmap_reload_map(rxr->ptag, rxbuf->pmap, addr); slot->flags &= ~NS_BUF_CHANGED; } - curr->read.pkt_addr = htole64(paddr); curr->wb.upper.status_error = 0; + curr->read.pkt_addr = htole64(paddr); bus_dmamap_sync(rxr->ptag, rxbuf->pmap, - BUS_DMASYNC_PREREAD); - j = (j == lim) ? 0 : j + 1; - l = (l == lim) ? 0 : l + 1; + BUS_DMASYNC_PREREAD); + nm_i = nm_next(nm_i, lim); + nic_i = nm_next(nic_i, lim); } kring->nr_hwavail -= n; - kring->nr_hwcur = k; + kring->nr_hwcur = cur; + bus_dmamap_sync(rxr->rxdma.dma_tag, rxr->rxdma.dma_map, - BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); + BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); /* * IMPORTANT: we must leave one free slot in the ring, - * so move l back by one unit + * so move nic_i back by one unit */ - l = (l == 0) ? lim : l - 1; - E1000_WRITE_REG(&adapter->hw, E1000_RDT(rxr->me), l); + nic_i = (nic_i == 0) ? lim : nic_i - 1; + E1000_WRITE_REG(&adapter->hw, E1000_RDT(rxr->me), nic_i); } - /* tell userspace that there are new packets */ + + /* tell userspace that there might be new packets */ ring->avail = kring->nr_hwavail - resvd; + return 0; + +ring_reset: + return netmap_ring_reinit(kring); } @@ -314,6 +321,8 @@ igb_netmap_attach(struct adapter *adapte na.nm_txsync = igb_netmap_txsync; na.nm_rxsync = igb_netmap_rxsync; na.nm_register = igb_netmap_reg; - netmap_attach(&na, adapter->num_queues); -} + na.num_tx_rings = na.num_rx_rings = adapter->num_queues; + netmap_attach(&na); +} + /* end of file */ Modified: head/sys/dev/netmap/if_lem_netmap.h ============================================================================== --- head/sys/dev/netmap/if_lem_netmap.h Sun Dec 15 07:07:13 2013 (r259411) +++ head/sys/dev/netmap/if_lem_netmap.h Sun Dec 15 08:37:24 2013 (r259412) @@ -27,11 +27,12 @@ /* * $FreeBSD$ * - * netmap support for "lem" + * netmap support for: lem * * For details on netmap support please see ixgbe_netmap.h */ + #include #include #include @@ -40,17 +41,13 @@ /* - * Register/unregister + * Register/unregister. We are already under netmap lock. */ static int -lem_netmap_reg(struct ifnet *ifp, int onoff) +lem_netmap_reg(struct netmap_adapter *na, int onoff) { + struct ifnet *ifp = na->ifp; struct adapter *adapter = ifp->if_softc; - struct netmap_adapter *na = NA(ifp); - int error = 0; - - if (na == NULL) - return EINVAL; EM_CORE_LOCK(adapter); @@ -64,24 +61,14 @@ lem_netmap_reg(struct ifnet *ifp, int on taskqueue_drain(adapter->tq, &adapter->rxtx_task); taskqueue_drain(adapter->tq, &adapter->link_task); #endif /* !EM_LEGCY_IRQ */ - if (onoff) { - ifp->if_capenable |= IFCAP_NETMAP; - na->if_transmit = ifp->if_transmit; - ifp->if_transmit = netmap_transmit; - - lem_init_locked(adapter); - if ((ifp->if_drv_flags & (IFF_DRV_RUNNING | IFF_DRV_OACTIVE)) == 0) { - error = ENOMEM; - goto fail; - } + /* enable or disable flags and callbacks in na and ifp */ + if (onoff) { + nm_set_native_flags(na); } else { -fail: - /* return to non-netmap mode */ - ifp->if_transmit = na->if_transmit; - ifp->if_capenable &= ~IFCAP_NETMAP; - lem_init_locked(adapter); /* also enable intr */ + nm_clear_native_flags(na); } + lem_init_locked(adapter); /* also enable intr */ #ifndef EM_LEGACY_IRQ taskqueue_unblock(adapter->tq); // XXX do we need this ? @@ -89,7 +76,7 @@ fail: EM_CORE_UNLOCK(adapter); - return (error); + return (ifp->if_drv_flags & IFF_DRV_RUNNING ? 0 : 1); } @@ -97,108 +84,102 @@ fail: * Reconcile kernel and user view of the transmit ring. */ static int -lem_netmap_txsync(struct ifnet *ifp, u_int ring_nr, int flags) +lem_netmap_txsync(struct netmap_adapter *na, u_int ring_nr, int flags) { - struct adapter *adapter = ifp->if_softc; - struct netmap_adapter *na = NA(ifp); + struct ifnet *ifp = na->ifp; struct netmap_kring *kring = &na->tx_rings[ring_nr]; struct netmap_ring *ring = kring->ring; - u_int j, k, l, n = 0, lim = kring->nkr_num_slots - 1; - + u_int nm_i; /* index into the netmap ring */ + u_int nic_i; /* index into the NIC ring */ + u_int n, new_slots; + u_int const lim = kring->nkr_num_slots - 1; + u_int const cur = nm_txsync_prologue(kring, &new_slots); /* generate an interrupt approximately every half ring */ - int report_frequency = kring->nkr_num_slots >> 1; + u_int report_frequency = kring->nkr_num_slots >> 1; + + /* device-specific */ + struct adapter *adapter = ifp->if_softc; - ND("%s: hwofs %d, hwcur %d hwavail %d lease %d cur %d avail %d", - ifp->if_xname, - kring->nkr_hwofs, kring->nr_hwcur, kring->nr_hwavail, - kring->nkr_hwlease, - ring->cur, ring->avail); - /* take a copy of ring->cur now, and never read it again */ - k = ring->cur; - if (k > lim) + if (cur > lim) /* error checking in nm_txsync_prologue() */ return netmap_ring_reinit(kring); bus_dmamap_sync(adapter->txdma.dma_tag, adapter->txdma.dma_map, BUS_DMASYNC_POSTREAD); + /* - * Process new packets to send. j is the current index in the - * netmap ring, l is the corresponding index in the NIC ring. + * First part: process new packets to send. */ - j = kring->nr_hwcur; - if (netmap_verbose > 255) - RD(5, "device %s send %d->%d", ifp->if_xname, j, k); - if (j != k) { /* we have new packets to send */ - l = netmap_idx_k2n(kring, j); - for (n = 0; j != k; n++) { - /* slot is the current slot in the netmap ring */ - struct netmap_slot *slot = &ring->slot[j]; - /* curr is the current slot in the nic ring */ - struct e1000_tx_desc *curr = &adapter->tx_desc_base[l]; - struct em_buffer *txbuf = &adapter->tx_buffer_area[l]; - int flags = ((slot->flags & NS_REPORT) || - j == 0 || j == report_frequency) ? - E1000_TXD_CMD_RS : 0; + + nm_i = kring->nr_hwcur; + if (nm_i != cur) { /* we have new packets to send */ + nic_i = netmap_idx_k2n(kring, nm_i); + for (n = 0; nm_i != cur; n++) { + struct netmap_slot *slot = &ring->slot[nm_i]; + u_int len = slot->len; uint64_t paddr; void *addr = PNMB(slot, &paddr); - u_int len = slot->len; - if (addr == netmap_buffer_base || len > NETMAP_BUF_SIZE) { - return netmap_ring_reinit(kring); - } - ND("slot %d NIC %d %s", j, l, nm_dump_buf(addr, len, 128, NULL)); + /* device-specific */ + struct e1000_tx_desc *curr = &adapter->tx_desc_base[nic_i]; + struct em_buffer *txbuf = &adapter->tx_buffer_area[nic_i]; + int flags = (slot->flags & NS_REPORT || + nic_i == 0 || nic_i == report_frequency) ? + E1000_TXD_CMD_RS : 0; - slot->flags &= ~NS_REPORT; - if (1 || slot->flags & NS_BUF_CHANGED) { + NM_CHECK_ADDR_LEN(addr, len); + + if (slot->flags & NS_BUF_CHANGED) { /* buffer has changed, reload map */ - netmap_reload_map(adapter->txtag, txbuf->map, addr); curr->buffer_addr = htole64(paddr); - slot->flags &= ~NS_BUF_CHANGED; + netmap_reload_map(adapter->txtag, txbuf->map, addr); } + slot->flags &= ~(NS_REPORT | NS_BUF_CHANGED); + + /* Fill the slot in the NIC ring. */ curr->upper.data = 0; - curr->lower.data = - htole32( adapter->txd_cmd | len | + curr->lower.data = htole32(adapter->txd_cmd | len | (E1000_TXD_CMD_EOP | flags) ); - - ND("len %d kring %d nic %d", len, j, l); *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***