Index: head/sys/dev/rt/if_rt.c =================================================================== --- head/sys/dev/rt/if_rt.c (revision 366507) +++ head/sys/dev/rt/if_rt.c (revision 366508) @@ -1,2964 +1,2966 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2015-2016, Stanislav Galabov * Copyright (c) 2014, Aleksandr A. Mityaev * Copyright (c) 2011, Aleksandr Rybalko * based on hard work * by Alexander Egorenkov * and by Damien Bergamini * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice unmodified, this list of conditions, and the following * disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include "if_rtvar.h" #include "if_rtreg.h" +#include + #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "opt_platform.h" #include "opt_rt305x.h" #ifdef FDT #include #include #include #endif #include #include #ifdef RT_MDIO #include #include #include "mdio_if.h" #endif #if 0 #include #include #endif #ifdef IF_RT_PHY_SUPPORT #include "miibus_if.h" #endif /* * Defines and macros */ #define RT_MAX_AGG_SIZE 3840 #define RT_TX_DATA_SEG0_SIZE MJUMPAGESIZE #define RT_MS(_v, _f) (((_v) & _f) >> _f##_S) #define RT_SM(_v, _f) (((_v) << _f##_S) & _f) #define RT_TX_WATCHDOG_TIMEOUT 5 #define RT_CHIPID_RT2880 0x2880 #define RT_CHIPID_RT3050 0x3050 #define RT_CHIPID_RT3883 0x3883 #define RT_CHIPID_RT5350 0x5350 #define RT_CHIPID_MT7620 0x7620 #define RT_CHIPID_MT7621 0x7621 #ifdef FDT /* more specific and new models should go first */ static const struct ofw_compat_data rt_compat_data[] = { { "ralink,rt2880-eth", RT_CHIPID_RT2880 }, { "ralink,rt3050-eth", RT_CHIPID_RT3050 }, { "ralink,rt3352-eth", RT_CHIPID_RT3050 }, { "ralink,rt3883-eth", RT_CHIPID_RT3883 }, { "ralink,rt5350-eth", RT_CHIPID_RT5350 }, { "ralink,mt7620a-eth", RT_CHIPID_MT7620 }, { "mediatek,mt7620-eth", RT_CHIPID_MT7620 }, { "ralink,mt7621-eth", RT_CHIPID_MT7621 }, { "mediatek,mt7621-eth", RT_CHIPID_MT7621 }, { NULL, 0 } }; #endif /* * Static function prototypes */ static int rt_probe(device_t dev); static int rt_attach(device_t dev); static int rt_detach(device_t dev); static int rt_shutdown(device_t dev); static int rt_suspend(device_t dev); static int rt_resume(device_t dev); static void rt_init_locked(void *priv); static void rt_init(void *priv); static void rt_stop_locked(void *priv); static void rt_stop(void *priv); static void rt_start(struct ifnet *ifp); static int rt_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data); static void rt_periodic(void *arg); static void rt_tx_watchdog(void *arg); static void rt_intr(void *arg); static void rt_rt5350_intr(void *arg); static void rt_tx_coherent_intr(struct rt_softc *sc); static void rt_rx_coherent_intr(struct rt_softc *sc); static void rt_rx_delay_intr(struct rt_softc *sc); static void rt_tx_delay_intr(struct rt_softc *sc); static void rt_rx_intr(struct rt_softc *sc, int qid); static void rt_tx_intr(struct rt_softc *sc, int qid); static void rt_rx_done_task(void *context, int pending); static void rt_tx_done_task(void *context, int pending); static void rt_periodic_task(void *context, int pending); static int rt_rx_eof(struct rt_softc *sc, struct rt_softc_rx_ring *ring, int limit); static void rt_tx_eof(struct rt_softc *sc, struct rt_softc_tx_ring *ring); static void rt_update_stats(struct rt_softc *sc); static void rt_watchdog(struct rt_softc *sc); static void rt_update_raw_counters(struct rt_softc *sc); static void rt_intr_enable(struct rt_softc *sc, uint32_t intr_mask); static void rt_intr_disable(struct rt_softc *sc, uint32_t intr_mask); static int rt_txrx_enable(struct rt_softc *sc); static int rt_alloc_rx_ring(struct rt_softc *sc, struct rt_softc_rx_ring *ring, int qid); static void rt_reset_rx_ring(struct rt_softc *sc, struct rt_softc_rx_ring *ring); static void rt_free_rx_ring(struct rt_softc *sc, struct rt_softc_rx_ring *ring); static int rt_alloc_tx_ring(struct rt_softc *sc, struct rt_softc_tx_ring *ring, int qid); static void rt_reset_tx_ring(struct rt_softc *sc, struct rt_softc_tx_ring *ring); static void rt_free_tx_ring(struct rt_softc *sc, struct rt_softc_tx_ring *ring); static void rt_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error); static void rt_sysctl_attach(struct rt_softc *sc); #ifdef IF_RT_PHY_SUPPORT void rt_miibus_statchg(device_t); #endif #if defined(IF_RT_PHY_SUPPORT) || defined(RT_MDIO) static int rt_miibus_readreg(device_t, int, int); static int rt_miibus_writereg(device_t, int, int, int); #endif static int rt_ifmedia_upd(struct ifnet *); static void rt_ifmedia_sts(struct ifnet *, struct ifmediareq *); static SYSCTL_NODE(_hw, OID_AUTO, rt, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "RT driver parameters"); #ifdef IF_RT_DEBUG static int rt_debug = 0; SYSCTL_INT(_hw_rt, OID_AUTO, debug, CTLFLAG_RWTUN, &rt_debug, 0, "RT debug level"); #endif static int rt_probe(device_t dev) { struct rt_softc *sc = device_get_softc(dev); char buf[80]; #ifdef FDT const struct ofw_compat_data * cd; cd = ofw_bus_search_compatible(dev, rt_compat_data); if (cd->ocd_data == 0) return (ENXIO); sc->rt_chipid = (unsigned int)(cd->ocd_data); #else #if defined(MT7620) sc->rt_chipid = RT_CHIPID_MT7620; #elif defined(MT7621) sc->rt_chipid = RT_CHIPID_MT7621; #elif defined(RT5350) sc->rt_chipid = RT_CHIPID_RT5350; #else sc->rt_chipid = RT_CHIPID_RT3050; #endif #endif snprintf(buf, sizeof(buf), "Ralink %cT%x onChip Ethernet driver", sc->rt_chipid >= 0x7600 ? 'M' : 'R', sc->rt_chipid); device_set_desc_copy(dev, buf); return (BUS_PROBE_GENERIC); } /* * macaddr_atoi - translate string MAC address to uint8_t array */ static int macaddr_atoi(const char *str, uint8_t *mac) { int count, i; unsigned int amac[ETHER_ADDR_LEN]; /* Aligned version */ count = sscanf(str, "%x%*c%x%*c%x%*c%x%*c%x%*c%x", &amac[0], &amac[1], &amac[2], &amac[3], &amac[4], &amac[5]); if (count < ETHER_ADDR_LEN) { memset(mac, 0, ETHER_ADDR_LEN); return (1); } /* Copy aligned to result */ for (i = 0; i < ETHER_ADDR_LEN; i ++) mac[i] = (amac[i] & 0xff); return (0); } #ifdef USE_GENERATED_MAC_ADDRESS /* * generate_mac(uin8_t *mac) * This is MAC address generator for cases when real device MAC address * unknown or not yet accessible. * Use 'b','s','d' signature and 3 octets from CRC32 on kenv. * MAC = 'b', 's', 'd', CRC[3]^CRC[2], CRC[1], CRC[0] * * Output - MAC address, that do not change between reboots, if hints or * bootloader info unchange. */ static void generate_mac(uint8_t *mac) { unsigned char *cp; int i = 0; uint32_t crc = 0xffffffff; /* Generate CRC32 on kenv */ for (cp = kenvp[0]; cp != NULL; cp = kenvp[++i]) { crc = calculate_crc32c(crc, cp, strlen(cp) + 1); } crc = ~crc; mac[0] = 'b'; mac[1] = 's'; mac[2] = 'd'; mac[3] = (crc >> 24) ^ ((crc >> 16) & 0xff); mac[4] = (crc >> 8) & 0xff; mac[5] = crc & 0xff; } #endif /* * ether_request_mac - try to find usable MAC address. */ static int ether_request_mac(device_t dev, uint8_t *mac) { char *var; /* * "ethaddr" is passed via envp on RedBoot platforms * "kmac" is passed via argv on RouterBOOT platforms */ #if defined(RT305X_UBOOT) || defined(__REDBOOT__) || defined(__ROUTERBOOT__) if ((var = kern_getenv("ethaddr")) != NULL || (var = kern_getenv("kmac")) != NULL ) { if(!macaddr_atoi(var, mac)) { printf("%s: use %s macaddr from KENV\n", device_get_nameunit(dev), var); freeenv(var); return (0); } freeenv(var); } #endif /* * Try from hints * hint.[dev].[unit].macaddr */ if (!resource_string_value(device_get_name(dev), device_get_unit(dev), "macaddr", (const char **)&var)) { if(!macaddr_atoi(var, mac)) { printf("%s: use %s macaddr from hints\n", device_get_nameunit(dev), var); return (0); } } #ifdef USE_GENERATED_MAC_ADDRESS generate_mac(mac); device_printf(dev, "use generated %02x:%02x:%02x:%02x:%02x:%02x " "macaddr\n", mac[0], mac[1], mac[2], mac[3], mac[4], mac[5]); #else /* Hardcoded */ mac[0] = 0x00; mac[1] = 0x18; mac[2] = 0xe7; mac[3] = 0xd5; mac[4] = 0x83; mac[5] = 0x90; device_printf(dev, "use hardcoded 00:18:e7:d5:83:90 macaddr\n"); #endif return (0); } /* * Reset hardware */ static void reset_freng(struct rt_softc *sc) { /* XXX hard reset kills everything so skip it ... */ return; } static int rt_attach(device_t dev) { struct rt_softc *sc; struct ifnet *ifp; int error, i; #ifdef FDT phandle_t node; char fdtval[32]; #endif sc = device_get_softc(dev); sc->dev = dev; #ifdef FDT node = ofw_bus_get_node(sc->dev); #endif mtx_init(&sc->lock, device_get_nameunit(dev), MTX_NETWORK_LOCK, MTX_DEF | MTX_RECURSE); sc->mem_rid = 0; sc->mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->mem_rid, RF_ACTIVE | RF_SHAREABLE); if (sc->mem == NULL) { device_printf(dev, "could not allocate memory resource\n"); error = ENXIO; goto fail; } sc->bst = rman_get_bustag(sc->mem); sc->bsh = rman_get_bushandle(sc->mem); sc->irq_rid = 0; sc->irq = bus_alloc_resource_any(dev, SYS_RES_IRQ, &sc->irq_rid, RF_ACTIVE); if (sc->irq == NULL) { device_printf(dev, "could not allocate interrupt resource\n"); error = ENXIO; goto fail; } #ifdef IF_RT_DEBUG sc->debug = rt_debug; SYSCTL_ADD_INT(device_get_sysctl_ctx(dev), SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO, "debug", CTLFLAG_RW, &sc->debug, 0, "rt debug level"); #endif /* Reset hardware */ reset_freng(sc); if (sc->rt_chipid == RT_CHIPID_MT7620) { sc->csum_fail_ip = MT7620_RXD_SRC_IP_CSUM_FAIL; sc->csum_fail_l4 = MT7620_RXD_SRC_L4_CSUM_FAIL; } else if (sc->rt_chipid == RT_CHIPID_MT7621) { sc->csum_fail_ip = MT7621_RXD_SRC_IP_CSUM_FAIL; sc->csum_fail_l4 = MT7621_RXD_SRC_L4_CSUM_FAIL; } else { sc->csum_fail_ip = RT305X_RXD_SRC_IP_CSUM_FAIL; sc->csum_fail_l4 = RT305X_RXD_SRC_L4_CSUM_FAIL; } /* Fill in soc-specific registers map */ switch(sc->rt_chipid) { case RT_CHIPID_MT7620: case RT_CHIPID_MT7621: sc->gdma1_base = MT7620_GDMA1_BASE; /* fallthrough */ case RT_CHIPID_RT5350: device_printf(dev, "%cT%x Ethernet MAC (rev 0x%08x)\n", sc->rt_chipid >= 0x7600 ? 'M' : 'R', sc->rt_chipid, sc->mac_rev); /* RT5350: No GDMA, PSE, CDMA, PPE */ RT_WRITE(sc, GE_PORT_BASE + 0x0C00, // UDPCS, TCPCS, IPCS=1 RT_READ(sc, GE_PORT_BASE + 0x0C00) | (0x7<<16)); sc->delay_int_cfg=RT5350_PDMA_BASE+RT5350_DELAY_INT_CFG; sc->fe_int_status=RT5350_FE_INT_STATUS; sc->fe_int_enable=RT5350_FE_INT_ENABLE; sc->pdma_glo_cfg=RT5350_PDMA_BASE+RT5350_PDMA_GLO_CFG; sc->pdma_rst_idx=RT5350_PDMA_BASE+RT5350_PDMA_RST_IDX; for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) { sc->tx_base_ptr[i]=RT5350_PDMA_BASE+RT5350_TX_BASE_PTR(i); sc->tx_max_cnt[i]=RT5350_PDMA_BASE+RT5350_TX_MAX_CNT(i); sc->tx_ctx_idx[i]=RT5350_PDMA_BASE+RT5350_TX_CTX_IDX(i); sc->tx_dtx_idx[i]=RT5350_PDMA_BASE+RT5350_TX_DTX_IDX(i); } sc->rx_ring_count=2; sc->rx_base_ptr[0]=RT5350_PDMA_BASE+RT5350_RX_BASE_PTR0; sc->rx_max_cnt[0]=RT5350_PDMA_BASE+RT5350_RX_MAX_CNT0; sc->rx_calc_idx[0]=RT5350_PDMA_BASE+RT5350_RX_CALC_IDX0; sc->rx_drx_idx[0]=RT5350_PDMA_BASE+RT5350_RX_DRX_IDX0; sc->rx_base_ptr[1]=RT5350_PDMA_BASE+RT5350_RX_BASE_PTR1; sc->rx_max_cnt[1]=RT5350_PDMA_BASE+RT5350_RX_MAX_CNT1; sc->rx_calc_idx[1]=RT5350_PDMA_BASE+RT5350_RX_CALC_IDX1; sc->rx_drx_idx[1]=RT5350_PDMA_BASE+RT5350_RX_DRX_IDX1; sc->int_rx_done_mask=RT5350_INT_RXQ0_DONE; sc->int_tx_done_mask=RT5350_INT_TXQ0_DONE; break; default: device_printf(dev, "RT305XF Ethernet MAC (rev 0x%08x)\n", sc->mac_rev); sc->gdma1_base = GDMA1_BASE; sc->delay_int_cfg=PDMA_BASE+DELAY_INT_CFG; sc->fe_int_status=GE_PORT_BASE+FE_INT_STATUS; sc->fe_int_enable=GE_PORT_BASE+FE_INT_ENABLE; sc->pdma_glo_cfg=PDMA_BASE+PDMA_GLO_CFG; sc->pdma_rst_idx=PDMA_BASE+PDMA_RST_IDX; for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) { sc->tx_base_ptr[i]=PDMA_BASE+TX_BASE_PTR(i); sc->tx_max_cnt[i]=PDMA_BASE+TX_MAX_CNT(i); sc->tx_ctx_idx[i]=PDMA_BASE+TX_CTX_IDX(i); sc->tx_dtx_idx[i]=PDMA_BASE+TX_DTX_IDX(i); } sc->rx_ring_count=1; sc->rx_base_ptr[0]=PDMA_BASE+RX_BASE_PTR0; sc->rx_max_cnt[0]=PDMA_BASE+RX_MAX_CNT0; sc->rx_calc_idx[0]=PDMA_BASE+RX_CALC_IDX0; sc->rx_drx_idx[0]=PDMA_BASE+RX_DRX_IDX0; sc->int_rx_done_mask=INT_RX_DONE; sc->int_tx_done_mask=INT_TXQ0_DONE; } if (sc->gdma1_base != 0) RT_WRITE(sc, sc->gdma1_base + GDMA_FWD_CFG, ( GDM_ICS_EN | /* Enable IP Csum */ GDM_TCS_EN | /* Enable TCP Csum */ GDM_UCS_EN | /* Enable UDP Csum */ GDM_STRPCRC | /* Strip CRC from packet */ GDM_DST_PORT_CPU << GDM_UFRC_P_SHIFT | /* fwd UCast to CPU */ GDM_DST_PORT_CPU << GDM_BFRC_P_SHIFT | /* fwd BCast to CPU */ GDM_DST_PORT_CPU << GDM_MFRC_P_SHIFT | /* fwd MCast to CPU */ GDM_DST_PORT_CPU << GDM_OFRC_P_SHIFT /* fwd Other to CPU */ )); #ifdef FDT if (sc->rt_chipid == RT_CHIPID_RT2880 || sc->rt_chipid == RT_CHIPID_RT3883) { if (OF_getprop(node, "port-mode", fdtval, sizeof(fdtval)) > 0 && strcmp(fdtval, "gigasw") == 0) RT_WRITE(sc, MDIO_CFG, MDIO_2880_GIGA_INIT); else RT_WRITE(sc, MDIO_CFG, MDIO_2880_100T_INIT); } #endif /* allocate Tx and Rx rings */ for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) { error = rt_alloc_tx_ring(sc, &sc->tx_ring[i], i); if (error != 0) { device_printf(dev, "could not allocate Tx ring #%d\n", i); goto fail; } } sc->tx_ring_mgtqid = 5; for (i = 0; i < sc->rx_ring_count; i++) { error = rt_alloc_rx_ring(sc, &sc->rx_ring[i], i); if (error != 0) { device_printf(dev, "could not allocate Rx ring\n"); goto fail; } } callout_init(&sc->periodic_ch, 0); callout_init_mtx(&sc->tx_watchdog_ch, &sc->lock, 0); ifp = sc->ifp = if_alloc(IFT_ETHER); if (ifp == NULL) { device_printf(dev, "could not if_alloc()\n"); error = ENOMEM; goto fail; } ifp->if_softc = sc; if_initname(ifp, device_get_name(sc->dev), device_get_unit(sc->dev)); ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; ifp->if_init = rt_init; ifp->if_ioctl = rt_ioctl; ifp->if_start = rt_start; #define RT_TX_QLEN 256 IFQ_SET_MAXLEN(&ifp->if_snd, RT_TX_QLEN); ifp->if_snd.ifq_drv_maxlen = RT_TX_QLEN; IFQ_SET_READY(&ifp->if_snd); #ifdef IF_RT_PHY_SUPPORT error = mii_attach(dev, &sc->rt_miibus, ifp, rt_ifmedia_upd, rt_ifmedia_sts, BMSR_DEFCAPMASK, MII_PHY_ANY, MII_OFFSET_ANY, 0); if (error != 0) { device_printf(dev, "attaching PHYs failed\n"); error = ENXIO; goto fail; } #else ifmedia_init(&sc->rt_ifmedia, 0, rt_ifmedia_upd, rt_ifmedia_sts); ifmedia_add(&sc->rt_ifmedia, IFM_ETHER | IFM_100_TX | IFM_FDX, 0, NULL); ifmedia_set(&sc->rt_ifmedia, IFM_ETHER | IFM_100_TX | IFM_FDX); #endif /* IF_RT_PHY_SUPPORT */ ether_request_mac(dev, sc->mac_addr); ether_ifattach(ifp, sc->mac_addr); /* * Tell the upper layer(s) we support long frames. */ ifp->if_hdrlen = sizeof(struct ether_vlan_header); ifp->if_capabilities |= IFCAP_VLAN_MTU; ifp->if_capenable |= IFCAP_VLAN_MTU; ifp->if_capabilities |= IFCAP_RXCSUM|IFCAP_TXCSUM; ifp->if_capenable |= IFCAP_RXCSUM|IFCAP_TXCSUM; /* init task queue */ NET_TASK_INIT(&sc->rx_done_task, 0, rt_rx_done_task, sc); TASK_INIT(&sc->tx_done_task, 0, rt_tx_done_task, sc); TASK_INIT(&sc->periodic_task, 0, rt_periodic_task, sc); sc->rx_process_limit = 100; sc->taskqueue = taskqueue_create("rt_taskq", M_NOWAIT, taskqueue_thread_enqueue, &sc->taskqueue); taskqueue_start_threads(&sc->taskqueue, 1, PI_NET, "%s taskq", device_get_nameunit(sc->dev)); rt_sysctl_attach(sc); /* set up interrupt */ error = bus_setup_intr(dev, sc->irq, INTR_TYPE_NET | INTR_MPSAFE, NULL, (sc->rt_chipid == RT_CHIPID_RT5350 || sc->rt_chipid == RT_CHIPID_MT7620 || sc->rt_chipid == RT_CHIPID_MT7621) ? rt_rt5350_intr : rt_intr, sc, &sc->irqh); if (error != 0) { printf("%s: could not set up interrupt\n", device_get_nameunit(dev)); goto fail; } #ifdef IF_RT_DEBUG device_printf(dev, "debug var at %#08x\n", (u_int)&(sc->debug)); #endif return (0); fail: /* free Tx and Rx rings */ for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) rt_free_tx_ring(sc, &sc->tx_ring[i]); for (i = 0; i < sc->rx_ring_count; i++) rt_free_rx_ring(sc, &sc->rx_ring[i]); mtx_destroy(&sc->lock); if (sc->mem != NULL) bus_release_resource(dev, SYS_RES_MEMORY, sc->mem_rid, sc->mem); if (sc->irq != NULL) bus_release_resource(dev, SYS_RES_IRQ, sc->irq_rid, sc->irq); return (error); } /* * Set media options. */ static int rt_ifmedia_upd(struct ifnet *ifp) { struct rt_softc *sc; #ifdef IF_RT_PHY_SUPPORT struct mii_data *mii; struct mii_softc *miisc; int error = 0; sc = ifp->if_softc; RT_SOFTC_LOCK(sc); mii = device_get_softc(sc->rt_miibus); LIST_FOREACH(miisc, &mii->mii_phys, mii_list) PHY_RESET(miisc); error = mii_mediachg(mii); RT_SOFTC_UNLOCK(sc); return (error); #else /* !IF_RT_PHY_SUPPORT */ struct ifmedia *ifm; struct ifmedia_entry *ife; sc = ifp->if_softc; ifm = &sc->rt_ifmedia; ife = ifm->ifm_cur; if (IFM_TYPE(ifm->ifm_media) != IFM_ETHER) return (EINVAL); if (IFM_SUBTYPE(ife->ifm_media) == IFM_AUTO) { device_printf(sc->dev, "AUTO is not supported for multiphy MAC"); return (EINVAL); } /* * Ignore everything */ return (0); #endif /* IF_RT_PHY_SUPPORT */ } /* * Report current media status. */ static void rt_ifmedia_sts(struct ifnet *ifp, struct ifmediareq *ifmr) { #ifdef IF_RT_PHY_SUPPORT struct rt_softc *sc; struct mii_data *mii; sc = ifp->if_softc; RT_SOFTC_LOCK(sc); mii = device_get_softc(sc->rt_miibus); mii_pollstat(mii); ifmr->ifm_active = mii->mii_media_active; ifmr->ifm_status = mii->mii_media_status; ifmr->ifm_active = IFM_ETHER | IFM_100_TX | IFM_FDX; ifmr->ifm_status = IFM_AVALID | IFM_ACTIVE; RT_SOFTC_UNLOCK(sc); #else /* !IF_RT_PHY_SUPPORT */ ifmr->ifm_status = IFM_AVALID | IFM_ACTIVE; ifmr->ifm_active = IFM_ETHER | IFM_100_TX | IFM_FDX; #endif /* IF_RT_PHY_SUPPORT */ } static int rt_detach(device_t dev) { struct rt_softc *sc; struct ifnet *ifp; int i; sc = device_get_softc(dev); ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_ANY, "detaching\n"); RT_SOFTC_LOCK(sc); ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); callout_stop(&sc->periodic_ch); callout_stop(&sc->tx_watchdog_ch); taskqueue_drain(sc->taskqueue, &sc->rx_done_task); taskqueue_drain(sc->taskqueue, &sc->tx_done_task); taskqueue_drain(sc->taskqueue, &sc->periodic_task); /* free Tx and Rx rings */ for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) rt_free_tx_ring(sc, &sc->tx_ring[i]); for (i = 0; i < sc->rx_ring_count; i++) rt_free_rx_ring(sc, &sc->rx_ring[i]); RT_SOFTC_UNLOCK(sc); #ifdef IF_RT_PHY_SUPPORT if (sc->rt_miibus != NULL) device_delete_child(dev, sc->rt_miibus); #endif ether_ifdetach(ifp); if_free(ifp); taskqueue_free(sc->taskqueue); mtx_destroy(&sc->lock); bus_generic_detach(dev); bus_teardown_intr(dev, sc->irq, sc->irqh); bus_release_resource(dev, SYS_RES_IRQ, sc->irq_rid, sc->irq); bus_release_resource(dev, SYS_RES_MEMORY, sc->mem_rid, sc->mem); return (0); } static int rt_shutdown(device_t dev) { struct rt_softc *sc; sc = device_get_softc(dev); RT_DPRINTF(sc, RT_DEBUG_ANY, "shutting down\n"); rt_stop(sc); return (0); } static int rt_suspend(device_t dev) { struct rt_softc *sc; sc = device_get_softc(dev); RT_DPRINTF(sc, RT_DEBUG_ANY, "suspending\n"); rt_stop(sc); return (0); } static int rt_resume(device_t dev) { struct rt_softc *sc; struct ifnet *ifp; sc = device_get_softc(dev); ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_ANY, "resuming\n"); if (ifp->if_flags & IFF_UP) rt_init(sc); return (0); } /* * rt_init_locked - Run initialization process having locked mtx. */ static void rt_init_locked(void *priv) { struct rt_softc *sc; struct ifnet *ifp; #ifdef IF_RT_PHY_SUPPORT struct mii_data *mii; #endif int i, ntries; uint32_t tmp; sc = priv; ifp = sc->ifp; #ifdef IF_RT_PHY_SUPPORT mii = device_get_softc(sc->rt_miibus); #endif RT_DPRINTF(sc, RT_DEBUG_ANY, "initializing\n"); RT_SOFTC_ASSERT_LOCKED(sc); /* hardware reset */ //RT_WRITE(sc, GE_PORT_BASE + FE_RST_GLO, PSE_RESET); //rt305x_sysctl_set(SYSCTL_RSTCTRL, SYSCTL_RSTCTRL_FRENG); /* Fwd to CPU (uni|broad|multi)cast and Unknown */ if (sc->gdma1_base != 0) RT_WRITE(sc, sc->gdma1_base + GDMA_FWD_CFG, ( GDM_ICS_EN | /* Enable IP Csum */ GDM_TCS_EN | /* Enable TCP Csum */ GDM_UCS_EN | /* Enable UDP Csum */ GDM_STRPCRC | /* Strip CRC from packet */ GDM_DST_PORT_CPU << GDM_UFRC_P_SHIFT | /* fwd UCast to CPU */ GDM_DST_PORT_CPU << GDM_BFRC_P_SHIFT | /* fwd BCast to CPU */ GDM_DST_PORT_CPU << GDM_MFRC_P_SHIFT | /* fwd MCast to CPU */ GDM_DST_PORT_CPU << GDM_OFRC_P_SHIFT /* fwd Other to CPU */ )); /* disable DMA engine */ RT_WRITE(sc, sc->pdma_glo_cfg, 0); RT_WRITE(sc, sc->pdma_rst_idx, 0xffffffff); /* wait while DMA engine is busy */ for (ntries = 0; ntries < 100; ntries++) { tmp = RT_READ(sc, sc->pdma_glo_cfg); if (!(tmp & (FE_TX_DMA_BUSY | FE_RX_DMA_BUSY))) break; DELAY(1000); } if (ntries == 100) { device_printf(sc->dev, "timeout waiting for DMA engine\n"); goto fail; } /* reset Rx and Tx rings */ tmp = FE_RST_DRX_IDX0 | FE_RST_DTX_IDX3 | FE_RST_DTX_IDX2 | FE_RST_DTX_IDX1 | FE_RST_DTX_IDX0; RT_WRITE(sc, sc->pdma_rst_idx, tmp); /* XXX switch set mac address */ for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) rt_reset_tx_ring(sc, &sc->tx_ring[i]); for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) { /* update TX_BASE_PTRx */ RT_WRITE(sc, sc->tx_base_ptr[i], sc->tx_ring[i].desc_phys_addr); RT_WRITE(sc, sc->tx_max_cnt[i], RT_SOFTC_TX_RING_DESC_COUNT); RT_WRITE(sc, sc->tx_ctx_idx[i], 0); } /* init Rx ring */ for (i = 0; i < sc->rx_ring_count; i++) rt_reset_rx_ring(sc, &sc->rx_ring[i]); /* update RX_BASE_PTRx */ for (i = 0; i < sc->rx_ring_count; i++) { RT_WRITE(sc, sc->rx_base_ptr[i], sc->rx_ring[i].desc_phys_addr); RT_WRITE(sc, sc->rx_max_cnt[i], RT_SOFTC_RX_RING_DATA_COUNT); RT_WRITE(sc, sc->rx_calc_idx[i], RT_SOFTC_RX_RING_DATA_COUNT - 1); } /* write back DDONE, 16byte burst enable RX/TX DMA */ tmp = FE_TX_WB_DDONE | FE_DMA_BT_SIZE16 | FE_RX_DMA_EN | FE_TX_DMA_EN; if (sc->rt_chipid == RT_CHIPID_MT7620 || sc->rt_chipid == RT_CHIPID_MT7621) tmp |= (1<<31); RT_WRITE(sc, sc->pdma_glo_cfg, tmp); /* disable interrupts mitigation */ RT_WRITE(sc, sc->delay_int_cfg, 0); /* clear pending interrupts */ RT_WRITE(sc, sc->fe_int_status, 0xffffffff); /* enable interrupts */ if (sc->rt_chipid == RT_CHIPID_RT5350 || sc->rt_chipid == RT_CHIPID_MT7620 || sc->rt_chipid == RT_CHIPID_MT7621) tmp = RT5350_INT_TX_COHERENT | RT5350_INT_RX_COHERENT | RT5350_INT_TXQ3_DONE | RT5350_INT_TXQ2_DONE | RT5350_INT_TXQ1_DONE | RT5350_INT_TXQ0_DONE | RT5350_INT_RXQ1_DONE | RT5350_INT_RXQ0_DONE; else tmp = CNT_PPE_AF | CNT_GDM_AF | PSE_P2_FC | GDM_CRC_DROP | PSE_BUF_DROP | GDM_OTHER_DROP | PSE_P1_FC | PSE_P0_FC | PSE_FQ_EMPTY | INT_TX_COHERENT | INT_RX_COHERENT | INT_TXQ3_DONE | INT_TXQ2_DONE | INT_TXQ1_DONE | INT_TXQ0_DONE | INT_RX_DONE; sc->intr_enable_mask = tmp; RT_WRITE(sc, sc->fe_int_enable, tmp); if (rt_txrx_enable(sc) != 0) goto fail; #ifdef IF_RT_PHY_SUPPORT if (mii) mii_mediachg(mii); #endif /* IF_RT_PHY_SUPPORT */ ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; ifp->if_drv_flags |= IFF_DRV_RUNNING; sc->periodic_round = 0; callout_reset(&sc->periodic_ch, hz / 10, rt_periodic, sc); return; fail: rt_stop_locked(sc); } /* * rt_init - lock and initialize device. */ static void rt_init(void *priv) { struct rt_softc *sc; sc = priv; RT_SOFTC_LOCK(sc); rt_init_locked(sc); RT_SOFTC_UNLOCK(sc); } /* * rt_stop_locked - stop TX/RX w/ lock */ static void rt_stop_locked(void *priv) { struct rt_softc *sc; struct ifnet *ifp; sc = priv; ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_ANY, "stopping\n"); RT_SOFTC_ASSERT_LOCKED(sc); sc->tx_timer = 0; ifp->if_drv_flags &= ~(IFF_DRV_RUNNING | IFF_DRV_OACTIVE); callout_stop(&sc->periodic_ch); callout_stop(&sc->tx_watchdog_ch); RT_SOFTC_UNLOCK(sc); taskqueue_block(sc->taskqueue); /* * Sometime rt_stop_locked called from isr and we get panic * When found, I fix it */ #ifdef notyet taskqueue_drain(sc->taskqueue, &sc->rx_done_task); taskqueue_drain(sc->taskqueue, &sc->tx_done_task); taskqueue_drain(sc->taskqueue, &sc->periodic_task); #endif RT_SOFTC_LOCK(sc); /* disable interrupts */ RT_WRITE(sc, sc->fe_int_enable, 0); if(sc->rt_chipid != RT_CHIPID_RT5350 && sc->rt_chipid != RT_CHIPID_MT7620 && sc->rt_chipid != RT_CHIPID_MT7621) { /* reset adapter */ RT_WRITE(sc, GE_PORT_BASE + FE_RST_GLO, PSE_RESET); } if (sc->gdma1_base != 0) RT_WRITE(sc, sc->gdma1_base + GDMA_FWD_CFG, ( GDM_ICS_EN | /* Enable IP Csum */ GDM_TCS_EN | /* Enable TCP Csum */ GDM_UCS_EN | /* Enable UDP Csum */ GDM_STRPCRC | /* Strip CRC from packet */ GDM_DST_PORT_CPU << GDM_UFRC_P_SHIFT | /* fwd UCast to CPU */ GDM_DST_PORT_CPU << GDM_BFRC_P_SHIFT | /* fwd BCast to CPU */ GDM_DST_PORT_CPU << GDM_MFRC_P_SHIFT | /* fwd MCast to CPU */ GDM_DST_PORT_CPU << GDM_OFRC_P_SHIFT /* fwd Other to CPU */ )); } static void rt_stop(void *priv) { struct rt_softc *sc; sc = priv; RT_SOFTC_LOCK(sc); rt_stop_locked(sc); RT_SOFTC_UNLOCK(sc); } /* * rt_tx_data - transmit packet. */ static int rt_tx_data(struct rt_softc *sc, struct mbuf *m, int qid) { struct ifnet *ifp; struct rt_softc_tx_ring *ring; struct rt_softc_tx_data *data; struct rt_txdesc *desc; struct mbuf *m_d; bus_dma_segment_t dma_seg[RT_SOFTC_MAX_SCATTER]; int error, ndmasegs, ndescs, i; KASSERT(qid >= 0 && qid < RT_SOFTC_TX_RING_COUNT, ("%s: Tx data: invalid qid=%d\n", device_get_nameunit(sc->dev), qid)); RT_SOFTC_TX_RING_ASSERT_LOCKED(&sc->tx_ring[qid]); ifp = sc->ifp; ring = &sc->tx_ring[qid]; desc = &ring->desc[ring->desc_cur]; data = &ring->data[ring->data_cur]; error = bus_dmamap_load_mbuf_sg(ring->data_dma_tag, data->dma_map, m, dma_seg, &ndmasegs, 0); if (error != 0) { /* too many fragments, linearize */ RT_DPRINTF(sc, RT_DEBUG_TX, "could not load mbuf DMA map, trying to linearize " "mbuf: ndmasegs=%d, len=%d, error=%d\n", ndmasegs, m->m_pkthdr.len, error); m_d = m_collapse(m, M_NOWAIT, 16); if (m_d == NULL) { m_freem(m); m = NULL; return (ENOMEM); } m = m_d; sc->tx_defrag_packets++; error = bus_dmamap_load_mbuf_sg(ring->data_dma_tag, data->dma_map, m, dma_seg, &ndmasegs, 0); if (error != 0) { device_printf(sc->dev, "could not load mbuf DMA map: " "ndmasegs=%d, len=%d, error=%d\n", ndmasegs, m->m_pkthdr.len, error); m_freem(m); return (error); } } if (m->m_pkthdr.len == 0) ndmasegs = 0; /* determine how many Tx descs are required */ ndescs = 1 + ndmasegs / 2; if ((ring->desc_queued + ndescs) > (RT_SOFTC_TX_RING_DESC_COUNT - 2)) { RT_DPRINTF(sc, RT_DEBUG_TX, "there are not enough Tx descs\n"); sc->no_tx_desc_avail++; bus_dmamap_unload(ring->data_dma_tag, data->dma_map); m_freem(m); return (EFBIG); } data->m = m; /* set up Tx descs */ for (i = 0; i < ndmasegs; i += 2) { /* TODO: this needs to be refined as MT7620 for example has * a different word3 layout than RT305x and RT5350 (the last * one doesn't use word3 at all). And so does MT7621... */ if (sc->rt_chipid != RT_CHIPID_MT7621) { /* Set destination */ if (sc->rt_chipid != RT_CHIPID_MT7620) desc->dst = (TXDSCR_DST_PORT_GDMA1); if ((ifp->if_capenable & IFCAP_TXCSUM) != 0) desc->dst |= (TXDSCR_IP_CSUM_GEN | TXDSCR_UDP_CSUM_GEN | TXDSCR_TCP_CSUM_GEN); /* Set queue id */ desc->qn = qid; /* No PPPoE */ desc->pppoe = 0; /* No VLAN */ desc->vid = 0; } else { desc->vid = 0; desc->pppoe = 0; desc->qn = 0; desc->dst = 2; } desc->sdp0 = htole32(dma_seg[i].ds_addr); desc->sdl0 = htole16(dma_seg[i].ds_len | ( ((i+1) == ndmasegs )?RT_TXDESC_SDL0_LASTSEG:0 )); if ((i+1) < ndmasegs) { desc->sdp1 = htole32(dma_seg[i+1].ds_addr); desc->sdl1 = htole16(dma_seg[i+1].ds_len | ( ((i+2) == ndmasegs )?RT_TXDESC_SDL1_LASTSEG:0 )); } else { desc->sdp1 = 0; desc->sdl1 = 0; } if ((i+2) < ndmasegs) { ring->desc_queued++; ring->desc_cur = (ring->desc_cur + 1) % RT_SOFTC_TX_RING_DESC_COUNT; } desc = &ring->desc[ring->desc_cur]; } RT_DPRINTF(sc, RT_DEBUG_TX, "sending data: len=%d, ndmasegs=%d, " "DMA ds_len=%d/%d/%d/%d/%d\n", m->m_pkthdr.len, ndmasegs, (int) dma_seg[0].ds_len, (int) dma_seg[1].ds_len, (int) dma_seg[2].ds_len, (int) dma_seg[3].ds_len, (int) dma_seg[4].ds_len); bus_dmamap_sync(ring->seg0_dma_tag, ring->seg0_dma_map, BUS_DMASYNC_PREWRITE); bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_PREWRITE); bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_PREWRITE); ring->desc_queued++; ring->desc_cur = (ring->desc_cur + 1) % RT_SOFTC_TX_RING_DESC_COUNT; ring->data_queued++; ring->data_cur = (ring->data_cur + 1) % RT_SOFTC_TX_RING_DATA_COUNT; /* kick Tx */ RT_WRITE(sc, sc->tx_ctx_idx[qid], ring->desc_cur); return (0); } /* * rt_start - start Transmit/Receive */ static void rt_start(struct ifnet *ifp) { struct rt_softc *sc; struct mbuf *m; int qid = 0 /* XXX must check QoS priority */; sc = ifp->if_softc; if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) return; for (;;) { IFQ_DRV_DEQUEUE(&ifp->if_snd, m); if (m == NULL) break; m->m_pkthdr.rcvif = NULL; RT_SOFTC_TX_RING_LOCK(&sc->tx_ring[qid]); if (sc->tx_ring[qid].data_queued >= RT_SOFTC_TX_RING_DATA_COUNT) { RT_SOFTC_TX_RING_UNLOCK(&sc->tx_ring[qid]); RT_DPRINTF(sc, RT_DEBUG_TX, "if_start: Tx ring with qid=%d is full\n", qid); m_freem(m); ifp->if_drv_flags |= IFF_DRV_OACTIVE; if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); sc->tx_data_queue_full[qid]++; break; } if (rt_tx_data(sc, m, qid) != 0) { RT_SOFTC_TX_RING_UNLOCK(&sc->tx_ring[qid]); if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); break; } RT_SOFTC_TX_RING_UNLOCK(&sc->tx_ring[qid]); sc->tx_timer = RT_TX_WATCHDOG_TIMEOUT; callout_reset(&sc->tx_watchdog_ch, hz, rt_tx_watchdog, sc); } } /* * rt_update_promisc - set/clear promiscuous mode. Unused yet, because * filtering done by attached Ethernet switch. */ static void rt_update_promisc(struct ifnet *ifp) { struct rt_softc *sc; sc = ifp->if_softc; printf("%s: %s promiscuous mode\n", device_get_nameunit(sc->dev), (ifp->if_flags & IFF_PROMISC) ? "entering" : "leaving"); } /* * rt_ioctl - ioctl handler. */ static int rt_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data) { struct rt_softc *sc; struct ifreq *ifr; #ifdef IF_RT_PHY_SUPPORT struct mii_data *mii; #endif /* IF_RT_PHY_SUPPORT */ int error, startall; sc = ifp->if_softc; ifr = (struct ifreq *) data; error = 0; switch (cmd) { case SIOCSIFFLAGS: startall = 0; RT_SOFTC_LOCK(sc); if (ifp->if_flags & IFF_UP) { if (ifp->if_drv_flags & IFF_DRV_RUNNING) { if ((ifp->if_flags ^ sc->if_flags) & IFF_PROMISC) rt_update_promisc(ifp); } else { rt_init_locked(sc); startall = 1; } } else { if (ifp->if_drv_flags & IFF_DRV_RUNNING) rt_stop_locked(sc); } sc->if_flags = ifp->if_flags; RT_SOFTC_UNLOCK(sc); break; case SIOCGIFMEDIA: case SIOCSIFMEDIA: #ifdef IF_RT_PHY_SUPPORT mii = device_get_softc(sc->rt_miibus); error = ifmedia_ioctl(ifp, ifr, &mii->mii_media, cmd); #else error = ifmedia_ioctl(ifp, ifr, &sc->rt_ifmedia, cmd); #endif /* IF_RT_PHY_SUPPORT */ break; default: error = ether_ioctl(ifp, cmd, data); break; } return (error); } /* * rt_periodic - Handler of PERIODIC interrupt */ static void rt_periodic(void *arg) { struct rt_softc *sc; sc = arg; RT_DPRINTF(sc, RT_DEBUG_PERIODIC, "periodic\n"); taskqueue_enqueue(sc->taskqueue, &sc->periodic_task); } /* * rt_tx_watchdog - Handler of TX Watchdog */ static void rt_tx_watchdog(void *arg) { struct rt_softc *sc; struct ifnet *ifp; sc = arg; ifp = sc->ifp; if (sc->tx_timer == 0) return; if (--sc->tx_timer == 0) { device_printf(sc->dev, "Tx watchdog timeout: resetting\n"); #ifdef notyet /* * XXX: Commented out, because reset break input. */ rt_stop_locked(sc); rt_init_locked(sc); #endif if_inc_counter(ifp, IFCOUNTER_OERRORS, 1); sc->tx_watchdog_timeouts++; } callout_reset(&sc->tx_watchdog_ch, hz, rt_tx_watchdog, sc); } /* * rt_cnt_ppe_af - Handler of PPE Counter Table Almost Full interrupt */ static void rt_cnt_ppe_af(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "PPE Counter Table Almost Full\n"); } /* * rt_cnt_gdm_af - Handler of GDMA 1 & 2 Counter Table Almost Full interrupt */ static void rt_cnt_gdm_af(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "GDMA 1 & 2 Counter Table Almost Full\n"); } /* * rt_pse_p2_fc - Handler of PSE port2 (GDMA 2) flow control interrupt */ static void rt_pse_p2_fc(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "PSE port2 (GDMA 2) flow control asserted.\n"); } /* * rt_gdm_crc_drop - Handler of GDMA 1/2 discard a packet due to CRC error * interrupt */ static void rt_gdm_crc_drop(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "GDMA 1 & 2 discard a packet due to CRC error\n"); } /* * rt_pse_buf_drop - Handler of buffer sharing limitation interrupt */ static void rt_pse_buf_drop(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "PSE discards a packet due to buffer sharing limitation\n"); } /* * rt_gdm_other_drop - Handler of discard on other reason interrupt */ static void rt_gdm_other_drop(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "GDMA 1 & 2 discard a packet due to other reason\n"); } /* * rt_pse_p1_fc - Handler of PSE port1 (GDMA 1) flow control interrupt */ static void rt_pse_p1_fc(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "PSE port1 (GDMA 1) flow control asserted.\n"); } /* * rt_pse_p0_fc - Handler of PSE port0 (CDMA) flow control interrupt */ static void rt_pse_p0_fc(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "PSE port0 (CDMA) flow control asserted.\n"); } /* * rt_pse_fq_empty - Handler of PSE free Q empty threshold reached interrupt */ static void rt_pse_fq_empty(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "PSE free Q empty threshold reached & forced drop " "condition occurred.\n"); } /* * rt_intr - main ISR */ static void rt_intr(void *arg) { struct rt_softc *sc; struct ifnet *ifp; uint32_t status; sc = arg; ifp = sc->ifp; /* acknowledge interrupts */ status = RT_READ(sc, sc->fe_int_status); RT_WRITE(sc, sc->fe_int_status, status); RT_DPRINTF(sc, RT_DEBUG_INTR, "interrupt: status=0x%08x\n", status); if (status == 0xffffffff || /* device likely went away */ status == 0) /* not for us */ return; sc->interrupts++; if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) return; if (status & CNT_PPE_AF) rt_cnt_ppe_af(sc); if (status & CNT_GDM_AF) rt_cnt_gdm_af(sc); if (status & PSE_P2_FC) rt_pse_p2_fc(sc); if (status & GDM_CRC_DROP) rt_gdm_crc_drop(sc); if (status & PSE_BUF_DROP) rt_pse_buf_drop(sc); if (status & GDM_OTHER_DROP) rt_gdm_other_drop(sc); if (status & PSE_P1_FC) rt_pse_p1_fc(sc); if (status & PSE_P0_FC) rt_pse_p0_fc(sc); if (status & PSE_FQ_EMPTY) rt_pse_fq_empty(sc); if (status & INT_TX_COHERENT) rt_tx_coherent_intr(sc); if (status & INT_RX_COHERENT) rt_rx_coherent_intr(sc); if (status & RX_DLY_INT) rt_rx_delay_intr(sc); if (status & TX_DLY_INT) rt_tx_delay_intr(sc); if (status & INT_RX_DONE) rt_rx_intr(sc, 0); if (status & INT_TXQ3_DONE) rt_tx_intr(sc, 3); if (status & INT_TXQ2_DONE) rt_tx_intr(sc, 2); if (status & INT_TXQ1_DONE) rt_tx_intr(sc, 1); if (status & INT_TXQ0_DONE) rt_tx_intr(sc, 0); } /* * rt_rt5350_intr - main ISR for Ralink 5350 SoC */ static void rt_rt5350_intr(void *arg) { struct rt_softc *sc; struct ifnet *ifp; uint32_t status; sc = arg; ifp = sc->ifp; /* acknowledge interrupts */ status = RT_READ(sc, sc->fe_int_status); RT_WRITE(sc, sc->fe_int_status, status); RT_DPRINTF(sc, RT_DEBUG_INTR, "interrupt: status=0x%08x\n", status); if (status == 0xffffffff || /* device likely went away */ status == 0) /* not for us */ return; sc->interrupts++; if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) return; if (status & RT5350_INT_TX_COHERENT) rt_tx_coherent_intr(sc); if (status & RT5350_INT_RX_COHERENT) rt_rx_coherent_intr(sc); if (status & RT5350_RX_DLY_INT) rt_rx_delay_intr(sc); if (status & RT5350_TX_DLY_INT) rt_tx_delay_intr(sc); if (status & RT5350_INT_RXQ1_DONE) rt_rx_intr(sc, 1); if (status & RT5350_INT_RXQ0_DONE) rt_rx_intr(sc, 0); if (status & RT5350_INT_TXQ3_DONE) rt_tx_intr(sc, 3); if (status & RT5350_INT_TXQ2_DONE) rt_tx_intr(sc, 2); if (status & RT5350_INT_TXQ1_DONE) rt_tx_intr(sc, 1); if (status & RT5350_INT_TXQ0_DONE) rt_tx_intr(sc, 0); } static void rt_tx_coherent_intr(struct rt_softc *sc) { uint32_t tmp; int i; RT_DPRINTF(sc, RT_DEBUG_INTR, "Tx coherent interrupt\n"); sc->tx_coherent_interrupts++; /* restart DMA engine */ tmp = RT_READ(sc, sc->pdma_glo_cfg); tmp &= ~(FE_TX_WB_DDONE | FE_TX_DMA_EN); RT_WRITE(sc, sc->pdma_glo_cfg, tmp); for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) rt_reset_tx_ring(sc, &sc->tx_ring[i]); for (i = 0; i < RT_SOFTC_TX_RING_COUNT; i++) { RT_WRITE(sc, sc->tx_base_ptr[i], sc->tx_ring[i].desc_phys_addr); RT_WRITE(sc, sc->tx_max_cnt[i], RT_SOFTC_TX_RING_DESC_COUNT); RT_WRITE(sc, sc->tx_ctx_idx[i], 0); } rt_txrx_enable(sc); } /* * rt_rx_coherent_intr */ static void rt_rx_coherent_intr(struct rt_softc *sc) { uint32_t tmp; int i; RT_DPRINTF(sc, RT_DEBUG_INTR, "Rx coherent interrupt\n"); sc->rx_coherent_interrupts++; /* restart DMA engine */ tmp = RT_READ(sc, sc->pdma_glo_cfg); tmp &= ~(FE_RX_DMA_EN); RT_WRITE(sc, sc->pdma_glo_cfg, tmp); /* init Rx ring */ for (i = 0; i < sc->rx_ring_count; i++) rt_reset_rx_ring(sc, &sc->rx_ring[i]); for (i = 0; i < sc->rx_ring_count; i++) { RT_WRITE(sc, sc->rx_base_ptr[i], sc->rx_ring[i].desc_phys_addr); RT_WRITE(sc, sc->rx_max_cnt[i], RT_SOFTC_RX_RING_DATA_COUNT); RT_WRITE(sc, sc->rx_calc_idx[i], RT_SOFTC_RX_RING_DATA_COUNT - 1); } rt_txrx_enable(sc); } /* * rt_rx_intr - a packet received */ static void rt_rx_intr(struct rt_softc *sc, int qid) { KASSERT(qid >= 0 && qid < sc->rx_ring_count, ("%s: Rx interrupt: invalid qid=%d\n", device_get_nameunit(sc->dev), qid)); RT_DPRINTF(sc, RT_DEBUG_INTR, "Rx interrupt\n"); sc->rx_interrupts[qid]++; RT_SOFTC_LOCK(sc); if (!(sc->intr_disable_mask & (sc->int_rx_done_mask << qid))) { rt_intr_disable(sc, (sc->int_rx_done_mask << qid)); taskqueue_enqueue(sc->taskqueue, &sc->rx_done_task); } sc->intr_pending_mask |= (sc->int_rx_done_mask << qid); RT_SOFTC_UNLOCK(sc); } static void rt_rx_delay_intr(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "Rx delay interrupt\n"); sc->rx_delay_interrupts++; } static void rt_tx_delay_intr(struct rt_softc *sc) { RT_DPRINTF(sc, RT_DEBUG_INTR, "Tx delay interrupt\n"); sc->tx_delay_interrupts++; } /* * rt_tx_intr - Transsmition of packet done */ static void rt_tx_intr(struct rt_softc *sc, int qid) { KASSERT(qid >= 0 && qid < RT_SOFTC_TX_RING_COUNT, ("%s: Tx interrupt: invalid qid=%d\n", device_get_nameunit(sc->dev), qid)); RT_DPRINTF(sc, RT_DEBUG_INTR, "Tx interrupt: qid=%d\n", qid); sc->tx_interrupts[qid]++; RT_SOFTC_LOCK(sc); if (!(sc->intr_disable_mask & (sc->int_tx_done_mask << qid))) { rt_intr_disable(sc, (sc->int_tx_done_mask << qid)); taskqueue_enqueue(sc->taskqueue, &sc->tx_done_task); } sc->intr_pending_mask |= (sc->int_tx_done_mask << qid); RT_SOFTC_UNLOCK(sc); } /* * rt_rx_done_task - run RX task */ static void rt_rx_done_task(void *context, int pending) { struct rt_softc *sc; struct ifnet *ifp; int again; sc = context; ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_RX, "Rx done task\n"); if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) return; sc->intr_pending_mask &= ~sc->int_rx_done_mask; again = rt_rx_eof(sc, &sc->rx_ring[0], sc->rx_process_limit); RT_SOFTC_LOCK(sc); if ((sc->intr_pending_mask & sc->int_rx_done_mask) || again) { RT_DPRINTF(sc, RT_DEBUG_RX, "Rx done task: scheduling again\n"); taskqueue_enqueue(sc->taskqueue, &sc->rx_done_task); } else { rt_intr_enable(sc, sc->int_rx_done_mask); } RT_SOFTC_UNLOCK(sc); } /* * rt_tx_done_task - check for pending TX task in all queues */ static void rt_tx_done_task(void *context, int pending) { struct rt_softc *sc; struct ifnet *ifp; uint32_t intr_mask; int i; sc = context; ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_TX, "Tx done task\n"); if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) return; for (i = RT_SOFTC_TX_RING_COUNT - 1; i >= 0; i--) { if (sc->intr_pending_mask & (sc->int_tx_done_mask << i)) { sc->intr_pending_mask &= ~(sc->int_tx_done_mask << i); rt_tx_eof(sc, &sc->tx_ring[i]); } } sc->tx_timer = 0; ifp->if_drv_flags &= ~IFF_DRV_OACTIVE; if(sc->rt_chipid == RT_CHIPID_RT5350 || sc->rt_chipid == RT_CHIPID_MT7620 || sc->rt_chipid == RT_CHIPID_MT7621) intr_mask = ( RT5350_INT_TXQ3_DONE | RT5350_INT_TXQ2_DONE | RT5350_INT_TXQ1_DONE | RT5350_INT_TXQ0_DONE); else intr_mask = ( INT_TXQ3_DONE | INT_TXQ2_DONE | INT_TXQ1_DONE | INT_TXQ0_DONE); RT_SOFTC_LOCK(sc); rt_intr_enable(sc, ~sc->intr_pending_mask & (sc->intr_disable_mask & intr_mask)); if (sc->intr_pending_mask & intr_mask) { RT_DPRINTF(sc, RT_DEBUG_TX, "Tx done task: scheduling again\n"); taskqueue_enqueue(sc->taskqueue, &sc->tx_done_task); } RT_SOFTC_UNLOCK(sc); if (!IFQ_IS_EMPTY(&ifp->if_snd)) rt_start(ifp); } /* * rt_periodic_task - run periodic task */ static void rt_periodic_task(void *context, int pending) { struct rt_softc *sc; struct ifnet *ifp; sc = context; ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_PERIODIC, "periodic task: round=%lu\n", sc->periodic_round); if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) return; RT_SOFTC_LOCK(sc); sc->periodic_round++; rt_update_stats(sc); if ((sc->periodic_round % 10) == 0) { rt_update_raw_counters(sc); rt_watchdog(sc); } RT_SOFTC_UNLOCK(sc); callout_reset(&sc->periodic_ch, hz / 10, rt_periodic, sc); } /* * rt_rx_eof - check for frames that done by DMA engine and pass it into * network subsystem. */ static int rt_rx_eof(struct rt_softc *sc, struct rt_softc_rx_ring *ring, int limit) { struct ifnet *ifp; /* struct rt_softc_rx_ring *ring; */ struct rt_rxdesc *desc; struct rt_softc_rx_data *data; struct mbuf *m, *mnew; bus_dma_segment_t segs[1]; bus_dmamap_t dma_map; uint32_t index, desc_flags; int error, nsegs, len, nframes; ifp = sc->ifp; /* ring = &sc->rx_ring[0]; */ nframes = 0; while (limit != 0) { index = RT_READ(sc, sc->rx_drx_idx[0]); if (ring->cur == index) break; desc = &ring->desc[ring->cur]; data = &ring->data[ring->cur]; bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); #ifdef IF_RT_DEBUG if ( sc->debug & RT_DEBUG_RX ) { printf("\nRX Descriptor[%#08x] dump:\n", (u_int)desc); hexdump(desc, 16, 0, 0); printf("-----------------------------------\n"); } #endif /* XXX Sometime device don`t set DDONE bit */ #ifdef DDONE_FIXED if (!(desc->sdl0 & htole16(RT_RXDESC_SDL0_DDONE))) { RT_DPRINTF(sc, RT_DEBUG_RX, "DDONE=0, try next\n"); break; } #endif len = le16toh(desc->sdl0) & 0x3fff; RT_DPRINTF(sc, RT_DEBUG_RX, "new frame len=%d\n", len); nframes++; mnew = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUMPAGESIZE); if (mnew == NULL) { sc->rx_mbuf_alloc_errors++; if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); goto skip; } mnew->m_len = mnew->m_pkthdr.len = MJUMPAGESIZE; error = bus_dmamap_load_mbuf_sg(ring->data_dma_tag, ring->spare_dma_map, mnew, segs, &nsegs, BUS_DMA_NOWAIT); if (error != 0) { RT_DPRINTF(sc, RT_DEBUG_RX, "could not load Rx mbuf DMA map: " "error=%d, nsegs=%d\n", error, nsegs); m_freem(mnew); sc->rx_mbuf_dmamap_errors++; if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); goto skip; } KASSERT(nsegs == 1, ("%s: too many DMA segments", device_get_nameunit(sc->dev))); bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_POSTREAD); bus_dmamap_unload(ring->data_dma_tag, data->dma_map); dma_map = data->dma_map; data->dma_map = ring->spare_dma_map; ring->spare_dma_map = dma_map; bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_PREREAD); m = data->m; desc_flags = desc->word3; data->m = mnew; /* Add 2 for proper align of RX IP header */ desc->sdp0 = htole32(segs[0].ds_addr+2); desc->sdl0 = htole32(segs[0].ds_len-2); desc->word3 = 0; RT_DPRINTF(sc, RT_DEBUG_RX, "Rx frame: rxdesc flags=0x%08x\n", desc_flags); m->m_pkthdr.rcvif = ifp; /* Add 2 to fix data align, after sdp0 = addr + 2 */ m->m_data += 2; m->m_pkthdr.len = m->m_len = len; /* check for crc errors */ if ((ifp->if_capenable & IFCAP_RXCSUM) != 0) { /*check for valid checksum*/ if (desc_flags & (sc->csum_fail_ip|sc->csum_fail_l4)) { RT_DPRINTF(sc, RT_DEBUG_RX, "rxdesc: crc error\n"); if_inc_counter(ifp, IFCOUNTER_IERRORS, 1); if (!(ifp->if_flags & IFF_PROMISC)) { m_freem(m); goto skip; } } if ((desc_flags & sc->csum_fail_ip) == 0) { m->m_pkthdr.csum_flags |= CSUM_IP_CHECKED; m->m_pkthdr.csum_flags |= CSUM_IP_VALID; m->m_pkthdr.csum_data = 0xffff; } m->m_flags &= ~M_HASFCS; } (*ifp->if_input)(ifp, m); skip: desc->sdl0 &= ~htole16(RT_RXDESC_SDL0_DDONE); bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); ring->cur = (ring->cur + 1) % RT_SOFTC_RX_RING_DATA_COUNT; limit--; } if (ring->cur == 0) RT_WRITE(sc, sc->rx_calc_idx[0], RT_SOFTC_RX_RING_DATA_COUNT - 1); else RT_WRITE(sc, sc->rx_calc_idx[0], ring->cur - 1); RT_DPRINTF(sc, RT_DEBUG_RX, "Rx eof: nframes=%d\n", nframes); sc->rx_packets += nframes; return (limit == 0); } /* * rt_tx_eof - check for successful transmitted frames and mark their * descriptor as free. */ static void rt_tx_eof(struct rt_softc *sc, struct rt_softc_tx_ring *ring) { struct ifnet *ifp; struct rt_txdesc *desc; struct rt_softc_tx_data *data; uint32_t index; int ndescs, nframes; ifp = sc->ifp; ndescs = 0; nframes = 0; for (;;) { index = RT_READ(sc, sc->tx_dtx_idx[ring->qid]); if (ring->desc_next == index) break; ndescs++; desc = &ring->desc[ring->desc_next]; bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE); if (desc->sdl0 & htole16(RT_TXDESC_SDL0_LASTSEG) || desc->sdl1 & htole16(RT_TXDESC_SDL1_LASTSEG)) { nframes++; data = &ring->data[ring->data_next]; bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->data_dma_tag, data->dma_map); m_freem(data->m); data->m = NULL; if_inc_counter(ifp, IFCOUNTER_OPACKETS, 1); RT_SOFTC_TX_RING_LOCK(ring); ring->data_queued--; ring->data_next = (ring->data_next + 1) % RT_SOFTC_TX_RING_DATA_COUNT; RT_SOFTC_TX_RING_UNLOCK(ring); } desc->sdl0 &= ~htole16(RT_TXDESC_SDL0_DDONE); bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); RT_SOFTC_TX_RING_LOCK(ring); ring->desc_queued--; ring->desc_next = (ring->desc_next + 1) % RT_SOFTC_TX_RING_DESC_COUNT; RT_SOFTC_TX_RING_UNLOCK(ring); } RT_DPRINTF(sc, RT_DEBUG_TX, "Tx eof: qid=%d, ndescs=%d, nframes=%d\n", ring->qid, ndescs, nframes); } /* * rt_update_stats - query statistics counters and update related variables. */ static void rt_update_stats(struct rt_softc *sc) { struct ifnet *ifp; ifp = sc->ifp; RT_DPRINTF(sc, RT_DEBUG_STATS, "update statistic: \n"); /* XXX do update stats here */ } /* * rt_watchdog - reinit device on watchdog event. */ static void rt_watchdog(struct rt_softc *sc) { uint32_t tmp; #ifdef notyet int ntries; #endif if(sc->rt_chipid != RT_CHIPID_RT5350 && sc->rt_chipid != RT_CHIPID_MT7620 && sc->rt_chipid != RT_CHIPID_MT7621) { tmp = RT_READ(sc, PSE_BASE + CDMA_OQ_STA); RT_DPRINTF(sc, RT_DEBUG_WATCHDOG, "watchdog: PSE_IQ_STA=0x%08x\n", tmp); } /* XXX: do not reset */ #ifdef notyet if (((tmp >> P0_IQ_PCNT_SHIFT) & 0xff) != 0) { sc->tx_queue_not_empty[0]++; for (ntries = 0; ntries < 10; ntries++) { tmp = RT_READ(sc, PSE_BASE + PSE_IQ_STA); if (((tmp >> P0_IQ_PCNT_SHIFT) & 0xff) == 0) break; DELAY(1); } } if (((tmp >> P1_IQ_PCNT_SHIFT) & 0xff) != 0) { sc->tx_queue_not_empty[1]++; for (ntries = 0; ntries < 10; ntries++) { tmp = RT_READ(sc, PSE_BASE + PSE_IQ_STA); if (((tmp >> P1_IQ_PCNT_SHIFT) & 0xff) == 0) break; DELAY(1); } } #endif } /* * rt_update_raw_counters - update counters. */ static void rt_update_raw_counters(struct rt_softc *sc) { sc->tx_bytes += RT_READ(sc, CNTR_BASE + GDMA_TX_GBCNT0); sc->tx_packets += RT_READ(sc, CNTR_BASE + GDMA_TX_GPCNT0); sc->tx_skip += RT_READ(sc, CNTR_BASE + GDMA_TX_SKIPCNT0); sc->tx_collision+= RT_READ(sc, CNTR_BASE + GDMA_TX_COLCNT0); sc->rx_bytes += RT_READ(sc, CNTR_BASE + GDMA_RX_GBCNT0); sc->rx_packets += RT_READ(sc, CNTR_BASE + GDMA_RX_GPCNT0); sc->rx_crc_err += RT_READ(sc, CNTR_BASE + GDMA_RX_CSUM_ERCNT0); sc->rx_short_err+= RT_READ(sc, CNTR_BASE + GDMA_RX_SHORT_ERCNT0); sc->rx_long_err += RT_READ(sc, CNTR_BASE + GDMA_RX_LONG_ERCNT0); sc->rx_phy_err += RT_READ(sc, CNTR_BASE + GDMA_RX_FERCNT0); sc->rx_fifo_overflows+= RT_READ(sc, CNTR_BASE + GDMA_RX_OERCNT0); } static void rt_intr_enable(struct rt_softc *sc, uint32_t intr_mask) { uint32_t tmp; sc->intr_disable_mask &= ~intr_mask; tmp = sc->intr_enable_mask & ~sc->intr_disable_mask; RT_WRITE(sc, sc->fe_int_enable, tmp); } static void rt_intr_disable(struct rt_softc *sc, uint32_t intr_mask) { uint32_t tmp; sc->intr_disable_mask |= intr_mask; tmp = sc->intr_enable_mask & ~sc->intr_disable_mask; RT_WRITE(sc, sc->fe_int_enable, tmp); } /* * rt_txrx_enable - enable TX/RX DMA */ static int rt_txrx_enable(struct rt_softc *sc) { struct ifnet *ifp; uint32_t tmp; int ntries; ifp = sc->ifp; /* enable Tx/Rx DMA engine */ for (ntries = 0; ntries < 200; ntries++) { tmp = RT_READ(sc, sc->pdma_glo_cfg); if (!(tmp & (FE_TX_DMA_BUSY | FE_RX_DMA_BUSY))) break; DELAY(1000); } if (ntries == 200) { device_printf(sc->dev, "timeout waiting for DMA engine\n"); return (-1); } DELAY(50); tmp |= FE_TX_WB_DDONE | FE_RX_DMA_EN | FE_TX_DMA_EN; RT_WRITE(sc, sc->pdma_glo_cfg, tmp); /* XXX set Rx filter */ return (0); } /* * rt_alloc_rx_ring - allocate RX DMA ring buffer */ static int rt_alloc_rx_ring(struct rt_softc *sc, struct rt_softc_rx_ring *ring, int qid) { struct rt_rxdesc *desc; struct rt_softc_rx_data *data; bus_dma_segment_t segs[1]; int i, nsegs, error; error = bus_dma_tag_create(bus_get_dma_tag(sc->dev), PAGE_SIZE, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, RT_SOFTC_RX_RING_DATA_COUNT * sizeof(struct rt_rxdesc), 1, RT_SOFTC_RX_RING_DATA_COUNT * sizeof(struct rt_rxdesc), 0, NULL, NULL, &ring->desc_dma_tag); if (error != 0) { device_printf(sc->dev, "could not create Rx desc DMA tag\n"); goto fail; } error = bus_dmamem_alloc(ring->desc_dma_tag, (void **) &ring->desc, BUS_DMA_NOWAIT | BUS_DMA_ZERO, &ring->desc_dma_map); if (error != 0) { device_printf(sc->dev, "could not allocate Rx desc DMA memory\n"); goto fail; } error = bus_dmamap_load(ring->desc_dma_tag, ring->desc_dma_map, ring->desc, RT_SOFTC_RX_RING_DATA_COUNT * sizeof(struct rt_rxdesc), rt_dma_map_addr, &ring->desc_phys_addr, 0); if (error != 0) { device_printf(sc->dev, "could not load Rx desc DMA map\n"); goto fail; } error = bus_dma_tag_create(bus_get_dma_tag(sc->dev), PAGE_SIZE, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, MJUMPAGESIZE, 1, MJUMPAGESIZE, 0, NULL, NULL, &ring->data_dma_tag); if (error != 0) { device_printf(sc->dev, "could not create Rx data DMA tag\n"); goto fail; } for (i = 0; i < RT_SOFTC_RX_RING_DATA_COUNT; i++) { desc = &ring->desc[i]; data = &ring->data[i]; error = bus_dmamap_create(ring->data_dma_tag, 0, &data->dma_map); if (error != 0) { device_printf(sc->dev, "could not create Rx data DMA " "map\n"); goto fail; } data->m = m_getjcl(M_NOWAIT, MT_DATA, M_PKTHDR, MJUMPAGESIZE); if (data->m == NULL) { device_printf(sc->dev, "could not allocate Rx mbuf\n"); error = ENOMEM; goto fail; } data->m->m_len = data->m->m_pkthdr.len = MJUMPAGESIZE; error = bus_dmamap_load_mbuf_sg(ring->data_dma_tag, data->dma_map, data->m, segs, &nsegs, BUS_DMA_NOWAIT); if (error != 0) { device_printf(sc->dev, "could not load Rx mbuf DMA map\n"); goto fail; } KASSERT(nsegs == 1, ("%s: too many DMA segments", device_get_nameunit(sc->dev))); /* Add 2 for proper align of RX IP header */ desc->sdp0 = htole32(segs[0].ds_addr+2); desc->sdl0 = htole32(segs[0].ds_len-2); } error = bus_dmamap_create(ring->data_dma_tag, 0, &ring->spare_dma_map); if (error != 0) { device_printf(sc->dev, "could not create Rx spare DMA map\n"); goto fail; } bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); ring->qid = qid; return (0); fail: rt_free_rx_ring(sc, ring); return (error); } /* * rt_reset_rx_ring - reset RX ring buffer */ static void rt_reset_rx_ring(struct rt_softc *sc, struct rt_softc_rx_ring *ring) { struct rt_rxdesc *desc; int i; for (i = 0; i < RT_SOFTC_RX_RING_DATA_COUNT; i++) { desc = &ring->desc[i]; desc->sdl0 &= ~htole16(RT_RXDESC_SDL0_DDONE); } bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE); ring->cur = 0; } /* * rt_free_rx_ring - free memory used by RX ring buffer */ static void rt_free_rx_ring(struct rt_softc *sc, struct rt_softc_rx_ring *ring) { struct rt_softc_rx_data *data; int i; if (ring->desc != NULL) { bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->desc_dma_tag, ring->desc_dma_map); bus_dmamem_free(ring->desc_dma_tag, ring->desc, ring->desc_dma_map); } if (ring->desc_dma_tag != NULL) bus_dma_tag_destroy(ring->desc_dma_tag); for (i = 0; i < RT_SOFTC_RX_RING_DATA_COUNT; i++) { data = &ring->data[i]; if (data->m != NULL) { bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_POSTREAD); bus_dmamap_unload(ring->data_dma_tag, data->dma_map); m_freem(data->m); } if (data->dma_map != NULL) bus_dmamap_destroy(ring->data_dma_tag, data->dma_map); } if (ring->spare_dma_map != NULL) bus_dmamap_destroy(ring->data_dma_tag, ring->spare_dma_map); if (ring->data_dma_tag != NULL) bus_dma_tag_destroy(ring->data_dma_tag); } /* * rt_alloc_tx_ring - allocate TX ring buffer */ static int rt_alloc_tx_ring(struct rt_softc *sc, struct rt_softc_tx_ring *ring, int qid) { struct rt_softc_tx_data *data; int error, i; mtx_init(&ring->lock, device_get_nameunit(sc->dev), NULL, MTX_DEF); error = bus_dma_tag_create(bus_get_dma_tag(sc->dev), PAGE_SIZE, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, RT_SOFTC_TX_RING_DESC_COUNT * sizeof(struct rt_txdesc), 1, RT_SOFTC_TX_RING_DESC_COUNT * sizeof(struct rt_txdesc), 0, NULL, NULL, &ring->desc_dma_tag); if (error != 0) { device_printf(sc->dev, "could not create Tx desc DMA tag\n"); goto fail; } error = bus_dmamem_alloc(ring->desc_dma_tag, (void **) &ring->desc, BUS_DMA_NOWAIT | BUS_DMA_ZERO, &ring->desc_dma_map); if (error != 0) { device_printf(sc->dev, "could not allocate Tx desc DMA memory\n"); goto fail; } error = bus_dmamap_load(ring->desc_dma_tag, ring->desc_dma_map, ring->desc, (RT_SOFTC_TX_RING_DESC_COUNT * sizeof(struct rt_txdesc)), rt_dma_map_addr, &ring->desc_phys_addr, 0); if (error != 0) { device_printf(sc->dev, "could not load Tx desc DMA map\n"); goto fail; } ring->desc_queued = 0; ring->desc_cur = 0; ring->desc_next = 0; error = bus_dma_tag_create(bus_get_dma_tag(sc->dev), PAGE_SIZE, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, RT_SOFTC_TX_RING_DATA_COUNT * RT_TX_DATA_SEG0_SIZE, 1, RT_SOFTC_TX_RING_DATA_COUNT * RT_TX_DATA_SEG0_SIZE, 0, NULL, NULL, &ring->seg0_dma_tag); if (error != 0) { device_printf(sc->dev, "could not create Tx seg0 DMA tag\n"); goto fail; } error = bus_dmamem_alloc(ring->seg0_dma_tag, (void **) &ring->seg0, BUS_DMA_NOWAIT | BUS_DMA_ZERO, &ring->seg0_dma_map); if (error != 0) { device_printf(sc->dev, "could not allocate Tx seg0 DMA memory\n"); goto fail; } error = bus_dmamap_load(ring->seg0_dma_tag, ring->seg0_dma_map, ring->seg0, RT_SOFTC_TX_RING_DATA_COUNT * RT_TX_DATA_SEG0_SIZE, rt_dma_map_addr, &ring->seg0_phys_addr, 0); if (error != 0) { device_printf(sc->dev, "could not load Tx seg0 DMA map\n"); goto fail; } error = bus_dma_tag_create(bus_get_dma_tag(sc->dev), PAGE_SIZE, 0, BUS_SPACE_MAXADDR_32BIT, BUS_SPACE_MAXADDR, NULL, NULL, MJUMPAGESIZE, RT_SOFTC_MAX_SCATTER, MJUMPAGESIZE, 0, NULL, NULL, &ring->data_dma_tag); if (error != 0) { device_printf(sc->dev, "could not create Tx data DMA tag\n"); goto fail; } for (i = 0; i < RT_SOFTC_TX_RING_DATA_COUNT; i++) { data = &ring->data[i]; error = bus_dmamap_create(ring->data_dma_tag, 0, &data->dma_map); if (error != 0) { device_printf(sc->dev, "could not create Tx data DMA " "map\n"); goto fail; } } ring->data_queued = 0; ring->data_cur = 0; ring->data_next = 0; ring->qid = qid; return (0); fail: rt_free_tx_ring(sc, ring); return (error); } /* * rt_reset_tx_ring - reset TX ring buffer to empty state */ static void rt_reset_tx_ring(struct rt_softc *sc, struct rt_softc_tx_ring *ring) { struct rt_softc_tx_data *data; struct rt_txdesc *desc; int i; for (i = 0; i < RT_SOFTC_TX_RING_DESC_COUNT; i++) { desc = &ring->desc[i]; desc->sdl0 = 0; desc->sdl1 = 0; } ring->desc_queued = 0; ring->desc_cur = 0; ring->desc_next = 0; bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_PREWRITE); bus_dmamap_sync(ring->seg0_dma_tag, ring->seg0_dma_map, BUS_DMASYNC_PREWRITE); for (i = 0; i < RT_SOFTC_TX_RING_DATA_COUNT; i++) { data = &ring->data[i]; if (data->m != NULL) { bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->data_dma_tag, data->dma_map); m_freem(data->m); data->m = NULL; } } ring->data_queued = 0; ring->data_cur = 0; ring->data_next = 0; } /* * rt_free_tx_ring - free RX ring buffer */ static void rt_free_tx_ring(struct rt_softc *sc, struct rt_softc_tx_ring *ring) { struct rt_softc_tx_data *data; int i; if (ring->desc != NULL) { bus_dmamap_sync(ring->desc_dma_tag, ring->desc_dma_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->desc_dma_tag, ring->desc_dma_map); bus_dmamem_free(ring->desc_dma_tag, ring->desc, ring->desc_dma_map); } if (ring->desc_dma_tag != NULL) bus_dma_tag_destroy(ring->desc_dma_tag); if (ring->seg0 != NULL) { bus_dmamap_sync(ring->seg0_dma_tag, ring->seg0_dma_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->seg0_dma_tag, ring->seg0_dma_map); bus_dmamem_free(ring->seg0_dma_tag, ring->seg0, ring->seg0_dma_map); } if (ring->seg0_dma_tag != NULL) bus_dma_tag_destroy(ring->seg0_dma_tag); for (i = 0; i < RT_SOFTC_TX_RING_DATA_COUNT; i++) { data = &ring->data[i]; if (data->m != NULL) { bus_dmamap_sync(ring->data_dma_tag, data->dma_map, BUS_DMASYNC_POSTWRITE); bus_dmamap_unload(ring->data_dma_tag, data->dma_map); m_freem(data->m); } if (data->dma_map != NULL) bus_dmamap_destroy(ring->data_dma_tag, data->dma_map); } if (ring->data_dma_tag != NULL) bus_dma_tag_destroy(ring->data_dma_tag); mtx_destroy(&ring->lock); } /* * rt_dma_map_addr - get address of busdma segment */ static void rt_dma_map_addr(void *arg, bus_dma_segment_t *segs, int nseg, int error) { if (error != 0) return; KASSERT(nseg == 1, ("too many DMA segments, %d should be 1", nseg)); *(bus_addr_t *) arg = segs[0].ds_addr; } /* * rt_sysctl_attach - attach sysctl nodes for NIC counters. */ static void rt_sysctl_attach(struct rt_softc *sc) { struct sysctl_ctx_list *ctx; struct sysctl_oid *tree; struct sysctl_oid *stats; ctx = device_get_sysctl_ctx(sc->dev); tree = device_get_sysctl_tree(sc->dev); /* statistic counters */ stats = SYSCTL_ADD_NODE(ctx, SYSCTL_CHILDREN(tree), OID_AUTO, "stats", CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "statistic"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "interrupts", CTLFLAG_RD, &sc->interrupts, "all interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_coherent_interrupts", CTLFLAG_RD, &sc->tx_coherent_interrupts, "Tx coherent interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_coherent_interrupts", CTLFLAG_RD, &sc->rx_coherent_interrupts, "Rx coherent interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_interrupts", CTLFLAG_RD, &sc->rx_interrupts[0], "Rx interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_delay_interrupts", CTLFLAG_RD, &sc->rx_delay_interrupts, "Rx delay interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ3_interrupts", CTLFLAG_RD, &sc->tx_interrupts[3], "Tx AC3 interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ2_interrupts", CTLFLAG_RD, &sc->tx_interrupts[2], "Tx AC2 interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ1_interrupts", CTLFLAG_RD, &sc->tx_interrupts[1], "Tx AC1 interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ0_interrupts", CTLFLAG_RD, &sc->tx_interrupts[0], "Tx AC0 interrupts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_delay_interrupts", CTLFLAG_RD, &sc->tx_delay_interrupts, "Tx delay interrupts"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ3_desc_queued", CTLFLAG_RD, &sc->tx_ring[3].desc_queued, 0, "Tx AC3 descriptors queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ3_data_queued", CTLFLAG_RD, &sc->tx_ring[3].data_queued, 0, "Tx AC3 data queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ2_desc_queued", CTLFLAG_RD, &sc->tx_ring[2].desc_queued, 0, "Tx AC2 descriptors queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ2_data_queued", CTLFLAG_RD, &sc->tx_ring[2].data_queued, 0, "Tx AC2 data queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ1_desc_queued", CTLFLAG_RD, &sc->tx_ring[1].desc_queued, 0, "Tx AC1 descriptors queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ1_data_queued", CTLFLAG_RD, &sc->tx_ring[1].data_queued, 0, "Tx AC1 data queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ0_desc_queued", CTLFLAG_RD, &sc->tx_ring[0].desc_queued, 0, "Tx AC0 descriptors queued"); SYSCTL_ADD_INT(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ0_data_queued", CTLFLAG_RD, &sc->tx_ring[0].data_queued, 0, "Tx AC0 data queued"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ3_data_queue_full", CTLFLAG_RD, &sc->tx_data_queue_full[3], "Tx AC3 data queue full"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ2_data_queue_full", CTLFLAG_RD, &sc->tx_data_queue_full[2], "Tx AC2 data queue full"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ1_data_queue_full", CTLFLAG_RD, &sc->tx_data_queue_full[1], "Tx AC1 data queue full"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "TXQ0_data_queue_full", CTLFLAG_RD, &sc->tx_data_queue_full[0], "Tx AC0 data queue full"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_watchdog_timeouts", CTLFLAG_RD, &sc->tx_watchdog_timeouts, "Tx watchdog timeouts"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_defrag_packets", CTLFLAG_RD, &sc->tx_defrag_packets, "Tx defragmented packets"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "no_tx_desc_avail", CTLFLAG_RD, &sc->no_tx_desc_avail, "no Tx descriptors available"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_mbuf_alloc_errors", CTLFLAG_RD, &sc->rx_mbuf_alloc_errors, "Rx mbuf allocation errors"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_mbuf_dmamap_errors", CTLFLAG_RD, &sc->rx_mbuf_dmamap_errors, "Rx mbuf DMA mapping errors"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_queue_0_not_empty", CTLFLAG_RD, &sc->tx_queue_not_empty[0], "Tx queue 0 not empty"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_queue_1_not_empty", CTLFLAG_RD, &sc->tx_queue_not_empty[1], "Tx queue 1 not empty"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_packets", CTLFLAG_RD, &sc->rx_packets, "Rx packets"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_crc_errors", CTLFLAG_RD, &sc->rx_crc_err, "Rx CRC errors"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_phy_errors", CTLFLAG_RD, &sc->rx_phy_err, "Rx PHY errors"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_dup_packets", CTLFLAG_RD, &sc->rx_dup_packets, "Rx duplicate packets"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_fifo_overflows", CTLFLAG_RD, &sc->rx_fifo_overflows, "Rx FIFO overflows"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_bytes", CTLFLAG_RD, &sc->rx_bytes, "Rx bytes"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_long_err", CTLFLAG_RD, &sc->rx_long_err, "Rx too long frame errors"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "rx_short_err", CTLFLAG_RD, &sc->rx_short_err, "Rx too short frame errors"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_bytes", CTLFLAG_RD, &sc->tx_bytes, "Tx bytes"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_packets", CTLFLAG_RD, &sc->tx_packets, "Tx packets"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_skip", CTLFLAG_RD, &sc->tx_skip, "Tx skip count for GDMA ports"); SYSCTL_ADD_ULONG(ctx, SYSCTL_CHILDREN(stats), OID_AUTO, "tx_collision", CTLFLAG_RD, &sc->tx_collision, "Tx collision count for GDMA ports"); } #if defined(IF_RT_PHY_SUPPORT) || defined(RT_MDIO) /* This code is only work RT2880 and same chip. */ /* TODO: make RT3052 and later support code. But nobody need it? */ static int rt_miibus_readreg(device_t dev, int phy, int reg) { struct rt_softc *sc = device_get_softc(dev); int dat; /* * PSEUDO_PHYAD is a special value for indicate switch attached. * No one PHY use PSEUDO_PHYAD (0x1e) address. */ #ifndef RT_MDIO if (phy == 31) { /* Fake PHY ID for bfeswitch attach */ switch (reg) { case MII_BMSR: return (BMSR_EXTSTAT|BMSR_MEDIAMASK); case MII_PHYIDR1: return (0x40); /* As result of faking */ case MII_PHYIDR2: /* PHY will detect as */ return (0x6250); /* bfeswitch */ } } #endif /* Wait prev command done if any */ while (RT_READ(sc, MDIO_ACCESS) & MDIO_CMD_ONGO); dat = ((phy << MDIO_PHY_ADDR_SHIFT) & MDIO_PHY_ADDR_MASK) | ((reg << MDIO_PHYREG_ADDR_SHIFT) & MDIO_PHYREG_ADDR_MASK); RT_WRITE(sc, MDIO_ACCESS, dat); RT_WRITE(sc, MDIO_ACCESS, dat | MDIO_CMD_ONGO); while (RT_READ(sc, MDIO_ACCESS) & MDIO_CMD_ONGO); return (RT_READ(sc, MDIO_ACCESS) & MDIO_PHY_DATA_MASK); } static int rt_miibus_writereg(device_t dev, int phy, int reg, int val) { struct rt_softc *sc = device_get_softc(dev); int dat; /* Wait prev command done if any */ while (RT_READ(sc, MDIO_ACCESS) & MDIO_CMD_ONGO); dat = MDIO_CMD_WR | ((phy << MDIO_PHY_ADDR_SHIFT) & MDIO_PHY_ADDR_MASK) | ((reg << MDIO_PHYREG_ADDR_SHIFT) & MDIO_PHYREG_ADDR_MASK) | (val & MDIO_PHY_DATA_MASK); RT_WRITE(sc, MDIO_ACCESS, dat); RT_WRITE(sc, MDIO_ACCESS, dat | MDIO_CMD_ONGO); while (RT_READ(sc, MDIO_ACCESS) & MDIO_CMD_ONGO); return (0); } #endif #ifdef IF_RT_PHY_SUPPORT void rt_miibus_statchg(device_t dev) { struct rt_softc *sc = device_get_softc(dev); struct mii_data *mii; mii = device_get_softc(sc->rt_miibus); if ((mii->mii_media_status & (IFM_ACTIVE | IFM_AVALID)) == (IFM_ACTIVE | IFM_AVALID)) { switch (IFM_SUBTYPE(mii->mii_media_active)) { case IFM_10_T: case IFM_100_TX: /* XXX check link here */ sc->flags |= 1; break; default: break; } } } #endif /* IF_RT_PHY_SUPPORT */ static device_method_t rt_dev_methods[] = { DEVMETHOD(device_probe, rt_probe), DEVMETHOD(device_attach, rt_attach), DEVMETHOD(device_detach, rt_detach), DEVMETHOD(device_shutdown, rt_shutdown), DEVMETHOD(device_suspend, rt_suspend), DEVMETHOD(device_resume, rt_resume), #ifdef IF_RT_PHY_SUPPORT /* MII interface */ DEVMETHOD(miibus_readreg, rt_miibus_readreg), DEVMETHOD(miibus_writereg, rt_miibus_writereg), DEVMETHOD(miibus_statchg, rt_miibus_statchg), #endif DEVMETHOD_END }; static driver_t rt_driver = { "rt", rt_dev_methods, sizeof(struct rt_softc) }; static devclass_t rt_dev_class; DRIVER_MODULE(rt, nexus, rt_driver, rt_dev_class, 0, 0); #ifdef FDT DRIVER_MODULE(rt, simplebus, rt_driver, rt_dev_class, 0, 0); #endif MODULE_DEPEND(rt, ether, 1, 1, 1); MODULE_DEPEND(rt, miibus, 1, 1, 1); #ifdef RT_MDIO MODULE_DEPEND(rt, mdio, 1, 1, 1); static int rtmdio_probe(device_t); static int rtmdio_attach(device_t); static int rtmdio_detach(device_t); static struct mtx miibus_mtx; MTX_SYSINIT(miibus_mtx, &miibus_mtx, "rt mii lock", MTX_DEF); /* * Declare an additional, separate driver for accessing the MDIO bus. */ static device_method_t rtmdio_methods[] = { /* Device interface */ DEVMETHOD(device_probe, rtmdio_probe), DEVMETHOD(device_attach, rtmdio_attach), DEVMETHOD(device_detach, rtmdio_detach), /* bus interface */ DEVMETHOD(bus_add_child, device_add_child_ordered), /* MDIO access */ DEVMETHOD(mdio_readreg, rt_miibus_readreg), DEVMETHOD(mdio_writereg, rt_miibus_writereg), }; DEFINE_CLASS_0(rtmdio, rtmdio_driver, rtmdio_methods, sizeof(struct rt_softc)); static devclass_t rtmdio_devclass; DRIVER_MODULE(miiproxy, rt, miiproxy_driver, miiproxy_devclass, 0, 0); DRIVER_MODULE(rtmdio, simplebus, rtmdio_driver, rtmdio_devclass, 0, 0); DRIVER_MODULE(mdio, rtmdio, mdio_driver, mdio_devclass, 0, 0); static int rtmdio_probe(device_t dev) { if (!ofw_bus_status_okay(dev)) return (ENXIO); if (!ofw_bus_is_compatible(dev, "ralink,rt2880-mdio")) return (ENXIO); device_set_desc(dev, "RT built-in ethernet interface, MDIO controller"); return(0); } static int rtmdio_attach(device_t dev) { struct rt_softc *sc; int error; sc = device_get_softc(dev); sc->dev = dev; sc->mem_rid = 0; sc->mem = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->mem_rid, RF_ACTIVE | RF_SHAREABLE); if (sc->mem == NULL) { device_printf(dev, "couldn't map memory\n"); error = ENXIO; goto fail; } sc->bst = rman_get_bustag(sc->mem); sc->bsh = rman_get_bushandle(sc->mem); bus_generic_probe(dev); bus_enumerate_hinted_children(dev); error = bus_generic_attach(dev); fail: return(error); } static int rtmdio_detach(device_t dev) { return(0); } #endif Index: head/sys/geom/eli/g_eli.c =================================================================== --- head/sys/geom/eli/g_eli.c (revision 366507) +++ head/sys/geom/eli/g_eli.c (revision 366508) @@ -1,1464 +1,1465 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2005-2019 Pawel Jakub Dawidek * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHORS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHORS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include #include +#include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include FEATURE(geom_eli, "GEOM crypto module"); MALLOC_DEFINE(M_ELI, "eli data", "GEOM_ELI Data"); SYSCTL_DECL(_kern_geom); SYSCTL_NODE(_kern_geom, OID_AUTO, eli, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "GEOM_ELI stuff"); static int g_eli_version = G_ELI_VERSION; SYSCTL_INT(_kern_geom_eli, OID_AUTO, version, CTLFLAG_RD, &g_eli_version, 0, "GELI version"); int g_eli_debug = 0; SYSCTL_INT(_kern_geom_eli, OID_AUTO, debug, CTLFLAG_RWTUN, &g_eli_debug, 0, "Debug level"); static u_int g_eli_tries = 3; SYSCTL_UINT(_kern_geom_eli, OID_AUTO, tries, CTLFLAG_RWTUN, &g_eli_tries, 0, "Number of tries for entering the passphrase"); static u_int g_eli_visible_passphrase = GETS_NOECHO; SYSCTL_UINT(_kern_geom_eli, OID_AUTO, visible_passphrase, CTLFLAG_RWTUN, &g_eli_visible_passphrase, 0, "Visibility of passphrase prompt (0 = invisible, 1 = visible, 2 = asterisk)"); u_int g_eli_overwrites = G_ELI_OVERWRITES; SYSCTL_UINT(_kern_geom_eli, OID_AUTO, overwrites, CTLFLAG_RWTUN, &g_eli_overwrites, 0, "Number of times on-disk keys should be overwritten when destroying them"); static u_int g_eli_threads = 0; SYSCTL_UINT(_kern_geom_eli, OID_AUTO, threads, CTLFLAG_RWTUN, &g_eli_threads, 0, "Number of threads doing crypto work"); u_int g_eli_batch = 0; SYSCTL_UINT(_kern_geom_eli, OID_AUTO, batch, CTLFLAG_RWTUN, &g_eli_batch, 0, "Use crypto operations batching"); /* * Passphrase cached during boot, in order to be more user-friendly if * there are multiple providers using the same passphrase. */ static char cached_passphrase[256]; static u_int g_eli_boot_passcache = 1; TUNABLE_INT("kern.geom.eli.boot_passcache", &g_eli_boot_passcache); SYSCTL_UINT(_kern_geom_eli, OID_AUTO, boot_passcache, CTLFLAG_RD, &g_eli_boot_passcache, 0, "Passphrases are cached during boot process for possible reuse"); static void fetch_loader_passphrase(void * dummy) { char * env_passphrase; KASSERT(dynamic_kenv, ("need dynamic kenv")); if ((env_passphrase = kern_getenv("kern.geom.eli.passphrase")) != NULL) { /* Extract passphrase from the environment. */ strlcpy(cached_passphrase, env_passphrase, sizeof(cached_passphrase)); freeenv(env_passphrase); /* Wipe the passphrase from the environment. */ kern_unsetenv("kern.geom.eli.passphrase"); } } SYSINIT(geli_fetch_loader_passphrase, SI_SUB_KMEM + 1, SI_ORDER_ANY, fetch_loader_passphrase, NULL); static void zero_boot_passcache(void) { explicit_bzero(cached_passphrase, sizeof(cached_passphrase)); } static void zero_geli_intake_keys(void) { struct keybuf *keybuf; int i; if ((keybuf = get_keybuf()) != NULL) { /* Scan the key buffer, clear all GELI keys. */ for (i = 0; i < keybuf->kb_nents; i++) { if (keybuf->kb_ents[i].ke_type == KEYBUF_TYPE_GELI) { explicit_bzero(keybuf->kb_ents[i].ke_data, sizeof(keybuf->kb_ents[i].ke_data)); keybuf->kb_ents[i].ke_type = KEYBUF_TYPE_NONE; } } } } static void zero_intake_passcache(void *dummy) { zero_boot_passcache(); zero_geli_intake_keys(); } EVENTHANDLER_DEFINE(mountroot, zero_intake_passcache, NULL, 0); static eventhandler_tag g_eli_pre_sync = NULL; static int g_eli_read_metadata_offset(struct g_class *mp, struct g_provider *pp, off_t offset, struct g_eli_metadata *md); static int g_eli_destroy_geom(struct gctl_req *req, struct g_class *mp, struct g_geom *gp); static void g_eli_init(struct g_class *mp); static void g_eli_fini(struct g_class *mp); static g_taste_t g_eli_taste; static g_dumpconf_t g_eli_dumpconf; struct g_class g_eli_class = { .name = G_ELI_CLASS_NAME, .version = G_VERSION, .ctlreq = g_eli_config, .taste = g_eli_taste, .destroy_geom = g_eli_destroy_geom, .init = g_eli_init, .fini = g_eli_fini }; /* * Code paths: * BIO_READ: * g_eli_start -> g_eli_crypto_read -> g_io_request -> g_eli_read_done -> g_eli_crypto_run -> g_eli_crypto_read_done -> g_io_deliver * BIO_WRITE: * g_eli_start -> g_eli_crypto_run -> g_eli_crypto_write_done -> g_io_request -> g_eli_write_done -> g_io_deliver */ /* * EAGAIN from crypto(9) means, that we were probably balanced to another crypto * accelerator or something like this. * The function updates the SID and rerun the operation. */ int g_eli_crypto_rerun(struct cryptop *crp) { struct g_eli_softc *sc; struct g_eli_worker *wr; struct bio *bp; int error; bp = (struct bio *)crp->crp_opaque; sc = bp->bio_to->geom->softc; LIST_FOREACH(wr, &sc->sc_workers, w_next) { if (wr->w_number == bp->bio_pflags) break; } KASSERT(wr != NULL, ("Invalid worker (%u).", bp->bio_pflags)); G_ELI_DEBUG(1, "Rerunning crypto %s request (sid: %p -> %p).", bp->bio_cmd == BIO_READ ? "READ" : "WRITE", wr->w_sid, crp->crp_session); wr->w_sid = crp->crp_session; crp->crp_etype = 0; error = crypto_dispatch(crp); if (error == 0) return (0); G_ELI_DEBUG(1, "%s: crypto_dispatch() returned %d.", __func__, error); crp->crp_etype = error; return (error); } static void g_eli_getattr_done(struct bio *bp) { if (bp->bio_error == 0 && !strcmp(bp->bio_attribute, "GEOM::physpath")) { strlcat(bp->bio_data, "/eli", bp->bio_length); } g_std_done(bp); } /* * The function is called afer reading encrypted data from the provider. * * g_eli_start -> g_eli_crypto_read -> g_io_request -> G_ELI_READ_DONE -> g_eli_crypto_run -> g_eli_crypto_read_done -> g_io_deliver */ void g_eli_read_done(struct bio *bp) { struct g_eli_softc *sc; struct bio *pbp; G_ELI_LOGREQ(2, bp, "Request done."); pbp = bp->bio_parent; if (pbp->bio_error == 0 && bp->bio_error != 0) pbp->bio_error = bp->bio_error; g_destroy_bio(bp); /* * Do we have all sectors already? */ pbp->bio_inbed++; if (pbp->bio_inbed < pbp->bio_children) return; sc = pbp->bio_to->geom->softc; if (pbp->bio_error != 0) { G_ELI_LOGREQ(0, pbp, "%s() failed (error=%d)", __func__, pbp->bio_error); pbp->bio_completed = 0; if (pbp->bio_driver2 != NULL) { free(pbp->bio_driver2, M_ELI); pbp->bio_driver2 = NULL; } g_io_deliver(pbp, pbp->bio_error); if (sc != NULL) atomic_subtract_int(&sc->sc_inflight, 1); return; } mtx_lock(&sc->sc_queue_mtx); bioq_insert_tail(&sc->sc_queue, pbp); mtx_unlock(&sc->sc_queue_mtx); wakeup(sc); } /* * The function is called after we encrypt and write data. * * g_eli_start -> g_eli_crypto_run -> g_eli_crypto_write_done -> g_io_request -> G_ELI_WRITE_DONE -> g_io_deliver */ void g_eli_write_done(struct bio *bp) { struct g_eli_softc *sc; struct bio *pbp; G_ELI_LOGREQ(2, bp, "Request done."); pbp = bp->bio_parent; if (pbp->bio_error == 0 && bp->bio_error != 0) pbp->bio_error = bp->bio_error; g_destroy_bio(bp); /* * Do we have all sectors already? */ pbp->bio_inbed++; if (pbp->bio_inbed < pbp->bio_children) return; free(pbp->bio_driver2, M_ELI); pbp->bio_driver2 = NULL; if (pbp->bio_error != 0) { G_ELI_LOGREQ(0, pbp, "%s() failed (error=%d)", __func__, pbp->bio_error); pbp->bio_completed = 0; } else pbp->bio_completed = pbp->bio_length; /* * Write is finished, send it up. */ sc = pbp->bio_to->geom->softc; g_io_deliver(pbp, pbp->bio_error); if (sc != NULL) atomic_subtract_int(&sc->sc_inflight, 1); } /* * This function should never be called, but GEOM made as it set ->orphan() * method for every geom. */ static void g_eli_orphan_spoil_assert(struct g_consumer *cp) { panic("Function %s() called for %s.", __func__, cp->geom->name); } static void g_eli_orphan(struct g_consumer *cp) { struct g_eli_softc *sc; g_topology_assert(); sc = cp->geom->softc; if (sc == NULL) return; g_eli_destroy(sc, TRUE); } static void g_eli_resize(struct g_consumer *cp) { struct g_eli_softc *sc; struct g_provider *epp, *pp; off_t oldsize; g_topology_assert(); sc = cp->geom->softc; if (sc == NULL) return; if ((sc->sc_flags & G_ELI_FLAG_AUTORESIZE) == 0) { G_ELI_DEBUG(0, "Autoresize is turned off, old size: %jd.", (intmax_t)sc->sc_provsize); return; } pp = cp->provider; if ((sc->sc_flags & G_ELI_FLAG_ONETIME) == 0) { struct g_eli_metadata md; u_char *sector; int error; sector = NULL; error = g_eli_read_metadata_offset(cp->geom->class, pp, sc->sc_provsize - pp->sectorsize, &md); if (error != 0) { G_ELI_DEBUG(0, "Cannot read metadata from %s (error=%d).", pp->name, error); goto iofail; } md.md_provsize = pp->mediasize; sector = malloc(pp->sectorsize, M_ELI, M_WAITOK | M_ZERO); eli_metadata_encode(&md, sector); error = g_write_data(cp, pp->mediasize - pp->sectorsize, sector, pp->sectorsize); if (error != 0) { G_ELI_DEBUG(0, "Cannot store metadata on %s (error=%d).", pp->name, error); goto iofail; } explicit_bzero(sector, pp->sectorsize); error = g_write_data(cp, sc->sc_provsize - pp->sectorsize, sector, pp->sectorsize); if (error != 0) { G_ELI_DEBUG(0, "Cannot clear old metadata from %s (error=%d).", pp->name, error); goto iofail; } iofail: explicit_bzero(&md, sizeof(md)); zfree(sector, M_ELI); } oldsize = sc->sc_mediasize; sc->sc_mediasize = eli_mediasize(sc, pp->mediasize, pp->sectorsize); g_eli_key_resize(sc); sc->sc_provsize = pp->mediasize; epp = LIST_FIRST(&sc->sc_geom->provider); g_resize_provider(epp, sc->sc_mediasize); G_ELI_DEBUG(0, "Device %s size changed from %jd to %jd.", epp->name, (intmax_t)oldsize, (intmax_t)sc->sc_mediasize); } /* * BIO_READ: * G_ELI_START -> g_eli_crypto_read -> g_io_request -> g_eli_read_done -> g_eli_crypto_run -> g_eli_crypto_read_done -> g_io_deliver * BIO_WRITE: * G_ELI_START -> g_eli_crypto_run -> g_eli_crypto_write_done -> g_io_request -> g_eli_write_done -> g_io_deliver */ static void g_eli_start(struct bio *bp) { struct g_eli_softc *sc; struct g_consumer *cp; struct bio *cbp; sc = bp->bio_to->geom->softc; KASSERT(sc != NULL, ("Provider's error should be set (error=%d)(device=%s).", bp->bio_to->error, bp->bio_to->name)); G_ELI_LOGREQ(2, bp, "Request received."); switch (bp->bio_cmd) { case BIO_READ: case BIO_WRITE: case BIO_GETATTR: case BIO_FLUSH: case BIO_ZONE: case BIO_SPEEDUP: break; case BIO_DELETE: /* * If the user hasn't set the NODELETE flag, we just pass * it down the stack and let the layers beneath us do (or * not) whatever they do with it. If they have, we * reject it. A possible extension would be an * additional flag to take it as a hint to shred the data * with [multiple?] overwrites. */ if (!(sc->sc_flags & G_ELI_FLAG_NODELETE)) break; default: g_io_deliver(bp, EOPNOTSUPP); return; } cbp = g_clone_bio(bp); if (cbp == NULL) { g_io_deliver(bp, ENOMEM); return; } bp->bio_driver1 = cbp; bp->bio_pflags = G_ELI_NEW_BIO; switch (bp->bio_cmd) { case BIO_READ: if (!(sc->sc_flags & G_ELI_FLAG_AUTH)) { g_eli_crypto_read(sc, bp, 0); break; } /* FALLTHROUGH */ case BIO_WRITE: mtx_lock(&sc->sc_queue_mtx); bioq_insert_tail(&sc->sc_queue, bp); mtx_unlock(&sc->sc_queue_mtx); wakeup(sc); break; case BIO_GETATTR: case BIO_FLUSH: case BIO_DELETE: case BIO_SPEEDUP: case BIO_ZONE: if (bp->bio_cmd == BIO_GETATTR) cbp->bio_done = g_eli_getattr_done; else cbp->bio_done = g_std_done; cp = LIST_FIRST(&sc->sc_geom->consumer); cbp->bio_to = cp->provider; G_ELI_LOGREQ(2, cbp, "Sending request."); g_io_request(cbp, cp); break; } } static int g_eli_newsession(struct g_eli_worker *wr) { struct g_eli_softc *sc; struct crypto_session_params csp; uint32_t caps; int error, new_crypto; void *key; sc = wr->w_softc; memset(&csp, 0, sizeof(csp)); csp.csp_mode = CSP_MODE_CIPHER; csp.csp_cipher_alg = sc->sc_ealgo; csp.csp_ivlen = g_eli_ivlen(sc->sc_ealgo); csp.csp_cipher_klen = sc->sc_ekeylen / 8; if (sc->sc_ealgo == CRYPTO_AES_XTS) csp.csp_cipher_klen <<= 1; if ((sc->sc_flags & G_ELI_FLAG_FIRST_KEY) != 0) { key = g_eli_key_hold(sc, 0, LIST_FIRST(&sc->sc_geom->consumer)->provider->sectorsize); csp.csp_cipher_key = key; } else { key = NULL; csp.csp_cipher_key = sc->sc_ekey; } if (sc->sc_flags & G_ELI_FLAG_AUTH) { csp.csp_mode = CSP_MODE_ETA; csp.csp_auth_alg = sc->sc_aalgo; csp.csp_auth_klen = G_ELI_AUTH_SECKEYLEN; } switch (sc->sc_crypto) { case G_ELI_CRYPTO_SW_ACCEL: case G_ELI_CRYPTO_SW: error = crypto_newsession(&wr->w_sid, &csp, CRYPTOCAP_F_SOFTWARE); break; case G_ELI_CRYPTO_HW: error = crypto_newsession(&wr->w_sid, &csp, CRYPTOCAP_F_HARDWARE); break; case G_ELI_CRYPTO_UNKNOWN: error = crypto_newsession(&wr->w_sid, &csp, CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE); if (error == 0) { caps = crypto_ses2caps(wr->w_sid); if (caps & CRYPTOCAP_F_HARDWARE) new_crypto = G_ELI_CRYPTO_HW; else if (caps & CRYPTOCAP_F_ACCEL_SOFTWARE) new_crypto = G_ELI_CRYPTO_SW_ACCEL; else new_crypto = G_ELI_CRYPTO_SW; mtx_lock(&sc->sc_queue_mtx); if (sc->sc_crypto == G_ELI_CRYPTO_UNKNOWN) sc->sc_crypto = new_crypto; mtx_unlock(&sc->sc_queue_mtx); } break; default: panic("%s: invalid condition", __func__); } if ((sc->sc_flags & G_ELI_FLAG_FIRST_KEY) != 0) { if (error) g_eli_key_drop(sc, key); else wr->w_first_key = key; } return (error); } static void g_eli_freesession(struct g_eli_worker *wr) { struct g_eli_softc *sc; crypto_freesession(wr->w_sid); if (wr->w_first_key != NULL) { sc = wr->w_softc; g_eli_key_drop(sc, wr->w_first_key); wr->w_first_key = NULL; } } static void g_eli_cancel(struct g_eli_softc *sc) { struct bio *bp; mtx_assert(&sc->sc_queue_mtx, MA_OWNED); while ((bp = bioq_takefirst(&sc->sc_queue)) != NULL) { KASSERT(bp->bio_pflags == G_ELI_NEW_BIO, ("Not new bio when canceling (bp=%p).", bp)); g_io_deliver(bp, ENXIO); } } static struct bio * g_eli_takefirst(struct g_eli_softc *sc) { struct bio *bp; mtx_assert(&sc->sc_queue_mtx, MA_OWNED); if (!(sc->sc_flags & G_ELI_FLAG_SUSPEND)) return (bioq_takefirst(&sc->sc_queue)); /* * Device suspended, so we skip new I/O requests. */ TAILQ_FOREACH(bp, &sc->sc_queue.queue, bio_queue) { if (bp->bio_pflags != G_ELI_NEW_BIO) break; } if (bp != NULL) bioq_remove(&sc->sc_queue, bp); return (bp); } /* * This is the main function for kernel worker thread when we don't have * hardware acceleration and we have to do cryptography in software. * Dedicated thread is needed, so we don't slow down g_up/g_down GEOM * threads with crypto work. */ static void g_eli_worker(void *arg) { struct g_eli_softc *sc; struct g_eli_worker *wr; struct bio *bp; int error; wr = arg; sc = wr->w_softc; #ifdef EARLY_AP_STARTUP MPASS(!sc->sc_cpubind || smp_started); #elif defined(SMP) /* Before sched_bind() to a CPU, wait for all CPUs to go on-line. */ if (sc->sc_cpubind) { while (!smp_started) tsleep(wr, 0, "geli:smp", hz / 4); } #endif thread_lock(curthread); sched_prio(curthread, PUSER); if (sc->sc_cpubind) sched_bind(curthread, wr->w_number % mp_ncpus); thread_unlock(curthread); G_ELI_DEBUG(1, "Thread %s started.", curthread->td_proc->p_comm); for (;;) { mtx_lock(&sc->sc_queue_mtx); again: bp = g_eli_takefirst(sc); if (bp == NULL) { if (sc->sc_flags & G_ELI_FLAG_DESTROY) { g_eli_cancel(sc); LIST_REMOVE(wr, w_next); g_eli_freesession(wr); free(wr, M_ELI); G_ELI_DEBUG(1, "Thread %s exiting.", curthread->td_proc->p_comm); wakeup(&sc->sc_workers); mtx_unlock(&sc->sc_queue_mtx); kproc_exit(0); } while (sc->sc_flags & G_ELI_FLAG_SUSPEND) { if (sc->sc_inflight > 0) { G_ELI_DEBUG(0, "inflight=%d", sc->sc_inflight); /* * We still have inflight BIOs, so * sleep and retry. */ msleep(sc, &sc->sc_queue_mtx, PRIBIO, "geli:inf", hz / 5); goto again; } /* * Suspend requested, mark the worker as * suspended and go to sleep. */ if (wr->w_active) { g_eli_freesession(wr); wr->w_active = FALSE; } wakeup(&sc->sc_workers); msleep(sc, &sc->sc_queue_mtx, PRIBIO, "geli:suspend", 0); if (!wr->w_active && !(sc->sc_flags & G_ELI_FLAG_SUSPEND)) { error = g_eli_newsession(wr); KASSERT(error == 0, ("g_eli_newsession() failed on resume (error=%d)", error)); wr->w_active = TRUE; } goto again; } msleep(sc, &sc->sc_queue_mtx, PDROP, "geli:w", 0); continue; } if (bp->bio_pflags == G_ELI_NEW_BIO) atomic_add_int(&sc->sc_inflight, 1); mtx_unlock(&sc->sc_queue_mtx); if (bp->bio_pflags == G_ELI_NEW_BIO) { bp->bio_pflags = 0; if (sc->sc_flags & G_ELI_FLAG_AUTH) { if (bp->bio_cmd == BIO_READ) g_eli_auth_read(sc, bp); else g_eli_auth_run(wr, bp); } else { if (bp->bio_cmd == BIO_READ) g_eli_crypto_read(sc, bp, 1); else g_eli_crypto_run(wr, bp); } } else { if (sc->sc_flags & G_ELI_FLAG_AUTH) g_eli_auth_run(wr, bp); else g_eli_crypto_run(wr, bp); } } } static int g_eli_read_metadata_offset(struct g_class *mp, struct g_provider *pp, off_t offset, struct g_eli_metadata *md) { struct g_geom *gp; struct g_consumer *cp; u_char *buf = NULL; int error; g_topology_assert(); gp = g_new_geomf(mp, "eli:taste"); gp->start = g_eli_start; gp->access = g_std_access; /* * g_eli_read_metadata() is always called from the event thread. * Our geom is created and destroyed in the same event, so there * could be no orphan nor spoil event in the meantime. */ gp->orphan = g_eli_orphan_spoil_assert; gp->spoiled = g_eli_orphan_spoil_assert; cp = g_new_consumer(gp); cp->flags |= G_CF_DIRECT_SEND | G_CF_DIRECT_RECEIVE; error = g_attach(cp, pp); if (error != 0) goto end; error = g_access(cp, 1, 0, 0); if (error != 0) goto end; g_topology_unlock(); buf = g_read_data(cp, offset, pp->sectorsize, &error); g_topology_lock(); if (buf == NULL) goto end; error = eli_metadata_decode(buf, md); if (error != 0) goto end; /* Metadata was read and decoded successfully. */ end: if (buf != NULL) g_free(buf); if (cp->provider != NULL) { if (cp->acr == 1) g_access(cp, -1, 0, 0); g_detach(cp); } g_destroy_consumer(cp); g_destroy_geom(gp); return (error); } int g_eli_read_metadata(struct g_class *mp, struct g_provider *pp, struct g_eli_metadata *md) { return (g_eli_read_metadata_offset(mp, pp, pp->mediasize - pp->sectorsize, md)); } /* * The function is called when we had last close on provider and user requested * to close it when this situation occur. */ static void g_eli_last_close(void *arg, int flags __unused) { struct g_geom *gp; char gpname[64]; int error; g_topology_assert(); gp = arg; strlcpy(gpname, gp->name, sizeof(gpname)); error = g_eli_destroy(gp->softc, TRUE); KASSERT(error == 0, ("Cannot detach %s on last close (error=%d).", gpname, error)); G_ELI_DEBUG(0, "Detached %s on last close.", gpname); } int g_eli_access(struct g_provider *pp, int dr, int dw, int de) { struct g_eli_softc *sc; struct g_geom *gp; gp = pp->geom; sc = gp->softc; if (dw > 0) { if (sc->sc_flags & G_ELI_FLAG_RO) { /* Deny write attempts. */ return (EROFS); } /* Someone is opening us for write, we need to remember that. */ sc->sc_flags |= G_ELI_FLAG_WOPEN; return (0); } /* Is this the last close? */ if (pp->acr + dr > 0 || pp->acw + dw > 0 || pp->ace + de > 0) return (0); /* * Automatically detach on last close if requested. */ if ((sc->sc_flags & G_ELI_FLAG_RW_DETACH) || (sc->sc_flags & G_ELI_FLAG_WOPEN)) { g_post_event(g_eli_last_close, gp, M_WAITOK, NULL); } return (0); } static int g_eli_cpu_is_disabled(int cpu) { #ifdef SMP return (CPU_ISSET(cpu, &hlt_cpus_mask)); #else return (0); #endif } struct g_geom * g_eli_create(struct gctl_req *req, struct g_class *mp, struct g_provider *bpp, const struct g_eli_metadata *md, const u_char *mkey, int nkey) { struct g_eli_softc *sc; struct g_eli_worker *wr; struct g_geom *gp; struct g_provider *pp; struct g_consumer *cp; struct g_geom_alias *gap; u_int i, threads; int dcw, error; G_ELI_DEBUG(1, "Creating device %s%s.", bpp->name, G_ELI_SUFFIX); KASSERT(eli_metadata_crypto_supported(md), ("%s: unsupported crypto for %s", __func__, bpp->name)); gp = g_new_geomf(mp, "%s%s", bpp->name, G_ELI_SUFFIX); sc = malloc(sizeof(*sc), M_ELI, M_WAITOK | M_ZERO); gp->start = g_eli_start; /* * Spoiling can happen even though we have the provider open * exclusively, e.g. through media change events. */ gp->spoiled = g_eli_orphan; gp->orphan = g_eli_orphan; gp->resize = g_eli_resize; gp->dumpconf = g_eli_dumpconf; /* * If detach-on-last-close feature is not enabled and we don't operate * on read-only provider, we can simply use g_std_access(). */ if (md->md_flags & (G_ELI_FLAG_WO_DETACH | G_ELI_FLAG_RO)) gp->access = g_eli_access; else gp->access = g_std_access; eli_metadata_softc(sc, md, bpp->sectorsize, bpp->mediasize); sc->sc_nkey = nkey; gp->softc = sc; sc->sc_geom = gp; bioq_init(&sc->sc_queue); mtx_init(&sc->sc_queue_mtx, "geli:queue", NULL, MTX_DEF); mtx_init(&sc->sc_ekeys_lock, "geli:ekeys", NULL, MTX_DEF); pp = NULL; cp = g_new_consumer(gp); cp->flags |= G_CF_DIRECT_SEND | G_CF_DIRECT_RECEIVE; error = g_attach(cp, bpp); if (error != 0) { if (req != NULL) { gctl_error(req, "Cannot attach to %s (error=%d).", bpp->name, error); } else { G_ELI_DEBUG(1, "Cannot attach to %s (error=%d).", bpp->name, error); } goto failed; } /* * Keep provider open all the time, so we can run critical tasks, * like Master Keys deletion, without wondering if we can open * provider or not. * We don't open provider for writing only when user requested read-only * access. */ dcw = (sc->sc_flags & G_ELI_FLAG_RO) ? 0 : 1; error = g_access(cp, 1, dcw, 1); if (error != 0) { if (req != NULL) { gctl_error(req, "Cannot access %s (error=%d).", bpp->name, error); } else { G_ELI_DEBUG(1, "Cannot access %s (error=%d).", bpp->name, error); } goto failed; } /* * Remember the keys in our softc structure. */ g_eli_mkey_propagate(sc, mkey); LIST_INIT(&sc->sc_workers); threads = g_eli_threads; if (threads == 0) threads = mp_ncpus; sc->sc_cpubind = (mp_ncpus > 1 && threads == mp_ncpus); for (i = 0; i < threads; i++) { if (g_eli_cpu_is_disabled(i)) { G_ELI_DEBUG(1, "%s: CPU %u disabled, skipping.", bpp->name, i); continue; } wr = malloc(sizeof(*wr), M_ELI, M_WAITOK | M_ZERO); wr->w_softc = sc; wr->w_number = i; wr->w_active = TRUE; error = g_eli_newsession(wr); if (error != 0) { free(wr, M_ELI); if (req != NULL) { gctl_error(req, "Cannot set up crypto session " "for %s (error=%d).", bpp->name, error); } else { G_ELI_DEBUG(1, "Cannot set up crypto session " "for %s (error=%d).", bpp->name, error); } goto failed; } error = kproc_create(g_eli_worker, wr, &wr->w_proc, 0, 0, "g_eli[%u] %s", i, bpp->name); if (error != 0) { g_eli_freesession(wr); free(wr, M_ELI); if (req != NULL) { gctl_error(req, "Cannot create kernel thread " "for %s (error=%d).", bpp->name, error); } else { G_ELI_DEBUG(1, "Cannot create kernel thread " "for %s (error=%d).", bpp->name, error); } goto failed; } LIST_INSERT_HEAD(&sc->sc_workers, wr, w_next); } /* * Create decrypted provider. */ pp = g_new_providerf(gp, "%s%s", bpp->name, G_ELI_SUFFIX); pp->flags |= G_PF_DIRECT_SEND | G_PF_DIRECT_RECEIVE; if (CRYPTO_HAS_VMPAGE) { /* * On DMAP architectures we can use unmapped I/O. But don't * use it with data integrity verification. That code hasn't * been written yet. */ if ((sc->sc_flags & G_ELI_FLAG_AUTH) == 0) pp->flags |= G_PF_ACCEPT_UNMAPPED; } pp->mediasize = sc->sc_mediasize; pp->sectorsize = sc->sc_sectorsize; LIST_FOREACH(gap, &bpp->aliases, ga_next) g_provider_add_alias(pp, "%s%s", gap->ga_alias, G_ELI_SUFFIX); g_error_provider(pp, 0); G_ELI_DEBUG(0, "Device %s created.", pp->name); G_ELI_DEBUG(0, "Encryption: %s %u", g_eli_algo2str(sc->sc_ealgo), sc->sc_ekeylen); if (sc->sc_flags & G_ELI_FLAG_AUTH) G_ELI_DEBUG(0, " Integrity: %s", g_eli_algo2str(sc->sc_aalgo)); G_ELI_DEBUG(0, " Crypto: %s", sc->sc_crypto == G_ELI_CRYPTO_SW_ACCEL ? "accelerated software" : sc->sc_crypto == G_ELI_CRYPTO_SW ? "software" : "hardware"); return (gp); failed: mtx_lock(&sc->sc_queue_mtx); sc->sc_flags |= G_ELI_FLAG_DESTROY; wakeup(sc); /* * Wait for kernel threads self destruction. */ while (!LIST_EMPTY(&sc->sc_workers)) { msleep(&sc->sc_workers, &sc->sc_queue_mtx, PRIBIO, "geli:destroy", 0); } mtx_destroy(&sc->sc_queue_mtx); if (cp->provider != NULL) { if (cp->acr == 1) g_access(cp, -1, -dcw, -1); g_detach(cp); } g_destroy_consumer(cp); g_destroy_geom(gp); g_eli_key_destroy(sc); zfree(sc, M_ELI); return (NULL); } int g_eli_destroy(struct g_eli_softc *sc, boolean_t force) { struct g_geom *gp; struct g_provider *pp; g_topology_assert(); if (sc == NULL) return (ENXIO); gp = sc->sc_geom; pp = LIST_FIRST(&gp->provider); if (pp != NULL && (pp->acr != 0 || pp->acw != 0 || pp->ace != 0)) { if (force) { G_ELI_DEBUG(1, "Device %s is still open, so it " "cannot be definitely removed.", pp->name); sc->sc_flags |= G_ELI_FLAG_RW_DETACH; gp->access = g_eli_access; g_wither_provider(pp, ENXIO); return (EBUSY); } else { G_ELI_DEBUG(1, "Device %s is still open (r%dw%de%d).", pp->name, pp->acr, pp->acw, pp->ace); return (EBUSY); } } mtx_lock(&sc->sc_queue_mtx); sc->sc_flags |= G_ELI_FLAG_DESTROY; wakeup(sc); while (!LIST_EMPTY(&sc->sc_workers)) { msleep(&sc->sc_workers, &sc->sc_queue_mtx, PRIBIO, "geli:destroy", 0); } mtx_destroy(&sc->sc_queue_mtx); gp->softc = NULL; g_eli_key_destroy(sc); zfree(sc, M_ELI); G_ELI_DEBUG(0, "Device %s destroyed.", gp->name); g_wither_geom_close(gp, ENXIO); return (0); } static int g_eli_destroy_geom(struct gctl_req *req __unused, struct g_class *mp __unused, struct g_geom *gp) { struct g_eli_softc *sc; sc = gp->softc; return (g_eli_destroy(sc, FALSE)); } static int g_eli_keyfiles_load(struct hmac_ctx *ctx, const char *provider) { u_char *keyfile, *data; char *file, name[64]; size_t size; int i; for (i = 0; ; i++) { snprintf(name, sizeof(name), "%s:geli_keyfile%d", provider, i); keyfile = preload_search_by_type(name); if (keyfile == NULL && i == 0) { /* * If there is only one keyfile, allow simpler name. */ snprintf(name, sizeof(name), "%s:geli_keyfile", provider); keyfile = preload_search_by_type(name); } if (keyfile == NULL) return (i); /* Return number of loaded keyfiles. */ data = preload_fetch_addr(keyfile); if (data == NULL) { G_ELI_DEBUG(0, "Cannot find key file data for %s.", name); return (0); } size = preload_fetch_size(keyfile); if (size == 0) { G_ELI_DEBUG(0, "Cannot find key file size for %s.", name); return (0); } file = preload_search_info(keyfile, MODINFO_NAME); if (file == NULL) { G_ELI_DEBUG(0, "Cannot find key file name for %s.", name); return (0); } G_ELI_DEBUG(1, "Loaded keyfile %s for %s (type: %s).", file, provider, name); g_eli_crypto_hmac_update(ctx, data, size); } } static void g_eli_keyfiles_clear(const char *provider) { u_char *keyfile, *data; char name[64]; size_t size; int i; for (i = 0; ; i++) { snprintf(name, sizeof(name), "%s:geli_keyfile%d", provider, i); keyfile = preload_search_by_type(name); if (keyfile == NULL) return; data = preload_fetch_addr(keyfile); size = preload_fetch_size(keyfile); if (data != NULL && size != 0) explicit_bzero(data, size); } } /* * Tasting is only made on boot. * We detect providers which should be attached before root is mounted. */ static struct g_geom * g_eli_taste(struct g_class *mp, struct g_provider *pp, int flags __unused) { struct g_eli_metadata md; struct g_geom *gp; struct hmac_ctx ctx; char passphrase[256]; u_char key[G_ELI_USERKEYLEN], mkey[G_ELI_DATAIVKEYLEN]; u_int i, nkey, nkeyfiles, tries, showpass; int error; struct keybuf *keybuf; g_trace(G_T_TOPOLOGY, "%s(%s, %s)", __func__, mp->name, pp->name); g_topology_assert(); if (root_mounted() || g_eli_tries == 0) return (NULL); G_ELI_DEBUG(3, "Tasting %s.", pp->name); error = g_eli_read_metadata(mp, pp, &md); if (error != 0) return (NULL); gp = NULL; if (strcmp(md.md_magic, G_ELI_MAGIC) != 0) return (NULL); if (md.md_version > G_ELI_VERSION) { printf("geom_eli.ko module is too old to handle %s.\n", pp->name); return (NULL); } if (md.md_provsize != pp->mediasize) return (NULL); /* Should we attach it on boot? */ if (!(md.md_flags & G_ELI_FLAG_BOOT) && !(md.md_flags & G_ELI_FLAG_GELIBOOT)) return (NULL); if (md.md_keys == 0x00) { G_ELI_DEBUG(0, "No valid keys on %s.", pp->name); return (NULL); } if (!eli_metadata_crypto_supported(&md)) { G_ELI_DEBUG(0, "%s uses invalid or unsupported algorithms\n", pp->name); return (NULL); } if (md.md_iterations == -1) { /* If there is no passphrase, we try only once. */ tries = 1; } else { /* Ask for the passphrase no more than g_eli_tries times. */ tries = g_eli_tries; } if ((keybuf = get_keybuf()) != NULL) { /* Scan the key buffer, try all GELI keys. */ for (i = 0; i < keybuf->kb_nents; i++) { if (keybuf->kb_ents[i].ke_type == KEYBUF_TYPE_GELI) { memcpy(key, keybuf->kb_ents[i].ke_data, sizeof(key)); if (g_eli_mkey_decrypt_any(&md, key, mkey, &nkey) == 0 ) { explicit_bzero(key, sizeof(key)); goto have_key; } } } } for (i = 0; i <= tries; i++) { g_eli_crypto_hmac_init(&ctx, NULL, 0); /* * Load all key files. */ nkeyfiles = g_eli_keyfiles_load(&ctx, pp->name); if (nkeyfiles == 0 && md.md_iterations == -1) { /* * No key files and no passphrase, something is * definitely wrong here. * geli(8) doesn't allow for such situation, so assume * that there was really no passphrase and in that case * key files are no properly defined in loader.conf. */ G_ELI_DEBUG(0, "Found no key files in loader.conf for %s.", pp->name); return (NULL); } /* Ask for the passphrase if defined. */ if (md.md_iterations >= 0) { /* Try first with cached passphrase. */ if (i == 0) { if (!g_eli_boot_passcache) continue; memcpy(passphrase, cached_passphrase, sizeof(passphrase)); } else { printf("Enter passphrase for %s: ", pp->name); showpass = g_eli_visible_passphrase; if ((md.md_flags & G_ELI_FLAG_GELIDISPLAYPASS) != 0) showpass = GETS_ECHOPASS; cngets(passphrase, sizeof(passphrase), showpass); memcpy(cached_passphrase, passphrase, sizeof(passphrase)); } } /* * Prepare Derived-Key from the user passphrase. */ if (md.md_iterations == 0) { g_eli_crypto_hmac_update(&ctx, md.md_salt, sizeof(md.md_salt)); g_eli_crypto_hmac_update(&ctx, passphrase, strlen(passphrase)); explicit_bzero(passphrase, sizeof(passphrase)); } else if (md.md_iterations > 0) { u_char dkey[G_ELI_USERKEYLEN]; pkcs5v2_genkey(dkey, sizeof(dkey), md.md_salt, sizeof(md.md_salt), passphrase, md.md_iterations); explicit_bzero(passphrase, sizeof(passphrase)); g_eli_crypto_hmac_update(&ctx, dkey, sizeof(dkey)); explicit_bzero(dkey, sizeof(dkey)); } g_eli_crypto_hmac_final(&ctx, key, 0); /* * Decrypt Master-Key. */ error = g_eli_mkey_decrypt_any(&md, key, mkey, &nkey); explicit_bzero(key, sizeof(key)); if (error == -1) { if (i == tries) { G_ELI_DEBUG(0, "Wrong key for %s. No tries left.", pp->name); g_eli_keyfiles_clear(pp->name); return (NULL); } if (i > 0) { G_ELI_DEBUG(0, "Wrong key for %s. Tries left: %u.", pp->name, tries - i); } /* Try again. */ continue; } else if (error > 0) { G_ELI_DEBUG(0, "Cannot decrypt Master Key for %s (error=%d).", pp->name, error); g_eli_keyfiles_clear(pp->name); return (NULL); } g_eli_keyfiles_clear(pp->name); G_ELI_DEBUG(1, "Using Master Key %u for %s.", nkey, pp->name); break; } have_key: /* * We have correct key, let's attach provider. */ gp = g_eli_create(NULL, mp, pp, &md, mkey, nkey); explicit_bzero(mkey, sizeof(mkey)); explicit_bzero(&md, sizeof(md)); if (gp == NULL) { G_ELI_DEBUG(0, "Cannot create device %s%s.", pp->name, G_ELI_SUFFIX); return (NULL); } return (gp); } static void g_eli_dumpconf(struct sbuf *sb, const char *indent, struct g_geom *gp, struct g_consumer *cp, struct g_provider *pp) { struct g_eli_softc *sc; g_topology_assert(); sc = gp->softc; if (sc == NULL) return; if (pp != NULL || cp != NULL) return; /* Nothing here. */ sbuf_printf(sb, "%s%ju\n", indent, (uintmax_t)sc->sc_ekeys_total); sbuf_printf(sb, "%s%ju\n", indent, (uintmax_t)sc->sc_ekeys_allocated); sbuf_printf(sb, "%s", indent); if (sc->sc_flags == 0) sbuf_cat(sb, "NONE"); else { int first = 1; #define ADD_FLAG(flag, name) do { \ if (sc->sc_flags & (flag)) { \ if (!first) \ sbuf_cat(sb, ", "); \ else \ first = 0; \ sbuf_cat(sb, name); \ } \ } while (0) ADD_FLAG(G_ELI_FLAG_SUSPEND, "SUSPEND"); ADD_FLAG(G_ELI_FLAG_SINGLE_KEY, "SINGLE-KEY"); ADD_FLAG(G_ELI_FLAG_NATIVE_BYTE_ORDER, "NATIVE-BYTE-ORDER"); ADD_FLAG(G_ELI_FLAG_ONETIME, "ONETIME"); ADD_FLAG(G_ELI_FLAG_BOOT, "BOOT"); ADD_FLAG(G_ELI_FLAG_WO_DETACH, "W-DETACH"); ADD_FLAG(G_ELI_FLAG_RW_DETACH, "RW-DETACH"); ADD_FLAG(G_ELI_FLAG_AUTH, "AUTH"); ADD_FLAG(G_ELI_FLAG_WOPEN, "W-OPEN"); ADD_FLAG(G_ELI_FLAG_DESTROY, "DESTROY"); ADD_FLAG(G_ELI_FLAG_RO, "READ-ONLY"); ADD_FLAG(G_ELI_FLAG_NODELETE, "NODELETE"); ADD_FLAG(G_ELI_FLAG_GELIBOOT, "GELIBOOT"); ADD_FLAG(G_ELI_FLAG_GELIDISPLAYPASS, "GELIDISPLAYPASS"); ADD_FLAG(G_ELI_FLAG_AUTORESIZE, "AUTORESIZE"); #undef ADD_FLAG } sbuf_cat(sb, "\n"); if (!(sc->sc_flags & G_ELI_FLAG_ONETIME)) { sbuf_printf(sb, "%s%u\n", indent, sc->sc_nkey); } sbuf_printf(sb, "%s%u\n", indent, sc->sc_version); sbuf_printf(sb, "%s", indent); switch (sc->sc_crypto) { case G_ELI_CRYPTO_HW: sbuf_cat(sb, "hardware"); break; case G_ELI_CRYPTO_SW: sbuf_cat(sb, "software"); break; case G_ELI_CRYPTO_SW_ACCEL: sbuf_cat(sb, "accelerated software"); break; default: sbuf_cat(sb, "UNKNOWN"); break; } sbuf_cat(sb, "\n"); if (sc->sc_flags & G_ELI_FLAG_AUTH) { sbuf_printf(sb, "%s%s\n", indent, g_eli_algo2str(sc->sc_aalgo)); } sbuf_printf(sb, "%s%u\n", indent, sc->sc_ekeylen); sbuf_printf(sb, "%s%s\n", indent, g_eli_algo2str(sc->sc_ealgo)); sbuf_printf(sb, "%s%s\n", indent, (sc->sc_flags & G_ELI_FLAG_SUSPEND) ? "SUSPENDED" : "ACTIVE"); } static void g_eli_shutdown_pre_sync(void *arg, int howto) { struct g_class *mp; struct g_geom *gp, *gp2; struct g_provider *pp; struct g_eli_softc *sc; int error; mp = arg; g_topology_lock(); LIST_FOREACH_SAFE(gp, &mp->geom, geom, gp2) { sc = gp->softc; if (sc == NULL) continue; pp = LIST_FIRST(&gp->provider); KASSERT(pp != NULL, ("No provider? gp=%p (%s)", gp, gp->name)); if (pp->acr != 0 || pp->acw != 0 || pp->ace != 0 || SCHEDULER_STOPPED()) { sc->sc_flags |= G_ELI_FLAG_RW_DETACH; gp->access = g_eli_access; } else { error = g_eli_destroy(sc, TRUE); } } g_topology_unlock(); } static void g_eli_init(struct g_class *mp) { g_eli_pre_sync = EVENTHANDLER_REGISTER(shutdown_pre_sync, g_eli_shutdown_pre_sync, mp, SHUTDOWN_PRI_FIRST); if (g_eli_pre_sync == NULL) G_ELI_DEBUG(0, "Warning! Cannot register shutdown event."); } static void g_eli_fini(struct g_class *mp) { if (g_eli_pre_sync != NULL) EVENTHANDLER_DEREGISTER(shutdown_pre_sync, g_eli_pre_sync); } DECLARE_GEOM_CLASS(g_eli_class, g_eli); MODULE_DEPEND(g_eli, crypto, 1, 1, 1); MODULE_VERSION(geom_eli, 0); Index: head/sys/kern/kern_environment.c =================================================================== --- head/sys/kern/kern_environment.c (revision 366507) +++ head/sys/kern/kern_environment.c (revision 366508) @@ -1,1079 +1,1080 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 1998 Michael Smith * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ /* * The unified bootloader passes us a pointer to a preserved copy of * bootstrap/kernel environment variables. We convert them to a * dynamic array of strings later when the VM subsystem is up. * * We make these available through the kenv(2) syscall for userland * and through kern_getenv()/freeenv() kern_setenv() kern_unsetenv() testenv() for * the kernel. */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include +#include #include #include #include #include #include #include #include #include static char *_getenv_dynamic_locked(const char *name, int *idx); static char *_getenv_dynamic(const char *name, int *idx); static char *kenv_acquire(const char *name); static void kenv_release(const char *buf); static MALLOC_DEFINE(M_KENV, "kenv", "kernel environment"); #define KENV_SIZE 512 /* Maximum number of environment strings */ static uma_zone_t kenv_zone; static int kenv_mvallen = KENV_MVALLEN; /* pointer to the config-generated static environment */ char *kern_envp; /* pointer to the md-static environment */ char *md_envp; static int md_env_len; static int md_env_pos; static char *kernenv_next(char *); /* dynamic environment variables */ char **kenvp; struct mtx kenv_lock; /* * No need to protect this with a mutex since SYSINITS are single threaded. */ bool dynamic_kenv; #define KENV_CHECK if (!dynamic_kenv) \ panic("%s: called before SI_SUB_KMEM", __func__) int sys_kenv(td, uap) struct thread *td; struct kenv_args /* { int what; const char *name; char *value; int len; } */ *uap; { char *name, *value, *buffer = NULL; size_t len, done, needed, buflen; int error, i; KASSERT(dynamic_kenv, ("kenv: dynamic_kenv = false")); error = 0; if (uap->what == KENV_DUMP) { #ifdef MAC error = mac_kenv_check_dump(td->td_ucred); if (error) return (error); #endif done = needed = 0; buflen = uap->len; if (buflen > KENV_SIZE * (KENV_MNAMELEN + kenv_mvallen + 2)) buflen = KENV_SIZE * (KENV_MNAMELEN + kenv_mvallen + 2); if (uap->len > 0 && uap->value != NULL) buffer = malloc(buflen, M_TEMP, M_WAITOK|M_ZERO); mtx_lock(&kenv_lock); for (i = 0; kenvp[i] != NULL; i++) { len = strlen(kenvp[i]) + 1; needed += len; len = min(len, buflen - done); /* * If called with a NULL or insufficiently large * buffer, just keep computing the required size. */ if (uap->value != NULL && buffer != NULL && len > 0) { bcopy(kenvp[i], buffer + done, len); done += len; } } mtx_unlock(&kenv_lock); if (buffer != NULL) { error = copyout(buffer, uap->value, done); free(buffer, M_TEMP); } td->td_retval[0] = ((done == needed) ? 0 : needed); return (error); } switch (uap->what) { case KENV_SET: error = priv_check(td, PRIV_KENV_SET); if (error) return (error); break; case KENV_UNSET: error = priv_check(td, PRIV_KENV_UNSET); if (error) return (error); break; } name = malloc(KENV_MNAMELEN + 1, M_TEMP, M_WAITOK); error = copyinstr(uap->name, name, KENV_MNAMELEN + 1, NULL); if (error) goto done; switch (uap->what) { case KENV_GET: #ifdef MAC error = mac_kenv_check_get(td->td_ucred, name); if (error) goto done; #endif value = kern_getenv(name); if (value == NULL) { error = ENOENT; goto done; } len = strlen(value) + 1; if (len > uap->len) len = uap->len; error = copyout(value, uap->value, len); freeenv(value); if (error) goto done; td->td_retval[0] = len; break; case KENV_SET: len = uap->len; if (len < 1) { error = EINVAL; goto done; } if (len > kenv_mvallen + 1) len = kenv_mvallen + 1; value = malloc(len, M_TEMP, M_WAITOK); error = copyinstr(uap->value, value, len, NULL); if (error) { free(value, M_TEMP); goto done; } #ifdef MAC error = mac_kenv_check_set(td->td_ucred, name, value); if (error == 0) #endif kern_setenv(name, value); free(value, M_TEMP); break; case KENV_UNSET: #ifdef MAC error = mac_kenv_check_unset(td->td_ucred, name); if (error) goto done; #endif error = kern_unsetenv(name); if (error) error = ENOENT; break; default: error = EINVAL; break; } done: free(name, M_TEMP); return (error); } /* * Populate the initial kernel environment. * * This is called very early in MD startup, either to provide a copy of the * environment obtained from a boot loader, or to provide an empty buffer into * which MD code can store an initial environment using kern_setenv() calls. * * kern_envp is set to the static_env generated by config(8). This implements * the env keyword described in config(5). * * If len is non-zero, the caller is providing an empty buffer. The caller will * subsequently use kern_setenv() to add up to len bytes of initial environment * before the dynamic environment is available. * * If len is zero, the caller is providing a pre-loaded buffer containing * environment strings. Additional strings cannot be added until the dynamic * environment is available. The memory pointed to must remain stable at least * until sysinit runs init_dynamic_kenv() and preferably until after SI_SUB_KMEM * is finished so that subr_hints routines may continue to use it until the * environments have been fully merged at the end of the pass. If no initial * environment is available from the boot loader, passing a NULL pointer allows * the static_env to be installed if it is configured. In this case, any call * to kern_setenv() prior to the setup of the dynamic environment will result in * a panic. */ void init_static_kenv(char *buf, size_t len) { KASSERT(!dynamic_kenv, ("kenv: dynamic_kenv already initialized")); /* * Suitably sized means it must be able to hold at least one empty * variable, otherwise things go belly up if a kern_getenv call is * made without a prior call to kern_setenv as we have a malformed * environment. */ KASSERT(len == 0 || len >= 2, ("kenv: static env must be initialized or suitably sized")); KASSERT(len == 0 || (*buf == '\0' && *(buf + 1) == '\0'), ("kenv: sized buffer must be initially empty")); /* * We may be called twice, with the second call needed to relocate * md_envp after enabling paging. md_envp is then garbage if it is * not null and the relocation will move it. Discard it so as to * not crash using its old value in our first call to kern_getenv(). * * The second call gives the same environment as the first except * in silly configurations where the static env disables itself. * * Other env calls don't handle possibly-garbage pointers, so must * not be made between enabling paging and calling here. */ md_envp = NULL; md_env_len = 0; md_env_pos = 0; /* * Give the static environment a chance to disable the loader(8) * environment first. This is done with loader_env.disabled=1. * * static_env and static_hints may both be disabled, but in slightly * different ways. For static_env, we just don't setup kern_envp and * it's as if a static env wasn't even provided. For static_hints, * we effectively zero out the buffer to stop the rest of the kernel * from being able to use it. * * We're intentionally setting this up so that static_hints.disabled may * be specified in either the MD env or the static env. This keeps us * consistent in our new world view. * * As a warning, the static environment may not be disabled in any way * if the static environment has disabled the loader environment. */ kern_envp = static_env; if (!getenv_is_true("loader_env.disabled")) { md_envp = buf; md_env_len = len; md_env_pos = 0; if (getenv_is_true("static_env.disabled")) { kern_envp[0] = '\0'; kern_envp[1] = '\0'; } } if (getenv_is_true("static_hints.disabled")) { static_hints[0] = '\0'; static_hints[1] = '\0'; } } static void init_dynamic_kenv_from(char *init_env, int *curpos) { char *cp, *cpnext, *eqpos, *found; size_t len; int i; if (init_env && *init_env != '\0') { found = NULL; i = *curpos; for (cp = init_env; cp != NULL; cp = cpnext) { cpnext = kernenv_next(cp); len = strlen(cp) + 1; if (len > KENV_MNAMELEN + 1 + kenv_mvallen + 1) { printf( "WARNING: too long kenv string, ignoring %s\n", cp); goto sanitize; } eqpos = strchr(cp, '='); if (eqpos == NULL) { printf( "WARNING: malformed static env value, ignoring %s\n", cp); goto sanitize; } *eqpos = 0; /* * De-dupe the environment as we go. We don't add the * duplicated assignments because config(8) will flip * the order of the static environment around to make * kernel processing match the order of specification * in the kernel config. */ found = _getenv_dynamic_locked(cp, NULL); *eqpos = '='; if (found != NULL) goto sanitize; if (i > KENV_SIZE) { printf( "WARNING: too many kenv strings, ignoring %s\n", cp); goto sanitize; } kenvp[i] = malloc(len, M_KENV, M_WAITOK); strcpy(kenvp[i++], cp); sanitize: explicit_bzero(cp, len - 1); } *curpos = i; } } /* * Setup the dynamic kernel environment. */ static void init_dynamic_kenv(void *data __unused) { int dynamic_envpos; int size; TUNABLE_INT_FETCH("kenv_mvallen", &kenv_mvallen); size = KENV_MNAMELEN + 1 + kenv_mvallen + 1; kenv_zone = uma_zcreate("kenv", size, NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); kenvp = malloc((KENV_SIZE + 1) * sizeof(char *), M_KENV, M_WAITOK | M_ZERO); dynamic_envpos = 0; init_dynamic_kenv_from(md_envp, &dynamic_envpos); init_dynamic_kenv_from(kern_envp, &dynamic_envpos); kenvp[dynamic_envpos] = NULL; mtx_init(&kenv_lock, "kernel environment", NULL, MTX_DEF); dynamic_kenv = true; } SYSINIT(kenv, SI_SUB_KMEM + 1, SI_ORDER_FIRST, init_dynamic_kenv, NULL); void freeenv(char *env) { if (dynamic_kenv && env != NULL) { explicit_bzero(env, strlen(env)); uma_zfree(kenv_zone, env); } } /* * Internal functions for string lookup. */ static char * _getenv_dynamic_locked(const char *name, int *idx) { char *cp; int len, i; len = strlen(name); for (cp = kenvp[0], i = 0; cp != NULL; cp = kenvp[++i]) { if ((strncmp(cp, name, len) == 0) && (cp[len] == '=')) { if (idx != NULL) *idx = i; return (cp + len + 1); } } return (NULL); } static char * _getenv_dynamic(const char *name, int *idx) { mtx_assert(&kenv_lock, MA_OWNED); return (_getenv_dynamic_locked(name, idx)); } static char * _getenv_static_from(char *chkenv, const char *name) { char *cp, *ep; int len; for (cp = chkenv; cp != NULL; cp = kernenv_next(cp)) { for (ep = cp; (*ep != '=') && (*ep != 0); ep++) ; if (*ep != '=') continue; len = ep - cp; ep++; if (!strncmp(name, cp, len) && name[len] == 0) return (ep); } return (NULL); } static char * _getenv_static(const char *name) { char *val; val = _getenv_static_from(md_envp, name); if (val != NULL) return (val); val = _getenv_static_from(kern_envp, name); if (val != NULL) return (val); return (NULL); } /* * Look up an environment variable by name. * Return a pointer to the string if found. * The pointer has to be freed with freeenv() * after use. */ char * kern_getenv(const char *name) { char *cp, *ret; int len; if (dynamic_kenv) { len = KENV_MNAMELEN + 1 + kenv_mvallen + 1; ret = uma_zalloc(kenv_zone, M_WAITOK | M_ZERO); mtx_lock(&kenv_lock); cp = _getenv_dynamic(name, NULL); if (cp != NULL) strlcpy(ret, cp, len); mtx_unlock(&kenv_lock); if (cp == NULL) { uma_zfree(kenv_zone, ret); ret = NULL; } } else ret = _getenv_static(name); return (ret); } /* * Test if an environment variable is defined. */ int testenv(const char *name) { char *cp; cp = kenv_acquire(name); kenv_release(cp); if (cp != NULL) return (1); return (0); } /* * Set an environment variable in the MD-static environment. This cannot * feasibly be done on config(8)-generated static environments as they don't * generally include space for extra variables. */ static int setenv_static(const char *name, const char *value) { int len; if (md_env_pos >= md_env_len) return (-1); /* Check space for x=y and two nuls */ len = strlen(name) + strlen(value); if (len + 3 < md_env_len - md_env_pos) { len = sprintf(&md_envp[md_env_pos], "%s=%s", name, value); md_env_pos += len+1; md_envp[md_env_pos] = '\0'; return (0); } else return (-1); } /* * Set an environment variable by name. */ int kern_setenv(const char *name, const char *value) { char *buf, *cp, *oldenv; int namelen, vallen, i; if (!dynamic_kenv && md_env_len > 0) return (setenv_static(name, value)); KENV_CHECK; namelen = strlen(name) + 1; if (namelen > KENV_MNAMELEN + 1) return (-1); vallen = strlen(value) + 1; if (vallen > kenv_mvallen + 1) return (-1); buf = malloc(namelen + vallen, M_KENV, M_WAITOK); sprintf(buf, "%s=%s", name, value); mtx_lock(&kenv_lock); cp = _getenv_dynamic(name, &i); if (cp != NULL) { oldenv = kenvp[i]; kenvp[i] = buf; mtx_unlock(&kenv_lock); free(oldenv, M_KENV); } else { /* We add the option if it wasn't found */ for (i = 0; (cp = kenvp[i]) != NULL; i++) ; /* Bounds checking */ if (i < 0 || i >= KENV_SIZE) { free(buf, M_KENV); mtx_unlock(&kenv_lock); return (-1); } kenvp[i] = buf; kenvp[i + 1] = NULL; mtx_unlock(&kenv_lock); } return (0); } /* * Unset an environment variable string. */ int kern_unsetenv(const char *name) { char *cp, *oldenv; int i, j; KENV_CHECK; mtx_lock(&kenv_lock); cp = _getenv_dynamic(name, &i); if (cp != NULL) { oldenv = kenvp[i]; for (j = i + 1; kenvp[j] != NULL; j++) kenvp[i++] = kenvp[j]; kenvp[i] = NULL; mtx_unlock(&kenv_lock); zfree(oldenv, M_KENV); return (0); } mtx_unlock(&kenv_lock); return (-1); } /* * Return the internal kenv buffer for the variable name, if it exists. * If the dynamic kenv is initialized and the name is present, return * with kenv_lock held. */ static char * kenv_acquire(const char *name) { char *value; if (dynamic_kenv) { mtx_lock(&kenv_lock); value = _getenv_dynamic(name, NULL); if (value == NULL) mtx_unlock(&kenv_lock); return (value); } else return (_getenv_static(name)); } /* * Undo a previous kenv_acquire() operation */ static void kenv_release(const char *buf) { if ((buf != NULL) && dynamic_kenv) mtx_unlock(&kenv_lock); } /* * Return a string value from an environment variable. */ int getenv_string(const char *name, char *data, int size) { char *cp; cp = kenv_acquire(name); if (cp != NULL) strlcpy(data, cp, size); kenv_release(cp); return (cp != NULL); } /* * Return an array of integers at the given type size and signedness. */ int getenv_array(const char *name, void *pdata, int size, int *psize, int type_size, bool allow_signed) { uint8_t shift; int64_t value; int64_t old; const char *buf; char *end; const char *ptr; int n; int rc; rc = 0; /* assume failure */ buf = kenv_acquire(name); if (buf == NULL) goto error; /* get maximum number of elements */ size /= type_size; n = 0; for (ptr = buf; *ptr != 0; ) { value = strtoq(ptr, &end, 0); /* check if signed numbers are allowed */ if (value < 0 && !allow_signed) goto error; /* check for invalid value */ if (ptr == end) goto error; /* check for valid suffix */ switch (*end) { case 't': case 'T': shift = 40; end++; break; case 'g': case 'G': shift = 30; end++; break; case 'm': case 'M': shift = 20; end++; break; case 'k': case 'K': shift = 10; end++; break; case ' ': case '\t': case ',': case 0: shift = 0; break; default: /* garbage after numeric value */ goto error; } /* skip till next value, if any */ while (*end == '\t' || *end == ',' || *end == ' ') end++; /* update pointer */ ptr = end; /* apply shift */ old = value; value <<= shift; /* overflow check */ if ((value >> shift) != old) goto error; /* check for buffer overflow */ if (n >= size) goto error; /* store value according to type size */ switch (type_size) { case 1: if (allow_signed) { if (value < SCHAR_MIN || value > SCHAR_MAX) goto error; } else { if (value < 0 || value > UCHAR_MAX) goto error; } ((uint8_t *)pdata)[n] = (uint8_t)value; break; case 2: if (allow_signed) { if (value < SHRT_MIN || value > SHRT_MAX) goto error; } else { if (value < 0 || value > USHRT_MAX) goto error; } ((uint16_t *)pdata)[n] = (uint16_t)value; break; case 4: if (allow_signed) { if (value < INT_MIN || value > INT_MAX) goto error; } else { if (value > UINT_MAX) goto error; } ((uint32_t *)pdata)[n] = (uint32_t)value; break; case 8: ((uint64_t *)pdata)[n] = (uint64_t)value; break; default: goto error; } n++; } *psize = n * type_size; if (n != 0) rc = 1; /* success */ error: kenv_release(buf); return (rc); } /* * Return an integer value from an environment variable. */ int getenv_int(const char *name, int *data) { quad_t tmp; int rval; rval = getenv_quad(name, &tmp); if (rval) *data = (int) tmp; return (rval); } /* * Return an unsigned integer value from an environment variable. */ int getenv_uint(const char *name, unsigned int *data) { quad_t tmp; int rval; rval = getenv_quad(name, &tmp); if (rval) *data = (unsigned int) tmp; return (rval); } /* * Return an int64_t value from an environment variable. */ int getenv_int64(const char *name, int64_t *data) { quad_t tmp; int64_t rval; rval = getenv_quad(name, &tmp); if (rval) *data = (int64_t) tmp; return (rval); } /* * Return an uint64_t value from an environment variable. */ int getenv_uint64(const char *name, uint64_t *data) { quad_t tmp; uint64_t rval; rval = getenv_quad(name, &tmp); if (rval) *data = (uint64_t) tmp; return (rval); } /* * Return a long value from an environment variable. */ int getenv_long(const char *name, long *data) { quad_t tmp; int rval; rval = getenv_quad(name, &tmp); if (rval) *data = (long) tmp; return (rval); } /* * Return an unsigned long value from an environment variable. */ int getenv_ulong(const char *name, unsigned long *data) { quad_t tmp; int rval; rval = getenv_quad(name, &tmp); if (rval) *data = (unsigned long) tmp; return (rval); } /* * Return a quad_t value from an environment variable. */ int getenv_quad(const char *name, quad_t *data) { const char *value; char suffix, *vtp; quad_t iv; value = kenv_acquire(name); if (value == NULL) { goto error; } iv = strtoq(value, &vtp, 0); if (vtp == value || (vtp[0] != '\0' && vtp[1] != '\0')) { goto error; } suffix = vtp[0]; kenv_release(value); switch (suffix) { case 't': case 'T': iv *= 1024; /* FALLTHROUGH */ case 'g': case 'G': iv *= 1024; /* FALLTHROUGH */ case 'm': case 'M': iv *= 1024; /* FALLTHROUGH */ case 'k': case 'K': iv *= 1024; case '\0': break; default: return (0); } *data = iv; return (1); error: kenv_release(value); return (0); } /* * Return a boolean value from an environment variable. This can be in * numerical or string form, i.e. "1" or "true". */ int getenv_bool(const char *name, bool *data) { char *val; int ret = 0; if (name == NULL) return (0); val = kern_getenv(name); if (val == NULL) return (0); if ((strcmp(val, "1") == 0) || (strcasecmp(val, "true") == 0)) { *data = true; ret = 1; } else if ((strcmp(val, "0") == 0) || (strcasecmp(val, "false") == 0)) { *data = false; ret = 1; } else { /* Spit out a warning for malformed boolean variables. */ printf("Environment variable %s has non-boolean value \"%s\"\n", name, val); } freeenv(val); return (ret); } /* * Wrapper around getenv_bool to easily check for true. */ bool getenv_is_true(const char *name) { bool val; if (getenv_bool(name, &val) != 0) return (val); return (false); } /* * Wrapper around getenv_bool to easily check for false. */ bool getenv_is_false(const char *name) { bool val; if (getenv_bool(name, &val) != 0) return (!val); return (false); } /* * Find the next entry after the one which (cp) falls within, return a * pointer to its start or NULL if there are no more. */ static char * kernenv_next(char *cp) { if (cp != NULL) { while (*cp != 0) cp++; cp++; if (*cp == 0) cp = NULL; } return (cp); } void tunable_int_init(void *data) { struct tunable_int *d = (struct tunable_int *)data; TUNABLE_INT_FETCH(d->path, d->var); } void tunable_long_init(void *data) { struct tunable_long *d = (struct tunable_long *)data; TUNABLE_LONG_FETCH(d->path, d->var); } void tunable_ulong_init(void *data) { struct tunable_ulong *d = (struct tunable_ulong *)data; TUNABLE_ULONG_FETCH(d->path, d->var); } void tunable_int64_init(void *data) { struct tunable_int64 *d = (struct tunable_int64 *)data; TUNABLE_INT64_FETCH(d->path, d->var); } void tunable_uint64_init(void *data) { struct tunable_uint64 *d = (struct tunable_uint64 *)data; TUNABLE_UINT64_FETCH(d->path, d->var); } void tunable_quad_init(void *data) { struct tunable_quad *d = (struct tunable_quad *)data; TUNABLE_QUAD_FETCH(d->path, d->var); } void tunable_bool_init(void *data) { struct tunable_bool *d = (struct tunable_bool *)data; TUNABLE_BOOL_FETCH(d->path, d->var); } void tunable_str_init(void *data) { struct tunable_str *d = (struct tunable_str *)data; TUNABLE_STR_FETCH(d->path, d->var, d->size); } Index: head/sys/kern/subr_hints.c =================================================================== --- head/sys/kern/subr_hints.c (revision 366507) +++ head/sys/kern/subr_hints.c (revision 366508) @@ -1,516 +1,517 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2000,2001 Peter Wemm * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include __FBSDID("$FreeBSD$"); #include #include +#include #include #include #include #include #include #include #define FBACK_MDENV 0 /* MD env (e.g. loader.conf) */ #define FBACK_STENV 1 /* Static env */ #define FBACK_STATIC 2 /* static_hints */ /* * We'll use hintenv_merged to indicate that the dynamic environment has been * properly prepared for hint usage. This implies that the dynamic environment * has already been setup (dynamic_kenv) and that we have added any supplied * static_hints to the dynamic environment. */ static bool hintenv_merged; /* Static environment and static hints cannot change, so we'll skip known bad */ static bool stenv_skip; static bool sthints_skip; /* * Access functions for device resources. */ static void static_hints_to_env(void *data __unused) { const char *cp; char *line, *eq; int eqidx, i; cp = static_hints; while (cp && *cp != '\0') { eq = strchr(cp, '='); if (eq == NULL) /* Bad hint value */ continue; eqidx = eq - cp; i = strlen(cp); line = malloc(i + 1, M_TEMP, M_WAITOK); strcpy(line, cp); line[eqidx] = line[i] = '\0'; /* * Before adding a hint to the dynamic environment, check if * another value for said hint has already been added. This is * needed because static environment overrides static hints and * dynamic environment overrides all. */ if (testenv(line) == 0) kern_setenv(line, line + eqidx + 1); free(line, M_TEMP); cp += i + 1; } hintenv_merged = true; } /* Any time after dynamic env is setup */ SYSINIT(hintenv, SI_SUB_KMEM + 1, SI_ORDER_SECOND, static_hints_to_env, NULL); /* * Checks the environment to see if we even have any hints. If it has no hints, * then res_find can take the hint that there's no point in searching it and * either move on to the next environment or fail early. */ static bool _res_checkenv(char *envp) { char *cp; cp = envp; while (cp) { if (strncmp(cp, "hint.", 5) == 0) return (true); while (*cp != '\0') cp++; cp++; if (*cp == '\0') break; } return (false); } /* * Evil wildcarding resource string lookup. * This walks the supplied env string table and returns a match. * The start point can be remembered for incremental searches. */ static int res_find(char **hintp_cookie, int *line, int *startln, const char *name, int *unit, const char *resname, const char *value, const char **ret_name, int *ret_namelen, int *ret_unit, const char **ret_resname, int *ret_resnamelen, const char **ret_value) { int fbacklvl = FBACK_MDENV, i = 0, n = 0; char r_name[32]; int r_unit; char r_resname[32]; char r_value[128]; const char *s, *cp; char *hintp, *p; bool dyn_used = false; /* * We are expecting that the caller will pass us a hintp_cookie that * they are tracking. Upon entry, if *hintp_cookie is *not* set, this * indicates to us that we should be figuring out based on the current * environment where to search. This keeps us sane throughout the * entirety of a single search. */ if (*hintp_cookie == NULL) { hintp = NULL; if (hintenv_merged) { /* * static_hints, if it was previously used, has * already been folded in to the environment * by this point. */ mtx_lock(&kenv_lock); cp = kenvp[0]; for (i = 0; cp != NULL; cp = kenvp[++i]) { if (!strncmp(cp, "hint.", 5)) { hintp = kenvp[0]; break; } } mtx_unlock(&kenv_lock); dyn_used = true; } else { /* * We'll have a chance to keep coming back here until * we've actually exhausted all of our possibilities. * We might have chosen the MD/Static env because it * had some kind of hints, but perhaps it didn't have * the hint we are looking for. We don't provide any * fallback when searching the dynamic environment. */ fallback: if (dyn_used || fbacklvl >= FBACK_STATIC) return (ENOENT); switch (fbacklvl) { case FBACK_MDENV: fbacklvl++; if (_res_checkenv(md_envp)) { hintp = md_envp; break; } /* FALLTHROUGH */ case FBACK_STENV: fbacklvl++; if (!stenv_skip && _res_checkenv(kern_envp)) { hintp = kern_envp; break; } else stenv_skip = true; /* FALLTHROUGH */ case FBACK_STATIC: fbacklvl++; /* We'll fallback to static_hints if needed/can */ if (!sthints_skip && _res_checkenv(static_hints)) hintp = static_hints; else sthints_skip = true; break; default: return (ENOENT); } } if (hintp == NULL) return (ENOENT); *hintp_cookie = hintp; } else { hintp = *hintp_cookie; if (hintenv_merged && hintp == kenvp[0]) dyn_used = true; else /* * If we aren't using the dynamic environment, we need * to run through the proper fallback procedure again. * This is so that we do continuations right if we're * working with *line and *startln. */ goto fallback; } if (dyn_used) { mtx_lock(&kenv_lock); i = 0; } cp = hintp; while (cp) { (*line)++; if (strncmp(cp, "hint.", 5) != 0) goto nexthint; n = sscanf(cp, "hint.%32[^.].%d.%32[^=]=%127s", r_name, &r_unit, r_resname, r_value); if (n != 4) { printf("CONFIG: invalid hint '%s'\n", cp); p = strchr(cp, 'h'); *p = 'H'; goto nexthint; } if (startln && *startln >= 0 && *line < *startln) goto nexthint; if (name && strcmp(name, r_name) != 0) goto nexthint; if (unit && *unit != r_unit) goto nexthint; if (resname && strcmp(resname, r_resname) != 0) goto nexthint; if (value && strcmp(value, r_value) != 0) goto nexthint; /* Successfully found a hint matching all criteria */ break; nexthint: if (dyn_used) { cp = kenvp[++i]; if (cp == NULL) break; } else { while (*cp != '\0') cp++; cp++; if (*cp == '\0') { cp = NULL; break; } } } if (dyn_used) mtx_unlock(&kenv_lock); if (cp == NULL) goto fallback; s = cp; /* This is a bit of a hack, but at least is reentrant */ /* Note that it returns some !unterminated! strings. */ s = strchr(s, '.') + 1; /* start of device */ if (ret_name) *ret_name = s; s = strchr(s, '.') + 1; /* start of unit */ if (ret_namelen && ret_name) *ret_namelen = s - *ret_name - 1; /* device length */ if (ret_unit) *ret_unit = r_unit; s = strchr(s, '.') + 1; /* start of resname */ if (ret_resname) *ret_resname = s; s = strchr(s, '=') + 1; /* start of value */ if (ret_resnamelen && ret_resname) *ret_resnamelen = s - *ret_resname - 1; /* value len */ if (ret_value) *ret_value = s; if (startln) /* line number for anchor */ *startln = *line + 1; return 0; } /* * Search all the data sources for matches to our query. We look for * dynamic hints first as overrides for static or fallback hints. */ static int resource_find(int *line, int *startln, const char *name, int *unit, const char *resname, const char *value, const char **ret_name, int *ret_namelen, int *ret_unit, const char **ret_resname, int *ret_resnamelen, const char **ret_value) { int i; int un; char *hintp; *line = 0; hintp = NULL; /* Search for exact unit matches first */ i = res_find(&hintp, line, startln, name, unit, resname, value, ret_name, ret_namelen, ret_unit, ret_resname, ret_resnamelen, ret_value); if (i == 0) return 0; if (unit == NULL) return ENOENT; /* If we are still here, search for wildcard matches */ un = -1; i = res_find(&hintp, line, startln, name, &un, resname, value, ret_name, ret_namelen, ret_unit, ret_resname, ret_resnamelen, ret_value); if (i == 0) return 0; return ENOENT; } int resource_int_value(const char *name, int unit, const char *resname, int *result) { int error; const char *str; char *op; unsigned long val; int line; line = 0; error = resource_find(&line, NULL, name, &unit, resname, NULL, NULL, NULL, NULL, NULL, NULL, &str); if (error) return error; if (*str == '\0') return EFTYPE; val = strtoul(str, &op, 0); if (*op != '\0') return EFTYPE; *result = val; return 0; } int resource_long_value(const char *name, int unit, const char *resname, long *result) { int error; const char *str; char *op; unsigned long val; int line; line = 0; error = resource_find(&line, NULL, name, &unit, resname, NULL, NULL, NULL, NULL, NULL, NULL, &str); if (error) return error; if (*str == '\0') return EFTYPE; val = strtoul(str, &op, 0); if (*op != '\0') return EFTYPE; *result = val; return 0; } int resource_string_value(const char *name, int unit, const char *resname, const char **result) { int error; const char *str; int line; line = 0; error = resource_find(&line, NULL, name, &unit, resname, NULL, NULL, NULL, NULL, NULL, NULL, &str); if (error) return error; *result = str; return 0; } /* * This is a bit nasty, but allows us to not modify the env strings. */ static const char * resource_string_copy(const char *s, int len) { static char stringbuf[256]; static int offset = 0; const char *ret; if (len == 0) len = strlen(s); if (len > 255) return NULL; if ((offset + len + 1) > 255) offset = 0; bcopy(s, &stringbuf[offset], len); stringbuf[offset + len] = '\0'; ret = &stringbuf[offset]; offset += len + 1; return ret; } /* * err = resource_find_match(&anchor, &name, &unit, resname, value) * Iteratively fetch a list of devices wired "at" something * res and value are restrictions. eg: "at", "scbus0". * For practical purposes, res = required, value = optional. * *name and *unit are set. * set *anchor to zero before starting. */ int resource_find_match(int *anchor, const char **name, int *unit, const char *resname, const char *value) { const char *found_name; int found_namelen; int found_unit; int ret; int newln; newln = *anchor; ret = resource_find(anchor, &newln, NULL, NULL, resname, value, &found_name, &found_namelen, &found_unit, NULL, NULL, NULL); if (ret == 0) { *name = resource_string_copy(found_name, found_namelen); *unit = found_unit; } *anchor = newln; return ret; } /* * err = resource_find_dev(&anchor, name, &unit, res, value); * Iterate through a list of devices, returning their unit numbers. * res and value are optional restrictions. eg: "at", "scbus0". * *unit is set to the value. * set *anchor to zero before starting. */ int resource_find_dev(int *anchor, const char *name, int *unit, const char *resname, const char *value) { int found_unit; int newln; int ret; newln = *anchor; ret = resource_find(anchor, &newln, name, NULL, resname, value, NULL, NULL, &found_unit, NULL, NULL, NULL); if (ret == 0) { *unit = found_unit; } *anchor = newln; return ret; } /* * Check to see if a device is disabled via a disabled hint. */ int resource_disabled(const char *name, int unit) { int error, value; error = resource_int_value(name, unit, "disabled", &value); if (error) return (0); return (value); } /* * Clear a value associated with a device by removing it from * the kernel environment. This only removes a hint for an * exact unit. */ int resource_unset_value(const char *name, int unit, const char *resname) { char varname[128]; const char *retname, *retvalue; int error, line; size_t len; line = 0; error = resource_find(&line, NULL, name, &unit, resname, NULL, &retname, NULL, NULL, NULL, NULL, &retvalue); if (error) return (error); retname -= strlen("hint."); len = retvalue - retname - 1; if (len > sizeof(varname) - 1) return (ENAMETOOLONG); memcpy(varname, retname, len); varname[len] = '\0'; return (kern_unsetenv(varname)); } Index: head/sys/sys/kenv.h =================================================================== --- head/sys/sys/kenv.h (revision 366507) +++ head/sys/sys/kenv.h (revision 366508) @@ -1,45 +1,59 @@ /*- * SPDX-License-Identifier: BSD-2-Clause-FreeBSD * * Copyright (c) 2002 Maxime Henrion * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * $FreeBSD$ */ #ifndef _SYS_KENV_H_ #define _SYS_KENV_H_ /* * Constants for the kenv(2) syscall */ #define KENV_GET 0 #define KENV_SET 1 #define KENV_UNSET 2 #define KENV_DUMP 3 #define KENV_MNAMELEN 128 /* Maximum name length (for the syscall) */ #define KENV_MVALLEN 128 /* Maximum value length (for the syscall) */ +#ifdef _KERNEL +/* + * Most of these variables should be const. + */ +extern bool dynamic_kenv; +extern struct mtx kenv_lock; +extern char *kern_envp; +extern char *md_envp; +extern char static_env[]; +extern char static_hints[]; /* by config for now */ + +extern char **kenvp; +#endif /* _KERNEL */ + #endif /* !_SYS_KENV_H_ */ Index: head/sys/sys/systm.h =================================================================== --- head/sys/sys/systm.h (revision 366507) +++ head/sys/sys/systm.h (revision 366508) @@ -1,649 +1,635 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1988, 1991, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. * All or some portions of this file are derived from material licensed * to the University of California by American Telephone and Telegraph * Co. or Unix System Laboratories, Inc. and are reproduced herein with * the permission of UNIX System Laboratories, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)systm.h 8.7 (Berkeley) 3/29/95 * $FreeBSD$ */ #ifndef _SYS_SYSTM_H_ #define _SYS_SYSTM_H_ #include #include #include #include #include #include /* for people using printf mainly */ __NULLABILITY_PRAGMA_PUSH #ifdef _KERNEL extern int cold; /* nonzero if we are doing a cold boot */ extern int suspend_blocked; /* block suspend due to pending shutdown */ extern int rebooting; /* kern_reboot() has been called. */ extern const char *panicstr; /* panic message */ extern bool panicked; #define KERNEL_PANICKED() __predict_false(panicked) extern char version[]; /* system version */ extern char compiler_version[]; /* compiler version */ extern char copyright[]; /* system copyright */ extern int kstack_pages; /* number of kernel stack pages */ extern u_long pagesizes[]; /* supported page sizes */ extern long physmem; /* physical memory */ extern long realmem; /* 'real' memory */ extern char *rootdevnames[2]; /* names of possible root devices */ extern int boothowto; /* reboot flags, from console subsystem */ extern int bootverbose; /* nonzero to print verbose messages */ extern int maxusers; /* system tune hint */ extern int ngroups_max; /* max # of supplemental groups */ extern int vm_guest; /* Running as virtual machine guest? */ /* * Detected virtual machine guest types. The intention is to expand * and/or add to the VM_GUEST_VM type if specific VM functionality is * ever implemented (e.g. vendor-specific paravirtualization features). * Keep in sync with vm_guest_sysctl_names[]. */ enum VM_GUEST { VM_GUEST_NO = 0, VM_GUEST_VM, VM_GUEST_XEN, VM_GUEST_HV, VM_GUEST_VMWARE, VM_GUEST_KVM, VM_GUEST_BHYVE, VM_GUEST_VBOX, VM_GUEST_PARALLELS, VM_LAST }; #ifdef INVARIANTS /* The option is always available */ #define VNASSERT(exp, vp, msg) do { \ if (__predict_false(!(exp))) { \ vn_printf(vp, "VNASSERT failed: %s not true at %s:%d (%s)\n",\ #exp, __FILE__, __LINE__, __func__); \ kassert_panic msg; \ } \ } while (0) #define VNPASS(exp, vp) do { \ const char *_exp = #exp; \ VNASSERT(exp, vp, ("condition %s not met at %s:%d (%s)", \ _exp, __FILE__, __LINE__, __func__)); \ } while (0) #define __assert_unreachable() \ panic("executing segment marked as unreachable at %s:%d (%s)\n", \ __FILE__, __LINE__, __func__) #else #define VNASSERT(exp, vp, msg) do { \ } while (0) #define VNPASS(exp, vp) do { \ } while (0) #define __assert_unreachable() __unreachable() #endif #ifndef CTASSERT /* Allow lint to override */ #define CTASSERT(x) _Static_assert(x, "compile-time assertion failed") #endif #endif /* KERNEL */ /* * These functions need to be declared before the KASSERT macro is invoked in * !KASSERT_PANIC_OPTIONAL builds, so their declarations are sort of out of * place compared to other function definitions in this header. On the other * hand, this header is a bit disorganized anyway. */ void panic(const char *, ...) __dead2 __printflike(1, 2); void vpanic(const char *, __va_list) __dead2 __printflike(1, 0); #if defined(_STANDALONE) /* * Until we have more experience with KASSERTS that are called * from the boot loader, they are off. The bootloader does this * a little differently than the kernel (we just call printf atm). * we avoid most of the common functions in the boot loader, so * declare printf() here too. */ int printf(const char *, ...) __printflike(1, 2); # define kassert_panic printf #else /* !_STANDALONE */ # if defined(WITNESS) || defined(INVARIANT_SUPPORT) # ifdef KASSERT_PANIC_OPTIONAL void kassert_panic(const char *fmt, ...) __printflike(1, 2); # else # define kassert_panic panic # endif /* KASSERT_PANIC_OPTIONAL */ # endif /* defined(WITNESS) || defined(INVARIANT_SUPPORT) */ #endif /* _STANDALONE */ #if defined(INVARIANTS) || defined(_STANDALONE) #define KASSERT(exp,msg) do { \ if (__predict_false(!(exp))) \ kassert_panic msg; \ } while (0) #else /* !INVARIANTS && !_STANDALONE */ #define KASSERT(exp,msg) do { \ } while (0) #endif /* INVARIANTS || _STANDALONE */ /* * Helpful macros for quickly coming up with assertions with informative * panic messages. */ #define MPASS(ex) MPASS4(ex, #ex, __FILE__, __LINE__) #define MPASS2(ex, what) MPASS4(ex, what, __FILE__, __LINE__) #define MPASS3(ex, file, line) MPASS4(ex, #ex, file, line) #define MPASS4(ex, what, file, line) \ KASSERT((ex), ("Assertion %s failed at %s:%d", what, file, line)) /* * Align variables. */ #define __read_mostly __section(".data.read_mostly") #define __read_frequently __section(".data.read_frequently") #define __exclusive_cache_line __aligned(CACHE_LINE_SIZE) \ __section(".data.exclusive_cache_line") #ifdef _KERNEL #include /* MAXCPU */ #include /* curthread */ #include /* * Assert that a pointer can be loaded from memory atomically. * * This assertion enforces stronger alignment than necessary. For example, * on some architectures, atomicity for unaligned loads will depend on * whether or not the load spans multiple cache lines. */ #define ASSERT_ATOMIC_LOAD_PTR(var, msg) \ KASSERT(sizeof(var) == sizeof(void *) && \ ((uintptr_t)&(var) & (sizeof(void *) - 1)) == 0, msg) /* * Assert that a thread is in critical(9) section. */ #define CRITICAL_ASSERT(td) \ KASSERT((td)->td_critnest >= 1, ("Not in critical section")); /* * If we have already panic'd and this is the thread that called * panic(), then don't block on any mutexes but silently succeed. * Otherwise, the kernel will deadlock since the scheduler isn't * going to run the thread that holds any lock we need. */ #define SCHEDULER_STOPPED_TD(td) ({ \ MPASS((td) == curthread); \ __predict_false((td)->td_stopsched); \ }) #define SCHEDULER_STOPPED() SCHEDULER_STOPPED_TD(curthread) -/* - * XXX the hints declarations are even more misplaced than most declarations - * in this file, since they are needed in one file (per arch) and only used - * in two files. - * XXX most of these variables should be const. - */ extern int osreldate; -extern bool dynamic_kenv; -extern struct mtx kenv_lock; -extern char *kern_envp; -extern char *md_envp; -extern char static_env[]; -extern char static_hints[]; /* by config for now */ - -extern char **kenvp; extern const void *zero_region; /* address space maps to a zeroed page */ extern int unmapped_buf_allowed; #ifdef __LP64__ #define IOSIZE_MAX iosize_max() #define DEVFS_IOSIZE_MAX devfs_iosize_max() #else #define IOSIZE_MAX SSIZE_MAX #define DEVFS_IOSIZE_MAX SSIZE_MAX #endif /* * General function declarations. */ struct inpcb; struct lock_object; struct malloc_type; struct mtx; struct proc; struct socket; struct thread; struct tty; struct ucred; struct uio; struct _jmp_buf; struct trapframe; struct eventtimer; int setjmp(struct _jmp_buf *) __returns_twice; void longjmp(struct _jmp_buf *, int) __dead2; int dumpstatus(vm_offset_t addr, off_t count); int nullop(void); int eopnotsupp(void); int ureadc(int, struct uio *); void hashdestroy(void *, struct malloc_type *, u_long); void *hashinit(int count, struct malloc_type *type, u_long *hashmask); void *hashinit_flags(int count, struct malloc_type *type, u_long *hashmask, int flags); #define HASH_NOWAIT 0x00000001 #define HASH_WAITOK 0x00000002 void *phashinit(int count, struct malloc_type *type, u_long *nentries); void *phashinit_flags(int count, struct malloc_type *type, u_long *nentries, int flags); void g_waitidle(void); void cpu_flush_dcache(void *, size_t); void cpu_rootconf(void); void critical_enter_KBI(void); void critical_exit_KBI(void); void critical_exit_preempt(void); void init_param1(void); void init_param2(long physpages); void init_static_kenv(char *, size_t); void tablefull(const char *); /* * Allocate per-thread "current" state in the linuxkpi */ extern int (*lkpi_alloc_current)(struct thread *, int); int linux_alloc_current_noop(struct thread *, int); #if defined(KLD_MODULE) || defined(KTR_CRITICAL) || !defined(_KERNEL) || defined(GENOFFSET) #define critical_enter() critical_enter_KBI() #define critical_exit() critical_exit_KBI() #else static __inline void critical_enter(void) { struct thread_lite *td; td = (struct thread_lite *)curthread; td->td_critnest++; __compiler_membar(); } static __inline void critical_exit(void) { struct thread_lite *td; td = (struct thread_lite *)curthread; KASSERT(td->td_critnest != 0, ("critical_exit: td_critnest == 0")); __compiler_membar(); td->td_critnest--; __compiler_membar(); if (__predict_false(td->td_owepreempt)) critical_exit_preempt(); } #endif #ifdef EARLY_PRINTF typedef void early_putc_t(int ch); extern early_putc_t *early_putc; #endif int kvprintf(char const *, void (*)(int, void*), void *, int, __va_list) __printflike(1, 0); void log(int, const char *, ...) __printflike(2, 3); void log_console(struct uio *); void vlog(int, const char *, __va_list) __printflike(2, 0); int asprintf(char **ret, struct malloc_type *mtp, const char *format, ...) __printflike(3, 4); int printf(const char *, ...) __printflike(1, 2); int snprintf(char *, size_t, const char *, ...) __printflike(3, 4); int sprintf(char *buf, const char *, ...) __printflike(2, 3); int uprintf(const char *, ...) __printflike(1, 2); int vprintf(const char *, __va_list) __printflike(1, 0); int vasprintf(char **ret, struct malloc_type *mtp, const char *format, __va_list ap) __printflike(3, 0); int vsnprintf(char *, size_t, const char *, __va_list) __printflike(3, 0); int vsnrprintf(char *, size_t, int, const char *, __va_list) __printflike(4, 0); int vsprintf(char *buf, const char *, __va_list) __printflike(2, 0); int sscanf(const char *, char const * _Nonnull, ...) __scanflike(2, 3); int vsscanf(const char * _Nonnull, char const * _Nonnull, __va_list) __scanflike(2, 0); long strtol(const char *, char **, int); u_long strtoul(const char *, char **, int); quad_t strtoq(const char *, char **, int); u_quad_t strtouq(const char *, char **, int); void tprintf(struct proc *p, int pri, const char *, ...) __printflike(3, 4); void vtprintf(struct proc *, int, const char *, __va_list) __printflike(3, 0); void hexdump(const void *ptr, int length, const char *hdr, int flags); #define HD_COLUMN_MASK 0xff #define HD_DELIM_MASK 0xff00 #define HD_OMIT_COUNT (1 << 16) #define HD_OMIT_HEX (1 << 17) #define HD_OMIT_CHARS (1 << 18) #define ovbcopy(f, t, l) bcopy((f), (t), (l)) void bcopy(const void * _Nonnull from, void * _Nonnull to, size_t len); void bzero(void * _Nonnull buf, size_t len); void explicit_bzero(void * _Nonnull, size_t); int bcmp(const void *b1, const void *b2, size_t len); void *memset(void * _Nonnull buf, int c, size_t len); void *memcpy(void * _Nonnull to, const void * _Nonnull from, size_t len); void *memmove(void * _Nonnull dest, const void * _Nonnull src, size_t n); int memcmp(const void *b1, const void *b2, size_t len); #ifdef KCSAN void *kcsan_memset(void *, int, size_t); void *kcsan_memcpy(void *, const void *, size_t); void *kcsan_memmove(void *, const void *, size_t); int kcsan_memcmp(const void *, const void *, size_t); #define bcopy(from, to, len) kcsan_memmove((to), (from), (len)) #define bzero(buf, len) kcsan_memset((buf), 0, (len)) #define bcmp(b1, b2, len) kcsan_memcmp((b1), (b2), (len)) #define memset(buf, c, len) kcsan_memset((buf), (c), (len)) #define memcpy(to, from, len) kcsan_memcpy((to), (from), (len)) #define memmove(dest, src, n) kcsan_memmove((dest), (src), (n)) #define memcmp(b1, b2, len) kcsan_memcmp((b1), (b2), (len)) #else #define bcopy(from, to, len) __builtin_memmove((to), (from), (len)) #define bzero(buf, len) __builtin_memset((buf), 0, (len)) #define bcmp(b1, b2, len) __builtin_memcmp((b1), (b2), (len)) #define memset(buf, c, len) __builtin_memset((buf), (c), (len)) #define memcpy(to, from, len) __builtin_memcpy((to), (from), (len)) #define memmove(dest, src, n) __builtin_memmove((dest), (src), (n)) #define memcmp(b1, b2, len) __builtin_memcmp((b1), (b2), (len)) #endif void *memset_early(void * _Nonnull buf, int c, size_t len); #define bzero_early(buf, len) memset_early((buf), 0, (len)) void *memcpy_early(void * _Nonnull to, const void * _Nonnull from, size_t len); void *memmove_early(void * _Nonnull dest, const void * _Nonnull src, size_t n); #define bcopy_early(from, to, len) memmove_early((to), (from), (len)) #define copystr(src, dst, len, outlen) ({ \ size_t __r, __len, *__outlen; \ \ __len = (len); \ __outlen = (outlen); \ __r = strlcpy((dst), (src), __len); \ if (__outlen != NULL) \ *__outlen = ((__r >= __len) ? __len : __r + 1); \ ((__r >= __len) ? ENAMETOOLONG : 0); \ }) int copyinstr(const void * __restrict udaddr, void * _Nonnull __restrict kaddr, size_t len, size_t * __restrict lencopied); int copyin(const void * __restrict udaddr, void * _Nonnull __restrict kaddr, size_t len); int copyin_nofault(const void * __restrict udaddr, void * _Nonnull __restrict kaddr, size_t len); int copyout(const void * _Nonnull __restrict kaddr, void * __restrict udaddr, size_t len); int copyout_nofault(const void * _Nonnull __restrict kaddr, void * __restrict udaddr, size_t len); #ifdef KCSAN int kcsan_copyin(const void *, void *, size_t); int kcsan_copyinstr(const void *, void *, size_t, size_t *); int kcsan_copyout(const void *, void *, size_t); #define copyin(u, k, l) kcsan_copyin((u), (k), (l)) #define copyinstr(u, k, l, lc) kcsan_copyinstr((u), (k), (l), (lc)) #define copyout(k, u, l) kcsan_copyout((k), (u), (l)) #endif int fubyte(volatile const void *base); long fuword(volatile const void *base); int fuword16(volatile const void *base); int32_t fuword32(volatile const void *base); int64_t fuword64(volatile const void *base); int fueword(volatile const void *base, long *val); int fueword32(volatile const void *base, int32_t *val); int fueword64(volatile const void *base, int64_t *val); int subyte(volatile void *base, int byte); int suword(volatile void *base, long word); int suword16(volatile void *base, int word); int suword32(volatile void *base, int32_t word); int suword64(volatile void *base, int64_t word); uint32_t casuword32(volatile uint32_t *base, uint32_t oldval, uint32_t newval); u_long casuword(volatile u_long *p, u_long oldval, u_long newval); int casueword32(volatile uint32_t *base, uint32_t oldval, uint32_t *oldvalp, uint32_t newval); int casueword(volatile u_long *p, u_long oldval, u_long *oldvalp, u_long newval); void realitexpire(void *); int sysbeep(int hertz, int period); void hardclock(int cnt, int usermode); void hardclock_sync(int cpu); void softclock(void *); void statclock(int cnt, int usermode); void profclock(int cnt, int usermode, uintfptr_t pc); int hardclockintr(void); void startprofclock(struct proc *); void stopprofclock(struct proc *); void cpu_startprofclock(void); void cpu_stopprofclock(void); void suspendclock(void); void resumeclock(void); sbintime_t cpu_idleclock(void); void cpu_activeclock(void); void cpu_new_callout(int cpu, sbintime_t bt, sbintime_t bt_opt); void cpu_et_frequency(struct eventtimer *et, uint64_t newfreq); extern int cpu_disable_c2_sleep; extern int cpu_disable_c3_sleep; char *kern_getenv(const char *name); void freeenv(char *env); int getenv_int(const char *name, int *data); int getenv_uint(const char *name, unsigned int *data); int getenv_long(const char *name, long *data); int getenv_ulong(const char *name, unsigned long *data); int getenv_string(const char *name, char *data, int size); int getenv_int64(const char *name, int64_t *data); int getenv_uint64(const char *name, uint64_t *data); int getenv_quad(const char *name, quad_t *data); int getenv_bool(const char *name, bool *data); bool getenv_is_true(const char *name); bool getenv_is_false(const char *name); int kern_setenv(const char *name, const char *value); int kern_unsetenv(const char *name); int testenv(const char *name); int getenv_array(const char *name, void *data, int size, int *psize, int type_size, bool allow_signed); #define GETENV_UNSIGNED false /* negative numbers not allowed */ #define GETENV_SIGNED true /* negative numbers allowed */ typedef uint64_t (cpu_tick_f)(void); void set_cputicker(cpu_tick_f *func, uint64_t freq, unsigned var); extern cpu_tick_f *cpu_ticks; uint64_t cpu_tickrate(void); uint64_t cputick2usec(uint64_t tick); #ifdef APM_FIXUP_CALLTODO struct timeval; void adjust_timeout_calltodo(struct timeval *time_change); #endif /* APM_FIXUP_CALLTODO */ #include /* Initialize the world */ void consinit(void); void cpu_initclocks(void); void cpu_initclocks_bsp(void); void cpu_initclocks_ap(void); void usrinfoinit(void); /* Finalize the world */ void kern_reboot(int) __dead2; void shutdown_nice(int); /* Stubs for obsolete functions that used to be for interrupt management */ static __inline intrmask_t splhigh(void) { return 0; } static __inline intrmask_t splimp(void) { return 0; } static __inline intrmask_t splnet(void) { return 0; } static __inline intrmask_t spltty(void) { return 0; } static __inline void splx(intrmask_t ipl __unused) { return; } /* * Common `proc' functions are declared here so that proc.h can be included * less often. */ int _sleep(const void * _Nonnull chan, struct lock_object *lock, int pri, const char *wmesg, sbintime_t sbt, sbintime_t pr, int flags); #define msleep(chan, mtx, pri, wmesg, timo) \ _sleep((chan), &(mtx)->lock_object, (pri), (wmesg), \ tick_sbt * (timo), 0, C_HARDCLOCK) #define msleep_sbt(chan, mtx, pri, wmesg, bt, pr, flags) \ _sleep((chan), &(mtx)->lock_object, (pri), (wmesg), (bt), (pr), \ (flags)) int msleep_spin_sbt(const void * _Nonnull chan, struct mtx *mtx, const char *wmesg, sbintime_t sbt, sbintime_t pr, int flags); #define msleep_spin(chan, mtx, wmesg, timo) \ msleep_spin_sbt((chan), (mtx), (wmesg), tick_sbt * (timo), \ 0, C_HARDCLOCK) int pause_sbt(const char *wmesg, sbintime_t sbt, sbintime_t pr, int flags); #define pause(wmesg, timo) \ pause_sbt((wmesg), tick_sbt * (timo), 0, C_HARDCLOCK) #define pause_sig(wmesg, timo) \ pause_sbt((wmesg), tick_sbt * (timo), 0, C_HARDCLOCK | C_CATCH) #define tsleep(chan, pri, wmesg, timo) \ _sleep((chan), NULL, (pri), (wmesg), tick_sbt * (timo), \ 0, C_HARDCLOCK) #define tsleep_sbt(chan, pri, wmesg, bt, pr, flags) \ _sleep((chan), NULL, (pri), (wmesg), (bt), (pr), (flags)) void wakeup(const void *chan); void wakeup_one(const void *chan); void wakeup_any(const void *chan); /* * Common `struct cdev *' stuff are declared here to avoid #include poisoning */ struct cdev; dev_t dev2udev(struct cdev *x); const char *devtoname(struct cdev *cdev); #ifdef __LP64__ size_t devfs_iosize_max(void); size_t iosize_max(void); #endif int poll_no_poll(int events); /* XXX: Should be void nanodelay(u_int nsec); */ void DELAY(int usec); /* Root mount holdback API */ struct root_hold_token { int flags; const char *who; TAILQ_ENTRY(root_hold_token) list; }; struct root_hold_token *root_mount_hold(const char *identifier); void root_mount_hold_token(const char *identifier, struct root_hold_token *h); void root_mount_rel(struct root_hold_token *h); int root_mounted(void); /* * Unit number allocation API. (kern/subr_unit.c) */ struct unrhdr; struct unrhdr *new_unrhdr(int low, int high, struct mtx *mutex); void init_unrhdr(struct unrhdr *uh, int low, int high, struct mtx *mutex); void delete_unrhdr(struct unrhdr *uh); void clear_unrhdr(struct unrhdr *uh); void clean_unrhdr(struct unrhdr *uh); void clean_unrhdrl(struct unrhdr *uh); int alloc_unr(struct unrhdr *uh); int alloc_unr_specific(struct unrhdr *uh, u_int item); int alloc_unrl(struct unrhdr *uh); void free_unr(struct unrhdr *uh, u_int item); #ifndef __LP64__ #define UNR64_LOCKED #endif struct unrhdr64 { uint64_t counter; }; static __inline void new_unrhdr64(struct unrhdr64 *unr64, uint64_t low) { unr64->counter = low; } #ifdef UNR64_LOCKED uint64_t alloc_unr64(struct unrhdr64 *); #else static __inline uint64_t alloc_unr64(struct unrhdr64 *unr64) { return (atomic_fetchadd_64(&unr64->counter, 1)); } #endif void intr_prof_stack_use(struct thread *td, struct trapframe *frame); void counted_warning(unsigned *counter, const char *msg); /* * APIs to manage deprecation and obsolescence. */ struct device; void _gone_in(int major, const char *msg); void _gone_in_dev(struct device *dev, int major, const char *msg); #ifdef NO_OBSOLETE_CODE #define __gone_ok(m, msg) \ _Static_assert(m < P_OSREL_MAJOR(__FreeBSD_version)), \ "Obsolete code: " msg); #else #define __gone_ok(m, msg) #endif #define gone_in(major, msg) __gone_ok(major, msg) _gone_in(major, msg) #define gone_in_dev(dev, major, msg) __gone_ok(major, msg) _gone_in_dev(dev, major, msg) #endif /* _KERNEL */ __NULLABILITY_PRAGMA_POP #endif /* !_SYS_SYSTM_H_ */