diff options
Diffstat (limited to 'sys')
70 files changed, 1876 insertions, 501 deletions
diff --git a/sys/arm64/arm64/elf32_machdep.c b/sys/arm64/arm64/elf32_machdep.c index 5c81c6cdce3d..8f8a934ad520 100644 --- a/sys/arm64/arm64/elf32_machdep.c +++ b/sys/arm64/arm64/elf32_machdep.c @@ -225,7 +225,7 @@ freebsd32_fetch_syscall_args(struct thread *td) sa->args[i] = ap[i]; if (narg > nap) { if (narg - nap > nitems(args)) - panic("Too many system call arguiments"); + panic("Too many system call arguments"); error = copyin((void *)td->td_frame->tf_x[13], args, (narg - nap) * sizeof(int)); if (error != 0) diff --git a/sys/arm64/arm64/trap.c b/sys/arm64/arm64/trap.c index bed58095201a..75c9b5f87892 100644 --- a/sys/arm64/arm64/trap.c +++ b/sys/arm64/arm64/trap.c @@ -246,6 +246,7 @@ external_abort(struct thread *td, struct trapframe *frame, uint64_t esr, print_registers(frame); print_gp_register("far", far); + printf(" esr: 0x%.16lx\n", esr); panic("Unhandled external data abort"); } diff --git a/sys/compat/linuxkpi/common/include/acpi/acpi.h b/sys/compat/linuxkpi/common/include/acpi/acpi.h index 1e398d05ba20..016c7ede0f6e 100644 --- a/sys/compat/linuxkpi/common/include/acpi/acpi.h +++ b/sys/compat/linuxkpi/common/include/acpi/acpi.h @@ -37,7 +37,7 @@ /* * LINUXKPI_WANT_LINUX_ACPI is a temporary workaround to allow drm-kmod * to update all needed branches without breaking builds. - * Once that happened and checks are implemented based on __FreeBSD_verison + * Once that happened and checks are implemented based on __FreeBSD_version * we will remove these conditions again. */ diff --git a/sys/compat/linuxkpi/common/include/linux/ieee80211.h b/sys/compat/linuxkpi/common/include/linux/ieee80211.h index b9161c586d07..33850359869f 100644 --- a/sys/compat/linuxkpi/common/include/linux/ieee80211.h +++ b/sys/compat/linuxkpi/common/include/linux/ieee80211.h @@ -524,24 +524,24 @@ struct ieee80211_mgmt { uint16_t beacon_int; uint16_t capab_info; uint8_t variable[0]; - } beacon; + } __packed beacon; /* 9.3.3.5 Association Request frame format */ struct { uint16_t capab_info; uint16_t listen_interval; uint8_t variable[0]; - } assoc_req; + } __packed assoc_req; /* 9.3.3.10 Probe Request frame format */ struct { uint8_t variable[0]; - } probe_req; + } __packed probe_req; /* 9.3.3.11 Probe Response frame format */ struct { uint64_t timestamp; uint16_t beacon_int; uint16_t capab_info; uint8_t variable[0]; - } probe_resp; + } __packed probe_resp; /* 9.3.3.14 Action frame format */ struct { /* 9.4.1.11 Action field */ @@ -557,7 +557,7 @@ struct ieee80211_mgmt { uint8_t tpc_elem_length; uint8_t tpc_elem_tx_power; uint8_t tpc_elem_link_margin; - } tpc_report; + } __packed tpc_report; /* 9.6.8.33 Fine Timing Measurement frame format */ struct { uint8_t dialog_token; @@ -567,7 +567,7 @@ struct ieee80211_mgmt { uint16_t tod_error; uint16_t toa_error; uint8_t variable[0]; - } ftm; + } __packed ftm; /* 802.11-2016, 9.6.5.2 ADDBA Request frame format */ struct { uint8_t action_code; @@ -577,16 +577,16 @@ struct ieee80211_mgmt { uint16_t start_seq_num; /* Optional follows... */ uint8_t variable[0]; - } addba_req; + } __packed addba_req; /* XXX */ struct { uint8_t dialog_token; - } wnm_timing_msr; + } __packed wnm_timing_msr; } u; - } action; + } __packed action; DECLARE_FLEX_ARRAY(uint8_t, body); } u; -}; +} __packed __aligned(2); struct ieee80211_cts { /* net80211::ieee80211_frame_cts */ __le16 frame_control; diff --git a/sys/compat/linuxkpi/common/include/linux/skbuff.h b/sys/compat/linuxkpi/common/include/linux/skbuff.h index c8ad90281e34..0f192ceab318 100644 --- a/sys/compat/linuxkpi/common/include/linux/skbuff.h +++ b/sys/compat/linuxkpi/common/include/linux/skbuff.h @@ -120,7 +120,7 @@ enum sk_checksum_flags { CHECKSUM_NONE = 0x00, CHECKSUM_UNNECESSARY = 0x01, CHECKSUM_PARTIAL = 0x02, - CHECKSUM_COMPLETE = 0x04, + CHECKSUM_COMPLETE = 0x03, }; struct skb_frag { @@ -170,7 +170,7 @@ struct sk_buff { }; }; uint16_t protocol; - uint8_t ip_summed; + uint8_t ip_summed; /* 2 bit only. */ /* uint8_t */ /* "Scratch" area for layers to store metadata. */ diff --git a/sys/compat/linuxkpi/common/include/linux/string_helpers.h b/sys/compat/linuxkpi/common/include/linux/string_helpers.h index 1bdff2730361..2c6fe0b1708d 100644 --- a/sys/compat/linuxkpi/common/include/linux/string_helpers.h +++ b/sys/compat/linuxkpi/common/include/linux/string_helpers.h @@ -66,4 +66,6 @@ str_enable_disable(bool value) return "disable"; } +#define str_disable_enable(_v) str_enable_disable(!(_v)) + #endif diff --git a/sys/compat/linuxkpi/common/src/linux_80211.c b/sys/compat/linuxkpi/common/src/linux_80211.c index e248588dd275..3f850653bc7c 100644 --- a/sys/compat/linuxkpi/common/src/linux_80211.c +++ b/sys/compat/linuxkpi/common/src/linux_80211.c @@ -77,6 +77,8 @@ #include <linux/rculist.h> #include "linux_80211.h" +/* #define LKPI_80211_USE_SCANLIST */ +/* #define LKPI_80211_BGSCAN */ #define LKPI_80211_WME #define LKPI_80211_HW_CRYPTO #define LKPI_80211_HT @@ -103,6 +105,10 @@ SYSCTL_DECL(_compat_linuxkpi); SYSCTL_NODE(_compat_linuxkpi, OID_AUTO, 80211, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "LinuxKPI 802.11 compatibility layer"); +static bool lkpi_order_scanlist = false; +SYSCTL_BOOL(_compat_linuxkpi_80211, OID_AUTO, order_scanlist, CTLFLAG_RW, + &lkpi_order_scanlist, 0, "Enable LinuxKPI 802.11 scan list shuffeling"); + #if defined(LKPI_80211_HW_CRYPTO) static bool lkpi_hwcrypto = false; SYSCTL_BOOL(_compat_linuxkpi_80211, OID_AUTO, hw_crypto, CTLFLAG_RDTUN, @@ -167,6 +173,7 @@ const struct cfg80211_ops linuxkpi_mac80211cfgops = { static struct lkpi_sta *lkpi_find_lsta_by_ni(struct lkpi_vif *, struct ieee80211_node *); #endif +static void lkpi_sw_scan_task(void *, int); static void lkpi_80211_txq_tx_one(struct lkpi_sta *, struct mbuf *); static void lkpi_80211_txq_task(void *, int); static void lkpi_80211_lhw_rxq_task(void *, int); @@ -955,6 +962,30 @@ lkpi_nl80211_band_to_net80211_band(enum nl80211_band band) return (0x00); } +#ifdef LINUXKPI_DEBUG_80211 +static const char * +lkpi_nl80211_band_name(enum nl80211_band band) +{ + switch (band) { + case NL80211_BAND_2GHZ: + return "2Ghz"; + break; + case NL80211_BAND_5GHZ: + return "5Ghz"; + break; + case NL80211_BAND_60GHZ: + return "60Ghz"; + break; + case NL80211_BAND_6GHZ: + return "6Ghz"; + break; + default: + panic("%s: unsupported band %u\n", __func__, band); + break; + } +} +#endif + #if 0 static enum ieee80211_ac_numbers lkpi_ac_net_to_l80211(int ac) @@ -1319,6 +1350,7 @@ lkpi_iv_key_delete(struct ieee80211vap *vap, const struct ieee80211_key *k) lhw = ic->ic_softc; hw = LHW_TO_HW(lhw); lvif = VAP_TO_LVIF(vap); + vif = LVIF_TO_VIF(lvif); /* * Make sure we do not make it here without going through @@ -1326,6 +1358,23 @@ lkpi_iv_key_delete(struct ieee80211vap *vap, const struct ieee80211_key *k) */ lockdep_assert_wiphy(hw->wiphy); + /* + * While we are assoc we may still send packets. We cannot delete the + * keys as otherwise packets could go out unencrypted. Some firmware + * does not like this and will fire an assert. + * net80211 needs to drive this better but given we want the disassoc + * frame out and have to unlock we are open to a race currently. + * This check should prevent problems. + * How to test: run 800Mbit/s UDP traffic and during that restart your + * supplicant. You want to survive that. + */ + if (vif->cfg.assoc) { + if (linuxkpi_debug_80211 & D80211_TRACE_HW_CRYPTO) + ic_printf(ic, "%d %lu %s: vif still assoc; not deleting keys\n", + curthread->td_tid, jiffies, __func__); + return (0); + } + if (IEEE80211_KEY_UNDEFINED(k)) { ic_printf(ic, "%s: vap %p key %p is undefined: %p %u\n", __func__, vap, k, k->wk_cipher, k->wk_keyix); @@ -1370,7 +1419,6 @@ lkpi_iv_key_delete(struct ieee80211vap *vap, const struct ieee80211_key *k) kc->keyidx, kc->hw_key_idx, kc->flags, IEEE80211_KEY_FLAG_BITS); #endif - vif = LVIF_TO_VIF(lvif); error = lkpi_80211_mo_set_key(hw, DISABLE_KEY, vif, sta, kc); if (error != 0) { ic_printf(ic, "%d %lu %s: set_key cmd %d(%s) for sta %6D failed: %d\n", @@ -1842,13 +1890,13 @@ lkpi_update_dtim_tsf(struct ieee80211_vif *vif, struct ieee80211_node *ni, vif->bss_conf.beacon_int = 16; bss_changed |= BSS_CHANGED_BEACON_INT; } - if (vif->bss_conf.dtim_period != vap->iv_dtim_period && - vap->iv_dtim_period > 0) { - vif->bss_conf.dtim_period = vap->iv_dtim_period; + if (vif->bss_conf.dtim_period != ni->ni_dtim_period && + ni->ni_dtim_period > 0) { + vif->bss_conf.dtim_period = ni->ni_dtim_period; bss_changed |= BSS_CHANGED_BEACON_INFO; } - vif->bss_conf.sync_dtim_count = vap->iv_dtim_count; + vif->bss_conf.sync_dtim_count = ni->ni_dtim_count; vif->bss_conf.sync_tsf = le64toh(ni->ni_tstamp.tsf); /* vif->bss_conf.sync_device_ts = set in linuxkpi_ieee80211_rx. */ @@ -1876,6 +1924,8 @@ lkpi_stop_hw_scan(struct lkpi_hw *lhw, struct ieee80211_vif *vif) int error; bool cancel; + TRACE_SCAN(lhw->ic, "scan_flags %b", lhw->scan_flags, LKPI_LHW_SCAN_BITS); + LKPI_80211_LHW_SCAN_LOCK(lhw); cancel = (lhw->scan_flags & LKPI_LHW_SCAN_RUNNING) != 0; LKPI_80211_LHW_SCAN_UNLOCK(lhw); @@ -2798,6 +2848,14 @@ _lkpi_sta_assoc_to_down(struct ieee80211vap *vap, enum ieee80211_state nstate, i bss_changed = 0; bss_changed |= lkpi_disassoc(sta, vif, lhw); +#ifdef LKPI_80211_HW_CRYPTO + /* + * In theory we remove keys here but there must not exist any for this + * state change until we clean them up again into small steps and no + * code duplication. + */ +#endif + lkpi_lsta_dump(lsta, ni, __func__, __LINE__); /* Adjust sta and change state (from NONE) to NOTEXIST. */ @@ -3333,6 +3391,16 @@ lkpi_sta_run_to_init(struct ieee80211vap *vap, enum ieee80211_state nstate, int #ifdef LKPI_80211_HW_CRYPTO if (lkpi_hwcrypto) { + /* + * In theory we only need to do this if we changed assoc. + * If we were not assoc, there should be no keys and we + * should not be here. + */ +#ifdef notyet + KASSERT((bss_changed & BSS_CHANGED_ASSOC) != 0, ("%s: " + "trying to remove keys but were not assoc: %#010jx, lvif %p\n", + __func__, (uintmax_t)bss_changed, lvif)); +#endif error = lkpi_sta_del_keys(hw, vif, lsta); if (error != 0) { ic_printf(vap->iv_ic, "%s:%d: lkpi_sta_del_keys " @@ -3394,6 +3462,9 @@ lkpi_sta_run_to_init(struct ieee80211vap *vap, enum ieee80211_state nstate, int * 4) call unassign_vif_chanctx * 5) call lkpi_hw_conf_idle * 6) call remove_chanctx + * + * Note: vif->driver_flags & IEEE80211_VIF_REMOVE_AP_AFTER_DISASSOC + * might change this. */ bss_changed |= lkpi_disassoc(sta, vif, lhw); @@ -3545,7 +3616,7 @@ lkpi_iv_newstate(struct ieee80211vap *vap, enum ieee80211_state nstate, int arg) vif = LVIF_TO_VIF(lvif); /* No need to replicate this in most state handlers. */ - if (ostate == IEEE80211_S_SCAN && nstate != IEEE80211_S_SCAN) + if (nstate > IEEE80211_S_SCAN) lkpi_stop_hw_scan(lhw, vif); s = sta_state_fsm; @@ -3824,6 +3895,7 @@ lkpi_ic_vap_create(struct ieee80211com *ic, const char name[IFNAMSIZ], lvif = malloc(len, M_80211_VAP, M_WAITOK | M_ZERO); mtx_init(&lvif->mtx, "lvif", NULL, MTX_DEF); + TASK_INIT(&lvif->sw_scan_task, 0, lkpi_sw_scan_task, lvif); INIT_LIST_HEAD(&lvif->lsta_list); lvif->lvif_bss = NULL; refcount_init(&lvif->nt_unlocked, 0); @@ -4062,6 +4134,8 @@ lkpi_ic_vap_delete(struct ieee80211vap *vap) /* Clear up per-VIF/VAP sysctls. */ sysctl_ctx_free(&lvif->sysctl_ctx); + ieee80211_draintask(ic, &lvif->sw_scan_task); + LKPI_80211_LHW_LVIF_LOCK(lhw); TAILQ_REMOVE(&lhw->lvif_head, lvif, lvif_entry); LKPI_80211_LHW_LVIF_UNLOCK(lhw); @@ -4303,6 +4377,113 @@ lkpi_scan_ies_add(uint8_t *p, struct ieee80211_scan_ies *scan_ies, } static void +lkpi_enable_hw_scan(struct lkpi_hw *lhw) +{ + + if (lhw->ops->hw_scan) { + /* + * Advertise full-offload scanning. + * + * Not limiting to SINGLE_SCAN_ON_ALL_BANDS here as otherwise + * we essentially disable hw_scan for all drivers not setting + * the flag. + */ + lhw->ic->ic_flags_ext |= IEEE80211_FEXT_SCAN_OFFLOAD; + lhw->scan_flags |= LKPI_LHW_SCAN_HW; + } +} + +#ifndef LKPI_80211_USE_SCANLIST +static const uint32_t chan_pri[] = { + 5180, 5500, 5745, + 5260, 5580, 5660, 5825, + 5220, 5300, 5540, 5620, 5700, 5785, 5865, + 2437, 2412, 2422, 2462, 2472, 2432, 2452 +}; + +static int +lkpi_scan_chan_list_idx(const struct linuxkpi_ieee80211_channel *lc) +{ + int i; + + for (i = 0; i < nitems(chan_pri); i++) { + if (lc->center_freq == chan_pri[i]) + return (i); + } + + return (-1); +} + +static int +lkpi_scan_chan_list_comp(const struct linuxkpi_ieee80211_channel *lc1, + const struct linuxkpi_ieee80211_channel *lc2) +{ + int idx1, idx2; + + /* Find index in list. */ + idx1 = lkpi_scan_chan_list_idx(lc1); + idx2 = lkpi_scan_chan_list_idx(lc2); + + if (idx1 == -1 && idx2 != -1) + return (1); + if (idx1 != -1 && idx2 == -1) + return (-1); + + /* Neither on the list, use center_freq. */ + if (idx1 == -1 && idx2 == -1) + return (lc1->center_freq - lc2->center_freq); + + /* Whichever is first in the list. */ + return (idx1 - idx2); +} + +static void +lkpi_scan_chan_list_resort(struct linuxkpi_ieee80211_channel **cpp, size_t nchan) +{ + struct linuxkpi_ieee80211_channel *lc, *nc; + size_t i, j; + int rc; + + for (i = (nchan - 1); i > 0; i--) { + for (j = i; j > 0 ; j--) { + lc = *(cpp + j); + nc = *(cpp + j - 1); + rc = lkpi_scan_chan_list_comp(lc, nc); + if (rc < 0) { + *(cpp + j) = nc; + *(cpp + j - 1) = lc; + } + } + } +} + +static bool +lkpi_scan_chan(struct linuxkpi_ieee80211_channel *c, + struct ieee80211com *ic, bool log) +{ + + if ((c->flags & IEEE80211_CHAN_DISABLED) != 0) { + if (log) + TRACE_SCAN(ic, "Skipping disabled chan " + "on band %s [%#x/%u/%#x]", + lkpi_nl80211_band_name(c->band), c->hw_value, + c->center_freq, c->flags); + return (false); + } + if (isclr(ic->ic_chan_active, ieee80211_mhz2ieee(c->center_freq, + lkpi_nl80211_band_to_net80211_band(c->band)))) { + if (log) + TRACE_SCAN(ic, "Skipping !active chan " + "on band %s [%#x/%u/%#x]", + lkpi_nl80211_band_name(c->band), c->hw_value, + c->center_freq, c->flags); + return (false); + } + return (true); +} +#endif + +static void lkpi_ic_scan_start(struct ieee80211com *ic) { struct lkpi_hw *lhw; @@ -4315,35 +4496,52 @@ lkpi_ic_scan_start(struct ieee80211com *ic) bool is_hw_scan; lhw = ic->ic_softc; + ss = ic->ic_scan; + vap = ss->ss_vap; + TRACE_SCAN(ic, "scan_flags %b", lhw->scan_flags, LKPI_LHW_SCAN_BITS); + LKPI_80211_LHW_SCAN_LOCK(lhw); if ((lhw->scan_flags & LKPI_LHW_SCAN_RUNNING) != 0) { /* A scan is still running. */ LKPI_80211_LHW_SCAN_UNLOCK(lhw); + TRACE_SCAN(ic, "Trying to start new scan while still running; " + "cancelling new net80211 scan; scan_flags %b", + lhw->scan_flags, LKPI_LHW_SCAN_BITS); + ieee80211_cancel_scan(vap); return; } is_hw_scan = (lhw->scan_flags & LKPI_LHW_SCAN_HW) != 0; LKPI_80211_LHW_SCAN_UNLOCK(lhw); - ss = ic->ic_scan; - vap = ss->ss_vap; +#if 0 if (vap->iv_state != IEEE80211_S_SCAN) { - IMPROVE("We need to be able to scan if not in S_SCAN"); + TODO("We need to be able to scan if not in S_SCAN"); + TRACE_SCAN(ic, "scan_flags %b iv_state %d", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, vap->iv_state); + ieee80211_cancel_scan(vap); return; } +#endif hw = LHW_TO_HW(lhw); if (!is_hw_scan) { /* If hw_scan is cleared clear FEXT_SCAN_OFFLOAD too. */ vap->iv_flags_ext &= ~IEEE80211_FEXT_SCAN_OFFLOAD; -sw_scan: + lvif = VAP_TO_LVIF(vap); vif = LVIF_TO_VIF(lvif); if (vap->iv_state == IEEE80211_S_SCAN) lkpi_hw_conf_idle(hw, false); + LKPI_80211_LHW_SCAN_LOCK(lhw); + lhw->scan_flags |= LKPI_LHW_SCAN_RUNNING; + LKPI_80211_LHW_SCAN_UNLOCK(lhw); + lkpi_update_mcast_filter(ic); + TRACE_SCAN(vap->iv_ic, "Starting SW_SCAN: scan_flags %b", + lhw->scan_flags, LKPI_LHW_SCAN_BITS); lkpi_80211_mo_sw_scan_start(hw, vif, vif->addr); /* net80211::scan_start() handled PS for us. */ IMPROVE(); @@ -4358,6 +4556,9 @@ sw_scan: struct cfg80211_scan_6ghz_params *s6gp; size_t chan_len, nchan, ssids_len, s6ghzlen; int band, i, ssid_count, common_ie_len; +#ifndef LKPI_80211_USE_SCANLIST + int n; +#endif uint32_t band_mask; uint8_t *ie, *ieend; bool running; @@ -4369,7 +4570,8 @@ sw_scan: band_mask = 0; nchan = 0; if (ieee80211_hw_check(hw, SINGLE_SCAN_ON_ALL_BANDS)) { -#if 0 /* Avoid net80211 scan lists until it has proper scan offload support. */ +#ifdef LKPI_80211_USE_SCANLIST + /* Avoid net80211 scan lists until it has proper scan offload support. */ for (i = ss->ss_next; i < ss->ss_last; i++) { nchan++; band = lkpi_net80211_chan_to_nl80211_band( @@ -4387,8 +4589,17 @@ sw_scan: continue; } if (hw->wiphy->bands[band] != NULL) { - nchan += hw->wiphy->bands[band]->n_channels; + struct linuxkpi_ieee80211_channel *channels; + int n; + band_mask |= (1 << band); + + channels = hw->wiphy->bands[band]->channels; + n = hw->wiphy->bands[band]->n_channels; + for (i = 0; i < n; i++) { + if (lkpi_scan_chan(&channels[i], ic, true)) + nchan++; + } } } #endif @@ -4427,11 +4638,32 @@ sw_scan: /* hw_req->req.wdev */ hw_req->req.wiphy = hw->wiphy; hw_req->req.no_cck = false; /* XXX */ -#if 0 - /* This seems to pessimise default scanning behaviour. */ - hw_req->req.duration_mandatory = TICKS_2_USEC(ss->ss_mindwell); - hw_req->req.duration = TICKS_2_USEC(ss->ss_maxdwell); -#endif + + /* + * In general setting duration[_mandatory] seems to pessimise + * default scanning behaviour. We only use it for BGSCANnig + * to keep the dwell times small. + * Setting duration_mandatory makes this the maximum dwell + * time (otherwise may be shorter). Duration is in TU. + */ + if ((ic->ic_flags_ext & IEEE80211_FEXT_BGSCAN) != 0) { + unsigned long dwell; + + if ((ic->ic_caps & IEEE80211_C_BGSCAN) == 0 || + (vap->iv_flags & IEEE80211_F_BGSCAN) == 0) + ic_printf(ic, "BGSCAN despite off: %b, %b, %b\n", + ic->ic_flags_ext, IEEE80211_FEXT_BITS, + vap->iv_flags, IEEE80211_F_BITS, + ic->ic_caps, IEEE80211_C_BITS); + + dwell = ss->ss_mindwell; + if (dwell == 0) + dwell = msecs_to_ticks(20); + + hw_req->req.duration_mandatory = true; + hw_req->req.duration = TICKS_2_USEC(dwell) / 1024; + } + #ifdef __notyet__ hw_req->req.flags |= NL80211_SCAN_FLAG_RANDOM_ADDR; memcpy(hw_req->req.mac_addr, xxx, IEEE80211_ADDR_LEN); @@ -4442,11 +4674,12 @@ sw_scan: hw_req->req.n_channels = nchan; cpp = (struct linuxkpi_ieee80211_channel **)(hw_req + 1); lc = (struct linuxkpi_ieee80211_channel *)(cpp + nchan); +#ifdef LKPI_80211_USE_SCANLIST for (i = 0; i < nchan; i++) { *(cpp + i) = (struct linuxkpi_ieee80211_channel *)(lc + i); } -#if 0 /* Avoid net80211 scan lists until it has proper scan offload support. */ + /* Avoid net80211 scan lists until it has proper scan offload support. */ for (i = 0; i < nchan; i++) { struct ieee80211_channel *c; @@ -4459,7 +4692,9 @@ sw_scan: lc++; } #else - for (band = 0; band < NUM_NL80211_BANDS; band++) { + /* Add bands in reverse order for scanning. */ + n = 0; + for (band = NUM_NL80211_BANDS - 1; band >= 0; band--) { struct ieee80211_supported_band *supband; struct linuxkpi_ieee80211_channel *channels; @@ -4474,11 +4709,30 @@ sw_scan: channels = supband->channels; for (i = 0; i < supband->n_channels; i++) { - *lc = channels[i]; - lc++; + if (lkpi_scan_chan(&channels[i], ic, false)) + *(cpp + n++) = &channels[i]; + } + } + if (lkpi_order_scanlist) + lkpi_scan_chan_list_resort(cpp, nchan); + + if ((linuxkpi_debug_80211 & D80211_SCAN) != 0) { + printf("%s:%d: %s SCAN Channel List (nchan=%zu): ", + __func__, __LINE__, ic->ic_name, nchan); + for (i = 0; i < nchan; i++) { + struct linuxkpi_ieee80211_channel *xc; + + xc = *(cpp + i); + printf(" %d(%d)", + ieee80211_mhz2ieee(xc->center_freq, + lkpi_nl80211_band_to_net80211_band( + xc->band)), + xc->center_freq); } + printf("\n"); } #endif + hw_req->req.n_ssids = ssid_count; if (hw_req->req.n_ssids > 0) { ssids = (struct cfg80211_ssid *)lc; @@ -4505,6 +4759,7 @@ sw_scan: ieend = lkpi_scan_ies_add(ie, &hw_req->ies, band_mask, vap, hw); hw_req->req.ie = ie; hw_req->req.ie_len = ieend - ie; + hw_req->req.scan_start = jiffies; lvif = VAP_TO_LVIF(vap); vif = LVIF_TO_VIF(lvif); @@ -4522,13 +4777,30 @@ sw_scan: LKPI_80211_LHW_SCAN_UNLOCK(lhw); if (running) { free(hw_req, M_LKPI80211); + TRACE_SCAN(ic, "Trying to start new scan while still " + "running (2); cancelling new net80211 scan; " + "scan_flags %b", + lhw->scan_flags, LKPI_LHW_SCAN_BITS); + ieee80211_cancel_scan(vap); return; } lkpi_update_mcast_filter(ic); + TRACE_SCAN(ic, "Starting HW_SCAN: scan_flags %b, " + "ie_len %d, n_ssids %d, n_chan %d, common_ie_len %d [%d, %d]", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, hw_req->req.ie_len, + hw_req->req.n_ssids, hw_req->req.n_channels, + hw_req->ies.common_ie_len, + hw_req->ies.len[NL80211_BAND_2GHZ], + hw_req->ies.len[NL80211_BAND_5GHZ]); error = lkpi_80211_mo_hw_scan(hw, vif, hw_req); if (error != 0) { + bool scan_done; + int e; + + TRACE_SCAN(ic, "hw_scan failed; scan_flags %b, error %d", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, error); ieee80211_cancel_scan(vap); /* @@ -4545,14 +4817,35 @@ sw_scan: * So we cannot rely on that behaviour and have to check * and balance between both code paths. */ + e = 0; + scan_done = true; LKPI_80211_LHW_SCAN_LOCK(lhw); if ((lhw->scan_flags & LKPI_LHW_SCAN_RUNNING) != 0) { + free(lhw->hw_req, M_LKPI80211); lhw->hw_req = NULL; + /* + * The ieee80211_cancel_scan() above runs in a + * taskq and it may take ages for the previous + * scan to clear; starting a new one right away + * we run into the problem that the old one is + * still active. + */ + e = msleep(lhw, &lhw->scan_mtx, 0, "lhwscanstop", hz); + scan_done = (lhw->scan_flags & LKPI_LHW_SCAN_RUNNING) != 0; + + /* + * Now we can clear running if no one else did. + */ lhw->scan_flags &= ~LKPI_LHW_SCAN_RUNNING; } LKPI_80211_LHW_SCAN_UNLOCK(lhw); lkpi_update_mcast_filter(ic); + if (!scan_done) { + ic_printf(ic, "ERROR: %s: timeout/error to wait " + "for ieee80211_cancel_scan: %d\n", __func__, e); + return; + } /* * XXX-SIGH magic number. @@ -4560,24 +4853,15 @@ sw_scan: * not possible. Fall back to sw scan in that case. */ if (error == 1) { - LKPI_80211_LHW_SCAN_LOCK(lhw); - lhw->scan_flags &= ~LKPI_LHW_SCAN_HW; - LKPI_80211_LHW_SCAN_UNLOCK(lhw); /* - * XXX If we clear this now and later a driver - * thinks it * can do a hw_scan again, we will - * currently not re-enable it? + * We need to put this into some defered context + * the net80211 scan may not be done yet + * (ic_flags & IEEE80211_F_SCAN) and we cannot + * wait here; if we do scan_curchan_task always + * runs after our timeout to finalize the scan. */ - vap->iv_flags_ext &= ~IEEE80211_FEXT_SCAN_OFFLOAD; - ieee80211_start_scan(vap, - IEEE80211_SCAN_ACTIVE | - IEEE80211_SCAN_NOPICK | - IEEE80211_SCAN_ONCE, - IEEE80211_SCAN_FOREVER, - ss->ss_mindwell ? ss->ss_mindwell : msecs_to_ticks(20), - ss->ss_maxdwell ? ss->ss_maxdwell : msecs_to_ticks(200), - vap->iv_des_nssid, vap->iv_des_ssid); - goto sw_scan; + ieee80211_runtask(ic, &lvif->sw_scan_task); + return; } ic_printf(ic, "ERROR: %s: hw_scan returned %d\n", @@ -4587,12 +4871,50 @@ sw_scan: } static void +lkpi_sw_scan_task(void *arg, int pending __unused) +{ + struct lkpi_hw *lhw; + struct lkpi_vif *lvif; + struct ieee80211vap *vap; + struct ieee80211_scan_state *ss; + + lvif = arg; + vap = LVIF_TO_VAP(lvif); + lhw = vap->iv_ic->ic_softc; + ss = vap->iv_ic->ic_scan; + + LKPI_80211_LHW_SCAN_LOCK(lhw); + /* + * We will re-enable this at scan_end calling lkpi_enable_hw_scan(). + * IEEE80211_FEXT_SCAN_OFFLOAD will be cleared by lkpi_ic_scan_start. + */ + lhw->scan_flags &= ~LKPI_LHW_SCAN_HW; + LKPI_80211_LHW_SCAN_UNLOCK(lhw); + + TRACE_SCAN(vap->iv_ic, "Triggering SW_SCAN: pending %d, scan_flags %b", + pending, lhw->scan_flags, LKPI_LHW_SCAN_BITS); + + /* + * This will call ic_scan_start() and we will get into the right path + * unless other scans started in between. + */ + ieee80211_start_scan(vap, + IEEE80211_SCAN_ONCE, + msecs_to_ticks(10000), /* 10000 ms (=~ 50 chan * 200 ms) */ + ss->ss_mindwell ? ss->ss_mindwell : msecs_to_ticks(20), + ss->ss_maxdwell ? ss->ss_maxdwell : msecs_to_ticks(200), + vap->iv_des_nssid, vap->iv_des_ssid); +} + +static void lkpi_ic_scan_end(struct ieee80211com *ic) { struct lkpi_hw *lhw; bool is_hw_scan; lhw = ic->ic_softc; + TRACE_SCAN(ic, "scan_flags %b", lhw->scan_flags, LKPI_LHW_SCAN_BITS); + LKPI_80211_LHW_SCAN_LOCK(lhw); if ((lhw->scan_flags & LKPI_LHW_SCAN_RUNNING) == 0) { LKPI_80211_LHW_SCAN_UNLOCK(lhw); @@ -4621,6 +4943,16 @@ lkpi_ic_scan_end(struct ieee80211com *ic) if (vap->iv_state == IEEE80211_S_SCAN) lkpi_hw_conf_idle(hw, true); } + + /* + * In case we disabled the hw_scan in lkpi_ic_scan_start() and + * switched to swscan, re-enable hw_scan if available. + */ + lkpi_enable_hw_scan(lhw); + + LKPI_80211_LHW_SCAN_LOCK(lhw); + wakeup(lhw); + LKPI_80211_LHW_SCAN_UNLOCK(lhw); } static void @@ -4631,6 +4963,10 @@ lkpi_ic_scan_curchan(struct ieee80211_scan_state *ss, bool is_hw_scan; lhw = ss->ss_ic->ic_softc; + TRACE_SCAN(ss->ss_ic, "scan_flags %b chan %d maxdwell %lu", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, + ss->ss_ic->ic_curchan->ic_ieee, maxdwell); + LKPI_80211_LHW_SCAN_LOCK(lhw); is_hw_scan = (lhw->scan_flags & LKPI_LHW_SCAN_HW) != 0; LKPI_80211_LHW_SCAN_UNLOCK(lhw); @@ -4645,6 +4981,10 @@ lkpi_ic_scan_mindwell(struct ieee80211_scan_state *ss) bool is_hw_scan; lhw = ss->ss_ic->ic_softc; + TRACE_SCAN(ss->ss_ic, "scan_flags %b chan %d mindwell %lu", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, + ss->ss_ic->ic_curchan->ic_ieee, ss->ss_mindwell); + LKPI_80211_LHW_SCAN_LOCK(lhw); is_hw_scan = (lhw->scan_flags & LKPI_LHW_SCAN_HW) != 0; LKPI_80211_LHW_SCAN_UNLOCK(lhw); @@ -6042,6 +6382,7 @@ linuxkpi_ieee80211_alloc_hw(size_t priv_len, const struct ieee80211_ops *ops) LKPI_80211_LHW_SCAN_LOCK_INIT(lhw); LKPI_80211_LHW_TXQ_LOCK_INIT(lhw); + spin_lock_init(&lhw->txq_lock); sx_init_flags(&lhw->lvif_sx, "lhw-lvif", SX_RECURSE | SX_DUPOK); LKPI_80211_LHW_MC_LOCK_INIT(lhw); TAILQ_INIT(&lhw->lvif_head); @@ -6147,6 +6488,7 @@ linuxkpi_ieee80211_iffree(struct ieee80211_hw *hw) LKPI_80211_LHW_MC_UNLOCK(lhw); /* Cleanup more of lhw here or in wiphy_free()? */ + spin_lock_destroy(&lhw->txq_lock); LKPI_80211_LHW_TXQ_LOCK_DESTROY(lhw); LKPI_80211_LHW_SCAN_LOCK_DESTROY(lhw); sx_destroy(&lhw->lvif_sx); @@ -6255,21 +6597,13 @@ linuxkpi_ieee80211_ifattach(struct ieee80211_hw *hw) IEEE80211_C_SHSLOT | /* short slot time supported */ IEEE80211_C_SHPREAMBLE /* short preamble supported */ ; -#if 0 - /* Scanning is a different kind of beast to re-work. */ - ic->ic_caps |= IEEE80211_C_BGSCAN; + +#ifdef LKPI_80211_BGSCAN + if (lhw->ops->hw_scan) + ic->ic_caps |= IEEE80211_C_BGSCAN; #endif - if (lhw->ops->hw_scan) { - /* - * Advertise full-offload scanning. - * - * Not limiting to SINGLE_SCAN_ON_ALL_BANDS here as otherwise - * we essentially disable hw_scan for all drivers not setting - * the flag. - */ - ic->ic_flags_ext |= IEEE80211_FEXT_SCAN_OFFLOAD; - lhw->scan_flags |= LKPI_LHW_SCAN_HW; - } + + lkpi_enable_hw_scan(lhw); /* Does HW support Fragmentation offload? */ if (ieee80211_hw_check(hw, SUPPORTS_TX_FRAG)) @@ -6726,13 +7060,19 @@ linuxkpi_ieee80211_scan_completed(struct ieee80211_hw *hw, ic = lhw->ic; ss = ic->ic_scan; + TRACE_SCAN(ic, "scan_flags %b info { %ju, %6D, aborted %d }", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, + (uintmax_t)info->scan_start_tsf, info->tsf_bssid, ":", + info->aborted); + ieee80211_scan_done(ss->ss_vap); LKPI_80211_LHW_SCAN_LOCK(lhw); free(lhw->hw_req, M_LKPI80211); lhw->hw_req = NULL; lhw->scan_flags &= ~LKPI_LHW_SCAN_RUNNING; - wakeup(lhw); + /* The wakeup(lhw) will be called from lkpi_ic_scan_end(). */ + /* wakeup(lhw); */ LKPI_80211_LHW_SCAN_UNLOCK(lhw); return; @@ -7002,6 +7342,63 @@ lkpi_convert_rx_status(struct ieee80211_hw *hw, struct lkpi_sta *lsta, } } +#ifdef LINUXKPI_DEBUG_80211 +static void +lkpi_rx_log_beacon(struct mbuf *m, struct lkpi_hw *lhw, + struct ieee80211_rx_status *rx_status) +{ + struct ieee80211_mgmt *f; + uint8_t *e; + char ssid[IEEE80211_NWID_LEN * 4 + 1]; + + memset(ssid, '\0', sizeof(ssid)); + + f = mtod(m, struct ieee80211_mgmt *); + e = f->u.beacon.variable; + /* + * Usually SSID is right after the fixed part and for debugging we will + * be fine should we miss it if it is not. + */ + while ((e - (uint8_t *)f) < m->m_len) { + if (*e == IEEE80211_ELEMID_SSID) + break; + e += (2 + *(e + 1)); + } + if (*e == IEEE80211_ELEMID_SSID) { + int i, len; + char *p; + + p = ssid; + len = m->m_len - ((e + 2) - (uint8_t *)f); + if (len > *(e + 1)) + len = *(e + 1); + e += 2; + for (i = 0; i < len; i++) { + /* Printable character? */ + if (*e >= 0x20 && *e < 0x7f) { + *p++ = *e++; + } else { + snprintf(p, 5, "%#04x", *e++); + p += 4; + } + } + *p = '\0'; + } + + /* We print skb, skb->data, m as we are seeing 'ghost beacons'. */ + TRACE_SCAN_BEACON(lhw->ic, "Beacon: scan_flags %b, band %s freq %u chan %-4d " + "len %d { %#06x %#06x %6D %6D %6D %#06x %ju %u %#06x SSID '%s' }", + lhw->scan_flags, LKPI_LHW_SCAN_BITS, + lkpi_nl80211_band_name(rx_status->band), rx_status->freq, + linuxkpi_ieee80211_frequency_to_channel(rx_status->freq, 0), + m->m_pkthdr.len, f->frame_control, f->duration_id, + f->da, ":", f->sa, ":", f->bssid, ":", f->seq_ctrl, + (uintmax_t)le64_to_cpu(f->u.beacon.timestamp), + le16_to_cpu(f->u.beacon.beacon_int), + le16_to_cpu(f->u.beacon.capab_info), ssid); +} +#endif + /* For %list see comment towards the end of the function. */ void linuxkpi_ieee80211_rx(struct ieee80211_hw *hw, struct sk_buff *skb, @@ -7058,7 +7455,15 @@ linuxkpi_ieee80211_rx(struct ieee80211_hw *hw, struct sk_buff *skb, is_beacon = ieee80211_is_beacon(hdr->frame_control); #ifdef LINUXKPI_DEBUG_80211 - if (is_beacon && (linuxkpi_debug_80211 & D80211_TRACE_RX_BEACONS) == 0) + /* + * We use the mbuf here as otherwise the variable part might + * be in skb frags. + */ + if (is_beacon && ((linuxkpi_debug_80211 & D80211_SCAN_BEACON) != 0)) + lkpi_rx_log_beacon(m, lhw, rx_status); + + if (is_beacon && (linuxkpi_debug_80211 & D80211_TRACE_RX_BEACONS) == 0 && + (linuxkpi_debug_80211 & D80211_SCAN_BEACON) == 0) goto no_trace_beacons; if (linuxkpi_debug_80211 & D80211_TRACE_RX) @@ -7073,7 +7478,8 @@ linuxkpi_ieee80211_rx(struct ieee80211_hw *hw, struct sk_buff *skb, hexdump(mtod(m, const void *), m->m_len, "RX (raw) ", 0); /* Implement a dump_rxcb() !!! */ - if (linuxkpi_debug_80211 & D80211_TRACE_RX) + if ((linuxkpi_debug_80211 & D80211_TRACE_RX) != 0 || + (linuxkpi_debug_80211 & D80211_SCAN_BEACON) != 0) printf("TRACE-RX: %s: RXCB: %ju %ju %u, %b, %u, %#0x, %#0x, " "%u band %u, %u { %d %d %d %d }, %d, %#x %#x %#x %#x %u %u %u\n", __func__, @@ -8124,21 +8530,30 @@ lkpi_ieee80211_wake_queues_locked(struct ieee80211_hw *hw) void linuxkpi_ieee80211_wake_queues(struct ieee80211_hw *hw) { - wiphy_lock(hw->wiphy); + struct lkpi_hw *lhw; + unsigned long flags; + + lhw = HW_TO_LHW(hw); + + spin_lock_irqsave(&lhw->txq_lock, flags); lkpi_ieee80211_wake_queues_locked(hw); - wiphy_unlock(hw->wiphy); + spin_unlock_irqrestore(&lhw->txq_lock, flags); } void linuxkpi_ieee80211_wake_queue(struct ieee80211_hw *hw, int qnum) { + struct lkpi_hw *lhw; + unsigned long flags; KASSERT(qnum < hw->queues, ("%s: qnum %d >= hw->queues %d, hw %p\n", __func__, qnum, hw->queues, hw)); - wiphy_lock(hw->wiphy); + lhw = HW_TO_LHW(hw); + + spin_lock_irqsave(&lhw->txq_lock, flags); lkpi_ieee80211_wake_queues(hw, qnum); - wiphy_unlock(hw->wiphy); + spin_unlock_irqrestore(&lhw->txq_lock, flags); } /* This is just hardware queues. */ diff --git a/sys/compat/linuxkpi/common/src/linux_80211.h b/sys/compat/linuxkpi/common/src/linux_80211.h index eaf6d804af4c..0dfcd7646c34 100644 --- a/sys/compat/linuxkpi/common/src/linux_80211.h +++ b/sys/compat/linuxkpi/common/src/linux_80211.h @@ -59,6 +59,8 @@ #define D80211_IMPROVE_TXQ 0x00000004 #define D80211_TRACE 0x00000010 #define D80211_TRACEOK 0x00000020 +#define D80211_SCAN 0x00000040 +#define D80211_SCAN_BEACON 0x00000080 #define D80211_TRACE_TX 0x00000100 #define D80211_TRACE_TX_DUMP 0x00000200 #define D80211_TRACE_RX 0x00001000 @@ -75,6 +77,20 @@ #define D80211_TRACE_MODE_HE 0x04000000 #define D80211_TRACE_MODE_EHT 0x08000000 +#ifdef LINUXKPI_DEBUG_80211 +#define TRACE_SCAN(ic, fmt, ...) \ + if (linuxkpi_debug_80211 & D80211_SCAN) \ + printf("%s:%d: %s SCAN " fmt "\n", \ + __func__, __LINE__, ic->ic_name, ##__VA_ARGS__) +#define TRACE_SCAN_BEACON(ic, fmt, ...) \ + if (linuxkpi_debug_80211 & D80211_SCAN_BEACON) \ + printf("%s:%d: %s SCAN " fmt "\n", \ + __func__, __LINE__, ic->ic_name, ##__VA_ARGS__) +#else +#define TRACE_SCAN(...) do {} while (0) +#define TRACE_SCAN_BEACON(...) do {} while (0) +#endif + #define IMPROVE_TXQ(...) \ if (linuxkpi_debug_80211 & D80211_IMPROVE_TXQ) \ printf("%s:%d: XXX LKPI80211 IMPROVE_TXQ\n", __func__, __LINE__) @@ -191,6 +207,7 @@ struct lkpi_vif { struct mbuf *, int, const struct ieee80211_rx_stats *, int, int); + struct task sw_scan_task; struct list_head lsta_list; @@ -236,6 +253,7 @@ struct lkpi_hw { /* name it mac80211_sc? */ struct mtx txq_mtx; uint32_t txq_generation[IEEE80211_NUM_ACS]; TAILQ_HEAD(, lkpi_txq) scheduled_txqs[IEEE80211_NUM_ACS]; + spinlock_t txq_lock; /* Deferred RX path. */ struct task rxq_task; @@ -298,6 +316,9 @@ struct lkpi_hw { /* name it mac80211_sc? */ #define LHW_TO_HW(_lhw) (&(_lhw)->hw) #define HW_TO_LHW(_hw) container_of(_hw, struct lkpi_hw, hw) +#define LKPI_LHW_SCAN_BITS \ + "\010\1RUNING\2HW" + struct lkpi_chanctx { struct list_head entry; diff --git a/sys/conf/files.x86 b/sys/conf/files.x86 index 9976e9cfec5d..953da7dd1284 100644 --- a/sys/conf/files.x86 +++ b/sys/conf/files.x86 @@ -146,6 +146,7 @@ dev/hyperv/vmbus/vmbus_et.c optional hyperv dev/hyperv/vmbus/vmbus_if.m optional hyperv dev/hyperv/vmbus/vmbus_res.c optional hyperv dev/hyperv/vmbus/vmbus_xact.c optional hyperv +dev/ichwd/i6300esbwd.c optional ichwd dev/ichwd/ichwd.c optional ichwd dev/imcsmb/imcsmb.c optional imcsmb dev/imcsmb/imcsmb_pci.c optional imcsmb pci diff --git a/sys/contrib/dev/acpica/components/executer/extrace.c b/sys/contrib/dev/acpica/components/executer/extrace.c index d54d4908ca65..b48a5fcb289b 100644 --- a/sys/contrib/dev/acpica/components/executer/extrace.c +++ b/sys/contrib/dev/acpica/components/executer/extrace.c @@ -301,7 +301,7 @@ AcpiExTraceArgs(ACPI_OPERAND_OBJECT **Params, UINT32 Count) switch (obj_desc->Common.Type) { case ACPI_TYPE_INTEGER: - ACPI_DEBUG_PRINT_RAW((ACPI_DB_TRACE_POINT, "%lx", obj_desc->Integer.Value)); + ACPI_DEBUG_PRINT_RAW((ACPI_DB_TRACE_POINT, "%jx", (uintmax_t)obj_desc->Integer.Value)); break; case ACPI_TYPE_STRING: diff --git a/sys/contrib/dev/rtw88/main.c b/sys/contrib/dev/rtw88/main.c index 021d076808e0..963b73f35350 100644 --- a/sys/contrib/dev/rtw88/main.c +++ b/sys/contrib/dev/rtw88/main.c @@ -57,6 +57,62 @@ module_param_named(support_vht, rtw_vht_support, bool, 0644); MODULE_PARM_DESC(support_vht, "Set to Y to enable VHT support"); #endif +#if defined(__FreeBSD__) +/* Macros based on rtw89::core.c. */ +#define RTW88_DEF_CHAN(_freq, _hw_val, _flags, _band) \ + { .center_freq = _freq, .hw_value = _hw_val, .flags = _flags, .band = _band, } +#define RTW88_DEF_CHAN_2G(_freq, _hw_val) \ + RTW88_DEF_CHAN(_freq, _hw_val, 0, NL80211_BAND_2GHZ) +#define RTW88_DEF_CHAN_5G(_freq, _hw_val) \ + RTW88_DEF_CHAN(_freq, _hw_val, 0, NL80211_BAND_5GHZ) +#define RTW88_DEF_CHAN_5G_NO_HT40MINUS(_freq, _hw_val) \ + RTW88_DEF_CHAN(_freq, _hw_val, IEEE80211_CHAN_NO_HT40MINUS, NL80211_BAND_5GHZ) + +static struct ieee80211_channel rtw_channeltable_2g[] = { + RTW88_DEF_CHAN_2G(2412, 1), + RTW88_DEF_CHAN_2G(2417, 2), + RTW88_DEF_CHAN_2G(2422, 3), + RTW88_DEF_CHAN_2G(2427, 4), + RTW88_DEF_CHAN_2G(2432, 5), + RTW88_DEF_CHAN_2G(2437, 6), + RTW88_DEF_CHAN_2G(2442, 7), + RTW88_DEF_CHAN_2G(2447, 8), + RTW88_DEF_CHAN_2G(2452, 9), + RTW88_DEF_CHAN_2G(2457, 10), + RTW88_DEF_CHAN_2G(2462, 11), + RTW88_DEF_CHAN_2G(2467, 12), + RTW88_DEF_CHAN_2G(2472, 13), + RTW88_DEF_CHAN_2G(2484, 14), +}; + +static struct ieee80211_channel rtw_channeltable_5g[] = { + RTW88_DEF_CHAN_5G(5180, 36), + RTW88_DEF_CHAN_5G(5200, 40), + RTW88_DEF_CHAN_5G(5220, 44), + RTW88_DEF_CHAN_5G(5240, 48), + RTW88_DEF_CHAN_5G(5260, 52), + RTW88_DEF_CHAN_5G(5280, 56), + RTW88_DEF_CHAN_5G(5300, 60), + RTW88_DEF_CHAN_5G(5320, 64), + RTW88_DEF_CHAN_5G(5500, 100), + RTW88_DEF_CHAN_5G(5520, 104), + RTW88_DEF_CHAN_5G(5540, 108), + RTW88_DEF_CHAN_5G(5560, 112), + RTW88_DEF_CHAN_5G(5580, 116), + RTW88_DEF_CHAN_5G(5600, 120), + RTW88_DEF_CHAN_5G(5620, 124), + RTW88_DEF_CHAN_5G(5640, 128), + RTW88_DEF_CHAN_5G(5660, 132), + RTW88_DEF_CHAN_5G(5680, 136), + RTW88_DEF_CHAN_5G(5700, 140), + RTW88_DEF_CHAN_5G(5720, 144), + RTW88_DEF_CHAN_5G(5745, 149), + RTW88_DEF_CHAN_5G(5765, 153), + RTW88_DEF_CHAN_5G(5785, 157), + RTW88_DEF_CHAN_5G(5805, 161), + RTW88_DEF_CHAN_5G_NO_HT40MINUS(5825, 165), +}; +#elif deifned(__linux__) static struct ieee80211_channel rtw_channeltable_2g[] = { {.center_freq = 2412, .hw_value = 1,}, {.center_freq = 2417, .hw_value = 2,}, @@ -102,6 +158,7 @@ static struct ieee80211_channel rtw_channeltable_5g[] = { {.center_freq = 5825, .hw_value = 165, .flags = IEEE80211_CHAN_NO_HT40MINUS}, }; +#endif static struct ieee80211_rate rtw_ratetable[] = { {.bitrate = 10, .hw_value = 0x00,}, diff --git a/sys/contrib/vchiq/interface/vchiq_arm/vchiq_2835_arm.c b/sys/contrib/vchiq/interface/vchiq_arm/vchiq_2835_arm.c index ab8981e25cb2..0150ce72f0a4 100644 --- a/sys/contrib/vchiq/interface/vchiq_arm/vchiq_2835_arm.c +++ b/sys/contrib/vchiq/interface/vchiq_arm/vchiq_2835_arm.c @@ -464,7 +464,8 @@ create_pagelist(char __user *buf, size_t count, unsigned short type, (type == PAGELIST_READ ? VM_PROT_WRITE : 0 ) | VM_PROT_READ, pages, num_pages); if (actual_pages != num_pages) { - vm_page_unhold_pages(pages, actual_pages); + if (actual_pages > 0) + vm_page_unhold_pages(pages, actual_pages); free(pagelist, M_VCPAGELIST); return (-ENOMEM); } diff --git a/sys/dev/amdgpio/amdgpio.c b/sys/dev/amdgpio/amdgpio.c index 2bd455c612b8..20589ff71b0b 100644 --- a/sys/dev/amdgpio/amdgpio.c +++ b/sys/dev/amdgpio/amdgpio.c @@ -3,6 +3,10 @@ * * Copyright (c) 2018 Advanced Micro Devices * All rights reserved. + * Copyright (c) 2025 The FreeBSD Foundation + * + * Portions of this software were developed by Aymeric Wibo + * <obiwac@freebsd.org> under sponsorship from the FreeBSD Foundation. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -51,11 +55,11 @@ #include <dev/acpica/acpivar.h> #include <dev/gpio/gpiobusvar.h> -#include "gpio_if.h" #include "amdgpio.h" static struct resource_spec amdgpio_spec[] = { - { SYS_RES_MEMORY, 0, RF_ACTIVE }, + { SYS_RES_MEMORY, 0, RF_ACTIVE }, + { SYS_RES_IRQ, 0, RF_ACTIVE | RF_SHAREABLE }, { -1, 0, 0 } }; @@ -196,7 +200,7 @@ static int amdgpio_pin_setflags(device_t dev, uint32_t pin, uint32_t flags) { struct amdgpio_softc *sc; - uint32_t reg, val, allowed; + uint32_t reg, val; sc = device_get_softc(dev); @@ -204,18 +208,19 @@ amdgpio_pin_setflags(device_t dev, uint32_t pin, uint32_t flags) if (!amdgpio_valid_pin(sc, pin)) return (EINVAL); - allowed = GPIO_PIN_INPUT | GPIO_PIN_OUTPUT; + if ((flags & ~AMDGPIO_DEFAULT_CAPS) != 0) { + device_printf(dev, "disallowed flags (0x%x) trying to be set " + "(allowed is 0x%x)\n", flags, AMDGPIO_DEFAULT_CAPS); + return (EINVAL); + } - /* - * Only directtion flag allowed - */ - if (flags & ~allowed) + /* Either input or output must be selected. */ + if ((flags & (GPIO_PIN_INPUT | GPIO_PIN_OUTPUT)) == 0) return (EINVAL); - /* - * Not both directions simultaneously - */ - if ((flags & allowed) == allowed) + /* Not both directions simultaneously. */ + if ((flags & (GPIO_PIN_INPUT | GPIO_PIN_OUTPUT)) == + (GPIO_PIN_INPUT | GPIO_PIN_OUTPUT)) return (EINVAL); /* Set the GPIO mode and state */ @@ -224,16 +229,21 @@ amdgpio_pin_setflags(device_t dev, uint32_t pin, uint32_t flags) reg = AMDGPIO_PIN_REGISTER(pin); val = amdgpio_read_4(sc, reg); - if (flags & GPIO_PIN_INPUT) { + if ((flags & GPIO_PIN_INPUT) != 0) val &= ~BIT(OUTPUT_ENABLE_OFF); - sc->sc_gpio_pins[pin].gp_flags = GPIO_PIN_INPUT; - } else { + else val |= BIT(OUTPUT_ENABLE_OFF); - sc->sc_gpio_pins[pin].gp_flags = GPIO_PIN_OUTPUT; - } + + val &= ~(BIT(PULL_DOWN_ENABLE_OFF) | BIT(PULL_UP_ENABLE_OFF)); + + if ((flags & GPIO_PIN_PULLDOWN) != 0) + val |= BIT(PULL_DOWN_ENABLE_OFF); + if ((flags & GPIO_PIN_PULLUP) != 0) + val |= BIT(PULL_UP_ENABLE_OFF); amdgpio_write_4(sc, reg, val); + sc->sc_gpio_pins[pin].gp_flags = flags; dprintf("pin %d flags 0x%x val 0x%x gp_flags 0x%x\n", pin, flags, val, sc->sc_gpio_pins[pin].gp_flags); @@ -359,11 +369,73 @@ amdgpio_probe(device_t dev) return (rv); } +static void +amdgpio_eoi_locked(struct amdgpio_softc *sc) +{ + uint32_t master_reg = amdgpio_read_4(sc, WAKE_INT_MASTER_REG); + + AMDGPIO_ASSERT_LOCKED(sc); + master_reg |= EOI_MASK; + amdgpio_write_4(sc, WAKE_INT_MASTER_REG, master_reg); +} + +static void +amdgpio_eoi(struct amdgpio_softc *sc) +{ + AMDGPIO_LOCK(sc); + amdgpio_eoi_locked(sc); + AMDGPIO_UNLOCK(sc); +} + +static int +amdgpio_intr_filter(void *arg) +{ + struct amdgpio_softc *sc = arg; + int off, rv = FILTER_STRAY; + uint32_t reg; + + /* We can lock in the filter routine as it is MTX_SPIN. */ + AMDGPIO_LOCK(sc); + + /* + * TODO Instead of just reading the registers of all pins, we should + * read WAKE_INT_STATUS_REG0/1. A bit set in here denotes a group of + * 4 pins where at least one has an interrupt for us. Then we can just + * iterate over those 4 pins. + * + * See GPIO_Interrupt_Status_Index_0 in BKDG. + */ + for (size_t pin = 0; pin < AMD_GPIO_PINS_EXPOSED; pin++) { + off = AMDGPIO_PIN_REGISTER(pin); + reg = amdgpio_read_4(sc, off); + if ((reg & UNSERVICED_INTERRUPT_MASK) == 0) + continue; + /* + * Must write 1's to wake/interrupt status bits to clear them. + * We can do this simply by writing back to the register. + */ + amdgpio_write_4(sc, off, reg); + } + + amdgpio_eoi_locked(sc); + AMDGPIO_UNLOCK(sc); + + rv = FILTER_HANDLED; + return (rv); +} + +static void +amdgpio_intr_handler(void *arg) +{ + /* TODO */ +} + static int amdgpio_attach(device_t dev) { struct amdgpio_softc *sc; - int i, pin, bank; + int i, pin, bank, reg; + uint32_t flags; sc = device_get_softc(dev); sc->sc_dev = dev; @@ -386,6 +458,14 @@ amdgpio_attach(device_t dev) sc->sc_bst = rman_get_bustag(sc->sc_res[0]); sc->sc_bsh = rman_get_bushandle(sc->sc_res[0]); + /* Set up interrupt handler. */ + if (bus_setup_intr(dev, sc->sc_res[1], INTR_TYPE_MISC | INTR_MPSAFE, + amdgpio_intr_filter, amdgpio_intr_handler, sc, &sc->sc_intr_handle) + != 0) { + device_printf(dev, "couldn't set up interrupt\n"); + goto err_intr; + } + /* Initialize all possible pins to be Invalid */ for (i = 0; i < AMD_GPIO_PINS_MAX ; i++) { snprintf(sc->sc_gpio_pins[i].gp_name, GPIOMAXNAME, @@ -395,7 +475,12 @@ amdgpio_attach(device_t dev) sc->sc_gpio_pins[i].gp_flags = 0; } - /* Initialize only driver exposed pins with appropriate capabilities */ + /* + * Initialize only driver exposed pins with appropriate capabilities. + * + * XXX Also mask and disable interrupts on all pins, since we don't + * support them at the moment. + */ for (i = 0; i < AMD_GPIO_PINS_EXPOSED ; i++) { pin = kernzp_pins[i].pin_num; bank = pin/AMD_GPIO_PINS_PER_BANK; @@ -406,7 +491,14 @@ amdgpio_attach(device_t dev) sc->sc_gpio_pins[pin].gp_flags = amdgpio_is_pin_output(sc, pin) ? GPIO_PIN_OUTPUT : GPIO_PIN_INPUT; + + reg = AMDGPIO_PIN_REGISTER(pin); + flags = amdgpio_read_4(sc, reg); + flags &= ~(1 << INTERRUPT_ENABLE_OFF); + flags &= ~(1 << INTERRUPT_MASK_OFF); + amdgpio_write_4(sc, reg, flags); } + amdgpio_eoi(sc); sc->sc_busdev = gpiobus_add_bus(dev); if (sc->sc_busdev == NULL) { @@ -418,8 +510,9 @@ amdgpio_attach(device_t dev) return (0); err_bus: + bus_teardown_intr(dev, sc->sc_res[1], sc->sc_intr_handle); +err_intr: bus_release_resources(dev, amdgpio_spec, sc->sc_res); - err_rsrc: AMDGPIO_LOCK_DESTROY(sc); @@ -434,7 +527,8 @@ amdgpio_detach(device_t dev) if (sc->sc_busdev) gpiobus_detach_bus(dev); - + if (sc->sc_intr_handle) + bus_teardown_intr(dev, sc->sc_res[1], sc->sc_intr_handle); bus_release_resources(dev, amdgpio_spec, sc->sc_res); AMDGPIO_LOCK_DESTROY(sc); diff --git a/sys/dev/amdgpio/amdgpio.h b/sys/dev/amdgpio/amdgpio.h index aca3039bfc98..3743eba23e17 100644 --- a/sys/dev/amdgpio/amdgpio.h +++ b/sys/dev/amdgpio/amdgpio.h @@ -50,7 +50,8 @@ AMD_GPIO_PINS_BANK1 + \ AMD_GPIO_PINS_BANK2 + \ AMD_GPIO_PINS_BANK3) -#define AMDGPIO_DEFAULT_CAPS (GPIO_PIN_INPUT | GPIO_PIN_OUTPUT) +#define AMDGPIO_DEFAULT_CAPS (GPIO_PIN_INPUT | GPIO_PIN_OUTPUT | \ + GPIO_PIN_PULLDOWN | GPIO_PIN_PULLUP) /* Register related macros */ #define AMDGPIO_PIN_REGISTER(pin) (pin * 4) @@ -84,6 +85,9 @@ #define INTERRUPT_STS_OFF 28 #define WAKE_STS_OFF 29 +#define UNSERVICED_INTERRUPT_MASK \ + ((1 << INTERRUPT_STS_OFF) | (1 << WAKE_STS_OFF)) + #define DB_TMR_OUT_MASK 0xFUL #define DB_CNTRL_MASK 0x3UL #define ACTIVE_LEVEL_MASK 0x3UL @@ -316,12 +320,13 @@ struct amdgpio_softc { int sc_npins; int sc_ngroups; struct mtx sc_mtx; - struct resource *sc_res[AMD_GPIO_NUM_PIN_BANK + 1]; + struct resource *sc_res[2]; bus_space_tag_t sc_bst; bus_space_handle_t sc_bsh; struct gpio_pin sc_gpio_pins[AMD_GPIO_PINS_MAX]; const struct pin_info *sc_pin_info; const struct amd_pingroup *sc_groups; + void *sc_intr_handle; }; struct amdgpio_sysctl { diff --git a/sys/dev/cxgbe/t4_main.c b/sys/dev/cxgbe/t4_main.c index 9e91250cb61c..9756a6945384 100644 --- a/sys/dev/cxgbe/t4_main.c +++ b/sys/dev/cxgbe/t4_main.c @@ -9016,7 +9016,7 @@ sysctl_loadavg(SYSCTL_HANDLER_ARGS) rc = begin_synchronized_op(sc, NULL, SLEEP_OK | INTR_OK, "t4lavg"); if (rc) return (rc); - if (hw_all_ok(sc)) + if (!hw_all_ok(sc)) rc = ENXIO; else { param = V_FW_PARAMS_MNEM(FW_PARAMS_MNEM_DEV) | diff --git a/sys/dev/gpio/gpio_if.m b/sys/dev/gpio/gpio_if.m index 5501b2b5c0e7..0b6988ceba79 100644 --- a/sys/dev/gpio/gpio_if.m +++ b/sys/dev/gpio/gpio_if.m @@ -62,6 +62,22 @@ CODE { return (0); } + + static int + gpio_default_get_pin_list(device_t dev, uint32_t *pin_list) + { + uint32_t maxpin; + int err; + + err = GPIO_PIN_MAX(dev, &maxpin); + if (err != 0) + return (ENXIO); + + for (int i = 0; i <= maxpin; i++) + pin_list[i] = i; + + return (0); + } }; HEADER { @@ -185,3 +201,13 @@ METHOD int pin_config_32 { uint32_t num_pins; uint32_t *pin_flags; } DEFAULT gpio_default_nosupport; + +# +# Get the controller's pin numbers. pin_list is expected to be an array with at +# least GPIO_PIN_MAX() elements. Populates pin_list from 0 to GPIO_PIN_MAX() by +# default. +# +METHOD int get_pin_list { + device_t dev; + uint32_t *pin_list; +} DEFAULT gpio_default_get_pin_list; diff --git a/sys/dev/gpio/gpiobus.c b/sys/dev/gpio/gpiobus.c index 5f1f6532a79b..698b5e5fdd01 100644 --- a/sys/dev/gpio/gpiobus.c +++ b/sys/dev/gpio/gpiobus.c @@ -319,10 +319,6 @@ gpiobus_add_bus(device_t dev) busdev = device_add_child(dev, "gpiobus", DEVICE_UNIT_ANY); if (busdev == NULL) return (NULL); - if (device_add_child(dev, "gpioc", DEVICE_UNIT_ANY) == NULL) { - device_delete_child(dev, busdev); - return (NULL); - } #ifdef FDT ofw_gpiobus_register_provider(dev); #endif @@ -372,6 +368,37 @@ gpiobus_init_softc(device_t dev) } int +gpiobus_add_gpioc(device_t dev) +{ + struct gpiobus_ivar *devi; + struct gpiobus_softc *sc; + device_t gpioc; + int err; + + gpioc = BUS_ADD_CHILD(dev, 0, "gpioc", DEVICE_UNIT_ANY); + if (gpioc == NULL) + return (ENXIO); + + sc = device_get_softc(dev); + devi = device_get_ivars(gpioc); + + devi->npins = sc->sc_npins; + err = gpiobus_alloc_ivars(devi); + if (err != 0) { + device_delete_child(dev, gpioc); + return (err); + } + + err = GPIO_GET_PIN_LIST(sc->sc_dev, devi->pins); + if (err != 0) { + device_delete_child(dev, gpioc); + gpiobus_free_ivars(devi); + } + + return (err); +} + +int gpiobus_alloc_ivars(struct gpiobus_ivar *devi) { @@ -562,6 +589,10 @@ gpiobus_attach(device_t dev) if (err != 0) return (err); + err = gpiobus_add_gpioc(dev); + if (err != 0) + return (err); + /* * Get parent's pins and mark them as unmapped */ @@ -961,7 +992,7 @@ gpiobus_pin_getflags(device_t dev, device_t child, uint32_t pin, if (pin >= devi->npins) return (EINVAL); - return GPIO_PIN_GETFLAGS(sc->sc_dev, devi->pins[pin], flags); + return (GPIO_PIN_GETFLAGS(sc->sc_dev, devi->pins[pin], flags)); } static int @@ -974,7 +1005,7 @@ gpiobus_pin_getcaps(device_t dev, device_t child, uint32_t pin, if (pin >= devi->npins) return (EINVAL); - return GPIO_PIN_GETCAPS(sc->sc_dev, devi->pins[pin], caps); + return (GPIO_PIN_GETCAPS(sc->sc_dev, devi->pins[pin], caps)); } static int @@ -987,7 +1018,7 @@ gpiobus_pin_set(device_t dev, device_t child, uint32_t pin, if (pin >= devi->npins) return (EINVAL); - return GPIO_PIN_SET(sc->sc_dev, devi->pins[pin], value); + return (GPIO_PIN_SET(sc->sc_dev, devi->pins[pin], value)); } static int @@ -1000,7 +1031,7 @@ gpiobus_pin_get(device_t dev, device_t child, uint32_t pin, if (pin >= devi->npins) return (EINVAL); - return GPIO_PIN_GET(sc->sc_dev, devi->pins[pin], value); + return (GPIO_PIN_GET(sc->sc_dev, devi->pins[pin], value)); } static int @@ -1012,7 +1043,57 @@ gpiobus_pin_toggle(device_t dev, device_t child, uint32_t pin) if (pin >= devi->npins) return (EINVAL); - return GPIO_PIN_TOGGLE(sc->sc_dev, devi->pins[pin]); + return (GPIO_PIN_TOGGLE(sc->sc_dev, devi->pins[pin])); +} + +/* + * Verify that a child has all the pins they are requesting + * to access in their ivars. + */ +static bool +gpiobus_pin_verify_32(struct gpiobus_ivar *devi, uint32_t first_pin, + uint32_t num_pins) +{ + if (first_pin + num_pins > devi->npins) + return (false); + + /* Make sure the pins are consecutive. */ + for (uint32_t pin = first_pin; pin < first_pin + num_pins - 1; pin++) { + if (devi->pins[pin] + 1 != devi->pins[pin + 1]) + return (false); + } + + return (true); +} + +static int +gpiobus_pin_access_32(device_t dev, device_t child, uint32_t first_pin, + uint32_t clear_pins, uint32_t change_pins, uint32_t *orig_pins) +{ + struct gpiobus_softc *sc = GPIOBUS_SOFTC(dev); + struct gpiobus_ivar *devi = GPIOBUS_IVAR(child); + + if (!gpiobus_pin_verify_32(devi, first_pin, 32)) + return (EINVAL); + + return (GPIO_PIN_ACCESS_32(sc->sc_dev, devi->pins[first_pin], + clear_pins, change_pins, orig_pins)); +} + +static int +gpiobus_pin_config_32(device_t dev, device_t child, uint32_t first_pin, + uint32_t num_pins, uint32_t *pin_flags) +{ + struct gpiobus_softc *sc = GPIOBUS_SOFTC(dev); + struct gpiobus_ivar *devi = GPIOBUS_IVAR(child); + + if (num_pins > 32) + return (EINVAL); + if (!gpiobus_pin_verify_32(devi, first_pin, num_pins)) + return (EINVAL); + + return (GPIO_PIN_CONFIG_32(sc->sc_dev, + devi->pins[first_pin], num_pins, pin_flags)); } static int @@ -1093,6 +1174,8 @@ static device_method_t gpiobus_methods[] = { DEVMETHOD(gpiobus_pin_get, gpiobus_pin_get), DEVMETHOD(gpiobus_pin_set, gpiobus_pin_set), DEVMETHOD(gpiobus_pin_toggle, gpiobus_pin_toggle), + DEVMETHOD(gpiobus_pin_access_32,gpiobus_pin_access_32), + DEVMETHOD(gpiobus_pin_config_32,gpiobus_pin_config_32), DEVMETHOD(gpiobus_pin_getname, gpiobus_pin_getname), DEVMETHOD(gpiobus_pin_setname, gpiobus_pin_setname), diff --git a/sys/dev/gpio/gpiobus_if.m b/sys/dev/gpio/gpiobus_if.m index 8bf29839ef4e..890738c4e809 100644 --- a/sys/dev/gpio/gpiobus_if.m +++ b/sys/dev/gpio/gpiobus_if.m @@ -107,6 +107,36 @@ METHOD int pin_setflags { }; # +# Simultaneously read and/or change up to 32 adjacent pins. +# If the device cannot change the pins simultaneously, returns EOPNOTSUPP. +# +# More details about using this interface can be found in sys/gpio.h +# +METHOD int pin_access_32 { + device_t dev; + device_t child; + uint32_t first_pin; + uint32_t clear_pins; + uint32_t change_pins; + uint32_t *orig_pins; +}; + +# +# Simultaneously configure up to 32 adjacent pins. +# This is intended to change the configuration of all the pins simultaneously, +# but unlike pin_access_32, this will not fail if the hardware can't do so. +# +# More details about using this interface can be found in sys/gpio.h +# +METHOD int pin_config_32 { + device_t dev; + device_t child; + uint32_t first_pin; + uint32_t num_pins; + uint32_t *pin_flags; +}; + +# # Get the pin name # METHOD int pin_getname { diff --git a/sys/dev/gpio/gpiobus_internal.h b/sys/dev/gpio/gpiobus_internal.h index c198e5f79989..58f862343403 100644 --- a/sys/dev/gpio/gpiobus_internal.h +++ b/sys/dev/gpio/gpiobus_internal.h @@ -44,6 +44,7 @@ int gpiobus_acquire_pin(device_t, uint32_t); void gpiobus_release_pin(device_t, uint32_t); int gpiobus_child_location(device_t, device_t, struct sbuf *); device_t gpiobus_add_child_common(device_t, u_int, const char *, int, size_t); +int gpiobus_add_gpioc(device_t); extern driver_t gpiobus_driver; #endif diff --git a/sys/dev/gpio/gpioc.c b/sys/dev/gpio/gpioc.c index 87fed38ebe3e..5a60f939dc78 100644 --- a/sys/dev/gpio/gpioc.c +++ b/sys/dev/gpio/gpioc.c @@ -45,7 +45,6 @@ #include <dev/gpio/gpiobusvar.h> -#include "gpio_if.h" #include "gpiobus_if.h" #undef GPIOC_DEBUG @@ -59,7 +58,7 @@ struct gpioc_softc { device_t sc_dev; /* gpiocX dev */ - device_t sc_pdev; /* gpioX dev */ + device_t sc_pdev; /* gpiobusX dev */ struct cdev *sc_ctl_dev; /* controller device */ int sc_unit; int sc_npins; @@ -69,6 +68,7 @@ struct gpioc_softc { struct gpioc_pin_intr { struct gpioc_softc *sc; gpio_pin_t pin; + uint32_t intr_mode; bool config_locked; int intr_rid; struct resource *intr_res; @@ -112,8 +112,10 @@ struct gpioc_pin_event { static MALLOC_DEFINE(M_GPIOC, "gpioc", "gpioc device data"); -static int gpioc_allocate_pin_intr(struct gpioc_pin_intr*, uint32_t); -static int gpioc_release_pin_intr(struct gpioc_pin_intr*); +static int gpioc_allocate_pin_intr(struct gpioc_softc*, + struct gpioc_pin_intr*, uint32_t, uint32_t); +static int gpioc_release_pin_intr(struct gpioc_softc*, + struct gpioc_pin_intr*); static int gpioc_attach_priv_pin(struct gpioc_cdevpriv*, struct gpioc_pin_intr*); static int gpioc_detach_priv_pin(struct gpioc_cdevpriv*, @@ -191,27 +193,36 @@ number_of_events(struct gpioc_cdevpriv *priv) } static int -gpioc_allocate_pin_intr(struct gpioc_pin_intr *intr_conf, uint32_t flags) +gpioc_allocate_pin_intr(struct gpioc_softc *sc, + struct gpioc_pin_intr *intr_conf, uint32_t pin, uint32_t flags) { int err; intr_conf->config_locked = true; mtx_unlock(&intr_conf->mtx); - intr_conf->intr_res = gpio_alloc_intr_resource(intr_conf->pin->dev, + MPASS(intr_conf->pin == NULL); + err = gpio_pin_get_by_bus_pinnum(sc->sc_pdev, pin, &intr_conf->pin); + if (err != 0) + goto error_exit; + + intr_conf->intr_res = gpio_alloc_intr_resource(sc->sc_dev, &intr_conf->intr_rid, RF_ACTIVE, intr_conf->pin, flags); if (intr_conf->intr_res == NULL) { err = ENXIO; - goto error_exit; + goto error_pin; } - err = bus_setup_intr(intr_conf->pin->dev, intr_conf->intr_res, + err = bus_setup_intr(sc->sc_dev, intr_conf->intr_res, INTR_TYPE_MISC | INTR_MPSAFE, NULL, gpioc_interrupt_handler, intr_conf, &intr_conf->intr_cookie); - if (err != 0) - goto error_exit; + if (err != 0) { + bus_release_resource(sc->sc_dev, intr_conf->intr_res); + intr_conf->intr_res = NULL; + goto error_pin; + } - intr_conf->pin->flags = flags; + intr_conf->intr_mode = flags; error_exit: mtx_lock(&intr_conf->mtx); @@ -219,10 +230,15 @@ error_exit: wakeup(&intr_conf->config_locked); return (err); + +error_pin: + gpio_pin_release(intr_conf->pin); + intr_conf->pin = NULL; + goto error_exit; } static int -gpioc_release_pin_intr(struct gpioc_pin_intr *intr_conf) +gpioc_release_pin_intr(struct gpioc_softc *sc, struct gpioc_pin_intr *intr_conf) { int err; @@ -230,8 +246,8 @@ gpioc_release_pin_intr(struct gpioc_pin_intr *intr_conf) mtx_unlock(&intr_conf->mtx); if (intr_conf->intr_cookie != NULL) { - err = bus_teardown_intr(intr_conf->pin->dev, - intr_conf->intr_res, intr_conf->intr_cookie); + err = bus_teardown_intr(sc->sc_dev, intr_conf->intr_res, + intr_conf->intr_cookie); if (err != 0) goto error_exit; else @@ -239,7 +255,7 @@ gpioc_release_pin_intr(struct gpioc_pin_intr *intr_conf) } if (intr_conf->intr_res != NULL) { - err = bus_release_resource(intr_conf->pin->dev, SYS_RES_IRQ, + err = bus_release_resource(sc->sc_dev, SYS_RES_IRQ, intr_conf->intr_rid, intr_conf->intr_res); if (err != 0) goto error_exit; @@ -249,7 +265,10 @@ gpioc_release_pin_intr(struct gpioc_pin_intr *intr_conf) } } - intr_conf->pin->flags = 0; + gpio_pin_release(intr_conf->pin); + intr_conf->pin = NULL; + + intr_conf->intr_mode = 0; err = 0; error_exit: @@ -386,7 +405,7 @@ gpioc_get_intr_config(struct gpioc_softc *sc, struct gpioc_cdevpriv *priv, struct gpioc_privs *priv_link; uint32_t flags; - flags = intr_conf->pin->flags; + flags = intr_conf->intr_mode; if (flags == 0) return (0); @@ -411,7 +430,7 @@ gpioc_set_intr_config(struct gpioc_softc *sc, struct gpioc_cdevpriv *priv, int res; res = 0; - if (intr_conf->pin->flags == 0 && flags == 0) { + if (intr_conf->intr_mode == 0 && flags == 0) { /* No interrupt configured and none requested: Do nothing. */ return (0); } @@ -419,17 +438,17 @@ gpioc_set_intr_config(struct gpioc_softc *sc, struct gpioc_cdevpriv *priv, while (intr_conf->config_locked == true) mtx_sleep(&intr_conf->config_locked, &intr_conf->mtx, 0, "gpicfg", 0); - if (intr_conf->pin->flags == 0 && flags != 0) { + if (intr_conf->intr_mode == 0 && flags != 0) { /* * No interrupt is configured, but one is requested: Allocate * and setup interrupt on the according pin. */ - res = gpioc_allocate_pin_intr(intr_conf, flags); + res = gpioc_allocate_pin_intr(sc, intr_conf, pin, flags); if (res == 0) res = gpioc_attach_priv_pin(priv, intr_conf); if (res == EEXIST) res = 0; - } else if (intr_conf->pin->flags == flags) { + } else if (intr_conf->intr_mode == flags) { /* * Same interrupt requested as already configured: Attach the * cdevpriv to the corresponding pin. @@ -437,14 +456,14 @@ gpioc_set_intr_config(struct gpioc_softc *sc, struct gpioc_cdevpriv *priv, res = gpioc_attach_priv_pin(priv, intr_conf); if (res == EEXIST) res = 0; - } else if (intr_conf->pin->flags != 0 && flags == 0) { + } else if (intr_conf->intr_mode != 0 && flags == 0) { /* * Interrupt configured, but none requested: Teardown and * release the pin when no other cdevpriv is attached. Otherwise * just detach pin and cdevpriv from each other. */ if (gpioc_intr_reconfig_allowed(priv, intr_conf)) { - res = gpioc_release_pin_intr(intr_conf); + res = gpioc_release_pin_intr(sc, intr_conf); } if (res == 0) res = gpioc_detach_priv_pin(priv, intr_conf); @@ -456,9 +475,10 @@ gpioc_set_intr_config(struct gpioc_softc *sc, struct gpioc_cdevpriv *priv, if (!gpioc_intr_reconfig_allowed(priv, intr_conf)) res = EBUSY; else { - res = gpioc_release_pin_intr(intr_conf); + res = gpioc_release_pin_intr(sc, intr_conf); if (res == 0) - res = gpioc_allocate_pin_intr(intr_conf, flags); + res = gpioc_allocate_pin_intr(sc, intr_conf, + pin, flags); if (res == 0) res = gpioc_attach_priv_pin(priv, intr_conf); if (res == EEXIST) @@ -475,18 +495,16 @@ gpioc_interrupt_handler(void *arg) { struct gpioc_pin_intr *intr_conf; struct gpioc_privs *privs; - struct gpioc_softc *sc; sbintime_t evtime; - uint32_t pin_state; + bool pin_state; intr_conf = arg; - sc = intr_conf->sc; /* Capture time and pin state first. */ evtime = sbinuptime(); - if (intr_conf->pin->flags & GPIO_INTR_EDGE_BOTH) - GPIO_PIN_GET(sc->sc_pdev, intr_conf->pin->pin, &pin_state); - else if (intr_conf->pin->flags & GPIO_INTR_EDGE_RISING) + if (intr_conf->intr_mode & GPIO_INTR_EDGE_BOTH) + gpio_pin_is_active(intr_conf->pin, &pin_state); + else if (intr_conf->intr_mode & GPIO_INTR_EDGE_RISING) pin_state = true; else pin_state = false; @@ -575,18 +593,11 @@ gpioc_attach(device_t dev) sc->sc_pdev = device_get_parent(dev); sc->sc_unit = device_get_unit(dev); - err = GPIO_PIN_MAX(sc->sc_pdev, &sc->sc_npins); - sc->sc_npins++; /* Number of pins is one more than max pin number. */ - if (err != 0) - return (err); + sc->sc_npins = gpiobus_get_npins(dev); sc->sc_pin_intr = malloc(sizeof(struct gpioc_pin_intr) * sc->sc_npins, M_GPIOC, M_WAITOK | M_ZERO); for (int i = 0; i < sc->sc_npins; i++) { - sc->sc_pin_intr[i].pin = malloc(sizeof(struct gpiobus_pin), - M_GPIOC, M_WAITOK | M_ZERO); sc->sc_pin_intr[i].sc = sc; - sc->sc_pin_intr[i].pin->pin = i; - sc->sc_pin_intr[i].pin->dev = sc->sc_pdev; mtx_init(&sc->sc_pin_intr[i].mtx, "gpioc pin", NULL, MTX_DEF); SLIST_INIT(&sc->sc_pin_intr[i].privs); } @@ -610,20 +621,16 @@ static int gpioc_detach(device_t dev) { struct gpioc_softc *sc = device_get_softc(dev); - int err; if (sc->sc_ctl_dev) destroy_dev(sc->sc_ctl_dev); for (int i = 0; i < sc->sc_npins; i++) { mtx_destroy(&sc->sc_pin_intr[i].mtx); - free(sc->sc_pin_intr[i].pin, M_GPIOC); + MPASS(sc->sc_pin_intr[i].pin == NULL); } free(sc->sc_pin_intr, M_GPIOC); - if ((err = bus_generic_detach(dev)) != 0) - return (err); - return (0); } @@ -655,7 +662,7 @@ gpioc_cdevpriv_dtor(void *data) KASSERT(consistency == 1, ("inconsistent links between pin config and cdevpriv")); if (gpioc_intr_reconfig_allowed(priv, pin_link->pin)) { - gpioc_release_pin_intr(pin_link->pin); + gpioc_release_pin_intr(priv->sc, pin_link->pin); } mtx_unlock(&pin_link->pin->mtx); SLIST_REMOVE(&priv->pins, pin_link, gpioc_pins, next); @@ -778,7 +785,6 @@ static int gpioc_ioctl(struct cdev *cdev, u_long cmd, caddr_t arg, int fflag, struct thread *td) { - device_t bus; int max_pin, res; struct gpioc_softc *sc = cdev->si_drv1; struct gpioc_cdevpriv *priv; @@ -789,30 +795,32 @@ gpioc_ioctl(struct cdev *cdev, u_long cmd, caddr_t arg, int fflag, struct gpio_event_config *evcfg; uint32_t caps, intrflags; - bus = GPIO_GET_BUS(sc->sc_pdev); - if (bus == NULL) - return (EINVAL); switch (cmd) { case GPIOMAXPIN: - max_pin = -1; - res = GPIO_PIN_MAX(sc->sc_pdev, &max_pin); + res = 0; + max_pin = sc->sc_npins - 1; bcopy(&max_pin, arg, sizeof(max_pin)); break; case GPIOGETCONFIG: bcopy(arg, &pin, sizeof(pin)); dprintf("get config pin %d\n", pin.gp_pin); - res = GPIO_PIN_GETFLAGS(sc->sc_pdev, pin.gp_pin, + res = GPIOBUS_PIN_GETFLAGS(sc->sc_pdev, sc->sc_dev, pin.gp_pin, &pin.gp_flags); /* Fail early */ - if (res) + if (res != 0) break; res = devfs_get_cdevpriv((void **)&priv); - if (res) + if (res != 0) break; pin.gp_flags |= gpioc_get_intr_config(sc, priv, pin.gp_pin); - GPIO_PIN_GETCAPS(sc->sc_pdev, pin.gp_pin, &pin.gp_caps); - GPIOBUS_PIN_GETNAME(bus, pin.gp_pin, pin.gp_name); + res = GPIOBUS_PIN_GETCAPS(sc->sc_pdev, sc->sc_dev, pin.gp_pin, + &pin.gp_caps); + if (res != 0) + break; + res = GPIOBUS_PIN_GETNAME(sc->sc_pdev, pin.gp_pin, pin.gp_name); + if (res != 0) + break; bcopy(&pin, arg, sizeof(pin)); break; case GPIOSETCONFIG: @@ -821,7 +829,8 @@ gpioc_ioctl(struct cdev *cdev, u_long cmd, caddr_t arg, int fflag, res = devfs_get_cdevpriv((void **)&priv); if (res != 0) break; - res = GPIO_PIN_GETCAPS(sc->sc_pdev, pin.gp_pin, &caps); + res = GPIOBUS_PIN_GETCAPS(sc->sc_pdev, sc->sc_dev, + pin.gp_pin, &caps); if (res != 0) break; res = gpio_check_flags(caps, pin.gp_flags); @@ -847,8 +856,8 @@ gpioc_ioctl(struct cdev *cdev, u_long cmd, caddr_t arg, int fflag, } if (res != 0) break; - res = GPIO_PIN_SETFLAGS(sc->sc_pdev, pin.gp_pin, - (pin.gp_flags & ~GPIO_INTR_MASK)); + res = GPIOBUS_PIN_SETFLAGS(sc->sc_pdev, sc->sc_dev, pin.gp_pin, + pin.gp_flags & ~GPIO_INTR_MASK); if (res != 0) break; res = gpioc_set_intr_config(sc, priv, pin.gp_pin, @@ -856,40 +865,43 @@ gpioc_ioctl(struct cdev *cdev, u_long cmd, caddr_t arg, int fflag, break; case GPIOGET: bcopy(arg, &req, sizeof(req)); - res = GPIO_PIN_GET(sc->sc_pdev, req.gp_pin, + res = GPIOBUS_PIN_GET(sc->sc_pdev, sc->sc_dev, req.gp_pin, &req.gp_value); - dprintf("read pin %d -> %d\n", + if (res != 0) + break; + dprintf("read pin %d -> %d\n", req.gp_pin, req.gp_value); bcopy(&req, arg, sizeof(req)); break; case GPIOSET: bcopy(arg, &req, sizeof(req)); - res = GPIO_PIN_SET(sc->sc_pdev, req.gp_pin, + res = GPIOBUS_PIN_SET(sc->sc_pdev, sc->sc_dev, req.gp_pin, req.gp_value); - dprintf("write pin %d -> %d\n", + dprintf("write pin %d -> %d\n", req.gp_pin, req.gp_value); break; case GPIOTOGGLE: bcopy(arg, &req, sizeof(req)); - dprintf("toggle pin %d\n", + dprintf("toggle pin %d\n", req.gp_pin); - res = GPIO_PIN_TOGGLE(sc->sc_pdev, req.gp_pin); + res = GPIOBUS_PIN_TOGGLE(sc->sc_pdev, sc->sc_dev, req.gp_pin); break; case GPIOSETNAME: bcopy(arg, &pin, sizeof(pin)); dprintf("set name on pin %d\n", pin.gp_pin); - res = GPIOBUS_PIN_SETNAME(bus, pin.gp_pin, + res = GPIOBUS_PIN_SETNAME(sc->sc_pdev, pin.gp_pin, pin.gp_name); break; case GPIOACCESS32: a32 = (struct gpio_access_32 *)arg; - res = GPIO_PIN_ACCESS_32(sc->sc_pdev, a32->first_pin, - a32->clear_pins, a32->change_pins, &a32->orig_pins); + res = GPIOBUS_PIN_ACCESS_32(sc->sc_pdev, sc->sc_dev, + a32->first_pin, a32->clear_pins, a32->change_pins, + &a32->orig_pins); break; case GPIOCONFIG32: c32 = (struct gpio_config_32 *)arg; - res = GPIO_PIN_CONFIG_32(sc->sc_pdev, c32->first_pin, - c32->num_pins, c32->pin_flags); + res = GPIOBUS_PIN_CONFIG_32(sc->sc_pdev, sc->sc_dev, + c32->first_pin, c32->num_pins, c32->pin_flags); break; case GPIOCONFIGEVENTS: evcfg = (struct gpio_event_config *)arg; @@ -1050,9 +1062,6 @@ static device_method_t gpioc_methods[] = { DEVMETHOD(device_probe, gpioc_probe), DEVMETHOD(device_attach, gpioc_attach), DEVMETHOD(device_detach, gpioc_detach), - DEVMETHOD(device_shutdown, bus_generic_shutdown), - DEVMETHOD(device_suspend, bus_generic_suspend), - DEVMETHOD(device_resume, bus_generic_resume), DEVMETHOD_END }; @@ -1063,5 +1072,5 @@ driver_t gpioc_driver = { sizeof(struct gpioc_softc) }; -DRIVER_MODULE(gpioc, gpio, gpioc_driver, 0, 0); +DRIVER_MODULE(gpioc, gpiobus, gpioc_driver, 0, 0); MODULE_VERSION(gpioc, 1); diff --git a/sys/dev/gpio/ofw_gpiobus.c b/sys/dev/gpio/ofw_gpiobus.c index b12b78fac18c..da1bfbc268b8 100644 --- a/sys/dev/gpio/ofw_gpiobus.c +++ b/sys/dev/gpio/ofw_gpiobus.c @@ -426,6 +426,9 @@ ofw_gpiobus_attach(device_t dev) err = gpiobus_init_softc(dev); if (err != 0) return (err); + err = gpiobus_add_gpioc(dev); + if (err != 0) + return (err); bus_identify_children(dev); bus_enumerate_hinted_children(dev); /* diff --git a/sys/dev/ichwd/i6300esbwd.c b/sys/dev/ichwd/i6300esbwd.c new file mode 100644 index 000000000000..d95aeb53c3f5 --- /dev/null +++ b/sys/dev/ichwd/i6300esbwd.c @@ -0,0 +1,245 @@ +/* + * Copyright (c) 2025 The FreeBSD Foundation + * + * SPDX-License-Identifier: BSD-2-Clause + */ + +/* + * Reference: Intel 6300ESB Controller Hub Datasheet Section 16 + */ + +#include <sys/param.h> +#include <sys/eventhandler.h> +#include <sys/kernel.h> +#include <sys/module.h> +#include <sys/sysctl.h> +#include <sys/errno.h> +#include <sys/systm.h> +#include <sys/bus.h> +#include <machine/bus.h> +#include <sys/rman.h> +#include <machine/resource.h> +#include <sys/watchdog.h> + +#include <dev/pci/pcireg.h> + +#include <dev/ichwd/ichwd.h> +#include <dev/ichwd/i6300esbwd.h> + +#include <x86/pci_cfgreg.h> +#include <dev/pci/pcivar.h> +#include <dev/pci/pci_private.h> + +struct i6300esbwd_softc { + device_t dev; + int res_id; + struct resource *res; + eventhandler_tag ev_tag; + bool locked; +}; + +static const struct i6300esbwd_pci_id { + uint16_t id; + const char *name; +} i6300esbwd_pci_devices[] = { + { DEVICEID_6300ESB_2, "6300ESB Watchdog Timer" }, +}; + +static uint16_t +i6300esbwd_cfg_read(struct i6300esbwd_softc *sc) +{ + return (pci_read_config(sc->dev, WDT_CONFIG_REG, 2)); +} + +static void +i6300esbwd_cfg_write(struct i6300esbwd_softc *sc, uint16_t val) +{ + pci_write_config(sc->dev, WDT_CONFIG_REG, val, 2); +} + +static uint8_t +i6300esbwd_lock_read(struct i6300esbwd_softc *sc) +{ + return (pci_read_config(sc->dev, WDT_LOCK_REG, 1)); +} + +static void +i6300esbwd_lock_write(struct i6300esbwd_softc *sc, uint8_t val) +{ + pci_write_config(sc->dev, WDT_LOCK_REG, val, 1); +} + +/* + * According to Intel 6300ESB I/O Controller Hub Datasheet 16.5.2, + * the resource should be unlocked before modifing any registers. + * The way to unlock is by write 0x80, 0x86 to the reload register. + */ +static void +i6300esbwd_unlock_res(struct i6300esbwd_softc *sc) +{ + bus_write_2(sc->res, WDT_RELOAD_REG, WDT_UNLOCK_SEQ_1_VAL); + bus_write_2(sc->res, WDT_RELOAD_REG, WDT_UNLOCK_SEQ_2_VAL); +} + +static int +i6300esbwd_sysctl_locked(SYSCTL_HANDLER_ARGS) +{ + struct i6300esbwd_softc *sc = (struct i6300esbwd_softc *)arg1; + int error; + int result; + + result = sc->locked; + error = sysctl_handle_int(oidp, &result, 0, req); + + if (error || !req->newptr) + return (error); + + if (result == 1 && !sc->locked) { + i6300esbwd_lock_write(sc, i6300esbwd_lock_read(sc) | WDT_LOCK); + sc->locked = true; + } + + return (0); +} + +static void +i6300esbwd_event(void *arg, unsigned int cmd, int *error) +{ + struct i6300esbwd_softc *sc = arg; + uint32_t timeout; + uint16_t regval; + + cmd &= WD_INTERVAL; + if (cmd != 0 && + (cmd < WD_TO_1MS || (cmd - WD_TO_1MS) >= WDT_PRELOAD_BIT)) { + *error = EINVAL; + return; + } + timeout = 1 << (cmd - WD_TO_1MS); + + /* reset the timer to prevent timeout a timeout is about to occur */ + i6300esbwd_unlock_res(sc); + bus_write_2(sc->res, WDT_RELOAD_REG, WDT_RELOAD); + + if (!cmd) { + /* + * when the lock is enabled, we are unable to overwrite LOCK + * register + */ + if (sc->locked) + *error = EPERM; + else + i6300esbwd_lock_write(sc, + i6300esbwd_lock_read(sc) & ~WDT_ENABLE); + return; + } + + i6300esbwd_unlock_res(sc); + bus_write_4(sc->res, WDT_PRELOAD_1_REG, timeout); + + i6300esbwd_unlock_res(sc); + bus_write_4(sc->res, WDT_PRELOAD_2_REG, timeout); + + i6300esbwd_unlock_res(sc); + bus_write_2(sc->res, WDT_RELOAD_REG, WDT_RELOAD); + + if (!sc->locked) { + i6300esbwd_lock_write(sc, WDT_ENABLE); + regval = i6300esbwd_lock_read(sc); + sc->locked = regval & WDT_LOCK; + } +} + +static int +i6300esbwd_probe(device_t dev) +{ + const struct i6300esbwd_pci_id *pci_id; + uint16_t pci_dev_id; + int err = ENXIO; + + if (pci_get_vendor(dev) != VENDORID_INTEL) + goto end; + + pci_dev_id = pci_get_device(dev); + for (pci_id = i6300esbwd_pci_devices; + pci_id < i6300esbwd_pci_devices + nitems(i6300esbwd_pci_devices); + ++pci_id) { + if (pci_id->id == pci_dev_id) { + device_set_desc(dev, pci_id->name); + err = BUS_PROBE_DEFAULT; + break; + } + } + +end: + return (err); +} + +static int +i6300esbwd_attach(device_t dev) +{ + struct i6300esbwd_softc *sc = device_get_softc(dev); + uint16_t regval; + + sc->dev = dev; + sc->res_id = PCIR_BAR(0); + sc->res = bus_alloc_resource_any(dev, SYS_RES_MEMORY, &sc->res_id, + RF_ACTIVE); + if (sc->res == NULL) { + device_printf(dev, "unable to map memory region\n"); + return (ENXIO); + } + + i6300esbwd_cfg_write(sc, WDT_INT_TYPE_DISABLED_VAL); + regval = i6300esbwd_lock_read(sc); + if (regval & WDT_LOCK) + sc->locked = true; + else { + sc->locked = false; + i6300esbwd_lock_write(sc, WDT_TOUT_CNF_WT_MODE); + } + + i6300esbwd_unlock_res(sc); + bus_write_2(sc->res, WDT_RELOAD_REG, WDT_RELOAD | WDT_TIMEOUT); + + sc->ev_tag = EVENTHANDLER_REGISTER(watchdog_list, i6300esbwd_event, sc, + 0); + + SYSCTL_ADD_PROC(device_get_sysctl_ctx(dev), + SYSCTL_CHILDREN(device_get_sysctl_tree(dev)), OID_AUTO, "locked", + CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, sc, 0, + i6300esbwd_sysctl_locked, "I", + "Lock the timer so that we cannot disable it"); + + return (0); +} + +static int +i6300esbwd_detach(device_t dev) +{ + struct i6300esbwd_softc *sc = device_get_softc(dev); + + if (sc->ev_tag) + EVENTHANDLER_DEREGISTER(watchdog_list, sc->ev_tag); + + if (sc->res) + bus_release_resource(dev, SYS_RES_MEMORY, sc->res_id, sc->res); + + return (0); +} + +static device_method_t i6300esbwd_methods[] = { + DEVMETHOD(device_probe, i6300esbwd_probe), + DEVMETHOD(device_attach, i6300esbwd_attach), + DEVMETHOD(device_detach, i6300esbwd_detach), + DEVMETHOD(device_shutdown, i6300esbwd_detach), + DEVMETHOD_END +}; + +static driver_t i6300esbwd_driver = { + "i6300esbwd", + i6300esbwd_methods, + sizeof(struct i6300esbwd_softc), +}; + +DRIVER_MODULE(i6300esbwd, pci, i6300esbwd_driver, NULL, NULL); diff --git a/sys/dev/ichwd/i6300esbwd.h b/sys/dev/ichwd/i6300esbwd.h new file mode 100644 index 000000000000..39ed5d5a84f6 --- /dev/null +++ b/sys/dev/ichwd/i6300esbwd.h @@ -0,0 +1,46 @@ +/* + * Copyright (c) 2025 The FreeBSD Foundation + * + * SPDX-License-Identifier: BSD-2-Clause + */ + +#ifndef _I6300ESBWD_H_ +#define _I6300ESBWD_H_ + +#define WDT_CONFIG_REG 0x60 +#define WDT_LOCK_REG 0x68 + +#define WDT_PRELOAD_1_REG 0x00 +#define WDT_PRELOAD_2_REG 0x04 +#define WDT_INTR_REG 0x08 +#define WDT_RELOAD_REG 0x0C + +/* For config register */ +#define WDT_OUTPUT_EN (0x1 << 5) +#define WDT_PRE_SEL (0x1 << 2) +#define WDT_INT_TYPE_BITS (0x3) +#define WDT_INT_TYPE_IRQ_VAL (0x0) +#define WDT_INT_TYPE_RES_VAL (0x1) +#define WDT_INT_TYPE_SMI_VAL (0x2) +#define WDT_INT_TYPE_DISABLED_VAL (0x3) + +/* For lock register */ +#define WDT_TOUT_CNF_WT_MODE (0x0 << 2) +#define WDT_TOUT_CNF_FR_MODE (0x1 << 2) +#define WDT_ENABLE (0x02) +#define WDT_LOCK (0x01) + +/* For preload 1/2 registers */ +#define WDT_PRELOAD_BIT 20 +#define WDT_PRELOAD_BITS ((0x1 << WDT_PRELOAD_BIT) - 1) + +/* For interrupt register */ +#define WDT_INTR_ACT (0x01 << 0) + +/* For reload register */ +#define WDT_TIMEOUT (0x01 << 9) +#define WDT_RELOAD (0x01 << 8) +#define WDT_UNLOCK_SEQ_1_VAL 0x80 +#define WDT_UNLOCK_SEQ_2_VAL 0x86 + +#endif /* _I6300ESBWD_H_ */ diff --git a/sys/dev/ichwd/ichwd.c b/sys/dev/ichwd/ichwd.c index cade2cc4fb45..5481553cc175 100644 --- a/sys/dev/ichwd/ichwd.c +++ b/sys/dev/ichwd/ichwd.c @@ -90,7 +90,7 @@ static struct ichwd_device ichwd_devices[] = { { DEVICEID_82801E, "Intel 82801E watchdog timer", 5, 1 }, { DEVICEID_82801EB, "Intel 82801EB watchdog timer", 5, 1 }, { DEVICEID_82801EBR, "Intel 82801EB/ER watchdog timer", 5, 1 }, - { DEVICEID_6300ESB, "Intel 6300ESB watchdog timer", 5, 1 }, + { DEVICEID_6300ESB_1, "Intel 6300ESB watchdog timer", 5, 1 }, { DEVICEID_82801FBR, "Intel 82801FB/FR watchdog timer", 6, 2 }, { DEVICEID_ICH6M, "Intel ICH6M watchdog timer", 6, 2 }, { DEVICEID_ICH6W, "Intel ICH6W watchdog timer", 6, 2 }, diff --git a/sys/dev/ichwd/ichwd.h b/sys/dev/ichwd/ichwd.h index 90fda08b74c1..72d0ca1cd6aa 100644 --- a/sys/dev/ichwd/ichwd.h +++ b/sys/dev/ichwd/ichwd.h @@ -151,7 +151,8 @@ struct ichwd_softc { #define DEVICEID_82801E 0x2450 #define DEVICEID_82801EB 0x24dc #define DEVICEID_82801EBR 0x24d0 -#define DEVICEID_6300ESB 0x25a1 +#define DEVICEID_6300ESB_1 0x25a1 +#define DEVICEID_6300ESB_2 0x25ab #define DEVICEID_82801FBR 0x2640 #define DEVICEID_ICH6M 0x2641 #define DEVICEID_ICH6W 0x2642 diff --git a/sys/dev/nvme/nvme.c b/sys/dev/nvme/nvme.c index 84f365024f13..ead91f0d01fe 100644 --- a/sys/dev/nvme/nvme.c +++ b/sys/dev/nvme/nvme.c @@ -295,7 +295,6 @@ nvme_register_consumer(nvme_cons_ns_fn_t ns_fn, nvme_cons_ctrlr_fn_t ctrlr_fn, void nvme_unregister_consumer(struct nvme_consumer *consumer) { - consumer->id = INVALID_CONSUMER_ID; } diff --git a/sys/dev/nvme/nvme_ahci.c b/sys/dev/nvme/nvme_ahci.c index 888207a454f7..b06661226d34 100644 --- a/sys/dev/nvme/nvme_ahci.c +++ b/sys/dev/nvme/nvme_ahci.c @@ -124,6 +124,5 @@ bad: static int nvme_ahci_detach(device_t dev) { - return (nvme_detach(dev)); } diff --git a/sys/dev/nvme/nvme_ctrlr.c b/sys/dev/nvme/nvme_ctrlr.c index fd7f00ced14b..49960b0f920a 100644 --- a/sys/dev/nvme/nvme_ctrlr.c +++ b/sys/dev/nvme/nvme_ctrlr.c @@ -597,7 +597,6 @@ nvme_ctrlr_construct_namespaces(struct nvme_controller *ctrlr) static bool is_log_page_id_valid(uint8_t page_id) { - switch (page_id) { case NVME_LOG_ERROR: case NVME_LOG_HEALTH_INFORMATION: @@ -653,7 +652,6 @@ static void nvme_ctrlr_log_critical_warnings(struct nvme_controller *ctrlr, uint8_t state) { - if (state & NVME_CRIT_WARN_ST_AVAILABLE_SPARE) nvme_printf(ctrlr, "SMART WARNING: available spare space below threshold\n"); @@ -781,7 +779,6 @@ nvme_ctrlr_configure_aer(struct nvme_controller *ctrlr) static void nvme_ctrlr_configure_int_coalescing(struct nvme_controller *ctrlr) { - ctrlr->int_coal_time = 0; TUNABLE_INT_FETCH("hw.nvme.int_coal_time", &ctrlr->int_coal_time); @@ -1420,8 +1417,8 @@ nvme_ctrlr_linux_passthru_cmd(struct nvme_controller *ctrlr, req->cmd.opc = npc->opcode; req->cmd.fuse = npc->flags; - req->cmd.rsvd2 = htole16(npc->cdw2); - req->cmd.rsvd3 = htole16(npc->cdw3); + req->cmd.rsvd2 = htole32(npc->cdw2); + req->cmd.rsvd3 = htole32(npc->cdw3); req->cmd.cdw10 = htole32(npc->cdw10); req->cmd.cdw11 = htole32(npc->cdw11); req->cmd.cdw12 = htole32(npc->cdw12); @@ -1776,7 +1773,6 @@ void nvme_ctrlr_submit_admin_request(struct nvme_controller *ctrlr, struct nvme_request *req) { - nvme_qpair_submit_request(&ctrlr->adminq, req); } @@ -1793,14 +1789,12 @@ nvme_ctrlr_submit_io_request(struct nvme_controller *ctrlr, device_t nvme_ctrlr_get_device(struct nvme_controller *ctrlr) { - return (ctrlr->dev); } const struct nvme_controller_data * nvme_ctrlr_get_data(struct nvme_controller *ctrlr) { - return (&ctrlr->cdata); } @@ -1853,7 +1847,6 @@ nvme_ctrlr_suspend(struct nvme_controller *ctrlr) int nvme_ctrlr_resume(struct nvme_controller *ctrlr) { - /* * Can't touch failed controllers, so nothing to do to resume. */ diff --git a/sys/dev/nvme/nvme_ctrlr_cmd.c b/sys/dev/nvme/nvme_ctrlr_cmd.c index 993a7718356d..5a44ed425acb 100644 --- a/sys/dev/nvme/nvme_ctrlr_cmd.c +++ b/sys/dev/nvme/nvme_ctrlr_cmd.c @@ -281,7 +281,6 @@ nvme_ctrlr_cmd_get_error_page(struct nvme_controller *ctrlr, struct nvme_error_information_entry *payload, uint32_t num_entries, nvme_cb_fn_t cb_fn, void *cb_arg) { - KASSERT(num_entries > 0, ("%s called with num_entries==0\n", __func__)); /* Controller's error log page entries is 0-based. */ @@ -302,7 +301,6 @@ nvme_ctrlr_cmd_get_health_information_page(struct nvme_controller *ctrlr, uint32_t nsid, struct nvme_health_information_page *payload, nvme_cb_fn_t cb_fn, void *cb_arg) { - nvme_ctrlr_cmd_get_log_page(ctrlr, NVME_LOG_HEALTH_INFORMATION, nsid, payload, sizeof(*payload), cb_fn, cb_arg); } @@ -311,7 +309,6 @@ void nvme_ctrlr_cmd_get_firmware_page(struct nvme_controller *ctrlr, struct nvme_firmware_page *payload, nvme_cb_fn_t cb_fn, void *cb_arg) { - nvme_ctrlr_cmd_get_log_page(ctrlr, NVME_LOG_FIRMWARE_SLOT, NVME_GLOBAL_NAMESPACE_TAG, payload, sizeof(*payload), cb_fn, cb_arg); diff --git a/sys/dev/nvme/nvme_ns.c b/sys/dev/nvme/nvme_ns.c index 3f29382fe42f..e84d2066930e 100644 --- a/sys/dev/nvme/nvme_ns.c +++ b/sys/dev/nvme/nvme_ns.c @@ -129,7 +129,6 @@ static int nvme_ns_close(struct cdev *dev __unused, int flags, int fmt __unused, struct thread *td) { - return (0); } @@ -231,7 +230,6 @@ nvme_ns_get_model_number(struct nvme_namespace *ns) const struct nvme_namespace_data * nvme_ns_get_data(struct nvme_namespace *ns) { - return (&ns->data); } @@ -631,7 +629,6 @@ nvme_ns_construct(struct nvme_namespace *ns, uint32_t id, void nvme_ns_destruct(struct nvme_namespace *ns) { - if (ns->cdev != NULL) { if (ns->cdev->si_drv2 != NULL) destroy_dev(ns->cdev->si_drv2); diff --git a/sys/dev/nvme/nvme_pci.c b/sys/dev/nvme/nvme_pci.c index 29b49b7df403..c07a68d2f0dc 100644 --- a/sys/dev/nvme/nvme_pci.c +++ b/sys/dev/nvme/nvme_pci.c @@ -151,7 +151,6 @@ nvme_pci_probe (device_t device) static int nvme_ctrlr_allocate_bar(struct nvme_controller *ctrlr) { - ctrlr->resource_id = PCIR_BAR(0); ctrlr->resource = bus_alloc_resource_any(ctrlr->dev, SYS_RES_MEMORY, diff --git a/sys/dev/nvme/nvme_private.h b/sys/dev/nvme/nvme_private.h index 36f00fedc48e..52f9e12f8f9a 100644 --- a/sys/dev/nvme/nvme_private.h +++ b/sys/dev/nvme/nvme_private.h @@ -459,8 +459,7 @@ int nvme_detach(device_t dev); * vast majority of these without waiting for a tick plus scheduling delays. Since * these are on startup, this drastically reduces startup time. */ -static __inline -void +static __inline void nvme_completion_poll(struct nvme_completion_poll_status *status) { int timeout = ticks + 10 * hz; diff --git a/sys/dev/nvme/nvme_qpair.c b/sys/dev/nvme/nvme_qpair.c index bd8626e32209..4f2c44da3b4f 100644 --- a/sys/dev/nvme/nvme_qpair.c +++ b/sys/dev/nvme/nvme_qpair.c @@ -793,7 +793,6 @@ nvme_admin_qpair_destroy(struct nvme_qpair *qpair) void nvme_io_qpair_destroy(struct nvme_qpair *qpair) { - nvme_qpair_destroy(qpair); } @@ -1202,7 +1201,6 @@ _nvme_qpair_submit_request(struct nvme_qpair *qpair, struct nvme_request *req) void nvme_qpair_submit_request(struct nvme_qpair *qpair, struct nvme_request *req) { - mtx_lock(&qpair->lock); _nvme_qpair_submit_request(qpair, req); mtx_unlock(&qpair->lock); @@ -1226,7 +1224,6 @@ nvme_qpair_enable(struct nvme_qpair *qpair) void nvme_qpair_reset(struct nvme_qpair *qpair) { - qpair->sq_head = qpair->sq_tail = qpair->cq_head = 0; /* diff --git a/sys/dev/nvme/nvme_sim.c b/sys/dev/nvme/nvme_sim.c index 4974bb718222..a06774a64761 100644 --- a/sys/dev/nvme/nvme_sim.c +++ b/sys/dev/nvme/nvme_sim.c @@ -301,7 +301,6 @@ nvme_sim_action(struct cam_sim *sim, union ccb *ccb) static void nvme_sim_poll(struct cam_sim *sim) { - nvme_ctrlr_poll(sim2ctrlr(sim)); } diff --git a/sys/dev/nvme/nvme_sysctl.c b/sys/dev/nvme/nvme_sysctl.c index a5a44721f9f9..50d19e730a16 100644 --- a/sys/dev/nvme/nvme_sysctl.c +++ b/sys/dev/nvme/nvme_sysctl.c @@ -153,7 +153,6 @@ nvme_sysctl_timeout_period(SYSCTL_HANDLER_ARGS) static void nvme_qpair_reset_stats(struct nvme_qpair *qpair) { - /* * Reset the values. Due to sanity checks in * nvme_qpair_process_completions, we reset the number of interrupt diff --git a/sys/dev/puc/pucdata.c b/sys/dev/puc/pucdata.c index e911a407cca9..436af76001da 100644 --- a/sys/dev/puc/pucdata.c +++ b/sys/dev/puc/pucdata.c @@ -64,6 +64,7 @@ static puc_config_f puc_config_quatech; static puc_config_f puc_config_syba; static puc_config_f puc_config_siig; static puc_config_f puc_config_sunix; +static puc_config_f puc_config_systembase; static puc_config_f puc_config_timedia; static puc_config_f puc_config_titan; @@ -1705,6 +1706,23 @@ const struct puc_cfg puc_pci_devices[] = { PUC_PORT_4S, 0x10, 0, 8, .config_function = puc_config_icbook }, + + /* + * Systembase cards using SB16C1050 UARTs: + */ + { 0x14a1, 0x0008, 0x14a1, 0x0008, + "Systembase SB16C1058", + DEFAULT_RCLK * 8, + PUC_PORT_8S, 0x10, 0, 8, + .config_function = puc_config_systembase, + }, + { 0x14a1, 0x0004, 0x14a1, 0x0004, + "Systembase SB16C1054", + DEFAULT_RCLK * 8, + PUC_PORT_4S, 0x10, 0, 8, + .config_function = puc_config_systembase, + }, + { 0xffff, 0, 0xffff, 0, NULL, 0 } }; @@ -2294,3 +2312,28 @@ puc_config_titan(struct puc_softc *sc __unused, enum puc_cfg_cmd cmd, } return (ENXIO); } + +static int +puc_config_systembase(struct puc_softc *sc __unused, + enum puc_cfg_cmd cmd, int port, intptr_t *res) +{ + struct puc_bar *bar; + + switch (cmd) { + case PUC_CFG_SETUP: + bar = puc_get_bar(sc, 0x14); + if (bar == NULL) + return (ENXIO); + + /* + * The Systembase SB16C1058 (and probably other devices + * based on the SB16C1050 UART core) require poking a + * register in the *other* RID to turn on interrupts. + */ + bus_write_1(bar->b_res, /* OPT_IMRREG0 */ 0xc, 0xff); + return (0); + default: + break; + } + return (ENXIO); +} diff --git a/sys/dev/tpm/tpm_tis_core.c b/sys/dev/tpm/tpm_tis_core.c index d8421f8156c9..4159de4daf3b 100644 --- a/sys/dev/tpm/tpm_tis_core.c +++ b/sys/dev/tpm/tpm_tis_core.c @@ -97,6 +97,7 @@ tpmtis_attach(device_t dev) { struct tpm_sc *sc; int result; + int poll = 0; sc = device_get_softc(dev); sc->dev = dev; @@ -105,6 +106,12 @@ tpmtis_attach(device_t dev) sx_init(&sc->dev_lock, "TPM driver lock"); sc->buf = malloc(TPM_BUFSIZE, M_TPM20, M_WAITOK); + resource_int_value("tpm", device_get_unit(dev), "use_polling", &poll); + if (poll != 0) { + device_printf(dev, "Using poll method to get TPM operation status \n"); + goto skip_irq; + } + sc->irq_rid = 0; sc->irq_res = bus_alloc_resource_any(dev, SYS_RES_IRQ, &sc->irq_rid, RF_ACTIVE | RF_SHAREABLE); diff --git a/sys/dev/uart/uart_bus_pci.c b/sys/dev/uart/uart_bus_pci.c index 14ac213066b8..22af8ee8663c 100644 --- a/sys/dev/uart/uart_bus_pci.c +++ b/sys/dev/uart/uart_bus_pci.c @@ -141,6 +141,8 @@ static const struct pci_id pci_ns8250_ids[] = { 0x10, 16384000 }, { 0x1415, 0xc120, 0xffff, 0, "Oxford Semiconductor OXPCIe952 PCIe 16950 UART", 0x10 }, +{ 0x14a1, 0x0008, 0x14a1, 0x0008, "Systembase SB16C1058", + 0x10, 8 * DEFAULT_RCLK, }, { 0x14e4, 0x160a, 0xffff, 0, "Broadcom TruManage UART", 0x10, 128 * DEFAULT_RCLK, 2}, { 0x14e4, 0x4344, 0xffff, 0, "Sony Ericsson GC89 PC Card", 0x10}, diff --git a/sys/dev/usb/usb_hub.c b/sys/dev/usb/usb_hub.c index e3509862ef54..ee9d8ab0c9bb 100644 --- a/sys/dev/usb/usb_hub.c +++ b/sys/dev/usb/usb_hub.c @@ -954,7 +954,8 @@ done: * packet. This function is called having the "bus_mtx" locked. *------------------------------------------------------------------------*/ void -uhub_root_intr(struct usb_bus *bus, const uint8_t *ptr, uint8_t len) +uhub_root_intr(struct usb_bus *bus, + const uint8_t *ptr __unused, uint8_t len __unused) { USB_BUS_LOCK_ASSERT(bus, MA_OWNED); diff --git a/sys/dev/virtio/network/if_vtnet.c b/sys/dev/virtio/network/if_vtnet.c index ecb3dbb370e5..867da80a53a8 100644 --- a/sys/dev/virtio/network/if_vtnet.c +++ b/sys/dev/virtio/network/if_vtnet.c @@ -2551,8 +2551,10 @@ vtnet_txq_enqueue_buf(struct vtnet_txq *txq, struct mbuf **m_head, error = sglist_append_mbuf(sg, m); if (error) { m = m_defrag(m, M_NOWAIT); - if (m == NULL) + if (m == NULL) { + sc->vtnet_stats.tx_defrag_failed++; goto fail; + } *m_head = m; sc->vtnet_stats.tx_defragged++; @@ -2568,7 +2570,6 @@ vtnet_txq_enqueue_buf(struct vtnet_txq *txq, struct mbuf **m_head, return (error); fail: - sc->vtnet_stats.tx_defrag_failed++; m_freem(*m_head); *m_head = NULL; @@ -3813,9 +3814,9 @@ vtnet_rx_filter_mac(struct vtnet_softc *sc) if_printf(ifp, "error setting host MAC filter table\n"); out: - if (promisc != 0 && vtnet_set_promisc(sc, true) != 0) + if (promisc && vtnet_set_promisc(sc, true) != 0) if_printf(ifp, "cannot enable promiscuous mode\n"); - if (allmulti != 0 && vtnet_set_allmulti(sc, true) != 0) + if (allmulti && vtnet_set_allmulti(sc, true) != 0) if_printf(ifp, "cannot enable all-multicast mode\n"); } @@ -4100,21 +4101,29 @@ vtnet_setup_rxq_sysctl(struct sysctl_ctx_list *ctx, stats = &rxq->vtnrx_stats; - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "ipackets", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "ipackets", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_ipackets, "Receive packets"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "ibytes", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "ibytes", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_ibytes, "Receive bytes"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "iqdrops", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "iqdrops", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_iqdrops, "Receive drops"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "ierrors", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "ierrors", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_ierrors, "Receive errors"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "csum", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "csum", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_csum, "Receive checksum offloaded"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "csum_failed", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "csum_failed", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_csum_failed, "Receive checksum offload failed"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "host_lro", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "host_lro", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_host_lro, "Receive host segmentation offloaded"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "rescheduled", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "rescheduled", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vrxs_rescheduled, "Receive interrupt handler rescheduled"); } @@ -4135,17 +4144,23 @@ vtnet_setup_txq_sysctl(struct sysctl_ctx_list *ctx, stats = &txq->vtntx_stats; - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "opackets", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "opackets", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vtxs_opackets, "Transmit packets"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "obytes", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "obytes", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vtxs_obytes, "Transmit bytes"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "omcasts", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "omcasts", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vtxs_omcasts, "Transmit multicasts"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "csum", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "csum", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vtxs_csum, "Transmit checksum offloaded"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "tso", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "tso", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vtxs_tso, "Transmit TCP segmentation offloaded"); - SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "rescheduled", CTLFLAG_RD, + SYSCTL_ADD_UQUAD(ctx, list, OID_AUTO, "rescheduled", + CTLFLAG_RD | CTLFLAG_STATS, &stats->vtxs_rescheduled, "Transmit interrupt handler rescheduled"); } @@ -4170,6 +4185,102 @@ vtnet_setup_queue_sysctl(struct vtnet_softc *sc) } } +static int +vtnet_sysctl_rx_csum_failed(SYSCTL_HANDLER_ARGS) +{ + struct vtnet_softc *sc = (struct vtnet_softc *)arg1; + struct vtnet_statistics *stats = &sc->vtnet_stats; + struct vtnet_rxq_stats *rxst; + int i; + + stats->rx_csum_failed = 0; + for (i = 0; i < sc->vtnet_max_vq_pairs; i++) { + rxst = &sc->vtnet_rxqs[i].vtnrx_stats; + stats->rx_csum_failed += rxst->vrxs_csum_failed; + } + return (sysctl_handle_64(oidp, NULL, stats->rx_csum_failed, req)); +} + +static int +vtnet_sysctl_rx_csum_offloaded(SYSCTL_HANDLER_ARGS) +{ + struct vtnet_softc *sc = (struct vtnet_softc *)arg1; + struct vtnet_statistics *stats = &sc->vtnet_stats; + struct vtnet_rxq_stats *rxst; + int i; + + stats->rx_csum_offloaded = 0; + for (i = 0; i < sc->vtnet_max_vq_pairs; i++) { + rxst = &sc->vtnet_rxqs[i].vtnrx_stats; + stats->rx_csum_offloaded += rxst->vrxs_csum; + } + return (sysctl_handle_64(oidp, NULL, stats->rx_csum_offloaded, req)); +} + +static int +vtnet_sysctl_rx_task_rescheduled(SYSCTL_HANDLER_ARGS) +{ + struct vtnet_softc *sc = (struct vtnet_softc *)arg1; + struct vtnet_statistics *stats = &sc->vtnet_stats; + struct vtnet_rxq_stats *rxst; + int i; + + stats->rx_task_rescheduled = 0; + for (i = 0; i < sc->vtnet_max_vq_pairs; i++) { + rxst = &sc->vtnet_rxqs[i].vtnrx_stats; + stats->rx_task_rescheduled += rxst->vrxs_rescheduled; + } + return (sysctl_handle_64(oidp, NULL, stats->rx_task_rescheduled, req)); +} + +static int +vtnet_sysctl_tx_csum_offloaded(SYSCTL_HANDLER_ARGS) +{ + struct vtnet_softc *sc = (struct vtnet_softc *)arg1; + struct vtnet_statistics *stats = &sc->vtnet_stats; + struct vtnet_txq_stats *txst; + int i; + + stats->tx_csum_offloaded = 0; + for (i = 0; i < sc->vtnet_max_vq_pairs; i++) { + txst = &sc->vtnet_txqs[i].vtntx_stats; + stats->tx_csum_offloaded += txst->vtxs_csum; + } + return (sysctl_handle_64(oidp, NULL, stats->tx_csum_offloaded, req)); +} + +static int +vtnet_sysctl_tx_tso_offloaded(SYSCTL_HANDLER_ARGS) +{ + struct vtnet_softc *sc = (struct vtnet_softc *)arg1; + struct vtnet_statistics *stats = &sc->vtnet_stats; + struct vtnet_txq_stats *txst; + int i; + + stats->tx_tso_offloaded = 0; + for (i = 0; i < sc->vtnet_max_vq_pairs; i++) { + txst = &sc->vtnet_txqs[i].vtntx_stats; + stats->tx_tso_offloaded += txst->vtxs_tso; + } + return (sysctl_handle_64(oidp, NULL, stats->tx_tso_offloaded, req)); +} + +static int +vtnet_sysctl_tx_task_rescheduled(SYSCTL_HANDLER_ARGS) +{ + struct vtnet_softc *sc = (struct vtnet_softc *)arg1; + struct vtnet_statistics *stats = &sc->vtnet_stats; + struct vtnet_txq_stats *txst; + int i; + + stats->tx_task_rescheduled = 0; + for (i = 0; i < sc->vtnet_max_vq_pairs; i++) { + txst = &sc->vtnet_txqs[i].vtntx_stats; + stats->tx_task_rescheduled += txst->vtxs_rescheduled; + } + return (sysctl_handle_64(oidp, NULL, stats->tx_task_rescheduled, req)); +} + static void vtnet_setup_stat_sysctl(struct sysctl_ctx_list *ctx, struct sysctl_oid_list *child, struct vtnet_softc *sc) @@ -4189,69 +4300,75 @@ vtnet_setup_stat_sysctl(struct sysctl_ctx_list *ctx, stats->tx_task_rescheduled = txaccum.vtxs_rescheduled; SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "mbuf_alloc_failed", - CTLFLAG_RD, &stats->mbuf_alloc_failed, + CTLFLAG_RD | CTLFLAG_STATS, &stats->mbuf_alloc_failed, "Mbuf cluster allocation failures"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_frame_too_large", - CTLFLAG_RD, &stats->rx_frame_too_large, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_frame_too_large, "Received frame larger than the mbuf chain"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_enq_replacement_failed", - CTLFLAG_RD, &stats->rx_enq_replacement_failed, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_enq_replacement_failed, "Enqueuing the replacement receive mbuf failed"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_mergeable_failed", - CTLFLAG_RD, &stats->rx_mergeable_failed, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_mergeable_failed, "Mergeable buffers receive failures"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_csum_bad_ethtype", - CTLFLAG_RD, &stats->rx_csum_bad_ethtype, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_csum_bad_ethtype, "Received checksum offloaded buffer with unsupported " "Ethernet type"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_csum_bad_ipproto", - CTLFLAG_RD, &stats->rx_csum_bad_ipproto, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_csum_bad_ipproto, "Received checksum offloaded buffer with incorrect IP protocol"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_csum_bad_offset", - CTLFLAG_RD, &stats->rx_csum_bad_offset, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_csum_bad_offset, "Received checksum offloaded buffer with incorrect offset"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_csum_bad_proto", - CTLFLAG_RD, &stats->rx_csum_bad_proto, + CTLFLAG_RD | CTLFLAG_STATS, &stats->rx_csum_bad_proto, "Received checksum offloaded buffer with incorrect protocol"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_csum_failed", - CTLFLAG_RD, &stats->rx_csum_failed, + SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_csum_failed", + CTLTYPE_U64 | CTLFLAG_RD | CTLFLAG_STATS, + sc, 0, vtnet_sysctl_rx_csum_failed, "QU", "Received buffer checksum offload failed"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_csum_offloaded", - CTLFLAG_RD, &stats->rx_csum_offloaded, + SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_csum_offloaded", + CTLTYPE_U64 | CTLFLAG_RD | CTLFLAG_STATS, + sc, 0, vtnet_sysctl_rx_csum_offloaded, "QU", "Received buffer checksum offload succeeded"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "rx_task_rescheduled", - CTLFLAG_RD, &stats->rx_task_rescheduled, + SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "rx_task_rescheduled", + CTLTYPE_U64 | CTLFLAG_RD | CTLFLAG_STATS, + sc, 0, vtnet_sysctl_rx_task_rescheduled, "QU", "Times the receive interrupt task rescheduled itself"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_csum_unknown_ethtype", - CTLFLAG_RD, &stats->tx_csum_unknown_ethtype, + CTLFLAG_RD | CTLFLAG_STATS, &stats->tx_csum_unknown_ethtype, "Aborted transmit of checksum offloaded buffer with unknown " "Ethernet type"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_csum_proto_mismatch", - CTLFLAG_RD, &stats->tx_csum_proto_mismatch, + CTLFLAG_RD | CTLFLAG_STATS, &stats->tx_csum_proto_mismatch, "Aborted transmit of checksum offloaded buffer because mismatched " "protocols"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_tso_not_tcp", - CTLFLAG_RD, &stats->tx_tso_not_tcp, + CTLFLAG_RD | CTLFLAG_STATS, &stats->tx_tso_not_tcp, "Aborted transmit of TSO buffer with non TCP protocol"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_tso_without_csum", - CTLFLAG_RD, &stats->tx_tso_without_csum, + CTLFLAG_RD | CTLFLAG_STATS, &stats->tx_tso_without_csum, "Aborted transmit of TSO buffer without TCP checksum offload"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_defragged", - CTLFLAG_RD, &stats->tx_defragged, + CTLFLAG_RD | CTLFLAG_STATS, &stats->tx_defragged, "Transmit mbufs defragged"); SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_defrag_failed", - CTLFLAG_RD, &stats->tx_defrag_failed, + CTLFLAG_RD | CTLFLAG_STATS, &stats->tx_defrag_failed, "Aborted transmit of buffer because defrag failed"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_csum_offloaded", - CTLFLAG_RD, &stats->tx_csum_offloaded, + SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "tx_csum_offloaded", + CTLTYPE_U64 | CTLFLAG_RD | CTLFLAG_STATS, + sc, 0, vtnet_sysctl_tx_csum_offloaded, "QU", "Offloaded checksum of transmitted buffer"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_tso_offloaded", - CTLFLAG_RD, &stats->tx_tso_offloaded, + SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "tx_tso_offloaded", + CTLTYPE_U64 | CTLFLAG_RD | CTLFLAG_STATS, + sc, 0, vtnet_sysctl_tx_tso_offloaded, "QU", "Segmentation offload of transmitted buffer"); - SYSCTL_ADD_UQUAD(ctx, child, OID_AUTO, "tx_task_rescheduled", - CTLFLAG_RD, &stats->tx_task_rescheduled, + SYSCTL_ADD_PROC(ctx, child, OID_AUTO, "tx_task_rescheduled", + CTLTYPE_U64 | CTLFLAG_RD | CTLFLAG_STATS, + sc, 0, vtnet_sysctl_tx_task_rescheduled, "QU", "Times the transmit interrupt task rescheduled itself"); } diff --git a/sys/fs/nfsclient/nfs_clvnops.c b/sys/fs/nfsclient/nfs_clvnops.c index a8b06fdb261b..eee571a04821 100644 --- a/sys/fs/nfsclient/nfs_clvnops.c +++ b/sys/fs/nfsclient/nfs_clvnops.c @@ -3474,7 +3474,7 @@ nfs_advlock(struct vop_advlock_args *ap) u_quad_t size; struct nfsmount *nmp; - error = NFSVOPLOCK(vp, LK_SHARED); + error = NFSVOPLOCK(vp, LK_EXCLUSIVE); if (error != 0) return (EBADF); nmp = VFSTONFS(vp->v_mount); @@ -3511,11 +3511,6 @@ nfs_advlock(struct vop_advlock_args *ap) cred = p->p_ucred; else cred = td->td_ucred; - NFSVOPLOCK(vp, LK_UPGRADE | LK_RETRY); - if (VN_IS_DOOMED(vp)) { - error = EBADF; - goto out; - } /* * If this is unlocking a write locked region, flush and diff --git a/sys/fs/tarfs/tarfs_vnops.c b/sys/fs/tarfs/tarfs_vnops.c index acf18de5ab51..c110107bb210 100644 --- a/sys/fs/tarfs/tarfs_vnops.c +++ b/sys/fs/tarfs/tarfs_vnops.c @@ -334,6 +334,10 @@ tarfs_readdir(struct vop_readdir_args *ap) tnp, tnp->name, uio->uio_offset, uio->uio_resid); if (uio->uio_offset == TARFS_COOKIE_EOF) { + if (eofflag != NULL) { + TARFS_DPF(VNODE, "%s: Setting EOF flag\n", __func__); + *eofflag = 1; + } TARFS_DPF(VNODE, "%s: EOF\n", __func__); return (0); } diff --git a/sys/kern/subr_witness.c b/sys/kern/subr_witness.c index ab47b6ad29a3..a65c3ca128d9 100644 --- a/sys/kern/subr_witness.c +++ b/sys/kern/subr_witness.c @@ -57,7 +57,7 @@ * b : public affirmation by word or example of usually * religious faith or conviction <the heroic witness to divine * life -- Pilot> - * 6 capitalized : a member of the Jehovah's Witnesses + * 6 capitalized : a member of the Jehovah's Witnesses */ /* @@ -131,7 +131,7 @@ #define LI_SLEEPABLE 0x00040000 /* Lock may be held while sleeping. */ #ifndef WITNESS_COUNT -#define WITNESS_COUNT 1536 +#define WITNESS_COUNT 1536 #endif #define WITNESS_HASH_SIZE 251 /* Prime, gives load factor < 2 */ #define WITNESS_PENDLIST (512 + (MAXCPU * 4)) @@ -158,20 +158,18 @@ * These flags go in the witness relationship matrix and describe the * relationship between any two struct witness objects. */ -#define WITNESS_UNRELATED 0x00 /* No lock order relation. */ -#define WITNESS_PARENT 0x01 /* Parent, aka direct ancestor. */ -#define WITNESS_ANCESTOR 0x02 /* Direct or indirect ancestor. */ -#define WITNESS_CHILD 0x04 /* Child, aka direct descendant. */ -#define WITNESS_DESCENDANT 0x08 /* Direct or indirect descendant. */ -#define WITNESS_ANCESTOR_MASK (WITNESS_PARENT | WITNESS_ANCESTOR) -#define WITNESS_DESCENDANT_MASK (WITNESS_CHILD | WITNESS_DESCENDANT) -#define WITNESS_RELATED_MASK \ - (WITNESS_ANCESTOR_MASK | WITNESS_DESCENDANT_MASK) -#define WITNESS_REVERSAL 0x10 /* A lock order reversal has been - * observed. */ -#define WITNESS_RESERVED1 0x20 /* Unused flag, reserved. */ -#define WITNESS_RESERVED2 0x40 /* Unused flag, reserved. */ -#define WITNESS_LOCK_ORDER_KNOWN 0x80 /* This lock order is known. */ +#define WITNESS_UNRELATED 0x00 /* No lock order relation. */ +#define WITNESS_PARENT 0x01 /* Parent, aka direct ancestor. */ +#define WITNESS_ANCESTOR 0x02 /* Direct or indirect ancestor. */ +#define WITNESS_CHILD 0x04 /* Child, aka direct descendant. */ +#define WITNESS_DESCENDANT 0x08 /* Direct or indirect descendant. */ +#define WITNESS_ANCESTOR_MASK (WITNESS_PARENT | WITNESS_ANCESTOR) +#define WITNESS_DESCENDANT_MASK (WITNESS_CHILD | WITNESS_DESCENDANT) +#define WITNESS_RELATED_MASK (WITNESS_ANCESTOR_MASK | WITNESS_DESCENDANT_MASK) +#define WITNESS_REVERSAL 0x10 /* A lock order reversal has been observed. */ +#define WITNESS_RESERVED1 0x20 /* Unused flag, reserved. */ +#define WITNESS_RESERVED2 0x40 /* Unused flag, reserved. */ +#define WITNESS_LOCK_ORDER_KNOWN 0x80 /* This lock order is known. */ /* Descendant to ancestor flags */ #define WITNESS_DTOA(x) (((x) & WITNESS_RELATED_MASK) >> 2) @@ -218,20 +216,18 @@ struct lock_list_entry { * (for example, "vnode interlock"). */ struct witness { - char w_name[MAX_W_NAME]; - uint32_t w_index; /* Index in the relationship matrix */ + char w_name[MAX_W_NAME]; + uint32_t w_index; /* Index in the relationship matrix */ struct lock_class *w_class; - STAILQ_ENTRY(witness) w_list; /* List of all witnesses. */ - STAILQ_ENTRY(witness) w_typelist; /* Witnesses of a type. */ - struct witness *w_hash_next; /* Linked list in hash buckets. */ - const char *w_file; /* File where last acquired */ - uint32_t w_line; /* Line where last acquired */ - uint32_t w_refcount; - uint16_t w_num_ancestors; /* direct/indirect - * ancestor count */ - uint16_t w_num_descendants; /* direct/indirect - * descendant count */ - int16_t w_ddb_level; + STAILQ_ENTRY(witness) w_list; /* List of all witnesses. */ + STAILQ_ENTRY(witness) w_typelist; /* Witnesses of a type. */ + struct witness *w_hash_next; /* Linked list in hash buckets. */ + const char *w_file; /* File where last acquired */ + uint32_t w_line; /* Line where last acquired */ + uint32_t w_refcount; + uint16_t w_num_ancestors; /* direct/indirect ancestor count */ + uint16_t w_num_descendants; /* direct/indirect descendant count */ + int16_t w_ddb_level; unsigned w_displayed:1; unsigned w_reversed:1; }; @@ -265,7 +261,7 @@ struct witness_lock_order_data { /* * The witness lock order data hash table. Keys are witness index tuples * (struct witness_lock_order_key), elements are lock order data objects - * (struct witness_lock_order_data). + * (struct witness_lock_order_data). */ struct witness_lock_order_hash { struct witness_lock_order_data *wloh_array[WITNESS_LO_HASH_SIZE]; @@ -295,7 +291,6 @@ struct witness_order_list_entry { static __inline int witness_lock_type_equal(struct witness *w1, struct witness *w2) { - return ((w1->w_class->lc_flags & (LC_SLEEPLOCK | LC_SPINLOCK)) == (w2->w_class->lc_flags & (LC_SLEEPLOCK | LC_SPINLOCK))); } @@ -304,7 +299,6 @@ static __inline int witness_lock_order_key_equal(const struct witness_lock_order_key *a, const struct witness_lock_order_key *b) { - return (a->from == b->from && a->to == b->to); } @@ -415,7 +409,7 @@ SYSCTL_INT(_debug_witness, OID_AUTO, skipspin, CTLFLAG_RDTUN, &witness_skipspin, int badstack_sbuf_size; int witness_count = WITNESS_COUNT; -SYSCTL_INT(_debug_witness, OID_AUTO, witness_count, CTLFLAG_RDTUN, +SYSCTL_INT(_debug_witness, OID_AUTO, witness_count, CTLFLAG_RDTUN, &witness_count, 0, ""); /* @@ -760,7 +754,6 @@ static int witness_spin_warn = 0; static const char * fixup_filename(const char *file) { - if (file == NULL) return (NULL); while (strncmp(file, "../", 3) == 0) @@ -835,7 +828,7 @@ witness_startup(void *mem) w_free_cnt--; for (i = 0; i < witness_count; i++) { - memset(w_rmatrix[i], 0, sizeof(*w_rmatrix[i]) * + memset(w_rmatrix[i], 0, sizeof(*w_rmatrix[i]) * (witness_count + 1)); } @@ -989,16 +982,16 @@ witness_ddb_display_descendants(int(*prnt)(const char *fmt, ...), { int i; - for (i = 0; i < indent; i++) - prnt(" "); + for (i = 0; i < indent; i++) + prnt(" "); prnt("%s (type: %s, depth: %d, active refs: %d)", w->w_name, w->w_class->lc_name, w->w_ddb_level, w->w_refcount); - if (w->w_displayed) { - prnt(" -- (already displayed)\n"); - return; - } - w->w_displayed = 1; + if (w->w_displayed) { + prnt(" -- (already displayed)\n"); + return; + } + w->w_displayed = 1; if (w->w_file != NULL && w->w_line != 0) prnt(" -- last acquired @ %s:%d\n", fixup_filename(w->w_file), w->w_line); @@ -1079,7 +1072,6 @@ witness_ddb_display(int(*prnt)(const char *fmt, ...)) int witness_defineorder(struct lock_object *lock1, struct lock_object *lock2) { - if (witness_watch == -1 || KERNEL_PANICKED()) return (0); @@ -1257,7 +1249,7 @@ witness_checkorder(struct lock_object *lock, int flags, const char *file, w->w_reversed = 1; mtx_unlock_spin(&w_mtx); witness_output( - "acquiring duplicate lock of same type: \"%s\"\n", + "acquiring duplicate lock of same type: \"%s\"\n", w->w_name); witness_output(" 1st %s @ %s:%d\n", plock->li_lock->lo_name, fixup_filename(plock->li_file), plock->li_line); @@ -1743,7 +1735,7 @@ found: /* * In order to reduce contention on w_mtx, we want to keep always an - * head object into lists so that frequent allocation from the + * head object into lists so that frequent allocation from the * free witness pool (and subsequent locking) is avoided. * In order to maintain the current code simple, when the head * object is totally unloaded it means also that we do not have @@ -1781,7 +1773,7 @@ witness_thread_exit(struct thread *td) n++; witness_list_lock(&lle->ll_children[i], witness_output); - + } kassert_panic( "Thread %p cannot exit while holding sleeplocks\n", td); @@ -1948,7 +1940,6 @@ found: static void depart(struct witness *w) { - MPASS(w->w_refcount == 0); if (w->w_class->lc_flags & LC_SLEEPLOCK) { w_sleep_cnt--; @@ -1999,18 +1990,18 @@ adopt(struct witness *parent, struct witness *child) child->w_num_ancestors++; } - /* - * Find each ancestor of 'pi'. Note that 'pi' itself is counted as + /* + * Find each ancestor of 'pi'. Note that 'pi' itself is counted as * an ancestor of 'pi' during this loop. */ for (i = 1; i <= w_max_used_index; i++) { - if ((w_rmatrix[i][pi] & WITNESS_ANCESTOR_MASK) == 0 && + if ((w_rmatrix[i][pi] & WITNESS_ANCESTOR_MASK) == 0 && (i != pi)) continue; /* Find each descendant of 'i' and mark it as a descendant. */ for (j = 1; j <= w_max_used_index; j++) { - /* + /* * Skip children that are already marked as * descendants of 'i'. */ @@ -2021,7 +2012,7 @@ adopt(struct witness *parent, struct witness *child) * We are only interested in descendants of 'ci'. Note * that 'ci' itself is counted as a descendant of 'ci'. */ - if ((w_rmatrix[ci][j] & WITNESS_ANCESTOR_MASK) == 0 && + if ((w_rmatrix[ci][j] & WITNESS_ANCESTOR_MASK) == 0 && (j != ci)) continue; w_rmatrix[i][j] |= WITNESS_ANCESTOR; @@ -2029,16 +2020,16 @@ adopt(struct witness *parent, struct witness *child) w_data[i].w_num_descendants++; w_data[j].w_num_ancestors++; - /* + /* * Make sure we aren't marking a node as both an - * ancestor and descendant. We should have caught + * ancestor and descendant. We should have caught * this as a lock order reversal earlier. */ if ((w_rmatrix[i][j] & WITNESS_ANCESTOR_MASK) && (w_rmatrix[i][j] & WITNESS_DESCENDANT_MASK)) { printf("witness rmatrix paradox! [%d][%d]=%d " "both ancestor and descendant\n", - i, j, w_rmatrix[i][j]); + i, j, w_rmatrix[i][j]); kdb_backtrace(); printf("Witness disabled.\n"); witness_watch = -1; @@ -2047,7 +2038,7 @@ adopt(struct witness *parent, struct witness *child) (w_rmatrix[j][i] & WITNESS_DESCENDANT_MASK)) { printf("witness rmatrix paradox! [%d][%d]=%d " "both ancestor and descendant\n", - j, i, w_rmatrix[j][i]); + j, i, w_rmatrix[j][i]); kdb_backtrace(); printf("Witness disabled.\n"); witness_watch = -1; @@ -2124,7 +2115,6 @@ _isitmyx(struct witness *w1, struct witness *w2, int rmask, const char *fname) static int isitmychild(struct witness *parent, struct witness *child) { - return (_isitmyx(parent, child, WITNESS_PARENT, __func__)); } @@ -2134,7 +2124,6 @@ isitmychild(struct witness *parent, struct witness *child) static int isitmydescendant(struct witness *ancestor, struct witness *descendant) { - return (_isitmyx(ancestor, descendant, WITNESS_ANCESTOR_MASK, __func__)); } @@ -2182,7 +2171,7 @@ witness_get(void) STAILQ_REMOVE_HEAD(&w_free, w_list); w_free_cnt--; index = w->w_index; - MPASS(index > 0 && index == w_max_used_index+1 && + MPASS(index > 0 && index == w_max_used_index + 1 && index < witness_count); bzero(w, sizeof(*w)); w->w_index = index; @@ -2194,7 +2183,6 @@ witness_get(void) static void witness_free(struct witness *w) { - STAILQ_INSERT_HEAD(&w_free, w, w_list); w_free_cnt++; } @@ -2219,11 +2207,10 @@ witness_lock_list_get(void) bzero(lle, sizeof(*lle)); return (lle); } - + static void witness_lock_list_free(struct lock_list_entry *lle) { - mtx_lock_spin(&w_mtx); lle->ll_next = w_lock_list_free; w_lock_list_free = lle; @@ -2297,7 +2284,6 @@ witness_voutput(const char *fmt, va_list ap) static int witness_thread_has_locks(struct thread *td) { - if (td->td_sleeplocks == NULL) return (0); return (td->td_sleeplocks->ll_count != 0); @@ -2573,14 +2559,12 @@ witness_setflag(struct lock_object *lock, int flag, int set) void witness_norelease(struct lock_object *lock) { - witness_setflag(lock, LI_NORELEASE, 1); } void witness_releaseok(struct lock_object *lock) { - witness_setflag(lock, LI_NORELEASE, 0); } @@ -2588,7 +2572,6 @@ witness_releaseok(struct lock_object *lock) static void witness_ddb_list(struct thread *td) { - KASSERT(witness_cold == 0, ("%s: witness_cold", __func__)); KASSERT(kdb_active, ("%s: not in the debugger", __func__)); @@ -2653,7 +2636,6 @@ DB_SHOW_ALIAS_FLAGS(alllocks, db_witness_list_all, DB_CMD_MEMSAFE); DB_SHOW_COMMAND_FLAGS(witness, db_witness_display, DB_CMD_MEMSAFE) { - witness_ddb_display(db_printf); } #endif @@ -2673,9 +2655,9 @@ sbuf_print_witness_badstacks(struct sbuf *sb, size_t *oldidx) /* Allocate and init temporary storage space. */ tmp_w1 = malloc(sizeof(struct witness), M_TEMP, M_WAITOK | M_ZERO); tmp_w2 = malloc(sizeof(struct witness), M_TEMP, M_WAITOK | M_ZERO); - tmp_data1 = malloc(sizeof(struct witness_lock_order_data), M_TEMP, + tmp_data1 = malloc(sizeof(struct witness_lock_order_data), M_TEMP, M_WAITOK | M_ZERO); - tmp_data2 = malloc(sizeof(struct witness_lock_order_data), M_TEMP, + tmp_data2 = malloc(sizeof(struct witness_lock_order_data), M_TEMP, M_WAITOK | M_ZERO); stack_zero(&tmp_data1->wlod_stack); stack_zero(&tmp_data2->wlod_stack); @@ -2750,12 +2732,12 @@ restart: sbuf_printf(sb, "\nLock order reversal between \"%s\"(%s) and \"%s\"(%s)!\n", - tmp_w1->w_name, tmp_w1->w_class->lc_name, + tmp_w1->w_name, tmp_w1->w_class->lc_name, tmp_w2->w_name, tmp_w2->w_class->lc_name); if (data1) { sbuf_printf(sb, "Lock order \"%s\"(%s) -> \"%s\"(%s) first seen at:\n", - tmp_w1->w_name, tmp_w1->w_class->lc_name, + tmp_w1->w_name, tmp_w1->w_class->lc_name, tmp_w2->w_name, tmp_w2->w_class->lc_name); stack_sbuf_print(sb, &tmp_data1->wlod_stack); sbuf_putc(sb, '\n'); @@ -2763,7 +2745,7 @@ restart: if (data2 && data2 != data1) { sbuf_printf(sb, "Lock order \"%s\"(%s) -> \"%s\"(%s) first seen at:\n", - tmp_w2->w_name, tmp_w2->w_class->lc_name, + tmp_w2->w_name, tmp_w2->w_class->lc_name, tmp_w1->w_name, tmp_w1->w_class->lc_name); stack_sbuf_print(sb, &tmp_data2->wlod_stack); sbuf_putc(sb, '\n'); @@ -2823,7 +2805,6 @@ sysctl_debug_witness_badstacks(SYSCTL_HANDLER_ARGS) static int sbuf_db_printf_drain(void *arg __unused, const char *data, int len) { - return (db_printf("%.*s", len, data)); } @@ -3068,7 +3049,7 @@ witness_lock_order_get(struct witness *parent, struct witness *child) & WITNESS_LOCK_ORDER_KNOWN) == 0) goto out; - hash = witness_hash_djb2((const char*)&key, + hash = witness_hash_djb2((const char *)&key, sizeof(key)) % w_lohash.wloh_size; data = w_lohash.wloh_array[hash]; while (data != NULL) { @@ -3089,7 +3070,6 @@ out: static int witness_lock_order_check(struct witness *parent, struct witness *child) { - if (parent != child && w_rmatrix[parent->w_index][child->w_index] & WITNESS_LOCK_ORDER_KNOWN && @@ -3115,7 +3095,7 @@ witness_lock_order_add(struct witness *parent, struct witness *child) & WITNESS_LOCK_ORDER_KNOWN) return (1); - hash = witness_hash_djb2((const char*)&key, + hash = witness_hash_djb2((const char *)&key, sizeof(key)) % w_lohash.wloh_size; w_rmatrix[parent->w_index][child->w_index] |= WITNESS_LOCK_ORDER_KNOWN; data = w_lofree; @@ -3134,7 +3114,6 @@ witness_lock_order_add(struct witness *parent, struct witness *child) static void witness_increment_graph_generation(void) { - if (witness_cold == 0) mtx_assert(&w_mtx, MA_OWNED); w_generation++; @@ -3143,7 +3122,6 @@ witness_increment_graph_generation(void) static int witness_output_drain(void *arg __unused, const char *data, int len) { - witness_output("%.*s", len, data); return (len); } diff --git a/sys/kern/vfs_init.c b/sys/kern/vfs_init.c index cd30d5cfae47..93ac001af8ad 100644 --- a/sys/kern/vfs_init.c +++ b/sys/kern/vfs_init.c @@ -103,6 +103,16 @@ struct vattr va_null; * Routines having to do with the management of the vnode table. */ +void +vfs_unref_vfsconf(struct vfsconf *vfsp) +{ + vfsconf_lock(); + KASSERT(vfsp->vfc_refcount > 0, + ("vfs %p refcount underflow %d", vfsp, vfsp->vfc_refcount)); + vfsp->vfc_refcount--; + vfsconf_unlock(); +} + static struct vfsconf * vfs_byname_locked(const char *name) { @@ -123,9 +133,11 @@ vfs_byname(const char *name) { struct vfsconf *vfsp; - vfsconf_slock(); + vfsconf_lock(); vfsp = vfs_byname_locked(name); - vfsconf_sunlock(); + if (vfsp != NULL) + vfsp->vfc_refcount++; + vfsconf_unlock(); return (vfsp); } diff --git a/sys/kern/vfs_mount.c b/sys/kern/vfs_mount.c index 8e64a7fe966b..13403acacc08 100644 --- a/sys/kern/vfs_mount.c +++ b/sys/kern/vfs_mount.c @@ -683,7 +683,6 @@ vfs_mount_alloc(struct vnode *vp, struct vfsconf *vfsp, const char *fspath, MPASSERT(mp->mnt_vfs_ops == 1, mp, ("vfs_ops should be 1 but %d found", mp->mnt_vfs_ops)); (void) vfs_busy(mp, MBF_NOWAIT); - atomic_add_acq_int(&vfsp->vfc_refcount, 1); mp->mnt_op = vfsp->vfc_vfsops; mp->mnt_vfc = vfsp; mp->mnt_stat.f_type = vfsp->vfc_typenum; @@ -731,7 +730,6 @@ vfs_mount_destroy(struct mount *mp) __FILE__, __LINE__)); MPPASS(mp->mnt_writeopcount == 0, mp); MPPASS(mp->mnt_secondary_writes == 0, mp); - atomic_subtract_rel_int(&mp->mnt_vfc->vfc_refcount, 1); if (!TAILQ_EMPTY(&mp->mnt_nvnodelist)) { struct vnode *vp; @@ -769,6 +767,9 @@ vfs_mount_destroy(struct mount *mp) vfs_free_addrlist(mp->mnt_export); free(mp->mnt_export, M_MOUNT); } + vfsconf_lock(); + mp->mnt_vfc->vfc_refcount--; + vfsconf_unlock(); crfree(mp->mnt_cred); uma_zfree(mount_zone, mp); } @@ -1133,6 +1134,7 @@ vfs_domount_first( if (jailed(td->td_ucred) && (!prison_allow(td->td_ucred, vfsp->vfc_prison_flag) || vp == td->td_ucred->cr_prison->pr_root)) { vput(vp); + vfs_unref_vfsconf(vfsp); return (EPERM); } @@ -1169,6 +1171,7 @@ vfs_domount_first( } if (error != 0) { vput(vp); + vfs_unref_vfsconf(vfsp); return (error); } vn_seqc_write_begin(vp); diff --git a/sys/modules/ichwd/Makefile b/sys/modules/ichwd/Makefile index 3c3bbc37eff5..27b4c38437ff 100644 --- a/sys/modules/ichwd/Makefile +++ b/sys/modules/ichwd/Makefile @@ -1,6 +1,6 @@ .PATH: ${SRCTOP}/sys/dev/ichwd KMOD= ichwd -SRCS= ichwd.c device_if.h bus_if.h pci_if.h isa_if.h +SRCS= i6300esbwd.c ichwd.c device_if.h bus_if.h pci_if.h isa_if.h .include <bsd.kmod.mk> diff --git a/sys/modules/irdma/Makefile b/sys/modules/irdma/Makefile index b2ffb67ca66f..a9ef6e63d3f2 100644 --- a/sys/modules/irdma/Makefile +++ b/sys/modules/irdma/Makefile @@ -1,8 +1,8 @@ .include <bsd.own.mk> -OFED_INC_DIR = ${.CURDIR}/../../ofed/include -ICE_DIR = ${.CURDIR}/../../dev/ice -.PATH: ${.CURDIR}/../../dev/irdma +OFED_INC_DIR = ${SRCTOP}/sys/ofed/include +ICE_DIR = ${SRCTOP}/sys/dev/ice +.PATH: ${SRCTOP}/sys/dev/irdma KMOD= irdma SRCS= icrdma.c diff --git a/sys/net/if.c b/sys/net/if.c index 79c883fd4a0a..0fc30488f1e5 100644 --- a/sys/net/if.c +++ b/sys/net/if.c @@ -74,7 +74,6 @@ #include <vm/uma.h> #include <net/bpf.h> -#include <net/ethernet.h> #include <net/if.h> #include <net/if_arp.h> #include <net/if_clone.h> @@ -2589,16 +2588,7 @@ ifhwioctl(u_long cmd, struct ifnet *ifp, caddr_t data, struct thread *td) * flip. They require special handling because in-kernel * consumers may indepdently toggle them. */ - if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { - if (new_flags & IFF_PPROMISC) - ifp->if_flags |= IFF_PROMISC; - else if (ifp->if_pcount == 0) - ifp->if_flags &= ~IFF_PROMISC; - if (log_promisc_mode_change) - if_printf(ifp, "permanently promiscuous mode %s\n", - ((new_flags & IFF_PPROMISC) ? - "enabled" : "disabled")); - } + if_setppromisc(ifp, new_flags & IFF_PPROMISC); if ((ifp->if_flags ^ new_flags) & IFF_PALLMULTI) { if (new_flags & IFF_PALLMULTI) ifp->if_flags |= IFF_ALLMULTI; @@ -4456,6 +4446,32 @@ if_getmtu_family(const if_t ifp, int family) return (ifp->if_mtu); } +void +if_setppromisc(if_t ifp, bool ppromisc) +{ + int new_flags; + + if (ppromisc) + new_flags = ifp->if_flags | IFF_PPROMISC; + else + new_flags = ifp->if_flags & ~IFF_PPROMISC; + if ((ifp->if_flags ^ new_flags) & IFF_PPROMISC) { + if (new_flags & IFF_PPROMISC) + new_flags |= IFF_PROMISC; + /* + * Only unset IFF_PROMISC if there are no more consumers of + * promiscuity, i.e. the ifp->if_pcount refcount is 0. + */ + else if (ifp->if_pcount == 0) + new_flags &= ~IFF_PROMISC; + if (log_promisc_mode_change) + if_printf(ifp, "permanently promiscuous mode %s\n", + ((new_flags & IFF_PPROMISC) ? + "enabled" : "disabled")); + } + ifp->if_flags = new_flags; +} + /* * Methods for drivers to access interface unicast and multicast * link level addresses. Driver shall not know 'struct ifaddr' neither diff --git a/sys/net/if_epair.c b/sys/net/if_epair.c index a213a84e17db..581c2434b8fb 100644 --- a/sys/net/if_epair.c +++ b/sys/net/if_epair.c @@ -67,7 +67,6 @@ #include <net/if_var.h> #include <net/if_clone.h> #include <net/if_media.h> -#include <net/if_var.h> #include <net/if_private.h> #include <net/if_types.h> #include <net/netisr.h> diff --git a/sys/net/if_var.h b/sys/net/if_var.h index 08435e7bd5f6..f2df612b19c1 100644 --- a/sys/net/if_var.h +++ b/sys/net/if_var.h @@ -622,6 +622,7 @@ int if_setmtu(if_t ifp, int mtu); int if_getmtu(const if_t ifp); int if_getmtu_family(const if_t ifp, int family); void if_notifymtu(if_t ifp); +void if_setppromisc(const if_t ifp, bool ppromisc); int if_setflagbits(if_t ifp, int set, int clear); int if_setflags(if_t ifp, int flags); int if_getflags(const if_t ifp); diff --git a/sys/net/pfvar.h b/sys/net/pfvar.h index d6c13470f2eb..cf6d2508cf65 100644 --- a/sys/net/pfvar.h +++ b/sys/net/pfvar.h @@ -645,6 +645,7 @@ struct pf_kpool { int tblidx; u_int16_t proxy_port[2]; u_int8_t opts; + sa_family_t ipv6_nexthop_af; }; struct pf_rule_actions { @@ -2680,7 +2681,7 @@ u_short pf_map_addr(sa_family_t, struct pf_krule *, struct pf_addr *, struct pf_kpool *); u_short pf_map_addr_sn(u_int8_t, struct pf_krule *, struct pf_addr *, struct pf_addr *, - sa_family_t *, struct pfi_kkif **nkif, + sa_family_t *, struct pfi_kkif **, struct pf_addr *, struct pf_kpool *, pf_sn_types_t); int pf_get_transaddr_af(struct pf_krule *, diff --git a/sys/netgraph/ng_nat.c b/sys/netgraph/ng_nat.c index defbe817becd..8b82d777caeb 100644 --- a/sys/netgraph/ng_nat.c +++ b/sys/netgraph/ng_nat.c @@ -818,7 +818,8 @@ ng_nat_rcvdata(hook_p hook, item_p item ) if (ip->ip_v != IPVERSION) goto send; /* other IP version, let it pass */ - if (m->m_pkthdr.len < ipofs + ntohs(ip->ip_len)) + uint16_t ip_len = ntohs(ip->ip_len); + if (m->m_pkthdr.len < (ipofs + ip_len)) goto send; /* packet too short (i.e. fragmented or broken) */ /* @@ -852,50 +853,68 @@ ng_nat_rcvdata(hook_p hook, item_p item ) if (rval == PKT_ALIAS_RESPOND) m->m_flags |= M_SKIP_FIREWALL; - m->m_pkthdr.len = m->m_len = ntohs(ip->ip_len) + ipofs; - if ((ip->ip_off & htons(IP_OFFMASK)) == 0 && - ip->ip_p == IPPROTO_TCP) { - struct tcphdr *th = (struct tcphdr *)((caddr_t)ip + - (ip->ip_hl << 2)); + /* Re-read just in case it has been updated */ + ip_len = ntohs(ip->ip_len); + int new_m_len = ip_len + ipofs; + if (new_m_len > (m->m_len + M_TRAILINGSPACE(m))) { /* - * Here is our terrible HACK. - * - * Sometimes LibAlias edits contents of TCP packet. - * In this case it needs to recompute full TCP - * checksum. However, the problem is that LibAlias - * doesn't have any idea about checksum offloading - * in kernel. To workaround this, we do not do - * checksumming in LibAlias, but only mark the - * packets with TH_RES1 in the th_x2 field. If we - * receive a marked packet, we calculate correct - * checksum for it aware of offloading. - * - * Why do I do such a terrible hack instead of - * recalculating checksum for each packet? - * Because the previous checksum was not checked! - * Recalculating checksums for EVERY packet will - * hide ALL transmission errors. Yes, marked packets - * still suffer from this problem. But, sigh, natd(8) - * has this problem, too. + * This is just a safety railguard to make sure LibAlias has not + * screwed the IP packet up somehow, should probably be KASSERT() + * at some point. Calling in_delayed_cksum() will parse IP packet + * again and reliably panic if there is less data than the IP + * header declares, there might be some other places too. */ + log(LOG_ERR, "ng_nat_rcvdata: outgoing packet corrupted, " + "not enough data: expected %d, available (%d - %d)\n", + ip_len, m->m_len + (int)M_TRAILINGSPACE(m), ipofs); + NG_FREE_ITEM(item); + return (ENXIO); + } + + m->m_pkthdr.len = m->m_len = new_m_len; - if (tcp_get_flags(th) & TH_RES1) { - uint16_t ip_len = ntohs(ip->ip_len); + if ((ip->ip_off & htons(IP_OFFMASK)) != 0 || ip->ip_p != IPPROTO_TCP) + goto send; - tcp_set_flags(th, tcp_get_flags(th) & ~TH_RES1); - th->th_sum = in_pseudo(ip->ip_src.s_addr, - ip->ip_dst.s_addr, htons(IPPROTO_TCP + - ip_len - (ip->ip_hl << 2))); + uint16_t pl_offset = ip->ip_hl << 2; + struct tcphdr *th = (struct tcphdr *)((caddr_t)ip + pl_offset); - if ((m->m_pkthdr.csum_flags & CSUM_TCP) == 0) { - m->m_pkthdr.csum_data = offsetof(struct tcphdr, - th_sum); - in_delayed_cksum(m); - } - } - } + /* + * Here is our terrible HACK. + * + * Sometimes LibAlias edits contents of TCP packet. + * In this case it needs to recompute full TCP + * checksum. However, the problem is that LibAlias + * doesn't have any idea about checksum offloading + * in kernel. To workaround this, we do not do + * checksumming in LibAlias, but only mark the + * packets with TH_RES1 in the th_x2 field. If we + * receive a marked packet, we calculate correct + * checksum for it aware of offloading. + * + * Why do I do such a terrible hack instead of + * recalculating checksum for each packet? + * Because the previous checksum was not checked! + * Recalculating checksums for EVERY packet will + * hide ALL transmission errors. Yes, marked packets + * still suffer from this problem. But, sigh, natd(8) + * has this problem, too. + */ + + if (!(tcp_get_flags(th) & TH_RES1)) + goto send; + + tcp_set_flags(th, tcp_get_flags(th) & ~TH_RES1); + th->th_sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, + htons(IPPROTO_TCP + ip_len - pl_offset)); + + if ((m->m_pkthdr.csum_flags & CSUM_TCP) != 0) + goto send; + + m->m_pkthdr.csum_data = offsetof(struct tcphdr, th_sum); + in_delayed_cksum_o(m, ipofs); send: if (hook == priv->in) diff --git a/sys/netinet/ip_output.c b/sys/netinet/ip_output.c index ec6ba8d92015..ef08b9cfd3d6 100644 --- a/sys/netinet/ip_output.c +++ b/sys/netinet/ip_output.c @@ -1044,14 +1044,14 @@ done: } void -in_delayed_cksum(struct mbuf *m) +in_delayed_cksum_o(struct mbuf *m, uint16_t iph_offset) { struct ip *ip; struct udphdr *uh; uint16_t cklen, csum, offset; - ip = mtod(m, struct ip *); - offset = ip->ip_hl << 2 ; + ip = (struct ip *)mtodo(m, iph_offset); + offset = iph_offset + (ip->ip_hl << 2); if (m->m_pkthdr.csum_flags & CSUM_UDP) { /* if udp header is not in the first mbuf copy udplen */ @@ -1078,6 +1078,13 @@ in_delayed_cksum(struct mbuf *m) *(u_short *)mtodo(m, offset) = csum; } +void +in_delayed_cksum(struct mbuf *m) +{ + + in_delayed_cksum_o(m, 0); +} + /* * IP socket option processing. */ diff --git a/sys/netinet/ip_var.h b/sys/netinet/ip_var.h index f782ebc53eb0..c113484079a3 100644 --- a/sys/netinet/ip_var.h +++ b/sys/netinet/ip_var.h @@ -271,6 +271,7 @@ VNET_DECLARE(struct pfil_head *, inet_local_pfil_head); #define PFIL_INET_LOCAL_NAME "inet-local" void in_delayed_cksum(struct mbuf *m); +void in_delayed_cksum_o(struct mbuf *m, uint16_t o); /* Hooks for ipfw, dummynet, divert etc. Most are declared in raw_ip.c */ /* diff --git a/sys/netinet/tcp_hpts.c b/sys/netinet/tcp_hpts.c index b77ebc928809..63bbe4bba11b 100644 --- a/sys/netinet/tcp_hpts.c +++ b/sys/netinet/tcp_hpts.c @@ -137,8 +137,6 @@ #include <netinet/in_kdtrace.h> #include <netinet/in_pcb.h> #include <netinet/ip.h> -#include <netinet/ip_icmp.h> /* required for icmp_var.h */ -#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */ #include <netinet/ip_var.h> #include <netinet/ip6.h> #include <netinet6/in6_pcb.h> diff --git a/sys/netinet/tcp_stacks/bbr.c b/sys/netinet/tcp_stacks/bbr.c index fed259f4d8e1..f2d7867df9b4 100644 --- a/sys/netinet/tcp_stacks/bbr.c +++ b/sys/netinet/tcp_stacks/bbr.c @@ -78,8 +78,6 @@ #include <netinet/in_kdtrace.h> #include <netinet/in_pcb.h> #include <netinet/ip.h> -#include <netinet/ip_icmp.h> /* required for icmp_var.h */ -#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */ #include <netinet/ip_var.h> #include <netinet/ip6.h> #include <netinet6/in6_pcb.h> diff --git a/sys/netinet/tcp_stacks/rack.c b/sys/netinet/tcp_stacks/rack.c index 71dd4de6baf9..11ef5ba706c5 100644 --- a/sys/netinet/tcp_stacks/rack.c +++ b/sys/netinet/tcp_stacks/rack.c @@ -77,8 +77,6 @@ #include <netinet/in_kdtrace.h> #include <netinet/in_pcb.h> #include <netinet/ip.h> -#include <netinet/ip_icmp.h> /* required for icmp_var.h */ -#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */ #include <netinet/ip_var.h> #include <netinet/ip6.h> #include <netinet6/in6_pcb.h> diff --git a/sys/netinet/tcp_stacks/rack_bbr_common.c b/sys/netinet/tcp_stacks/rack_bbr_common.c index fc12672a45f7..4a0a5fc118f6 100644 --- a/sys/netinet/tcp_stacks/rack_bbr_common.c +++ b/sys/netinet/tcp_stacks/rack_bbr_common.c @@ -76,8 +76,6 @@ #include <netinet/in_kdtrace.h> #include <netinet/in_pcb.h> #include <netinet/ip.h> -#include <netinet/ip_icmp.h> /* required for icmp_var.h */ -#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */ #include <netinet/ip_var.h> #include <netinet/ip6.h> #include <netinet6/in6_pcb.h> diff --git a/sys/netinet/tcp_stacks/rack_pcm.c b/sys/netinet/tcp_stacks/rack_pcm.c index 759bfda98357..1a51097f627c 100644 --- a/sys/netinet/tcp_stacks/rack_pcm.c +++ b/sys/netinet/tcp_stacks/rack_pcm.c @@ -78,8 +78,6 @@ #include <netinet/in_kdtrace.h> #include <netinet/in_pcb.h> #include <netinet/ip.h> -#include <netinet/ip_icmp.h> /* required for icmp_var.h */ -#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */ #include <netinet/ip_var.h> #include <netinet/ip6.h> #include <netinet6/in6_pcb.h> diff --git a/sys/netinet/tcp_stacks/tailq_hash.c b/sys/netinet/tcp_stacks/tailq_hash.c index 5ba3e7cd36c0..ff01640524b6 100644 --- a/sys/netinet/tcp_stacks/tailq_hash.c +++ b/sys/netinet/tcp_stacks/tailq_hash.c @@ -51,8 +51,6 @@ #include <netinet/in_kdtrace.h> #include <netinet/in_pcb.h> #include <netinet/ip.h> -#include <netinet/ip_icmp.h> /* required for icmp_var.h */ -#include <netinet/icmp_var.h> /* for ICMP_BANDLIM */ #include <netinet/ip_var.h> #include <netinet/ip6.h> #include <netinet6/in6_pcb.h> diff --git a/sys/netinet/tcp_syncache.c b/sys/netinet/tcp_syncache.c index 80e6b53d10df..d617d0ed4aac 100644 --- a/sys/netinet/tcp_syncache.c +++ b/sys/netinet/tcp_syncache.c @@ -1294,7 +1294,6 @@ syncache_expand(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th, } else TCPSTAT_INC(tcps_sc_completed); -/* how do we find the inp for the new socket? */ if (sc != &scs) syncache_free(sc); return (1); diff --git a/sys/netlink/route/iface_drivers.c b/sys/netlink/route/iface_drivers.c index 21db3017df18..4f1540740ead 100644 --- a/sys/netlink/route/iface_drivers.c +++ b/sys/netlink/route/iface_drivers.c @@ -82,28 +82,39 @@ _nl_modify_ifp_generic(struct ifnet *ifp, struct nl_parsed_link *lattrs, } } - if ((lattrs->ifi_change & IFF_UP) && (lattrs->ifi_flags & IFF_UP) == 0) { - /* Request to down the interface */ - if_down(ifp); + if ((lattrs->ifi_change & IFF_UP) != 0 || lattrs->ifi_change == 0) { + /* Request to up or down the interface */ + if (lattrs->ifi_flags & IFF_UP) + if_up(ifp); + else + if_down(ifp); } if (lattrs->ifla_mtu > 0) { if (nlp_has_priv(npt->nlp, PRIV_NET_SETIFMTU)) { struct ifreq ifr = { .ifr_mtu = lattrs->ifla_mtu }; - error = ifhwioctl(SIOCSIFMTU, ifp, (char *)&ifr, curthread); + error = ifhwioctl(SIOCSIFMTU, ifp, (char *)&ifr, + curthread); + if (error != 0) { + nlmsg_report_err_msg(npt, "Failed to set mtu"); + return (error); + } } else { nlmsg_report_err_msg(npt, "Not enough privileges to set mtu"); return (EPERM); } } - if (lattrs->ifi_change & IFF_PROMISC) { - error = ifpromisc(ifp, lattrs->ifi_flags & IFF_PROMISC); - if (error != 0) { - nlmsg_report_err_msg(npt, "unable to set promisc"); - return (error); - } - } + if ((lattrs->ifi_change & IFF_PROMISC) != 0 || + lattrs->ifi_change == 0) + /* + * When asking for IFF_PROMISC, set permanent flag instead + * (IFF_PPROMISC) as we have no way of doing promiscuity + * reference counting through ifpromisc(). Every call to this + * function either sets or unsets IFF_PROMISC, and ifi_change + * is usually set to 0xFFFFFFFF. + */ + if_setppromisc(ifp, (lattrs->ifi_flags & IFF_PROMISC) != 0); if (lattrs->ifla_address != NULL) { if (nlp_has_priv(npt->nlp, PRIV_NET_SETIFMAC)) { diff --git a/sys/netpfil/pf/if_pfsync.c b/sys/netpfil/pf/if_pfsync.c index 585c196391c0..cfc300d99396 100644 --- a/sys/netpfil/pf/if_pfsync.c +++ b/sys/netpfil/pf/if_pfsync.c @@ -605,7 +605,8 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) rt_kif = rpool_first->kif; /* * Guess the AF of the route address, FreeBSD 13 does - * not support af-to so it should be safe. + * not support af-to nor prefer-ipv6-nexthop + * so it should be safe. */ rt_af = r->af; } else if (!PF_AZERO(&sp->pfs_1301.rt_addr, sp->pfs_1301.af)) { @@ -634,8 +635,9 @@ pfsync_state_import(union pfsync_state_union *sp, int flags, int msg_version) } rt = sp->pfs_1400.rt; /* - * Guess the AF of the route address, FreeBSD 13 does - * not support af-to so it should be safe. + * Guess the AF of the route address, FreeBSD 14 does + * not support af-to nor prefer-ipv6-nexthop + * so it should be safe. */ rt_af = sp->pfs_1400.af; } diff --git a/sys/netpfil/pf/pf.c b/sys/netpfil/pf/pf.c index 8cd4fff95b15..4325835c7671 100644 --- a/sys/netpfil/pf/pf.c +++ b/sys/netpfil/pf/pf.c @@ -5960,7 +5960,9 @@ pf_test_rule(struct pf_krule **rm, struct pf_kstate **sm, if (r->rt) { /* * Set act.rt here instead of in pf_rule_to_actions() because - * it is applied only from the last pass rule. + * it is applied only from the last pass rule. For rules + * with the prefer-ipv6-nexthop option act.rt_af is a hint + * about AF of the forwarded packet and might be changed. */ pd->act.rt = r->rt; if (r->rt == PF_REPLYTO) @@ -8974,9 +8976,10 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, struct pf_kstate *s, struct pf_pdesc *pd, struct inpcb *inp) { struct mbuf *m0, *m1, *md; - struct route ro; - const struct sockaddr *gw = &ro.ro_dst; - struct sockaddr_in *dst; + struct route_in6 ro; + union sockaddr_union rt_gw; + const union sockaddr_union *gw = (const union sockaddr_union *)&ro.ro_dst; + union sockaddr_union *dst; struct ip *ip; struct ifnet *ifp = NULL; int error = 0; @@ -9071,10 +9074,35 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, ip = mtod(m0, struct ip *); bzero(&ro, sizeof(ro)); - dst = (struct sockaddr_in *)&ro.ro_dst; - dst->sin_family = AF_INET; - dst->sin_len = sizeof(struct sockaddr_in); - dst->sin_addr.s_addr = pd->act.rt_addr.v4.s_addr; + dst = (union sockaddr_union *)&ro.ro_dst; + dst->sin.sin_family = AF_INET; + dst->sin.sin_len = sizeof(struct sockaddr_in); + dst->sin.sin_addr = ip->ip_dst; + if (ifp) { /* Only needed in forward direction and route-to */ + bzero(&rt_gw, sizeof(rt_gw)); + ro.ro_flags |= RT_HAS_GW; + gw = &rt_gw; + switch (pd->act.rt_af) { +#ifdef INET + case AF_INET: + rt_gw.sin.sin_family = AF_INET; + rt_gw.sin.sin_len = sizeof(struct sockaddr_in); + rt_gw.sin.sin_addr.s_addr = pd->act.rt_addr.v4.s_addr; + break; +#endif /* INET */ +#ifdef INET6 + case AF_INET6: + rt_gw.sin6.sin6_family = AF_INET6; + rt_gw.sin6.sin6_len = sizeof(struct sockaddr_in6); + pf_addrcpy((struct pf_addr *)&rt_gw.sin6.sin6_addr, + &pd->act.rt_addr, AF_INET6); + break; +#endif /* INET6 */ + default: + /* Normal af-to without route-to */ + break; + } + } if (pd->dir == PF_IN) { if (ip->ip_ttl <= IPTTLDEC) { @@ -9098,10 +9126,10 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, /* Use the gateway if needed. */ if (nh->nh_flags & NHF_GATEWAY) { - gw = &nh->gw_sa; + gw = (const union sockaddr_union *)&nh->gw_sa; ro.ro_flags |= RT_HAS_GW; } else { - dst->sin_addr = ip->ip_dst; + dst->sin.sin_addr = ip->ip_dst; } /* @@ -9126,6 +9154,9 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, PF_STATE_UNLOCK(s); } + /* It must have been either set from rt_af or from fib4_lookup */ + KASSERT(gw->sin.sin_family != 0, ("%s: gw address family undetermined", __func__)); + if (ifp == NULL) { m0 = pd->m; pd->m = NULL; @@ -9210,9 +9241,11 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, m_clrprotoflags(m0); /* Avoid confusing lower layers. */ md = m0; - error = pf_dummynet_route(pd, s, r, ifp, gw, &md); + error = pf_dummynet_route(pd, s, r, ifp, + (const struct sockaddr *)gw, &md); if (md != NULL) { - error = (*ifp->if_output)(ifp, md, gw, &ro); + error = (*ifp->if_output)(ifp, md, + (const struct sockaddr *)gw, (struct route *)&ro); SDT_PROBE2(pf, ip, route_to, output, ifp, error); } goto done; @@ -9253,9 +9286,11 @@ pf_route(struct pf_krule *r, struct ifnet *oifp, md = m0; pd->pf_mtag = pf_find_mtag(md); error = pf_dummynet_route(pd, s, r, ifp, - gw, &md); + (const struct sockaddr *)gw, &md); if (md != NULL) { - error = (*ifp->if_output)(ifp, md, gw, &ro); + error = (*ifp->if_output)(ifp, md, + (const struct sockaddr *)gw, + (struct route *)&ro); SDT_PROBE2(pf, ip, route_to, output, ifp, error); } } else diff --git a/sys/netpfil/pf/pf.h b/sys/netpfil/pf/pf.h index 51b3fd6390e1..8edd5a5110a1 100644 --- a/sys/netpfil/pf/pf.h +++ b/sys/netpfil/pf/pf.h @@ -131,6 +131,7 @@ enum { PF_ADDR_ADDRMASK, PF_ADDR_NOROUTE, PF_ADDR_DYNIFTL, #define PF_POOL_TYPEMASK 0x0f #define PF_POOL_STICKYADDR 0x20 #define PF_POOL_ENDPI 0x40 +#define PF_POOL_IPV6NH 0x80 #define PF_WSCALE_FLAG 0x80 #define PF_WSCALE_MASK 0x0f diff --git a/sys/netpfil/pf/pf_ioctl.c b/sys/netpfil/pf/pf_ioctl.c index e5da05a958f6..d395730d6a54 100644 --- a/sys/netpfil/pf/pf_ioctl.c +++ b/sys/netpfil/pf/pf_ioctl.c @@ -2276,6 +2276,7 @@ pf_ioctl_addrule(struct pf_krule *rule, uint32_t ticket, rule->nat.cur = TAILQ_FIRST(&rule->nat.list); rule->rdr.cur = TAILQ_FIRST(&rule->rdr.list); rule->route.cur = TAILQ_FIRST(&rule->route.list); + rule->route.ipv6_nexthop_af = AF_INET6; TAILQ_INSERT_TAIL(ruleset->rules[rs_num].inactive.ptr, rule, entries); ruleset->rules[rs_num].inactive.rcount++; diff --git a/sys/netpfil/pf/pf_lb.c b/sys/netpfil/pf/pf_lb.c index bc9e1dc72902..b8b5157c9b15 100644 --- a/sys/netpfil/pf/pf_lb.c +++ b/sys/netpfil/pf/pf_lb.c @@ -545,11 +545,18 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, uint64_t hashidx; int cnt; sa_family_t wanted_af; + u_int8_t pool_type; + bool prefer_ipv6_nexthop = rpool->opts & PF_POOL_IPV6NH; KASSERT(saf != 0, ("%s: saf == 0", __func__)); KASSERT(naf != NULL, ("%s: naf = NULL", __func__)); KASSERT((*naf) != 0, ("%s: *naf = 0", __func__)); + /* + * Given (*naf) is a hint about AF of the forwarded packet. + * It might be changed if prefer_ipv6_nexthop is enabled and + * the combination of nexthop AF and packet AF allows for it. + */ wanted_af = (*naf); mtx_lock(&rpool->mtx); @@ -594,19 +601,38 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, } else { raddr = &rpool->cur->addr.v.a.addr; rmask = &rpool->cur->addr.v.a.mask; - /* - * For single addresses check their address family. Unless they - * have none, which happens when addresses are added with - * the old ioctl mechanism. In such case trust that the address - * has the proper AF. - */ - if (rpool->cur->af && rpool->cur->af != wanted_af) { - reason = PFRES_MAPFAILED; - goto done_pool_mtx; + } + + /* + * For pools with a single host with the prefer-ipv6-nexthop option + * we can return pool address of any AF, unless the forwarded packet + * is IPv6, then we can return only if pool address is IPv6. + * For non-prefer-ipv6-nexthop we can return pool address only + * of wanted AF, unless the pool address'es AF is unknown, which + * happens in case old ioctls have been used to set up the pool. + * + * Round-robin pools have their own logic for retrying next addresses. + */ + pool_type = rpool->opts & PF_POOL_TYPEMASK; + if (pool_type == PF_POOL_NONE || pool_type == PF_POOL_BITMASK || + ((pool_type == PF_POOL_RANDOM || pool_type == PF_POOL_SRCHASH) && + rpool->cur->addr.type != PF_ADDR_TABLE && + rpool->cur->addr.type != PF_ADDR_DYNIFTL)) { + if (prefer_ipv6_nexthop) { + if (rpool->cur->af == AF_INET && (*naf) == AF_INET6) { + reason = PFRES_MAPFAILED; + goto done_pool_mtx; + } + wanted_af = rpool->cur->af; + } else { + if (rpool->cur->af != 0 && rpool->cur->af != (*naf)) { + reason = PFRES_MAPFAILED; + goto done_pool_mtx; + } } } - switch (rpool->opts & PF_POOL_TYPEMASK) { + switch (pool_type) { case PF_POOL_NONE: pf_addrcpy(naddr, raddr, wanted_af); break; @@ -631,10 +657,22 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, else rpool->tblidx = (int)arc4random_uniform(cnt); memset(&rpool->counter, 0, sizeof(rpool->counter)); + if (prefer_ipv6_nexthop) + wanted_af = AF_INET6; + retry_other_af_random: if (pfr_pool_get(kt, &rpool->tblidx, &rpool->counter, wanted_af, pf_islinklocal, false)) { - reason = PFRES_MAPFAILED; - goto done_pool_mtx; /* unsupported */ + /* Retry with IPv4 nexthop for IPv4 traffic */ + if (prefer_ipv6_nexthop && + wanted_af == AF_INET6 && + (*naf) == AF_INET) { + wanted_af = AF_INET; + goto retry_other_af_random; + } else { + /* no hosts in wanted AF */ + reason = PFRES_MAPFAILED; + goto done_pool_mtx; + } } pf_addrcpy(naddr, &rpool->counter, wanted_af); } else if (init_addr != NULL && PF_AZERO(init_addr, @@ -702,10 +740,22 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, else rpool->tblidx = (int)(hashidx % cnt); memset(&rpool->counter, 0, sizeof(rpool->counter)); + if (prefer_ipv6_nexthop) + wanted_af = AF_INET6; + retry_other_af_srchash: if (pfr_pool_get(kt, &rpool->tblidx, &rpool->counter, wanted_af, pf_islinklocal, false)) { - reason = PFRES_MAPFAILED; - goto done_pool_mtx; /* unsupported */ + /* Retry with IPv4 nexthop for IPv4 traffic */ + if (prefer_ipv6_nexthop && + wanted_af == AF_INET6 && + (*naf) == AF_INET) { + wanted_af = AF_INET; + goto retry_other_af_srchash; + } else { + /* no hosts in wanted AF */ + reason = PFRES_MAPFAILED; + goto done_pool_mtx; + } } pf_addrcpy(naddr, &rpool->counter, wanted_af); } else { @@ -718,6 +768,9 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, { struct pf_kpooladdr *acur = rpool->cur; + retry_other_af_rr: + if (prefer_ipv6_nexthop) + wanted_af = rpool->ipv6_nexthop_af; if (rpool->cur->addr.type == PF_ADDR_TABLE) { if (!pfr_pool_get(rpool->cur->addr.p.tbl, &rpool->tblidx, &rpool->counter, wanted_af, @@ -728,46 +781,55 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, &rpool->tblidx, &rpool->counter, wanted_af, pf_islinklocal, true)) goto get_addr; - } else if (pf_match_addr(0, raddr, rmask, &rpool->counter, - wanted_af)) + } else if (rpool->cur->af == wanted_af && + pf_match_addr(0, raddr, rmask, &rpool->counter, wanted_af)) goto get_addr; - + if (prefer_ipv6_nexthop && + (*naf) == AF_INET && wanted_af == AF_INET6) { + /* Reset table index when changing wanted AF. */ + rpool->tblidx = -1; + rpool->ipv6_nexthop_af = AF_INET; + goto retry_other_af_rr; + } try_next: + /* Reset prefer-ipv6-nexthop search to IPv6 when iterating pools. */ + rpool->ipv6_nexthop_af = AF_INET6; if (TAILQ_NEXT(rpool->cur, entries) == NULL) rpool->cur = TAILQ_FIRST(&rpool->list); else rpool->cur = TAILQ_NEXT(rpool->cur, entries); + try_next_ipv6_nexthop_rr: + /* Reset table index when iterating pools or changing wanted AF. */ rpool->tblidx = -1; + if (prefer_ipv6_nexthop) + wanted_af = rpool->ipv6_nexthop_af; if (rpool->cur->addr.type == PF_ADDR_TABLE) { - if (pfr_pool_get(rpool->cur->addr.p.tbl, + if (!pfr_pool_get(rpool->cur->addr.p.tbl, &rpool->tblidx, &rpool->counter, wanted_af, NULL, - true)) { - /* table contains no address of type 'wanted_af' */ - if (rpool->cur != acur) - goto try_next; - reason = PFRES_MAPFAILED; - goto done_pool_mtx; - } + true)) + goto get_addr; } else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) { - if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, - &rpool->tblidx, &rpool->counter, wanted_af, - pf_islinklocal, true)) { - /* interface has no address of type 'wanted_af' */ - if (rpool->cur != acur) - goto try_next; - reason = PFRES_MAPFAILED; - goto done_pool_mtx; - } + if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt, + &rpool->tblidx, &rpool->counter, wanted_af, pf_islinklocal, + true)) + goto get_addr; } else { - raddr = &rpool->cur->addr.v.a.addr; - rmask = &rpool->cur->addr.v.a.mask; - if (rpool->cur->af && rpool->cur->af != wanted_af) { - reason = PFRES_MAPFAILED; - goto done_pool_mtx; + if (rpool->cur->af == wanted_af) { + raddr = &rpool->cur->addr.v.a.addr; + rmask = &rpool->cur->addr.v.a.mask; + pf_addrcpy(&rpool->counter, raddr, wanted_af); + goto get_addr; } - pf_addrcpy(&rpool->counter, raddr, wanted_af); } - + if (prefer_ipv6_nexthop && + (*naf) == AF_INET && wanted_af == AF_INET6) { + rpool->ipv6_nexthop_af = AF_INET; + goto try_next_ipv6_nexthop_rr; + } + if (rpool->cur != acur) + goto try_next; + reason = PFRES_MAPFAILED; + goto done_pool_mtx; get_addr: pf_addrcpy(naddr, &rpool->counter, wanted_af); if (init_addr != NULL && PF_AZERO(init_addr, wanted_af)) @@ -777,9 +839,16 @@ pf_map_addr(sa_family_t saf, struct pf_krule *r, struct pf_addr *saddr, } } + if (wanted_af == 0) { + reason = PFRES_MAPFAILED; + goto done_pool_mtx; + } + if (nkif) *nkif = rpool->cur->kif; + (*naf) = wanted_af; + done_pool_mtx: mtx_unlock(&rpool->mtx); diff --git a/sys/sys/mount.h b/sys/sys/mount.h index f6480b173a5c..18f85192f6c3 100644 --- a/sys/sys/mount.h +++ b/sys/sys/mount.h @@ -1007,6 +1007,7 @@ struct mntarg *mount_argsu(struct mntarg *ma, const char *name, const void *val, void statfs_scale_blocks(struct statfs *sf, long max_size); struct vfsconf *vfs_byname(const char *); struct vfsconf *vfs_byname_kld(const char *, struct thread *td, int *); +void vfs_unref_vfsconf(struct vfsconf *vfsp); void vfs_mount_destroy(struct mount *); void vfs_event_signal(fsid_t *, u_int32_t, intptr_t); void vfs_freeopts(struct vfsoptlist *opts); diff --git a/sys/vm/vm_extern.h b/sys/vm/vm_extern.h index 93ec6014c27d..2b15a01206a2 100644 --- a/sys/vm/vm_extern.h +++ b/sys/vm/vm_extern.h @@ -91,6 +91,8 @@ void vm_fault_copy_entry(vm_map_t, vm_map_t, vm_map_entry_t, vm_map_entry_t, vm_ooffset_t *); int vm_fault_disable_pagefaults(void); void vm_fault_enable_pagefaults(int save); +int vm_fault_hold_pages_e(vm_map_t map, vm_offset_t addr, vm_size_t len, + vm_prot_t prot, vm_page_t *ma, int max_count, int *ppages_count); int vm_fault_quick_hold_pages(vm_map_t map, vm_offset_t addr, vm_size_t len, vm_prot_t prot, vm_page_t *ma, int max_count); int vm_fault_trap(vm_map_t map, vm_offset_t vaddr, vm_prot_t fault_type, diff --git a/sys/vm/vm_fault.c b/sys/vm/vm_fault.c index 3e57e8d4f1d0..2a27f3e674a4 100644 --- a/sys/vm/vm_fault.c +++ b/sys/vm/vm_fault.c @@ -1995,32 +1995,43 @@ vm_fault_prefault(const struct faultstate *fs, vm_offset_t addra, } /* - * Hold each of the physical pages that are mapped by the specified range of - * virtual addresses, ["addr", "addr" + "len"), if those mappings are valid - * and allow the specified types of access, "prot". If all of the implied - * pages are successfully held, then the number of held pages is returned - * together with pointers to those pages in the array "ma". However, if any - * of the pages cannot be held, -1 is returned. + * Hold each of the physical pages that are mapped by the specified + * range of virtual addresses, ["addr", "addr" + "len"), if those + * mappings are valid and allow the specified types of access, "prot". + * If all of the implied pages are successfully held, then the number + * of held pages is assigned to *ppages_count, together with pointers + * to those pages in the array "ma". The returned value is zero. + * + * However, if any of the pages cannot be held, an error is returned, + * and no pages are held. + * Error values: + * ENOMEM - the range is not valid + * EINVAL - the provided vm_page array is too small to hold all pages + * EAGAIN - a page was not mapped, and the thread is in nofaulting mode + * EFAULT - a page with requested permissions cannot be mapped + * (more detailed result from vm_fault() is lost) */ int -vm_fault_quick_hold_pages(vm_map_t map, vm_offset_t addr, vm_size_t len, - vm_prot_t prot, vm_page_t *ma, int max_count) +vm_fault_hold_pages_e(vm_map_t map, vm_offset_t addr, vm_size_t len, + vm_prot_t prot, vm_page_t *ma, int max_count, int *ppages_count) { vm_offset_t end, va; vm_page_t *mp; - int count; + int count, error; boolean_t pmap_failed; - if (len == 0) + if (len == 0) { + *ppages_count = 0; return (0); + } end = round_page(addr + len); addr = trunc_page(addr); if (!vm_map_range_valid(map, addr, end)) - return (-1); + return (ENOMEM); if (atop(end - addr) > max_count) - panic("vm_fault_quick_hold_pages: count > max_count"); + return (EINVAL); count = atop(end - addr); /* @@ -2062,19 +2073,49 @@ vm_fault_quick_hold_pages(vm_map_t map, vm_offset_t addr, vm_size_t len, * the proper behaviour explicitly. */ if ((prot & VM_PROT_QUICK_NOFAULT) != 0 && - (curthread->td_pflags & TDP_NOFAULTING) != 0) - goto error; - for (mp = ma, va = addr; va < end; mp++, va += PAGE_SIZE) + (curthread->td_pflags & TDP_NOFAULTING) != 0) { + error = EAGAIN; + goto fail; + } + for (mp = ma, va = addr; va < end; mp++, va += PAGE_SIZE) { if (*mp == NULL && vm_fault(map, va, prot, - VM_FAULT_NORMAL, mp) != KERN_SUCCESS) - goto error; + VM_FAULT_NORMAL, mp) != KERN_SUCCESS) { + error = EFAULT; + goto fail; + } + } } - return (count); -error: + *ppages_count = count; + return (0); +fail: for (mp = ma; mp < ma + count; mp++) if (*mp != NULL) vm_page_unwire(*mp, PQ_INACTIVE); - return (-1); + return (error); +} + + /* + * Hold each of the physical pages that are mapped by the specified range of + * virtual addresses, ["addr", "addr" + "len"), if those mappings are valid + * and allow the specified types of access, "prot". If all of the implied + * pages are successfully held, then the number of held pages is returned + * together with pointers to those pages in the array "ma". However, if any + * of the pages cannot be held, -1 is returned. + */ +int +vm_fault_quick_hold_pages(vm_map_t map, vm_offset_t addr, vm_size_t len, + vm_prot_t prot, vm_page_t *ma, int max_count) +{ + int error, pages_count; + + error = vm_fault_hold_pages_e(map, addr, len, prot, ma, + max_count, &pages_count); + if (error != 0) { + if (error == EINVAL) + panic("vm_fault_quick_hold_pages: count > max_count"); + return (-1); + } + return (pages_count); } /* |