/*- * SPDX-License-Identifier: BSD-2-Clause * * Copyright (c) 2013 Tycho Nightingale <tycho.nightingale@pluribusnetworks.com> * Copyright (c) 2013 Neel Natu <neel@freebsd.org> * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include <sys/cdefs.h> #include "opt_bhyve_snapshot.h" #include <sys/param.h> #include <sys/lock.h> #include <sys/mutex.h> #include <sys/kernel.h> #include <sys/malloc.h> #include <sys/systm.h> #include <dev/acpica/acpi_hpet.h> #include <machine/vmm.h> #include <machine/vmm_dev.h> #include <machine/vmm_snapshot.h> #include "vmm_lapic.h" #include "vatpic.h" #include "vioapic.h" #include "vhpet.h" #include "vmm_ktr.h" static MALLOC_DEFINE(M_VHPET, "vhpet", "bhyve virtual hpet"); #define HPET_FREQ 16777216 /* 16.7 (2^24) Mhz */ #define FS_PER_S 1000000000000000ul /* Timer N Configuration and Capabilities Register */ #define HPET_TCAP_RO_MASK (HPET_TCAP_INT_ROUTE | \ HPET_TCAP_FSB_INT_DEL | \ HPET_TCAP_SIZE | \ HPET_TCAP_PER_INT) /* * HPET requires at least 3 timers and up to 32 timers per block. */ #define VHPET_NUM_TIMERS 8 CTASSERT(VHPET_NUM_TIMERS >= 3 && VHPET_NUM_TIMERS <= 32); struct vhpet_callout_arg { struct vhpet *vhpet; int timer_num; }; struct vhpet { struct vm *vm; struct mtx mtx; sbintime_t freq_sbt; uint64_t config; /* Configuration */ uint64_t isr; /* Interrupt Status */ uint32_t countbase; /* HPET counter base value */ sbintime_t countbase_sbt; /* uptime corresponding to base value */ struct { uint64_t cap_config; /* Configuration */ uint64_t msireg; /* FSB interrupt routing */ uint32_t compval; /* Comparator */ uint32_t comprate; struct callout callout; sbintime_t callout_sbt; /* time when counter==compval */ struct vhpet_callout_arg arg; } timer[VHPET_NUM_TIMERS]; }; #define VHPET_LOCK(vhp) mtx_lock(&((vhp)->mtx)) #define VHPET_UNLOCK(vhp) mtx_unlock(&((vhp)->mtx)) static void vhpet_start_timer(struct vhpet *vhpet, int n, uint32_t counter, sbintime_t now); static uint64_t vhpet_capabilities(void) { uint64_t cap = 0; cap |= 0x8086 << 16; /* vendor id */ cap |= (VHPET_NUM_TIMERS - 1) << 8; /* number of timers */ cap |= 1; /* revision */ cap &= ~HPET_CAP_COUNT_SIZE; /* 32-bit timer */ cap &= 0xffffffff; cap |= (FS_PER_S / HPET_FREQ) << 32; /* tick period in fs */ return (cap); } static __inline bool vhpet_counter_enabled(struct vhpet *vhpet) { return ((vhpet->config & HPET_CNF_ENABLE) ? true : false); } static __inline bool vhpet_timer_msi_enabled(struct vhpet *vhpet, int n) { const uint64_t msi_enable = HPET_TCAP_FSB_INT_DEL | HPET_TCNF_FSB_EN; if ((vhpet->timer[n].cap_config & msi_enable) == msi_enable) return (true); else return (false); } static __inline int vhpet_timer_ioapic_pin(struct vhpet *vhpet, int n) { /* * If the timer is configured to use MSI then treat it as if the * timer is not connected to the ioapic. */ if (vhpet_timer_msi_enabled(vhpet, n)) return (0); return ((vhpet->timer[n].cap_config & HPET_TCNF_INT_ROUTE) >> 9); } static uint32_t vhpet_counter(struct vhpet *vhpet, sbintime_t *nowptr) { uint32_t val; sbintime_t now, delta; val = vhpet->countbase; if (vhpet_counter_enabled(vhpet)) { now = sbinuptime(); delta = now - vhpet->countbase_sbt; KASSERT(delta >= 0, ("vhpet_counter: uptime went backwards: " "%#lx to %#lx", vhpet->countbase_sbt, now)); val += delta / vhpet->freq_sbt; if (nowptr != NULL) *nowptr = now; } else { /* * The sbinuptime corresponding to the 'countbase' is * meaningless when the counter is disabled. Make sure * that the caller doesn't want to use it. */ KASSERT(nowptr == NULL, ("vhpet_counter: nowptr must be NULL")); } return (val); } static void vhpet_timer_clear_isr(struct vhpet *vhpet, int n) { int pin; if (vhpet->isr & (1 << n)) { pin = vhpet_timer_ioapic_pin(vhpet, n); KASSERT(pin != 0, ("vhpet timer %d irq incorrectly routed", n)); vioapic_deassert_irq(vhpet->vm, pin); vhpet->isr &= ~(1 << n); } } static __inline bool vhpet_periodic_timer(struct vhpet *vhpet, int n) { return ((vhpet->timer[n].cap_config & HPET_TCNF_TYPE) != 0); } static __inline bool vhpet_timer_interrupt_enabled(struct vhpet *vhpet, int n) { return ((vhpet->timer[n].cap_config & HPET_TCNF_INT_ENB) != 0); } static __inline bool vhpet_timer_edge_trig(struct vhpet *vhpet, int n) { KASSERT(!vhpet_timer_msi_enabled(vhpet, n), ("vhpet_timer_edge_trig: " "timer %d is using MSI", n)); if ((vhpet->timer[n].cap_config & HPET_TCNF_INT_TYPE) == 0) return (true); else return (false); } static void vhpet_timer_interrupt(struct vhpet *vhpet, int n) { int pin; /* If interrupts are not enabled for this timer then just return. */ if (!vhpet_timer_interrupt_enabled(vhpet, n)) return; /* * If a level triggered interrupt is already asserted then just return. */ if ((vhpet->isr & (1 << n)) != 0) { VM_CTR1(vhpet->vm, "hpet t%d intr is already asserted", n); return; } if (vhpet_timer_msi_enabled(vhpet, n)) { lapic_intr_msi(vhpet->vm, vhpet->timer[n].msireg >> 32, vhpet->timer[n].msireg & 0xffffffff); return; } pin = vhpet_timer_ioapic_pin(vhpet, n); if (pin == 0) { VM_CTR1(vhpet->vm, "hpet t%d intr is not routed to ioapic", n); return; } if (vhpet_timer_edge_trig(vhpet, n)) { vioapic_pulse_irq(vhpet->vm, pin); } else { vhpet->isr |= 1 << n; vioapic_assert_irq(vhpet->vm, pin); } } static void vhpet_adjust_compval(struct vhpet *vhpet, int n, uint32_t counter) { uint32_t compval, comprate, compnext; KASSERT(vhpet->timer[n].comprate != 0, ("hpet t%d is not periodic", n)); compval = vhpet->timer[n].compval; comprate = vhpet->timer[n].comprate; /* * Calculate the comparator value to be used for the next periodic * interrupt. * * This function is commonly called from the callout handler. * In this scenario the 'counter' is ahead of 'compval'. To find * the next value to program into the accumulator we divide the * number space between 'compval' and 'counter' into 'comprate' * sized units. The 'compval' is rounded up such that is "ahead" * of 'counter'. */ compnext = compval + ((counter - compval) / comprate + 1) * comprate; vhpet->timer[n].compval = compnext; } static void vhpet_handler(void *a) { int n; uint32_t counter; sbintime_t now; struct vhpet *vhpet; struct callout *callout; struct vhpet_callout_arg *arg; arg = a; vhpet = arg->vhpet; n = arg->timer_num; callout = &vhpet->timer[n].callout; VM_CTR1(vhpet->vm, "hpet t%d fired", n); VHPET_LOCK(vhpet); if (callout_pending(callout)) /* callout was reset */ goto done; if (!callout_active(callout)) /* callout was stopped */ goto done; callout_deactivate(callout); if (!vhpet_counter_enabled(vhpet)) panic("vhpet(%p) callout with counter disabled", vhpet); counter = vhpet_counter(vhpet, &now); vhpet_start_timer(vhpet, n, counter, now); vhpet_timer_interrupt(vhpet, n); done: VHPET_UNLOCK(vhpet); return; } static void vhpet_stop_timer(struct vhpet *vhpet, int n, sbintime_t now) { VM_CTR1(vhpet->vm, "hpet t%d stopped", n); callout_stop(&vhpet->timer[n].callout); /* * If the callout was scheduled to expire in the past but hasn't * had a chance to execute yet then trigger the timer interrupt * here. Failing to do so will result in a missed timer interrupt * in the guest. This is especially bad in one-shot mode because * the next interrupt has to wait for the counter to wrap around. */ if (vhpet->timer[n].callout_sbt < now) { VM_CTR1(vhpet->vm, "hpet t%d interrupt triggered after " "stopping timer", n); vhpet_timer_interrupt(vhpet, n); } } static void vhpet_start_timer(struct vhpet *vhpet, int n, uint32_t counter, sbintime_t now) { sbintime_t delta, precision; if (vhpet->timer[n].comprate != 0) vhpet_adjust_compval(vhpet, n, counter); else { /* * In one-shot mode it is the guest's responsibility to make * sure that the comparator value is not in the "past". The * hardware doesn't have any belt-and-suspenders to deal with * this so we don't either. */ } delta = (vhpet->timer[n].compval - counter) * vhpet->freq_sbt; precision = delta >> tc_precexp; vhpet->timer[n].callout_sbt = now + delta; callout_reset_sbt(&vhpet->timer[n].callout, vhpet->timer[n].callout_sbt, precision, vhpet_handler, &vhpet->timer[n].arg, C_ABSOLUTE); } static void vhpet_start_counting(struct vhpet *vhpet) { int i; vhpet->countbase_sbt = sbinuptime(); for (i = 0; i < VHPET_NUM_TIMERS; i++) { /* * Restart the timers based on the value of the main counter * when it stopped counting. */ vhpet_start_timer(vhpet, i, vhpet->countbase, vhpet->countbase_sbt); } } static void vhpet_stop_counting(struct vhpet *vhpet, uint32_t counter, sbintime_t now) { int i; vhpet->countbase = counter; for (i = 0; i < VHPET_NUM_TIMERS; i++) vhpet_stop_timer(vhpet, i, now); } static __inline void update_register(uint64_t *regptr, uint64_t data, uint64_t mask) { *regptr &= ~mask; *regptr |= (data & mask); } static void vhpet_timer_update_config(struct vhpet *vhpet, int n, uint64_t data, uint64_t mask) { bool clear_isr; int old_pin, new_pin; uint32_t allowed_irqs; uint64_t oldval, newval; if (vhpet_timer_msi_enabled(vhpet, n) || vhpet_timer_edge_trig(vhpet, n)) { if (vhpet->isr & (1 << n)) panic("vhpet timer %d isr should not be asserted", n); } old_pin = vhpet_timer_ioapic_pin(vhpet, n); oldval = vhpet->timer[n].cap_config; newval = oldval; update_register(&newval, data, mask); newval &= ~(HPET_TCAP_RO_MASK | HPET_TCNF_32MODE); newval |= oldval & HPET_TCAP_RO_MASK; if (newval == oldval) return; vhpet->timer[n].cap_config = newval; VM_CTR2(vhpet->vm, "hpet t%d cap_config set to 0x%016x", n, newval); /* * Validate the interrupt routing in the HPET_TCNF_INT_ROUTE field. * If it does not match the bits set in HPET_TCAP_INT_ROUTE then set * it to the default value of 0. */ allowed_irqs = vhpet->timer[n].cap_config >> 32; new_pin = vhpet_timer_ioapic_pin(vhpet, n); if (new_pin != 0 && (allowed_irqs & (1 << new_pin)) == 0) { VM_CTR3(vhpet->vm, "hpet t%d configured invalid irq %d, " "allowed_irqs 0x%08x", n, new_pin, allowed_irqs); new_pin = 0; vhpet->timer[n].cap_config &= ~HPET_TCNF_INT_ROUTE; } if (!vhpet_periodic_timer(vhpet, n)) vhpet->timer[n].comprate = 0; /* * If the timer's ISR bit is set then clear it in the following cases: * - interrupt is disabled * - interrupt type is changed from level to edge or fsb. * - interrupt routing is changed * * This is to ensure that this timer's level triggered interrupt does * not remain asserted forever. */ if (vhpet->isr & (1 << n)) { KASSERT(old_pin != 0, ("timer %d isr asserted to ioapic pin %d", n, old_pin)); if (!vhpet_timer_interrupt_enabled(vhpet, n)) clear_isr = true; else if (vhpet_timer_msi_enabled(vhpet, n)) clear_isr = true; else if (vhpet_timer_edge_trig(vhpet, n)) clear_isr = true; else if (vhpet_timer_ioapic_pin(vhpet, n) != old_pin) clear_isr = true; else clear_isr = false; if (clear_isr) { VM_CTR1(vhpet->vm, "hpet t%d isr cleared due to " "configuration change", n); vioapic_deassert_irq(vhpet->vm, old_pin); vhpet->isr &= ~(1 << n); } } } int vhpet_mmio_write(struct vcpu *vcpu, uint64_t gpa, uint64_t val, int size, void *arg) { struct vhpet *vhpet; uint64_t data, mask, oldval, val64; uint32_t isr_clear_mask, old_compval, old_comprate, counter; sbintime_t now, *nowptr; int i, offset; vhpet = vm_hpet(vcpu_vm(vcpu)); offset = gpa - VHPET_BASE; VHPET_LOCK(vhpet); /* Accesses to the HPET should be 4 or 8 bytes wide */ switch (size) { case 8: mask = 0xffffffffffffffff; data = val; break; case 4: mask = 0xffffffff; data = val; if ((offset & 0x4) != 0) { mask <<= 32; data <<= 32; } break; default: VM_CTR2(vhpet->vm, "hpet invalid mmio write: " "offset 0x%08x, size %d", offset, size); goto done; } /* Access to the HPET should be naturally aligned to its width */ if (offset & (size - 1)) { VM_CTR2(vhpet->vm, "hpet invalid mmio write: " "offset 0x%08x, size %d", offset, size); goto done; } if (offset == HPET_CONFIG || offset == HPET_CONFIG + 4) { /* * Get the most recent value of the counter before updating * the 'config' register. If the HPET is going to be disabled * then we need to update 'countbase' with the value right * before it is disabled. */ nowptr = vhpet_counter_enabled(vhpet) ? &now : NULL; counter = vhpet_counter(vhpet, nowptr); oldval = vhpet->config; update_register(&vhpet->config, data, mask); /* * LegacyReplacement Routing is not supported so clear the * bit explicitly. */ vhpet->config &= ~HPET_CNF_LEG_RT; if ((oldval ^ vhpet->config) & HPET_CNF_ENABLE) { if (vhpet_counter_enabled(vhpet)) { vhpet_start_counting(vhpet); VM_CTR0(vhpet->vm, "hpet enabled"); } else { vhpet_stop_counting(vhpet, counter, now); VM_CTR0(vhpet->vm, "hpet disabled"); } } goto done; } if (offset == HPET_ISR || offset == HPET_ISR + 4) { isr_clear_mask = vhpet->isr & data; for (i = 0; i < VHPET_NUM_TIMERS; i++) { if ((isr_clear_mask & (1 << i)) != 0) { VM_CTR1(vhpet->vm, "hpet t%d isr cleared", i); vhpet_timer_clear_isr(vhpet, i); } } goto done; } if (offset == HPET_MAIN_COUNTER || offset == HPET_MAIN_COUNTER + 4) { /* Zero-extend the counter to 64-bits before updating it */ val64 = vhpet_counter(vhpet, NULL); update_register(&val64, data, mask); vhpet->countbase = val64; if (vhpet_counter_enabled(vhpet)) vhpet_start_counting(vhpet); goto done; } for (i = 0; i < VHPET_NUM_TIMERS; i++) { if (offset == HPET_TIMER_CAP_CNF(i) || offset == HPET_TIMER_CAP_CNF(i) + 4) { vhpet_timer_update_config(vhpet, i, data, mask); break; } if (offset == HPET_TIMER_COMPARATOR(i) || offset == HPET_TIMER_COMPARATOR(i) + 4) { old_compval = vhpet->timer[i].compval; old_comprate = vhpet->timer[i].comprate; if (vhpet_periodic_timer(vhpet, i)) { /* * In periodic mode writes to the comparator * change the 'compval' register only if the * HPET_TCNF_VAL_SET bit is set in the config * register. */ val64 = vhpet->timer[i].comprate; update_register(&val64, data, mask); vhpet->timer[i].comprate = val64; if ((vhpet->timer[i].cap_config & HPET_TCNF_VAL_SET) != 0) { vhpet->timer[i].compval = val64; } } else { KASSERT(vhpet->timer[i].comprate == 0, ("vhpet one-shot timer %d has invalid " "rate %u", i, vhpet->timer[i].comprate)); val64 = vhpet->timer[i].compval; update_register(&val64, data, mask); vhpet->timer[i].compval = val64; } vhpet->timer[i].cap_config &= ~HPET_TCNF_VAL_SET; if (vhpet->timer[i].compval != old_compval || vhpet->timer[i].comprate != old_comprate) { if (vhpet_counter_enabled(vhpet)) { counter = vhpet_counter(vhpet, &now); vhpet_start_timer(vhpet, i, counter, now); } } break; } if (offset == HPET_TIMER_FSB_VAL(i) || offset == HPET_TIMER_FSB_ADDR(i)) { update_register(&vhpet->timer[i].msireg, data, mask); break; } } done: VHPET_UNLOCK(vhpet); return (0); } int vhpet_mmio_read(struct vcpu *vcpu, uint64_t gpa, uint64_t *rval, int size, void *arg) { int i, offset; struct vhpet *vhpet; uint64_t data; vhpet = vm_hpet(vcpu_vm(vcpu)); offset = gpa - VHPET_BASE; VHPET_LOCK(vhpet); /* Accesses to the HPET should be 4 or 8 bytes wide */ if (size != 4 && size != 8) { VM_CTR2(vhpet->vm, "hpet invalid mmio read: " "offset 0x%08x, size %d", offset, size); data = 0; goto done; } /* Access to the HPET should be naturally aligned to its width */ if (offset & (size - 1)) { VM_CTR2(vhpet->vm, "hpet invalid mmio read: " "offset 0x%08x, size %d", offset, size); data = 0; goto done; } if (offset == HPET_CAPABILITIES || offset == HPET_CAPABILITIES + 4) { data = vhpet_capabilities(); goto done; } if (offset == HPET_CONFIG || offset == HPET_CONFIG + 4) { data = vhpet->config; goto done; } if (offset == HPET_ISR || offset == HPET_ISR + 4) { data = vhpet->isr; goto done; } if (offset == HPET_MAIN_COUNTER || offset == HPET_MAIN_COUNTER + 4) { data = vhpet_counter(vhpet, NULL); goto done; } for (i = 0; i < VHPET_NUM_TIMERS; i++) { if (offset == HPET_TIMER_CAP_CNF(i) || offset == HPET_TIMER_CAP_CNF(i) + 4) { data = vhpet->timer[i].cap_config; break; } if (offset == HPET_TIMER_COMPARATOR(i) || offset == HPET_TIMER_COMPARATOR(i) + 4) { data = vhpet->timer[i].compval; break; } if (offset == HPET_TIMER_FSB_VAL(i) || offset == HPET_TIMER_FSB_ADDR(i)) { data = vhpet->timer[i].msireg; break; } } if (i >= VHPET_NUM_TIMERS) data = 0; done: VHPET_UNLOCK(vhpet); if (size == 4) { if (offset & 0x4) data >>= 32; } *rval = data; return (0); } struct vhpet * vhpet_init(struct vm *vm) { int i, pincount; struct vhpet *vhpet; uint64_t allowed_irqs; struct vhpet_callout_arg *arg; struct bintime bt; vhpet = malloc(sizeof(struct vhpet), M_VHPET, M_WAITOK | M_ZERO); vhpet->vm = vm; mtx_init(&vhpet->mtx, "vhpet lock", NULL, MTX_DEF); FREQ2BT(HPET_FREQ, &bt); vhpet->freq_sbt = bttosbt(bt); pincount = vioapic_pincount(vm); if (pincount >= 32) allowed_irqs = 0xff000000; /* irqs 24-31 */ else if (pincount >= 20) allowed_irqs = 0xf << (pincount - 4); /* 4 upper irqs */ else allowed_irqs = 0; /* * Initialize HPET timer hardware state. */ for (i = 0; i < VHPET_NUM_TIMERS; i++) { vhpet->timer[i].cap_config = allowed_irqs << 32; vhpet->timer[i].cap_config |= HPET_TCAP_PER_INT; vhpet->timer[i].cap_config |= HPET_TCAP_FSB_INT_DEL; vhpet->timer[i].compval = 0xffffffff; callout_init(&vhpet->timer[i].callout, 1); arg = &vhpet->timer[i].arg; arg->vhpet = vhpet; arg->timer_num = i; } return (vhpet); } void vhpet_cleanup(struct vhpet *vhpet) { int i; for (i = 0; i < VHPET_NUM_TIMERS; i++) callout_drain(&vhpet->timer[i].callout); mtx_destroy(&vhpet->mtx); free(vhpet, M_VHPET); } int vhpet_getcap(struct vm_hpet_cap *cap) { cap->capabilities = vhpet_capabilities(); return (0); } #ifdef BHYVE_SNAPSHOT int vhpet_snapshot(struct vhpet *vhpet, struct vm_snapshot_meta *meta) { int i, ret; uint32_t countbase; SNAPSHOT_VAR_OR_LEAVE(vhpet->freq_sbt, meta, ret, done); SNAPSHOT_VAR_OR_LEAVE(vhpet->config, meta, ret, done); SNAPSHOT_VAR_OR_LEAVE(vhpet->isr, meta, ret, done); /* at restore time the countbase should have the value it had when the * snapshot was created; since the value is not directly kept in * vhpet->countbase, but rather computed relative to the current system * uptime using countbase_sbt, save the value retured by vhpet_counter */ if (meta->op == VM_SNAPSHOT_SAVE) countbase = vhpet_counter(vhpet, NULL); SNAPSHOT_VAR_OR_LEAVE(countbase, meta, ret, done); if (meta->op == VM_SNAPSHOT_RESTORE) vhpet->countbase = countbase; for (i = 0; i < nitems(vhpet->timer); i++) { SNAPSHOT_VAR_OR_LEAVE(vhpet->timer[i].cap_config, meta, ret, done); SNAPSHOT_VAR_OR_LEAVE(vhpet->timer[i].msireg, meta, ret, done); SNAPSHOT_VAR_OR_LEAVE(vhpet->timer[i].compval, meta, ret, done); SNAPSHOT_VAR_OR_LEAVE(vhpet->timer[i].comprate, meta, ret, done); SNAPSHOT_VAR_OR_LEAVE(vhpet->timer[i].callout_sbt, meta, ret, done); } done: return (ret); } int vhpet_restore_time(struct vhpet *vhpet) { if (vhpet_counter_enabled(vhpet)) vhpet_start_counting(vhpet); return (0); } #endif