From: Roger Pau Monné Subject: x86/vpt: fix race when migrating timers between vCPUs The current vPT code will migrate the emulated timers between vCPUs (change the pt->vcpu field) while just holding the destination lock, either from create_periodic_time or pt_adjust_global_vcpu_target if the global target is adjusted. Changing the periodic_timer vCPU field in this way creates a race where a third party could grab the lock in the unlocked region of pt_adjust_global_vcpu_target (or before create_periodic_time performs the vcpu change) and then release the lock from a different vCPU, creating a locking imbalance. Introduce a per-domain rwlock in order to protect periodic_time migration between vCPU lists. Taking the lock in read mode prevents any timer from being migrated to a different vCPU, while taking it in write mode allows performing migration of timers across vCPUs. The per-vcpu locks are still used to protect all the other fields from the periodic_timer struct. Note that such migration shouldn't happen frequently, and hence there's no performance drop as a result of such locking. This is XSA-336. Reported-by: Igor Druzhinin Tested-by: Igor Druzhinin Signed-off-by: Roger Pau Monné Reviewed-by: Jan Beulich --- Changes since v2: - Re-order pt_adjust_vcpu to remove one if. - Fix pt_lock to not call pt_vcpu_lock, as we might end up using a stale value of pt->vcpu when taking the per-vcpu lock. Changes since v1: - Use a per-domain rwlock to protect timer vCPU migration. --- a/xen/arch/x86/hvm/hvm.c +++ b/xen/arch/x86/hvm/hvm.c @@ -658,6 +658,8 @@ int hvm_domain_initialise(struct domain /* need link to containing domain */ d->arch.hvm.pl_time->domain = d; + rwlock_init(&d->arch.hvm.pl_time->pt_migrate); + /* Set the default IO Bitmap. */ if ( is_hardware_domain(d) ) { --- a/xen/arch/x86/hvm/vpt.c +++ b/xen/arch/x86/hvm/vpt.c @@ -153,23 +153,32 @@ static int pt_irq_masked(struct periodic return 1; } -static void pt_lock(struct periodic_time *pt) +static void pt_vcpu_lock(struct vcpu *v) { - struct vcpu *v; + read_lock(&v->domain->arch.hvm.pl_time->pt_migrate); + spin_lock(&v->arch.hvm.tm_lock); +} - for ( ; ; ) - { - v = pt->vcpu; - spin_lock(&v->arch.hvm.tm_lock); - if ( likely(pt->vcpu == v) ) - break; - spin_unlock(&v->arch.hvm.tm_lock); - } +static void pt_vcpu_unlock(struct vcpu *v) +{ + spin_unlock(&v->arch.hvm.tm_lock); + read_unlock(&v->domain->arch.hvm.pl_time->pt_migrate); +} + +static void pt_lock(struct periodic_time *pt) +{ + /* + * We cannot use pt_vcpu_lock here, because we need to acquire the + * per-domain lock first and then (re-)fetch the value of pt->vcpu, or + * else we might be using a stale value of pt->vcpu. + */ + read_lock(&pt->vcpu->domain->arch.hvm.pl_time->pt_migrate); + spin_lock(&pt->vcpu->arch.hvm.tm_lock); } static void pt_unlock(struct periodic_time *pt) { - spin_unlock(&pt->vcpu->arch.hvm.tm_lock); + pt_vcpu_unlock(pt->vcpu); } static void pt_process_missed_ticks(struct periodic_time *pt) @@ -219,7 +228,7 @@ void pt_save_timer(struct vcpu *v) if ( v->pause_flags & VPF_blocked ) return; - spin_lock(&v->arch.hvm.tm_lock); + pt_vcpu_lock(v); list_for_each_entry ( pt, head, list ) if ( !pt->do_not_freeze ) @@ -227,7 +236,7 @@ void pt_save_timer(struct vcpu *v) pt_freeze_time(v); - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); } void pt_restore_timer(struct vcpu *v) @@ -235,7 +244,7 @@ void pt_restore_timer(struct vcpu *v) struct list_head *head = &v->arch.hvm.tm_list; struct periodic_time *pt; - spin_lock(&v->arch.hvm.tm_lock); + pt_vcpu_lock(v); list_for_each_entry ( pt, head, list ) { @@ -248,7 +257,7 @@ void pt_restore_timer(struct vcpu *v) pt_thaw_time(v); - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); } static void pt_timer_fn(void *data) @@ -309,7 +318,7 @@ int pt_update_irq(struct vcpu *v) int irq, pt_vector = -1; bool level; - spin_lock(&v->arch.hvm.tm_lock); + pt_vcpu_lock(v); earliest_pt = NULL; max_lag = -1ULL; @@ -339,7 +348,7 @@ int pt_update_irq(struct vcpu *v) if ( earliest_pt == NULL ) { - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); return -1; } @@ -347,7 +356,7 @@ int pt_update_irq(struct vcpu *v) irq = earliest_pt->irq; level = earliest_pt->level; - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); switch ( earliest_pt->source ) { @@ -394,7 +403,7 @@ int pt_update_irq(struct vcpu *v) time_cb *cb = NULL; void *cb_priv; - spin_lock(&v->arch.hvm.tm_lock); + pt_vcpu_lock(v); /* Make sure the timer is still on the list. */ list_for_each_entry ( pt, &v->arch.hvm.tm_list, list ) if ( pt == earliest_pt ) @@ -404,7 +413,7 @@ int pt_update_irq(struct vcpu *v) cb_priv = pt->priv; break; } - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); if ( cb != NULL ) cb(v, cb_priv); @@ -441,12 +450,12 @@ void pt_intr_post(struct vcpu *v, struct if ( intack.source == hvm_intsrc_vector ) return; - spin_lock(&v->arch.hvm.tm_lock); + pt_vcpu_lock(v); pt = is_pt_irq(v, intack); if ( pt == NULL ) { - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); return; } @@ -455,7 +464,7 @@ void pt_intr_post(struct vcpu *v, struct cb = pt->cb; cb_priv = pt->priv; - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); if ( cb != NULL ) cb(v, cb_priv); @@ -466,12 +475,12 @@ void pt_migrate(struct vcpu *v) struct list_head *head = &v->arch.hvm.tm_list; struct periodic_time *pt; - spin_lock(&v->arch.hvm.tm_lock); + pt_vcpu_lock(v); list_for_each_entry ( pt, head, list ) migrate_timer(&pt->timer, v->processor); - spin_unlock(&v->arch.hvm.tm_lock); + pt_vcpu_unlock(v); } void create_periodic_time( @@ -490,7 +499,7 @@ void create_periodic_time( destroy_periodic_time(pt); - spin_lock(&v->arch.hvm.tm_lock); + write_lock(&v->domain->arch.hvm.pl_time->pt_migrate); pt->pending_intr_nr = 0; pt->do_not_freeze = 0; @@ -540,7 +549,7 @@ void create_periodic_time( init_timer(&pt->timer, pt_timer_fn, pt, v->processor); set_timer(&pt->timer, pt->scheduled); - spin_unlock(&v->arch.hvm.tm_lock); + write_unlock(&v->domain->arch.hvm.pl_time->pt_migrate); } void destroy_periodic_time(struct periodic_time *pt) @@ -565,30 +574,20 @@ void destroy_periodic_time(struct period static void pt_adjust_vcpu(struct periodic_time *pt, struct vcpu *v) { - int on_list; - ASSERT(pt->source == PTSRC_isa || pt->source == PTSRC_ioapic); if ( pt->vcpu == NULL ) return; - pt_lock(pt); - on_list = pt->on_list; - if ( pt->on_list ) - list_del(&pt->list); - pt->on_list = 0; - pt_unlock(pt); - - spin_lock(&v->arch.hvm.tm_lock); + write_lock(&pt->vcpu->domain->arch.hvm.pl_time->pt_migrate); pt->vcpu = v; - if ( on_list ) + if ( pt->on_list ) { - pt->on_list = 1; + list_del(&pt->list); list_add(&pt->list, &v->arch.hvm.tm_list); - migrate_timer(&pt->timer, v->processor); } - spin_unlock(&v->arch.hvm.tm_lock); + write_unlock(&pt->vcpu->domain->arch.hvm.pl_time->pt_migrate); } void pt_adjust_global_vcpu_target(struct vcpu *v) --- a/xen/include/asm-x86/hvm/vpt.h +++ b/xen/include/asm-x86/hvm/vpt.h @@ -128,6 +128,13 @@ struct pl_time { /* platform time */ struct RTCState vrtc; struct HPETState vhpet; struct PMTState vpmt; + /* + * rwlock to prevent periodic_time vCPU migration. Take the lock in read + * mode in order to prevent the vcpu field of periodic_time from changing. + * Lock must be taken in write mode when changes to the vcpu field are + * performed, as it allows exclusive access to all the timers of a domain. + */ + rwlock_t pt_migrate; /* guest_time = Xen sys time + stime_offset */ int64_t stime_offset; /* Ensures monotonicity in appropriate timer modes. */