Message ID | 20170614165223.7543-2-andre.przywara@arm.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Hi Andre, On 06/14/2017 05:51 PM, Andre Przywara wrote: > When reading the priority value of a virtual interrupt, we were taking > the respective rank lock so far. > However for forwarded interrupts (Dom0 only so far) this may lead to a > deadlock with the following call chain: > - MMIO access to change the IRQ affinity, calling the ITARGETSR handler > - this handler takes the appropriate rank lock and calls vgic_store_itargetsr() > - vgic_store_itargetsr() will eventually call vgic_migrate_irq() > - if this IRQ is already in-flight, it will remove it from the old > VCPU and inject it into the new one, by calling vgic_vcpu_inject_irq() > - vgic_vcpu_inject_irq will call vgic_get_virq_priority() > - vgic_get_virq_priority() tries to take the rank lock - again! > It seems like this code path has never been exercised before. > > Fix this by avoiding taking the lock in vgic_get_virq_priority() (like we > do in vgic_get_target_vcpu()). > Actually we are just reading one byte, and priority changes while > interrupts are handled are a benign race that can happen on real hardware > too. So it is safe to just prevent the compiler from reading from the > struct more than once. > > Signed-off-by: Andre Przywara <andre.przywara@arm.com> Reviewed-by: Julien Grall <julien.grall@arm.com> Cheers,
Hi, On 14/06/17 18:55, Julien Grall wrote: > Hi Andre, > > On 06/14/2017 05:51 PM, Andre Przywara wrote: >> When reading the priority value of a virtual interrupt, we were taking >> the respective rank lock so far. >> However for forwarded interrupts (Dom0 only so far) this may lead to a >> deadlock with the following call chain: >> - MMIO access to change the IRQ affinity, calling the ITARGETSR handler >> - this handler takes the appropriate rank lock and calls >> vgic_store_itargetsr() >> - vgic_store_itargetsr() will eventually call vgic_migrate_irq() >> - if this IRQ is already in-flight, it will remove it from the old >> VCPU and inject it into the new one, by calling vgic_vcpu_inject_irq() >> - vgic_vcpu_inject_irq will call vgic_get_virq_priority() >> - vgic_get_virq_priority() tries to take the rank lock - again! >> It seems like this code path has never been exercised before. >> >> Fix this by avoiding taking the lock in vgic_get_virq_priority() (like we >> do in vgic_get_target_vcpu()). >> Actually we are just reading one byte, and priority changes while >> interrupts are handled are a benign race that can happen on real hardware >> too. So it is safe to just prevent the compiler from reading from the >> struct more than once. >> >> Signed-off-by: Andre Przywara <andre.przywara@arm.com> > > Reviewed-by: Julien Grall <julien.grall@arm.com> As you mentioned in the cover letter, I think it would be good to get this fixed in Xen 4.9. Stefano what do you think? Cheers,
On Mon, 19 Jun 2017, Julien Grall wrote: > Hi, > > On 14/06/17 18:55, Julien Grall wrote: > > Hi Andre, > > > > On 06/14/2017 05:51 PM, Andre Przywara wrote: > > > When reading the priority value of a virtual interrupt, we were taking > > > the respective rank lock so far. > > > However for forwarded interrupts (Dom0 only so far) this may lead to a > > > deadlock with the following call chain: > > > - MMIO access to change the IRQ affinity, calling the ITARGETSR handler > > > - this handler takes the appropriate rank lock and calls > > > vgic_store_itargetsr() > > > - vgic_store_itargetsr() will eventually call vgic_migrate_irq() > > > - if this IRQ is already in-flight, it will remove it from the old > > > VCPU and inject it into the new one, by calling vgic_vcpu_inject_irq() > > > - vgic_vcpu_inject_irq will call vgic_get_virq_priority() > > > - vgic_get_virq_priority() tries to take the rank lock - again! > > > It seems like this code path has never been exercised before. > > > > > > Fix this by avoiding taking the lock in vgic_get_virq_priority() (like we > > > do in vgic_get_target_vcpu()). > > > Actually we are just reading one byte, and priority changes while > > > interrupts are handled are a benign race that can happen on real hardware > > > too. So it is safe to just prevent the compiler from reading from the > > > struct more than once. > > > > > > Signed-off-by: Andre Przywara <andre.przywara@arm.com> > > > > Reviewed-by: Julien Grall <julien.grall@arm.com> > > As you mentioned in the cover letter, I think it would be good to get this > fixed in Xen 4.9. Stefano what do you think? Yes, I think it makes sense. Do you release-ack it?
Hi, On 06/19/2017 11:44 PM, Stefano Stabellini wrote: > On Mon, 19 Jun 2017, Julien Grall wrote: >> Hi, >> >> On 14/06/17 18:55, Julien Grall wrote: >>> Hi Andre, >>> >>> On 06/14/2017 05:51 PM, Andre Przywara wrote: >>>> When reading the priority value of a virtual interrupt, we were taking >>>> the respective rank lock so far. >>>> However for forwarded interrupts (Dom0 only so far) this may lead to a >>>> deadlock with the following call chain: >>>> - MMIO access to change the IRQ affinity, calling the ITARGETSR handler >>>> - this handler takes the appropriate rank lock and calls >>>> vgic_store_itargetsr() >>>> - vgic_store_itargetsr() will eventually call vgic_migrate_irq() >>>> - if this IRQ is already in-flight, it will remove it from the old >>>> VCPU and inject it into the new one, by calling vgic_vcpu_inject_irq() >>>> - vgic_vcpu_inject_irq will call vgic_get_virq_priority() >>>> - vgic_get_virq_priority() tries to take the rank lock - again! >>>> It seems like this code path has never been exercised before. >>>> >>>> Fix this by avoiding taking the lock in vgic_get_virq_priority() (like we >>>> do in vgic_get_target_vcpu()). >>>> Actually we are just reading one byte, and priority changes while >>>> interrupts are handled are a benign race that can happen on real hardware >>>> too. So it is safe to just prevent the compiler from reading from the >>>> struct more than once. >>>> >>>> Signed-off-by: Andre Przywara <andre.przywara@arm.com> >>> >>> Reviewed-by: Julien Grall <julien.grall@arm.com> >> >> As you mentioned in the cover letter, I think it would be good to get this >> fixed in Xen 4.9. Stefano what do you think? > > Yes, I think it makes sense. Do you release-ack it? Release-acked-by: Julien Grall <julien.grall@arm.com> Cheers,
diff --git a/xen/arch/arm/vgic-v2.c b/xen/arch/arm/vgic-v2.c index dc9f95b..9fa42e1 100644 --- a/xen/arch/arm/vgic-v2.c +++ b/xen/arch/arm/vgic-v2.c @@ -252,15 +252,15 @@ static int vgic_v2_distr_mmio_read(struct vcpu *v, mmio_info_t *info, case VRANGE32(GICD_IPRIORITYR, GICD_IPRIORITYRN): { uint32_t ipriorityr; + uint8_t rank_index; if ( dabt.size != DABT_BYTE && dabt.size != DABT_WORD ) goto bad_width; rank = vgic_rank_offset(v, 8, gicd_reg - GICD_IPRIORITYR, DABT_WORD); if ( rank == NULL ) goto read_as_zero; + rank_index = REG_RANK_INDEX(8, gicd_reg - GICD_IPRIORITYR, DABT_WORD); vgic_lock_rank(v, rank, flags); - ipriorityr = rank->ipriorityr[REG_RANK_INDEX(8, - gicd_reg - GICD_IPRIORITYR, - DABT_WORD)]; + ipriorityr = ACCESS_ONCE(rank->ipriorityr[rank_index]); vgic_unlock_rank(v, rank, flags); *r = vgic_reg32_extract(ipriorityr, info); @@ -499,7 +499,7 @@ static int vgic_v2_distr_mmio_write(struct vcpu *v, mmio_info_t *info, case VRANGE32(GICD_IPRIORITYR, GICD_IPRIORITYRN): { - uint32_t *ipriorityr; + uint32_t *ipriorityr, priority; if ( dabt.size != DABT_BYTE && dabt.size != DABT_WORD ) goto bad_width; rank = vgic_rank_offset(v, 8, gicd_reg - GICD_IPRIORITYR, DABT_WORD); @@ -508,7 +508,10 @@ static int vgic_v2_distr_mmio_write(struct vcpu *v, mmio_info_t *info, ipriorityr = &rank->ipriorityr[REG_RANK_INDEX(8, gicd_reg - GICD_IPRIORITYR, DABT_WORD)]; - vgic_reg32_update(ipriorityr, r, info); + priority = ACCESS_ONCE(*ipriorityr); + vgic_reg32_update(&priority, r, info); + ACCESS_ONCE(*ipriorityr) = priority; + vgic_unlock_rank(v, rank, flags); return 1; } diff --git a/xen/arch/arm/vgic-v3.c b/xen/arch/arm/vgic-v3.c index d10757a..9018ddc 100644 --- a/xen/arch/arm/vgic-v3.c +++ b/xen/arch/arm/vgic-v3.c @@ -515,14 +515,15 @@ static int __vgic_v3_distr_common_mmio_read(const char *name, struct vcpu *v, case VRANGE32(GICD_IPRIORITYR, GICD_IPRIORITYRN): { uint32_t ipriorityr; + uint8_t rank_index; if ( dabt.size != DABT_BYTE && dabt.size != DABT_WORD ) goto bad_width; rank = vgic_rank_offset(v, 8, reg - GICD_IPRIORITYR, DABT_WORD); if ( rank == NULL ) goto read_as_zero; + rank_index = REG_RANK_INDEX(8, reg - GICD_IPRIORITYR, DABT_WORD); vgic_lock_rank(v, rank, flags); - ipriorityr = rank->ipriorityr[REG_RANK_INDEX(8, reg - GICD_IPRIORITYR, - DABT_WORD)]; + ipriorityr = ACCESS_ONCE(rank->ipriorityr[rank_index]); vgic_unlock_rank(v, rank, flags); *r = vgic_reg32_extract(ipriorityr, info); @@ -630,7 +631,7 @@ static int __vgic_v3_distr_common_mmio_write(const char *name, struct vcpu *v, case VRANGE32(GICD_IPRIORITYR, GICD_IPRIORITYRN): { - uint32_t *ipriorityr; + uint32_t *ipriorityr, priority; if ( dabt.size != DABT_BYTE && dabt.size != DABT_WORD ) goto bad_width; rank = vgic_rank_offset(v, 8, reg - GICD_IPRIORITYR, DABT_WORD); @@ -638,7 +639,9 @@ static int __vgic_v3_distr_common_mmio_write(const char *name, struct vcpu *v, vgic_lock_rank(v, rank, flags); ipriorityr = &rank->ipriorityr[REG_RANK_INDEX(8, reg - GICD_IPRIORITYR, DABT_WORD)]; - vgic_reg32_update(ipriorityr, r, info); + priority = ACCESS_ONCE(*ipriorityr); + vgic_reg32_update(&priority, r, info); + ACCESS_ONCE(*ipriorityr) = priority; vgic_unlock_rank(v, rank, flags); return 1; } diff --git a/xen/arch/arm/vgic.c b/xen/arch/arm/vgic.c index 83569b0..18fe420 100644 --- a/xen/arch/arm/vgic.c +++ b/xen/arch/arm/vgic.c @@ -227,14 +227,8 @@ struct vcpu *vgic_get_target_vcpu(struct vcpu *v, unsigned int virq) static int vgic_get_virq_priority(struct vcpu *v, unsigned int virq) { struct vgic_irq_rank *rank = vgic_rank_irq(v, virq); - unsigned long flags; - int priority; - - vgic_lock_rank(v, rank, flags); - priority = rank->priority[virq & INTERRUPT_RANK_MASK]; - vgic_unlock_rank(v, rank, flags); - return priority; + return ACCESS_ONCE(rank->priority[virq & INTERRUPT_RANK_MASK]); } bool vgic_migrate_irq(struct vcpu *old, struct vcpu *new, unsigned int irq)
When reading the priority value of a virtual interrupt, we were taking the respective rank lock so far. However for forwarded interrupts (Dom0 only so far) this may lead to a deadlock with the following call chain: - MMIO access to change the IRQ affinity, calling the ITARGETSR handler - this handler takes the appropriate rank lock and calls vgic_store_itargetsr() - vgic_store_itargetsr() will eventually call vgic_migrate_irq() - if this IRQ is already in-flight, it will remove it from the old VCPU and inject it into the new one, by calling vgic_vcpu_inject_irq() - vgic_vcpu_inject_irq will call vgic_get_virq_priority() - vgic_get_virq_priority() tries to take the rank lock - again! It seems like this code path has never been exercised before. Fix this by avoiding taking the lock in vgic_get_virq_priority() (like we do in vgic_get_target_vcpu()). Actually we are just reading one byte, and priority changes while interrupts are handled are a benign race that can happen on real hardware too. So it is safe to just prevent the compiler from reading from the struct more than once. Signed-off-by: Andre Przywara <andre.przywara@arm.com> --- xen/arch/arm/vgic-v2.c | 13 ++++++++----- xen/arch/arm/vgic-v3.c | 11 +++++++---- xen/arch/arm/vgic.c | 8 +------- 3 files changed, 16 insertions(+), 16 deletions(-)