iommu/vt-d: Don't queue_iova() if there is no flush queue
authorDmitry Safonov <dima@arista.com>
Tue, 16 Jul 2019 21:38:05 +0000 (22:38 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 4 Aug 2019 07:30:54 +0000 (09:30 +0200)
commit effa467870c7612012885df4e246bdb8ffd8e44c upstream.

Intel VT-d driver was reworked to use common deferred flushing
implementation. Previously there was one global per-cpu flush queue,
afterwards - one per domain.

Before deferring a flush, the queue should be allocated and initialized.

Currently only domains with IOMMU_DOMAIN_DMA type initialize their flush
queue. It's probably worth to init it for static or unmanaged domains
too, but it may be arguable - I'm leaving it to iommu folks.

Prevent queuing an iova flush if the domain doesn't have a queue.
The defensive check seems to be worth to keep even if queue would be
initialized for all kinds of domains. And is easy backportable.

On 4.19.43 stable kernel it has a user-visible effect: previously for
devices in si domain there were crashes, on sata devices:

 BUG: spinlock bad magic on CPU#6, swapper/0/1
  lock: 0xffff88844f582008, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
 CPU: 6 PID: 1 Comm: swapper/0 Not tainted 4.19.43 #1
 Call Trace:
  <IRQ>
  dump_stack+0x61/0x7e
  spin_bug+0x9d/0xa3
  do_raw_spin_lock+0x22/0x8e
  _raw_spin_lock_irqsave+0x32/0x3a
  queue_iova+0x45/0x115
  intel_unmap+0x107/0x113
  intel_unmap_sg+0x6b/0x76
  __ata_qc_complete+0x7f/0x103
  ata_qc_complete+0x9b/0x26a
  ata_qc_complete_multiple+0xd0/0xe3
  ahci_handle_port_interrupt+0x3ee/0x48a
  ahci_handle_port_intr+0x73/0xa9
  ahci_single_level_irq_intr+0x40/0x60
  __handle_irq_event_percpu+0x7f/0x19a
  handle_irq_event_percpu+0x32/0x72
  handle_irq_event+0x38/0x56
  handle_edge_irq+0x102/0x121
  handle_irq+0x147/0x15c
  do_IRQ+0x66/0xf2
  common_interrupt+0xf/0xf
 RIP: 0010:__do_softirq+0x8c/0x2df

The same for usb devices that use ehci-pci:
 BUG: spinlock bad magic on CPU#0, swapper/0/1
  lock: 0xffff88844f402008, .magic: 00000000, .owner: <none>/-1, .owner_cpu: 0
 CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.19.43 #4
 Call Trace:
  <IRQ>
  dump_stack+0x61/0x7e
  spin_bug+0x9d/0xa3
  do_raw_spin_lock+0x22/0x8e
  _raw_spin_lock_irqsave+0x32/0x3a
  queue_iova+0x77/0x145
  intel_unmap+0x107/0x113
  intel_unmap_page+0xe/0x10
  usb_hcd_unmap_urb_setup_for_dma+0x53/0x9d
  usb_hcd_unmap_urb_for_dma+0x17/0x100
  unmap_urb_for_dma+0x22/0x24
  __usb_hcd_giveback_urb+0x51/0xc3
  usb_giveback_urb_bh+0x97/0xde
  tasklet_action_common.isra.4+0x5f/0xa1
  tasklet_action+0x2d/0x30
  __do_softirq+0x138/0x2df
  irq_exit+0x7d/0x8b
  smp_apic_timer_interrupt+0x10f/0x151
  apic_timer_interrupt+0xf/0x20
  </IRQ>
 RIP: 0010:_raw_spin_unlock_irqrestore+0x17/0x39

Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Lu Baolu <baolu.lu@linux.intel.com>
Cc: iommu@lists.linux-foundation.org
Cc: <stable@vger.kernel.org> # 4.14+
Fixes: 13cf01744608 ("iommu/vt-d: Make use of iova deferred flushing")
Signed-off-by: Dmitry Safonov <dima@arista.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Joerg Roedel <jroedel@suse.de>
[v4.14-port notes:
o minor conflict with untrusted IOMMU devices check under if-condition]
Signed-off-by: Dmitry Safonov <dima@arista.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/iommu/intel-iommu.c
drivers/iommu/iova.c
include/linux/iova.h

index c143901..b9af241 100644 (file)
@@ -3721,7 +3721,7 @@ static void intel_unmap(struct device *dev, dma_addr_t dev_addr, size_t size)
 
        freelist = domain_unmap(domain, start_pfn, last_pfn);
 
-       if (intel_iommu_strict) {
+       if (intel_iommu_strict || !has_iova_flush_queue(&domain->iovad)) {
                iommu_flush_iotlb_psi(iommu, domain, start_pfn,
                                      nrpages, !freelist, 0);
                /* free iova */
index 83fe262..60348d7 100644 (file)
@@ -65,9 +65,14 @@ init_iova_domain(struct iova_domain *iovad, unsigned long granule,
 }
 EXPORT_SYMBOL_GPL(init_iova_domain);
 
+bool has_iova_flush_queue(struct iova_domain *iovad)
+{
+       return !!iovad->fq;
+}
+
 static void free_iova_flush_queue(struct iova_domain *iovad)
 {
-       if (!iovad->fq)
+       if (!has_iova_flush_queue(iovad))
                return;
 
        if (timer_pending(&iovad->fq_timer))
@@ -85,13 +90,14 @@ static void free_iova_flush_queue(struct iova_domain *iovad)
 int init_iova_flush_queue(struct iova_domain *iovad,
                          iova_flush_cb flush_cb, iova_entry_dtor entry_dtor)
 {
+       struct iova_fq __percpu *queue;
        int cpu;
 
        atomic64_set(&iovad->fq_flush_start_cnt,  0);
        atomic64_set(&iovad->fq_flush_finish_cnt, 0);
 
-       iovad->fq = alloc_percpu(struct iova_fq);
-       if (!iovad->fq)
+       queue = alloc_percpu(struct iova_fq);
+       if (!queue)
                return -ENOMEM;
 
        iovad->flush_cb   = flush_cb;
@@ -100,13 +106,17 @@ int init_iova_flush_queue(struct iova_domain *iovad,
        for_each_possible_cpu(cpu) {
                struct iova_fq *fq;
 
-               fq = per_cpu_ptr(iovad->fq, cpu);
+               fq = per_cpu_ptr(queue, cpu);
                fq->head = 0;
                fq->tail = 0;
 
                spin_lock_init(&fq->lock);
        }
 
+       smp_wmb();
+
+       iovad->fq = queue;
+
        timer_setup(&iovad->fq_timer, fq_flush_timeout, 0);
        atomic_set(&iovad->fq_timer_on, 0);
 
index 928442d..073dc27 100644 (file)
@@ -156,6 +156,7 @@ struct iova *reserve_iova(struct iova_domain *iovad, unsigned long pfn_lo,
 void copy_reserved_iova(struct iova_domain *from, struct iova_domain *to);
 void init_iova_domain(struct iova_domain *iovad, unsigned long granule,
        unsigned long start_pfn);
+bool has_iova_flush_queue(struct iova_domain *iovad);
 int init_iova_flush_queue(struct iova_domain *iovad,
                          iova_flush_cb flush_cb, iova_entry_dtor entry_dtor);
 struct iova *find_iova(struct iova_domain *iovad, unsigned long pfn);
@@ -236,6 +237,11 @@ static inline void init_iova_domain(struct iova_domain *iovad,
 {
 }
 
+bool has_iova_flush_queue(struct iova_domain *iovad)
+{
+       return false;
+}
+
 static inline int init_iova_flush_queue(struct iova_domain *iovad,
                                        iova_flush_cb flush_cb,
                                        iova_entry_dtor entry_dtor)