All of lore.kernel.org
 help / color / mirror / Atom feed
* [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct
@ 2016-09-23  5:02 Peter Xu
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps Peter Xu
                   ` (3 more replies)
  0 siblings, 4 replies; 7+ messages in thread
From: Peter Xu @ 2016-09-23  5:02 UTC (permalink / raw)
  To: qemu-devel
  Cc: dgibson, jasowang, mst, pbonzini, cornelia.huck, wexu, vkaplans,
	alex.williamson, peterx

V7:
- add comments to memory_region_notify_iommu() to better clarify the
  interface [David]
- vfio_iommu_map_notify(): remove pointless "IOMMUTLBEntry *iotlb =
  data" [David]
- typo fix on English [David]

V6:
- use IOMMUTLBEntry directly for IOMMUNotifier.notify() [David]
- add explicit comment for patch 1 that CHANGE should be treated as a
  MAP then an UNMAP [David]
- add a-b for David on patch 2

V5:
- squash spapr_tce_notify_{started|stopped} into
  spapr_tce_notify_flag_changed [David]
- in spapr_tce_notify_flag_changed: check flags against "!=
  IOMMU_NOTIFIER_NONE", but not "== IOMMU_NOTIFIER_ALL" [David]
- add r-b for David on patch 3

V4:
- change "notifier_caps" into "notifier_flags" [David]
- rename IOMMU_NOTIFIER_{CHANGE|INVALIDATION} with MAP/UNMAP [David]
- introduce IOMMUOps.notify_flag_changed, to replace notify_started
  and notify_stopped [David, Paolo]

V3:
- use QLIST instead of embedding Notifier into IOMMUNotifier [Paolo]
- fix a build error for ppc64-softmmu

The idea originates from one of Alex's reply:

  https://lists.gnu.org/archive/html/qemu-devel/2016-09/msg00254.html

But after further discussions, it seems that only adding a simple type
for notifier is not enough. This series introduced IOMMUNotifier
struct to replace the old Notifier interface. Along with it, we can
provide registration for one (or multiple) of the IOMMU notifications:

- cache invalidations
- entry changes

This is a support material for Jason's vhost dmar patchset.

Please read commit messages for detailed information. Thanks,

Peter Xu (3):
  memory: introduce IOMMUNotifier and its caps
  memory: introduce IOMMUOps.notify_flag_changed
  intel_iommu: allow UNMAP notifiers

 hw/i386/intel_iommu.c         | 18 ++++++++-----
 hw/ppc/spapr_iommu.c          | 18 +++++++------
 hw/vfio/common.c              |  4 +--
 include/exec/memory.h         | 56 ++++++++++++++++++++++++++++++++--------
 include/hw/vfio/vfio-common.h |  2 +-
 memory.c                      | 60 +++++++++++++++++++++++++++++++++----------
 6 files changed, 116 insertions(+), 42 deletions(-)

-- 
2.7.4

^ permalink raw reply	[flat|nested] 7+ messages in thread

* [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps
  2016-09-23  5:02 [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct Peter Xu
@ 2016-09-23  5:02 ` Peter Xu
  2016-09-23  7:20   ` David Gibson
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 2/3] memory: introduce IOMMUOps.notify_flag_changed Peter Xu
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 7+ messages in thread
From: Peter Xu @ 2016-09-23  5:02 UTC (permalink / raw)
  To: qemu-devel
  Cc: dgibson, jasowang, mst, pbonzini, cornelia.huck, wexu, vkaplans,
	alex.williamson, peterx

IOMMU Notifier list is used for notifying IO address mapping changes.
Currently VFIO is the only user.

However it is possible that future consumer like vhost would like to
only listen to part of its notifications (e.g., cache invalidations).

This patch introduced IOMMUNotifier and IOMMUNotfierFlag bits for a
finer grained control of it.

IOMMUNotifier contains a bitfield for the notify consumer describing
what kind of notification it is interested in. Currently two kinds of
notifications are defined:

- IOMMU_NOTIFIER_MAP:    for newly mapped entries (additions)
- IOMMU_NOTIFIER_UNMAP:  for entries to be removed (cache invalidates)

When registering the IOMMU notifier, we need to specify one or multiple
types of messages to listen to.

When notifications are triggered, its type will be checked against the
notifier's type bits, and only notifiers with registered bits will be
notified.

(For any IOMMU implementation, an in-place mapping change should be
 notified with an UNMAP followed by a MAP.)

Signed-off-by: Peter Xu <peterx@redhat.com>
---
 hw/vfio/common.c              |  4 ++--
 include/exec/memory.h         | 47 ++++++++++++++++++++++++++++++++++++-------
 include/hw/vfio/vfio-common.h |  2 +-
 memory.c                      | 37 +++++++++++++++++++++++++---------
 4 files changed, 71 insertions(+), 19 deletions(-)

diff --git a/hw/vfio/common.c b/hw/vfio/common.c
index b313e7c..29188a1 100644
--- a/hw/vfio/common.c
+++ b/hw/vfio/common.c
@@ -293,11 +293,10 @@ static bool vfio_listener_skipped_section(MemoryRegionSection *section)
            section->offset_within_address_space & (1ULL << 63);
 }
 
-static void vfio_iommu_map_notify(Notifier *n, void *data)
+static void vfio_iommu_map_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
 {
     VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
     VFIOContainer *container = giommu->container;
-    IOMMUTLBEntry *iotlb = data;
     hwaddr iova = iotlb->iova + giommu->iommu_offset;
     MemoryRegion *mr;
     hwaddr xlat;
@@ -454,6 +453,7 @@ static void vfio_listener_region_add(MemoryListener *listener,
                                section->offset_within_region;
         giommu->container = container;
         giommu->n.notify = vfio_iommu_map_notify;
+        giommu->n.notifier_flags = IOMMU_NOTIFIER_ALL;
         QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
 
         memory_region_register_iommu_notifier(giommu->iommu, &giommu->n);
diff --git a/include/exec/memory.h b/include/exec/memory.h
index 3e4d416..14cda67 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -67,6 +67,27 @@ struct IOMMUTLBEntry {
     IOMMUAccessFlags perm;
 };
 
+/*
+ * Bitmap for different IOMMUNotifier capabilities. Each notifier can
+ * register with one or multiple IOMMU Notifier capability bit(s).
+ */
+typedef enum {
+    IOMMU_NOTIFIER_NONE = 0,
+    /* Notify cache invalidations */
+    IOMMU_NOTIFIER_UNMAP = 0x1,
+    /* Notify entry changes (newly created entries) */
+    IOMMU_NOTIFIER_MAP = 0x2,
+} IOMMUNotifierFlag;
+
+#define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
+
+struct IOMMUNotifier {
+    void (*notify)(struct IOMMUNotifier *notifier, IOMMUTLBEntry *data);
+    IOMMUNotifierFlag notifier_flags;
+    QLIST_ENTRY(IOMMUNotifier) node;
+};
+typedef struct IOMMUNotifier IOMMUNotifier;
+
 /* New-style MMIO accessors can indicate that the transaction failed.
  * A zero (MEMTX_OK) response means success; anything else is a failure
  * of some kind. The memory subsystem will bitwise-OR together results
@@ -201,7 +222,7 @@ struct MemoryRegion {
     const char *name;
     unsigned ioeventfd_nb;
     MemoryRegionIoeventfd *ioeventfds;
-    NotifierList iommu_notify;
+    QLIST_HEAD(, IOMMUNotifier) iommu_notify;
 };
 
 /**
@@ -607,6 +628,15 @@ uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr);
 /**
  * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
  *
+ * The notification type will be decided by entry.perm bits:
+ *
+ * - For UNMAP (cache invalidation) notifies: set entry.perm to IOMMU_NONE.
+ * - For MAP (newly added entry) notifies: set entry.perm to the
+ *   permission of the page (which is definitely !IOMMU_NONE).
+ *
+ * Note: for any IOMMU implementation, an in-place mapping change
+ * should be notified with an UNMAP followed by a MAP.
+ *
  * @mr: the memory region that was changed
  * @entry: the new entry in the IOMMU translation table.  The entry
  *         replaces all old entries for the same virtual I/O address range.
@@ -620,11 +650,12 @@ void memory_region_notify_iommu(MemoryRegion *mr,
  * IOMMU translation entries.
  *
  * @mr: the memory region to observe
- * @n: the notifier to be added; the notifier receives a pointer to an
- *     #IOMMUTLBEntry as the opaque value; the pointer ceases to be
- *     valid on exit from the notifier.
+ * @n: the IOMMUNotifier to be added; the notify callback receives a
+ *     pointer to an #IOMMUTLBEntry as the opaque value; the pointer
+ *     ceases to be valid on exit from the notifier.
  */
-void memory_region_register_iommu_notifier(MemoryRegion *mr, Notifier *n);
+void memory_region_register_iommu_notifier(MemoryRegion *mr,
+                                           IOMMUNotifier *n);
 
 /**
  * memory_region_iommu_replay: replay existing IOMMU translations to
@@ -636,7 +667,8 @@ void memory_region_register_iommu_notifier(MemoryRegion *mr, Notifier *n);
  * @is_write: Whether to treat the replay as a translate "write"
  *     through the iommu
  */
-void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write);
+void memory_region_iommu_replay(MemoryRegion *mr, IOMMUNotifier *n,
+                                bool is_write);
 
 /**
  * memory_region_unregister_iommu_notifier: unregister a notifier for
@@ -646,7 +678,8 @@ void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write);
  *      needs to be called
  * @n: the notifier to be removed.
  */
-void memory_region_unregister_iommu_notifier(MemoryRegion *mr, Notifier *n);
+void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
+                                             IOMMUNotifier *n);
 
 /**
  * memory_region_name: get a memory region's name
diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
index 94dfae3..c17602e 100644
--- a/include/hw/vfio/vfio-common.h
+++ b/include/hw/vfio/vfio-common.h
@@ -93,7 +93,7 @@ typedef struct VFIOGuestIOMMU {
     VFIOContainer *container;
     MemoryRegion *iommu;
     hwaddr iommu_offset;
-    Notifier n;
+    IOMMUNotifier n;
     QLIST_ENTRY(VFIOGuestIOMMU) giommu_next;
 } VFIOGuestIOMMU;
 
diff --git a/memory.c b/memory.c
index 1a1baf5..69d9d9a 100644
--- a/memory.c
+++ b/memory.c
@@ -1413,7 +1413,7 @@ void memory_region_init_iommu(MemoryRegion *mr,
     memory_region_init(mr, owner, name, size);
     mr->iommu_ops = ops,
     mr->terminates = true;  /* then re-forwards */
-    notifier_list_init(&mr->iommu_notify);
+    QLIST_INIT(&mr->iommu_notify);
 }
 
 static void memory_region_finalize(Object *obj)
@@ -1508,13 +1508,16 @@ bool memory_region_is_logging(MemoryRegion *mr, uint8_t client)
     return memory_region_get_dirty_log_mask(mr) & (1 << client);
 }
 
-void memory_region_register_iommu_notifier(MemoryRegion *mr, Notifier *n)
+void memory_region_register_iommu_notifier(MemoryRegion *mr,
+                                           IOMMUNotifier *n)
 {
+    /* We need to register for at least one bitfield */
+    assert(n->notifier_flags != IOMMU_NOTIFIER_NONE);
     if (mr->iommu_ops->notify_started &&
-        QLIST_EMPTY(&mr->iommu_notify.notifiers)) {
+        QLIST_EMPTY(&mr->iommu_notify)) {
         mr->iommu_ops->notify_started(mr);
     }
-    notifier_list_add(&mr->iommu_notify, n);
+    QLIST_INSERT_HEAD(&mr->iommu_notify, n, node);
 }
 
 uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr)
@@ -1526,7 +1529,8 @@ uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr)
     return TARGET_PAGE_SIZE;
 }
 
-void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write)
+void memory_region_iommu_replay(MemoryRegion *mr, IOMMUNotifier *n,
+                                bool is_write)
 {
     hwaddr addr, granularity;
     IOMMUTLBEntry iotlb;
@@ -1547,11 +1551,12 @@ void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write)
     }
 }
 
-void memory_region_unregister_iommu_notifier(MemoryRegion *mr, Notifier *n)
+void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
+                                             IOMMUNotifier *n)
 {
-    notifier_remove(n);
+    QLIST_REMOVE(n, node);
     if (mr->iommu_ops->notify_stopped &&
-        QLIST_EMPTY(&mr->iommu_notify.notifiers)) {
+        QLIST_EMPTY(&mr->iommu_notify)) {
         mr->iommu_ops->notify_stopped(mr);
     }
 }
@@ -1559,8 +1564,22 @@ void memory_region_unregister_iommu_notifier(MemoryRegion *mr, Notifier *n)
 void memory_region_notify_iommu(MemoryRegion *mr,
                                 IOMMUTLBEntry entry)
 {
+    IOMMUNotifier *iommu_notifier;
+    IOMMUNotifierFlag request_flags;
+
     assert(memory_region_is_iommu(mr));
-    notifier_list_notify(&mr->iommu_notify, &entry);
+
+    if (entry.perm & IOMMU_RW) {
+        request_flags = IOMMU_NOTIFIER_MAP;
+    } else {
+        request_flags = IOMMU_NOTIFIER_UNMAP;
+    }
+
+    QLIST_FOREACH(iommu_notifier, &mr->iommu_notify, node) {
+        if (iommu_notifier->notifier_flags & request_flags) {
+            iommu_notifier->notify(iommu_notifier, &entry);
+        }
+    }
 }
 
 void memory_region_set_log(MemoryRegion *mr, bool log, unsigned client)
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [Qemu-devel] [PATCH v7 2/3] memory: introduce IOMMUOps.notify_flag_changed
  2016-09-23  5:02 [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct Peter Xu
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps Peter Xu
@ 2016-09-23  5:02 ` Peter Xu
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 3/3] intel_iommu: allow UNMAP notifiers Peter Xu
  2016-09-23  7:22 ` [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct David Gibson
  3 siblings, 0 replies; 7+ messages in thread
From: Peter Xu @ 2016-09-23  5:02 UTC (permalink / raw)
  To: qemu-devel
  Cc: dgibson, jasowang, mst, pbonzini, cornelia.huck, wexu, vkaplans,
	alex.williamson, peterx

The new interface can be used to replace the old notify_started() and
notify_stopped(). Meanwhile it provides explicit flags so that IOMMUs
can know what kind of notifications it is requested for.

Acked-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Peter Xu <peterx@redhat.com>
---
 hw/i386/intel_iommu.c |  6 ++++--
 hw/ppc/spapr_iommu.c  | 18 ++++++++++--------
 include/exec/memory.h |  9 +++++----
 memory.c              | 29 +++++++++++++++++++++--------
 4 files changed, 40 insertions(+), 22 deletions(-)

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index 28c31a2..9d49be7 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -1974,7 +1974,9 @@ static IOMMUTLBEntry vtd_iommu_translate(MemoryRegion *iommu, hwaddr addr,
     return ret;
 }
 
-static void vtd_iommu_notify_started(MemoryRegion *iommu)
+static void vtd_iommu_notify_flag_changed(MemoryRegion *iommu,
+                                          IOMMUNotifierFlag old,
+                                          IOMMUNotifierFlag new)
 {
     VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
 
@@ -2348,7 +2350,7 @@ static void vtd_init(IntelIOMMUState *s)
     memset(s->womask, 0, DMAR_REG_SIZE);
 
     s->iommu_ops.translate = vtd_iommu_translate;
-    s->iommu_ops.notify_started = vtd_iommu_notify_started;
+    s->iommu_ops.notify_flag_changed = vtd_iommu_notify_flag_changed;
     s->root = 0;
     s->root_extended = false;
     s->dmar_enabled = false;
diff --git a/hw/ppc/spapr_iommu.c b/hw/ppc/spapr_iommu.c
index f20b0b8..ae30bbe 100644
--- a/hw/ppc/spapr_iommu.c
+++ b/hw/ppc/spapr_iommu.c
@@ -156,14 +156,17 @@ static uint64_t spapr_tce_get_min_page_size(MemoryRegion *iommu)
     return 1ULL << tcet->page_shift;
 }
 
-static void spapr_tce_notify_started(MemoryRegion *iommu)
+static void spapr_tce_notify_flag_changed(MemoryRegion *iommu,
+                                          IOMMUNotifierFlag old,
+                                          IOMMUNotifierFlag new)
 {
-    spapr_tce_set_need_vfio(container_of(iommu, sPAPRTCETable, iommu), true);
-}
+    struct sPAPRTCETable *tbl = container_of(iommu, sPAPRTCETable, iommu);
 
-static void spapr_tce_notify_stopped(MemoryRegion *iommu)
-{
-    spapr_tce_set_need_vfio(container_of(iommu, sPAPRTCETable, iommu), false);
+    if (old == IOMMU_NOTIFIER_NONE && new != IOMMU_NOTIFIER_NONE) {
+        spapr_tce_set_need_vfio(tbl, true);
+    } else if (old != IOMMU_NOTIFIER_NONE && new == IOMMU_NOTIFIER_NONE) {
+        spapr_tce_set_need_vfio(tbl, false);
+    }
 }
 
 static int spapr_tce_table_post_load(void *opaque, int version_id)
@@ -246,8 +249,7 @@ static const VMStateDescription vmstate_spapr_tce_table = {
 static MemoryRegionIOMMUOps spapr_iommu_ops = {
     .translate = spapr_tce_translate_iommu,
     .get_min_page_size = spapr_tce_get_min_page_size,
-    .notify_started = spapr_tce_notify_started,
-    .notify_stopped = spapr_tce_notify_stopped,
+    .notify_flag_changed = spapr_tce_notify_flag_changed,
 };
 
 static int spapr_tce_table_realize(DeviceState *dev)
diff --git a/include/exec/memory.h b/include/exec/memory.h
index 14cda67..a3f988b 100644
--- a/include/exec/memory.h
+++ b/include/exec/memory.h
@@ -174,10 +174,10 @@ struct MemoryRegionIOMMUOps {
     IOMMUTLBEntry (*translate)(MemoryRegion *iommu, hwaddr addr, bool is_write);
     /* Returns minimum supported page size */
     uint64_t (*get_min_page_size)(MemoryRegion *iommu);
-    /* Called when the first notifier is set */
-    void (*notify_started)(MemoryRegion *iommu);
-    /* Called when the last notifier is removed */
-    void (*notify_stopped)(MemoryRegion *iommu);
+    /* Called when IOMMU Notifier flag changed */
+    void (*notify_flag_changed)(MemoryRegion *iommu,
+                                IOMMUNotifierFlag old_flags,
+                                IOMMUNotifierFlag new_flags);
 };
 
 typedef struct CoalescedMemoryRange CoalescedMemoryRange;
@@ -223,6 +223,7 @@ struct MemoryRegion {
     unsigned ioeventfd_nb;
     MemoryRegionIoeventfd *ioeventfds;
     QLIST_HEAD(, IOMMUNotifier) iommu_notify;
+    IOMMUNotifierFlag iommu_notify_flags;
 };
 
 /**
diff --git a/memory.c b/memory.c
index 69d9d9a..27a3f2f 100644
--- a/memory.c
+++ b/memory.c
@@ -1414,6 +1414,7 @@ void memory_region_init_iommu(MemoryRegion *mr,
     mr->iommu_ops = ops,
     mr->terminates = true;  /* then re-forwards */
     QLIST_INIT(&mr->iommu_notify);
+    mr->iommu_notify_flags = IOMMU_NOTIFIER_NONE;
 }
 
 static void memory_region_finalize(Object *obj)
@@ -1508,16 +1509,31 @@ bool memory_region_is_logging(MemoryRegion *mr, uint8_t client)
     return memory_region_get_dirty_log_mask(mr) & (1 << client);
 }
 
+static void memory_region_update_iommu_notify_flags(MemoryRegion *mr)
+{
+    IOMMUNotifierFlag flags = IOMMU_NOTIFIER_NONE;
+    IOMMUNotifier *iommu_notifier;
+
+    QLIST_FOREACH(iommu_notifier, &mr->iommu_notify, node) {
+        flags |= iommu_notifier->notifier_flags;
+    }
+
+    if (flags != mr->iommu_notify_flags &&
+        mr->iommu_ops->notify_flag_changed) {
+        mr->iommu_ops->notify_flag_changed(mr, mr->iommu_notify_flags,
+                                           flags);
+    }
+
+    mr->iommu_notify_flags = flags;
+}
+
 void memory_region_register_iommu_notifier(MemoryRegion *mr,
                                            IOMMUNotifier *n)
 {
     /* We need to register for at least one bitfield */
     assert(n->notifier_flags != IOMMU_NOTIFIER_NONE);
-    if (mr->iommu_ops->notify_started &&
-        QLIST_EMPTY(&mr->iommu_notify)) {
-        mr->iommu_ops->notify_started(mr);
-    }
     QLIST_INSERT_HEAD(&mr->iommu_notify, n, node);
+    memory_region_update_iommu_notify_flags(mr);
 }
 
 uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr)
@@ -1555,10 +1571,7 @@ void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
                                              IOMMUNotifier *n)
 {
     QLIST_REMOVE(n, node);
-    if (mr->iommu_ops->notify_stopped &&
-        QLIST_EMPTY(&mr->iommu_notify)) {
-        mr->iommu_ops->notify_stopped(mr);
-    }
+    memory_region_update_iommu_notify_flags(mr);
 }
 
 void memory_region_notify_iommu(MemoryRegion *mr,
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 7+ messages in thread

* [Qemu-devel] [PATCH v7 3/3] intel_iommu: allow UNMAP notifiers
  2016-09-23  5:02 [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct Peter Xu
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps Peter Xu
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 2/3] memory: introduce IOMMUOps.notify_flag_changed Peter Xu
@ 2016-09-23  5:02 ` Peter Xu
  2016-09-23  7:22 ` [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct David Gibson
  3 siblings, 0 replies; 7+ messages in thread
From: Peter Xu @ 2016-09-23  5:02 UTC (permalink / raw)
  To: qemu-devel
  Cc: dgibson, jasowang, mst, pbonzini, cornelia.huck, wexu, vkaplans,
	alex.williamson, peterx

Intel vIOMMU is still lacking of a complete IOMMU notifier mechanism.
Before that is achieved, let's open a door for vhost DMAR support, which
only requires cache invalidations (UNMAP operations).

Meanwhile, converting hw_error() to error_report() and exit(1), to make
the error messages clean and obvious (so no CPU registers will be
dumped).

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Peter Xu <peterx@redhat.com>
---
 hw/i386/intel_iommu.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/hw/i386/intel_iommu.c b/hw/i386/intel_iommu.c
index 9d49be7..e4c3681 100644
--- a/hw/i386/intel_iommu.c
+++ b/hw/i386/intel_iommu.c
@@ -1980,10 +1980,14 @@ static void vtd_iommu_notify_flag_changed(MemoryRegion *iommu,
 {
     VTDAddressSpace *vtd_as = container_of(iommu, VTDAddressSpace, iommu);
 
-    hw_error("Device at bus %s addr %02x.%d requires iommu notifier which "
-             "is currently not supported by intel-iommu emulation",
-             vtd_as->bus->qbus.name, PCI_SLOT(vtd_as->devfn),
-             PCI_FUNC(vtd_as->devfn));
+    if (new & IOMMU_NOTIFIER_MAP) {
+        error_report("Device at bus %s addr %02x.%d requires iommu "
+                     "notifier which is currently not supported by "
+                     "intel-iommu emulation",
+                     vtd_as->bus->qbus.name, PCI_SLOT(vtd_as->devfn),
+                     PCI_FUNC(vtd_as->devfn));
+        exit(1);
+    }
 }
 
 static const VMStateDescription vtd_vmstate = {
-- 
2.7.4

^ permalink raw reply related	[flat|nested] 7+ messages in thread

* Re: [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps Peter Xu
@ 2016-09-23  7:20   ` David Gibson
  0 siblings, 0 replies; 7+ messages in thread
From: David Gibson @ 2016-09-23  7:20 UTC (permalink / raw)
  To: Peter Xu
  Cc: qemu-devel, mst, jasowang, vkaplans, alex.williamson, wexu,
	pbonzini, cornelia.huck, dgibson

[-- Attachment #1: Type: text/plain, Size: 10762 bytes --]

On Fri, Sep 23, 2016 at 01:02:26PM +0800, Peter Xu wrote:
> IOMMU Notifier list is used for notifying IO address mapping changes.
> Currently VFIO is the only user.
> 
> However it is possible that future consumer like vhost would like to
> only listen to part of its notifications (e.g., cache invalidations).
> 
> This patch introduced IOMMUNotifier and IOMMUNotfierFlag bits for a
> finer grained control of it.
> 
> IOMMUNotifier contains a bitfield for the notify consumer describing
> what kind of notification it is interested in. Currently two kinds of
> notifications are defined:
> 
> - IOMMU_NOTIFIER_MAP:    for newly mapped entries (additions)
> - IOMMU_NOTIFIER_UNMAP:  for entries to be removed (cache invalidates)
> 
> When registering the IOMMU notifier, we need to specify one or multiple
> types of messages to listen to.
> 
> When notifications are triggered, its type will be checked against the
> notifier's type bits, and only notifiers with registered bits will be
> notified.
> 
> (For any IOMMU implementation, an in-place mapping change should be
>  notified with an UNMAP followed by a MAP.)
> 
> Signed-off-by: Peter Xu <peterx@redhat.com>

Reviewed-by: David Gibson <david@gibson.dropbear.id.au>

> ---
>  hw/vfio/common.c              |  4 ++--
>  include/exec/memory.h         | 47 ++++++++++++++++++++++++++++++++++++-------
>  include/hw/vfio/vfio-common.h |  2 +-
>  memory.c                      | 37 +++++++++++++++++++++++++---------
>  4 files changed, 71 insertions(+), 19 deletions(-)
> 
> diff --git a/hw/vfio/common.c b/hw/vfio/common.c
> index b313e7c..29188a1 100644
> --- a/hw/vfio/common.c
> +++ b/hw/vfio/common.c
> @@ -293,11 +293,10 @@ static bool vfio_listener_skipped_section(MemoryRegionSection *section)
>             section->offset_within_address_space & (1ULL << 63);
>  }
>  
> -static void vfio_iommu_map_notify(Notifier *n, void *data)
> +static void vfio_iommu_map_notify(IOMMUNotifier *n, IOMMUTLBEntry *iotlb)
>  {
>      VFIOGuestIOMMU *giommu = container_of(n, VFIOGuestIOMMU, n);
>      VFIOContainer *container = giommu->container;
> -    IOMMUTLBEntry *iotlb = data;
>      hwaddr iova = iotlb->iova + giommu->iommu_offset;
>      MemoryRegion *mr;
>      hwaddr xlat;
> @@ -454,6 +453,7 @@ static void vfio_listener_region_add(MemoryListener *listener,
>                                 section->offset_within_region;
>          giommu->container = container;
>          giommu->n.notify = vfio_iommu_map_notify;
> +        giommu->n.notifier_flags = IOMMU_NOTIFIER_ALL;
>          QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next);
>  
>          memory_region_register_iommu_notifier(giommu->iommu, &giommu->n);
> diff --git a/include/exec/memory.h b/include/exec/memory.h
> index 3e4d416..14cda67 100644
> --- a/include/exec/memory.h
> +++ b/include/exec/memory.h
> @@ -67,6 +67,27 @@ struct IOMMUTLBEntry {
>      IOMMUAccessFlags perm;
>  };
>  
> +/*
> + * Bitmap for different IOMMUNotifier capabilities. Each notifier can
> + * register with one or multiple IOMMU Notifier capability bit(s).
> + */
> +typedef enum {
> +    IOMMU_NOTIFIER_NONE = 0,
> +    /* Notify cache invalidations */
> +    IOMMU_NOTIFIER_UNMAP = 0x1,
> +    /* Notify entry changes (newly created entries) */
> +    IOMMU_NOTIFIER_MAP = 0x2,
> +} IOMMUNotifierFlag;
> +
> +#define IOMMU_NOTIFIER_ALL (IOMMU_NOTIFIER_MAP | IOMMU_NOTIFIER_UNMAP)
> +
> +struct IOMMUNotifier {
> +    void (*notify)(struct IOMMUNotifier *notifier, IOMMUTLBEntry *data);
> +    IOMMUNotifierFlag notifier_flags;
> +    QLIST_ENTRY(IOMMUNotifier) node;
> +};
> +typedef struct IOMMUNotifier IOMMUNotifier;
> +
>  /* New-style MMIO accessors can indicate that the transaction failed.
>   * A zero (MEMTX_OK) response means success; anything else is a failure
>   * of some kind. The memory subsystem will bitwise-OR together results
> @@ -201,7 +222,7 @@ struct MemoryRegion {
>      const char *name;
>      unsigned ioeventfd_nb;
>      MemoryRegionIoeventfd *ioeventfds;
> -    NotifierList iommu_notify;
> +    QLIST_HEAD(, IOMMUNotifier) iommu_notify;
>  };
>  
>  /**
> @@ -607,6 +628,15 @@ uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr);
>  /**
>   * memory_region_notify_iommu: notify a change in an IOMMU translation entry.
>   *
> + * The notification type will be decided by entry.perm bits:
> + *
> + * - For UNMAP (cache invalidation) notifies: set entry.perm to IOMMU_NONE.
> + * - For MAP (newly added entry) notifies: set entry.perm to the
> + *   permission of the page (which is definitely !IOMMU_NONE).
> + *
> + * Note: for any IOMMU implementation, an in-place mapping change
> + * should be notified with an UNMAP followed by a MAP.
> + *
>   * @mr: the memory region that was changed
>   * @entry: the new entry in the IOMMU translation table.  The entry
>   *         replaces all old entries for the same virtual I/O address range.
> @@ -620,11 +650,12 @@ void memory_region_notify_iommu(MemoryRegion *mr,
>   * IOMMU translation entries.
>   *
>   * @mr: the memory region to observe
> - * @n: the notifier to be added; the notifier receives a pointer to an
> - *     #IOMMUTLBEntry as the opaque value; the pointer ceases to be
> - *     valid on exit from the notifier.
> + * @n: the IOMMUNotifier to be added; the notify callback receives a
> + *     pointer to an #IOMMUTLBEntry as the opaque value; the pointer
> + *     ceases to be valid on exit from the notifier.
>   */
> -void memory_region_register_iommu_notifier(MemoryRegion *mr, Notifier *n);
> +void memory_region_register_iommu_notifier(MemoryRegion *mr,
> +                                           IOMMUNotifier *n);
>  
>  /**
>   * memory_region_iommu_replay: replay existing IOMMU translations to
> @@ -636,7 +667,8 @@ void memory_region_register_iommu_notifier(MemoryRegion *mr, Notifier *n);
>   * @is_write: Whether to treat the replay as a translate "write"
>   *     through the iommu
>   */
> -void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write);
> +void memory_region_iommu_replay(MemoryRegion *mr, IOMMUNotifier *n,
> +                                bool is_write);
>  
>  /**
>   * memory_region_unregister_iommu_notifier: unregister a notifier for
> @@ -646,7 +678,8 @@ void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write);
>   *      needs to be called
>   * @n: the notifier to be removed.
>   */
> -void memory_region_unregister_iommu_notifier(MemoryRegion *mr, Notifier *n);
> +void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
> +                                             IOMMUNotifier *n);
>  
>  /**
>   * memory_region_name: get a memory region's name
> diff --git a/include/hw/vfio/vfio-common.h b/include/hw/vfio/vfio-common.h
> index 94dfae3..c17602e 100644
> --- a/include/hw/vfio/vfio-common.h
> +++ b/include/hw/vfio/vfio-common.h
> @@ -93,7 +93,7 @@ typedef struct VFIOGuestIOMMU {
>      VFIOContainer *container;
>      MemoryRegion *iommu;
>      hwaddr iommu_offset;
> -    Notifier n;
> +    IOMMUNotifier n;
>      QLIST_ENTRY(VFIOGuestIOMMU) giommu_next;
>  } VFIOGuestIOMMU;
>  
> diff --git a/memory.c b/memory.c
> index 1a1baf5..69d9d9a 100644
> --- a/memory.c
> +++ b/memory.c
> @@ -1413,7 +1413,7 @@ void memory_region_init_iommu(MemoryRegion *mr,
>      memory_region_init(mr, owner, name, size);
>      mr->iommu_ops = ops,
>      mr->terminates = true;  /* then re-forwards */
> -    notifier_list_init(&mr->iommu_notify);
> +    QLIST_INIT(&mr->iommu_notify);
>  }
>  
>  static void memory_region_finalize(Object *obj)
> @@ -1508,13 +1508,16 @@ bool memory_region_is_logging(MemoryRegion *mr, uint8_t client)
>      return memory_region_get_dirty_log_mask(mr) & (1 << client);
>  }
>  
> -void memory_region_register_iommu_notifier(MemoryRegion *mr, Notifier *n)
> +void memory_region_register_iommu_notifier(MemoryRegion *mr,
> +                                           IOMMUNotifier *n)
>  {
> +    /* We need to register for at least one bitfield */
> +    assert(n->notifier_flags != IOMMU_NOTIFIER_NONE);
>      if (mr->iommu_ops->notify_started &&
> -        QLIST_EMPTY(&mr->iommu_notify.notifiers)) {
> +        QLIST_EMPTY(&mr->iommu_notify)) {
>          mr->iommu_ops->notify_started(mr);
>      }
> -    notifier_list_add(&mr->iommu_notify, n);
> +    QLIST_INSERT_HEAD(&mr->iommu_notify, n, node);
>  }
>  
>  uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr)
> @@ -1526,7 +1529,8 @@ uint64_t memory_region_iommu_get_min_page_size(MemoryRegion *mr)
>      return TARGET_PAGE_SIZE;
>  }
>  
> -void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write)
> +void memory_region_iommu_replay(MemoryRegion *mr, IOMMUNotifier *n,
> +                                bool is_write)
>  {
>      hwaddr addr, granularity;
>      IOMMUTLBEntry iotlb;
> @@ -1547,11 +1551,12 @@ void memory_region_iommu_replay(MemoryRegion *mr, Notifier *n, bool is_write)
>      }
>  }
>  
> -void memory_region_unregister_iommu_notifier(MemoryRegion *mr, Notifier *n)
> +void memory_region_unregister_iommu_notifier(MemoryRegion *mr,
> +                                             IOMMUNotifier *n)
>  {
> -    notifier_remove(n);
> +    QLIST_REMOVE(n, node);
>      if (mr->iommu_ops->notify_stopped &&
> -        QLIST_EMPTY(&mr->iommu_notify.notifiers)) {
> +        QLIST_EMPTY(&mr->iommu_notify)) {
>          mr->iommu_ops->notify_stopped(mr);
>      }
>  }
> @@ -1559,8 +1564,22 @@ void memory_region_unregister_iommu_notifier(MemoryRegion *mr, Notifier *n)
>  void memory_region_notify_iommu(MemoryRegion *mr,
>                                  IOMMUTLBEntry entry)
>  {
> +    IOMMUNotifier *iommu_notifier;
> +    IOMMUNotifierFlag request_flags;
> +
>      assert(memory_region_is_iommu(mr));
> -    notifier_list_notify(&mr->iommu_notify, &entry);
> +
> +    if (entry.perm & IOMMU_RW) {
> +        request_flags = IOMMU_NOTIFIER_MAP;
> +    } else {
> +        request_flags = IOMMU_NOTIFIER_UNMAP;
> +    }
> +
> +    QLIST_FOREACH(iommu_notifier, &mr->iommu_notify, node) {
> +        if (iommu_notifier->notifier_flags & request_flags) {
> +            iommu_notifier->notify(iommu_notifier, &entry);
> +        }
> +    }
>  }
>  
>  void memory_region_set_log(MemoryRegion *mr, bool log, unsigned client)

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct
  2016-09-23  5:02 [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct Peter Xu
                   ` (2 preceding siblings ...)
  2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 3/3] intel_iommu: allow UNMAP notifiers Peter Xu
@ 2016-09-23  7:22 ` David Gibson
  2016-09-23  7:28   ` Peter Xu
  3 siblings, 1 reply; 7+ messages in thread
From: David Gibson @ 2016-09-23  7:22 UTC (permalink / raw)
  To: Peter Xu
  Cc: qemu-devel, mst, jasowang, vkaplans, alex.williamson, wexu,
	pbonzini, cornelia.huck, dgibson

[-- Attachment #1: Type: text/plain, Size: 2120 bytes --]

On Fri, Sep 23, 2016 at 01:02:25PM +0800, Peter Xu wrote:
> V7:
> - add comments to memory_region_notify_iommu() to better clarify the
>   interface [David]
> - vfio_iommu_map_notify(): remove pointless "IOMMUTLBEntry *iotlb =
>   data" [David]
> - typo fix on English [David]
> 
> V6:
> - use IOMMUTLBEntry directly for IOMMUNotifier.notify() [David]
> - add explicit comment for patch 1 that CHANGE should be treated as a
>   MAP then an UNMAP [David]
> - add a-b for David on patch 2
> 
> V5:
> - squash spapr_tce_notify_{started|stopped} into
>   spapr_tce_notify_flag_changed [David]
> - in spapr_tce_notify_flag_changed: check flags against "!=
>   IOMMU_NOTIFIER_NONE", but not "== IOMMU_NOTIFIER_ALL" [David]
> - add r-b for David on patch 3
> 
> V4:
> - change "notifier_caps" into "notifier_flags" [David]
> - rename IOMMU_NOTIFIER_{CHANGE|INVALIDATION} with MAP/UNMAP [David]
> - introduce IOMMUOps.notify_flag_changed, to replace notify_started
>   and notify_stopped [David, Paolo]
> 
> V3:
> - use QLIST instead of embedding Notifier into IOMMUNotifier [Paolo]
> - fix a build error for ppc64-softmmu
> 
> The idea originates from one of Alex's reply:
> 
>   https://lists.gnu.org/archive/html/qemu-devel/2016-09/msg00254.html
> 
> But after further discussions, it seems that only adding a simple type
> for notifier is not enough. This series introduced IOMMUNotifier
> struct to replace the old Notifier interface. Along with it, we can
> provide registration for one (or multiple) of the IOMMU notifications:
> 
> - cache invalidations
> - entry changes
> 
> This is a support material for Jason's vhost dmar patchset.
> 
> Please read commit messages for detailed information. Thanks,

I think we're ready to go with this.  I still think 3/3 is premature,
but I don't care that much.

Peter, thanks for your patience with my nitpicking of the interface.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 819 bytes --]

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct
  2016-09-23  7:22 ` [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct David Gibson
@ 2016-09-23  7:28   ` Peter Xu
  0 siblings, 0 replies; 7+ messages in thread
From: Peter Xu @ 2016-09-23  7:28 UTC (permalink / raw)
  To: David Gibson
  Cc: qemu-devel, mst, jasowang, vkaplans, alex.williamson, wexu,
	pbonzini, cornelia.huck, dgibson

On Fri, Sep 23, 2016 at 05:22:03PM +1000, David Gibson wrote:
> I think we're ready to go with this.  I still think 3/3 is premature,
> but I don't care that much.
> 
> Peter, thanks for your patience with my nitpicking of the interface.

My pleasure to have discussion with you on this. And (always) thank
you for helping review this work (apart from the heavy POWER tree
maintainance :).

-- peterx

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2016-09-23  7:28 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-09-23  5:02 [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct Peter Xu
2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 1/3] memory: introduce IOMMUNotifier and its caps Peter Xu
2016-09-23  7:20   ` David Gibson
2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 2/3] memory: introduce IOMMUOps.notify_flag_changed Peter Xu
2016-09-23  5:02 ` [Qemu-devel] [PATCH v7 3/3] intel_iommu: allow UNMAP notifiers Peter Xu
2016-09-23  7:22 ` [Qemu-devel] [PATCH v7 0/3] Introduce IOMMUNotifier struct David Gibson
2016-09-23  7:28   ` Peter Xu

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.