| Message ID | 1442495357-26547-7-git-send-email-david@gibson.dropbear.id.au |
|---|---|
| State | New |
| Headers | show |
On Thu, 2015-09-17 at 23:09 +1000, David Gibson wrote: > At present the memory listener used by vfio to keep host IOMMU mappings > in sync with the guest memory image assumes that if a guest IOMMU > appears, then it has no existing mappings. > > This may not be true if a VFIO device is hotplugged onto a guest bus > which didn't previously include a VFIO device, and which has existing > guest IOMMU mappings. > > Therefore, use the memory_region_register_iommu_notifier_replay() > function in order to fix this case, replaying existing guest IOMMU > mappings, bringing the host IOMMU into sync with the guest IOMMU. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- > hw/vfio/common.c | 34 +++++++++++++++++++--------------- > 1 file changed, 19 insertions(+), 15 deletions(-) > > diff --git a/hw/vfio/common.c b/hw/vfio/common.c > index daaac48..543c38e 100644 > --- a/hw/vfio/common.c > +++ b/hw/vfio/common.c > @@ -312,6 +312,22 @@ out: > rcu_read_unlock(); > } > > +static hwaddr vfio_container_granularity(VFIOContainer *container) > +{ > + uint64_t pgsize; > + > + assert(container->iommu_data.iova_pgsizes); return (hwaddr)1 << (ffsl(container->iommu_data.iova_pgsizes) - 1; ? > + > + /* Find the smallest page size supported by the IOMMU */ > + for (pgsize = 1; pgsize; pgsize <<= 1) { > + if (pgsize & container->iommu_data.iova_pgsizes) { > + return pgsize; > + } > + } > + /* Can't happen */ > + assert(0); > +} > + > static void vfio_listener_region_add(MemoryListener *listener, > MemoryRegionSection *section) > { > @@ -371,26 +387,14 @@ static void vfio_listener_region_add(MemoryListener *listener, > * would be the right place to wire that up (tell the KVM > * device emulation the VFIO iommu handles to use). > */ > - /* > - * This assumes that the guest IOMMU is empty of > - * mappings at this point. > - * > - * One way of doing this is: > - * 1. Avoid sharing IOMMUs between emulated devices or different > - * IOMMU groups. > - * 2. Implement VFIO_IOMMU_ENABLE in the host kernel to fail if > - * there are some mappings in IOMMU. > - * > - * VFIO on SPAPR does that. Other IOMMU models may do that different, > - * they must make sure there are no existing mappings or > - * loop through existing mappings to map them into VFIO. > - */ > giommu = g_malloc0(sizeof(*giommu)); > giommu->iommu = section->mr; > giommu->container = container; > giommu->n.notify = vfio_iommu_map_notify; > QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next); > - memory_region_register_iommu_notifier(giommu->iommu, &giommu->n); > + > + memory_region_register_iommu_notifier_replay(giommu->iommu, &giommu->n, > + vfio_container_granularity(container), false); > > return; > }
On Thu, Sep 17, 2015 at 10:54:24AM -0600, Alex Williamson wrote: > On Thu, 2015-09-17 at 23:09 +1000, David Gibson wrote: > > At present the memory listener used by vfio to keep host IOMMU mappings > > in sync with the guest memory image assumes that if a guest IOMMU > > appears, then it has no existing mappings. > > > > This may not be true if a VFIO device is hotplugged onto a guest bus > > which didn't previously include a VFIO device, and which has existing > > guest IOMMU mappings. > > > > Therefore, use the memory_region_register_iommu_notifier_replay() > > function in order to fix this case, replaying existing guest IOMMU > > mappings, bringing the host IOMMU into sync with the guest IOMMU. > > > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > > --- > > hw/vfio/common.c | 34 +++++++++++++++++++--------------- > > 1 file changed, 19 insertions(+), 15 deletions(-) > > > > diff --git a/hw/vfio/common.c b/hw/vfio/common.c > > index daaac48..543c38e 100644 > > --- a/hw/vfio/common.c > > +++ b/hw/vfio/common.c > > @@ -312,6 +312,22 @@ out: > > rcu_read_unlock(); > > } > > > > +static hwaddr vfio_container_granularity(VFIOContainer *container) > > +{ > > + uint64_t pgsize; > > + > > + assert(container->iommu_data.iova_pgsizes); > > return (hwaddr)1 << (ffsl(container->iommu_data.iova_pgsizes) - 1; Ah, yes, that should work. I didn't do it that way mostly because I tend to confuse myself when I try to remember exactly how ffs semantics work.
On 18/09/15 01:31, David Gibson wrote: > On Thu, Sep 17, 2015 at 10:54:24AM -0600, Alex Williamson wrote: >> On Thu, 2015-09-17 at 23:09 +1000, David Gibson wrote: >>> At present the memory listener used by vfio to keep host IOMMU mappings >>> in sync with the guest memory image assumes that if a guest IOMMU >>> appears, then it has no existing mappings. >>> >>> This may not be true if a VFIO device is hotplugged onto a guest bus >>> which didn't previously include a VFIO device, and which has existing >>> guest IOMMU mappings. >>> >>> Therefore, use the memory_region_register_iommu_notifier_replay() >>> function in order to fix this case, replaying existing guest IOMMU >>> mappings, bringing the host IOMMU into sync with the guest IOMMU. >>> >>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au> >>> --- >>> hw/vfio/common.c | 34 +++++++++++++++++++--------------- >>> 1 file changed, 19 insertions(+), 15 deletions(-) >>> >>> diff --git a/hw/vfio/common.c b/hw/vfio/common.c >>> index daaac48..543c38e 100644 >>> --- a/hw/vfio/common.c >>> +++ b/hw/vfio/common.c >>> @@ -312,6 +312,22 @@ out: >>> rcu_read_unlock(); >>> } >>> >>> +static hwaddr vfio_container_granularity(VFIOContainer *container) >>> +{ >>> + uint64_t pgsize; >>> + >>> + assert(container->iommu_data.iova_pgsizes); >> >> return (hwaddr)1 << (ffsl(container->iommu_data.iova_pgsizes) - 1; > > Ah, yes, that should work. I didn't do it that way mostly because I > tend to confuse myself when I try to remember exactly how ffs > semantics work. Maybe use ffsll instead of ffsl, in case the code ever runs on a 32-bit host instead of a 64-bit host ? Thomas
On 17/09/2015 15:09, David Gibson wrote: > At present the memory listener used by vfio to keep host IOMMU mappings > in sync with the guest memory image assumes that if a guest IOMMU > appears, then it has no existing mappings. > > This may not be true if a VFIO device is hotplugged onto a guest bus > which didn't previously include a VFIO device, and which has existing > guest IOMMU mappings. > > Therefore, use the memory_region_register_iommu_notifier_replay() > function in order to fix this case, replaying existing guest IOMMU > mappings, bringing the host IOMMU into sync with the guest IOMMU. > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > --- > hw/vfio/common.c | 34 +++++++++++++++++++--------------- > 1 file changed, 19 insertions(+), 15 deletions(-) > > diff --git a/hw/vfio/common.c b/hw/vfio/common.c > index daaac48..543c38e 100644 > --- a/hw/vfio/common.c > +++ b/hw/vfio/common.c > @@ -312,6 +312,22 @@ out: > rcu_read_unlock(); > } > > +static hwaddr vfio_container_granularity(VFIOContainer *container) > +{ > + uint64_t pgsize; > + > + assert(container->iommu_data.iova_pgsizes); > + > + /* Find the smallest page size supported by the IOMMU */ > + for (pgsize = 1; pgsize; pgsize <<= 1) { > + if (pgsize & container->iommu_data.iova_pgsizes) { > + return pgsize; > + } > + } Perhaps we can use gcc builtin ? return 1 << ctz64(container->iommu_data.iova_pgsizes); > + /* Can't happen */ > + assert(0); > +} > + > static void vfio_listener_region_add(MemoryListener *listener, > MemoryRegionSection *section) > { > @@ -371,26 +387,14 @@ static void vfio_listener_region_add(MemoryListener *listener, > * would be the right place to wire that up (tell the KVM > * device emulation the VFIO iommu handles to use). > */ > - /* > - * This assumes that the guest IOMMU is empty of > - * mappings at this point. > - * > - * One way of doing this is: > - * 1. Avoid sharing IOMMUs between emulated devices or different > - * IOMMU groups. > - * 2. Implement VFIO_IOMMU_ENABLE in the host kernel to fail if > - * there are some mappings in IOMMU. > - * > - * VFIO on SPAPR does that. Other IOMMU models may do that different, > - * they must make sure there are no existing mappings or > - * loop through existing mappings to map them into VFIO. > - */ > giommu = g_malloc0(sizeof(*giommu)); > giommu->iommu = section->mr; > giommu->container = container; > giommu->n.notify = vfio_iommu_map_notify; > QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next); > - memory_region_register_iommu_notifier(giommu->iommu, &giommu->n); > + > + memory_region_register_iommu_notifier_replay(giommu->iommu, &giommu->n, > + vfio_container_granularity(container), false); > > return; > } >
On Wed, Sep 23, 2015 at 01:02:32PM +0200, Thomas Huth wrote: > On 18/09/15 01:31, David Gibson wrote: > > On Thu, Sep 17, 2015 at 10:54:24AM -0600, Alex Williamson wrote: > >> On Thu, 2015-09-17 at 23:09 +1000, David Gibson wrote: > >>> At present the memory listener used by vfio to keep host IOMMU mappings > >>> in sync with the guest memory image assumes that if a guest IOMMU > >>> appears, then it has no existing mappings. > >>> > >>> This may not be true if a VFIO device is hotplugged onto a guest bus > >>> which didn't previously include a VFIO device, and which has existing > >>> guest IOMMU mappings. > >>> > >>> Therefore, use the memory_region_register_iommu_notifier_replay() > >>> function in order to fix this case, replaying existing guest IOMMU > >>> mappings, bringing the host IOMMU into sync with the guest IOMMU. > >>> > >>> Signed-off-by: David Gibson <david@gibson.dropbear.id.au> > >>> --- > >>> hw/vfio/common.c | 34 +++++++++++++++++++--------------- > >>> 1 file changed, 19 insertions(+), 15 deletions(-) > >>> > >>> diff --git a/hw/vfio/common.c b/hw/vfio/common.c > >>> index daaac48..543c38e 100644 > >>> --- a/hw/vfio/common.c > >>> +++ b/hw/vfio/common.c > >>> @@ -312,6 +312,22 @@ out: > >>> rcu_read_unlock(); > >>> } > >>> > >>> +static hwaddr vfio_container_granularity(VFIOContainer *container) > >>> +{ > >>> + uint64_t pgsize; > >>> + > >>> + assert(container->iommu_data.iova_pgsizes); > >> > >> return (hwaddr)1 << (ffsl(container->iommu_data.iova_pgsizes) - 1; > > > > Ah, yes, that should work. I didn't do it that way mostly because I > > tend to confuse myself when I try to remember exactly how ffs > > semantics work. > > Maybe use ffsll instead of ffsl, in case the code ever runs on a 32-bit > host instead of a 64-bit host ? Already done :)
diff --git a/hw/vfio/common.c b/hw/vfio/common.c index daaac48..543c38e 100644 --- a/hw/vfio/common.c +++ b/hw/vfio/common.c @@ -312,6 +312,22 @@ out: rcu_read_unlock(); } +static hwaddr vfio_container_granularity(VFIOContainer *container) +{ + uint64_t pgsize; + + assert(container->iommu_data.iova_pgsizes); + + /* Find the smallest page size supported by the IOMMU */ + for (pgsize = 1; pgsize; pgsize <<= 1) { + if (pgsize & container->iommu_data.iova_pgsizes) { + return pgsize; + } + } + /* Can't happen */ + assert(0); +} + static void vfio_listener_region_add(MemoryListener *listener, MemoryRegionSection *section) { @@ -371,26 +387,14 @@ static void vfio_listener_region_add(MemoryListener *listener, * would be the right place to wire that up (tell the KVM * device emulation the VFIO iommu handles to use). */ - /* - * This assumes that the guest IOMMU is empty of - * mappings at this point. - * - * One way of doing this is: - * 1. Avoid sharing IOMMUs between emulated devices or different - * IOMMU groups. - * 2. Implement VFIO_IOMMU_ENABLE in the host kernel to fail if - * there are some mappings in IOMMU. - * - * VFIO on SPAPR does that. Other IOMMU models may do that different, - * they must make sure there are no existing mappings or - * loop through existing mappings to map them into VFIO. - */ giommu = g_malloc0(sizeof(*giommu)); giommu->iommu = section->mr; giommu->container = container; giommu->n.notify = vfio_iommu_map_notify; QLIST_INSERT_HEAD(&container->giommu_list, giommu, giommu_next); - memory_region_register_iommu_notifier(giommu->iommu, &giommu->n); + + memory_region_register_iommu_notifier_replay(giommu->iommu, &giommu->n, + vfio_container_granularity(container), false); return; }
At present the memory listener used by vfio to keep host IOMMU mappings in sync with the guest memory image assumes that if a guest IOMMU appears, then it has no existing mappings. This may not be true if a VFIO device is hotplugged onto a guest bus which didn't previously include a VFIO device, and which has existing guest IOMMU mappings. Therefore, use the memory_region_register_iommu_notifier_replay() function in order to fix this case, replaying existing guest IOMMU mappings, bringing the host IOMMU into sync with the guest IOMMU. Signed-off-by: David Gibson <david@gibson.dropbear.id.au> --- hw/vfio/common.c | 34 +++++++++++++++++++--------------- 1 file changed, 19 insertions(+), 15 deletions(-)