Message ID | 20210219094230.231-3-jiangkunkun@huawei.com |
---|---|
State | New |
Headers | show |
Series | Add migration support for VFIO PCI devices in SMMUv3 nested stage mode | expand |
Hi Kunkun, On 2/19/21 10:42 AM, Kunkun Jiang wrote: > On Intel, the DMA mapped through the host single stage. Instead > we set up the stage 2 and stage 1 separately in nested mode as there > is no "Caching Mode". You need to rewrite the above sentences, Missing ARM and also the 1st sentences misses a verb. > > Legacy vfio_listener_log_sync cannot be used in nested stage as we > don't need to pay close attention to stage 1 mapping. This patch adds > vfio_prereg_listener_log_sync to mark dirty pages in nested mode. > > Signed-off-by: Kunkun Jiang <jiangkunkun@huawei.com> > --- > hw/vfio/common.c | 25 +++++++++++++++++++++++++ > 1 file changed, 25 insertions(+) > > diff --git a/hw/vfio/common.c b/hw/vfio/common.c > index 7c50905856..af333e0dee 100644 > --- a/hw/vfio/common.c > +++ b/hw/vfio/common.c > @@ -1216,6 +1216,22 @@ static int vfio_dma_sync_ram_section_dirty_bitmap(VFIOContainer *container, > int128_get64(section->size), ram_addr); > } > > +static void vfio_prereg_listener_log_sync(MemoryListener *listener, > + MemoryRegionSection *section) > +{ > + VFIOContainer *container = > + container_of(listener, VFIOContainer, prereg_listener); > + > + if (!memory_region_is_ram(section->mr) || > + !container->dirty_pages_supported) { > + return; > + } > + > + if (vfio_devices_all_saving(container)) { I fail to see where is this defined? > + vfio_dma_sync_ram_section_dirty_bitmap(container, section); > + } > +} > + > typedef struct { > IOMMUNotifier n; > VFIOGuestIOMMU *giommu; > @@ -1260,6 +1276,14 @@ static int vfio_sync_dirty_bitmap(VFIOContainer *container, > if (memory_region_is_iommu(section->mr)) { > VFIOGuestIOMMU *giommu; > > + /* > + * In nested mode, stage 2 and stage 1 are set up separately. We > + * only need to focus on stage 2 mapping when marking dirty pages. > + */ > + if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) { > + return 0; > + } > + > QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) { > if (MEMORY_REGION(giommu->iommu) == section->mr && > giommu->n.start == section->offset_within_region) { > @@ -1312,6 +1336,7 @@ static const MemoryListener vfio_memory_listener = { > static MemoryListener vfio_memory_prereg_listener = { > .region_add = vfio_prereg_listener_region_add, > .region_del = vfio_prereg_listener_region_del, > + .log_sync = vfio_prereg_listener_log_sync, > }; > > static void vfio_listener_release(VFIOContainer *container) > Thanks Eric
Hi Eric, On 2021/4/8 21:56, Auger Eric wrote: > Hi Kunkun, > > On 2/19/21 10:42 AM, Kunkun Jiang wrote: >> On Intel, the DMA mapped through the host single stage. Instead >> we set up the stage 2 and stage 1 separately in nested mode as there >> is no "Caching Mode". > You need to rewrite the above sentences, Missing ARM and also the 1st > sentences misses a verb. Thanks for your review! I will fix it in the next version. >> Legacy vfio_listener_log_sync cannot be used in nested stage as we >> don't need to pay close attention to stage 1 mapping. This patch adds >> vfio_prereg_listener_log_sync to mark dirty pages in nested mode. >> >> Signed-off-by: Kunkun Jiang <jiangkunkun@huawei.com> >> --- >> hw/vfio/common.c | 25 +++++++++++++++++++++++++ >> 1 file changed, 25 insertions(+) >> >> diff --git a/hw/vfio/common.c b/hw/vfio/common.c >> index 7c50905856..af333e0dee 100644 >> --- a/hw/vfio/common.c >> +++ b/hw/vfio/common.c >> @@ -1216,6 +1216,22 @@ static int vfio_dma_sync_ram_section_dirty_bitmap(VFIOContainer *container, >> int128_get64(section->size), ram_addr); >> } >> >> +static void vfio_prereg_listener_log_sync(MemoryListener *listener, >> + MemoryRegionSection *section) >> +{ >> + VFIOContainer *container = >> + container_of(listener, VFIOContainer, prereg_listener); >> + >> + if (!memory_region_is_ram(section->mr) || >> + !container->dirty_pages_supported) { >> + return; >> + } >> + >> + if (vfio_devices_all_saving(container)) { > I fail to see where is this defined? Keqian modified vfio_devices_all_saving to vfio_devices_all_dirty_tracking in 758b96b61d5. When I posted this series patches, it was vfio_devices_all_saving. In v2[1], I have updated it based on the lasted qemu. [1] https://lore.kernel.org/qemu-devel/20210331101259.2153-3-jiangkunkun@huawei.com/ Thanks, Kunkun Jiang >> + vfio_dma_sync_ram_section_dirty_bitmap(container, section); >> + } >> +} >> + >> typedef struct { >> IOMMUNotifier n; >> VFIOGuestIOMMU *giommu; >> @@ -1260,6 +1276,14 @@ static int vfio_sync_dirty_bitmap(VFIOContainer *container, >> if (memory_region_is_iommu(section->mr)) { >> VFIOGuestIOMMU *giommu; >> >> + /* >> + * In nested mode, stage 2 and stage 1 are set up separately. We >> + * only need to focus on stage 2 mapping when marking dirty pages. >> + */ >> + if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) { >> + return 0; >> + } >> + >> QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) { >> if (MEMORY_REGION(giommu->iommu) == section->mr && >> giommu->n.start == section->offset_within_region) { >> @@ -1312,6 +1336,7 @@ static const MemoryListener vfio_memory_listener = { >> static MemoryListener vfio_memory_prereg_listener = { >> .region_add = vfio_prereg_listener_region_add, >> .region_del = vfio_prereg_listener_region_del, >> + .log_sync = vfio_prereg_listener_log_sync, >> }; >> >> static void vfio_listener_release(VFIOContainer *container) >> > Thanks > > Eric > > .
diff --git a/hw/vfio/common.c b/hw/vfio/common.c index 7c50905856..af333e0dee 100644 --- a/hw/vfio/common.c +++ b/hw/vfio/common.c @@ -1216,6 +1216,22 @@ static int vfio_dma_sync_ram_section_dirty_bitmap(VFIOContainer *container, int128_get64(section->size), ram_addr); } +static void vfio_prereg_listener_log_sync(MemoryListener *listener, + MemoryRegionSection *section) +{ + VFIOContainer *container = + container_of(listener, VFIOContainer, prereg_listener); + + if (!memory_region_is_ram(section->mr) || + !container->dirty_pages_supported) { + return; + } + + if (vfio_devices_all_saving(container)) { + vfio_dma_sync_ram_section_dirty_bitmap(container, section); + } +} + typedef struct { IOMMUNotifier n; VFIOGuestIOMMU *giommu; @@ -1260,6 +1276,14 @@ static int vfio_sync_dirty_bitmap(VFIOContainer *container, if (memory_region_is_iommu(section->mr)) { VFIOGuestIOMMU *giommu; + /* + * In nested mode, stage 2 and stage 1 are set up separately. We + * only need to focus on stage 2 mapping when marking dirty pages. + */ + if (container->iommu_type == VFIO_TYPE1_NESTING_IOMMU) { + return 0; + } + QLIST_FOREACH(giommu, &container->giommu_list, giommu_next) { if (MEMORY_REGION(giommu->iommu) == section->mr && giommu->n.start == section->offset_within_region) { @@ -1312,6 +1336,7 @@ static const MemoryListener vfio_memory_listener = { static MemoryListener vfio_memory_prereg_listener = { .region_add = vfio_prereg_listener_region_add, .region_del = vfio_prereg_listener_region_del, + .log_sync = vfio_prereg_listener_log_sync, }; static void vfio_listener_release(VFIOContainer *container)
On Intel, the DMA mapped through the host single stage. Instead we set up the stage 2 and stage 1 separately in nested mode as there is no "Caching Mode". Legacy vfio_listener_log_sync cannot be used in nested stage as we don't need to pay close attention to stage 1 mapping. This patch adds vfio_prereg_listener_log_sync to mark dirty pages in nested mode. Signed-off-by: Kunkun Jiang <jiangkunkun@huawei.com> --- hw/vfio/common.c | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+)