| Message ID | 1426234057-16165-27-git-send-email-aik@ozlabs.ru (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Benjamin Herrenschmidt |
| Headers | show |
On Fri, 2015-03-13 at 19:07 +1100, Alexey Kardashevskiy wrote: > The existing IOMMU requires VFIO_IOMMU_ENABLE call to enable actual use > of the container (i.e. call DMA map/unmap) and this is where we check > the rlimit for locked pages. It assumes that only as much memory > as a default DMA window can be mapped. Every DMA map/unmap request will > do pinning/unpinning of physical pages. > > New IOMMU will split physical pages pinning and TCE table update. > It will require guest pages to be registered first and consequent > map/unmap requests to work only with pre-registered memory. > For the default single window case this means that the entire guest > (instead of 2GB) needs to be pinned before using VFIO. > However when a huge DMA window is added, no additional pinning will be > required, otherwise it would be guest RAM + 2GB. > > This advertises v2 SPAPR TCE IOMMU and restricts what the userspace > can do with v1 or v2 IOMMUs. > > Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> > --- > Changes: > v6: > * enforced limitations imposed by the SPAPR TCE IOMMU version > --- > drivers/vfio/vfio_iommu_spapr_tce.c | 18 +++++++++++++++++- > include/uapi/linux/vfio.h | 2 ++ > 2 files changed, 19 insertions(+), 1 deletion(-) > > diff --git a/drivers/vfio/vfio_iommu_spapr_tce.c b/drivers/vfio/vfio_iommu_spapr_tce.c > index 9d240b4..e191438 100644 > --- a/drivers/vfio/vfio_iommu_spapr_tce.c > +++ b/drivers/vfio/vfio_iommu_spapr_tce.c > @@ -95,6 +95,7 @@ struct tce_container { > bool enabled; > unsigned long locked_pages; > struct list_head mem_list; > + bool v2; > }; > > struct tce_memory { > @@ -398,7 +399,7 @@ static void *tce_iommu_open(unsigned long arg) > { > struct tce_container *container; > > - if (arg != VFIO_SPAPR_TCE_IOMMU) { > + if ((arg != VFIO_SPAPR_TCE_IOMMU) && (arg != VFIO_SPAPR_TCE_v2_IOMMU)) { > pr_err("tce_vfio: Wrong IOMMU type\n"); > return ERR_PTR(-EINVAL); > } > @@ -410,6 +411,8 @@ static void *tce_iommu_open(unsigned long arg) > mutex_init(&container->lock); > INIT_LIST_HEAD_RCU(&container->mem_list); > > + container->v2 = arg == VFIO_SPAPR_TCE_v2_IOMMU; > + > return container; > } > > @@ -580,6 +583,7 @@ static long tce_iommu_ioctl(void *iommu_data, > case VFIO_CHECK_EXTENSION: > switch (arg) { > case VFIO_SPAPR_TCE_IOMMU: > + case VFIO_SPAPR_TCE_v2_IOMMU: > ret = 1; > break; > default: > @@ -719,6 +723,9 @@ static long tce_iommu_ioctl(void *iommu_data, > case VFIO_IOMMU_SPAPR_REGISTER_MEMORY: { > struct vfio_iommu_spapr_register_memory param; > > + if (!container->v2) > + return -EPERM; > + > minsz = offsetofend(struct vfio_iommu_spapr_register_memory, > size); > > @@ -741,6 +748,9 @@ static long tce_iommu_ioctl(void *iommu_data, > case VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY: { > struct vfio_iommu_spapr_register_memory param; > > + if (!container->v2) > + return -EPERM; > + > minsz = offsetofend(struct vfio_iommu_spapr_register_memory, > size); > > @@ -761,6 +771,9 @@ static long tce_iommu_ioctl(void *iommu_data, > return 0; > } > case VFIO_IOMMU_ENABLE: > + if (container->v2) > + return -EPERM; > + > mutex_lock(&container->lock); > ret = tce_iommu_enable(container); > mutex_unlock(&container->lock); > @@ -768,6 +781,9 @@ static long tce_iommu_ioctl(void *iommu_data, > > > case VFIO_IOMMU_DISABLE: > + if (container->v2) > + return -EPERM; > + > mutex_lock(&container->lock); > tce_iommu_disable(container); > mutex_unlock(&container->lock); I wouldn't have guessed; nothing in the documentation suggests these ioctls are deprecated in v2 (ie. please document). If the ioctl doesn't exist for the IOMMU type, why not simply break and let it fall out at -ENOTTY? Same for the above, v1 would have previously returned -ENOTTY for those ioctls, why change to -EPERM? > diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h > index b17e120..fbc5286 100644 > --- a/include/uapi/linux/vfio.h > +++ b/include/uapi/linux/vfio.h > @@ -36,6 +36,8 @@ > /* Two-stage IOMMU */ > #define VFIO_TYPE1_NESTING_IOMMU 6 /* Implies v2 */ > > +#define VFIO_SPAPR_TCE_v2_IOMMU 7 > + > /* > * The IOCTL interface is designed for extensibility by embedding the > * structure length (argsz) and flags into structures passed between
On 03/17/2015 06:45 AM, Alex Williamson wrote: > On Fri, 2015-03-13 at 19:07 +1100, Alexey Kardashevskiy wrote: >> The existing IOMMU requires VFIO_IOMMU_ENABLE call to enable actual use >> of the container (i.e. call DMA map/unmap) and this is where we check >> the rlimit for locked pages. It assumes that only as much memory >> as a default DMA window can be mapped. Every DMA map/unmap request will >> do pinning/unpinning of physical pages. >> >> New IOMMU will split physical pages pinning and TCE table update. >> It will require guest pages to be registered first and consequent >> map/unmap requests to work only with pre-registered memory. >> For the default single window case this means that the entire guest >> (instead of 2GB) needs to be pinned before using VFIO. >> However when a huge DMA window is added, no additional pinning will be >> required, otherwise it would be guest RAM + 2GB. >> >> This advertises v2 SPAPR TCE IOMMU and restricts what the userspace >> can do with v1 or v2 IOMMUs. >> >> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> >> --- >> Changes: >> v6: >> * enforced limitations imposed by the SPAPR TCE IOMMU version >> --- >> drivers/vfio/vfio_iommu_spapr_tce.c | 18 +++++++++++++++++- >> include/uapi/linux/vfio.h | 2 ++ >> 2 files changed, 19 insertions(+), 1 deletion(-) >> >> diff --git a/drivers/vfio/vfio_iommu_spapr_tce.c b/drivers/vfio/vfio_iommu_spapr_tce.c >> index 9d240b4..e191438 100644 >> --- a/drivers/vfio/vfio_iommu_spapr_tce.c >> +++ b/drivers/vfio/vfio_iommu_spapr_tce.c >> @@ -95,6 +95,7 @@ struct tce_container { >> bool enabled; >> unsigned long locked_pages; >> struct list_head mem_list; >> + bool v2; >> }; >> >> struct tce_memory { >> @@ -398,7 +399,7 @@ static void *tce_iommu_open(unsigned long arg) >> { >> struct tce_container *container; >> >> - if (arg != VFIO_SPAPR_TCE_IOMMU) { >> + if ((arg != VFIO_SPAPR_TCE_IOMMU) && (arg != VFIO_SPAPR_TCE_v2_IOMMU)) { >> pr_err("tce_vfio: Wrong IOMMU type\n"); >> return ERR_PTR(-EINVAL); >> } >> @@ -410,6 +411,8 @@ static void *tce_iommu_open(unsigned long arg) >> mutex_init(&container->lock); >> INIT_LIST_HEAD_RCU(&container->mem_list); >> >> + container->v2 = arg == VFIO_SPAPR_TCE_v2_IOMMU; >> + >> return container; >> } >> >> @@ -580,6 +583,7 @@ static long tce_iommu_ioctl(void *iommu_data, >> case VFIO_CHECK_EXTENSION: >> switch (arg) { >> case VFIO_SPAPR_TCE_IOMMU: >> + case VFIO_SPAPR_TCE_v2_IOMMU: >> ret = 1; >> break; >> default: >> @@ -719,6 +723,9 @@ static long tce_iommu_ioctl(void *iommu_data, >> case VFIO_IOMMU_SPAPR_REGISTER_MEMORY: { >> struct vfio_iommu_spapr_register_memory param; >> >> + if (!container->v2) >> + return -EPERM; >> + >> minsz = offsetofend(struct vfio_iommu_spapr_register_memory, >> size); >> >> @@ -741,6 +748,9 @@ static long tce_iommu_ioctl(void *iommu_data, >> case VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY: { >> struct vfio_iommu_spapr_register_memory param; >> >> + if (!container->v2) >> + return -EPERM; >> + >> minsz = offsetofend(struct vfio_iommu_spapr_register_memory, >> size); >> >> @@ -761,6 +771,9 @@ static long tce_iommu_ioctl(void *iommu_data, >> return 0; >> } >> case VFIO_IOMMU_ENABLE: >> + if (container->v2) >> + return -EPERM; >> + >> mutex_lock(&container->lock); >> ret = tce_iommu_enable(container); >> mutex_unlock(&container->lock); >> @@ -768,6 +781,9 @@ static long tce_iommu_ioctl(void *iommu_data, >> >> >> case VFIO_IOMMU_DISABLE: >> + if (container->v2) >> + return -EPERM; >> + >> mutex_lock(&container->lock); >> tce_iommu_disable(container); >> mutex_unlock(&container->lock); > > > I wouldn't have guessed; nothing in the documentation suggests these > ioctls are deprecated in v2 (ie. please document). If the ioctl doesn't > exist for the IOMMU type, why not simply break and let it fall out at > -ENOTTY? Same for the above, v1 would have previously returned -ENOTTY > for those ioctls, why change to -EPERM? Good points. I'll fix them and merge this patch to "vfio: powerpc/spapr: Register memory" as this is where it actually belongs to. Agree? Thanks for the review!
diff --git a/drivers/vfio/vfio_iommu_spapr_tce.c b/drivers/vfio/vfio_iommu_spapr_tce.c index 9d240b4..e191438 100644 --- a/drivers/vfio/vfio_iommu_spapr_tce.c +++ b/drivers/vfio/vfio_iommu_spapr_tce.c @@ -95,6 +95,7 @@ struct tce_container { bool enabled; unsigned long locked_pages; struct list_head mem_list; + bool v2; }; struct tce_memory { @@ -398,7 +399,7 @@ static void *tce_iommu_open(unsigned long arg) { struct tce_container *container; - if (arg != VFIO_SPAPR_TCE_IOMMU) { + if ((arg != VFIO_SPAPR_TCE_IOMMU) && (arg != VFIO_SPAPR_TCE_v2_IOMMU)) { pr_err("tce_vfio: Wrong IOMMU type\n"); return ERR_PTR(-EINVAL); } @@ -410,6 +411,8 @@ static void *tce_iommu_open(unsigned long arg) mutex_init(&container->lock); INIT_LIST_HEAD_RCU(&container->mem_list); + container->v2 = arg == VFIO_SPAPR_TCE_v2_IOMMU; + return container; } @@ -580,6 +583,7 @@ static long tce_iommu_ioctl(void *iommu_data, case VFIO_CHECK_EXTENSION: switch (arg) { case VFIO_SPAPR_TCE_IOMMU: + case VFIO_SPAPR_TCE_v2_IOMMU: ret = 1; break; default: @@ -719,6 +723,9 @@ static long tce_iommu_ioctl(void *iommu_data, case VFIO_IOMMU_SPAPR_REGISTER_MEMORY: { struct vfio_iommu_spapr_register_memory param; + if (!container->v2) + return -EPERM; + minsz = offsetofend(struct vfio_iommu_spapr_register_memory, size); @@ -741,6 +748,9 @@ static long tce_iommu_ioctl(void *iommu_data, case VFIO_IOMMU_SPAPR_UNREGISTER_MEMORY: { struct vfio_iommu_spapr_register_memory param; + if (!container->v2) + return -EPERM; + minsz = offsetofend(struct vfio_iommu_spapr_register_memory, size); @@ -761,6 +771,9 @@ static long tce_iommu_ioctl(void *iommu_data, return 0; } case VFIO_IOMMU_ENABLE: + if (container->v2) + return -EPERM; + mutex_lock(&container->lock); ret = tce_iommu_enable(container); mutex_unlock(&container->lock); @@ -768,6 +781,9 @@ static long tce_iommu_ioctl(void *iommu_data, case VFIO_IOMMU_DISABLE: + if (container->v2) + return -EPERM; + mutex_lock(&container->lock); tce_iommu_disable(container); mutex_unlock(&container->lock); diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h index b17e120..fbc5286 100644 --- a/include/uapi/linux/vfio.h +++ b/include/uapi/linux/vfio.h @@ -36,6 +36,8 @@ /* Two-stage IOMMU */ #define VFIO_TYPE1_NESTING_IOMMU 6 /* Implies v2 */ +#define VFIO_SPAPR_TCE_v2_IOMMU 7 + /* * The IOCTL interface is designed for extensibility by embedding the * structure length (argsz) and flags into structures passed between
The existing IOMMU requires VFIO_IOMMU_ENABLE call to enable actual use of the container (i.e. call DMA map/unmap) and this is where we check the rlimit for locked pages. It assumes that only as much memory as a default DMA window can be mapped. Every DMA map/unmap request will do pinning/unpinning of physical pages. New IOMMU will split physical pages pinning and TCE table update. It will require guest pages to be registered first and consequent map/unmap requests to work only with pre-registered memory. For the default single window case this means that the entire guest (instead of 2GB) needs to be pinned before using VFIO. However when a huge DMA window is added, no additional pinning will be required, otherwise it would be guest RAM + 2GB. This advertises v2 SPAPR TCE IOMMU and restricts what the userspace can do with v1 or v2 IOMMUs. Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru> --- Changes: v6: * enforced limitations imposed by the SPAPR TCE IOMMU version --- drivers/vfio/vfio_iommu_spapr_tce.c | 18 +++++++++++++++++- include/uapi/linux/vfio.h | 2 ++ 2 files changed, 19 insertions(+), 1 deletion(-)