| Message ID | 1671141424-81853-5-git-send-email-steven.sistare@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | fixes for virtual address update | expand |
On Thu, Dec 15, 2022 at 01:57:01PM -0800, Steve Sistare wrote: > When a vfio container is preserved across exec or fork-exec, the new > task's mm has a locked_vm count of 0. After a dma vaddr is updated using > VFIO_DMA_MAP_FLAG_VADDR, locked_vm remains 0, and the pinned memory does > not count against the task's RLIMIT_MEMLOCK. > > To restore the correct locked_vm count, when VFIO_DMA_MAP_FLAG_VADDR is > used and the dma's mm has changed, add the mapping's pinned page count to > the new mm->locked_vm, subject to the rlimit. Now that mediated devices > are excluded when using VFIO_UPDATE_VADDR, the amount of pinned memory > equals the size of the mapping less the reserved page count. > > Fixes: c3cbab24db38 ("vfio/type1: implement interfaces to update vaddr") > > Signed-off-by: Steve Sistare <steven.sistare@oracle.com> > --- > drivers/vfio/vfio_iommu_type1.c | 35 +++++++++++++++++++++++++++++++++++ > 1 file changed, 35 insertions(+) > > diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c > index add87cd..70b52e9 100644 > --- a/drivers/vfio/vfio_iommu_type1.c > +++ b/drivers/vfio/vfio_iommu_type1.c > @@ -1588,6 +1588,38 @@ static bool vfio_iommu_iova_dma_valid(struct vfio_iommu *iommu, > return list_empty(iova); > } > > +static int vfio_change_dma_owner(struct vfio_dma *dma) > +{ > + struct task_struct *new_task = current->group_leader; > + > + if (new_task->mm != dma->mm) { > + long npage = (dma->size >> PAGE_SHIFT) - dma->reserved_pages; > + bool new_lock_cap = capable(CAP_IPC_LOCK); > + int ret = mmap_write_lock_killable(new_task->mm); > + > + if (ret) > + return ret; > + > + ret = __account_locked_vm(new_task->mm, npage, true, > + new_task, new_lock_cap); > + mmap_write_unlock(new_task->mm); > + if (ret) > + return ret; > + > + vfio_lock_acct(dma, -npage, true); > + if (dma->task != new_task) { > + put_task_struct(dma->task); > + dma->task = get_task_struct(new_task); > + } > + mmdrop(dma->mm); > + dma->mm = new_task->mm; This also should be current->mm not current->group_leader->mm Jason
diff --git a/drivers/vfio/vfio_iommu_type1.c b/drivers/vfio/vfio_iommu_type1.c index add87cd..70b52e9 100644 --- a/drivers/vfio/vfio_iommu_type1.c +++ b/drivers/vfio/vfio_iommu_type1.c @@ -1588,6 +1588,38 @@ static bool vfio_iommu_iova_dma_valid(struct vfio_iommu *iommu, return list_empty(iova); } +static int vfio_change_dma_owner(struct vfio_dma *dma) +{ + struct task_struct *new_task = current->group_leader; + + if (new_task->mm != dma->mm) { + long npage = (dma->size >> PAGE_SHIFT) - dma->reserved_pages; + bool new_lock_cap = capable(CAP_IPC_LOCK); + int ret = mmap_write_lock_killable(new_task->mm); + + if (ret) + return ret; + + ret = __account_locked_vm(new_task->mm, npage, true, + new_task, new_lock_cap); + mmap_write_unlock(new_task->mm); + if (ret) + return ret; + + vfio_lock_acct(dma, -npage, true); + if (dma->task != new_task) { + put_task_struct(dma->task); + dma->task = get_task_struct(new_task); + } + mmdrop(dma->mm); + dma->mm = new_task->mm; + mmgrab(dma->mm); + dma->lock_cap = new_lock_cap; + } + + return 0; +} + static int vfio_dma_do_map(struct vfio_iommu *iommu, struct vfio_iommu_type1_dma_map *map) { @@ -1637,6 +1669,9 @@ static int vfio_dma_do_map(struct vfio_iommu *iommu, dma->size != size) { ret = -EINVAL; } else { + ret = vfio_change_dma_owner(dma); + if (ret) + goto out_unlock; dma->vaddr = vaddr; dma->vaddr_invalid = false; iommu->vaddr_invalid_count--;
When a vfio container is preserved across exec or fork-exec, the new task's mm has a locked_vm count of 0. After a dma vaddr is updated using VFIO_DMA_MAP_FLAG_VADDR, locked_vm remains 0, and the pinned memory does not count against the task's RLIMIT_MEMLOCK. To restore the correct locked_vm count, when VFIO_DMA_MAP_FLAG_VADDR is used and the dma's mm has changed, add the mapping's pinned page count to the new mm->locked_vm, subject to the rlimit. Now that mediated devices are excluded when using VFIO_UPDATE_VADDR, the amount of pinned memory equals the size of the mapping less the reserved page count. Fixes: c3cbab24db38 ("vfio/type1: implement interfaces to update vaddr") Signed-off-by: Steve Sistare <steven.sistare@oracle.com> --- drivers/vfio/vfio_iommu_type1.c | 35 +++++++++++++++++++++++++++++++++++ 1 file changed, 35 insertions(+)