Jason Wang
2022-Jul-20 09:08 UTC
[PATCH v3 4/5] vduse: Support registering userspace memory for IOVA regions
On Wed, Jul 20, 2022 at 12:42 PM Xie Yongji <xieyongji at bytedance.com> wrote:> > Introduce two ioctls: VDUSE_IOTLB_REG_UMEM and > VDUSE_IOTLB_DEREG_UMEM to support registering > and de-registering userspace memory for IOVA > regions. > > Now it only supports registering userspace memory > for bounce buffer region in virtio-vdpa case. > > Signed-off-by: Xie Yongji <xieyongji at bytedance.com>Acked-by: Jason Wang <jasowang at redhat.com>> --- > drivers/vdpa/vdpa_user/vduse_dev.c | 141 +++++++++++++++++++++++++++++ > include/uapi/linux/vduse.h | 23 +++++ > 2 files changed, 164 insertions(+) > > diff --git a/drivers/vdpa/vdpa_user/vduse_dev.c b/drivers/vdpa/vdpa_user/vduse_dev.c > index 3bc27de58f46..eedff0a3885a 100644 > --- a/drivers/vdpa/vdpa_user/vduse_dev.c > +++ b/drivers/vdpa/vdpa_user/vduse_dev.c > @@ -21,6 +21,8 @@ > #include <linux/uio.h> > #include <linux/vdpa.h> > #include <linux/nospec.h> > +#include <linux/vmalloc.h> > +#include <linux/sched/mm.h> > #include <uapi/linux/vduse.h> > #include <uapi/linux/vdpa.h> > #include <uapi/linux/virtio_config.h> > @@ -64,6 +66,13 @@ struct vduse_vdpa { > struct vduse_dev *dev; > }; > > +struct vduse_umem { > + unsigned long iova; > + unsigned long npages; > + struct page **pages; > + struct mm_struct *mm; > +}; > + > struct vduse_dev { > struct vduse_vdpa *vdev; > struct device *dev; > @@ -95,6 +104,8 @@ struct vduse_dev { > u8 status; > u32 vq_num; > u32 vq_align; > + struct vduse_umem *umem; > + struct mutex mem_lock; > }; > > struct vduse_dev_msg { > @@ -917,6 +928,102 @@ static int vduse_dev_queue_irq_work(struct vduse_dev *dev, > return ret; > } > > +static int vduse_dev_dereg_umem(struct vduse_dev *dev, > + u64 iova, u64 size) > +{ > + int ret; > + > + mutex_lock(&dev->mem_lock); > + ret = -ENOENT; > + if (!dev->umem) > + goto unlock; > + > + ret = -EINVAL; > + if (dev->umem->iova != iova || size != dev->domain->bounce_size) > + goto unlock; > + > + vduse_domain_remove_user_bounce_pages(dev->domain); > + unpin_user_pages_dirty_lock(dev->umem->pages, > + dev->umem->npages, true); > + atomic64_sub(dev->umem->npages, &dev->umem->mm->pinned_vm); > + mmdrop(dev->umem->mm); > + vfree(dev->umem->pages); > + kfree(dev->umem); > + dev->umem = NULL; > + ret = 0; > +unlock: > + mutex_unlock(&dev->mem_lock); > + return ret; > +} > + > +static int vduse_dev_reg_umem(struct vduse_dev *dev, > + u64 iova, u64 uaddr, u64 size) > +{ > + struct page **page_list = NULL; > + struct vduse_umem *umem = NULL; > + long pinned = 0; > + unsigned long npages, lock_limit; > + int ret; > + > + if (!dev->domain->bounce_map || > + size != dev->domain->bounce_size || > + iova != 0 || uaddr & ~PAGE_MASK) > + return -EINVAL; > + > + mutex_lock(&dev->mem_lock); > + ret = -EEXIST; > + if (dev->umem) > + goto unlock; > + > + ret = -ENOMEM; > + npages = size >> PAGE_SHIFT; > + page_list = __vmalloc(array_size(npages, sizeof(struct page *)), > + GFP_KERNEL_ACCOUNT); > + umem = kzalloc(sizeof(*umem), GFP_KERNEL); > + if (!page_list || !umem) > + goto unlock; > + > + mmap_read_lock(current->mm); > + > + lock_limit = PFN_DOWN(rlimit(RLIMIT_MEMLOCK)); > + if (npages + atomic64_read(¤t->mm->pinned_vm) > lock_limit) > + goto out; > + > + pinned = pin_user_pages(uaddr, npages, FOLL_LONGTERM | FOLL_WRITE, > + page_list, NULL); > + if (pinned != npages) { > + ret = pinned < 0 ? pinned : -ENOMEM; > + goto out; > + } > + > + ret = vduse_domain_add_user_bounce_pages(dev->domain, > + page_list, pinned); > + if (ret) > + goto out; > + > + atomic64_add(npages, ¤t->mm->pinned_vm); > + > + umem->pages = page_list; > + umem->npages = pinned; > + umem->iova = iova; > + umem->mm = current->mm; > + mmgrab(current->mm); > + > + dev->umem = umem; > +out: > + if (ret && pinned > 0) > + unpin_user_pages(page_list, pinned); > + > + mmap_read_unlock(current->mm); > +unlock: > + if (ret) { > + vfree(page_list); > + kfree(umem); > + } > + mutex_unlock(&dev->mem_lock); > + return ret; > +} > + > static long vduse_dev_ioctl(struct file *file, unsigned int cmd, > unsigned long arg) > { > @@ -1089,6 +1196,38 @@ static long vduse_dev_ioctl(struct file *file, unsigned int cmd, > ret = vduse_dev_queue_irq_work(dev, &dev->vqs[index].inject); > break; > } > + case VDUSE_IOTLB_REG_UMEM: { > + struct vduse_iova_umem umem; > + > + ret = -EFAULT; > + if (copy_from_user(&umem, argp, sizeof(umem))) > + break; > + > + ret = -EINVAL; > + if (!is_mem_zero((const char *)umem.reserved, > + sizeof(umem.reserved))) > + break; > + > + ret = vduse_dev_reg_umem(dev, umem.iova, > + umem.uaddr, umem.size); > + break; > + } > + case VDUSE_IOTLB_DEREG_UMEM: { > + struct vduse_iova_umem umem; > + > + ret = -EFAULT; > + if (copy_from_user(&umem, argp, sizeof(umem))) > + break; > + > + ret = -EINVAL; > + if (!is_mem_zero((const char *)umem.reserved, > + sizeof(umem.reserved))) > + break; > + > + ret = vduse_dev_dereg_umem(dev, umem.iova, > + umem.size); > + break; > + } > default: > ret = -ENOIOCTLCMD; > break; > @@ -1101,6 +1240,7 @@ static int vduse_dev_release(struct inode *inode, struct file *file) > { > struct vduse_dev *dev = file->private_data; > > + vduse_dev_dereg_umem(dev, 0, dev->domain->bounce_size); > spin_lock(&dev->msg_lock); > /* Make sure the inflight messages can processed after reconncection */ > list_splice_init(&dev->recv_list, &dev->send_list); > @@ -1163,6 +1303,7 @@ static struct vduse_dev *vduse_dev_create(void) > return NULL; > > mutex_init(&dev->lock); > + mutex_init(&dev->mem_lock); > spin_lock_init(&dev->msg_lock); > INIT_LIST_HEAD(&dev->send_list); > INIT_LIST_HEAD(&dev->recv_list); > diff --git a/include/uapi/linux/vduse.h b/include/uapi/linux/vduse.h > index 7cfe1c1280c0..9885e0571f09 100644 > --- a/include/uapi/linux/vduse.h > +++ b/include/uapi/linux/vduse.h > @@ -210,6 +210,29 @@ struct vduse_vq_eventfd { > */ > #define VDUSE_VQ_INJECT_IRQ _IOW(VDUSE_BASE, 0x17, __u32) > > +/** > + * struct vduse_iova_umem - userspace memory configuration for one IOVA region > + * @uaddr: start address of userspace memory, it must be aligned to page size > + * @iova: start of the IOVA region > + * @size: size of the IOVA region > + * @reserved: for future use, needs to be initialized to zero > + * > + * Structure used by VDUSE_IOTLB_REG_UMEM and VDUSE_IOTLB_DEREG_UMEM > + * ioctls to register/de-register userspace memory for IOVA regions > + */ > +struct vduse_iova_umem { > + __u64 uaddr; > + __u64 iova; > + __u64 size; > + __u64 reserved[3]; > +}; > + > +/* Register userspace memory for IOVA regions */ > +#define VDUSE_IOTLB_REG_UMEM _IOW(VDUSE_BASE, 0x18, struct vduse_iova_umem) > + > +/* De-register the userspace memory. Caller should set iova and size field. */ > +#define VDUSE_IOTLB_DEREG_UMEM _IOW(VDUSE_BASE, 0x19, struct vduse_iova_umem) > + > /* The control messages definition for read(2)/write(2) on /dev/vduse/$NAME */ > > /** > -- > 2.20.1 >