From ab31523c2fcac557226bac72cbdf5fafe01f9a26 Mon Sep 17 00:00:00 2001 From: Gerd Hoffmann Date: Tue, 14 Dec 2010 18:40:46 +0000 Subject: xen/gntdev: allow usermode to map granted pages The gntdev driver allows usermode to map granted pages from other domains. This is typically used to implement a Xen backend driver in user mode. Signed-off-by: Gerd Hoffmann Signed-off-by: Stefano Stabellini Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/Kconfig | 7 + drivers/xen/Makefile | 2 + drivers/xen/gntdev.c | 635 +++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 644 insertions(+) create mode 100644 drivers/xen/gntdev.c (limited to 'drivers') diff --git a/drivers/xen/Kconfig b/drivers/xen/Kconfig index 6e6180ccd72..bd3095f0dff 100644 --- a/drivers/xen/Kconfig +++ b/drivers/xen/Kconfig @@ -62,6 +62,13 @@ config XEN_SYS_HYPERVISOR virtual environment, /sys/hypervisor will still be present, but will have no xen contents. +config XEN_GNTDEV + tristate "userspace grant access device driver" + depends on XEN + select MMU_NOTIFIER + help + Allows userspace processes to use grants. + config XEN_PLATFORM_PCI tristate "xen platform pci device driver" depends on XEN_PVHVM diff --git a/drivers/xen/Makefile b/drivers/xen/Makefile index 533a199e7a3..674fdb5fee4 100644 --- a/drivers/xen/Makefile +++ b/drivers/xen/Makefile @@ -9,6 +9,7 @@ obj-$(CONFIG_HOTPLUG_CPU) += cpu_hotplug.o obj-$(CONFIG_XEN_XENCOMM) += xencomm.o obj-$(CONFIG_XEN_BALLOON) += balloon.o obj-$(CONFIG_XEN_DEV_EVTCHN) += xen-evtchn.o +obj-$(CONFIG_XEN_GNTDEV) += xen-gntdev.o obj-$(CONFIG_XENFS) += xenfs/ obj-$(CONFIG_XEN_SYS_HYPERVISOR) += sys-hypervisor.o obj-$(CONFIG_XEN_PLATFORM_PCI) += platform-pci.o @@ -16,4 +17,5 @@ obj-$(CONFIG_SWIOTLB_XEN) += swiotlb-xen.o obj-$(CONFIG_XEN_DOM0) += pci.o xen-evtchn-y := evtchn.o +xen-gntdev-y := gntdev.o diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c new file mode 100644 index 00000000000..cfe063372aa --- /dev/null +++ b/drivers/xen/gntdev.c @@ -0,0 +1,635 @@ +/****************************************************************************** + * gntdev.c + * + * Device for accessing (in user-space) pages that have been granted by other + * domains. + * + * Copyright (c) 2006-2007, D G Murray. + * (c) 2009 Gerd Hoffmann + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#undef DEBUG + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include +#include +#include +#include + +MODULE_LICENSE("GPL"); +MODULE_AUTHOR("Derek G. Murray , " + "Gerd Hoffmann "); +MODULE_DESCRIPTION("User-space granted page access driver"); + +static int limit = 1024; +module_param(limit, int, 0644); +MODULE_PARM_DESC(limit, "Maximum number of grants that may be mapped at " + "once by a gntdev instance"); + +struct gntdev_priv { + struct list_head maps; + uint32_t used; + uint32_t limit; + /* lock protects maps from concurrent changes */ + spinlock_t lock; + struct mm_struct *mm; + struct mmu_notifier mn; +}; + +struct grant_map { + struct list_head next; + struct gntdev_priv *priv; + struct vm_area_struct *vma; + int index; + int count; + int flags; + int is_mapped; + struct ioctl_gntdev_grant_ref *grants; + struct gnttab_map_grant_ref *map_ops; + struct gnttab_unmap_grant_ref *unmap_ops; +}; + +/* ------------------------------------------------------------------ */ + +static void gntdev_print_maps(struct gntdev_priv *priv, + char *text, int text_index) +{ +#ifdef DEBUG + struct grant_map *map; + + pr_debug("maps list (priv %p, usage %d/%d)\n", + priv, priv->used, priv->limit); + + list_for_each_entry(map, &priv->maps, next) + pr_debug(" index %2d, count %2d %s\n", + map->index, map->count, + map->index == text_index && text ? text : ""); +#endif +} + +static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count) +{ + struct grant_map *add; + + add = kzalloc(sizeof(struct grant_map), GFP_KERNEL); + if (NULL == add) + return NULL; + + add->grants = kzalloc(sizeof(add->grants[0]) * count, GFP_KERNEL); + add->map_ops = kzalloc(sizeof(add->map_ops[0]) * count, GFP_KERNEL); + add->unmap_ops = kzalloc(sizeof(add->unmap_ops[0]) * count, GFP_KERNEL); + if (NULL == add->grants || + NULL == add->map_ops || + NULL == add->unmap_ops) + goto err; + + add->index = 0; + add->count = count; + add->priv = priv; + + if (add->count + priv->used > priv->limit) + goto err; + + return add; + +err: + kfree(add->grants); + kfree(add->map_ops); + kfree(add->unmap_ops); + kfree(add); + return NULL; +} + +static void gntdev_add_map(struct gntdev_priv *priv, struct grant_map *add) +{ + struct grant_map *map; + + list_for_each_entry(map, &priv->maps, next) { + if (add->index + add->count < map->index) { + list_add_tail(&add->next, &map->next); + goto done; + } + add->index = map->index + map->count; + } + list_add_tail(&add->next, &priv->maps); + +done: + priv->used += add->count; + gntdev_print_maps(priv, "[new]", add->index); +} + +static struct grant_map *gntdev_find_map_index(struct gntdev_priv *priv, + int index, int count) +{ + struct grant_map *map; + + list_for_each_entry(map, &priv->maps, next) { + if (map->index != index) + continue; + if (map->count != count) + continue; + return map; + } + return NULL; +} + +static struct grant_map *gntdev_find_map_vaddr(struct gntdev_priv *priv, + unsigned long vaddr) +{ + struct grant_map *map; + + list_for_each_entry(map, &priv->maps, next) { + if (!map->vma) + continue; + if (vaddr < map->vma->vm_start) + continue; + if (vaddr >= map->vma->vm_end) + continue; + return map; + } + return NULL; +} + +static int gntdev_del_map(struct grant_map *map) +{ + int i; + + if (map->vma) + return -EBUSY; + for (i = 0; i < map->count; i++) + if (map->unmap_ops[i].handle) + return -EBUSY; + + map->priv->used -= map->count; + list_del(&map->next); + return 0; +} + +static void gntdev_free_map(struct grant_map *map) +{ + if (!map) + return; + kfree(map->grants); + kfree(map->map_ops); + kfree(map->unmap_ops); + kfree(map); +} + +/* ------------------------------------------------------------------ */ + +static int find_grant_ptes(pte_t *pte, pgtable_t token, + unsigned long addr, void *data) +{ + struct grant_map *map = data; + unsigned int pgnr = (addr - map->vma->vm_start) >> PAGE_SHIFT; + u64 pte_maddr; + + BUG_ON(pgnr >= map->count); + pte_maddr = (u64)pfn_to_mfn(page_to_pfn(token)) << PAGE_SHIFT; + pte_maddr += (unsigned long)pte & ~PAGE_MASK; + gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr, map->flags, + map->grants[pgnr].ref, + map->grants[pgnr].domid); + gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr, map->flags, + 0 /* handle */); + return 0; +} + +static int map_grant_pages(struct grant_map *map) +{ + int i, err = 0; + + pr_debug("map %d+%d\n", map->index, map->count); + err = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, + map->map_ops, map->count); + if (err) + return err; + + for (i = 0; i < map->count; i++) { + if (map->map_ops[i].status) + err = -EINVAL; + map->unmap_ops[i].handle = map->map_ops[i].handle; + } + return err; +} + +static int unmap_grant_pages(struct grant_map *map, int offset, int pages) +{ + int i, err = 0; + + pr_debug("map %d+%d [%d+%d]\n", map->index, map->count, offset, pages); + err = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, + map->unmap_ops + offset, pages); + if (err) + return err; + + for (i = 0; i < pages; i++) { + if (map->unmap_ops[offset+i].status) + err = -EINVAL; + map->unmap_ops[offset+i].handle = 0; + } + return err; +} + +/* ------------------------------------------------------------------ */ + +static void gntdev_vma_close(struct vm_area_struct *vma) +{ + struct grant_map *map = vma->vm_private_data; + + pr_debug("close %p\n", vma); + map->is_mapped = 0; + map->vma = NULL; + vma->vm_private_data = NULL; +} + +static int gntdev_vma_fault(struct vm_area_struct *vma, struct vm_fault *vmf) +{ + pr_debug("vaddr %p, pgoff %ld (shouldn't happen)\n", + vmf->virtual_address, vmf->pgoff); + vmf->flags = VM_FAULT_ERROR; + return 0; +} + +static struct vm_operations_struct gntdev_vmops = { + .close = gntdev_vma_close, + .fault = gntdev_vma_fault, +}; + +/* ------------------------------------------------------------------ */ + +static void mn_invl_range_start(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, unsigned long end) +{ + struct gntdev_priv *priv = container_of(mn, struct gntdev_priv, mn); + struct grant_map *map; + unsigned long mstart, mend; + int err; + + spin_lock(&priv->lock); + list_for_each_entry(map, &priv->maps, next) { + if (!map->vma) + continue; + if (!map->is_mapped) + continue; + if (map->vma->vm_start >= end) + continue; + if (map->vma->vm_end <= start) + continue; + mstart = max(start, map->vma->vm_start); + mend = min(end, map->vma->vm_end); + pr_debug("map %d+%d (%lx %lx), range %lx %lx, mrange %lx %lx\n", + map->index, map->count, + map->vma->vm_start, map->vma->vm_end, + start, end, mstart, mend); + err = unmap_grant_pages(map, + (mstart - map->vma->vm_start) >> PAGE_SHIFT, + (mend - mstart) >> PAGE_SHIFT); + WARN_ON(err); + } + spin_unlock(&priv->lock); +} + +static void mn_invl_page(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long address) +{ + mn_invl_range_start(mn, mm, address, address + PAGE_SIZE); +} + +static void mn_release(struct mmu_notifier *mn, + struct mm_struct *mm) +{ + struct gntdev_priv *priv = container_of(mn, struct gntdev_priv, mn); + struct grant_map *map; + int err; + + spin_lock(&priv->lock); + list_for_each_entry(map, &priv->maps, next) { + if (!map->vma) + continue; + pr_debug("map %d+%d (%lx %lx)\n", + map->index, map->count, + map->vma->vm_start, map->vma->vm_end); + err = unmap_grant_pages(map, /* offset */ 0, map->count); + WARN_ON(err); + } + spin_unlock(&priv->lock); +} + +struct mmu_notifier_ops gntdev_mmu_ops = { + .release = mn_release, + .invalidate_page = mn_invl_page, + .invalidate_range_start = mn_invl_range_start, +}; + +/* ------------------------------------------------------------------ */ + +static int gntdev_open(struct inode *inode, struct file *flip) +{ + struct gntdev_priv *priv; + int ret = 0; + + priv = kzalloc(sizeof(*priv), GFP_KERNEL); + if (!priv) + return -ENOMEM; + + INIT_LIST_HEAD(&priv->maps); + spin_lock_init(&priv->lock); + priv->limit = limit; + + priv->mm = get_task_mm(current); + if (!priv->mm) { + kfree(priv); + return -ENOMEM; + } + priv->mn.ops = &gntdev_mmu_ops; + ret = mmu_notifier_register(&priv->mn, priv->mm); + mmput(priv->mm); + + if (ret) { + kfree(priv); + return ret; + } + + flip->private_data = priv; + pr_debug("priv %p\n", priv); + + return 0; +} + +static int gntdev_release(struct inode *inode, struct file *flip) +{ + struct gntdev_priv *priv = flip->private_data; + struct grant_map *map; + int err; + + pr_debug("priv %p\n", priv); + + spin_lock(&priv->lock); + while (!list_empty(&priv->maps)) { + map = list_entry(priv->maps.next, struct grant_map, next); + err = gntdev_del_map(map); + if (WARN_ON(err)) + gntdev_free_map(map); + + } + spin_unlock(&priv->lock); + + mmu_notifier_unregister(&priv->mn, priv->mm); + kfree(priv); + return 0; +} + +static long gntdev_ioctl_map_grant_ref(struct gntdev_priv *priv, + struct ioctl_gntdev_map_grant_ref __user *u) +{ + struct ioctl_gntdev_map_grant_ref op; + struct grant_map *map; + int err; + + if (copy_from_user(&op, u, sizeof(op)) != 0) + return -EFAULT; + pr_debug("priv %p, add %d\n", priv, op.count); + if (unlikely(op.count <= 0)) + return -EINVAL; + if (unlikely(op.count > priv->limit)) + return -EINVAL; + + err = -ENOMEM; + map = gntdev_alloc_map(priv, op.count); + if (!map) + return err; + if (copy_from_user(map->grants, &u->refs, + sizeof(map->grants[0]) * op.count) != 0) { + gntdev_free_map(map); + return err; + } + + spin_lock(&priv->lock); + gntdev_add_map(priv, map); + op.index = map->index << PAGE_SHIFT; + spin_unlock(&priv->lock); + + if (copy_to_user(u, &op, sizeof(op)) != 0) { + spin_lock(&priv->lock); + gntdev_del_map(map); + spin_unlock(&priv->lock); + gntdev_free_map(map); + return err; + } + return 0; +} + +static long gntdev_ioctl_unmap_grant_ref(struct gntdev_priv *priv, + struct ioctl_gntdev_unmap_grant_ref __user *u) +{ + struct ioctl_gntdev_unmap_grant_ref op; + struct grant_map *map; + int err = -ENOENT; + + if (copy_from_user(&op, u, sizeof(op)) != 0) + return -EFAULT; + pr_debug("priv %p, del %d+%d\n", priv, (int)op.index, (int)op.count); + + spin_lock(&priv->lock); + map = gntdev_find_map_index(priv, op.index >> PAGE_SHIFT, op.count); + if (map) + err = gntdev_del_map(map); + spin_unlock(&priv->lock); + if (!err) + gntdev_free_map(map); + return err; +} + +static long gntdev_ioctl_get_offset_for_vaddr(struct gntdev_priv *priv, + struct ioctl_gntdev_get_offset_for_vaddr __user *u) +{ + struct ioctl_gntdev_get_offset_for_vaddr op; + struct grant_map *map; + + if (copy_from_user(&op, u, sizeof(op)) != 0) + return -EFAULT; + pr_debug("priv %p, offset for vaddr %lx\n", priv, (unsigned long)op.vaddr); + + spin_lock(&priv->lock); + map = gntdev_find_map_vaddr(priv, op.vaddr); + if (map == NULL || + map->vma->vm_start != op.vaddr) { + spin_unlock(&priv->lock); + return -EINVAL; + } + op.offset = map->index << PAGE_SHIFT; + op.count = map->count; + spin_unlock(&priv->lock); + + if (copy_to_user(u, &op, sizeof(op)) != 0) + return -EFAULT; + return 0; +} + +static long gntdev_ioctl_set_max_grants(struct gntdev_priv *priv, + struct ioctl_gntdev_set_max_grants __user *u) +{ + struct ioctl_gntdev_set_max_grants op; + + if (copy_from_user(&op, u, sizeof(op)) != 0) + return -EFAULT; + pr_debug("priv %p, limit %d\n", priv, op.count); + if (op.count > limit) + return -E2BIG; + + spin_lock(&priv->lock); + priv->limit = op.count; + spin_unlock(&priv->lock); + return 0; +} + +static long gntdev_ioctl(struct file *flip, + unsigned int cmd, unsigned long arg) +{ + struct gntdev_priv *priv = flip->private_data; + void __user *ptr = (void __user *)arg; + + switch (cmd) { + case IOCTL_GNTDEV_MAP_GRANT_REF: + return gntdev_ioctl_map_grant_ref(priv, ptr); + + case IOCTL_GNTDEV_UNMAP_GRANT_REF: + return gntdev_ioctl_unmap_grant_ref(priv, ptr); + + case IOCTL_GNTDEV_GET_OFFSET_FOR_VADDR: + return gntdev_ioctl_get_offset_for_vaddr(priv, ptr); + + case IOCTL_GNTDEV_SET_MAX_GRANTS: + return gntdev_ioctl_set_max_grants(priv, ptr); + + default: + pr_debug("priv %p, unknown cmd %x\n", priv, cmd); + return -ENOIOCTLCMD; + } + + return 0; +} + +static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma) +{ + struct gntdev_priv *priv = flip->private_data; + int index = vma->vm_pgoff; + int count = (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; + struct grant_map *map; + int err = -EINVAL; + + if ((vma->vm_flags & VM_WRITE) && !(vma->vm_flags & VM_SHARED)) + return -EINVAL; + + pr_debug("map %d+%d at %lx (pgoff %lx)\n", + index, count, vma->vm_start, vma->vm_pgoff); + + spin_lock(&priv->lock); + map = gntdev_find_map_index(priv, index, count); + if (!map) + goto unlock_out; + if (map->vma) + goto unlock_out; + if (priv->mm != vma->vm_mm) { + printk(KERN_WARNING "Huh? Other mm?\n"); + goto unlock_out; + } + + vma->vm_ops = &gntdev_vmops; + + vma->vm_flags |= VM_RESERVED|VM_DONTCOPY|VM_DONTEXPAND; + + vma->vm_private_data = map; + map->vma = vma; + + map->flags = GNTMAP_host_map | GNTMAP_application_map | GNTMAP_contains_pte; + if (!(vma->vm_flags & VM_WRITE)) + map->flags |= GNTMAP_readonly; + + err = apply_to_page_range(vma->vm_mm, vma->vm_start, + vma->vm_end - vma->vm_start, + find_grant_ptes, map); + if (err) { + printk(KERN_WARNING "find_grant_ptes() failure.\n"); + goto unlock_out; + } + + err = map_grant_pages(map); + if (err) { + printk(KERN_WARNING "map_grant_pages() failure.\n"); + goto unlock_out; + } + map->is_mapped = 1; + +unlock_out: + spin_unlock(&priv->lock); + return err; +} + +static const struct file_operations gntdev_fops = { + .owner = THIS_MODULE, + .open = gntdev_open, + .release = gntdev_release, + .mmap = gntdev_mmap, + .unlocked_ioctl = gntdev_ioctl +}; + +static struct miscdevice gntdev_miscdev = { + .minor = MISC_DYNAMIC_MINOR, + .name = "xen/gntdev", + .fops = &gntdev_fops, +}; + +/* ------------------------------------------------------------------ */ + +static int __init gntdev_init(void) +{ + int err; + + if (!xen_domain()) + return -ENODEV; + + err = misc_register(&gntdev_miscdev); + if (err != 0) { + printk(KERN_ERR "Could not register gntdev device\n"); + return err; + } + return 0; +} + +static void __exit gntdev_exit(void) +{ + misc_deregister(&gntdev_miscdev); +} + +module_init(gntdev_init); +module_exit(gntdev_exit); + +/* ------------------------------------------------------------------ */ -- cgit v1.2.3-70-g09d2 From 8d3eaea24609c7cd6fb0e6471f46a52f9e5d0202 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Thu, 11 Nov 2010 14:39:12 -0800 Subject: xen/gntdev: add VM_PFNMAP to vma These pages are from other domains, so don't have any local PFN. VM_PFNMAP is the closest concept Linux has to this. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/gntdev.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index cfe063372aa..fa6355a9708 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -564,7 +564,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma) vma->vm_ops = &gntdev_vmops; - vma->vm_flags |= VM_RESERVED|VM_DONTCOPY|VM_DONTEXPAND; + vma->vm_flags |= VM_RESERVED|VM_DONTCOPY|VM_DONTEXPAND|VM_PFNMAP; vma->vm_private_data = map; map->vma = vma; -- cgit v1.2.3-70-g09d2 From 9329e7604fe915fd0201633d3c38adae307d56a5 Mon Sep 17 00:00:00 2001 From: Ian Campbell Date: Wed, 8 Dec 2010 11:57:40 +0000 Subject: xen: gntdev: move use of GNTMAP_contains_pte next to the map_op This flag controls the meaning of gnttab_map_grant_ref.host_addr and specifies that the field contains a reference to the pte entry to be used to perform the mapping. Therefore move the use of this flag to the point at which we actually use a reference to the pte instead of something else, splitting up the usage of the flag in this way is confusing and potentially error prone. The other flags are all properties of the mapping itself as opposed to properties of the hypercall arguments and therefore it make sense to continue to pass them round in map->flags. Signed-off-by: Ian Campbell Cc: Stefano Stabellini Cc: Derek G. Murray Cc: Gerd Hoffmann Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/gntdev.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'drivers') diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index fa6355a9708..888d76307d5 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -211,10 +211,12 @@ static int find_grant_ptes(pte_t *pte, pgtable_t token, BUG_ON(pgnr >= map->count); pte_maddr = (u64)pfn_to_mfn(page_to_pfn(token)) << PAGE_SHIFT; pte_maddr += (unsigned long)pte & ~PAGE_MASK; - gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr, map->flags, + gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr, + GNTMAP_contains_pte | map->flags, map->grants[pgnr].ref, map->grants[pgnr].domid); - gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr, map->flags, + gnttab_set_unmap_op(&map->unmap_ops[pgnr], pte_maddr, + GNTMAP_contains_pte | map->flags, 0 /* handle */); return 0; } @@ -569,7 +571,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma) vma->vm_private_data = map; map->vma = vma; - map->flags = GNTMAP_host_map | GNTMAP_application_map | GNTMAP_contains_pte; + map->flags = GNTMAP_host_map | GNTMAP_application_map; if (!(vma->vm_flags & VM_WRITE)) map->flags |= GNTMAP_readonly; -- cgit v1.2.3-70-g09d2 From ba5d1012292403c8037adf4a54c4ec50dfe846c4 Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Wed, 8 Dec 2010 10:54:32 -0800 Subject: xen/gntdev: stop using "token" argument It's the struct page of the L1 pte page. But we can get its mfn by simply doing an arbitrary_virt_to_machine() on it anyway (which is the safe conservative choice; since we no longer allow HIGHPTE pages, we would never expect to be operating on a mapped pte page). Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/gntdev.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index 888d76307d5..a2ea5335e15 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -209,8 +209,8 @@ static int find_grant_ptes(pte_t *pte, pgtable_t token, u64 pte_maddr; BUG_ON(pgnr >= map->count); - pte_maddr = (u64)pfn_to_mfn(page_to_pfn(token)) << PAGE_SHIFT; - pte_maddr += (unsigned long)pte & ~PAGE_MASK; + pte_maddr = arbitrary_virt_to_machine(pte).maddr; + gnttab_set_map_op(&map->map_ops[pgnr], pte_maddr, GNTMAP_contains_pte | map->flags, map->grants[pgnr].ref, -- cgit v1.2.3-70-g09d2 From f0a70c882ea546bbd802643990ceded32c39facc Mon Sep 17 00:00:00 2001 From: Daniel De Graaf Date: Fri, 7 Jan 2011 11:51:47 +0000 Subject: xen/gntdev: Fix circular locking dependency apply_to_page_range will acquire PTE lock while priv->lock is held, and mn_invl_range_start tries to acquire priv->lock with PTE already held. Fix by not holding priv->lock during the entire map operation. This is safe because map->vma is set nonzero while the lock is held, which will cause subsequent maps to fail and will cause the unmap ioctl (and other users of gntdev_del_map) to return -EBUSY until the area is unmapped. It is similarly impossible for gntdev_vma_close to be called while the vma is still being created. Signed-off-by: Daniel De Graaf Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/gntdev.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index a2ea5335e15..aba76d437ea 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -575,21 +575,26 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma) if (!(vma->vm_flags & VM_WRITE)) map->flags |= GNTMAP_readonly; + spin_unlock(&priv->lock); + err = apply_to_page_range(vma->vm_mm, vma->vm_start, vma->vm_end - vma->vm_start, find_grant_ptes, map); if (err) { printk(KERN_WARNING "find_grant_ptes() failure.\n"); - goto unlock_out; + return err; } err = map_grant_pages(map); if (err) { printk(KERN_WARNING "map_grant_pages() failure.\n"); - goto unlock_out; + return err; } + map->is_mapped = 1; + return 0; + unlock_out: spin_unlock(&priv->lock); return err; -- cgit v1.2.3-70-g09d2 From 289b777eac19c811b474593b4d2fd14e46340c23 Mon Sep 17 00:00:00 2001 From: Stefano Stabellini Date: Fri, 10 Dec 2010 14:54:44 +0000 Subject: xen: introduce gnttab_map_refs and gnttab_unmap_refs gnttab_map_refs maps some grant refs and uses the new m2p override to set a proper m2p mapping for the granted pages. gnttab_unmap_refs unmaps the granted refs and removes th mappings from the m2p override. Signed-off-by: Stefano Stabellini Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/grant-table.c | 36 ++++++++++++++++++++++++++++++++++++ include/xen/grant_table.h | 5 +++++ 2 files changed, 41 insertions(+) (limited to 'drivers') diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c index 6c453181649..1afd5690858 100644 --- a/drivers/xen/grant-table.c +++ b/drivers/xen/grant-table.c @@ -447,6 +447,42 @@ unsigned int gnttab_max_grant_frames(void) } EXPORT_SYMBOL_GPL(gnttab_max_grant_frames); +int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops, + struct page **pages, unsigned int count) +{ + int i, ret; + pte_t *pte; + unsigned long mfn; + + ret = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, map_ops, count); + + for (i = 0; i < count; i++) { + /* m2p override only supported for GNTMAP_contains_pte mappings */ + if (!(map_ops[i].flags & GNTMAP_contains_pte)) + continue; + pte = (pte_t *) (mfn_to_virt(PFN_DOWN(map_ops[i].host_addr)) + + (map_ops[i].host_addr & ~PAGE_MASK)); + mfn = pte_mfn(*pte); + m2p_add_override(mfn, pages[i]); + } + + return ret; +} +EXPORT_SYMBOL_GPL(gnttab_map_refs); + +int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops, + struct page **pages, unsigned int count) +{ + int i, ret; + + ret = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap_ops, count); + for (i = 0; i < count; i++) + m2p_remove_override(pages[i]); + + return ret; +} +EXPORT_SYMBOL_GPL(gnttab_unmap_refs); + static int gnttab_map(unsigned int start_idx, unsigned int end_idx) { struct gnttab_setup_table setup; diff --git a/include/xen/grant_table.h b/include/xen/grant_table.h index 1821aa103d8..b1fab6b5b3e 100644 --- a/include/xen/grant_table.h +++ b/include/xen/grant_table.h @@ -155,4 +155,9 @@ unsigned int gnttab_max_grant_frames(void); #define gnttab_map_vaddr(map) ((void *)(map.host_virt_addr)) +int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops, + struct page **pages, unsigned int count); +int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops, + struct page **pages, unsigned int count); + #endif /* __ASM_GNTTAB_H__ */ -- cgit v1.2.3-70-g09d2 From a12b4eb34bb1ea16046c5b61e7a887e252cc1cce Mon Sep 17 00:00:00 2001 From: Stefano Stabellini Date: Fri, 10 Dec 2010 14:56:42 +0000 Subject: xen gntdev: use gnttab_map_refs and gnttab_unmap_refs Use gnttab_map_refs and gnttab_unmap_refs to map and unmap the grant ref, so that we can have a corresponding struct page. Signed-off-by: Stefano Stabellini Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Konrad Rzeszutek Wilk --- drivers/xen/gntdev.c | 37 ++++++++++++++++++++++++++++++------- 1 file changed, 30 insertions(+), 7 deletions(-) (limited to 'drivers') diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c index aba76d437ea..1e31cdcdae1 100644 --- a/drivers/xen/gntdev.c +++ b/drivers/xen/gntdev.c @@ -71,6 +71,7 @@ struct grant_map { struct ioctl_gntdev_grant_ref *grants; struct gnttab_map_grant_ref *map_ops; struct gnttab_unmap_grant_ref *unmap_ops; + struct page **pages; }; /* ------------------------------------------------------------------ */ @@ -94,6 +95,7 @@ static void gntdev_print_maps(struct gntdev_priv *priv, static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count) { struct grant_map *add; + int i; add = kzalloc(sizeof(struct grant_map), GFP_KERNEL); if (NULL == add) @@ -102,11 +104,19 @@ static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count) add->grants = kzalloc(sizeof(add->grants[0]) * count, GFP_KERNEL); add->map_ops = kzalloc(sizeof(add->map_ops[0]) * count, GFP_KERNEL); add->unmap_ops = kzalloc(sizeof(add->unmap_ops[0]) * count, GFP_KERNEL); - if (NULL == add->grants || - NULL == add->map_ops || - NULL == add->unmap_ops) + add->pages = kzalloc(sizeof(add->pages[0]) * count, GFP_KERNEL); + if (NULL == add->grants || + NULL == add->map_ops || + NULL == add->unmap_ops || + NULL == add->pages) goto err; + for (i = 0; i < count; i++) { + add->pages[i] = alloc_page(GFP_KERNEL | __GFP_HIGHMEM); + if (add->pages[i] == NULL) + goto err; + } + add->index = 0; add->count = count; add->priv = priv; @@ -117,6 +127,12 @@ static struct grant_map *gntdev_alloc_map(struct gntdev_priv *priv, int count) return add; err: + if (add->pages) + for (i = 0; i < count; i++) { + if (add->pages[i]) + __free_page(add->pages[i]); + } + kfree(add->pages); kfree(add->grants); kfree(add->map_ops); kfree(add->unmap_ops); @@ -191,8 +207,17 @@ static int gntdev_del_map(struct grant_map *map) static void gntdev_free_map(struct grant_map *map) { + int i; + if (!map) return; + + if (map->pages) + for (i = 0; i < map->count; i++) { + if (map->pages[i]) + __free_page(map->pages[i]); + } + kfree(map->pages); kfree(map->grants); kfree(map->map_ops); kfree(map->unmap_ops); @@ -226,8 +251,7 @@ static int map_grant_pages(struct grant_map *map) int i, err = 0; pr_debug("map %d+%d\n", map->index, map->count); - err = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, - map->map_ops, map->count); + err = gnttab_map_refs(map->map_ops, map->pages, map->count); if (err) return err; @@ -244,8 +268,7 @@ static int unmap_grant_pages(struct grant_map *map, int offset, int pages) int i, err = 0; pr_debug("map %d+%d [%d+%d]\n", map->index, map->count, offset, pages); - err = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, - map->unmap_ops + offset, pages); + err = gnttab_unmap_refs(map->unmap_ops + offset, map->pages, pages); if (err) return err; -- cgit v1.2.3-70-g09d2 From 87f1d40a706bdebdc8f959b9ac291d0d8fdfcc7e Mon Sep 17 00:00:00 2001 From: Jeremy Fitzhardinge Date: Mon, 13 Dec 2010 14:42:30 +0000 Subject: xen p2m: clear the old pte when adding a page to m2p_override When adding a page to m2p_override we change the p2m of the page so we need to also clear the old pte of the kernel linear mapping because it doesn't correspond anymore. When we remove the page from m2p_override we restore the original p2m of the page and we also restore the old pte of the kernel linear mapping. Before changing the p2m mappings in m2p_add_override and m2p_remove_override, check that the page passed as argument is valid and return an error if it is not. Signed-off-by: Jeremy Fitzhardinge Signed-off-by: Stefano Stabellini Signed-off-by: Konrad Rzeszutek Wilk --- arch/x86/include/asm/xen/page.h | 4 ++-- arch/x86/xen/p2m.c | 49 +++++++++++++++++++++++++++++++++++++---- drivers/xen/grant-table.c | 16 +++++++++++--- 3 files changed, 60 insertions(+), 9 deletions(-) (limited to 'drivers') diff --git a/arch/x86/include/asm/xen/page.h b/arch/x86/include/asm/xen/page.h index 50f0a0f6bd6..f25bdf238a3 100644 --- a/arch/x86/include/asm/xen/page.h +++ b/arch/x86/include/asm/xen/page.h @@ -42,8 +42,8 @@ extern unsigned int machine_to_phys_order; extern unsigned long get_phys_to_machine(unsigned long pfn); extern bool set_phys_to_machine(unsigned long pfn, unsigned long mfn); -extern void m2p_add_override(unsigned long mfn, struct page *page); -extern void m2p_remove_override(struct page *page); +extern int m2p_add_override(unsigned long mfn, struct page *page); +extern int m2p_remove_override(struct page *page); extern struct page *m2p_find_override(unsigned long mfn); extern unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn); diff --git a/arch/x86/xen/p2m.c b/arch/x86/xen/p2m.c index b3b19d43b95..40d51225ff0 100644 --- a/arch/x86/xen/p2m.c +++ b/arch/x86/xen/p2m.c @@ -29,6 +29,7 @@ #include #include #include +#include #include #include @@ -404,34 +405,74 @@ static unsigned long mfn_hash(unsigned long mfn) } /* Add an MFN override for a particular page */ -void m2p_add_override(unsigned long mfn, struct page *page) +int m2p_add_override(unsigned long mfn, struct page *page) { unsigned long flags; - unsigned long pfn = page_to_pfn(page); + unsigned long pfn; + unsigned long address; + unsigned level; + pte_t *ptep = NULL; + + pfn = page_to_pfn(page); + if (!PageHighMem(page)) { + address = (unsigned long)__va(pfn << PAGE_SHIFT); + ptep = lookup_address(address, &level); + + if (WARN(ptep == NULL || level != PG_LEVEL_4K, + "m2p_add_override: pfn %lx not mapped", pfn)) + return -EINVAL; + } + page->private = mfn; page->index = pfn_to_mfn(pfn); __set_phys_to_machine(pfn, FOREIGN_FRAME(mfn)); + if (!PageHighMem(page)) + /* Just zap old mapping for now */ + pte_clear(&init_mm, address, ptep); + spin_lock_irqsave(&m2p_override_lock, flags); list_add(&page->lru, &m2p_overrides[mfn_hash(mfn)]); spin_unlock_irqrestore(&m2p_override_lock, flags); + + return 0; } -void m2p_remove_override(struct page *page) +int m2p_remove_override(struct page *page) { unsigned long flags; unsigned long mfn; unsigned long pfn; + unsigned long address; + unsigned level; + pte_t *ptep = NULL; pfn = page_to_pfn(page); mfn = get_phys_to_machine(pfn); if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT)) - return; + return -EINVAL; + + if (!PageHighMem(page)) { + address = (unsigned long)__va(pfn << PAGE_SHIFT); + ptep = lookup_address(address, &level); + + if (WARN(ptep == NULL || level != PG_LEVEL_4K, + "m2p_remove_override: pfn %lx not mapped", pfn)) + return -EINVAL; + } spin_lock_irqsave(&m2p_override_lock, flags); list_del(&page->lru); spin_unlock_irqrestore(&m2p_override_lock, flags); __set_phys_to_machine(pfn, page->index); + + if (!PageHighMem(page)) + set_pte_at(&init_mm, address, ptep, + pfn_pte(pfn, PAGE_KERNEL)); + /* No tlb flush necessary because the caller already + * left the pte unmapped. */ + + return 0; } struct page *m2p_find_override(unsigned long mfn) diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c index 1afd5690858..9ef54ebc119 100644 --- a/drivers/xen/grant-table.c +++ b/drivers/xen/grant-table.c @@ -455,6 +455,8 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops, unsigned long mfn; ret = HYPERVISOR_grant_table_op(GNTTABOP_map_grant_ref, map_ops, count); + if (ret) + return ret; for (i = 0; i < count; i++) { /* m2p override only supported for GNTMAP_contains_pte mappings */ @@ -463,7 +465,9 @@ int gnttab_map_refs(struct gnttab_map_grant_ref *map_ops, pte = (pte_t *) (mfn_to_virt(PFN_DOWN(map_ops[i].host_addr)) + (map_ops[i].host_addr & ~PAGE_MASK)); mfn = pte_mfn(*pte); - m2p_add_override(mfn, pages[i]); + ret = m2p_add_override(mfn, pages[i]); + if (ret) + return ret; } return ret; @@ -476,8 +480,14 @@ int gnttab_unmap_refs(struct gnttab_unmap_grant_ref *unmap_ops, int i, ret; ret = HYPERVISOR_grant_table_op(GNTTABOP_unmap_grant_ref, unmap_ops, count); - for (i = 0; i < count; i++) - m2p_remove_override(pages[i]); + if (ret) + return ret; + + for (i = 0; i < count; i++) { + ret = m2p_remove_override(pages[i]); + if (ret) + return ret; + } return ret; } -- cgit v1.2.3-70-g09d2