iommufd/fault: Use a separate spinlock to protect fault->deliver list

The fault->mutex serializes the fault read()/write() fops and the
iommufd_fault_auto_response_faults(), mainly for fault->response. Also, it
was conveniently used to fence the fault->deliver in poll() fop and
iommufd_fault_iopf_handler().

However, copy_from/to_user() may sleep if pagefaults are enabled. Thus,
they could take a long time to wait for user pages to swap in, blocking
iommufd_fault_iopf_handler() and its caller that is typically a shared IRQ
handler of an IOMMU driver, resulting in a potential global DOS.

Instead of reusing the mutex to protect the fault->deliver list, add a
separate spinlock, nested under the mutex, to do the job.
iommufd_fault_iopf_handler() would no longer be blocked by
copy_from/to_user().

Add a free_list in iommufd_auto_response_faults(), so the spinlock can
simply fence a fast list_for_each_entry_safe routine.

Provide two deliver list helpers for iommufd_fault_fops_read() to use:
 - Fetch the first iopf_group out of the fault->deliver list
 - Restore an iopf_group back to the head of the fault->deliver list

Lastly, move the mutex closer to the response in the fault structure,
and update its kdoc accordingly.

Fixes: 07838f7 ("iommufd: Add iommufd fault object")
Link: https://patch.msgid.link/r/[email protected]
Cc: [email protected]
Suggested-by: Jason Gunthorpe <[email protected]>
Reviewed-by: Kevin Tian <[email protected]>
Reviewed-by: Lu Baolu <[email protected]>
Reviewed-by: Jason Gunthorpe <[email protected]>
Signed-off-by: Nicolin Chen <[email protected]>
Signed-off-by: Jason Gunthorpe <[email protected]>

Author: nicolinc

drivers/iommu/iommufd/fault.c

@@ -103,15 +103,23 @@ static void iommufd_auto_response_faults(struct iommufd_hw_pagetable *hwpt,
 {
 	struct iommufd_fault *fault = hwpt->fault;
 	struct iopf_group *group, *next;
+	struct list_head free_list;
 	unsigned long index;
 
 	if (!fault)
 		return;
+	INIT_LIST_HEAD(&free_list);
 
 	mutex_lock(&fault->mutex);
+	spin_lock(&fault->lock);
 	list_for_each_entry_safe(group, next, &fault->deliver, node) {
 		if (group->attach_handle != &handle->handle)
 			continue;
+		list_move(&group->node, &free_list);
+	}
+	spin_unlock(&fault->lock);
+
+	list_for_each_entry_safe(group, next, &free_list, node) {
 		list_del(&group->node);
 		iopf_group_response(group, IOMMU_PAGE_RESP_INVALID);
 		iopf_free_group(group);
@@ -266,17 +274,19 @@ static ssize_t iommufd_fault_fops_read(struct file *filep, char __user *buf,
 		return -ESPIPE;
 
 	mutex_lock(&fault->mutex);
-	while (!list_empty(&fault->deliver) && count > done) {
-		group = list_first_entry(&fault->deliver,
-					 struct iopf_group, node);
-
-		if (group->fault_count * fault_size > count - done)
+	while ((group = iommufd_fault_deliver_fetch(fault))) {
+		if (done >= count ||
+		    group->fault_count * fault_size > count - done) {
+			iommufd_fault_deliver_restore(fault, group);
 			break;
+		}
 
 		rc = xa_alloc(&fault->response, &group->cookie, group,
 			      xa_limit_32b, GFP_KERNEL);
-		if (rc)
+		if (rc) {
+			iommufd_fault_deliver_restore(fault, group);
 			break;
+		}
 
 		idev = to_iommufd_handle(group->attach_handle)->idev;
 		list_for_each_entry(iopf, &group->faults, list) {
@@ -285,13 +295,12 @@ static ssize_t iommufd_fault_fops_read(struct file *filep, char __user *buf,
 						      group->cookie);
 			if (copy_to_user(buf + done, &data, fault_size)) {
 				xa_erase(&fault->response, group->cookie);
+				iommufd_fault_deliver_restore(fault, group);
 				rc = -EFAULT;
 				break;
 			}
 			done += fault_size;
 		}
-
-		list_del(&group->node);
 	}
 	mutex_unlock(&fault->mutex);
 
@@ -349,10 +358,10 @@ static __poll_t iommufd_fault_fops_poll(struct file *filep,
 	__poll_t pollflags = EPOLLOUT;
 
 	poll_wait(filep, &fault->wait_queue, wait);
-	mutex_lock(&fault->mutex);
+	spin_lock(&fault->lock);
 	if (!list_empty(&fault->deliver))
 		pollflags |= EPOLLIN | EPOLLRDNORM;
-	mutex_unlock(&fault->mutex);
+	spin_unlock(&fault->lock);
 
 	return pollflags;
 }
@@ -394,6 +403,7 @@ int iommufd_fault_alloc(struct iommufd_ucmd *ucmd)
 	INIT_LIST_HEAD(&fault->deliver);
 	xa_init_flags(&fault->response, XA_FLAGS_ALLOC1);
 	mutex_init(&fault->mutex);
+	spin_lock_init(&fault->lock);
 	init_waitqueue_head(&fault->wait_queue);
 
 	filep = anon_inode_getfile("[iommufd-pgfault]", &iommufd_fault_fops,
@@ -442,9 +452,9 @@ int iommufd_fault_iopf_handler(struct iopf_group *group)
 	hwpt = group->attach_handle->domain->fault_data;
 	fault = hwpt->fault;
 
-	mutex_lock(&fault->mutex);
+	spin_lock(&fault->lock);
 	list_add_tail(&group->node, &fault->deliver);
-	mutex_unlock(&fault->mutex);
+	spin_unlock(&fault->lock);
 
 	wake_up_interruptible(&fault->wait_queue);
 

drivers/iommu/iommufd/iommufd_private.h

@@ -443,14 +443,39 @@ struct iommufd_fault {
 	struct iommufd_ctx *ictx;
 	struct file *filep;
 
-	/* The lists of outstanding faults protected by below mutex. */
-	struct mutex mutex;
+	spinlock_t lock; /* protects the deliver list */
 	struct list_head deliver;
+	struct mutex mutex; /* serializes response flows */
 	struct xarray response;
 
 	struct wait_queue_head wait_queue;
 };
 
+/* Fetch the first node out of the fault->deliver list */
+static inline struct iopf_group *
+iommufd_fault_deliver_fetch(struct iommufd_fault *fault)
+{
+	struct list_head *list = &fault->deliver;
+	struct iopf_group *group = NULL;
+
+	spin_lock(&fault->lock);
+	if (!list_empty(list)) {
+		group = list_first_entry(list, struct iopf_group, node);
+		list_del(&group->node);
+	}
+	spin_unlock(&fault->lock);
+	return group;
+}
+
+/* Restore a node back to the head of the fault->deliver list */
+static inline void iommufd_fault_deliver_restore(struct iommufd_fault *fault,
+						 struct iopf_group *group)
+{
+	spin_lock(&fault->lock);
+	list_add(&group->node, &fault->deliver);
+	spin_unlock(&fault->lock);
+}
+
 struct iommufd_attach_handle {
 	struct iommu_attach_handle handle;
 	struct iommufd_device *idev;