syscall: msg: locking of vm-region is now handled by channel_read_msg

when locking a pair of objects, the object with the lesser memory address
is always locked first. the pair is unlocked in the opposite order.

include/kernel/object.h

@@ -31,6 +31,20 @@ extern "C" {
 		unsigned long flags)                                           \
 	{                                                                      \
 		object_unlock_irqrestore(&p->base, flags);                     \
+	}                                                                      \
+	static inline void object_name##_lock_pair_irqsave(                    \
+		struct object_name *a,                                         \
+		struct object_name *b,                                         \
+		unsigned long *flags)                                          \
+	{                                                                      \
+		object_lock_pair_irqsave(&a->base, &b->base, flags);           \
+	}                                                                      \
+	static inline void object_name##_unlock_pair_irqrestore(               \
+		struct object_name *a,                                         \
+		struct object_name *b,                                         \
+		unsigned long flags)                                           \
+	{                                                                      \
+		object_unlock_pair_irqrestore(&a->base, &b->base, flags);      \
 	}
 
 #define OBJECT_MAGIC    0xBADDCAFE
@@ -92,6 +106,15 @@ extern void object_unlock(struct object *obj);
 extern void object_lock_irqsave(struct object *obj, unsigned long *flags);
 extern void object_unlock_irqrestore(struct object *obj, unsigned long flags);
 
+extern void object_lock_pair_irqsave(
+	struct object *a,
+	struct object *b,
+	unsigned long *flags);
+extern void object_unlock_pair_irqrestore(
+	struct object *a,
+	struct object *b,
+	unsigned long flags);
+
 #ifdef __cplusplus
 }
 #endif

kernel/channel.c

@@ -1,5 +1,6 @@
 #include <kernel/channel.h>
 #include <kernel/msg.h>
+#include <kernel/port.h>
 #include <kernel/util.h>
 #include <kernel/vm-region.h>
 
@@ -80,6 +81,7 @@ static void kmsg_reply_error(
 	unsigned long *lock_flags)
 {
 	msg->msg_status = KMSG_REPLY_SENT;
+	msg->msg_sender_port->p_status = PORT_READY;
 	msg->msg_result = status;
 	thread_awaken(msg->msg_sender_thread);
 	spin_unlock_irqrestore(&msg->msg_lock, *lock_flags);
@@ -95,6 +97,7 @@ static struct kmsg *get_next_msg(
 		spin_lock_irqsave(&msg->msg_lock, lock_flags);
 		if (msg->msg_status == KMSG_WAIT_RECEIVE) {
 			msg->msg_status = KMSG_WAIT_REPLY;
+			msg->msg_sender_port->p_status = PORT_REPLY_BLOCKED;
 			return msg;
 		}
 
@@ -149,17 +152,24 @@ extern kern_status_t channel_recv_msg(
 	struct task *sender = msg->msg_sender_thread->tr_parent;
 	struct task *receiver = self->tr_parent;
 
+	struct vm_region *src = sender->t_address_space,
+			 *dst = receiver->t_address_space;
+	unsigned long f;
+	vm_region_lock_pair_irqsave(src, dst, &f);
+
 	kern_status_t status = vm_region_memmove_v(
-		receiver->t_address_space,
+		dst,
 		0,
 		out_msg->msg_data,
 		out_msg->msg_data_count,
-		sender->t_address_space,
+		src,
 		0,
 		msg->msg_req.msg_data,
 		msg->msg_req.msg_data_count,
 		VM_REGION_COPY_ALL,
 		NULL);
+	vm_region_unlock_pair_irqrestore(src, dst, f);
+
 	if (status != KERN_OK) {
 		kmsg_reply_error(msg, status, &msg_lock_flags);
 		return status;
@@ -208,17 +218,24 @@ extern kern_status_t channel_reply_msg(
 	/* the task that is about to send the response */
 	struct task *sender = self->tr_parent;
 
+	struct vm_region *src = sender->t_address_space,
+			 *dst = receiver->t_address_space;
+	unsigned long f;
+	vm_region_lock_pair_irqsave(src, dst, &f);
+
 	kern_status_t status = vm_region_memmove_v(
-		receiver->t_address_space,
+		dst,
 		0,
 		msg->msg_resp.msg_data,
 		msg->msg_resp.msg_data_count,
-		sender->t_address_space,
+		src,
 		0,
 		resp->msg_data,
 		resp->msg_data_count,
 		VM_REGION_COPY_ALL,
 		NULL);
+	vm_region_unlock_pair_irqrestore(src, dst, f);
+
 	if (status != KERN_OK) {
 		kmsg_reply_error(msg, status, &msg_lock_flags);
 		return status;
@@ -262,17 +279,24 @@ extern kern_status_t channel_read_msg(
 		return KERN_INVALID_ARGUMENT;
 	}
 
+	struct vm_region *src_region
+		= msg->msg_sender_thread->tr_parent->t_address_space;
+
+	unsigned long f;
+	vm_region_lock_pair_irqsave(src_region, dest_region, &f);
+
 	kern_status_t status = vm_region_memmove_v(
 		dest_region,
 		0,
 		dest_iov,
 		dest_iov_count,
-		msg->msg_sender_thread->tr_parent->t_address_space,
+		src_region,
 		offset,
 		msg->msg_req.msg_data,
 		msg->msg_req.msg_data_count,
 		VM_REGION_COPY_ALL,
 		nr_read);
+	vm_region_unlock_pair_irqrestore(src_region, dest_region, f);
 
 	spin_unlock_irqrestore(&msg->msg_lock, msg_lock_flags);
 

kernel/object.c

@@ -178,6 +178,38 @@ void object_unlock_irqrestore(struct object *obj, unsigned long flags)
 	spin_unlock_irqrestore(&obj->ob_lock, flags);
 }
 
+void object_lock_pair_irqsave(
+	struct object *a,
+	struct object *b,
+	unsigned long *flags)
+{
+	if (a == b) {
+		object_lock_irqsave(a, flags);
+	} else if (a < b) {
+		object_lock_irqsave(a, flags);
+		object_lock(b);
+	} else {
+		object_lock_irqsave(b, flags);
+		object_lock(a);
+	}
+}
+
+void object_unlock_pair_irqrestore(
+	struct object *a,
+	struct object *b,
+	unsigned long flags)
+{
+	if (a == b) {
+		object_unlock_irqrestore(a, flags);
+	} else if (a < b) {
+		object_unlock(b);
+		object_unlock_irqrestore(a, flags);
+	} else {
+		object_unlock(a);
+		object_unlock_irqrestore(b, flags);
+	}
+}
+
 void *object_data(struct object *obj)
 {
 	return (char *)obj + sizeof *obj;

kernel/port.c

@@ -103,5 +103,9 @@ kern_status_t port_send_msg(
 
 	wait_for_reply(msg, lock_flags);
 
+	channel_lock_irqsave(port->p_remote, &flags);
+	btree_delete(&port->p_remote->c_msg, &msg->msg_node);
+	channel_unlock_irqrestore(port->p_remote, flags);
+
 	return msg->msg_result;
 }

syscall/msg.c

@@ -417,7 +417,6 @@ kern_status_t sys_msg_read(
 	}
 
 	channel_lock_irqsave(channel, &flags);
-	vm_region_lock(self->t_address_space);
 	status = channel_read_msg(
 		channel,
 		id,
@@ -426,7 +425,6 @@ kern_status_t sys_msg_read(
 		iov,
 		iov_count,
 		nr_read);
-	vm_region_unlock(self->t_address_space);
 	channel_unlock_irqrestore(channel, flags);
 	object_unref(channel_obj);
 

vm/vm-region.c

@@ -36,6 +36,15 @@
 	region_find_free_area_linear(region, length)
 #endif
 
+#define unlock_mapping_parent(p, root)                                         \
+	do {                                                                   \
+		struct vm_region *parent                                       \
+			= region_from_entry(p->m_entry.e_parent);              \
+		if (parent != root) {                                          \
+			vm_region_unlock(parent);                              \
+		}                                                              \
+	} while (0)
+
 /* iterates over a range of mapped virtual memory in a region, and provides
  * a moving buffer through which the memory can be accessed */
 struct vm_iterator {
@@ -296,12 +305,13 @@ static struct vm_region *region_get_child_region_recursive(
  * this function should be called with `region` locked. if a mapping is found,
  * it will be returned with its immediate parent locked. */
 static struct vm_region_mapping *region_get_mapping_recursive(
-	struct vm_region *region,
+	struct vm_region *root,
 	off_t *offp,
 	size_t len)
 {
 	off_t offset = *offp;
-	region = region_get_child_region_recursive(region, &offset, len);
+	struct vm_region *region
+		= region_get_child_region_recursive(root, &offset, len);
 	if (!region) {
 		return NULL;
 	}
@@ -311,6 +321,14 @@ static struct vm_region_mapping *region_get_mapping_recursive(
 	struct vm_region_entry *entry = region_get_entry(region, offset, len);
 	*offp = offset;
 
+	if (!entry) {
+		if (region != root) {
+			vm_region_unlock(region);
+		}
+
+		return NULL;
+	}
+
 	/* return the mapping with the parent region still locked */
 	return mapping_from_entry(entry);
 }
@@ -593,7 +611,12 @@ static void vm_iterator_begin(
 
 	off_t offset = base - vm_region_get_base_address(region);
 	it->it_mapping = region_get_mapping_recursive(region, &offset, 1);
-	if (!it->it_mapping || (it->it_mapping->m_prot & prot) != prot) {
+	if (!it->it_mapping) {
+		return;
+	}
+
+	if ((it->it_mapping->m_prot & prot) != prot) {
+		unlock_mapping_parent(it->it_mapping, region);
 		return;
 	}
 
@@ -612,6 +635,7 @@ static void vm_iterator_begin(
 	}
 
 	if (!pg) {
+		unlock_mapping_parent(it->it_mapping, region);
 		return;
 	}
 
@@ -643,15 +667,6 @@ static void vm_iterator_begin(
 
 static kern_status_t vm_iterator_seek(struct vm_iterator *it, size_t nr_bytes)
 {
-#define UNLOCK_MAPPING_PARENT(p)                                               \
-	do {                                                                   \
-		struct vm_region *parent                                       \
-			= region_from_entry(p->m_entry.e_parent);              \
-		if (parent != it->it_region) {                                 \
-			vm_region_unlock(parent);                              \
-		}                                                              \
-	} while (0)
-
 	if (nr_bytes < it->it_max) {
 		it->it_base += nr_bytes;
 		it->it_buf = (char *)it->it_buf + nr_bytes;
@@ -661,7 +676,7 @@ static kern_status_t vm_iterator_seek(struct vm_iterator *it, size_t nr_bytes)
 
 	/* the parent region of it->it_mapping is locked here. if it is
 	 * different from it->it_region, it must be unlocked */
-	UNLOCK_MAPPING_PARENT(it->it_mapping);
+	unlock_mapping_parent(it->it_mapping, it->it_region);
 
 	it->it_base += nr_bytes;
 	off_t offset = it->it_base - vm_region_get_base_address(it->it_region);
@@ -674,13 +689,13 @@ static kern_status_t vm_iterator_seek(struct vm_iterator *it, size_t nr_bytes)
 		return KERN_MEMORY_FAULT;
 	}
 
-	/* past this point, if we encounter an error, must remember to unlock
-	 * the parent region of next_mapping */
+	/* past this point, if we encounter an error, must remember to
+	 * unlock the parent region of next_mapping */
 
 	if ((next_mapping->m_prot & it->it_prot) != it->it_prot) {
 		it->it_buf = NULL;
 		it->it_max = 0;
-		UNLOCK_MAPPING_PARENT(next_mapping);
+		unlock_mapping_parent(next_mapping, it->it_region);
 		return KERN_MEMORY_FAULT;
 	}
 
@@ -699,7 +714,7 @@ static kern_status_t vm_iterator_seek(struct vm_iterator *it, size_t nr_bytes)
 	}
 
 	if (!pg) {
-		UNLOCK_MAPPING_PARENT(next_mapping);
+		unlock_mapping_parent(next_mapping, it->it_region);
 		return KERN_NO_MEMORY;
 	}
 
@@ -730,9 +745,20 @@ static kern_status_t vm_iterator_seek(struct vm_iterator *it, size_t nr_bytes)
 	return KERN_OK;
 }
 
-/* this function must be called with `root` locked. `root` will be the first
- * entry visited by the iterator. from there, child entries are visited in
- * depth-first order. */
+/* this function must be called when you are finished with a
+ * vm_iterator, to ensure that all held locks are released. */
+static void vm_iterator_finish(struct vm_iterator *it)
+{
+	if (it->it_mapping) {
+		unlock_mapping_parent(it->it_mapping, it->it_region);
+	}
+
+	memset(it, 0x0, sizeof *it);
+}
+
+/* this function must be called with `root` locked. `root` will be the
+ * first entry visited by the iterator. from there, child entries are
+ * visited in depth-first order. */
 static void entry_iterator_begin(
 	struct entry_iterator *it,
 	struct vm_region *root)
@@ -742,8 +768,8 @@ static void entry_iterator_begin(
 	it->it_entry = &root->vr_entry;
 }
 
-/* this function must be called when you are finished with an entry_iterator,
- * to ensure that all held locks are released. */
+/* this function must be called when you are finished with an
+ * entry_iterator, to ensure that all held locks are released. */
 static void entry_iterator_finish(struct entry_iterator *it)
 {
 	struct vm_region_entry *cur = it->it_entry;
@@ -771,10 +797,10 @@ static void entry_iterator_finish(struct entry_iterator *it)
 /* move to the next entry in the traversal order.
  * when this function returns:
  *   1. if the visited entry is a region, it will be locked.
- *   2. if the visited entry is a mapping, its parent region will be locked.
- * a region will remain locked until all of its children and n-grand-children
- * have been visited. once iteration is finished, only `it->it_root` will be
- * locked.
+ *   2. if the visited entry is a mapping, its parent region will be
+ * locked. a region will remain locked until all of its children and
+ * n-grand-children have been visited. once iteration is finished, only
+ * `it->it_root` will be locked.
  */
 static void entry_iterator_move_next(struct entry_iterator *it)
 {
@@ -791,8 +817,9 @@ static void entry_iterator_move_next(struct entry_iterator *it)
 		if (entry->e_type == VM_REGION_ENTRY_REGION) {
 			struct vm_region *child_region
 				= region_from_entry(entry);
-			/* since `region` is locked, interrupts are already
-			 * disabled, so don't use lock_irq() here */
+			/* since `region` is locked, interrupts are
+			 * already disabled, so don't use lock_irq()
+			 * here */
 			vm_region_lock(child_region);
 		}
 
@@ -839,8 +866,9 @@ static void entry_iterator_move_next(struct entry_iterator *it)
 	}
 }
 
-/* erase the current entry and move to the next entry in the traversal order.
- * the current entry MUST be a mapping, otherwise nothing will happen.
+/* erase the current entry and move to the next entry in the traversal
+ * order. the current entry MUST be a mapping, otherwise nothing will
+ * happen.
  */
 static void entry_iterator_erase(struct entry_iterator *it)
 {
@@ -999,7 +1027,8 @@ struct vm_region *vm_region_cast(struct object *obj)
 	return VM_REGION_CAST(obj);
 }
 
-/* this function should be called with `parent` locked (if parent is non-NULL)
+/* this function should be called with `parent` locked (if parent is
+ * non-NULL)
  */
 kern_status_t vm_region_create(
 	struct vm_region *parent,
@@ -1088,9 +1117,9 @@ kern_status_t vm_region_kill(
 			= region_from_entry(region->vr_entry.e_parent);
 		region->vr_entry.e_parent = NULL;
 
-		/* locks must be acquired in parent->child order. since we're
-		 * going backwards here, unlock `region` before locking its
-		 * parent */
+		/* locks must be acquired in parent->child order. since
+		 * we're going backwards here, unlock `region` before
+		 * locking its parent */
 		vm_region_unlock_irqrestore(region, *lock_flags);
 		vm_region_lock_irqsave(parent, lock_flags);
 		btree_delete(&parent->vr_entries, &region->vr_entry.e_node);
@@ -1175,8 +1204,8 @@ kern_status_t vm_region_map_object(
 			root,
 			&region_offset,
 			length);
-		/* if `region` != `root`, it will need to be unlocked at the end
-		 * of the function */
+		/* if `region` != `root`, it will need to be unlocked at
+		 * the end of the function */
 	}
 
 	if (region->vr_status != VM_REGION_ONLINE) {
@@ -1257,8 +1286,8 @@ kern_status_t vm_region_map_object(
 	return KERN_OK;
 }
 
-/* unmap some pages in the middle of a mapping, splitting it into two separate
- * mappings */
+/* unmap some pages in the middle of a mapping, splitting it into two
+ * separate mappings */
 static kern_status_t split_mapping(
 	struct vm_region_mapping *mapping,
 	struct vm_region *root,
@@ -1488,7 +1517,8 @@ kern_status_t vm_region_unmap(
 				unmap_area_offset,
 				unmap_area_limit);
 		} else {
-			panic("don't know what to do with this mapping");
+			panic("don't know what to do with this "
+			      "mapping");
 		}
 
 		if (delete) {
@@ -1557,15 +1587,18 @@ bool vm_region_validate_access(
 			return false;
 		}
 
-		if ((mapping->m_prot & prot) != prot) {
-			return false;
-		}
+		bool ok = (mapping->m_prot & prot) == prot;
 
 		struct vm_region *parent
 			= region_from_entry(mapping->m_entry.e_parent);
+
 		if (parent != region) {
 			vm_region_unlock(parent);
 		}
+
+		if (!ok) {
+			return false;
+		}
 	}
 
 	return true;
@@ -1671,6 +1704,8 @@ kern_status_t vm_region_read_kernel(
 		dest += to_move;
 	}
 
+	vm_iterator_finish(&src);
+
 	if (nr_read) {
 		*nr_read = r;
 	}
@@ -1727,6 +1762,9 @@ kern_status_t vm_region_memmove(
 		r += to_move;
 	}
 
+	vm_iterator_finish(&src);
+	vm_iterator_finish(&dest);
+
 	if (nr_moved) {
 		*nr_moved = r;
 	}