Merge tag 'ceph-for-4.11-rc1' of git://github.com/ceph/ceph-client

Pull ceph updates from Ilya Dryomov:
 "This time around we have:

   - support for rbd data-pool feature, which enables rbd images on
     erasure-coded pools (myself). CEPH_PG_MAX_SIZE has been bumped to
     allow erasure-coded profiles with k+m up to 32.

   - a patch for ceph_d_revalidate() performance regression introduced
     in 4.9, along with some cleanups in the area (Jeff Layton)

   - a set of fixes for unsafe ->d_parent accesses in CephFS (Jeff
     Layton)

   - buffered reads are now processed in rsize windows instead of rasize
     windows (Andreas Gerstmayr). The new default for rsize mount option
     is 64M.

   - ack vs commit distinction is gone, greatly simplifying ->fsync()
     and MOSDOpReply handling code (myself)

  ... also a few filesystem bug fixes from Zheng, a CRUSH sync up (CRUSH
  computations are still serialized though) and several minor fixes and
  cleanups all over"

* tag 'ceph-for-4.11-rc1' of git://github.com/ceph/ceph-client: (52 commits)
  libceph, rbd, ceph: WRITE | ONDISK -> WRITE
  libceph: get rid of ack vs commit
  ceph: remove special ack vs commit behavior
  ceph: tidy some white space in get_nonsnap_parent()
  crush: fix dprintk compilation
  crush: do is_out test only if we do not collide
  ceph: remove req from unsafe list when unregistering it
  rbd: constify device_type structure
  rbd: kill obj_request->object_name and rbd_segment_name_cache
  rbd: store and use obj_request->object_no
  rbd: RBD_V{1,2}_DATA_FORMAT macros
  rbd: factor out __rbd_osd_req_create()
  rbd: set offset and length outside of rbd_obj_request_create()
  rbd: support for data-pool feature
  rbd: introduce rbd_init_layout()
  rbd: use rbd_obj_bytes() more
  rbd: remove now unused rbd_obj_request_wait() and helpers
  rbd: switch rbd_obj_method_sync() to ceph_osdc_call()
  libceph: pass reply buffer length through ceph_osdc_call()
  rbd: do away with obj_request in rbd_obj_read_sync()
  ...

Documentation/filesystems/ceph.txt

@@ -98,11 +98,10 @@ Mount Options
 	size.
 
   rsize=X
-	Specify the maximum read size in bytes.  By default there is no
-	maximum.
+	Specify the maximum read size in bytes.  Default: 64 MB.
 
   rasize=X
-	Specify the maximum readahead.
+	Specify the maximum readahead.  Default: 8 MB.
 
   mount_timeout=X
 	Specify the timeout value for mount (in seconds), in the case

drivers/block/rbd.c

@@ -123,9 +123,11 @@ static int atomic_dec_return_safe(atomic_t *v)
 #define RBD_FEATURE_LAYERING	(1<<0)
 #define RBD_FEATURE_STRIPINGV2	(1<<1)
 #define RBD_FEATURE_EXCLUSIVE_LOCK (1<<2)
+#define RBD_FEATURE_DATA_POOL (1<<7)
 #define RBD_FEATURES_ALL	(RBD_FEATURE_LAYERING |		\
 				 RBD_FEATURE_STRIPINGV2 |	\
-				 RBD_FEATURE_EXCLUSIVE_LOCK)
+				 RBD_FEATURE_EXCLUSIVE_LOCK |	\
+				 RBD_FEATURE_DATA_POOL)
 
 /* Features supported by this (client software) implementation. */
 
@@ -144,10 +146,9 @@ struct rbd_image_header {
 	/* These six fields never change for a given rbd image */
 	char *object_prefix;
 	__u8 obj_order;
-	__u8 crypt_type;
-	__u8 comp_type;
 	u64 stripe_unit;
 	u64 stripe_count;
+	s64 data_pool_id;
 	u64 features;		/* Might be changeable someday? */
 
 	/* The remaining fields need to be updated occasionally */
@@ -230,7 +231,7 @@ enum obj_req_flags {
 };
 
 struct rbd_obj_request {
-	const char		*object_name;
+	u64			object_no;
 	u64			offset;		/* object start byte */
 	u64			length;		/* bytes from offset */
 	unsigned long		flags;
@@ -438,7 +439,6 @@ static DEFINE_SPINLOCK(rbd_client_list_lock);
 
 static struct kmem_cache	*rbd_img_request_cache;
 static struct kmem_cache	*rbd_obj_request_cache;
-static struct kmem_cache	*rbd_segment_name_cache;
 
 static int rbd_major;
 static DEFINE_IDA(rbd_dev_id_ida);
@@ -972,6 +972,30 @@ static bool rbd_dev_ondisk_valid(struct rbd_image_header_ondisk *ondisk)
 	return true;
 }
 
+/*
+ * returns the size of an object in the image
+ */
+static u32 rbd_obj_bytes(struct rbd_image_header *header)
+{
+	return 1U << header->obj_order;
+}
+
+static void rbd_init_layout(struct rbd_device *rbd_dev)
+{
+	if (rbd_dev->header.stripe_unit == 0 ||
+	    rbd_dev->header.stripe_count == 0) {
+		rbd_dev->header.stripe_unit = rbd_obj_bytes(&rbd_dev->header);
+		rbd_dev->header.stripe_count = 1;
+	}
+
+	rbd_dev->layout.stripe_unit = rbd_dev->header.stripe_unit;
+	rbd_dev->layout.stripe_count = rbd_dev->header.stripe_count;
+	rbd_dev->layout.object_size = rbd_obj_bytes(&rbd_dev->header);
+	rbd_dev->layout.pool_id = rbd_dev->header.data_pool_id == CEPH_NOPOOL ?
+			  rbd_dev->spec->pool_id : rbd_dev->header.data_pool_id;
+	RCU_INIT_POINTER(rbd_dev->layout.pool_ns, NULL);
+}
+
 /*
  * Fill an rbd image header with information from the given format 1
  * on-disk header.
@@ -992,15 +1016,11 @@ static int rbd_header_from_disk(struct rbd_device *rbd_dev,
 	/* Allocate this now to avoid having to handle failure below */
 
 	if (first_time) {
-		size_t len;
-
-		len = strnlen(ondisk->object_prefix,
-				sizeof (ondisk->object_prefix));
-		object_prefix = kmalloc(len + 1, GFP_KERNEL);
+		object_prefix = kstrndup(ondisk->object_prefix,
+					 sizeof(ondisk->object_prefix),
+					 GFP_KERNEL);
 		if (!object_prefix)
 			return -ENOMEM;
-		memcpy(object_prefix, ondisk->object_prefix, len);
-		object_prefix[len] = '\0';
 	}
 
 	/* Allocate the snapshot context and fill it in */
@@ -1051,12 +1071,7 @@ static int rbd_header_from_disk(struct rbd_device *rbd_dev,
 	if (first_time) {
 		header->object_prefix = object_prefix;
 		header->obj_order = ondisk->options.order;
-		header->crypt_type = ondisk->options.crypt_type;
-		header->comp_type = ondisk->options.comp_type;
-		/* The rest aren't used for format 1 images */
-		header->stripe_unit = 0;
-		header->stripe_count = 0;
-		header->features = 0;
+		rbd_init_layout(rbd_dev);
 	} else {
 		ceph_put_snap_context(header->snapc);
 		kfree(header->snap_names);
@@ -1232,42 +1247,9 @@ static void rbd_dev_mapping_clear(struct rbd_device *rbd_dev)
 	rbd_dev->mapping.features = 0;
 }
 
-static void rbd_segment_name_free(const char *name)
-{
-	/* The explicit cast here is needed to drop the const qualifier */
-
-	kmem_cache_free(rbd_segment_name_cache, (void *)name);
-}
-
-static const char *rbd_segment_name(struct rbd_device *rbd_dev, u64 offset)
-{
-	char *name;
-	u64 segment;
-	int ret;
-	char *name_format;
-
-	name = kmem_cache_alloc(rbd_segment_name_cache, GFP_NOIO);
-	if (!name)
-		return NULL;
-	segment = offset >> rbd_dev->header.obj_order;
-	name_format = "%s.%012llx";
-	if (rbd_dev->image_format == 2)
-		name_format = "%s.%016llx";
-	ret = snprintf(name, CEPH_MAX_OID_NAME_LEN + 1, name_format,
-			rbd_dev->header.object_prefix, segment);
-	if (ret < 0 || ret > CEPH_MAX_OID_NAME_LEN) {
-		pr_err("error formatting segment name for #%llu (%d)\n",
-			segment, ret);
-		rbd_segment_name_free(name);
-		name = NULL;
-	}
-
-	return name;
-}
-
 static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset)
 {
-	u64 segment_size = (u64) 1 << rbd_dev->header.obj_order;
+	u64 segment_size = rbd_obj_bytes(&rbd_dev->header);
 
 	return offset & (segment_size - 1);
 }
@@ -1275,7 +1257,7 @@ static u64 rbd_segment_offset(struct rbd_device *rbd_dev, u64 offset)
 static u64 rbd_segment_length(struct rbd_device *rbd_dev,
 				u64 offset, u64 length)
 {
-	u64 segment_size = (u64) 1 << rbd_dev->header.obj_order;
+	u64 segment_size = rbd_obj_bytes(&rbd_dev->header);
 
 	offset &= segment_size - 1;
 
@@ -1286,14 +1268,6 @@ static u64 rbd_segment_length(struct rbd_device *rbd_dev,
 	return length;
 }
 
-/*
- * returns the size of an object in the image
- */
-static u64 rbd_obj_bytes(struct rbd_image_header *header)
-{
-	return 1 << header->obj_order;
-}
-
 /*
  * bio helpers
  */
@@ -1623,7 +1597,9 @@ static void rbd_obj_request_submit(struct rbd_obj_request *obj_request)
 {
 	struct ceph_osd_request *osd_req = obj_request->osd_req;
 
-	dout("%s %p osd_req %p\n", __func__, obj_request, osd_req);
+	dout("%s %p object_no %016llx %llu~%llu osd_req %p\n", __func__,
+	     obj_request, obj_request->object_no, obj_request->offset,
+	     obj_request->length, osd_req);
 	if (obj_request_img_data_test(obj_request)) {
 		WARN_ON(obj_request->callback != rbd_img_obj_callback);
 		rbd_img_request_get(obj_request->img_request);
@@ -1631,44 +1607,6 @@ static void rbd_obj_request_submit(struct rbd_obj_request *obj_request)
 	ceph_osdc_start_request(osd_req->r_osdc, osd_req, false);
 }
 
-static void rbd_obj_request_end(struct rbd_obj_request *obj_request)
-{
-	dout("%s %p\n", __func__, obj_request);
-	ceph_osdc_cancel_request(obj_request->osd_req);
-}
-
-/*
- * Wait for an object request to complete.  If interrupted, cancel the
- * underlying osd request.
- *
- * @timeout: in jiffies, 0 means "wait forever"
- */
-static int __rbd_obj_request_wait(struct rbd_obj_request *obj_request,
-				  unsigned long timeout)
-{
-	long ret;
-
-	dout("%s %p\n", __func__, obj_request);
-	ret = wait_for_completion_interruptible_timeout(
-					&obj_request->completion,
-					ceph_timeout_jiffies(timeout));
-	if (ret <= 0) {
-		if (ret == 0)
-			ret = -ETIMEDOUT;
-		rbd_obj_request_end(obj_request);
-	} else {
-		ret = 0;
-	}
-
-	dout("%s %p ret %d\n", __func__, obj_request, (int)ret);
-	return ret;
-}
-
-static int rbd_obj_request_wait(struct rbd_obj_request *obj_request)
-{
-	return __rbd_obj_request_wait(obj_request, 0);
-}
-
 static void rbd_img_request_complete(struct rbd_img_request *img_request)
 {
 
@@ -1955,8 +1893,8 @@ static void rbd_osd_req_callback(struct ceph_osd_request *osd_req)
 		rbd_osd_call_callback(obj_request);
 		break;
 	default:
-		rbd_warn(NULL, "%s: unsupported op %hu",
-			obj_request->object_name, (unsigned short) opcode);
+		rbd_warn(NULL, "unexpected OSD op: object_no %016llx opcode %d",
+			 obj_request->object_no, opcode);
 		break;
 	}
 
@@ -1980,6 +1918,40 @@ static void rbd_osd_req_format_write(struct rbd_obj_request *obj_request)
 	osd_req->r_data_offset = obj_request->offset;
 }
 
+static struct ceph_osd_request *
+__rbd_osd_req_create(struct rbd_device *rbd_dev,
+		     struct ceph_snap_context *snapc,
+		     int num_ops, unsigned int flags,
+		     struct rbd_obj_request *obj_request)
+{
+	struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
+	struct ceph_osd_request *req;
+	const char *name_format = rbd_dev->image_format == 1 ?
+				      RBD_V1_DATA_FORMAT : RBD_V2_DATA_FORMAT;
+
+	req = ceph_osdc_alloc_request(osdc, snapc, num_ops, false, GFP_NOIO);
+	if (!req)
+		return NULL;
+
+	req->r_flags = flags;
+	req->r_callback = rbd_osd_req_callback;
+	req->r_priv = obj_request;
+
+	req->r_base_oloc.pool = rbd_dev->layout.pool_id;
+	if (ceph_oid_aprintf(&req->r_base_oid, GFP_NOIO, name_format,
+			rbd_dev->header.object_prefix, obj_request->object_no))
+		goto err_req;
+
+	if (ceph_osdc_alloc_messages(req, GFP_NOIO))
+		goto err_req;
+
+	return req;
+
+err_req:
+	ceph_osdc_put_request(req);
+	return NULL;
+}
+
 /*
  * Create an osd request.  A read request has one osd op (read).
  * A write request has either one (watch) or two (hint+write) osd ops.
@@ -1993,8 +1965,6 @@ static struct ceph_osd_request *rbd_osd_req_create(
 					struct rbd_obj_request *obj_request)
 {
 	struct ceph_snap_context *snapc = NULL;
-	struct ceph_osd_client *osdc;
-	struct ceph_osd_request *osd_req;
 
 	if (obj_request_img_data_test(obj_request) &&
 		(op_type == OBJ_OP_DISCARD || op_type == OBJ_OP_WRITE)) {
@@ -2009,35 +1979,9 @@ static struct ceph_osd_request *rbd_osd_req_create(
 
 	rbd_assert(num_ops == 1 || ((op_type == OBJ_OP_WRITE) && num_ops == 2));
 
-	/* Allocate and initialize the request, for the num_ops ops */
-
-	osdc = &rbd_dev->rbd_client->client->osdc;
-	osd_req = ceph_osdc_alloc_request(osdc, snapc, num_ops, false,
-					  GFP_NOIO);
-	if (!osd_req)
-		goto fail;
-
-	if (op_type == OBJ_OP_WRITE || op_type == OBJ_OP_DISCARD)
-		osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK;
-	else
-		osd_req->r_flags = CEPH_OSD_FLAG_READ;
-
-	osd_req->r_callback = rbd_osd_req_callback;
-	osd_req->r_priv = obj_request;
-
-	osd_req->r_base_oloc.pool = rbd_dev->layout.pool_id;
-	if (ceph_oid_aprintf(&osd_req->r_base_oid, GFP_NOIO, "%s",
-			     obj_request->object_name))
-		goto fail;
-
-	if (ceph_osdc_alloc_messages(osd_req, GFP_NOIO))
-		goto fail;
-
-	return osd_req;
-
-fail:
-	ceph_osdc_put_request(osd_req);
-	return NULL;
+	return __rbd_osd_req_create(rbd_dev, snapc, num_ops,
+	    (op_type == OBJ_OP_WRITE || op_type == OBJ_OP_DISCARD) ?
+	    CEPH_OSD_FLAG_WRITE : CEPH_OSD_FLAG_READ, obj_request);
 }
 
 /*
@@ -2050,10 +1994,6 @@ static struct ceph_osd_request *
 rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request)
 {
 	struct rbd_img_request *img_request;
-	struct ceph_snap_context *snapc;
-	struct rbd_device *rbd_dev;
-	struct ceph_osd_client *osdc;
-	struct ceph_osd_request *osd_req;
 	int num_osd_ops = 3;
 
 	rbd_assert(obj_request_img_data_test(obj_request));
@@ -2065,77 +2005,34 @@ rbd_osd_req_create_copyup(struct rbd_obj_request *obj_request)
 	if (img_request_discard_test(img_request))
 		num_osd_ops = 2;
 
-	/* Allocate and initialize the request, for all the ops */
-
-	snapc = img_request->snapc;
-	rbd_dev = img_request->rbd_dev;
-	osdc = &rbd_dev->rbd_client->client->osdc;
-	osd_req = ceph_osdc_alloc_request(osdc, snapc, num_osd_ops,
-						false, GFP_NOIO);
-	if (!osd_req)
-		goto fail;
-
-	osd_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK;
-	osd_req->r_callback = rbd_osd_req_callback;
-	osd_req->r_priv = obj_request;
-
-	osd_req->r_base_oloc.pool = rbd_dev->layout.pool_id;
-	if (ceph_oid_aprintf(&osd_req->r_base_oid, GFP_NOIO, "%s",
-			     obj_request->object_name))
-		goto fail;
-
-	if (ceph_osdc_alloc_messages(osd_req, GFP_NOIO))
-		goto fail;
-
-	return osd_req;
-
-fail:
-	ceph_osdc_put_request(osd_req);
-	return NULL;
+	return __rbd_osd_req_create(img_request->rbd_dev,
+				    img_request->snapc, num_osd_ops,
+				    CEPH_OSD_FLAG_WRITE, obj_request);
 }
 
-
 static void rbd_osd_req_destroy(struct ceph_osd_request *osd_req)
 {
 	ceph_osdc_put_request(osd_req);
 }
 
-/* object_name is assumed to be a non-null pointer and NUL-terminated */
-
-static struct rbd_obj_request *rbd_obj_request_create(const char *object_name,
-						u64 offset, u64 length,
-						enum obj_request_type type)
+static struct rbd_obj_request *
+rbd_obj_request_create(enum obj_request_type type)
 {
 	struct rbd_obj_request *obj_request;
-	size_t size;
-	char *name;
 
 	rbd_assert(obj_request_type_valid(type));
 
-	size = strlen(object_name) + 1;
-	name = kmalloc(size, GFP_NOIO);
-	if (!name)
-		return NULL;
-
 	obj_request = kmem_cache_zalloc(rbd_obj_request_cache, GFP_NOIO);
-	if (!obj_request) {
-		kfree(name);
+	if (!obj_request)
 		return NULL;
-	}
 
-	obj_request->object_name = memcpy(name, object_name, size);
-	obj_request->offset = offset;
-	obj_request->length = length;
-	obj_request->flags = 0;
 	obj_request->which = BAD_WHICH;
 	obj_request->type = type;
 	INIT_LIST_HEAD(&obj_request->links);
 	init_completion(&obj_request->completion);
 	kref_init(&obj_request->kref);
 
-	dout("%s: \"%s\" %llu/%llu %d -> obj %p\n", __func__, object_name,
-		offset, length, (int)type, obj_request);
-
+	dout("%s %p\n", __func__, obj_request);
 	return obj_request;
 }
 
@@ -2170,8 +2067,6 @@ static void rbd_obj_request_destroy(struct kref *kref)
 		break;
 	}
 
-	kfree(obj_request->object_name);
-	obj_request->object_name = NULL;
 	kmem_cache_free(rbd_obj_request_cache, obj_request);
 }
 
@@ -2546,22 +2441,18 @@ static int rbd_img_request_fill(struct rbd_img_request *img_request,
 
 	while (resid) {
 		struct ceph_osd_request *osd_req;
-		const char *object_name;
-		u64 offset;
-		u64 length;
+		u64 object_no = img_offset >> rbd_dev->header.obj_order;
+		u64 offset = rbd_segment_offset(rbd_dev, img_offset);
+		u64 length = rbd_segment_length(rbd_dev, img_offset, resid);
 
-		object_name = rbd_segment_name(rbd_dev, img_offset);
-		if (!object_name)
-			goto out_unwind;
-		offset = rbd_segment_offset(rbd_dev, img_offset);
-		length = rbd_segment_length(rbd_dev, img_offset, resid);
-		obj_request = rbd_obj_request_create(object_name,
-						offset, length, type);
-		/* object request has its own copy of the object name */
-		rbd_segment_name_free(object_name);
+		obj_request = rbd_obj_request_create(type);
 		if (!obj_request)
 			goto out_unwind;
 
+		obj_request->object_no = object_no;
+		obj_request->offset = offset;
+		obj_request->length = length;
+
 		/*
 		 * set obj_request->img_request before creating the
 		 * osd_request so that it gets the right snapc
@@ -2771,7 +2662,7 @@ static int rbd_img_obj_parent_read_full(struct rbd_obj_request *obj_request)
 	 * child image to which the original request was to be sent.
 	 */
 	img_offset = obj_request->img_offset - obj_request->offset;
-	length = (u64)1 << rbd_dev->header.obj_order;
+	length = rbd_obj_bytes(&rbd_dev->header);
 
 	/*
 	 * There is no defined parent data beyond the parent
@@ -2900,11 +2791,12 @@ static int rbd_img_obj_exists_submit(struct rbd_obj_request *obj_request)
 	size_t size;
 	int ret;
 
-	stat_request = rbd_obj_request_create(obj_request->object_name, 0, 0,
-					      OBJ_REQUEST_PAGES);
+	stat_request = rbd_obj_request_create(OBJ_REQUEST_PAGES);
 	if (!stat_request)
 		return -ENOMEM;
 
+	stat_request->object_no = obj_request->object_no;
+
 	stat_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1,
 						   stat_request);
 	if (!stat_request->osd_req) {
@@ -3983,17 +3875,17 @@ out:
  * returned in the outbound buffer, or a negative error code.
  */
 static int rbd_obj_method_sync(struct rbd_device *rbd_dev,
-			     const char *object_name,
-			     const char *class_name,
+			     struct ceph_object_id *oid,
+			     struct ceph_object_locator *oloc,
 			     const char *method_name,
 			     const void *outbound,
 			     size_t outbound_size,
 			     void *inbound,
 			     size_t inbound_size)
 {
-	struct rbd_obj_request *obj_request;
-	struct page **pages;
-	u32 page_count;
+	struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
+	struct page *req_page = NULL;
+	struct page *reply_page;
 	int ret;
 
 	/*
@@ -4003,61 +3895,35 @@ static int rbd_obj_method_sync(struct rbd_device *rbd_dev,
 	 * method.  Currently if this is present it will be a
 	 * snapshot id.
 	 */
-	page_count = (u32)calc_pages_for(0, inbound_size);
-	pages = ceph_alloc_page_vector(page_count, GFP_KERNEL);
-	if (IS_ERR(pages))
-		return PTR_ERR(pages);
-
-	ret = -ENOMEM;
-	obj_request = rbd_obj_request_create(object_name, 0, inbound_size,
-							OBJ_REQUEST_PAGES);
-	if (!obj_request)
-		goto out;
+	if (outbound) {
+		if (outbound_size > PAGE_SIZE)
+			return -E2BIG;
 
-	obj_request->pages = pages;
-	obj_request->page_count = page_count;
-
-	obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1,
-						  obj_request);
-	if (!obj_request->osd_req)
-		goto out;
-
-	osd_req_op_cls_init(obj_request->osd_req, 0, CEPH_OSD_OP_CALL,
-					class_name, method_name);
-	if (outbound_size) {
-		struct ceph_pagelist *pagelist;
-
-		pagelist = kmalloc(sizeof (*pagelist), GFP_NOFS);
-		if (!pagelist)
-			goto out;
+		req_page = alloc_page(GFP_KERNEL);
+		if (!req_page)
+			return -ENOMEM;
 
-		ceph_pagelist_init(pagelist);
-		ceph_pagelist_append(pagelist, outbound, outbound_size);
-		osd_req_op_cls_request_data_pagelist(obj_request->osd_req, 0,
-						pagelist);
+		memcpy(page_address(req_page), outbound, outbound_size);
 	}
-	osd_req_op_cls_response_data_pages(obj_request->osd_req, 0,
-					obj_request->pages, inbound_size,
-					0, false, false);
-
-	rbd_obj_request_submit(obj_request);
-	ret = rbd_obj_request_wait(obj_request);
-	if (ret)
-		goto out;
 
-	ret = obj_request->result;
-	if (ret < 0)
-		goto out;
+	reply_page = alloc_page(GFP_KERNEL);
+	if (!reply_page) {
+		if (req_page)
+			__free_page(req_page);
+		return -ENOMEM;
+	}
 
-	rbd_assert(obj_request->xferred < (u64)INT_MAX);
-	ret = (int)obj_request->xferred;
-	ceph_copy_from_page_vector(pages, inbound, 0, obj_request->xferred);
-out:
-	if (obj_request)
-		rbd_obj_request_put(obj_request);
-	else
-		ceph_release_page_vector(pages, page_count);
+	ret = ceph_osdc_call(osdc, oid, oloc, RBD_DRV_NAME, method_name,
+			     CEPH_OSD_FLAG_READ, req_page, outbound_size,
+			     reply_page, &inbound_size);
+	if (!ret) {
+		memcpy(inbound, page_address(reply_page), inbound_size);
+		ret = inbound_size;
+	}
 
+	if (req_page)
+		__free_page(req_page);
+	__free_page(reply_page);
 	return ret;
 }
 
@@ -4256,63 +4122,46 @@ static void rbd_free_disk(struct rbd_device *rbd_dev)
 }
 
 static int rbd_obj_read_sync(struct rbd_device *rbd_dev,
-				const char *object_name,
-				u64 offset, u64 length, void *buf)
+			     struct ceph_object_id *oid,
+			     struct ceph_object_locator *oloc,
+			     void *buf, int buf_len)
 
 {
-	struct rbd_obj_request *obj_request;
-	struct page **pages = NULL;
-	u32 page_count;
-	size_t size;
+	struct ceph_osd_client *osdc = &rbd_dev->rbd_client->client->osdc;
+	struct ceph_osd_request *req;
+	struct page **pages;
+	int num_pages = calc_pages_for(0, buf_len);
 	int ret;
 
-	page_count = (u32) calc_pages_for(offset, length);
-	pages = ceph_alloc_page_vector(page_count, GFP_KERNEL);
-	if (IS_ERR(pages))
-		return PTR_ERR(pages);
-
-	ret = -ENOMEM;
-	obj_request = rbd_obj_request_create(object_name, offset, length,
-							OBJ_REQUEST_PAGES);
-	if (!obj_request)
-		goto out;
-
-	obj_request->pages = pages;
-	obj_request->page_count = page_count;
-
-	obj_request->osd_req = rbd_osd_req_create(rbd_dev, OBJ_OP_READ, 1,
-						  obj_request);
-	if (!obj_request->osd_req)
-		goto out;
+	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_KERNEL);
+	if (!req)
+		return -ENOMEM;
 
-	osd_req_op_extent_init(obj_request->osd_req, 0, CEPH_OSD_OP_READ,
-					offset, length, 0, 0);
-	osd_req_op_extent_osd_data_pages(obj_request->osd_req, 0,
-					obj_request->pages,
-					obj_request->length,
-					obj_request->offset & ~PAGE_MASK,
-					false, false);
+	ceph_oid_copy(&req->r_base_oid, oid);
+	ceph_oloc_copy(&req->r_base_oloc, oloc);
+	req->r_flags = CEPH_OSD_FLAG_READ;
 
-	rbd_obj_request_submit(obj_request);
-	ret = rbd_obj_request_wait(obj_request);
+	ret = ceph_osdc_alloc_messages(req, GFP_KERNEL);
 	if (ret)
-		goto out;
+		goto out_req;
 
-	ret = obj_request->result;
-	if (ret < 0)
-		goto out;
+	pages = ceph_alloc_page_vector(num_pages, GFP_KERNEL);
+	if (IS_ERR(pages)) {
+		ret = PTR_ERR(pages);
+		goto out_req;
+	}
 
-	rbd_assert(obj_request->xferred <= (u64) SIZE_MAX);
-	size = (size_t) obj_request->xferred;
-	ceph_copy_from_page_vector(pages, buf, 0, size);
-	rbd_assert(size <= (size_t)INT_MAX);
-	ret = (int)size;
-out:
-	if (obj_request)
-		rbd_obj_request_put(obj_request);
-	else
-		ceph_release_page_vector(pages, page_count);
+	osd_req_op_extent_init(req, 0, CEPH_OSD_OP_READ, 0, buf_len, 0, 0);
+	osd_req_op_extent_osd_data_pages(req, 0, pages, buf_len, 0, false,
+					 true);
+
+	ceph_osdc_start_request(osdc, req, false);
+	ret = ceph_osdc_wait_request(osdc, req);
+	if (ret >= 0)
+		ceph_copy_from_page_vector(pages, buf, 0, ret);
 
+out_req:
+	ceph_osdc_put_request(req);
 	return ret;
 }
 
@@ -4348,8 +4197,8 @@ static int rbd_dev_v1_header_info(struct rbd_device *rbd_dev)
 		if (!ondisk)
 			return -ENOMEM;
 
-		ret = rbd_obj_read_sync(rbd_dev, rbd_dev->header_oid.name,
-				       0, size, ondisk);
+		ret = rbd_obj_read_sync(rbd_dev, &rbd_dev->header_oid,
+					&rbd_dev->header_oloc, ondisk, size);
 		if (ret < 0)
 			goto out;
 		if ((size_t)ret < size) {
@@ -4781,7 +4630,7 @@ static const struct attribute_group *rbd_attr_groups[] = {
 
 static void rbd_dev_release(struct device *dev);
 
-static struct device_type rbd_device_type = {
+static const struct device_type rbd_device_type = {
 	.name		= "rbd",
 	.groups		= rbd_attr_groups,
 	.release	= rbd_dev_release,
@@ -4876,8 +4725,9 @@ static struct rbd_device *__rbd_dev_create(struct rbd_client *rbdc,
 	INIT_LIST_HEAD(&rbd_dev->node);
 	init_rwsem(&rbd_dev->header_rwsem);
 
+	rbd_dev->header.data_pool_id = CEPH_NOPOOL;
 	ceph_oid_init(&rbd_dev->header_oid);
-	ceph_oloc_init(&rbd_dev->header_oloc);
+	rbd_dev->header_oloc.pool = spec->pool_id;
 
 	mutex_init(&rbd_dev->watch_mutex);
 	rbd_dev->watch_state = RBD_WATCH_STATE_UNREGISTERED;
@@ -4899,12 +4749,6 @@ static struct rbd_device *__rbd_dev_create(struct rbd_client *rbdc,
 	rbd_dev->rbd_client = rbdc;
 	rbd_dev->spec = spec;
 
-	rbd_dev->layout.stripe_unit = 1 << RBD_MAX_OBJ_ORDER;
-	rbd_dev->layout.stripe_count = 1;
-	rbd_dev->layout.object_size = 1 << RBD_MAX_OBJ_ORDER;
-	rbd_dev->layout.pool_id = spec->pool_id;
-	RCU_INIT_POINTER(rbd_dev->layout.pool_ns, NULL);
-
 	return rbd_dev;
 }
 
@@ -4970,10 +4814,10 @@ static int _rbd_dev_v2_snap_size(struct rbd_device *rbd_dev, u64 snap_id,
 		__le64 size;
 	} __attribute__ ((packed)) size_buf = { 0 };
 
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_size",
-				&snapid, sizeof (snapid),
-				&size_buf, sizeof (size_buf));
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_size",
+				  &snapid, sizeof(snapid),
+				  &size_buf, sizeof(size_buf));
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0)
 		return ret;
@@ -5010,9 +4854,9 @@ static int rbd_dev_v2_object_prefix(struct rbd_device *rbd_dev)
 	if (!reply_buf)
 		return -ENOMEM;
 
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_object_prefix", NULL, 0,
-				reply_buf, RBD_OBJ_PREFIX_LEN_MAX);
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_object_prefix",
+				  NULL, 0, reply_buf, RBD_OBJ_PREFIX_LEN_MAX);
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0)
 		goto out;
@@ -5045,10 +4889,10 @@ static int _rbd_dev_v2_snap_features(struct rbd_device *rbd_dev, u64 snap_id,
 	u64 unsup;
 	int ret;
 
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_features",
-				&snapid, sizeof (snapid),
-				&features_buf, sizeof (features_buf));
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_features",
+				  &snapid, sizeof(snapid),
+				  &features_buf, sizeof(features_buf));
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0)
 		return ret;
@@ -5107,10 +4951,9 @@ static int rbd_dev_v2_parent_info(struct rbd_device *rbd_dev)
 	}
 
 	snapid = cpu_to_le64(rbd_dev->spec->snap_id);
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_parent",
-				&snapid, sizeof (snapid),
-				reply_buf, size);
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_parent",
+				  &snapid, sizeof(snapid), reply_buf, size);
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0)
 		goto out_err;
@@ -5210,9 +5053,9 @@ static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev)
 	u64 stripe_count;
 	int ret;
 
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_stripe_unit_count", NULL, 0,
-				(char *)&striping_info_buf, size);
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				&rbd_dev->header_oloc, "get_stripe_unit_count",
+				NULL, 0, &striping_info_buf, size);
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0)
 		return ret;
@@ -5226,7 +5069,7 @@ static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev)
 	 * out, and only fail if the image has non-default values.
 	 */
 	ret = -EINVAL;
-	obj_size = (u64)1 << rbd_dev->header.obj_order;
+	obj_size = rbd_obj_bytes(&rbd_dev->header);
 	p = &striping_info_buf;
 	stripe_unit = ceph_decode_64(&p);
 	if (stripe_unit != obj_size) {
@@ -5247,8 +5090,27 @@ static int rbd_dev_v2_striping_info(struct rbd_device *rbd_dev)
 	return 0;
 }
 
+static int rbd_dev_v2_data_pool(struct rbd_device *rbd_dev)
+{
+	__le64 data_pool_id;
+	int ret;
+
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_data_pool",
+				  NULL, 0, &data_pool_id, sizeof(data_pool_id));
+	if (ret < 0)
+		return ret;
+	if (ret < sizeof(data_pool_id))
+		return -EBADMSG;
+
+	rbd_dev->header.data_pool_id = le64_to_cpu(data_pool_id);
+	WARN_ON(rbd_dev->header.data_pool_id == CEPH_NOPOOL);
+	return 0;
+}
+
 static char *rbd_dev_image_name(struct rbd_device *rbd_dev)
 {
+	CEPH_DEFINE_OID_ONSTACK(oid);
 	size_t image_id_size;
 	char *image_id;
 	void *p;
@@ -5276,10 +5138,10 @@ static char *rbd_dev_image_name(struct rbd_device *rbd_dev)
 	if (!reply_buf)
 		goto out;
 
-	ret = rbd_obj_method_sync(rbd_dev, RBD_DIRECTORY,
-				"rbd", "dir_get_name",
-				image_id, image_id_size,
-				reply_buf, size);
+	ceph_oid_printf(&oid, "%s", RBD_DIRECTORY);
+	ret = rbd_obj_method_sync(rbd_dev, &oid, &rbd_dev->header_oloc,
+				  "dir_get_name", image_id, image_id_size,
+				  reply_buf, size);
 	if (ret < 0)
 		goto out;
 	p = reply_buf;
@@ -5458,9 +5320,9 @@ static int rbd_dev_v2_snap_context(struct rbd_device *rbd_dev)
 	if (!reply_buf)
 		return -ENOMEM;
 
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_snapcontext", NULL, 0,
-				reply_buf, size);
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_snapcontext",
+				  NULL, 0, reply_buf, size);
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0)
 		goto out;
@@ -5523,10 +5385,9 @@ static const char *rbd_dev_v2_snap_name(struct rbd_device *rbd_dev,
 		return ERR_PTR(-ENOMEM);
 
 	snapid = cpu_to_le64(snap_id);
-	ret = rbd_obj_method_sync(rbd_dev, rbd_dev->header_oid.name,
-				"rbd", "get_snapshot_name",
-				&snapid, sizeof (snapid),
-				reply_buf, size);
+	ret = rbd_obj_method_sync(rbd_dev, &rbd_dev->header_oid,
+				  &rbd_dev->header_oloc, "get_snapshot_name",
+				  &snapid, sizeof(snapid), reply_buf, size);
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret < 0) {
 		snap_name = ERR_PTR(ret);
@@ -5833,7 +5694,7 @@ static int rbd_dev_image_id(struct rbd_device *rbd_dev)
 {
 	int ret;
 	size_t size;
-	char *object_name;
+	CEPH_DEFINE_OID_ONSTACK(oid);
 	void *response;
 	char *image_id;
 
@@ -5853,12 +5714,12 @@ static int rbd_dev_image_id(struct rbd_device *rbd_dev)
 	 * First, see if the format 2 image id file exists, and if
 	 * so, get the image's persistent id from it.
 	 */
-	size = sizeof (RBD_ID_PREFIX) + strlen(rbd_dev->spec->image_name);
-	object_name = kmalloc(size, GFP_NOIO);
-	if (!object_name)
-		return -ENOMEM;
-	sprintf(object_name, "%s%s", RBD_ID_PREFIX, rbd_dev->spec->image_name);
-	dout("rbd id object name is %s\n", object_name);
+	ret = ceph_oid_aprintf(&oid, GFP_KERNEL, "%s%s", RBD_ID_PREFIX,
+			       rbd_dev->spec->image_name);
+	if (ret)
+		return ret;
+
+	dout("rbd id object name is %s\n", oid.name);
 
 	/* Response will be an encoded string, which includes a length */
 
@@ -5871,9 +5732,9 @@ static int rbd_dev_image_id(struct rbd_device *rbd_dev)
 
 	/* If it doesn't exist we'll assume it's a format 1 image */
 
-	ret = rbd_obj_method_sync(rbd_dev, object_name,
-				"rbd", "get_id", NULL, 0,
-				response, RBD_IMAGE_ID_LEN_MAX);
+	ret = rbd_obj_method_sync(rbd_dev, &oid, &rbd_dev->header_oloc,
+				  "get_id", NULL, 0,
+				  response, RBD_IMAGE_ID_LEN_MAX);
 	dout("%s: rbd_obj_method_sync returned %d\n", __func__, ret);
 	if (ret == -ENOENT) {
 		image_id = kstrdup("", GFP_KERNEL);
@@ -5896,8 +5757,7 @@ static int rbd_dev_image_id(struct rbd_device *rbd_dev)
 	}
 out:
 	kfree(response);
-	kfree(object_name);
-
+	ceph_oid_destroy(&oid);
 	return ret;
 }
 
@@ -5944,14 +5804,20 @@ static int rbd_dev_v2_header_onetime(struct rbd_device *rbd_dev)
 		if (ret < 0)
 			goto out_err;
 	}
-	/* No support for crypto and compression type format 2 images */
 
+	if (rbd_dev->header.features & RBD_FEATURE_DATA_POOL) {
+		ret = rbd_dev_v2_data_pool(rbd_dev);
+		if (ret)
+			goto out_err;
+	}
+
+	rbd_init_layout(rbd_dev);
 	return 0;
+
 out_err:
 	rbd_dev->header.features = 0;
 	kfree(rbd_dev->header.object_prefix);
 	rbd_dev->header.object_prefix = NULL;
-
 	return ret;
 }
 
@@ -6077,8 +5943,6 @@ static int rbd_dev_header_name(struct rbd_device *rbd_dev)
 	/* Record the header object name for this rbd image. */
 
 	rbd_assert(rbd_image_format_valid(rbd_dev->image_format));
-
-	rbd_dev->header_oloc.pool = rbd_dev->layout.pool_id;
 	if (rbd_dev->image_format == 1)
 		ret = ceph_oid_aprintf(&rbd_dev->header_oid, GFP_KERNEL, "%s%s",
 				       spec->image_name, RBD_SUFFIX);
@@ -6471,27 +6335,16 @@ static int rbd_slab_init(void)
 	if (!rbd_obj_request_cache)
 		goto out_err;
 
-	rbd_assert(!rbd_segment_name_cache);
-	rbd_segment_name_cache = kmem_cache_create("rbd_segment_name",
-					CEPH_MAX_OID_NAME_LEN + 1, 1, 0, NULL);
-	if (rbd_segment_name_cache)
-		return 0;
-out_err:
-	kmem_cache_destroy(rbd_obj_request_cache);
-	rbd_obj_request_cache = NULL;
+	return 0;
 
+out_err:
 	kmem_cache_destroy(rbd_img_request_cache);
 	rbd_img_request_cache = NULL;
-
 	return -ENOMEM;
 }
 
 static void rbd_slab_exit(void)
 {
-	rbd_assert(rbd_segment_name_cache);
-	kmem_cache_destroy(rbd_segment_name_cache);
-	rbd_segment_name_cache = NULL;
-
 	rbd_assert(rbd_obj_request_cache);
 	kmem_cache_destroy(rbd_obj_request_cache);
 	rbd_obj_request_cache = NULL;

drivers/block/rbd_types.h

@@ -25,8 +25,8 @@
  */
 
 #define RBD_HEADER_PREFIX      "rbd_header."
-#define RBD_DATA_PREFIX        "rbd_data."
 #define RBD_ID_PREFIX          "rbd_id."
+#define RBD_V2_DATA_FORMAT     "%s.%016llx"
 
 #define RBD_LOCK_NAME          "rbd_lock"
 #define RBD_LOCK_TAG           "internal"
@@ -42,13 +42,14 @@ enum rbd_notify_op {
 /*
  * For format version 1, rbd image 'foo' consists of objects
  *   foo.rbd		- image metadata
- *   rb.<idhi>.<idlo>.00000000
- *   rb.<idhi>.<idlo>.00000001
+ *   rb.<idhi>.<idlo>.<extra>.000000000000
+ *   rb.<idhi>.<idlo>.<extra>.000000000001
  *   ...		- data
  * There is no notion of a persistent image id in rbd format 1.
  */
 
 #define RBD_SUFFIX		".rbd"
+#define RBD_V1_DATA_FORMAT	"%s.%012llx"
 
 #define RBD_DIRECTORY           "rbd_directory"
 #define RBD_INFO                "rbd_info"
@@ -57,9 +58,6 @@ enum rbd_notify_op {
 #define RBD_MIN_OBJ_ORDER       16
 #define RBD_MAX_OBJ_ORDER       30
 
-#define RBD_COMP_NONE		0
-#define RBD_CRYPT_NONE		0
-
 #define RBD_HEADER_TEXT		"<<< Rados Block Device Image >>>\n"
 #define RBD_HEADER_SIGNATURE	"RBD"
 #define RBD_HEADER_VERSION	"001.005"

fs/ceph/addr.c

@@ -391,6 +391,7 @@ static int start_read(struct inode *inode, struct list_head *page_list, int max)
 			nr_pages = i;
 			if (nr_pages > 0) {
 				len = nr_pages << PAGE_SHIFT;
+				osd_req_op_extent_update(req, 0, len);
 				break;
 			}
 			goto out_pages;
@@ -771,7 +772,7 @@ static int ceph_writepages_start(struct address_space *mapping,
 	     wbc->sync_mode == WB_SYNC_NONE ? "NONE" :
 	     (wbc->sync_mode == WB_SYNC_ALL ? "ALL" : "HOLD"));
 
-	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
+	if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
 		if (ci->i_wrbuffer_ref > 0) {
 			pr_warn_ratelimited(
 				"writepage_start %p %lld forced umount\n",
@@ -1017,8 +1018,7 @@ new_request:
 					&ci->i_layout, vino,
 					offset, &len, 0, num_ops,
 					CEPH_OSD_OP_WRITE,
-					CEPH_OSD_FLAG_WRITE |
-					CEPH_OSD_FLAG_ONDISK,
+					CEPH_OSD_FLAG_WRITE,
 					snapc, truncate_seq,
 					truncate_size, false);
 		if (IS_ERR(req)) {
@@ -1028,8 +1028,7 @@ new_request:
 						min(num_ops,
 						    CEPH_OSD_SLAB_OPS),
 						CEPH_OSD_OP_WRITE,
-						CEPH_OSD_FLAG_WRITE |
-						CEPH_OSD_FLAG_ONDISK,
+						CEPH_OSD_FLAG_WRITE,
 						snapc, truncate_seq,
 						truncate_size, true);
 			BUG_ON(IS_ERR(req));
@@ -1194,7 +1193,7 @@ static int ceph_update_writeable_page(struct file *file,
 	int r;
 	struct ceph_snap_context *snapc, *oldest;
 
-	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
+	if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
 		dout(" page %p forced umount\n", page);
 		unlock_page(page);
 		return -EIO;
@@ -1681,8 +1680,7 @@ int ceph_uninline_data(struct file *filp, struct page *locked_page)
 
 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
 				    ceph_vino(inode), 0, &len, 0, 1,
-				    CEPH_OSD_OP_CREATE,
-				    CEPH_OSD_FLAG_ONDISK | CEPH_OSD_FLAG_WRITE,
+				    CEPH_OSD_OP_CREATE, CEPH_OSD_FLAG_WRITE,
 				    NULL, 0, 0, false);
 	if (IS_ERR(req)) {
 		err = PTR_ERR(req);
@@ -1699,8 +1697,7 @@ int ceph_uninline_data(struct file *filp, struct page *locked_page)
 
 	req = ceph_osdc_new_request(&fsc->client->osdc, &ci->i_layout,
 				    ceph_vino(inode), 0, &len, 1, 3,
-				    CEPH_OSD_OP_WRITE,
-				    CEPH_OSD_FLAG_ONDISK | CEPH_OSD_FLAG_WRITE,
+				    CEPH_OSD_OP_WRITE, CEPH_OSD_FLAG_WRITE,
 				    NULL, ci->i_truncate_seq,
 				    ci->i_truncate_size, false);
 	if (IS_ERR(req)) {
@@ -1873,7 +1870,7 @@ static int __ceph_pool_perm_get(struct ceph_inode_info *ci,
 		goto out_unlock;
 	}
 
-	wr_req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ACK;
+	wr_req->r_flags = CEPH_OSD_FLAG_WRITE;
 	osd_req_op_init(wr_req, 0, CEPH_OSD_OP_CREATE, CEPH_OSD_OP_FLAG_EXCL);
 	ceph_oloc_copy(&wr_req->r_base_oloc, &rd_req->r_base_oloc);
 	ceph_oid_copy(&wr_req->r_base_oid, &rd_req->r_base_oid);

fs/ceph/cache.c

@@ -234,7 +234,7 @@ void ceph_fscache_file_set_cookie(struct inode *inode, struct file *filp)
 		fscache_enable_cookie(ci->fscache, ceph_fscache_can_enable,
 				inode);
 		if (fscache_cookie_enabled(ci->fscache)) {
-			dout("fscache_file_set_cookie %p %p enabing cache\n",
+			dout("fscache_file_set_cookie %p %p enabling cache\n",
 			     inode, filp);
 		}
 	}

fs/ceph/caps.c

@@ -867,7 +867,7 @@ int __ceph_caps_file_wanted(struct ceph_inode_info *ci)
 /*
  * Return caps we have registered with the MDS(s) as 'wanted'.
  */
-int __ceph_caps_mds_wanted(struct ceph_inode_info *ci)
+int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check)
 {
 	struct ceph_cap *cap;
 	struct rb_node *p;
@@ -875,7 +875,7 @@ int __ceph_caps_mds_wanted(struct ceph_inode_info *ci)
 
 	for (p = rb_first(&ci->i_caps); p; p = rb_next(p)) {
 		cap = rb_entry(p, struct ceph_cap, ci_node);
-		if (!__cap_is_valid(cap))
+		if (check && !__cap_is_valid(cap))
 			continue;
 		if (cap == ci->i_auth_cap)
 			mds_wanted |= cap->mds_wanted;
@@ -1184,6 +1184,13 @@ static int __send_cap(struct ceph_mds_client *mdsc, struct ceph_cap *cap,
 		delayed = 1;
 	}
 	ci->i_ceph_flags &= ~(CEPH_I_NODELAY | CEPH_I_FLUSH);
+	if (want & ~cap->mds_wanted) {
+		/* user space may open/close single file frequently.
+		 * This avoids droping mds_wanted immediately after
+		 * requesting new mds_wanted.
+		 */
+		__cap_set_timeouts(mdsc, ci);
+	}
 
 	cap->issued &= retain;  /* drop bits we don't want */
 	if (cap->implemented & ~cap->issued) {
@@ -2084,8 +2091,6 @@ int ceph_fsync(struct file *file, loff_t start, loff_t end, int datasync)
 
 	dout("fsync %p%s\n", inode, datasync ? " datasync" : "");
 
-	ceph_sync_write_wait(inode);
-
 	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
 	if (ret < 0)
 		goto out;
@@ -2477,23 +2482,22 @@ again:
 
 		if (ci->i_ceph_flags & CEPH_I_CAP_DROPPED) {
 			int mds_wanted;
-			if (ACCESS_ONCE(mdsc->fsc->mount_state) ==
+			if (READ_ONCE(mdsc->fsc->mount_state) ==
 			    CEPH_MOUNT_SHUTDOWN) {
 				dout("get_cap_refs %p forced umount\n", inode);
 				*err = -EIO;
 				ret = 1;
 				goto out_unlock;
 			}
-			mds_wanted = __ceph_caps_mds_wanted(ci);
-			if ((mds_wanted & need) != need) {
+			mds_wanted = __ceph_caps_mds_wanted(ci, false);
+			if (need & ~(mds_wanted & need)) {
 				dout("get_cap_refs %p caps were dropped"
 				     " (session killed?)\n", inode);
 				*err = -ESTALE;
 				ret = 1;
 				goto out_unlock;
 			}
-			if ((mds_wanted & file_wanted) ==
-			    (file_wanted & (CEPH_CAP_FILE_RD|CEPH_CAP_FILE_WR)))
+			if (!(file_wanted & ~mds_wanted))
 				ci->i_ceph_flags &= ~CEPH_I_CAP_DROPPED;
 		}
 
@@ -3404,6 +3408,7 @@ retry:
 			tcap->implemented |= issued;
 			if (cap == ci->i_auth_cap)
 				ci->i_auth_cap = tcap;
+
 			if (!list_empty(&ci->i_cap_flush_list) &&
 			    ci->i_auth_cap == tcap) {
 				spin_lock(&mdsc->cap_dirty_lock);
@@ -3417,9 +3422,18 @@ retry:
 	} else if (tsession) {
 		/* add placeholder for the export tagert */
 		int flag = (cap == ci->i_auth_cap) ? CEPH_CAP_FLAG_AUTH : 0;
+		tcap = new_cap;
 		ceph_add_cap(inode, tsession, t_cap_id, -1, issued, 0,
 			     t_seq - 1, t_mseq, (u64)-1, flag, &new_cap);
 
+		if (!list_empty(&ci->i_cap_flush_list) &&
+		    ci->i_auth_cap == tcap) {
+			spin_lock(&mdsc->cap_dirty_lock);
+			list_move_tail(&ci->i_flushing_item,
+				       &tcap->session->s_cap_flushing);
+			spin_unlock(&mdsc->cap_dirty_lock);
+		}
+
 		__ceph_remove_cap(cap, false);
 		goto out_unlock;
 	}
@@ -3924,9 +3938,10 @@ int ceph_encode_inode_release(void **p, struct inode *inode,
 }
 
 int ceph_encode_dentry_release(void **p, struct dentry *dentry,
+			       struct inode *dir,
 			       int mds, int drop, int unless)
 {
-	struct inode *dir = d_inode(dentry->d_parent);
+	struct dentry *parent = NULL;
 	struct ceph_mds_request_release *rel = *p;
 	struct ceph_dentry_info *di = ceph_dentry(dentry);
 	int force = 0;
@@ -3941,9 +3956,14 @@ int ceph_encode_dentry_release(void **p, struct dentry *dentry,
 	spin_lock(&dentry->d_lock);
 	if (di->lease_session && di->lease_session->s_mds == mds)
 		force = 1;
+	if (!dir) {
+		parent = dget(dentry->d_parent);
+		dir = d_inode(parent);
+	}
 	spin_unlock(&dentry->d_lock);
 
 	ret = ceph_encode_inode_release(p, dir, mds, drop, unless, force);
+	dput(parent);
 
 	spin_lock(&dentry->d_lock);
 	if (ret && di->lease_session && di->lease_session->s_mds == mds) {

fs/ceph/debugfs.c

@@ -70,7 +70,7 @@ static int mdsc_show(struct seq_file *s, void *p)
 
 		seq_printf(s, "%s", ceph_mds_op_name(req->r_op));
 
-		if (req->r_got_unsafe)
+		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
 			seq_puts(s, "\t(unsafe)");
 		else
 			seq_puts(s, "\t");

fs/ceph/dir.c

@@ -371,7 +371,7 @@ more:
 		/* hints to request -> mds selection code */
 		req->r_direct_mode = USE_AUTH_MDS;
 		req->r_direct_hash = ceph_frag_value(frag);
-		req->r_direct_is_hash = true;
+		__set_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
 		if (fi->last_name) {
 			req->r_path2 = kstrdup(fi->last_name, GFP_KERNEL);
 			if (!req->r_path2) {
@@ -417,7 +417,7 @@ more:
 		fi->frag = frag;
 		fi->last_readdir = req;
 
-		if (req->r_did_prepopulate) {
+		if (test_bit(CEPH_MDS_R_DID_PREPOPULATE, &req->r_req_flags)) {
 			fi->readdir_cache_idx = req->r_readdir_cache_idx;
 			if (fi->readdir_cache_idx < 0) {
 				/* preclude from marking dir ordered */
@@ -752,7 +752,8 @@ static struct dentry *ceph_lookup(struct inode *dir, struct dentry *dentry,
 		mask |= CEPH_CAP_XATTR_SHARED;
 	req->r_args.getattr.mask = cpu_to_le32(mask);
 
-	req->r_locked_dir = dir;
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	err = ceph_mdsc_do_request(mdsc, NULL, req);
 	err = ceph_handle_snapdir(req, dentry, err);
 	dentry = ceph_finish_lookup(req, dentry, err);
@@ -813,7 +814,8 @@ static int ceph_mknod(struct inode *dir, struct dentry *dentry,
 	}
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
-	req->r_locked_dir = dir;
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_args.mknod.mode = cpu_to_le32(mode);
 	req->r_args.mknod.rdev = cpu_to_le32(rdev);
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
@@ -864,7 +866,8 @@ static int ceph_symlink(struct inode *dir, struct dentry *dentry,
 		ceph_mdsc_put_request(req);
 		goto out;
 	}
-	req->r_locked_dir = dir;
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
@@ -913,7 +916,8 @@ static int ceph_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
 
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
-	req->r_locked_dir = dir;
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_args.mkdir.mode = cpu_to_le32(mode);
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
@@ -957,7 +961,8 @@ static int ceph_link(struct dentry *old_dentry, struct inode *dir,
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
 	req->r_old_dentry = dget(old_dentry);
-	req->r_locked_dir = dir;
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	/* release LINK_SHARED on source inode (mds will lock it) */
@@ -1023,7 +1028,8 @@ static int ceph_unlink(struct inode *dir, struct dentry *dentry)
 	}
 	req->r_dentry = dget(dentry);
 	req->r_num_caps = 2;
-	req->r_locked_dir = dir;
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
 	req->r_dentry_unless = CEPH_CAP_FILE_EXCL;
 	req->r_inode_drop = drop_caps_for_unlink(inode);
@@ -1066,7 +1072,8 @@ static int ceph_rename(struct inode *old_dir, struct dentry *old_dentry,
 	req->r_num_caps = 2;
 	req->r_old_dentry = dget(old_dentry);
 	req->r_old_dentry_dir = old_dir;
-	req->r_locked_dir = new_dir;
+	req->r_parent = new_dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_old_dentry_drop = CEPH_CAP_FILE_SHARED;
 	req->r_old_dentry_unless = CEPH_CAP_FILE_EXCL;
 	req->r_dentry_drop = CEPH_CAP_FILE_SHARED;
@@ -1194,7 +1201,7 @@ static int ceph_d_revalidate(struct dentry *dentry, unsigned int flags)
 	struct inode *dir;
 
 	if (flags & LOOKUP_RCU) {
-		parent = ACCESS_ONCE(dentry->d_parent);
+		parent = READ_ONCE(dentry->d_parent);
 		dir = d_inode_rcu(parent);
 		if (!dir)
 			return -ECHILD;
@@ -1237,11 +1244,12 @@ static int ceph_d_revalidate(struct dentry *dentry, unsigned int flags)
 			return -ECHILD;
 
 		op = ceph_snap(dir) == CEPH_SNAPDIR ?
-			CEPH_MDS_OP_LOOKUPSNAP : CEPH_MDS_OP_GETATTR;
+			CEPH_MDS_OP_LOOKUPSNAP : CEPH_MDS_OP_LOOKUP;
 		req = ceph_mdsc_create_request(mdsc, op, USE_ANY_MDS);
 		if (!IS_ERR(req)) {
 			req->r_dentry = dget(dentry);
-			req->r_num_caps = op == CEPH_MDS_OP_GETATTR ? 1 : 2;
+			req->r_num_caps = 2;
+			req->r_parent = dir;
 
 			mask = CEPH_STAT_CAP_INODE | CEPH_CAP_AUTH_SHARED;
 			if (ceph_security_xattr_wanted(dir))

fs/ceph/export.c

@@ -207,7 +207,8 @@ static int ceph_get_name(struct dentry *parent, char *name,
 	req->r_inode = d_inode(child);
 	ihold(d_inode(child));
 	req->r_ino2 = ceph_vino(d_inode(parent));
-	req->r_locked_dir = d_inode(parent);
+	req->r_parent = d_inode(parent);
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	req->r_num_caps = 2;
 	err = ceph_mdsc_do_request(mdsc, NULL, req);
 

fs/ceph/file.c

@@ -283,7 +283,7 @@ int ceph_open(struct inode *inode, struct file *file)
 	spin_lock(&ci->i_ceph_lock);
 	if (__ceph_is_any_real_caps(ci) &&
 	    (((fmode & CEPH_FILE_MODE_WR) == 0) || ci->i_auth_cap)) {
-		int mds_wanted = __ceph_caps_mds_wanted(ci);
+		int mds_wanted = __ceph_caps_mds_wanted(ci, true);
 		int issued = __ceph_caps_issued(ci, NULL);
 
 		dout("open %p fmode %d want %s issued %s using existing\n",
@@ -379,7 +379,8 @@ int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
                mask |= CEPH_CAP_XATTR_SHARED;
        req->r_args.open.mask = cpu_to_le32(mask);
 
-	req->r_locked_dir = dir;           /* caller holds dir->i_mutex */
+	req->r_parent = dir;
+	set_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags);
 	err = ceph_mdsc_do_request(mdsc,
 				   (flags & (O_CREAT|O_TRUNC)) ? dir : NULL,
 				   req);
@@ -758,9 +759,7 @@ static void ceph_aio_retry_work(struct work_struct *work)
 		goto out;
 	}
 
-	req->r_flags =	CEPH_OSD_FLAG_ORDERSNAP |
-			CEPH_OSD_FLAG_ONDISK |
-			CEPH_OSD_FLAG_WRITE;
+	req->r_flags = CEPH_OSD_FLAG_ORDERSNAP | CEPH_OSD_FLAG_WRITE;
 	ceph_oloc_copy(&req->r_base_oloc, &orig_req->r_base_oloc);
 	ceph_oid_copy(&req->r_base_oid, &orig_req->r_base_oid);
 
@@ -794,89 +793,6 @@ out:
 	kfree(aio_work);
 }
 
-/*
- * Write commit request unsafe callback, called to tell us when a
- * request is unsafe (that is, in flight--has been handed to the
- * messenger to send to its target osd).  It is called again when
- * we've received a response message indicating the request is
- * "safe" (its CEPH_OSD_FLAG_ONDISK flag is set), or when a request
- * is completed early (and unsuccessfully) due to a timeout or
- * interrupt.
- *
- * This is used if we requested both an ACK and ONDISK commit reply
- * from the OSD.
- */
-static void ceph_sync_write_unsafe(struct ceph_osd_request *req, bool unsafe)
-{
-	struct ceph_inode_info *ci = ceph_inode(req->r_inode);
-
-	dout("%s %p tid %llu %ssafe\n", __func__, req, req->r_tid,
-		unsafe ? "un" : "");
-	if (unsafe) {
-		ceph_get_cap_refs(ci, CEPH_CAP_FILE_WR);
-		spin_lock(&ci->i_unsafe_lock);
-		list_add_tail(&req->r_unsafe_item,
-			      &ci->i_unsafe_writes);
-		spin_unlock(&ci->i_unsafe_lock);
-
-		complete_all(&req->r_completion);
-	} else {
-		spin_lock(&ci->i_unsafe_lock);
-		list_del_init(&req->r_unsafe_item);
-		spin_unlock(&ci->i_unsafe_lock);
-		ceph_put_cap_refs(ci, CEPH_CAP_FILE_WR);
-	}
-}
-
-/*
- * Wait on any unsafe replies for the given inode.  First wait on the
- * newest request, and make that the upper bound.  Then, if there are
- * more requests, keep waiting on the oldest as long as it is still older
- * than the original request.
- */
-void ceph_sync_write_wait(struct inode *inode)
-{
-	struct ceph_inode_info *ci = ceph_inode(inode);
-	struct list_head *head = &ci->i_unsafe_writes;
-	struct ceph_osd_request *req;
-	u64 last_tid;
-
-	if (!S_ISREG(inode->i_mode))
-		return;
-
-	spin_lock(&ci->i_unsafe_lock);
-	if (list_empty(head))
-		goto out;
-
-	/* set upper bound as _last_ entry in chain */
-
-	req = list_last_entry(head, struct ceph_osd_request,
-			      r_unsafe_item);
-	last_tid = req->r_tid;
-
-	do {
-		ceph_osdc_get_request(req);
-		spin_unlock(&ci->i_unsafe_lock);
-
-		dout("sync_write_wait on tid %llu (until %llu)\n",
-		     req->r_tid, last_tid);
-		wait_for_completion(&req->r_done_completion);
-		ceph_osdc_put_request(req);
-
-		spin_lock(&ci->i_unsafe_lock);
-		/*
-		 * from here on look at first entry in chain, since we
-		 * only want to wait for anything older than last_tid
-		 */
-		if (list_empty(head))
-			break;
-		req = list_first_entry(head, struct ceph_osd_request,
-				       r_unsafe_item);
-	} while (req->r_tid < last_tid);
-out:
-	spin_unlock(&ci->i_unsafe_lock);
-}
-
 static ssize_t
 ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter,
 		       struct ceph_snap_context *snapc,
@@ -915,9 +831,7 @@ ceph_direct_read_write(struct kiocb *iocb, struct iov_iter *iter,
 		if (ret2 < 0)
 			dout("invalidate_inode_pages2_range returned %d\n", ret2);
 
-		flags = CEPH_OSD_FLAG_ORDERSNAP |
-			CEPH_OSD_FLAG_ONDISK |
-			CEPH_OSD_FLAG_WRITE;
+		flags = CEPH_OSD_FLAG_ORDERSNAP | CEPH_OSD_FLAG_WRITE;
 	} else {
 		flags = CEPH_OSD_FLAG_READ;
 	}
@@ -1116,10 +1030,7 @@ ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos,
 	if (ret < 0)
 		dout("invalidate_inode_pages2_range returned %d\n", ret);
 
-	flags = CEPH_OSD_FLAG_ORDERSNAP |
-		CEPH_OSD_FLAG_ONDISK |
-		CEPH_OSD_FLAG_WRITE |
-		CEPH_OSD_FLAG_ACK;
+	flags = CEPH_OSD_FLAG_ORDERSNAP | CEPH_OSD_FLAG_WRITE;
 
 	while ((len = iov_iter_count(from)) > 0) {
 		size_t left;
@@ -1165,8 +1076,6 @@ ceph_sync_write(struct kiocb *iocb, struct iov_iter *from, loff_t pos,
 			goto out;
 		}
 
-		/* get a second commit callback */
-		req->r_unsafe_callback = ceph_sync_write_unsafe;
 		req->r_inode = inode;
 
 		osd_req_op_extent_osd_data_pages(req, 0, pages, len, 0,
@@ -1616,8 +1525,7 @@ static int ceph_zero_partial_object(struct inode *inode,
 					ceph_vino(inode),
 					offset, length,
 					0, 1, op,
-					CEPH_OSD_FLAG_WRITE |
-					CEPH_OSD_FLAG_ONDISK,
+					CEPH_OSD_FLAG_WRITE,
 					NULL, 0, 0, false);
 	if (IS_ERR(req)) {
 		ret = PTR_ERR(req);

fs/ceph/inode.c

@@ -499,7 +499,6 @@ struct inode *ceph_alloc_inode(struct super_block *sb)
 	ci->i_rdcache_gen = 0;
 	ci->i_rdcache_revoking = 0;
 
-	INIT_LIST_HEAD(&ci->i_unsafe_writes);
 	INIT_LIST_HEAD(&ci->i_unsafe_dirops);
 	INIT_LIST_HEAD(&ci->i_unsafe_iops);
 	spin_lock_init(&ci->i_unsafe_lock);
@@ -583,14 +582,6 @@ int ceph_drop_inode(struct inode *inode)
 	return 1;
 }
 
-void ceph_evict_inode(struct inode *inode)
-{
-	/* wait unsafe sync writes */
-	ceph_sync_write_wait(inode);
-	truncate_inode_pages_final(&inode->i_data);
-	clear_inode(inode);
-}
-
 static inline blkcnt_t calc_inode_blocks(u64 size)
 {
 	return (size + (1<<9) - 1) >> 9;
@@ -1016,7 +1007,9 @@ out:
 static void update_dentry_lease(struct dentry *dentry,
 				struct ceph_mds_reply_lease *lease,
 				struct ceph_mds_session *session,
-				unsigned long from_time)
+				unsigned long from_time,
+				struct ceph_vino *tgt_vino,
+				struct ceph_vino *dir_vino)
 {
 	struct ceph_dentry_info *di = ceph_dentry(dentry);
 	long unsigned duration = le32_to_cpu(lease->duration_ms);
@@ -1024,13 +1017,27 @@ static void update_dentry_lease(struct dentry *dentry,
 	long unsigned half_ttl = from_time + (duration * HZ / 2) / 1000;
 	struct inode *dir;
 
+	/*
+	 * Make sure dentry's inode matches tgt_vino. NULL tgt_vino means that
+	 * we expect a negative dentry.
+	 */
+	if (!tgt_vino && d_really_is_positive(dentry))
+		return;
+
+	if (tgt_vino && (d_really_is_negative(dentry) ||
+			!ceph_ino_compare(d_inode(dentry), tgt_vino)))
+		return;
+
 	spin_lock(&dentry->d_lock);
 	dout("update_dentry_lease %p duration %lu ms ttl %lu\n",
 	     dentry, duration, ttl);
 
-	/* make lease_rdcache_gen match directory */
 	dir = d_inode(dentry->d_parent);
 
+	/* make sure parent matches dir_vino */
+	if (!ceph_ino_compare(dir, dir_vino))
+		goto out_unlock;
+
 	/* only track leases on regular dentries */
 	if (ceph_snap(dir) != CEPH_NOSNAP)
 		goto out_unlock;
@@ -1108,61 +1115,27 @@ out:
  *
  * Called with snap_rwsem (read).
  */
-int ceph_fill_trace(struct super_block *sb, struct ceph_mds_request *req,
-		    struct ceph_mds_session *session)
+int ceph_fill_trace(struct super_block *sb, struct ceph_mds_request *req)
 {
+	struct ceph_mds_session *session = req->r_session;
 	struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
 	struct inode *in = NULL;
-	struct ceph_vino vino;
+	struct ceph_vino tvino, dvino;
 	struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
 	int err = 0;
 
 	dout("fill_trace %p is_dentry %d is_target %d\n", req,
 	     rinfo->head->is_dentry, rinfo->head->is_target);
 
-#if 0
-	/*
-	 * Debugging hook:
-	 *
-	 * If we resend completed ops to a recovering mds, we get no
-	 * trace.  Since that is very rare, pretend this is the case
-	 * to ensure the 'no trace' handlers in the callers behave.
-	 *
-	 * Fill in inodes unconditionally to avoid breaking cap
-	 * invariants.
-	 */
-	if (rinfo->head->op & CEPH_MDS_OP_WRITE) {
-		pr_info("fill_trace faking empty trace on %lld %s\n",
-			req->r_tid, ceph_mds_op_name(rinfo->head->op));
-		if (rinfo->head->is_dentry) {
-			rinfo->head->is_dentry = 0;
-			err = fill_inode(req->r_locked_dir,
-					 &rinfo->diri, rinfo->dirfrag,
-					 session, req->r_request_started, -1);
-		}
-		if (rinfo->head->is_target) {
-			rinfo->head->is_target = 0;
-			ininfo = rinfo->targeti.in;
-			vino.ino = le64_to_cpu(ininfo->ino);
-			vino.snap = le64_to_cpu(ininfo->snapid);
-			in = ceph_get_inode(sb, vino);
-			err = fill_inode(in, &rinfo->targeti, NULL,
-					 session, req->r_request_started,
-					 req->r_fmode);
-			iput(in);
-		}
-	}
-#endif
-
 	if (!rinfo->head->is_target && !rinfo->head->is_dentry) {
 		dout("fill_trace reply is empty!\n");
-		if (rinfo->head->result == 0 && req->r_locked_dir)
+		if (rinfo->head->result == 0 && req->r_parent)
 			ceph_invalidate_dir_request(req);
 		return 0;
 	}
 
 	if (rinfo->head->is_dentry) {
-		struct inode *dir = req->r_locked_dir;
+		struct inode *dir = req->r_parent;
 
 		if (dir) {
 			err = fill_inode(dir, NULL,
@@ -1188,8 +1161,8 @@ int ceph_fill_trace(struct super_block *sb, struct ceph_mds_request *req,
 			dname.name = rinfo->dname;
 			dname.len = rinfo->dname_len;
 			dname.hash = full_name_hash(parent, dname.name, dname.len);
-			vino.ino = le64_to_cpu(rinfo->targeti.in->ino);
-			vino.snap = le64_to_cpu(rinfo->targeti.in->snapid);
+			tvino.ino = le64_to_cpu(rinfo->targeti.in->ino);
+			tvino.snap = le64_to_cpu(rinfo->targeti.in->snapid);
 retry_lookup:
 			dn = d_lookup(parent, &dname);
 			dout("d_lookup on parent=%p name=%.*s got %p\n",
@@ -1206,8 +1179,8 @@ retry_lookup:
 				}
 				err = 0;
 			} else if (d_really_is_positive(dn) &&
-				   (ceph_ino(d_inode(dn)) != vino.ino ||
-				    ceph_snap(d_inode(dn)) != vino.snap)) {
+				   (ceph_ino(d_inode(dn)) != tvino.ino ||
+				    ceph_snap(d_inode(dn)) != tvino.snap)) {
 				dout(" dn %p points to wrong inode %p\n",
 				     dn, d_inode(dn));
 				d_delete(dn);
@@ -1221,10 +1194,10 @@ retry_lookup:
 	}
 
 	if (rinfo->head->is_target) {
-		vino.ino = le64_to_cpu(rinfo->targeti.in->ino);
-		vino.snap = le64_to_cpu(rinfo->targeti.in->snapid);
+		tvino.ino = le64_to_cpu(rinfo->targeti.in->ino);
+		tvino.snap = le64_to_cpu(rinfo->targeti.in->snapid);
 
-		in = ceph_get_inode(sb, vino);
+		in = ceph_get_inode(sb, tvino);
 		if (IS_ERR(in)) {
 			err = PTR_ERR(in);
 			goto done;
@@ -1233,8 +1206,8 @@ retry_lookup:
 
 		err = fill_inode(in, req->r_locked_page, &rinfo->targeti, NULL,
 				session, req->r_request_started,
-				(!req->r_aborted && rinfo->head->result == 0) ?
-				req->r_fmode : -1,
+				(!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags) &&
+				rinfo->head->result == 0) ?  req->r_fmode : -1,
 				&req->r_caps_reservation);
 		if (err < 0) {
 			pr_err("fill_inode badness %p %llx.%llx\n",
@@ -1247,8 +1220,9 @@ retry_lookup:
 	 * ignore null lease/binding on snapdir ENOENT, or else we
 	 * will have trouble splicing in the virtual snapdir later
 	 */
-	if (rinfo->head->is_dentry && !req->r_aborted &&
-	    req->r_locked_dir &&
+	if (rinfo->head->is_dentry &&
+            !test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags) &&
+	    test_bit(CEPH_MDS_R_PARENT_LOCKED, &req->r_req_flags) &&
 	    (rinfo->head->is_target || strncmp(req->r_dentry->d_name.name,
 					       fsc->mount_options->snapdir_name,
 					       req->r_dentry->d_name.len))) {
@@ -1257,17 +1231,19 @@ retry_lookup:
 		 * mknod symlink mkdir  : null -> new inode
 		 * unlink               : linked -> null
 		 */
-		struct inode *dir = req->r_locked_dir;
+		struct inode *dir = req->r_parent;
 		struct dentry *dn = req->r_dentry;
 		bool have_dir_cap, have_lease;
 
 		BUG_ON(!dn);
 		BUG_ON(!dir);
 		BUG_ON(d_inode(dn->d_parent) != dir);
-		BUG_ON(ceph_ino(dir) !=
-		       le64_to_cpu(rinfo->diri.in->ino));
-		BUG_ON(ceph_snap(dir) !=
-		       le64_to_cpu(rinfo->diri.in->snapid));
+
+		dvino.ino = le64_to_cpu(rinfo->diri.in->ino);
+		dvino.snap = le64_to_cpu(rinfo->diri.in->snapid);
+
+		BUG_ON(ceph_ino(dir) != dvino.ino);
+		BUG_ON(ceph_snap(dir) != dvino.snap);
 
 		/* do we have a lease on the whole dir? */
 		have_dir_cap =
@@ -1319,12 +1295,13 @@ retry_lookup:
 				ceph_dir_clear_ordered(dir);
 				dout("d_delete %p\n", dn);
 				d_delete(dn);
-			} else {
-				if (have_lease && d_unhashed(dn))
+			} else if (have_lease) {
+				if (d_unhashed(dn))
 					d_add(dn, NULL);
 				update_dentry_lease(dn, rinfo->dlease,
 						    session,
-						    req->r_request_started);
+						    req->r_request_started,
+						    NULL, &dvino);
 			}
 			goto done;
 		}
@@ -1347,15 +1324,19 @@ retry_lookup:
 			have_lease = false;
 		}
 
-		if (have_lease)
+		if (have_lease) {
+			tvino.ino = le64_to_cpu(rinfo->targeti.in->ino);
+			tvino.snap = le64_to_cpu(rinfo->targeti.in->snapid);
 			update_dentry_lease(dn, rinfo->dlease, session,
-					    req->r_request_started);
+					    req->r_request_started,
+					    &tvino, &dvino);
+		}
 		dout(" final dn %p\n", dn);
-	} else if (!req->r_aborted &&
-		   (req->r_op == CEPH_MDS_OP_LOOKUPSNAP ||
-		    req->r_op == CEPH_MDS_OP_MKSNAP)) {
+	} else if ((req->r_op == CEPH_MDS_OP_LOOKUPSNAP ||
+		    req->r_op == CEPH_MDS_OP_MKSNAP) &&
+		   !test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
 		struct dentry *dn = req->r_dentry;
-		struct inode *dir = req->r_locked_dir;
+		struct inode *dir = req->r_parent;
 
 		/* fill out a snapdir LOOKUPSNAP dentry */
 		BUG_ON(!dn);
@@ -1370,6 +1351,26 @@ retry_lookup:
 			goto done;
 		}
 		req->r_dentry = dn;  /* may have spliced */
+	} else if (rinfo->head->is_dentry) {
+		struct ceph_vino *ptvino = NULL;
+
+		if ((le32_to_cpu(rinfo->diri.in->cap.caps) & CEPH_CAP_FILE_SHARED) ||
+		    le32_to_cpu(rinfo->dlease->duration_ms)) {
+			dvino.ino = le64_to_cpu(rinfo->diri.in->ino);
+			dvino.snap = le64_to_cpu(rinfo->diri.in->snapid);
+
+			if (rinfo->head->is_target) {
+				tvino.ino = le64_to_cpu(rinfo->targeti.in->ino);
+				tvino.snap = le64_to_cpu(rinfo->targeti.in->snapid);
+				ptvino = &tvino;
+			}
+
+			update_dentry_lease(req->r_dentry, rinfo->dlease,
+				session, req->r_request_started, ptvino,
+				&dvino);
+		} else {
+			dout("%s: no dentry lease or dir cap\n", __func__);
+		}
 	}
 done:
 	dout("fill_trace done err=%d\n", err);
@@ -1478,7 +1479,7 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 	u32 fpos_offset;
 	struct ceph_readdir_cache_control cache_ctl = {};
 
-	if (req->r_aborted)
+	if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
 		return readdir_prepopulate_inodes_only(req, session);
 
 	if (rinfo->hash_order && req->r_path2) {
@@ -1523,14 +1524,14 @@ int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 	/* FIXME: release caps/leases if error occurs */
 	for (i = 0; i < rinfo->dir_nr; i++) {
 		struct ceph_mds_reply_dir_entry *rde = rinfo->dir_entries + i;
-		struct ceph_vino vino;
+		struct ceph_vino tvino, dvino;
 
 		dname.name = rde->name;
 		dname.len = rde->name_len;
 		dname.hash = full_name_hash(parent, dname.name, dname.len);
 
-		vino.ino = le64_to_cpu(rde->inode.in->ino);
-		vino.snap = le64_to_cpu(rde->inode.in->snapid);
+		tvino.ino = le64_to_cpu(rde->inode.in->ino);
+		tvino.snap = le64_to_cpu(rde->inode.in->snapid);
 
 		if (rinfo->hash_order) {
 			u32 hash = ceph_str_hash(ci->i_dir_layout.dl_dir_hash,
@@ -1559,8 +1560,8 @@ retry_lookup:
 				goto out;
 			}
 		} else if (d_really_is_positive(dn) &&
-			   (ceph_ino(d_inode(dn)) != vino.ino ||
-			    ceph_snap(d_inode(dn)) != vino.snap)) {
+			   (ceph_ino(d_inode(dn)) != tvino.ino ||
+			    ceph_snap(d_inode(dn)) != tvino.snap)) {
 			dout(" dn %p points to wrong inode %p\n",
 			     dn, d_inode(dn));
 			d_delete(dn);
@@ -1572,7 +1573,7 @@ retry_lookup:
 		if (d_really_is_positive(dn)) {
 			in = d_inode(dn);
 		} else {
-			in = ceph_get_inode(parent->d_sb, vino);
+			in = ceph_get_inode(parent->d_sb, tvino);
 			if (IS_ERR(in)) {
 				dout("new_inode badness\n");
 				d_drop(dn);
@@ -1617,8 +1618,9 @@ retry_lookup:
 
 		ceph_dentry(dn)->offset = rde->offset;
 
+		dvino = ceph_vino(d_inode(parent));
 		update_dentry_lease(dn, rde->lease, req->r_session,
-				    req->r_request_started);
+				    req->r_request_started, &tvino, &dvino);
 
 		if (err == 0 && skipped == 0 && cache_ctl.index >= 0) {
 			ret = fill_readdir_cache(d_inode(parent), dn,
@@ -1632,7 +1634,7 @@ next_item:
 	}
 out:
 	if (err == 0 && skipped == 0) {
-		req->r_did_prepopulate = true;
+		set_bit(CEPH_MDS_R_DID_PREPOPULATE, &req->r_req_flags);
 		req->r_readdir_cache_idx = cache_ctl.index;
 	}
 	ceph_readdir_cache_release(&cache_ctl);
@@ -1720,7 +1722,7 @@ static void ceph_invalidate_work(struct work_struct *work)
 
 	mutex_lock(&ci->i_truncate_mutex);
 
-	if (ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
+	if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
 		pr_warn_ratelimited("invalidate_pages %p %lld forced umount\n",
 				    inode, ceph_ino(inode));
 		mapping_set_error(inode->i_mapping, -EIO);

fs/ceph/ioctl.c

@@ -25,7 +25,7 @@ static long ceph_ioctl_get_layout(struct file *file, void __user *arg)
 		l.stripe_count = ci->i_layout.stripe_count;
 		l.object_size = ci->i_layout.object_size;
 		l.data_pool = ci->i_layout.pool_id;
-		l.preferred_osd = (s32)-1;
+		l.preferred_osd = -1;
 		if (copy_to_user(arg, &l, sizeof(l)))
 			return -EFAULT;
 	}
@@ -97,7 +97,7 @@ static long ceph_ioctl_set_layout(struct file *file, void __user *arg)
 		nl.data_pool = ci->i_layout.pool_id;
 
 	/* this is obsolete, and always -1 */
-	nl.preferred_osd = le64_to_cpu(-1);
+	nl.preferred_osd = -1;
 
 	err = __validate_layout(mdsc, &nl);
 	if (err)

fs/ceph/mds_client.c

@@ -547,8 +547,8 @@ void ceph_mdsc_release_request(struct kref *kref)
 		ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
 		iput(req->r_inode);
 	}
-	if (req->r_locked_dir)
-		ceph_put_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
+	if (req->r_parent)
+		ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
 	iput(req->r_target_inode);
 	if (req->r_dentry)
 		dput(req->r_dentry);
@@ -628,6 +628,9 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 {
 	dout("__unregister_request %p tid %lld\n", req, req->r_tid);
 
+	/* Never leave an unregistered request on an unsafe list! */
+	list_del_init(&req->r_unsafe_item);
+
 	if (req->r_tid == mdsc->oldest_tid) {
 		struct rb_node *p = rb_next(&req->r_node);
 		mdsc->oldest_tid = 0;
@@ -644,13 +647,15 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 
 	erase_request(&mdsc->request_tree, req);
 
-	if (req->r_unsafe_dir && req->r_got_unsafe) {
+	if (req->r_unsafe_dir  &&
+	    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 		struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
 		spin_lock(&ci->i_unsafe_lock);
 		list_del_init(&req->r_unsafe_dir_item);
 		spin_unlock(&ci->i_unsafe_lock);
 	}
-	if (req->r_target_inode && req->r_got_unsafe) {
+	if (req->r_target_inode &&
+	    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 		struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
 		spin_lock(&ci->i_unsafe_lock);
 		list_del_init(&req->r_unsafe_target_item);
@@ -667,6 +672,28 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
 	ceph_mdsc_put_request(req);
 }
 
+/*
+ * Walk back up the dentry tree until we hit a dentry representing a
+ * non-snapshot inode. We do this using the rcu_read_lock (which must be held
+ * when calling this) to ensure that the objects won't disappear while we're
+ * working with them. Once we hit a candidate dentry, we attempt to take a
+ * reference to it, and return that as the result.
+ */
+static struct inode *get_nonsnap_parent(struct dentry *dentry)
+{
+	struct inode *inode = NULL;
+
+	while (dentry && !IS_ROOT(dentry)) {
+		inode = d_inode_rcu(dentry);
+		if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
+			break;
+		dentry = dentry->d_parent;
+	}
+	if (inode)
+		inode = igrab(inode);
+	return inode;
+}
+
 /*
  * Choose mds to send request to next.  If there is a hint set in the
  * request (e.g., due to a prior forward hint from the mds), use that.
@@ -675,19 +702,6 @@ static void __unregister_request(struct ceph_mds_client *mdsc,
  *
  * Called under mdsc->mutex.
  */
-static struct dentry *get_nonsnap_parent(struct dentry *dentry)
-{
-	/*
-	 * we don't need to worry about protecting the d_parent access
-	 * here because we never renaming inside the snapped namespace
-	 * except to resplice to another snapdir, and either the old or new
-	 * result is a valid result.
-	 */
-	while (!IS_ROOT(dentry) && ceph_snap(d_inode(dentry)) != CEPH_NOSNAP)
-		dentry = dentry->d_parent;
-	return dentry;
-}
-
 static int __choose_mds(struct ceph_mds_client *mdsc,
 			struct ceph_mds_request *req)
 {
@@ -697,7 +711,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 	int mode = req->r_direct_mode;
 	int mds = -1;
 	u32 hash = req->r_direct_hash;
-	bool is_hash = req->r_direct_is_hash;
+	bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
 
 	/*
 	 * is there a specific mds we should try?  ignore hint if we have
@@ -717,30 +731,39 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 	inode = NULL;
 	if (req->r_inode) {
 		inode = req->r_inode;
+		ihold(inode);
 	} else if (req->r_dentry) {
 		/* ignore race with rename; old or new d_parent is okay */
-		struct dentry *parent = req->r_dentry->d_parent;
-		struct inode *dir = d_inode(parent);
+		struct dentry *parent;
+		struct inode *dir;
+
+		rcu_read_lock();
+		parent = req->r_dentry->d_parent;
+		dir = req->r_parent ? : d_inode_rcu(parent);
 
-		if (dir->i_sb != mdsc->fsc->sb) {
-			/* not this fs! */
+		if (!dir || dir->i_sb != mdsc->fsc->sb) {
+			/*  not this fs or parent went negative */
 			inode = d_inode(req->r_dentry);
+			if (inode)
+				ihold(inode);
 		} else if (ceph_snap(dir) != CEPH_NOSNAP) {
 			/* direct snapped/virtual snapdir requests
 			 * based on parent dir inode */
-			struct dentry *dn = get_nonsnap_parent(parent);
-			inode = d_inode(dn);
+			inode = get_nonsnap_parent(parent);
 			dout("__choose_mds using nonsnap parent %p\n", inode);
 		} else {
 			/* dentry target */
 			inode = d_inode(req->r_dentry);
 			if (!inode || mode == USE_AUTH_MDS) {
 				/* dir + name */
-				inode = dir;
+				inode = igrab(dir);
 				hash = ceph_dentry_hash(dir, req->r_dentry);
 				is_hash = true;
+			} else {
+				ihold(inode);
 			}
 		}
+		rcu_read_unlock();
 	}
 
 	dout("__choose_mds %p is_hash=%d (%d) mode %d\n", inode, (int)is_hash,
@@ -769,7 +792,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 				     (int)r, frag.ndist);
 				if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
 				    CEPH_MDS_STATE_ACTIVE)
-					return mds;
+					goto out;
 			}
 
 			/* since this file/dir wasn't known to be
@@ -784,7 +807,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 				     inode, ceph_vinop(inode), frag.frag, mds);
 				if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
 				    CEPH_MDS_STATE_ACTIVE)
-					return mds;
+					goto out;
 			}
 		}
 	}
@@ -797,6 +820,7 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 		cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
 	if (!cap) {
 		spin_unlock(&ci->i_ceph_lock);
+		iput(inode);
 		goto random;
 	}
 	mds = cap->session->s_mds;
@@ -804,6 +828,8 @@ static int __choose_mds(struct ceph_mds_client *mdsc,
 	     inode, ceph_vinop(inode), mds,
 	     cap == ci->i_auth_cap ? "auth " : "", cap);
 	spin_unlock(&ci->i_ceph_lock);
+out:
+	iput(inode);
 	return mds;
 
 random:
@@ -1036,7 +1062,6 @@ static void cleanup_session_requests(struct ceph_mds_client *mdsc,
 	while (!list_empty(&session->s_unsafe)) {
 		req = list_first_entry(&session->s_unsafe,
 				       struct ceph_mds_request, r_unsafe_item);
-		list_del_init(&req->r_unsafe_item);
 		pr_warn_ratelimited(" dropping unsafe request %llu\n",
 				    req->r_tid);
 		__unregister_request(mdsc, req);
@@ -1146,7 +1171,7 @@ static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
 		ci->i_ceph_flags |= CEPH_I_CAP_DROPPED;
 
 		if (ci->i_wrbuffer_ref > 0 &&
-		    ACCESS_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
+		    READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
 			invalidate = true;
 
 		while (!list_empty(&ci->i_cap_flush_list)) {
@@ -1775,18 +1800,23 @@ retry:
 	return path;
 }
 
-static int build_dentry_path(struct dentry *dentry,
+static int build_dentry_path(struct dentry *dentry, struct inode *dir,
 			     const char **ppath, int *ppathlen, u64 *pino,
 			     int *pfreepath)
 {
 	char *path;
 
-	if (ceph_snap(d_inode(dentry->d_parent)) == CEPH_NOSNAP) {
-		*pino = ceph_ino(d_inode(dentry->d_parent));
+	rcu_read_lock();
+	if (!dir)
+		dir = d_inode_rcu(dentry->d_parent);
+	if (dir && ceph_snap(dir) == CEPH_NOSNAP) {
+		*pino = ceph_ino(dir);
+		rcu_read_unlock();
 		*ppath = dentry->d_name.name;
 		*ppathlen = dentry->d_name.len;
 		return 0;
 	}
+	rcu_read_unlock();
 	path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
 	if (IS_ERR(path))
 		return PTR_ERR(path);
@@ -1822,8 +1852,8 @@ static int build_inode_path(struct inode *inode,
  * an explicit ino+path.
  */
 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
-				  const char *rpath, u64 rino,
-				  const char **ppath, int *pathlen,
+				  struct inode *rdiri, const char *rpath,
+				  u64 rino, const char **ppath, int *pathlen,
 				  u64 *ino, int *freepath)
 {
 	int r = 0;
@@ -1833,7 +1863,8 @@ static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
 		dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
 		     ceph_snap(rinode));
 	} else if (rdentry) {
-		r = build_dentry_path(rdentry, ppath, pathlen, ino, freepath);
+		r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
+					freepath);
 		dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
 		     *ppath);
 	} else if (rpath || rino) {
@@ -1866,7 +1897,7 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
 	int ret;
 
 	ret = set_request_path_attr(req->r_inode, req->r_dentry,
-			      req->r_path1, req->r_ino1.ino,
+			      req->r_parent, req->r_path1, req->r_ino1.ino,
 			      &path1, &pathlen1, &ino1, &freepath1);
 	if (ret < 0) {
 		msg = ERR_PTR(ret);
@@ -1874,6 +1905,7 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
 	}
 
 	ret = set_request_path_attr(NULL, req->r_old_dentry,
+			      req->r_old_dentry_dir,
 			      req->r_path2, req->r_ino2.ino,
 			      &path2, &pathlen2, &ino2, &freepath2);
 	if (ret < 0) {
@@ -1927,10 +1959,13 @@ static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
 		      mds, req->r_inode_drop, req->r_inode_unless, 0);
 	if (req->r_dentry_drop)
 		releases += ceph_encode_dentry_release(&p, req->r_dentry,
-		       mds, req->r_dentry_drop, req->r_dentry_unless);
+				req->r_parent, mds, req->r_dentry_drop,
+				req->r_dentry_unless);
 	if (req->r_old_dentry_drop)
 		releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
-		       mds, req->r_old_dentry_drop, req->r_old_dentry_unless);
+				req->r_old_dentry_dir, mds,
+				req->r_old_dentry_drop,
+				req->r_old_dentry_unless);
 	if (req->r_old_inode_drop)
 		releases += ceph_encode_inode_release(&p,
 		      d_inode(req->r_old_dentry),
@@ -2012,7 +2047,7 @@ static int __prepare_send_request(struct ceph_mds_client *mdsc,
 	dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
 	     req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
 
-	if (req->r_got_unsafe) {
+	if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 		void *p;
 		/*
 		 * Replay.  Do not regenerate message (and rebuild
@@ -2061,16 +2096,16 @@ static int __prepare_send_request(struct ceph_mds_client *mdsc,
 
 	rhead = msg->front.iov_base;
 	rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
-	if (req->r_got_unsafe)
+	if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
 		flags |= CEPH_MDS_FLAG_REPLAY;
-	if (req->r_locked_dir)
+	if (req->r_parent)
 		flags |= CEPH_MDS_FLAG_WANT_DENTRY;
 	rhead->flags = cpu_to_le32(flags);
 	rhead->num_fwd = req->r_num_fwd;
 	rhead->num_retry = req->r_attempts - 1;
 	rhead->ino = 0;
 
-	dout(" r_locked_dir = %p\n", req->r_locked_dir);
+	dout(" r_parent = %p\n", req->r_parent);
 	return 0;
 }
 
@@ -2084,8 +2119,8 @@ static int __do_request(struct ceph_mds_client *mdsc,
 	int mds = -1;
 	int err = 0;
 
-	if (req->r_err || req->r_got_result) {
-		if (req->r_aborted)
+	if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
+		if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
 			__unregister_request(mdsc, req);
 		goto out;
 	}
@@ -2096,12 +2131,12 @@ static int __do_request(struct ceph_mds_client *mdsc,
 		err = -EIO;
 		goto finish;
 	}
-	if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
 		dout("do_request forced umount\n");
 		err = -EIO;
 		goto finish;
 	}
-	if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
 		if (mdsc->mdsmap_err) {
 			err = mdsc->mdsmap_err;
 			dout("do_request mdsmap err %d\n", err);
@@ -2215,7 +2250,7 @@ static void kick_requests(struct ceph_mds_client *mdsc, int mds)
 	while (p) {
 		req = rb_entry(p, struct ceph_mds_request, r_node);
 		p = rb_next(p);
-		if (req->r_got_unsafe)
+		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
 			continue;
 		if (req->r_attempts > 0)
 			continue; /* only new requests */
@@ -2250,11 +2285,11 @@ int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
 
 	dout("do_request on %p\n", req);
 
-	/* take CAP_PIN refs for r_inode, r_locked_dir, r_old_dentry */
+	/* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
 	if (req->r_inode)
 		ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
-	if (req->r_locked_dir)
-		ceph_get_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
+	if (req->r_parent)
+		ceph_get_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
 	if (req->r_old_dentry_dir)
 		ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
 				  CEPH_CAP_PIN);
@@ -2289,7 +2324,7 @@ int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
 	mutex_lock(&mdsc->mutex);
 
 	/* only abort if we didn't race with a real reply */
-	if (req->r_got_result) {
+	if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
 		err = le32_to_cpu(req->r_reply_info.head->result);
 	} else if (err < 0) {
 		dout("aborted request %lld with %d\n", req->r_tid, err);
@@ -2301,10 +2336,10 @@ int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
 		 */
 		mutex_lock(&req->r_fill_mutex);
 		req->r_err = err;
-		req->r_aborted = true;
+		set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
 		mutex_unlock(&req->r_fill_mutex);
 
-		if (req->r_locked_dir &&
+		if (req->r_parent &&
 		    (req->r_op & CEPH_MDS_OP_WRITE))
 			ceph_invalidate_dir_request(req);
 	} else {
@@ -2323,7 +2358,7 @@ out:
  */
 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
 {
-	struct inode *inode = req->r_locked_dir;
+	struct inode *inode = req->r_parent;
 
 	dout("invalidate_dir_request %p (complete, lease(s))\n", inode);
 
@@ -2379,14 +2414,14 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	}
 
 	/* dup? */
-	if ((req->r_got_unsafe && !head->safe) ||
-	    (req->r_got_safe && head->safe)) {
+	if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
+	    (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
 		pr_warn("got a dup %s reply on %llu from mds%d\n",
 			   head->safe ? "safe" : "unsafe", tid, mds);
 		mutex_unlock(&mdsc->mutex);
 		goto out;
 	}
-	if (req->r_got_safe) {
+	if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
 		pr_warn("got unsafe after safe on %llu from mds%d\n",
 			   tid, mds);
 		mutex_unlock(&mdsc->mutex);
@@ -2425,10 +2460,10 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 
 
 	if (head->safe) {
-		req->r_got_safe = true;
+		set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
 		__unregister_request(mdsc, req);
 
-		if (req->r_got_unsafe) {
+		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 			/*
 			 * We already handled the unsafe response, now do the
 			 * cleanup.  No need to examine the response; the MDS
@@ -2437,7 +2472,6 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 			 * useful we could do with a revised return value.
 			 */
 			dout("got safe reply %llu, mds%d\n", tid, mds);
-			list_del_init(&req->r_unsafe_item);
 
 			/* last unsafe request during umount? */
 			if (mdsc->stopping && !__get_oldest_req(mdsc))
@@ -2446,7 +2480,7 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 			goto out;
 		}
 	} else {
-		req->r_got_unsafe = true;
+		set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
 		list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
 		if (req->r_unsafe_dir) {
 			struct ceph_inode_info *ci =
@@ -2486,7 +2520,7 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	/* insert trace into our cache */
 	mutex_lock(&req->r_fill_mutex);
 	current->journal_info = req;
-	err = ceph_fill_trace(mdsc->fsc->sb, req, req->r_session);
+	err = ceph_fill_trace(mdsc->fsc->sb, req);
 	if (err == 0) {
 		if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
 				    req->r_op == CEPH_MDS_OP_LSSNAP))
@@ -2500,7 +2534,8 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	if (realm)
 		ceph_put_snap_realm(mdsc, realm);
 
-	if (err == 0 && req->r_got_unsafe && req->r_target_inode) {
+	if (err == 0 && req->r_target_inode &&
+	    test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 		struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
 		spin_lock(&ci->i_unsafe_lock);
 		list_add_tail(&req->r_unsafe_target_item, &ci->i_unsafe_iops);
@@ -2508,12 +2543,12 @@ static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
 	}
 out_err:
 	mutex_lock(&mdsc->mutex);
-	if (!req->r_aborted) {
+	if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
 		if (err) {
 			req->r_err = err;
 		} else {
 			req->r_reply =  ceph_msg_get(msg);
-			req->r_got_result = true;
+			set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
 		}
 	} else {
 		dout("reply arrived after request %lld was aborted\n", tid);
@@ -2557,7 +2592,7 @@ static void handle_forward(struct ceph_mds_client *mdsc,
 		goto out;  /* dup reply? */
 	}
 
-	if (req->r_aborted) {
+	if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
 		dout("forward tid %llu aborted, unregistering\n", tid);
 		__unregister_request(mdsc, req);
 	} else if (fwd_seq <= req->r_num_fwd) {
@@ -2567,7 +2602,7 @@ static void handle_forward(struct ceph_mds_client *mdsc,
 		/* resend. forward race not possible; mds would drop */
 		dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
 		BUG_ON(req->r_err);
-		BUG_ON(req->r_got_result);
+		BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
 		req->r_attempts = 0;
 		req->r_num_fwd = fwd_seq;
 		req->r_resend_mds = next_mds;
@@ -2732,7 +2767,7 @@ static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
 	while (p) {
 		req = rb_entry(p, struct ceph_mds_request, r_node);
 		p = rb_next(p);
-		if (req->r_got_unsafe)
+		if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
 			continue;
 		if (req->r_attempts == 0)
 			continue; /* only old requests */
@@ -3556,7 +3591,7 @@ void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
 {
 	u64 want_tid, want_flush;
 
-	if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
 		return;
 
 	dout("sync\n");
@@ -3587,7 +3622,7 @@ void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
  */
 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
 {
-	if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
+	if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
 		return true;
 	return atomic_read(&mdsc->num_sessions) <= skipped;
 }

fs/ceph/mds_client.h

@@ -202,9 +202,18 @@ struct ceph_mds_request {
 	char *r_path1, *r_path2;
 	struct ceph_vino r_ino1, r_ino2;
 
-	struct inode *r_locked_dir; /* dir (if any) i_mutex locked by vfs */
+	struct inode *r_parent;		    /* parent dir inode */
 	struct inode *r_target_inode;       /* resulting inode */
 
+#define CEPH_MDS_R_DIRECT_IS_HASH	(1) /* r_direct_hash is valid */
+#define CEPH_MDS_R_ABORTED		(2) /* call was aborted */
+#define CEPH_MDS_R_GOT_UNSAFE		(3) /* got an unsafe reply */
+#define CEPH_MDS_R_GOT_SAFE		(4) /* got a safe reply */
+#define CEPH_MDS_R_GOT_RESULT		(5) /* got a result */
+#define CEPH_MDS_R_DID_PREPOPULATE	(6) /* prepopulated readdir */
+#define CEPH_MDS_R_PARENT_LOCKED	(7) /* is r_parent->i_rwsem wlocked? */
+	unsigned long	r_req_flags;
+
 	struct mutex r_fill_mutex;
 
 	union ceph_mds_request_args r_args;
@@ -216,7 +225,6 @@ struct ceph_mds_request {
 	/* for choosing which mds to send this request to */
 	int r_direct_mode;
 	u32 r_direct_hash;      /* choose dir frag based on this dentry hash */
-	bool r_direct_is_hash;  /* true if r_direct_hash is valid */
 
 	/* data payload is used for xattr ops */
 	struct ceph_pagelist *r_pagelist;
@@ -234,7 +242,6 @@ struct ceph_mds_request {
 	struct ceph_mds_reply_info_parsed r_reply_info;
 	struct page *r_locked_page;
 	int r_err;
-	bool r_aborted;
 
 	unsigned long r_timeout;  /* optional.  jiffies, 0 is "wait forever" */
 	unsigned long r_started;  /* start time to measure timeout against */
@@ -262,9 +269,7 @@ struct ceph_mds_request {
 	ceph_mds_request_callback_t r_callback;
 	ceph_mds_request_wait_callback_t r_wait_for_completion;
 	struct list_head  r_unsafe_item;  /* per-session unsafe list item */
-	bool		  r_got_unsafe, r_got_safe, r_got_result;
 
-	bool              r_did_prepopulate;
 	long long	  r_dir_release_cnt;
 	long long	  r_dir_ordered_cnt;
 	int		  r_readdir_cache_idx;

fs/ceph/super.c

@@ -757,7 +757,6 @@ static const struct super_operations ceph_super_ops = {
 	.destroy_inode	= ceph_destroy_inode,
 	.write_inode    = ceph_write_inode,
 	.drop_inode	= ceph_drop_inode,
-	.evict_inode	= ceph_evict_inode,
 	.sync_fs        = ceph_sync_fs,
 	.put_super	= ceph_put_super,
 	.show_options   = ceph_show_options,
@@ -952,6 +951,14 @@ static int ceph_register_bdi(struct super_block *sb,
 		fsc->backing_dev_info.ra_pages =
 			VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
 
+	if (fsc->mount_options->rsize > fsc->mount_options->rasize &&
+	    fsc->mount_options->rsize >= PAGE_SIZE)
+		fsc->backing_dev_info.io_pages =
+			(fsc->mount_options->rsize + PAGE_SIZE - 1)
+			>> PAGE_SHIFT;
+	else if (fsc->mount_options->rsize == 0)
+		fsc->backing_dev_info.io_pages = ULONG_MAX;
+
 	err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%ld",
 			   atomic_long_inc_return(&bdi_seq));
 	if (!err)

fs/ceph/super.h

@@ -45,8 +45,8 @@
 #define ceph_test_mount_opt(fsc, opt) \
 	(!!((fsc)->mount_options->flags & CEPH_MOUNT_OPT_##opt))
 
-#define CEPH_RSIZE_DEFAULT             0           /* max read size */
-#define CEPH_RASIZE_DEFAULT            (8192*1024) /* readahead */
+#define CEPH_RSIZE_DEFAULT              (64*1024*1024) /* max read size */
+#define CEPH_RASIZE_DEFAULT             (8192*1024)    /* max readahead */
 #define CEPH_MAX_READDIR_DEFAULT        1024
 #define CEPH_MAX_READDIR_BYTES_DEFAULT  (512*1024)
 #define CEPH_SNAPDIRNAME_DEFAULT        ".snap"
@@ -343,7 +343,6 @@ struct ceph_inode_info {
 	u32 i_rdcache_gen;      /* incremented each time we get FILE_CACHE. */
 	u32 i_rdcache_revoking; /* RDCACHE gen to async invalidate, if any */
 
-	struct list_head i_unsafe_writes; /* uncommitted sync writes */
 	struct list_head i_unsafe_dirops; /* uncommitted mds dir ops */
 	struct list_head i_unsafe_iops;   /* uncommitted mds inode ops */
 	spinlock_t i_unsafe_lock;
@@ -602,7 +601,7 @@ static inline int __ceph_caps_wanted(struct ceph_inode_info *ci)
 }
 
 /* what the mds thinks we want */
-extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci);
+extern int __ceph_caps_mds_wanted(struct ceph_inode_info *ci, bool check);
 
 extern void ceph_caps_init(struct ceph_mds_client *mdsc);
 extern void ceph_caps_finalize(struct ceph_mds_client *mdsc);
@@ -753,7 +752,6 @@ extern const struct inode_operations ceph_file_iops;
 extern struct inode *ceph_alloc_inode(struct super_block *sb);
 extern void ceph_destroy_inode(struct inode *inode);
 extern int ceph_drop_inode(struct inode *inode);
-extern void ceph_evict_inode(struct inode *inode);
 
 extern struct inode *ceph_get_inode(struct super_block *sb,
 				    struct ceph_vino vino);
@@ -764,8 +762,7 @@ extern void ceph_fill_file_time(struct inode *inode, int issued,
 				u64 time_warp_seq, struct timespec *ctime,
 				struct timespec *mtime, struct timespec *atime);
 extern int ceph_fill_trace(struct super_block *sb,
-			   struct ceph_mds_request *req,
-			   struct ceph_mds_session *session);
+			   struct ceph_mds_request *req);
 extern int ceph_readdir_prepopulate(struct ceph_mds_request *req,
 				    struct ceph_mds_session *session);
 
@@ -904,6 +901,7 @@ extern void ceph_flush_dirty_caps(struct ceph_mds_client *mdsc);
 extern int ceph_encode_inode_release(void **p, struct inode *inode,
 				     int mds, int drop, int unless, int force);
 extern int ceph_encode_dentry_release(void **p, struct dentry *dn,
+				      struct inode *dir,
 				      int mds, int drop, int unless);
 
 extern int ceph_get_caps(struct ceph_inode_info *ci, int need, int want,
@@ -933,7 +931,7 @@ extern int ceph_atomic_open(struct inode *dir, struct dentry *dentry,
 extern int ceph_release(struct inode *inode, struct file *filp);
 extern void ceph_fill_inline_data(struct inode *inode, struct page *locked_page,
 				  char *data, size_t len);
-extern void ceph_sync_write_wait(struct inode *inode);
+
 /* dir.c */
 extern const struct file_operations ceph_dir_fops;
 extern const struct file_operations ceph_snapdir_fops;

include/linux/ceph/osd_client.h

@@ -22,7 +22,6 @@ struct ceph_osd_client;
  * completion callback for async writepages
  */
 typedef void (*ceph_osdc_callback_t)(struct ceph_osd_request *);
-typedef void (*ceph_osdc_unsafe_callback_t)(struct ceph_osd_request *, bool);
 
 #define CEPH_HOMELESS_OSD	-1
 
@@ -170,15 +169,12 @@ struct ceph_osd_request {
 	unsigned int		r_num_ops;
 
 	int               r_result;
-	bool              r_got_reply;
 
 	struct ceph_osd_client *r_osdc;
 	struct kref       r_kref;
 	bool              r_mempool;
-	struct completion r_completion;
-	struct completion r_done_completion;  /* fsync waiter */
+	struct completion r_completion;       /* private to osd_client.c */
 	ceph_osdc_callback_t r_callback;
-	ceph_osdc_unsafe_callback_t r_unsafe_callback;
 	struct list_head  r_unsafe_item;
 
 	struct inode *r_inode;         	      /* for use by callbacks */

include/linux/ceph/osdmap.h

@@ -57,7 +57,7 @@ static inline bool ceph_can_shift_osds(struct ceph_pg_pool_info *pool)
 	case CEPH_POOL_TYPE_EC:
 		return false;
 	default:
-		BUG_ON(1);
+		BUG();
 	}
 }
 
@@ -81,13 +81,6 @@ void ceph_oloc_copy(struct ceph_object_locator *dest,
 		    const struct ceph_object_locator *src);
 void ceph_oloc_destroy(struct ceph_object_locator *oloc);
 
-/*
- * Maximum supported by kernel client object name length
- *
- * (probably outdated: must be >= RBD_MAX_MD_NAME_LEN -- currently 100)
- */
-#define CEPH_MAX_OID_NAME_LEN 100
-
 /*
  * 51-char inline_name is long enough for all cephfs and all but one
  * rbd requests: <imgname> in "<imgname>.rbd"/"rbd_id.<imgname>" can be
@@ -173,8 +166,8 @@ struct ceph_osdmap {
 	 * the list of osds that store+replicate them. */
 	struct crush_map *crush;
 
-	struct mutex crush_scratch_mutex;
-	int crush_scratch_ary[CEPH_PG_MAX_SIZE * 3];
+	struct mutex crush_workspace_mutex;
+	void *crush_workspace;
 };
 
 static inline bool ceph_osd_exists(struct ceph_osdmap *map, int osd)

include/linux/ceph/rados.h

@@ -50,7 +50,7 @@ struct ceph_timespec {
 #define CEPH_PG_LAYOUT_LINEAR 2
 #define CEPH_PG_LAYOUT_HYBRID 3
 
-#define CEPH_PG_MAX_SIZE      16  /* max # osds in a single pg */
+#define CEPH_PG_MAX_SIZE      32  /* max # osds in a single pg */
 
 /*
  * placement group.

include/linux/crush/crush.h

@@ -135,13 +135,6 @@ struct crush_bucket {
 	__u32 size;      /* num items */
 	__s32 *items;
 
-	/*
-	 * cached random permutation: used for uniform bucket and for
-	 * the linear search fallback for the other bucket types.
-	 */
-	__u32 perm_x;  /* @x for which *perm is defined */
-	__u32 perm_n;  /* num elements of *perm that are permuted/defined */
-	__u32 *perm;
 };
 
 struct crush_bucket_uniform {
@@ -211,6 +204,21 @@ struct crush_map {
 	 * device fails. */
 	__u8 chooseleaf_stable;
 
+	/*
+	 * This value is calculated after decode or construction by
+	 * the builder. It is exposed here (rather than having a
+	 * 'build CRUSH working space' function) so that callers can
+	 * reserve a static buffer, allocate space on the stack, or
+	 * otherwise avoid calling into the heap allocator if they
+	 * want to. The size of the working space depends on the map,
+	 * while the size of the scratch vector passed to the mapper
+	 * depends on the size of the desired result set.
+	 *
+	 * Nothing stops the caller from allocating both in one swell
+	 * foop and passing in two points, though.
+	 */
+	size_t working_size;
+
 #ifndef __KERNEL__
 	/*
 	 * version 0 (original) of straw_calc has various flaws.  version 1
@@ -248,4 +256,23 @@ static inline int crush_calc_tree_node(int i)
 	return ((i+1) << 1)-1;
 }
 
+/*
+ * These data structures are private to the CRUSH implementation. They
+ * are exposed in this header file because builder needs their
+ * definitions to calculate the total working size.
+ *
+ * Moving this out of the crush map allow us to treat the CRUSH map as
+ * immutable within the mapper and removes the requirement for a CRUSH
+ * map lock.
+ */
+struct crush_work_bucket {
+	__u32 perm_x; /* @x for which *perm is defined */
+	__u32 perm_n; /* num elements of *perm that are permuted/defined */
+	__u32 *perm;  /* Permutation of the bucket's items */
+};
+
+struct crush_work {
+	struct crush_work_bucket **work; /* Per-bucket working store */
+};
+
 #endif

include/linux/crush/mapper.h

@@ -15,6 +15,20 @@ extern int crush_do_rule(const struct crush_map *map,
 			 int ruleno,
 			 int x, int *result, int result_max,
 			 const __u32 *weights, int weight_max,
-			 int *scratch);
+			 void *cwin);
+
+/*
+ * Returns the exact amount of workspace that will need to be used
+ * for a given combination of crush_map and result_max. The caller can
+ * then allocate this much on its own, either on the stack, in a
+ * per-thread long-lived buffer, or however it likes.
+ */
+static inline size_t crush_work_size(const struct crush_map *map,
+				     int result_max)
+{
+	return map->working_size + result_max * 3 * sizeof(__u32);
+}
+
+void crush_init_workspace(const struct crush_map *map, void *v);
 
 #endif

net/ceph/cls_lock_client.c

@@ -69,8 +69,8 @@ int ceph_cls_lock(struct ceph_osd_client *osdc,
 	dout("%s lock_name %s type %d cookie %s tag %s desc %s flags 0x%x\n",
 	     __func__, lock_name, type, cookie, tag, desc, flags);
 	ret = ceph_osdc_call(osdc, oid, oloc, "lock", "lock",
-			     CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK,
-			     lock_op_page, lock_op_buf_size, NULL, NULL);
+			     CEPH_OSD_FLAG_WRITE, lock_op_page,
+			     lock_op_buf_size, NULL, NULL);
 
 	dout("%s: status %d\n", __func__, ret);
 	__free_page(lock_op_page);
@@ -117,8 +117,8 @@ int ceph_cls_unlock(struct ceph_osd_client *osdc,
 
 	dout("%s lock_name %s cookie %s\n", __func__, lock_name, cookie);
 	ret = ceph_osdc_call(osdc, oid, oloc, "lock", "unlock",
-			     CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK,
-			     unlock_op_page, unlock_op_buf_size, NULL, NULL);
+			     CEPH_OSD_FLAG_WRITE, unlock_op_page,
+			     unlock_op_buf_size, NULL, NULL);
 
 	dout("%s: status %d\n", __func__, ret);
 	__free_page(unlock_op_page);
@@ -170,8 +170,8 @@ int ceph_cls_break_lock(struct ceph_osd_client *osdc,
 	dout("%s lock_name %s cookie %s locker %s%llu\n", __func__, lock_name,
 	     cookie, ENTITY_NAME(*locker));
 	ret = ceph_osdc_call(osdc, oid, oloc, "lock", "break_lock",
-			     CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK,
-			     break_op_page, break_op_buf_size, NULL, NULL);
+			     CEPH_OSD_FLAG_WRITE, break_op_page,
+			     break_op_buf_size, NULL, NULL);
 
 	dout("%s: status %d\n", __func__, ret);
 	__free_page(break_op_page);
@@ -278,7 +278,7 @@ int ceph_cls_lock_info(struct ceph_osd_client *osdc,
 	int get_info_op_buf_size;
 	int name_len = strlen(lock_name);
 	struct page *get_info_op_page, *reply_page;
-	size_t reply_len;
+	size_t reply_len = PAGE_SIZE;
 	void *p, *end;
 	int ret;
 

net/ceph/crush/crush.c

@@ -45,7 +45,6 @@ int crush_get_bucket_item_weight(const struct crush_bucket *b, int p)
 
 void crush_destroy_bucket_uniform(struct crush_bucket_uniform *b)
 {
-	kfree(b->h.perm);
 	kfree(b->h.items);
 	kfree(b);
 }
@@ -54,14 +53,12 @@ void crush_destroy_bucket_list(struct crush_bucket_list *b)
 {
 	kfree(b->item_weights);
 	kfree(b->sum_weights);
-	kfree(b->h.perm);
 	kfree(b->h.items);
 	kfree(b);
 }
 
 void crush_destroy_bucket_tree(struct crush_bucket_tree *b)
 {
-	kfree(b->h.perm);
 	kfree(b->h.items);
 	kfree(b->node_weights);
 	kfree(b);
@@ -71,7 +68,6 @@ void crush_destroy_bucket_straw(struct crush_bucket_straw *b)
 {
 	kfree(b->straws);
 	kfree(b->item_weights);
-	kfree(b->h.perm);
 	kfree(b->h.items);
 	kfree(b);
 }
@@ -79,7 +75,6 @@ void crush_destroy_bucket_straw(struct crush_bucket_straw *b)
 void crush_destroy_bucket_straw2(struct crush_bucket_straw2 *b)
 {
 	kfree(b->item_weights);
-	kfree(b->h.perm);
 	kfree(b->h.items);
 	kfree(b);
 }

net/ceph/crush/mapper.c

@@ -54,7 +54,6 @@ int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size
 	return -1;
 }
 
-
 /*
  * bucket choose methods
  *
@@ -72,59 +71,60 @@ int crush_find_rule(const struct crush_map *map, int ruleset, int type, int size
  * Since this is expensive, we optimize for the r=0 case, which
  * captures the vast majority of calls.
  */
-static int bucket_perm_choose(struct crush_bucket *bucket,
+static int bucket_perm_choose(const struct crush_bucket *bucket,
+			      struct crush_work_bucket *work,
 			      int x, int r)
 {
 	unsigned int pr = r % bucket->size;
 	unsigned int i, s;
 
 	/* start a new permutation if @x has changed */
-	if (bucket->perm_x != (__u32)x || bucket->perm_n == 0) {
+	if (work->perm_x != (__u32)x || work->perm_n == 0) {
 		dprintk("bucket %d new x=%d\n", bucket->id, x);
-		bucket->perm_x = x;
+		work->perm_x = x;
 
 		/* optimize common r=0 case */
 		if (pr == 0) {
 			s = crush_hash32_3(bucket->hash, x, bucket->id, 0) %
 				bucket->size;
-			bucket->perm[0] = s;
-			bucket->perm_n = 0xffff;   /* magic value, see below */
+			work->perm[0] = s;
+			work->perm_n = 0xffff;   /* magic value, see below */
 			goto out;
 		}
 
 		for (i = 0; i < bucket->size; i++)
-			bucket->perm[i] = i;
-		bucket->perm_n = 0;
-	} else if (bucket->perm_n == 0xffff) {
+			work->perm[i] = i;
+		work->perm_n = 0;
+	} else if (work->perm_n == 0xffff) {
 		/* clean up after the r=0 case above */
 		for (i = 1; i < bucket->size; i++)
-			bucket->perm[i] = i;
-		bucket->perm[bucket->perm[0]] = 0;
-		bucket->perm_n = 1;
+			work->perm[i] = i;
+		work->perm[work->perm[0]] = 0;
+		work->perm_n = 1;
 	}
 
 	/* calculate permutation up to pr */
-	for (i = 0; i < bucket->perm_n; i++)
-		dprintk(" perm_choose have %d: %d\n", i, bucket->perm[i]);
-	while (bucket->perm_n <= pr) {
-		unsigned int p = bucket->perm_n;
+	for (i = 0; i < work->perm_n; i++)
+		dprintk(" perm_choose have %d: %d\n", i, work->perm[i]);
+	while (work->perm_n <= pr) {
+		unsigned int p = work->perm_n;
 		/* no point in swapping the final entry */
 		if (p < bucket->size - 1) {
 			i = crush_hash32_3(bucket->hash, x, bucket->id, p) %
 				(bucket->size - p);
 			if (i) {
-				unsigned int t = bucket->perm[p + i];
-				bucket->perm[p + i] = bucket->perm[p];
-				bucket->perm[p] = t;
+				unsigned int t = work->perm[p + i];
+				work->perm[p + i] = work->perm[p];
+				work->perm[p] = t;
 			}
 			dprintk(" perm_choose swap %d with %d\n", p, p+i);
 		}
-		bucket->perm_n++;
+		work->perm_n++;
 	}
 	for (i = 0; i < bucket->size; i++)
-		dprintk(" perm_choose  %d: %d\n", i, bucket->perm[i]);
+		dprintk(" perm_choose  %d: %d\n", i, work->perm[i]);
 
-	s = bucket->perm[pr];
+	s = work->perm[pr];
 out:
 	dprintk(" perm_choose %d sz=%d x=%d r=%d (%d) s=%d\n", bucket->id,
 		bucket->size, x, r, pr, s);
@@ -132,14 +132,14 @@ out:
 }
 
 /* uniform */
-static int bucket_uniform_choose(struct crush_bucket_uniform *bucket,
-				 int x, int r)
+static int bucket_uniform_choose(const struct crush_bucket_uniform *bucket,
+				 struct crush_work_bucket *work, int x, int r)
 {
-	return bucket_perm_choose(&bucket->h, x, r);
+	return bucket_perm_choose(&bucket->h, work, x, r);
 }
 
 /* list */
-static int bucket_list_choose(struct crush_bucket_list *bucket,
+static int bucket_list_choose(const struct crush_bucket_list *bucket,
 			      int x, int r)
 {
 	int i;
@@ -155,8 +155,9 @@ static int bucket_list_choose(struct crush_bucket_list *bucket,
 		w *= bucket->sum_weights[i];
 		w = w >> 16;
 		/*dprintk(" scaled %llx\n", w);*/
-		if (w < bucket->item_weights[i])
+		if (w < bucket->item_weights[i]) {
 			return bucket->h.items[i];
+		}
 	}
 
 	dprintk("bad list sums for bucket %d\n", bucket->h.id);
@@ -192,7 +193,7 @@ static int terminal(int x)
 	return x & 1;
 }
 
-static int bucket_tree_choose(struct crush_bucket_tree *bucket,
+static int bucket_tree_choose(const struct crush_bucket_tree *bucket,
 			      int x, int r)
 {
 	int n;
@@ -224,7 +225,7 @@ static int bucket_tree_choose(struct crush_bucket_tree *bucket,
 
 /* straw */
 
-static int bucket_straw_choose(struct crush_bucket_straw *bucket,
+static int bucket_straw_choose(const struct crush_bucket_straw *bucket,
 			       int x, int r)
 {
 	__u32 i;
@@ -301,7 +302,7 @@ static __u64 crush_ln(unsigned int xin)
  *
  */
 
-static int bucket_straw2_choose(struct crush_bucket_straw2 *bucket,
+static int bucket_straw2_choose(const struct crush_bucket_straw2 *bucket,
 				int x, int r)
 {
 	unsigned int i, high = 0;
@@ -344,37 +345,42 @@ static int bucket_straw2_choose(struct crush_bucket_straw2 *bucket,
 			high_draw = draw;
 		}
 	}
+
 	return bucket->h.items[high];
 }
 
 
-static int crush_bucket_choose(struct crush_bucket *in, int x, int r)
+static int crush_bucket_choose(const struct crush_bucket *in,
+			       struct crush_work_bucket *work,
+			       int x, int r)
 {
 	dprintk(" crush_bucket_choose %d x=%d r=%d\n", in->id, x, r);
 	BUG_ON(in->size == 0);
 	switch (in->alg) {
 	case CRUSH_BUCKET_UNIFORM:
-		return bucket_uniform_choose((struct crush_bucket_uniform *)in,
-					  x, r);
+		return bucket_uniform_choose(
+			(const struct crush_bucket_uniform *)in,
+			work, x, r);
 	case CRUSH_BUCKET_LIST:
-		return bucket_list_choose((struct crush_bucket_list *)in,
+		return bucket_list_choose((const struct crush_bucket_list *)in,
 					  x, r);
 	case CRUSH_BUCKET_TREE:
-		return bucket_tree_choose((struct crush_bucket_tree *)in,
+		return bucket_tree_choose((const struct crush_bucket_tree *)in,
 					  x, r);
 	case CRUSH_BUCKET_STRAW:
-		return bucket_straw_choose((struct crush_bucket_straw *)in,
-					   x, r);
+		return bucket_straw_choose(
+			(const struct crush_bucket_straw *)in,
+			x, r);
 	case CRUSH_BUCKET_STRAW2:
-		return bucket_straw2_choose((struct crush_bucket_straw2 *)in,
-					    x, r);
+		return bucket_straw2_choose(
+			(const struct crush_bucket_straw2 *)in,
+			x, r);
 	default:
 		dprintk("unknown bucket %d alg %d\n", in->id, in->alg);
 		return in->items[0];
 	}
 }
 
-
 /*
  * true if device is marked "out" (failed, fully offloaded)
  * of the cluster
@@ -416,7 +422,8 @@ static int is_out(const struct crush_map *map,
  * @parent_r: r value passed from the parent
  */
 static int crush_choose_firstn(const struct crush_map *map,
-			       struct crush_bucket *bucket,
+			       struct crush_work *work,
+			       const struct crush_bucket *bucket,
 			       const __u32 *weight, int weight_max,
 			       int x, int numrep, int type,
 			       int *out, int outpos,
@@ -434,7 +441,7 @@ static int crush_choose_firstn(const struct crush_map *map,
 	int rep;
 	unsigned int ftotal, flocal;
 	int retry_descent, retry_bucket, skip_rep;
-	struct crush_bucket *in = bucket;
+	const struct crush_bucket *in = bucket;
 	int r;
 	int i;
 	int item = 0;
@@ -473,9 +480,13 @@ static int crush_choose_firstn(const struct crush_map *map,
 				if (local_fallback_retries > 0 &&
 				    flocal >= (in->size>>1) &&
 				    flocal > local_fallback_retries)
-					item = bucket_perm_choose(in, x, r);
+					item = bucket_perm_choose(
+						in, work->work[-1-in->id],
+						x, r);
 				else
-					item = crush_bucket_choose(in, x, r);
+					item = crush_bucket_choose(
+						in, work->work[-1-in->id],
+						x, r);
 				if (item >= map->max_devices) {
 					dprintk("   bad item %d\n", item);
 					skip_rep = 1;
@@ -518,19 +529,21 @@ static int crush_choose_firstn(const struct crush_map *map,
 							sub_r = r >> (vary_r-1);
 						else
 							sub_r = 0;
-						if (crush_choose_firstn(map,
-							 map->buckets[-1-item],
-							 weight, weight_max,
-							 x, stable ? 1 : outpos+1, 0,
-							 out2, outpos, count,
-							 recurse_tries, 0,
-							 local_retries,
-							 local_fallback_retries,
-							 0,
-							 vary_r,
-							 stable,
-							 NULL,
-							 sub_r) <= outpos)
+						if (crush_choose_firstn(
+							    map,
+							    work,
+							    map->buckets[-1-item],
+							    weight, weight_max,
+							    x, stable ? 1 : outpos+1, 0,
+							    out2, outpos, count,
+							    recurse_tries, 0,
+							    local_retries,
+							    local_fallback_retries,
+							    0,
+							    vary_r,
+							    stable,
+							    NULL,
+							    sub_r) <= outpos)
 							/* didn't get leaf */
 							reject = 1;
 					} else {
@@ -539,14 +552,12 @@ static int crush_choose_firstn(const struct crush_map *map,
 					}
 				}
 
-				if (!reject) {
+				if (!reject && !collide) {
 					/* out? */
 					if (itemtype == 0)
 						reject = is_out(map, weight,
 								weight_max,
 								item, x);
-					else
-						reject = 0;
 				}
 
 reject:
@@ -600,7 +611,8 @@ reject:
  *
  */
 static void crush_choose_indep(const struct crush_map *map,
-			       struct crush_bucket *bucket,
+			       struct crush_work *work,
+			       const struct crush_bucket *bucket,
 			       const __u32 *weight, int weight_max,
 			       int x, int left, int numrep, int type,
 			       int *out, int outpos,
@@ -610,7 +622,7 @@ static void crush_choose_indep(const struct crush_map *map,
 			       int *out2,
 			       int parent_r)
 {
-	struct crush_bucket *in = bucket;
+	const struct crush_bucket *in = bucket;
 	int endpos = outpos + left;
 	int rep;
 	unsigned int ftotal;
@@ -678,7 +690,9 @@ static void crush_choose_indep(const struct crush_map *map,
 					break;
 				}
 
-				item = crush_bucket_choose(in, x, r);
+				item = crush_bucket_choose(
+					in, work->work[-1-in->id],
+					x, r);
 				if (item >= map->max_devices) {
 					dprintk("   bad item %d\n", item);
 					out[rep] = CRUSH_ITEM_NONE;
@@ -724,13 +738,15 @@ static void crush_choose_indep(const struct crush_map *map,
 
 				if (recurse_to_leaf) {
 					if (item < 0) {
-						crush_choose_indep(map,
-						   map->buckets[-1-item],
-						   weight, weight_max,
-						   x, 1, numrep, 0,
-						   out2, rep,
-						   recurse_tries, 0,
-						   0, NULL, r);
+						crush_choose_indep(
+							map,
+							work,
+							map->buckets[-1-item],
+							weight, weight_max,
+							x, 1, numrep, 0,
+							out2, rep,
+							recurse_tries, 0,
+							0, NULL, r);
 						if (out2[rep] == CRUSH_ITEM_NONE) {
 							/* placed nothing; no leaf */
 							break;
@@ -781,6 +797,53 @@ static void crush_choose_indep(const struct crush_map *map,
 #endif
 }
 
+
+/*
+ * This takes a chunk of memory and sets it up to be a shiny new
+ * working area for a CRUSH placement computation. It must be called
+ * on any newly allocated memory before passing it in to
+ * crush_do_rule. It may be used repeatedly after that, so long as the
+ * map has not changed. If the map /has/ changed, you must make sure
+ * the working size is no smaller than what was allocated and re-run
+ * crush_init_workspace.
+ *
+ * If you do retain the working space between calls to crush, make it
+ * thread-local.
+ */
+void crush_init_workspace(const struct crush_map *map, void *v)
+{
+	struct crush_work *w = v;
+	__s32 b;
+
+	/*
+	 * We work by moving through the available space and setting
+	 * values and pointers as we go.
+	 *
+	 * It's a bit like Forth's use of the 'allot' word since we
+	 * set the pointer first and then reserve the space for it to
+	 * point to by incrementing the point.
+	 */
+	v += sizeof(struct crush_work *);
+	w->work = v;
+	v += map->max_buckets * sizeof(struct crush_work_bucket *);
+	for (b = 0; b < map->max_buckets; ++b) {
+		if (!map->buckets[b])
+			continue;
+
+		w->work[b] = v;
+		switch (map->buckets[b]->alg) {
+		default:
+			v += sizeof(struct crush_work_bucket);
+			break;
+		}
+		w->work[b]->perm_x = 0;
+		w->work[b]->perm_n = 0;
+		w->work[b]->perm = v;
+		v += map->buckets[b]->size * sizeof(__u32);
+	}
+	BUG_ON(v - (void *)w != map->working_size);
+}
+
 /**
  * crush_do_rule - calculate a mapping with the given input and rule
  * @map: the crush_map
@@ -790,24 +853,25 @@ static void crush_choose_indep(const struct crush_map *map,
  * @result_max: maximum result size
  * @weight: weight vector (for map leaves)
  * @weight_max: size of weight vector
- * @scratch: scratch vector for private use; must be >= 3 * result_max
+ * @cwin: pointer to at least crush_work_size() bytes of memory
  */
 int crush_do_rule(const struct crush_map *map,
 		  int ruleno, int x, int *result, int result_max,
 		  const __u32 *weight, int weight_max,
-		  int *scratch)
+		  void *cwin)
 {
 	int result_len;
-	int *a = scratch;
-	int *b = scratch + result_max;
-	int *c = scratch + result_max*2;
+	struct crush_work *cw = cwin;
+	int *a = cwin + map->working_size;
+	int *b = a + result_max;
+	int *c = b + result_max;
+	int *w = a;
+	int *o = b;
 	int recurse_to_leaf;
-	int *w;
 	int wsize = 0;
-	int *o;
 	int osize;
 	int *tmp;
-	struct crush_rule *rule;
+	const struct crush_rule *rule;
 	__u32 step;
 	int i, j;
 	int numrep;
@@ -835,12 +899,10 @@ int crush_do_rule(const struct crush_map *map,
 
 	rule = map->rules[ruleno];
 	result_len = 0;
-	w = a;
-	o = b;
 
 	for (step = 0; step < rule->len; step++) {
 		int firstn = 0;
-		struct crush_rule_step *curstep = &rule->steps[step];
+		const struct crush_rule_step *curstep = &rule->steps[step];
 
 		switch (curstep->op) {
 		case CRUSH_RULE_TAKE:
@@ -936,6 +998,7 @@ int crush_do_rule(const struct crush_map *map,
 						recurse_tries = choose_tries;
 					osize += crush_choose_firstn(
 						map,
+						cw,
 						map->buckets[bno],
 						weight, weight_max,
 						x, numrep,
@@ -956,6 +1019,7 @@ int crush_do_rule(const struct crush_map *map,
 						    numrep : (result_max-osize));
 					crush_choose_indep(
 						map,
+						cw,
 						map->buckets[bno],
 						weight, weight_max,
 						x, out_size, numrep,
@@ -997,5 +1061,6 @@ int crush_do_rule(const struct crush_map *map,
 			break;
 		}
 	}
+
 	return result_len;
 }

net/ceph/crypto.c

@@ -3,6 +3,7 @@
 
 #include <linux/err.h>
 #include <linux/scatterlist.h>
+#include <linux/sched.h>
 #include <linux/slab.h>
 #include <crypto/aes.h>
 #include <crypto/skcipher.h>

net/ceph/osd_client.c

@@ -460,7 +460,6 @@ static void request_init(struct ceph_osd_request *req)
 
 	kref_init(&req->r_kref);
 	init_completion(&req->r_completion);
-	init_completion(&req->r_done_completion);
 	RB_CLEAR_NODE(&req->r_node);
 	RB_CLEAR_NODE(&req->r_mc_node);
 	INIT_LIST_HEAD(&req->r_unsafe_item);
@@ -672,7 +671,8 @@ void osd_req_op_extent_update(struct ceph_osd_request *osd_req,
 	BUG_ON(length > previous);
 
 	op->extent.length = length;
-	op->indata_len -= previous - length;
+	if (op->op == CEPH_OSD_OP_WRITE || op->op == CEPH_OSD_OP_WRITEFULL)
+		op->indata_len -= previous - length;
 }
 EXPORT_SYMBOL(osd_req_op_extent_update);
 
@@ -1636,7 +1636,7 @@ static void __submit_request(struct ceph_osd_request *req, bool wrlocked)
 	bool need_send = false;
 	bool promoted = false;
 
-	WARN_ON(req->r_tid || req->r_got_reply);
+	WARN_ON(req->r_tid);
 	dout("%s req %p wrlocked %d\n", __func__, req, wrlocked);
 
 again:
@@ -1704,17 +1704,10 @@ promote:
 
 static void account_request(struct ceph_osd_request *req)
 {
-	unsigned int mask = CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK;
+	WARN_ON(req->r_flags & (CEPH_OSD_FLAG_ACK | CEPH_OSD_FLAG_ONDISK));
+	WARN_ON(!(req->r_flags & (CEPH_OSD_FLAG_READ | CEPH_OSD_FLAG_WRITE)));
 
-	if (req->r_flags & CEPH_OSD_FLAG_READ) {
-		WARN_ON(req->r_flags & mask);
-		req->r_flags |= CEPH_OSD_FLAG_ACK;
-	} else if (req->r_flags & CEPH_OSD_FLAG_WRITE)
-		WARN_ON(!(req->r_flags & mask));
-	else
-		WARN_ON(1);
-
-	WARN_ON(req->r_unsafe_callback && (req->r_flags & mask) != mask);
+	req->r_flags |= CEPH_OSD_FLAG_ONDISK;
 	atomic_inc(&req->r_osdc->num_requests);
 }
 
@@ -1749,15 +1742,15 @@ static void finish_request(struct ceph_osd_request *req)
 
 static void __complete_request(struct ceph_osd_request *req)
 {
-	if (req->r_callback)
+	if (req->r_callback) {
+		dout("%s req %p tid %llu cb %pf result %d\n", __func__, req,
+		     req->r_tid, req->r_callback, req->r_result);
 		req->r_callback(req);
-	else
-		complete_all(&req->r_completion);
+	}
 }
 
 /*
- * Note that this is open-coded in handle_reply(), which has to deal
- * with ack vs commit, dup acks, etc.
+ * This is open-coded in handle_reply().
  */
 static void complete_request(struct ceph_osd_request *req, int err)
 {
@@ -1766,7 +1759,7 @@ static void complete_request(struct ceph_osd_request *req, int err)
 	req->r_result = err;
 	finish_request(req);
 	__complete_request(req);
-	complete_all(&req->r_done_completion);
+	complete_all(&req->r_completion);
 	ceph_osdc_put_request(req);
 }
 
@@ -1792,7 +1785,7 @@ static void cancel_request(struct ceph_osd_request *req)
 
 	cancel_map_check(req);
 	finish_request(req);
-	complete_all(&req->r_done_completion);
+	complete_all(&req->r_completion);
 	ceph_osdc_put_request(req);
 }
 
@@ -2169,7 +2162,6 @@ static void linger_commit_cb(struct ceph_osd_request *req)
 	mutex_lock(&lreq->lock);
 	dout("%s lreq %p linger_id %llu result %d\n", __func__, lreq,
 	     lreq->linger_id, req->r_result);
-	WARN_ON(!__linger_registered(lreq));
 	linger_reg_commit_complete(lreq, req->r_result);
 	lreq->committed = true;
 
@@ -2785,31 +2777,8 @@ e_inval:
 }
 
 /*
- * We are done with @req if
- *   - @m is a safe reply, or
- *   - @m is an unsafe reply and we didn't want a safe one
- */
-static bool done_request(const struct ceph_osd_request *req,
-			 const struct MOSDOpReply *m)
-{
-	return (m->result < 0 ||
-		(m->flags & CEPH_OSD_FLAG_ONDISK) ||
-		!(req->r_flags & CEPH_OSD_FLAG_ONDISK));
-}
-
-/*
- * handle osd op reply.  either call the callback if it is specified,
- * or do the completion to wake up the waiting thread.
- *
- * ->r_unsafe_callback is set?	yes			no
- *
- * first reply is OK (needed	r_cb/r_completion,	r_cb/r_completion,
- * any or needed/got safe)	r_done_completion	r_done_completion
- *
- * first reply is unsafe	r_unsafe_cb(true)	(nothing)
- *
- * when we get the safe reply	r_unsafe_cb(false),	r_cb/r_completion,
- *				r_done_completion	r_done_completion
+ * Handle MOSDOpReply.  Set ->r_result and call the callback if it is
+ * specified.
  */
 static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
 {
@@ -2818,7 +2787,6 @@ static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
 	struct MOSDOpReply m;
 	u64 tid = le64_to_cpu(msg->hdr.tid);
 	u32 data_len = 0;
-	bool already_acked;
 	int ret;
 	int i;
 
@@ -2897,50 +2865,22 @@ static void handle_reply(struct ceph_osd *osd, struct ceph_msg *msg)
 		       le32_to_cpu(msg->hdr.data_len), req->r_tid);
 		goto fail_request;
 	}
-	dout("%s req %p tid %llu acked %d result %d data_len %u\n", __func__,
-	     req, req->r_tid, req->r_got_reply, m.result, data_len);
-
-	already_acked = req->r_got_reply;
-	if (!already_acked) {
-		req->r_result = m.result ?: data_len;
-		req->r_replay_version = m.replay_version; /* struct */
-		req->r_got_reply = true;
-	} else if (!(m.flags & CEPH_OSD_FLAG_ONDISK)) {
-		dout("req %p tid %llu dup ack\n", req, req->r_tid);
-		goto out_unlock_session;
-	}
-
-	if (done_request(req, &m)) {
-		finish_request(req);
-		if (req->r_linger) {
-			WARN_ON(req->r_unsafe_callback);
-			dout("req %p tid %llu cb (locked)\n", req, req->r_tid);
-			__complete_request(req);
-		}
-	}
+	dout("%s req %p tid %llu result %d data_len %u\n", __func__,
+	     req, req->r_tid, m.result, data_len);
 
+	/*
+	 * Since we only ever request ONDISK, we should only ever get
+	 * one (type of) reply back.
+	 */
+	WARN_ON(!(m.flags & CEPH_OSD_FLAG_ONDISK));
+	req->r_result = m.result ?: data_len;
+	finish_request(req);
 	mutex_unlock(&osd->lock);
 	up_read(&osdc->lock);
 
-	if (done_request(req, &m)) {
-		if (already_acked && req->r_unsafe_callback) {
-			dout("req %p tid %llu safe-cb\n", req, req->r_tid);
-			req->r_unsafe_callback(req, false);
-		} else if (!req->r_linger) {
-			dout("req %p tid %llu cb\n", req, req->r_tid);
-			__complete_request(req);
-		}
-		complete_all(&req->r_done_completion);
-		ceph_osdc_put_request(req);
-	} else {
-		if (req->r_unsafe_callback) {
-			dout("req %p tid %llu unsafe-cb\n", req, req->r_tid);
-			req->r_unsafe_callback(req, true);
-		} else {
-			WARN_ON(1);
-		}
-	}
-
+	__complete_request(req);
+	complete_all(&req->r_completion);
+	ceph_osdc_put_request(req);
 	return;
 
 fail_request:
@@ -3540,7 +3480,7 @@ again:
 			up_read(&osdc->lock);
 			dout("%s waiting on req %p tid %llu last_tid %llu\n",
 			     __func__, req, req->r_tid, last_tid);
-			wait_for_completion(&req->r_done_completion);
+			wait_for_completion(&req->r_completion);
 			ceph_osdc_put_request(req);
 			goto again;
 		}
@@ -3599,7 +3539,7 @@ ceph_osdc_watch(struct ceph_osd_client *osdc,
 
 	ceph_oid_copy(&lreq->t.base_oid, oid);
 	ceph_oloc_copy(&lreq->t.base_oloc, oloc);
-	lreq->t.flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK;
+	lreq->t.flags = CEPH_OSD_FLAG_WRITE;
 	lreq->mtime = CURRENT_TIME;
 
 	lreq->reg_req = alloc_linger_request(lreq);
@@ -3657,7 +3597,7 @@ int ceph_osdc_unwatch(struct ceph_osd_client *osdc,
 
 	ceph_oid_copy(&req->r_base_oid, &lreq->t.base_oid);
 	ceph_oloc_copy(&req->r_base_oloc, &lreq->t.base_oloc);
-	req->r_flags = CEPH_OSD_FLAG_WRITE | CEPH_OSD_FLAG_ONDISK;
+	req->r_flags = CEPH_OSD_FLAG_WRITE;
 	req->r_mtime = CURRENT_TIME;
 	osd_req_op_watch_init(req, 0, lreq->linger_id,
 			      CEPH_OSD_WATCH_OP_UNWATCH);
@@ -4022,7 +3962,7 @@ EXPORT_SYMBOL(ceph_osdc_maybe_request_map);
  * Execute an OSD class method on an object.
  *
  * @flags: CEPH_OSD_FLAG_*
- * @resp_len: out param for reply length
+ * @resp_len: in/out param for reply length
  */
 int ceph_osdc_call(struct ceph_osd_client *osdc,
 		   struct ceph_object_id *oid,
@@ -4035,6 +3975,9 @@ int ceph_osdc_call(struct ceph_osd_client *osdc,
 	struct ceph_osd_request *req;
 	int ret;
 
+	if (req_len > PAGE_SIZE || (resp_page && *resp_len > PAGE_SIZE))
+		return -E2BIG;
+
 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
 	if (!req)
 		return -ENOMEM;
@@ -4053,7 +3996,7 @@ int ceph_osdc_call(struct ceph_osd_client *osdc,
 						  0, false, false);
 	if (resp_page)
 		osd_req_op_cls_response_data_pages(req, 0, &resp_page,
-						   PAGE_SIZE, 0, false, false);
+						   *resp_len, 0, false, false);
 
 	ceph_osdc_start_request(osdc, req, false);
 	ret = ceph_osdc_wait_request(osdc, req);
@@ -4220,8 +4163,7 @@ int ceph_osdc_writepages(struct ceph_osd_client *osdc, struct ceph_vino vino,
 	int page_align = off & ~PAGE_MASK;
 
 	req = ceph_osdc_new_request(osdc, layout, vino, off, &len, 0, 1,
-				    CEPH_OSD_OP_WRITE,
-				    CEPH_OSD_FLAG_ONDISK | CEPH_OSD_FLAG_WRITE,
+				    CEPH_OSD_OP_WRITE, CEPH_OSD_FLAG_WRITE,
 				    snapc, truncate_seq, truncate_size,
 				    true);
 	if (IS_ERR(req))

net/ceph/osdmap.c

@@ -153,6 +153,32 @@ bad:
         return -EINVAL;
 }
 
+static void crush_finalize(struct crush_map *c)
+{
+	__s32 b;
+
+	/* Space for the array of pointers to per-bucket workspace */
+	c->working_size = sizeof(struct crush_work) +
+	    c->max_buckets * sizeof(struct crush_work_bucket *);
+
+	for (b = 0; b < c->max_buckets; b++) {
+		if (!c->buckets[b])
+			continue;
+
+		switch (c->buckets[b]->alg) {
+		default:
+			/*
+			 * The base case, permutation variables and
+			 * the pointer to the permutation array.
+			 */
+			c->working_size += sizeof(struct crush_work_bucket);
+			break;
+		}
+		/* Every bucket has a permutation array. */
+		c->working_size += c->buckets[b]->size * sizeof(__u32);
+	}
+}
+
 static struct crush_map *crush_decode(void *pbyval, void *end)
 {
 	struct crush_map *c;
@@ -246,10 +272,6 @@ static struct crush_map *crush_decode(void *pbyval, void *end)
 		b->items = kcalloc(b->size, sizeof(__s32), GFP_NOFS);
 		if (b->items == NULL)
 			goto badmem;
-		b->perm = kcalloc(b->size, sizeof(u32), GFP_NOFS);
-		if (b->perm == NULL)
-			goto badmem;
-		b->perm_n = 0;
 
 		ceph_decode_need(p, end, b->size*sizeof(u32), bad);
 		for (j = 0; j < b->size; j++)
@@ -368,6 +390,8 @@ static struct crush_map *crush_decode(void *pbyval, void *end)
 	dout("crush decode tunable chooseleaf_stable = %d\n",
 	     c->chooseleaf_stable);
 
+	crush_finalize(c);
+
 done:
 	dout("crush_decode success\n");
 	return c;
@@ -719,7 +743,7 @@ struct ceph_osdmap *ceph_osdmap_alloc(void)
 	map->pool_max = -1;
 	map->pg_temp = RB_ROOT;
 	map->primary_temp = RB_ROOT;
-	mutex_init(&map->crush_scratch_mutex);
+	mutex_init(&map->crush_workspace_mutex);
 
 	return map;
 }
@@ -753,6 +777,7 @@ void ceph_osdmap_destroy(struct ceph_osdmap *map)
 	kfree(map->osd_weight);
 	kfree(map->osd_addr);
 	kfree(map->osd_primary_affinity);
+	kfree(map->crush_workspace);
 	kfree(map);
 }
 
@@ -808,6 +833,31 @@ static int osdmap_set_max_osd(struct ceph_osdmap *map, int max)
 	return 0;
 }
 
+static int osdmap_set_crush(struct ceph_osdmap *map, struct crush_map *crush)
+{
+	void *workspace;
+	size_t work_size;
+
+	if (IS_ERR(crush))
+		return PTR_ERR(crush);
+
+	work_size = crush_work_size(crush, CEPH_PG_MAX_SIZE);
+	dout("%s work_size %zu bytes\n", __func__, work_size);
+	workspace = kmalloc(work_size, GFP_NOIO);
+	if (!workspace) {
+		crush_destroy(crush);
+		return -ENOMEM;
+	}
+	crush_init_workspace(crush, workspace);
+
+	if (map->crush)
+		crush_destroy(map->crush);
+	kfree(map->crush_workspace);
+	map->crush = crush;
+	map->crush_workspace = workspace;
+	return 0;
+}
+
 #define OSDMAP_WRAPPER_COMPAT_VER	7
 #define OSDMAP_CLIENT_DATA_COMPAT_VER	1
 
@@ -1214,13 +1264,9 @@ static int osdmap_decode(void **p, void *end, struct ceph_osdmap *map)
 
 	/* crush */
 	ceph_decode_32_safe(p, end, len, e_inval);
-	map->crush = crush_decode(*p, min(*p + len, end));
-	if (IS_ERR(map->crush)) {
-		err = PTR_ERR(map->crush);
-		map->crush = NULL;
+	err = osdmap_set_crush(map, crush_decode(*p, min(*p + len, end)));
+	if (err)
 		goto bad;
-	}
-	*p += len;
 
 	/* ignore the rest */
 	*p = end;
@@ -1375,7 +1421,6 @@ e_inval:
 struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 					     struct ceph_osdmap *map)
 {
-	struct crush_map *newcrush = NULL;
 	struct ceph_fsid fsid;
 	u32 epoch = 0;
 	struct ceph_timespec modified;
@@ -1414,12 +1459,10 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 	/* new crush? */
 	ceph_decode_32_safe(p, end, len, e_inval);
 	if (len > 0) {
-		newcrush = crush_decode(*p, min(*p+len, end));
-		if (IS_ERR(newcrush)) {
-			err = PTR_ERR(newcrush);
-			newcrush = NULL;
+		err = osdmap_set_crush(map,
+				       crush_decode(*p, min(*p + len, end)));
+		if (err)
 			goto bad;
-		}
 		*p += len;
 	}
 
@@ -1439,12 +1482,6 @@ struct ceph_osdmap *osdmap_apply_incremental(void **p, void *end,
 
 	map->epoch++;
 	map->modified = modified;
-	if (newcrush) {
-		if (map->crush)
-			crush_destroy(map->crush);
-		map->crush = newcrush;
-		newcrush = NULL;
-	}
 
 	/* new_pools */
 	err = decode_new_pools(p, end, map);
@@ -1505,8 +1542,6 @@ bad:
 	print_hex_dump(KERN_DEBUG, "osdmap: ",
 		       DUMP_PREFIX_OFFSET, 16, 1,
 		       start, end - start, true);
-	if (newcrush)
-		crush_destroy(newcrush);
 	return ERR_PTR(err);
 }
 
@@ -1942,10 +1977,10 @@ static int do_crush(struct ceph_osdmap *map, int ruleno, int x,
 
 	BUG_ON(result_max > CEPH_PG_MAX_SIZE);
 
-	mutex_lock(&map->crush_scratch_mutex);
+	mutex_lock(&map->crush_workspace_mutex);
 	r = crush_do_rule(map->crush, ruleno, x, result, result_max,
-			  weight, weight_max, map->crush_scratch_ary);
-	mutex_unlock(&map->crush_scratch_mutex);
+			  weight, weight_max, map->crush_workspace);
+	mutex_unlock(&map->crush_workspace_mutex);
 
 	return r;
 }
@@ -1978,8 +2013,14 @@ static void pg_to_raw_osds(struct ceph_osdmap *osdmap,
 		return;
 	}
 
-	len = do_crush(osdmap, ruleno, pps, raw->osds,
-		       min_t(int, pi->size, ARRAY_SIZE(raw->osds)),
+	if (pi->size > ARRAY_SIZE(raw->osds)) {
+		pr_err_ratelimited("pool %lld ruleset %d type %d too wide: size %d > %zu\n",
+		       pi->id, pi->crush_ruleset, pi->type, pi->size,
+		       ARRAY_SIZE(raw->osds));
+		return;
+	}
+
+	len = do_crush(osdmap, ruleno, pps, raw->osds, pi->size,
 		       osdmap->osd_weight, osdmap->max_osd);
 	if (len < 0) {
 		pr_err("error %d from crush rule %d: pool %lld ruleset %d type %d size %d\n",

net/ceph/snapshot.c

@@ -18,8 +18,6 @@
  * 02110-1301, USA.
  */
 
-#include <stddef.h>
-
 #include <linux/types.h>
 #include <linux/export.h>
 #include <linux/ceph/libceph.h>