Merge branch 'layoutstats'

* layoutstats:
  pnfs/flexfiles: protect ktime manipulation with mirror lock
  nfs: provide pnfs_report_layoutstat when NFS42 is disabled
  pnfs/flexfiles: report layoutstat regularly
  nfs42: serialize LAYOUTSTATS calls of the same file
  pnfs/flexfiles: encode LAYOUTSTATS flexfiles specific data
  pnfs/flexfiles: add ff_layout_prepare_layoutstats
  pNFS/flexfiles: track when layout is first used
  pNFS/flexfiles: add layoutstats tracking
  pNFS/flexfiles: Remove unused struct members user_name, group_name
  pnfs: add pnfs_report_layoutstat helper function
  pNFS: fill in nfs42_layoutstat_ops
  NFSv.2/pnfs Add a LAYOUTSTATS rpc function

fs/nfs/flexfilelayout/flexfilelayout.c

@@ -20,6 +20,7 @@
 #include "../nfs4trace.h"
 #include "../iostat.h"
 #include "../nfs.h"
+#include "../nfs42.h"
 
 #define NFSDBG_FACILITY         NFSDBG_PNFS_LD
 
@@ -271,6 +272,7 @@ ff_layout_alloc_lseg(struct pnfs_layout_hdr *lh,
 
 		spin_lock_init(&fls->mirror_array[i]->lock);
 		fls->mirror_array[i]->ds_count = ds_count;
+		fls->mirror_array[i]->lseg = &fls->generic_hdr;
 
 		/* deviceid */
 		rc = decode_deviceid(&stream, &devid);
@@ -412,6 +414,146 @@ ff_layout_get_lseg_count(struct nfs4_ff_layout_segment *fls)
 	return 1;
 }
 
+static void
+nfs4_ff_start_busy_timer(struct nfs4_ff_busy_timer *timer)
+{
+	/* first IO request? */
+	if (atomic_inc_return(&timer->n_ops) == 1) {
+		timer->start_time = ktime_get();
+	}
+}
+
+static ktime_t
+nfs4_ff_end_busy_timer(struct nfs4_ff_busy_timer *timer)
+{
+	ktime_t start, now;
+
+	if (atomic_dec_return(&timer->n_ops) < 0)
+		WARN_ON_ONCE(1);
+
+	now = ktime_get();
+	start = timer->start_time;
+	timer->start_time = now;
+	return ktime_sub(now, start);
+}
+
+static ktime_t
+nfs4_ff_layout_calc_completion_time(struct rpc_task *task)
+{
+	return ktime_sub(ktime_get(), task->tk_start);
+}
+
+static bool
+nfs4_ff_layoutstat_start_io(struct nfs4_ff_layout_mirror *mirror,
+			    struct nfs4_ff_layoutstat *layoutstat)
+{
+	static const ktime_t notime = {0};
+	ktime_t now = ktime_get();
+
+	nfs4_ff_start_busy_timer(&layoutstat->busy_timer);
+	if (ktime_equal(mirror->start_time, notime))
+		mirror->start_time = now;
+	if (ktime_equal(mirror->last_report_time, notime))
+		mirror->last_report_time = now;
+	if (ktime_to_ms(ktime_sub(now, mirror->last_report_time)) >=
+			FF_LAYOUTSTATS_REPORT_INTERVAL) {
+		mirror->last_report_time = now;
+		return true;
+	}
+
+	return false;
+}
+
+static void
+nfs4_ff_layout_stat_io_update_requested(struct nfs4_ff_layoutstat *layoutstat,
+		__u64 requested)
+{
+	struct nfs4_ff_io_stat *iostat = &layoutstat->io_stat;
+
+	iostat->ops_requested++;
+	iostat->bytes_requested += requested;
+}
+
+static void
+nfs4_ff_layout_stat_io_update_completed(struct nfs4_ff_layoutstat *layoutstat,
+		__u64 requested,
+		__u64 completed,
+		ktime_t time_completed)
+{
+	struct nfs4_ff_io_stat *iostat = &layoutstat->io_stat;
+	ktime_t timer;
+
+	iostat->ops_completed++;
+	iostat->bytes_completed += completed;
+	iostat->bytes_not_delivered += requested - completed;
+
+	timer = nfs4_ff_end_busy_timer(&layoutstat->busy_timer);
+	iostat->total_busy_time =
+			ktime_add(iostat->total_busy_time, timer);
+	iostat->aggregate_completion_time =
+			ktime_add(iostat->aggregate_completion_time, time_completed);
+}
+
+static void
+nfs4_ff_layout_stat_io_start_read(struct nfs4_ff_layout_mirror *mirror,
+		__u64 requested)
+{
+	bool report;
+
+	spin_lock(&mirror->lock);
+	report = nfs4_ff_layoutstat_start_io(mirror, &mirror->read_stat);
+	nfs4_ff_layout_stat_io_update_requested(&mirror->read_stat, requested);
+	spin_unlock(&mirror->lock);
+
+	if (report)
+		pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode);
+}
+
+static void
+nfs4_ff_layout_stat_io_end_read(struct rpc_task *task,
+		struct nfs4_ff_layout_mirror *mirror,
+		__u64 requested,
+		__u64 completed)
+{
+	spin_lock(&mirror->lock);
+	nfs4_ff_layout_stat_io_update_completed(&mirror->read_stat,
+			requested, completed,
+			nfs4_ff_layout_calc_completion_time(task));
+	spin_unlock(&mirror->lock);
+}
+
+static void
+nfs4_ff_layout_stat_io_start_write(struct nfs4_ff_layout_mirror *mirror,
+		__u64 requested)
+{
+	bool report;
+
+	spin_lock(&mirror->lock);
+	report = nfs4_ff_layoutstat_start_io(mirror , &mirror->write_stat);
+	nfs4_ff_layout_stat_io_update_requested(&mirror->write_stat, requested);
+	spin_unlock(&mirror->lock);
+
+	if (report)
+		pnfs_report_layoutstat(mirror->lseg->pls_layout->plh_inode);
+}
+
+static void
+nfs4_ff_layout_stat_io_end_write(struct rpc_task *task,
+		struct nfs4_ff_layout_mirror *mirror,
+		__u64 requested,
+		__u64 completed,
+		enum nfs3_stable_how committed)
+{
+	if (committed == NFS_UNSTABLE)
+		requested = completed = 0;
+
+	spin_lock(&mirror->lock);
+	nfs4_ff_layout_stat_io_update_completed(&mirror->write_stat,
+			requested, completed,
+			nfs4_ff_layout_calc_completion_time(task));
+	spin_unlock(&mirror->lock);
+}
+
 static int
 ff_layout_alloc_commit_info(struct pnfs_layout_segment *lseg,
 			    struct nfs_commit_info *cinfo,
@@ -906,6 +1048,10 @@ ff_layout_reset_to_mds(struct pnfs_layout_segment *lseg, int idx)
 static int ff_layout_read_prepare_common(struct rpc_task *task,
 					 struct nfs_pgio_header *hdr)
 {
+	nfs4_ff_layout_stat_io_start_read(
+			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
+			hdr->args.count);
+
 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags))) {
 		rpc_exit(task, -EIO);
 		return -EIO;
@@ -959,15 +1105,15 @@ static void ff_layout_read_prepare_v4(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
 
-	if (ff_layout_read_prepare_common(task, hdr))
-		return;
-
 	if (ff_layout_setup_sequence(hdr->ds_clp,
 				     &hdr->args.seq_args,
 				     &hdr->res.seq_res,
 				     task))
 		return;
 
+	if (ff_layout_read_prepare_common(task, hdr))
+		return;
+
 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
 			hdr->args.lock_context, FMODE_READ) == -EIO)
 		rpc_exit(task, -EIO); /* lost lock, terminate I/O */
@@ -979,6 +1125,10 @@ static void ff_layout_read_call_done(struct rpc_task *task, void *data)
 
 	dprintk("--> %s task->tk_status %d\n", __func__, task->tk_status);
 
+	nfs4_ff_layout_stat_io_end_read(task,
+			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
+			hdr->args.count, hdr->res.count);
+
 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags) &&
 	    task->tk_status == 0) {
 		nfs4_sequence_done(task, &hdr->res.seq_res);
@@ -1080,6 +1230,10 @@ static int ff_layout_commit_done_cb(struct rpc_task *task,
 static int ff_layout_write_prepare_common(struct rpc_task *task,
 					  struct nfs_pgio_header *hdr)
 {
+	nfs4_ff_layout_stat_io_start_write(
+			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
+			hdr->args.count);
+
 	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags))) {
 		rpc_exit(task, -EIO);
 		return -EIO;
@@ -1113,15 +1267,15 @@ static void ff_layout_write_prepare_v4(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
 
-	if (ff_layout_write_prepare_common(task, hdr))
-		return;
-
 	if (ff_layout_setup_sequence(hdr->ds_clp,
 				     &hdr->args.seq_args,
 				     &hdr->res.seq_res,
 				     task))
 		return;
 
+	if (ff_layout_write_prepare_common(task, hdr))
+		return;
+
 	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
 			hdr->args.lock_context, FMODE_WRITE) == -EIO)
 		rpc_exit(task, -EIO); /* lost lock, terminate I/O */
@@ -1131,6 +1285,11 @@ static void ff_layout_write_call_done(struct rpc_task *task, void *data)
 {
 	struct nfs_pgio_header *hdr = data;
 
+	nfs4_ff_layout_stat_io_end_write(task,
+			FF_LAYOUT_COMP(hdr->lseg, hdr->pgio_mirror_idx),
+			hdr->args.count, hdr->res.count,
+			hdr->res.verf->committed);
+
 	if (test_bit(NFS_IOHDR_REDO, &hdr->flags) &&
 	    task->tk_status == 0) {
 		nfs4_sequence_done(task, &hdr->res.seq_res);
@@ -1149,8 +1308,17 @@ static void ff_layout_write_count_stats(struct rpc_task *task, void *data)
 	    &NFS_CLIENT(hdr->inode)->cl_metrics[NFSPROC4_CLNT_WRITE]);
 }
 
+static void ff_layout_commit_prepare_common(struct rpc_task *task,
+		struct nfs_commit_data *cdata)
+{
+	nfs4_ff_layout_stat_io_start_write(
+			FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index),
+			0);
+}
+
 static void ff_layout_commit_prepare_v3(struct rpc_task *task, void *data)
 {
+	ff_layout_commit_prepare_common(task, data);
 	rpc_call_start(task);
 }
 
@@ -1158,10 +1326,30 @@ static void ff_layout_commit_prepare_v4(struct rpc_task *task, void *data)
 {
 	struct nfs_commit_data *wdata = data;
 
-	ff_layout_setup_sequence(wdata->ds_clp,
+	if (ff_layout_setup_sequence(wdata->ds_clp,
 				 &wdata->args.seq_args,
 				 &wdata->res.seq_res,
-				 task);
+				 task))
+		return;
+	ff_layout_commit_prepare_common(task, data);
+}
+
+static void ff_layout_commit_done(struct rpc_task *task, void *data)
+{
+	struct nfs_commit_data *cdata = data;
+	struct nfs_page *req;
+	__u64 count = 0;
+
+	if (task->tk_status == 0) {
+		list_for_each_entry(req, &cdata->pages, wb_list)
+			count += req->wb_bytes;
+	}
+
+	nfs4_ff_layout_stat_io_end_write(task,
+			FF_LAYOUT_COMP(cdata->lseg, cdata->ds_commit_index),
+			count, count, NFS_FILE_SYNC);
+
+	pnfs_generic_write_commit_done(task, data);
 }
 
 static void ff_layout_commit_count_stats(struct rpc_task *task, void *data)
@@ -1202,14 +1390,14 @@ static const struct rpc_call_ops ff_layout_write_call_ops_v4 = {
 
 static const struct rpc_call_ops ff_layout_commit_call_ops_v3 = {
 	.rpc_call_prepare = ff_layout_commit_prepare_v3,
-	.rpc_call_done = pnfs_generic_write_commit_done,
+	.rpc_call_done = ff_layout_commit_done,
 	.rpc_count_stats = ff_layout_commit_count_stats,
 	.rpc_release = pnfs_generic_commit_release,
 };
 
 static const struct rpc_call_ops ff_layout_commit_call_ops_v4 = {
 	.rpc_call_prepare = ff_layout_commit_prepare_v4,
-	.rpc_call_done = pnfs_generic_write_commit_done,
+	.rpc_call_done = ff_layout_commit_done,
 	.rpc_count_stats = ff_layout_commit_count_stats,
 	.rpc_release = pnfs_generic_commit_release,
 };
@@ -1253,7 +1441,6 @@ ff_layout_read_pagelist(struct nfs_pgio_header *hdr)
 	fh = nfs4_ff_layout_select_ds_fh(lseg, idx);
 	if (fh)
 		hdr->args.fh = fh;
-
 	/*
 	 * Note that if we ever decide to split across DSes,
 	 * then we may need to handle dense-like offsets.
@@ -1382,6 +1569,7 @@ static int ff_layout_initiate_commit(struct nfs_commit_data *data, int how)
 	fh = select_ds_fh_from_commit(lseg, data->ds_commit_index);
 	if (fh)
 		data->args.fh = fh;
+
 	return nfs_initiate_commit(ds_clnt, data, ds->ds_clp->rpc_ops,
 				   vers == 3 ? &ff_layout_commit_call_ops_v3 :
 					       &ff_layout_commit_call_ops_v4,
@@ -1485,6 +1673,247 @@ ff_layout_encode_layoutreturn(struct pnfs_layout_hdr *lo,
 	dprintk("%s: Return\n", __func__);
 }
 
+static int
+ff_layout_ntop4(const struct sockaddr *sap, char *buf, const size_t buflen)
+{
+	const struct sockaddr_in *sin = (struct sockaddr_in *)sap;
+
+	return snprintf(buf, buflen, "%pI4", &sin->sin_addr);
+}
+
+static size_t
+ff_layout_ntop6_noscopeid(const struct sockaddr *sap, char *buf,
+			  const int buflen)
+{
+	const struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
+	const struct in6_addr *addr = &sin6->sin6_addr;
+
+	/*
+	 * RFC 4291, Section 2.2.2
+	 *
+	 * Shorthanded ANY address
+	 */
+	if (ipv6_addr_any(addr))
+		return snprintf(buf, buflen, "::");
+
+	/*
+	 * RFC 4291, Section 2.2.2
+	 *
+	 * Shorthanded loopback address
+	 */
+	if (ipv6_addr_loopback(addr))
+		return snprintf(buf, buflen, "::1");
+
+	/*
+	 * RFC 4291, Section 2.2.3
+	 *
+	 * Special presentation address format for mapped v4
+	 * addresses.
+	 */
+	if (ipv6_addr_v4mapped(addr))
+		return snprintf(buf, buflen, "::ffff:%pI4",
+					&addr->s6_addr32[3]);
+
+	/*
+	 * RFC 4291, Section 2.2.1
+	 */
+	return snprintf(buf, buflen, "%pI6c", addr);
+}
+
+/* Derived from rpc_sockaddr2uaddr */
+static void
+ff_layout_encode_netaddr(struct xdr_stream *xdr, struct nfs4_pnfs_ds_addr *da)
+{
+	struct sockaddr *sap = (struct sockaddr *)&da->da_addr;
+	char portbuf[RPCBIND_MAXUADDRPLEN];
+	char addrbuf[RPCBIND_MAXUADDRLEN];
+	char *netid;
+	unsigned short port;
+	int len, netid_len;
+	__be32 *p;
+
+	switch (sap->sa_family) {
+	case AF_INET:
+		if (ff_layout_ntop4(sap, addrbuf, sizeof(addrbuf)) == 0)
+			return;
+		port = ntohs(((struct sockaddr_in *)sap)->sin_port);
+		netid = "tcp";
+		netid_len = 3;
+		break;
+	case AF_INET6:
+		if (ff_layout_ntop6_noscopeid(sap, addrbuf, sizeof(addrbuf)) == 0)
+			return;
+		port = ntohs(((struct sockaddr_in6 *)sap)->sin6_port);
+		netid = "tcp6";
+		netid_len = 4;
+		break;
+	default:
+		/* we only support tcp and tcp6 */
+		WARN_ON_ONCE(1);
+		return;
+	}
+
+	snprintf(portbuf, sizeof(portbuf), ".%u.%u", port >> 8, port & 0xff);
+	len = strlcat(addrbuf, portbuf, sizeof(addrbuf));
+
+	p = xdr_reserve_space(xdr, 4 + netid_len);
+	xdr_encode_opaque(p, netid, netid_len);
+
+	p = xdr_reserve_space(xdr, 4 + len);
+	xdr_encode_opaque(p, addrbuf, len);
+}
+
+static void
+ff_layout_encode_nfstime(struct xdr_stream *xdr,
+			 ktime_t t)
+{
+	struct timespec64 ts;
+	__be32 *p;
+
+	p = xdr_reserve_space(xdr, 12);
+	ts = ktime_to_timespec64(t);
+	p = xdr_encode_hyper(p, ts.tv_sec);
+	*p++ = cpu_to_be32(ts.tv_nsec);
+}
+
+static void
+ff_layout_encode_io_latency(struct xdr_stream *xdr,
+			    struct nfs4_ff_io_stat *stat)
+{
+	__be32 *p;
+
+	p = xdr_reserve_space(xdr, 5 * 8);
+	p = xdr_encode_hyper(p, stat->ops_requested);
+	p = xdr_encode_hyper(p, stat->bytes_requested);
+	p = xdr_encode_hyper(p, stat->ops_completed);
+	p = xdr_encode_hyper(p, stat->bytes_completed);
+	p = xdr_encode_hyper(p, stat->bytes_not_delivered);
+	ff_layout_encode_nfstime(xdr, stat->total_busy_time);
+	ff_layout_encode_nfstime(xdr, stat->aggregate_completion_time);
+}
+
+static void
+ff_layout_encode_layoutstats(struct xdr_stream *xdr,
+			     struct nfs42_layoutstat_args *args,
+			     struct nfs42_layoutstat_devinfo *devinfo)
+{
+	struct nfs4_ff_layout_mirror *mirror = devinfo->layout_private;
+	struct nfs4_pnfs_ds_addr *da;
+	struct nfs4_pnfs_ds *ds = mirror->mirror_ds->ds;
+	struct nfs_fh *fh = &mirror->fh_versions[0];
+	__be32 *p, *start;
+
+	da = list_first_entry(&ds->ds_addrs, struct nfs4_pnfs_ds_addr, da_node);
+	dprintk("%s: DS %s: encoding address %s\n",
+		__func__, ds->ds_remotestr, da->da_remotestr);
+	/* layoutupdate length */
+	start = xdr_reserve_space(xdr, 4);
+	/* netaddr4 */
+	ff_layout_encode_netaddr(xdr, da);
+	/* nfs_fh4 */
+	p = xdr_reserve_space(xdr, 4 + fh->size);
+	xdr_encode_opaque(p, fh->data, fh->size);
+	/* ff_io_latency4 read */
+	spin_lock(&mirror->lock);
+	ff_layout_encode_io_latency(xdr, &mirror->read_stat.io_stat);
+	/* ff_io_latency4 write */
+	ff_layout_encode_io_latency(xdr, &mirror->write_stat.io_stat);
+	spin_unlock(&mirror->lock);
+	/* nfstime4 */
+	ff_layout_encode_nfstime(xdr, ktime_sub(ktime_get(), mirror->start_time));
+	/* bool */
+	p = xdr_reserve_space(xdr, 4);
+	*p = cpu_to_be32(false);
+
+	*start = cpu_to_be32((xdr->p - start - 1) * 4);
+}
+
+static bool
+ff_layout_mirror_prepare_stats(struct nfs42_layoutstat_args *args,
+			       struct pnfs_layout_segment *pls,
+			       int *dev_count, int dev_limit)
+{
+	struct nfs4_ff_layout_mirror *mirror;
+	struct nfs4_deviceid_node *dev;
+	struct nfs42_layoutstat_devinfo *devinfo;
+	int i;
+
+	for (i = 0; i <= FF_LAYOUT_MIRROR_COUNT(pls); i++) {
+		if (*dev_count >= dev_limit)
+			break;
+		mirror = FF_LAYOUT_COMP(pls, i);
+		if (!mirror || !mirror->mirror_ds)
+			continue;
+		dev = FF_LAYOUT_DEVID_NODE(pls, i);
+		devinfo = &args->devinfo[*dev_count];
+		memcpy(&devinfo->dev_id, &dev->deviceid, NFS4_DEVICEID4_SIZE);
+		devinfo->offset = pls->pls_range.offset;
+		devinfo->length = pls->pls_range.length;
+		/* well, we don't really know if IO is continuous or not! */
+		devinfo->read_count = mirror->read_stat.io_stat.bytes_completed;
+		devinfo->read_bytes = mirror->read_stat.io_stat.bytes_completed;
+		devinfo->write_count = mirror->write_stat.io_stat.bytes_completed;
+		devinfo->write_bytes = mirror->write_stat.io_stat.bytes_completed;
+		devinfo->layout_type = LAYOUT_FLEX_FILES;
+		devinfo->layoutstats_encode = ff_layout_encode_layoutstats;
+		devinfo->layout_private = mirror;
+		/* lseg refcount put in cleanup_layoutstats */
+		pnfs_get_lseg(pls);
+
+		++(*dev_count);
+	}
+
+	return *dev_count < dev_limit;
+}
+
+static int
+ff_layout_prepare_layoutstats(struct nfs42_layoutstat_args *args)
+{
+	struct pnfs_layout_segment *pls;
+	int dev_count = 0;
+
+	spin_lock(&args->inode->i_lock);
+	list_for_each_entry(pls, &NFS_I(args->inode)->layout->plh_segs, pls_list) {
+		dev_count += FF_LAYOUT_MIRROR_COUNT(pls);
+	}
+	spin_unlock(&args->inode->i_lock);
+	/* For now, send at most PNFS_LAYOUTSTATS_MAXDEV statistics */
+	if (dev_count > PNFS_LAYOUTSTATS_MAXDEV) {
+		dprintk("%s: truncating devinfo to limit (%d:%d)\n",
+			__func__, dev_count, PNFS_LAYOUTSTATS_MAXDEV);
+		dev_count = PNFS_LAYOUTSTATS_MAXDEV;
+	}
+	args->devinfo = kmalloc(dev_count * sizeof(*args->devinfo), GFP_KERNEL);
+	if (!args->devinfo)
+		return -ENOMEM;
+
+	dev_count = 0;
+	spin_lock(&args->inode->i_lock);
+	list_for_each_entry(pls, &NFS_I(args->inode)->layout->plh_segs, pls_list) {
+		if (!ff_layout_mirror_prepare_stats(args, pls, &dev_count,
+						    PNFS_LAYOUTSTATS_MAXDEV)) {
+			break;
+		}
+	}
+	spin_unlock(&args->inode->i_lock);
+	args->num_dev = dev_count;
+
+	return 0;
+}
+
+static void
+ff_layout_cleanup_layoutstats(struct nfs42_layoutstat_data *data)
+{
+	struct nfs4_ff_layout_mirror *mirror;
+	int i;
+
+	for (i = 0; i < data->args.num_dev; i++) {
+		mirror = data->args.devinfo[i].layout_private;
+		data->args.devinfo[i].layout_private = NULL;
+		pnfs_put_lseg(mirror->lseg);
+	}
+}
+
 static struct pnfs_layoutdriver_type flexfilelayout_type = {
 	.id			= LAYOUT_FLEX_FILES,
 	.name			= "LAYOUT_FLEX_FILES",
@@ -1507,6 +1936,8 @@ static struct pnfs_layoutdriver_type flexfilelayout_type = {
 	.alloc_deviceid_node    = ff_layout_alloc_deviceid_node,
 	.encode_layoutreturn    = ff_layout_encode_layoutreturn,
 	.sync			= pnfs_nfs_generic_sync,
+	.prepare_layoutstats	= ff_layout_prepare_layoutstats,
+	.cleanup_layoutstats	= ff_layout_cleanup_layoutstats,
 };
 
 static int __init nfs4flexfilelayout_init(void)

fs/nfs/flexfilelayout/flexfilelayout.h

@@ -15,6 +15,9 @@
  * due to network error etc. */
 #define NFS4_FLEXFILE_LAYOUT_MAX_MIRROR_CNT 4096
 
+/* LAYOUTSTATS report interval in ms */
+#define FF_LAYOUTSTATS_REPORT_INTERVAL (60000L)
+
 struct nfs4_ff_ds_version {
 	u32				version;
 	u32				minor_version;
@@ -41,19 +44,42 @@ struct nfs4_ff_layout_ds_err {
 	struct nfs4_deviceid		deviceid;
 };
 
+struct nfs4_ff_io_stat {
+	__u64				ops_requested;
+	__u64				bytes_requested;
+	__u64				ops_completed;
+	__u64				bytes_completed;
+	__u64				bytes_not_delivered;
+	ktime_t				total_busy_time;
+	ktime_t				aggregate_completion_time;
+};
+
+struct nfs4_ff_busy_timer {
+	ktime_t start_time;
+	atomic_t n_ops;
+};
+
+struct nfs4_ff_layoutstat {
+	struct nfs4_ff_io_stat io_stat;
+	struct nfs4_ff_busy_timer busy_timer;
+};
+
 struct nfs4_ff_layout_mirror {
+	struct pnfs_layout_segment	*lseg; /* back pointer */
 	u32				ds_count;
 	u32				efficiency;
 	struct nfs4_ff_layout_ds	*mirror_ds;
 	u32				fh_versions_cnt;
 	struct nfs_fh			*fh_versions;
 	nfs4_stateid			stateid;
-	struct nfs4_string		user_name;
-	struct nfs4_string		group_name;
 	u32				uid;
 	u32				gid;
 	struct rpc_cred			*cred;
 	spinlock_t			lock;
+	struct nfs4_ff_layoutstat	read_stat;
+	struct nfs4_ff_layoutstat	write_stat;
+	ktime_t				start_time;
+	ktime_t				last_report_time;
 };
 
 struct nfs4_ff_layout_segment {

fs/nfs/nfs42.h

@@ -5,11 +5,18 @@
 #ifndef __LINUX_FS_NFS_NFS4_2_H
 #define __LINUX_FS_NFS_NFS4_2_H
 
+/*
+ * FIXME:  four LAYOUTSTATS calls per compound at most! Do we need to support
+ * more? Need to consider not to pre-alloc too much for a compound.
+ */
+#define PNFS_LAYOUTSTATS_MAXDEV (4)
+
 /* nfs4.2proc.c */
 int nfs42_proc_allocate(struct file *, loff_t, loff_t);
 int nfs42_proc_deallocate(struct file *, loff_t, loff_t);
 loff_t nfs42_proc_llseek(struct file *, loff_t, int);
-
+int nfs42_proc_layoutstats_generic(struct nfs_server *,
+				   struct nfs42_layoutstat_data *);
 /* nfs4.2xdr.h */
 extern struct rpc_procinfo nfs4_2_procedures[];
 

fs/nfs/nfs42proc.c

@@ -10,6 +10,11 @@
 #include <linux/nfs_fs.h>
 #include "nfs4_fs.h"
 #include "nfs42.h"
+#include "iostat.h"
+#include "pnfs.h"
+#include "internal.h"
+
+#define NFSDBG_FACILITY NFSDBG_PNFS
 
 static int nfs42_set_rw_stateid(nfs4_stateid *dst, struct file *file,
 				fmode_t fmode)
@@ -165,3 +170,79 @@ loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
 
 	return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
 }
+
+static void
+nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
+{
+	struct nfs42_layoutstat_data *data = calldata;
+	struct nfs_server *server = NFS_SERVER(data->args.inode);
+
+	nfs41_setup_sequence(nfs4_get_session(server), &data->args.seq_args,
+			     &data->res.seq_res, task);
+}
+
+static void
+nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
+{
+	struct nfs42_layoutstat_data *data = calldata;
+
+	if (!nfs4_sequence_done(task, &data->res.seq_res))
+		return;
+
+	/* well, we don't care about errors at all! */
+	if (task->tk_status)
+		dprintk("%s server returns %d\n", __func__, task->tk_status);
+}
+
+static void
+nfs42_layoutstat_release(void *calldata)
+{
+	struct nfs42_layoutstat_data *data = calldata;
+	struct nfs_server *nfss = NFS_SERVER(data->args.inode);
+
+	if (nfss->pnfs_curr_ld->cleanup_layoutstats)
+		nfss->pnfs_curr_ld->cleanup_layoutstats(data);
+
+	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
+	smp_mb__before_atomic();
+	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
+	smp_mb__after_atomic();
+	nfs_iput_and_deactive(data->inode);
+	kfree(data->args.devinfo);
+	kfree(data);
+}
+
+static const struct rpc_call_ops nfs42_layoutstat_ops = {
+	.rpc_call_prepare = nfs42_layoutstat_prepare,
+	.rpc_call_done = nfs42_layoutstat_done,
+	.rpc_release = nfs42_layoutstat_release,
+};
+
+int nfs42_proc_layoutstats_generic(struct nfs_server *server,
+				   struct nfs42_layoutstat_data *data)
+{
+	struct rpc_message msg = {
+		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
+		.rpc_argp = &data->args,
+		.rpc_resp = &data->res,
+	};
+	struct rpc_task_setup task_setup = {
+		.rpc_client = server->client,
+		.rpc_message = &msg,
+		.callback_ops = &nfs42_layoutstat_ops,
+		.callback_data = data,
+		.flags = RPC_TASK_ASYNC,
+	};
+	struct rpc_task *task;
+
+	data->inode = nfs_igrab_and_active(data->args.inode);
+	if (!data->inode) {
+		nfs42_layoutstat_release(data);
+		return -EAGAIN;
+	}
+	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0);
+	task = rpc_run_task(&task_setup);
+	if (IS_ERR(task))
+		return PTR_ERR(task);
+	return 0;
+}

fs/nfs/nfs42xdr.c

@@ -4,6 +4,8 @@
 #ifndef __LINUX_FS_NFS_NFS4_2XDR_H
 #define __LINUX_FS_NFS_NFS4_2XDR_H
 
+#include "nfs42.h"
+
 #define encode_fallocate_maxsz		(encode_stateid_maxsz + \
 					 2 /* offset */ + \
 					 2 /* length */)
@@ -22,6 +24,16 @@
 					 1 /* whence */ + \
 					 2 /* offset */ + \
 					 2 /* length */)
+#define encode_io_info_maxsz		4
+#define encode_layoutstats_maxsz	(op_decode_hdr_maxsz + \
+					2 /* offset */ + \
+					2 /* length */ + \
+					encode_stateid_maxsz + \
+					encode_io_info_maxsz + \
+					encode_io_info_maxsz + \
+					1 /* opaque devaddr4 length */ + \
+					XDR_QUADLEN(PNFS_LAYOUTSTATS_MAXSIZE))
+#define decode_layoutstats_maxsz	(op_decode_hdr_maxsz)
 
 #define NFS4_enc_allocate_sz		(compound_encode_hdr_maxsz + \
 					 encode_putfh_maxsz + \
@@ -45,6 +57,14 @@
 #define NFS4_dec_seek_sz		(compound_decode_hdr_maxsz + \
 					 decode_putfh_maxsz + \
 					 decode_seek_maxsz)
+#define NFS4_enc_layoutstats_sz		(compound_encode_hdr_maxsz + \
+					 encode_sequence_maxsz + \
+					 encode_putfh_maxsz + \
+					 PNFS_LAYOUTSTATS_MAXDEV * encode_layoutstats_maxsz)
+#define NFS4_dec_layoutstats_sz		(compound_decode_hdr_maxsz + \
+					 decode_sequence_maxsz + \
+					 decode_putfh_maxsz + \
+					 PNFS_LAYOUTSTATS_MAXDEV * decode_layoutstats_maxsz)
 
 
 static void encode_fallocate(struct xdr_stream *xdr,
@@ -81,6 +101,33 @@ static void encode_seek(struct xdr_stream *xdr,
 	encode_uint32(xdr, args->sa_what);
 }
 
+static void encode_layoutstats(struct xdr_stream *xdr,
+			       struct nfs42_layoutstat_args *args,
+			       struct nfs42_layoutstat_devinfo *devinfo,
+			       struct compound_hdr *hdr)
+{
+	__be32 *p;
+
+	encode_op_hdr(xdr, OP_LAYOUTSTATS, decode_layoutstats_maxsz, hdr);
+	p = reserve_space(xdr, 8 + 8);
+	p = xdr_encode_hyper(p, devinfo->offset);
+	p = xdr_encode_hyper(p, devinfo->length);
+	encode_nfs4_stateid(xdr, &args->stateid);
+	p = reserve_space(xdr, 4*8 + NFS4_DEVICEID4_SIZE + 4);
+	p = xdr_encode_hyper(p, devinfo->read_count);
+	p = xdr_encode_hyper(p, devinfo->read_bytes);
+	p = xdr_encode_hyper(p, devinfo->write_count);
+	p = xdr_encode_hyper(p, devinfo->write_bytes);
+	p = xdr_encode_opaque_fixed(p, devinfo->dev_id.data,
+			NFS4_DEVICEID4_SIZE);
+	/* Encode layoutupdate4 */
+	*p++ = cpu_to_be32(devinfo->layout_type);
+	if (devinfo->layoutstats_encode != NULL)
+		devinfo->layoutstats_encode(xdr, args, devinfo);
+	else
+		encode_uint32(xdr, 0);
+}
+
 /*
  * Encode ALLOCATE request
  */
@@ -137,6 +184,28 @@ static void nfs4_xdr_enc_seek(struct rpc_rqst *req,
 	encode_nops(&hdr);
 }
 
+/*
+ * Encode LAYOUTSTATS request
+ */
+static void nfs4_xdr_enc_layoutstats(struct rpc_rqst *req,
+				     struct xdr_stream *xdr,
+				     struct nfs42_layoutstat_args *args)
+{
+	int i;
+
+	struct compound_hdr hdr = {
+		.minorversion = nfs4_xdr_minorversion(&args->seq_args),
+	};
+
+	encode_compound_hdr(xdr, req, &hdr);
+	encode_sequence(xdr, &args->seq_args, &hdr);
+	encode_putfh(xdr, args->fh, &hdr);
+	WARN_ON(args->num_dev > PNFS_LAYOUTSTATS_MAXDEV);
+	for (i = 0; i < args->num_dev; i++)
+		encode_layoutstats(xdr, args, &args->devinfo[i], &hdr);
+	encode_nops(&hdr);
+}
+
 static int decode_allocate(struct xdr_stream *xdr, struct nfs42_falloc_res *res)
 {
 	return decode_op_hdr(xdr, OP_ALLOCATE);
@@ -169,6 +238,28 @@ static int decode_seek(struct xdr_stream *xdr, struct nfs42_seek_res *res)
 	return -EIO;
 }
 
+static int decode_layoutstats(struct xdr_stream *xdr,
+			      struct nfs42_layoutstat_res *res)
+{
+	int status;
+	__be32 *p;
+
+	status = decode_op_hdr(xdr, OP_LAYOUTSTATS);
+	if (status)
+		return status;
+
+	p = xdr_inline_decode(xdr, 4);
+	if (unlikely(!p))
+		goto out_overflow;
+
+	res->rpc_status = be32_to_cpup(p++);
+	return 0;
+
+out_overflow:
+	print_overflow_msg(__func__, xdr);
+	return -EIO;
+}
+
 /*
  * Decode ALLOCATE request
  */
@@ -246,4 +337,35 @@ static int nfs4_xdr_dec_seek(struct rpc_rqst *rqstp,
 out:
 	return status;
 }
+
+/*
+ * Decode LAYOUTSTATS request
+ */
+static int nfs4_xdr_dec_layoutstats(struct rpc_rqst *rqstp,
+				    struct xdr_stream *xdr,
+				    struct nfs42_layoutstat_res *res)
+{
+	struct compound_hdr hdr;
+	int status, i;
+
+	status = decode_compound_hdr(xdr, &hdr);
+	if (status)
+		goto out;
+	status = decode_sequence(xdr, &res->seq_res, rqstp);
+	if (status)
+		goto out;
+	status = decode_putfh(xdr);
+	if (status)
+		goto out;
+	WARN_ON(res->num_dev > PNFS_LAYOUTSTATS_MAXDEV);
+	for (i = 0; i < res->num_dev; i++) {
+		status = decode_layoutstats(xdr, res);
+		if (status)
+			goto out;
+	}
+out:
+	res->rpc_status = status;
+	return status;
+}
+
 #endif /* __LINUX_FS_NFS_NFS4_2XDR_H */

fs/nfs/nfs4_fs.h

@@ -233,6 +233,7 @@ extern int nfs4_handle_exception(struct nfs_server *, int, struct nfs4_exception
 extern int nfs4_call_sync(struct rpc_clnt *, struct nfs_server *,
 			  struct rpc_message *, struct nfs4_sequence_args *,
 			  struct nfs4_sequence_res *, int);
+extern void nfs4_init_sequence(struct nfs4_sequence_args *, struct nfs4_sequence_res *, int);
 extern int nfs4_proc_setclientid(struct nfs_client *, u32, unsigned short, struct rpc_cred *, struct nfs4_setclientid_res *);
 extern int nfs4_proc_setclientid_confirm(struct nfs_client *, struct nfs4_setclientid_res *arg, struct rpc_cred *);
 extern int nfs4_proc_get_rootfh(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *, bool);

fs/nfs/nfs4proc.c

@@ -476,7 +476,7 @@ struct nfs4_call_sync_data {
 	struct nfs4_sequence_res *seq_res;
 };
 
-static void nfs4_init_sequence(struct nfs4_sequence_args *args,
+void nfs4_init_sequence(struct nfs4_sequence_args *args,
 			struct nfs4_sequence_res *res, int cache_reply)
 {
 	args->sa_slot = NULL;

fs/nfs/nfs4xdr.c

@@ -7431,6 +7431,7 @@ struct rpc_procinfo	nfs4_procedures[] = {
 	PROC(SEEK,		enc_seek,		dec_seek),
 	PROC(ALLOCATE,		enc_allocate,		dec_allocate),
 	PROC(DEALLOCATE,	enc_deallocate,		dec_deallocate),
+	PROC(LAYOUTSTATS,	enc_layoutstats,	dec_layoutstats),
 #endif /* CONFIG_NFS_V4_2 */
 };
 

fs/nfs/pnfs.c

@@ -35,6 +35,7 @@
 #include "iostat.h"
 #include "nfs4trace.h"
 #include "delegation.h"
+#include "nfs42.h"
 
 #define NFSDBG_FACILITY		NFSDBG_PNFS
 #define PNFS_LAYOUTGET_RETRY_TIMEOUT (120*HZ)
@@ -2250,3 +2251,60 @@ struct nfs4_threshold *pnfs_mdsthreshold_alloc(void)
 	}
 	return thp;
 }
+
+#if IS_ENABLED(CONFIG_NFS_V4_2)
+int
+pnfs_report_layoutstat(struct inode *inode)
+{
+	struct pnfs_layoutdriver_type *ld = NFS_SERVER(inode)->pnfs_curr_ld;
+	struct nfs_server *server = NFS_SERVER(inode);
+	struct nfs_inode *nfsi = NFS_I(inode);
+	struct nfs42_layoutstat_data *data;
+	struct pnfs_layout_hdr *hdr;
+	int status = 0;
+
+	if (!pnfs_enabled_sb(server) || !ld->prepare_layoutstats)
+		goto out;
+
+	if (test_and_set_bit(NFS_INO_LAYOUTSTATS, &nfsi->flags))
+		goto out;
+
+	spin_lock(&inode->i_lock);
+	if (!NFS_I(inode)->layout) {
+		spin_unlock(&inode->i_lock);
+		goto out;
+	}
+	hdr = NFS_I(inode)->layout;
+	pnfs_get_layout_hdr(hdr);
+	spin_unlock(&inode->i_lock);
+
+	data = kzalloc(sizeof(*data), GFP_KERNEL);
+	if (!data) {
+		status = -ENOMEM;
+		goto out_put;
+	}
+
+	data->args.fh = NFS_FH(inode);
+	data->args.inode = inode;
+	nfs4_stateid_copy(&data->args.stateid, &hdr->plh_stateid);
+	status = ld->prepare_layoutstats(&data->args);
+	if (status)
+		goto out_free;
+
+	status = nfs42_proc_layoutstats_generic(NFS_SERVER(inode), data);
+
+out:
+	dprintk("%s returns %d\n", __func__, status);
+	return status;
+
+out_free:
+	kfree(data);
+out_put:
+	pnfs_put_layout_hdr(hdr);
+	smp_mb__before_atomic();
+	clear_bit(NFS_INO_LAYOUTSTATS, &nfsi->flags);
+	smp_mb__after_atomic();
+	goto out;
+}
+EXPORT_SYMBOL_GPL(pnfs_report_layoutstat);
+#endif

fs/nfs/pnfs.h

@@ -178,6 +178,8 @@ struct pnfs_layoutdriver_type {
 	void (*encode_layoutcommit) (struct pnfs_layout_hdr *lo,
 				     struct xdr_stream *xdr,
 				     const struct nfs4_layoutcommit_args *args);
+	int (*prepare_layoutstats) (struct nfs42_layoutstat_args *args);
+	void (*cleanup_layoutstats) (struct nfs42_layoutstat_data *data);
 };
 
 struct pnfs_layout_hdr {
@@ -290,7 +292,6 @@ int pnfs_write_done_resend_to_mds(struct nfs_pgio_header *);
 struct nfs4_threshold *pnfs_mdsthreshold_alloc(void);
 void pnfs_error_mark_layout_for_return(struct inode *inode,
 				       struct pnfs_layout_segment *lseg);
-
 /* nfs4_deviceid_flags */
 enum {
 	NFS_DEVICEID_INVALID = 0,       /* set when MDS clientid recalled */
@@ -689,4 +690,14 @@ static inline void nfs4_pnfs_v3_ds_connect_unload(void)
 
 #endif /* CONFIG_NFS_V4_1 */
 
+#if IS_ENABLED(CONFIG_NFS_V4_2)
+int pnfs_report_layoutstat(struct inode *inode);
+#else
+static inline int
+pnfs_report_layoutstat(struct inode *inode)
+{
+	return 0;
+}
+#endif
+
 #endif /* FS_NFS_PNFS_H */

include/linux/nfs4.h

@@ -500,6 +500,7 @@ enum {
 	NFSPROC4_CLNT_SEEK,
 	NFSPROC4_CLNT_ALLOCATE,
 	NFSPROC4_CLNT_DEALLOCATE,
+	NFSPROC4_CLNT_LAYOUTSTATS,
 };
 
 /* nfs41 types */

include/linux/nfs_fs.h

@@ -219,6 +219,7 @@ struct nfs_inode {
 #define NFS_INO_COMMIT		(7)		/* inode is committing unstable writes */
 #define NFS_INO_LAYOUTCOMMIT	(9)		/* layoutcommit required */
 #define NFS_INO_LAYOUTCOMMITTING (10)		/* layoutcommit inflight */
+#define NFS_INO_LAYOUTSTATS	(11)		/* layoutstats inflight */
 
 static inline struct nfs_inode *NFS_I(const struct inode *inode)
 {

include/linux/nfs_xdr.h

@@ -316,6 +316,49 @@ struct nfs4_layoutreturn {
 	int rpc_status;
 };
 
+#define PNFS_LAYOUTSTATS_MAXSIZE 256
+
+struct nfs42_layoutstat_args;
+struct nfs42_layoutstat_devinfo;
+typedef	void (*layoutstats_encode_t)(struct xdr_stream *,
+		struct nfs42_layoutstat_args *,
+		struct nfs42_layoutstat_devinfo *);
+
+/* Per file per deviceid layoutstats */
+struct nfs42_layoutstat_devinfo {
+	struct nfs4_deviceid dev_id;
+	__u64 offset;
+	__u64 length;
+	__u64 read_count;
+	__u64 read_bytes;
+	__u64 write_count;
+	__u64 write_bytes;
+	__u32 layout_type;
+	layoutstats_encode_t layoutstats_encode;
+	void *layout_private;
+};
+
+struct nfs42_layoutstat_args {
+	struct nfs4_sequence_args seq_args;
+	struct nfs_fh *fh;
+	struct inode *inode;
+	nfs4_stateid stateid;
+	int num_dev;
+	struct nfs42_layoutstat_devinfo *devinfo;
+};
+
+struct nfs42_layoutstat_res {
+	struct nfs4_sequence_res seq_res;
+	int num_dev;
+	int rpc_status;
+};
+
+struct nfs42_layoutstat_data {
+	struct inode *inode;
+	struct nfs42_layoutstat_args args;
+	struct nfs42_layoutstat_res res;
+};
+
 struct stateowner_id {
 	__u64	create_time;
 	__u32	uniquifier;