Merge branch 'nfs-for-3.1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

* 'nfs-for-3.1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (28 commits)
pnfsblock: write_pagelist handle zero invalid extents
pnfsblock: note written INVAL areas for layoutcommit
pnfsblock: bl_write_pagelist
pnfsblock: bl_read_pagelist
pnfsblock: cleanup_layoutcommit
pnfsblock: encode_layoutcommit
pnfsblock: merge rw extents
pnfsblock: add extent manipulation functions
pnfsblock: bl_find_get_extent
pnfsblock: xdr decode pnfs_block_layout4
pnfsblock: call and parse getdevicelist
pnfsblock: merge extents
pnfsblock: lseg alloc and free
pnfsblock: remove device operations
pnfsblock: add device operations
pnfsblock: basic extent code
pnfsblock: use pageio_ops api
pnfsblock: add blocklayout Kconfig option, Makefile, and stubs
pnfs: cleanup_layoutcommit
pnfs: ask for layout_blksize and save it in nfs_server
...

fs/nfs/Kconfig

···
 	depends on NFS_FS && NFS_V4 && EXPERIMENTAL
 	select SUNRPC_BACKCHANNEL
 	select PNFS_FILE_LAYOUT
+	select PNFS_BLOCK
+	select MD
+	select BLK_DEV_DM
 	help
 	  This option enables support for minor version 1 of the NFSv4 protocol
-	  (RFC 5661) in the kernel's NFS client.
+	  (RFC 5661 and RFC 5663) in the kernel's NFS client.
 
 	  If unsure, say N.
 
 config PNFS_FILE_LAYOUT
+	tristate
+
+config PNFS_BLOCK
 	tristate
 
 config PNFS_OBJLAYOUT

fs/nfs/Makefile

···
 nfs_layout_nfsv41_files-y := nfs4filelayout.o nfs4filelayoutdev.o
 
 obj-$(CONFIG_PNFS_OBJLAYOUT) += objlayout/
+obj-$(CONFIG_PNFS_BLOCK) += blocklayout/

fs/nfs/blocklayout/Makefile

···
+#
+# Makefile for the pNFS block layout driver kernel module
+#
+obj-$(CONFIG_PNFS_BLOCK) += blocklayoutdriver.o
+blocklayoutdriver-objs := blocklayout.o extents.o blocklayoutdev.o blocklayoutdm.o

fs/nfs/blocklayout/blocklayout.c

···
+/*
+ *  linux/fs/nfs/blocklayout/blocklayout.c
+ *
+ *  Module for the NFSv4.1 pNFS block layout driver.
+ *
+ *  Copyright (c) 2006 The Regents of the University of Michigan.
+ *  All rights reserved.
+ *
+ *  Andy Adamson <[e-mail]>
+ *  Fred Isaman <[e-mail]>
+ *
+ * permission is granted to use, copy, create derivative works and
+ * redistribute this software and such derivative works for any purpose,
+ * so long as the name of the university of michigan is not used in
+ * any advertising or publicity pertaining to the use or distribution
+ * of this software without specific, written prior authorization.  if
+ * the above copyright notice or any other identification of the
+ * university of michigan is included in any copy of any portion of
+ * this software, then the disclaimer below must also be included.
+ *
+ * this software is provided as is, without representation from the
+ * university of michigan as to its fitness for any purpose, and without
+ * warranty by the university of michigan of any kind, either express
+ * or implied, including without limitation the implied warranties of
+ * merchantability and fitness for a particular purpose.  the regents
+ * of the university of michigan shall not be liable for any damages,
+ * including special, indirect, incidental, or consequential damages,
+ * with respect to any claim arising out or in connection with the use
+ * of the software, even if it has been or is hereafter advised of the
+ * possibility of such damages.
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/mount.h>
+#include <linux/namei.h>
+#include <linux/bio.h>		/* struct bio */
+#include <linux/buffer_head.h>	/* various write calls */
+
+#include "blocklayout.h"
+
+#define NFSDBG_FACILITY	NFSDBG_PNFS_LD
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Andy Adamson <[e-mail]>");
+MODULE_DESCRIPTION("The NFSv4.1 pNFS Block layout driver");
+
+struct dentry *bl_device_pipe;
+wait_queue_head_t bl_wq;
+
+static void print_page(struct page *page)
+{
+	dprintk("PRINTPAGE page %p\n", page);
+	dprintk("	PagePrivate %d\n", PagePrivate(page));
+	dprintk("	PageUptodate %d\n", PageUptodate(page));
+	dprintk("	PageError %d\n", PageError(page));
+	dprintk("	PageDirty %d\n", PageDirty(page));
+	dprintk("	PageReferenced %d\n", PageReferenced(page));
+	dprintk("	PageLocked %d\n", PageLocked(page));
+	dprintk("	PageWriteback %d\n", PageWriteback(page));
+	dprintk("	PageMappedToDisk %d\n", PageMappedToDisk(page));
+	dprintk("\n");
+}
+
+/* Given the be associated with isect, determine if page data needs to be
+ * initialized.
+ */
+static int is_hole(struct pnfs_block_extent *be, sector_t isect)
+{
+	if (be->be_state == PNFS_BLOCK_NONE_DATA)
+		return 1;
+	else if (be->be_state != PNFS_BLOCK_INVALID_DATA)
+		return 0;
+	else
+		return !bl_is_sector_init(be->be_inval, isect);
+}
+
+/* Given the be associated with isect, determine if page data can be
+ * written to disk.
+ */
+static int is_writable(struct pnfs_block_extent *be, sector_t isect)
+{
+	return (be->be_state == PNFS_BLOCK_READWRITE_DATA ||
+		be->be_state == PNFS_BLOCK_INVALID_DATA);
+}
+
+/* The data we are handed might be spread across several bios.  We need
+ * to track when the last one is finished.
+ */
+struct parallel_io {
+	struct kref refcnt;
+	struct rpc_call_ops call_ops;
+	void (*pnfs_callback) (void *data);
+	void *data;
+};
+
+static inline struct parallel_io *alloc_parallel(void *data)
+{
+	struct parallel_io *rv;
+
+	rv  = kmalloc(sizeof(*rv), GFP_NOFS);
+	if (rv) {
+		rv->data = data;
+		kref_init(&rv->refcnt);
+	}
+	return rv;
+}
+
+static inline void get_parallel(struct parallel_io *p)
+{
+	kref_get(&p->refcnt);
+}
+
+static void destroy_parallel(struct kref *kref)
+{
+	struct parallel_io *p = container_of(kref, struct parallel_io, refcnt);
+
+	dprintk("%s enter\n", __func__);
+	p->pnfs_callback(p->data);
+	kfree(p);
+}
+
+static inline void put_parallel(struct parallel_io *p)
+{
+	kref_put(&p->refcnt, destroy_parallel);
+}
+
+static struct bio *
+bl_submit_bio(int rw, struct bio *bio)
+{
+	if (bio) {
+		get_parallel(bio->bi_private);
+		dprintk("%s submitting %s bio %u@%llu\n", __func__,
+			rw == READ ? "read" : "write",
+			bio->bi_size, (unsigned long long)bio->bi_sector);
+		submit_bio(rw, bio);
+	}
+	return NULL;
+}
+
+static struct bio *bl_alloc_init_bio(int npg, sector_t isect,
+				     struct pnfs_block_extent *be,
+				     void (*end_io)(struct bio *, int err),
+				     struct parallel_io *par)
+{
+	struct bio *bio;
+
+	bio = bio_alloc(GFP_NOIO, npg);
+	if (!bio)
+		return NULL;
+
+	bio->bi_sector = isect - be->be_f_offset + be->be_v_offset;
+	bio->bi_bdev = be->be_mdev;
+	bio->bi_end_io = end_io;
+	bio->bi_private = par;
+	return bio;
+}
+
+static struct bio *bl_add_page_to_bio(struct bio *bio, int npg, int rw,
+				      sector_t isect, struct page *page,
+				      struct pnfs_block_extent *be,
+				      void (*end_io)(struct bio *, int err),
+				      struct parallel_io *par)
+{
+retry:
+	if (!bio) {
+		bio = bl_alloc_init_bio(npg, isect, be, end_io, par);
+		if (!bio)
+			return ERR_PTR(-ENOMEM);
+	}
+	if (bio_add_page(bio, page, PAGE_CACHE_SIZE, 0) < PAGE_CACHE_SIZE) {
+		bio = bl_submit_bio(rw, bio);
+		goto retry;
+	}
+	return bio;
+}
+
+static void bl_set_lo_fail(struct pnfs_layout_segment *lseg)
+{
+	if (lseg->pls_range.iomode == IOMODE_RW) {
+		dprintk("%s Setting layout IOMODE_RW fail bit\n", __func__);
+		set_bit(lo_fail_bit(IOMODE_RW), &lseg->pls_layout->plh_flags);
+	} else {
+		dprintk("%s Setting layout IOMODE_READ fail bit\n", __func__);
+		set_bit(lo_fail_bit(IOMODE_READ), &lseg->pls_layout->plh_flags);
+	}
+}
+
+/* This is basically copied from mpage_end_io_read */
+static void bl_end_io_read(struct bio *bio, int err)
+{
+	struct parallel_io *par = bio->bi_private;
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+	struct nfs_read_data *rdata = (struct nfs_read_data *)par->data;
+
+	do {
+		struct page *page = bvec->bv_page;
+
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+		if (uptodate)
+			SetPageUptodate(page);
+	} while (bvec >= bio->bi_io_vec);
+	if (!uptodate) {
+		if (!rdata->pnfs_error)
+			rdata->pnfs_error = -EIO;
+		bl_set_lo_fail(rdata->lseg);
+	}
+	bio_put(bio);
+	put_parallel(par);
+}
+
+static void bl_read_cleanup(struct work_struct *work)
+{
+	struct rpc_task *task;
+	struct nfs_read_data *rdata;
+	dprintk("%s enter\n", __func__);
+	task = container_of(work, struct rpc_task, u.tk_work);
+	rdata = container_of(task, struct nfs_read_data, task);
+	pnfs_ld_read_done(rdata);
+}
+
+static void
+bl_end_par_io_read(void *data)
+{
+	struct nfs_read_data *rdata = data;
+
+	INIT_WORK(&rdata->task.u.tk_work, bl_read_cleanup);
+	schedule_work(&rdata->task.u.tk_work);
+}
+
+/* We don't want normal .rpc_call_done callback used, so we replace it
+ * with this stub.
+ */
+static void bl_rpc_do_nothing(struct rpc_task *task, void *calldata)
+{
+	return;
+}
+
+static enum pnfs_try_status
+bl_read_pagelist(struct nfs_read_data *rdata)
+{
+	int i, hole;
+	struct bio *bio = NULL;
+	struct pnfs_block_extent *be = NULL, *cow_read = NULL;
+	sector_t isect, extent_length = 0;
+	struct parallel_io *par;
+	loff_t f_offset = rdata->args.offset;
+	size_t count = rdata->args.count;
+	struct page **pages = rdata->args.pages;
+	int pg_index = rdata->args.pgbase >> PAGE_CACHE_SHIFT;
+
+	dprintk("%s enter nr_pages %u offset %lld count %Zd\n", __func__,
+	       rdata->npages, f_offset, count);
+
+	par = alloc_parallel(rdata);
+	if (!par)
+		goto use_mds;
+	par->call_ops = *rdata->mds_ops;
+	par->call_ops.rpc_call_done = bl_rpc_do_nothing;
+	par->pnfs_callback = bl_end_par_io_read;
+	/* At this point, we can no longer jump to use_mds */
+
+	isect = (sector_t) (f_offset >> SECTOR_SHIFT);
+	/* Code assumes extents are page-aligned */
+	for (i = pg_index; i < rdata->npages; i++) {
+		if (!extent_length) {
+			/* We've used up the previous extent */
+			bl_put_extent(be);
+			bl_put_extent(cow_read);
+			bio = bl_submit_bio(READ, bio);
+			/* Get the next one */
+			be = bl_find_get_extent(BLK_LSEG2EXT(rdata->lseg),
+					     isect, &cow_read);
+			if (!be) {
+				rdata->pnfs_error = -EIO;
+				goto out;
+			}
+			extent_length = be->be_length -
+				(isect - be->be_f_offset);
+			if (cow_read) {
+				sector_t cow_length = cow_read->be_length -
+					(isect - cow_read->be_f_offset);
+				extent_length = min(extent_length, cow_length);
+			}
+		}
+		hole = is_hole(be, isect);
+		if (hole && !cow_read) {
+			bio = bl_submit_bio(READ, bio);
+			/* Fill hole w/ zeroes w/o accessing device */
+			dprintk("%s Zeroing page for hole\n", __func__);
+			zero_user_segment(pages[i], 0, PAGE_CACHE_SIZE);
+			print_page(pages[i]);
+			SetPageUptodate(pages[i]);
+		} else {
+			struct pnfs_block_extent *be_read;
+
+			be_read = (hole && cow_read) ? cow_read : be;
+			bio = bl_add_page_to_bio(bio, rdata->npages - i, READ,
+						 isect, pages[i], be_read,
+						 bl_end_io_read, par);
+			if (IS_ERR(bio)) {
+				rdata->pnfs_error = PTR_ERR(bio);
+				goto out;
+			}
+		}
+		isect += PAGE_CACHE_SECTORS;
+		extent_length -= PAGE_CACHE_SECTORS;
+	}
+	if ((isect << SECTOR_SHIFT) >= rdata->inode->i_size) {
+		rdata->res.eof = 1;
+		rdata->res.count = rdata->inode->i_size - f_offset;
+	} else {
+		rdata->res.count = (isect << SECTOR_SHIFT) - f_offset;
+	}
+out:
+	bl_put_extent(be);
+	bl_put_extent(cow_read);
+	bl_submit_bio(READ, bio);
+	put_parallel(par);
+	return PNFS_ATTEMPTED;
+
+ use_mds:
+	dprintk("Giving up and using normal NFS\n");
+	return PNFS_NOT_ATTEMPTED;
+}
+
+static void mark_extents_written(struct pnfs_block_layout *bl,
+				 __u64 offset, __u32 count)
+{
+	sector_t isect, end;
+	struct pnfs_block_extent *be;
+
+	dprintk("%s(%llu, %u)\n", __func__, offset, count);
+	if (count == 0)
+		return;
+	isect = (offset & (long)(PAGE_CACHE_MASK)) >> SECTOR_SHIFT;
+	end = (offset + count + PAGE_CACHE_SIZE - 1) & (long)(PAGE_CACHE_MASK);
+	end >>= SECTOR_SHIFT;
+	while (isect < end) {
+		sector_t len;
+		be = bl_find_get_extent(bl, isect, NULL);
+		BUG_ON(!be); /* FIXME */
+		len = min(end, be->be_f_offset + be->be_length) - isect;
+		if (be->be_state == PNFS_BLOCK_INVALID_DATA)
+			bl_mark_for_commit(be, isect, len); /* What if fails? */
+		isect += len;
+		bl_put_extent(be);
+	}
+}
+
+static void bl_end_io_write_zero(struct bio *bio, int err)
+{
+	struct parallel_io *par = bio->bi_private;
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct bio_vec *bvec = bio->bi_io_vec + bio->bi_vcnt - 1;
+	struct nfs_write_data *wdata = (struct nfs_write_data *)par->data;
+
+	do {
+		struct page *page = bvec->bv_page;
+
+		if (--bvec >= bio->bi_io_vec)
+			prefetchw(&bvec->bv_page->flags);
+		/* This is the zeroing page we added */
+		end_page_writeback(page);
+		page_cache_release(page);
+	} while (bvec >= bio->bi_io_vec);
+	if (!uptodate) {
+		if (!wdata->pnfs_error)
+			wdata->pnfs_error = -EIO;
+		bl_set_lo_fail(wdata->lseg);
+	}
+	bio_put(bio);
+	put_parallel(par);
+}
+
+/* This is basically copied from mpage_end_io_read */
+static void bl_end_io_write(struct bio *bio, int err)
+{
+	struct parallel_io *par = bio->bi_private;
+	const int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct nfs_write_data *wdata = (struct nfs_write_data *)par->data;
+
+	if (!uptodate) {
+		if (!wdata->pnfs_error)
+			wdata->pnfs_error = -EIO;
+		bl_set_lo_fail(wdata->lseg);
+	}
+	bio_put(bio);
+	put_parallel(par);
+}
+
+/* Function scheduled for call during bl_end_par_io_write,
+ * it marks sectors as written and extends the commitlist.
+ */
+static void bl_write_cleanup(struct work_struct *work)
+{
+	struct rpc_task *task;
+	struct nfs_write_data *wdata;
+	dprintk("%s enter\n", __func__);
+	task = container_of(work, struct rpc_task, u.tk_work);
+	wdata = container_of(task, struct nfs_write_data, task);
+	if (!wdata->pnfs_error) {
+		/* Marks for LAYOUTCOMMIT */
+		mark_extents_written(BLK_LSEG2EXT(wdata->lseg),
+				     wdata->args.offset, wdata->args.count);
+	}
+	pnfs_ld_write_done(wdata);
+}
+
+/* Called when last of bios associated with a bl_write_pagelist call finishes */
+static void bl_end_par_io_write(void *data)
+{
+	struct nfs_write_data *wdata = data;
+
+	wdata->task.tk_status = 0;
+	wdata->verf.committed = NFS_FILE_SYNC;
+	INIT_WORK(&wdata->task.u.tk_work, bl_write_cleanup);
+	schedule_work(&wdata->task.u.tk_work);
+}
+
+/* FIXME STUB - mark intersection of layout and page as bad, so is not
+ * used again.
+ */
+static void mark_bad_read(void)
+{
+	return;
+}
+
+/*
+ * map_block:  map a requested I/0 block (isect) into an offset in the LVM
+ * block_device
+ */
+static void
+map_block(struct buffer_head *bh, sector_t isect, struct pnfs_block_extent *be)
+{
+	dprintk("%s enter be=%p\n", __func__, be);
+
+	set_buffer_mapped(bh);
+	bh->b_bdev = be->be_mdev;
+	bh->b_blocknr = (isect - be->be_f_offset + be->be_v_offset) >>
+	    (be->be_mdev->bd_inode->i_blkbits - SECTOR_SHIFT);
+
+	dprintk("%s isect %llu, bh->b_blocknr %ld, using bsize %Zd\n",
+		__func__, (unsigned long long)isect, (long)bh->b_blocknr,
+		bh->b_size);
+	return;
+}
+
+/* Given an unmapped page, zero it or read in page for COW, page is locked
+ * by caller.
+ */
+static int
+init_page_for_write(struct page *page, struct pnfs_block_extent *cow_read)
+{
+	struct buffer_head *bh = NULL;
+	int ret = 0;
+	sector_t isect;
+
+	dprintk("%s enter, %p\n", __func__, page);
+	BUG_ON(PageUptodate(page));
+	if (!cow_read) {
+		zero_user_segment(page, 0, PAGE_SIZE);
+		SetPageUptodate(page);
+		goto cleanup;
+	}
+
+	bh = alloc_page_buffers(page, PAGE_CACHE_SIZE, 0);
+	if (!bh) {
+		ret = -ENOMEM;
+		goto cleanup;
+	}
+
+	isect = (sector_t) page->index << PAGE_CACHE_SECTOR_SHIFT;
+	map_block(bh, isect, cow_read);
+	if (!bh_uptodate_or_lock(bh))
+		ret = bh_submit_read(bh);
+	if (ret)
+		goto cleanup;
+	SetPageUptodate(page);
+
+cleanup:
+	bl_put_extent(cow_read);
+	if (bh)
+		free_buffer_head(bh);
+	if (ret) {
+		/* Need to mark layout with bad read...should now
+		 * just use nfs4 for reads and writes.
+		 */
+		mark_bad_read();
+	}
+	return ret;
+}
+
+static enum pnfs_try_status
+bl_write_pagelist(struct nfs_write_data *wdata, int sync)
+{
+	int i, ret, npg_zero, pg_index, last = 0;
+	struct bio *bio = NULL;
+	struct pnfs_block_extent *be = NULL, *cow_read = NULL;
+	sector_t isect, last_isect = 0, extent_length = 0;
+	struct parallel_io *par;
+	loff_t offset = wdata->args.offset;
+	size_t count = wdata->args.count;
+	struct page **pages = wdata->args.pages;
+	struct page *page;
+	pgoff_t index;
+	u64 temp;
+	int npg_per_block =
+	    NFS_SERVER(wdata->inode)->pnfs_blksize >> PAGE_CACHE_SHIFT;
+
+	dprintk("%s enter, %Zu@%lld\n", __func__, count, offset);
+	/* At this point, wdata->pages is a (sequential) list of nfs_pages.
+	 * We want to write each, and if there is an error set pnfs_error
+	 * to have it redone using nfs.
+	 */
+	par = alloc_parallel(wdata);
+	if (!par)
+		return PNFS_NOT_ATTEMPTED;
+	par->call_ops = *wdata->mds_ops;
+	par->call_ops.rpc_call_done = bl_rpc_do_nothing;
+	par->pnfs_callback = bl_end_par_io_write;
+	/* At this point, have to be more careful with error handling */
+
+	isect = (sector_t) ((offset & (long)PAGE_CACHE_MASK) >> SECTOR_SHIFT);
+	be = bl_find_get_extent(BLK_LSEG2EXT(wdata->lseg), isect, &cow_read);
+	if (!be || !is_writable(be, isect)) {
+		dprintk("%s no matching extents!\n", __func__);
+		wdata->pnfs_error = -EINVAL;
+		goto out;
+	}
+
+	/* First page inside INVALID extent */
+	if (be->be_state == PNFS_BLOCK_INVALID_DATA) {
+		temp = offset >> PAGE_CACHE_SHIFT;
+		npg_zero = do_div(temp, npg_per_block);
+		isect = (sector_t) (((offset - npg_zero * PAGE_CACHE_SIZE) &
+				     (long)PAGE_CACHE_MASK) >> SECTOR_SHIFT);
+		extent_length = be->be_length - (isect - be->be_f_offset);
+
+fill_invalid_ext:
+		dprintk("%s need to zero %d pages\n", __func__, npg_zero);
+		for (;npg_zero > 0; npg_zero--) {
+			/* page ref released in bl_end_io_write_zero */
+			index = isect >> PAGE_CACHE_SECTOR_SHIFT;
+			dprintk("%s zero %dth page: index %lu isect %llu\n",
+				__func__, npg_zero, index,
+				(unsigned long long)isect);
+			page =
+			    find_or_create_page(wdata->inode->i_mapping, index,
+						GFP_NOFS);
+			if (!page) {
+				dprintk("%s oom\n", __func__);
+				wdata->pnfs_error = -ENOMEM;
+				goto out;
+			}
+
+			/* PageDirty: Other will write this out
+			 * PageWriteback: Other is writing this out
+			 * PageUptodate: It was read before
+			 * sector_initialized: already written out
+			 */
+			if (PageDirty(page) || PageWriteback(page) ||
+			    bl_is_sector_init(be->be_inval, isect)) {
+				print_page(page);
+				unlock_page(page);
+				page_cache_release(page);
+				goto next_page;
+			}
+			if (!PageUptodate(page)) {
+				/* New page, readin or zero it */
+				init_page_for_write(page, cow_read);
+			}
+			set_page_writeback(page);
+			unlock_page(page);
+
+			ret = bl_mark_sectors_init(be->be_inval, isect,
+						       PAGE_CACHE_SECTORS,
+						       NULL);
+			if (unlikely(ret)) {
+				dprintk("%s bl_mark_sectors_init fail %d\n",
+					__func__, ret);
+				end_page_writeback(page);
+				page_cache_release(page);
+				wdata->pnfs_error = ret;
+				goto out;
+			}
+			bio = bl_add_page_to_bio(bio, npg_zero, WRITE,
+						 isect, page, be,
+						 bl_end_io_write_zero, par);
+			if (IS_ERR(bio)) {
+				wdata->pnfs_error = PTR_ERR(bio);
+				goto out;
+			}
+			/* FIXME: This should be done in bi_end_io */
+			mark_extents_written(BLK_LSEG2EXT(wdata->lseg),
+					     page->index << PAGE_CACHE_SHIFT,
+					     PAGE_CACHE_SIZE);
+next_page:
+			isect += PAGE_CACHE_SECTORS;
+			extent_length -= PAGE_CACHE_SECTORS;
+		}
+		if (last)
+			goto write_done;
+	}
+	bio = bl_submit_bio(WRITE, bio);
+
+	/* Middle pages */
+	pg_index = wdata->args.pgbase >> PAGE_CACHE_SHIFT;
+	for (i = pg_index; i < wdata->npages; i++) {
+		if (!extent_length) {
+			/* We've used up the previous extent */
+			bl_put_extent(be);
+			bio = bl_submit_bio(WRITE, bio);
+			/* Get the next one */
+			be = bl_find_get_extent(BLK_LSEG2EXT(wdata->lseg),
+					     isect, NULL);
+			if (!be || !is_writable(be, isect)) {
+				wdata->pnfs_error = -EINVAL;
+				goto out;
+			}
+			extent_length = be->be_length -
+			    (isect - be->be_f_offset);
+		}
+		if (be->be_state == PNFS_BLOCK_INVALID_DATA) {
+			ret = bl_mark_sectors_init(be->be_inval, isect,
+						       PAGE_CACHE_SECTORS,
+						       NULL);
+			if (unlikely(ret)) {
+				dprintk("%s bl_mark_sectors_init fail %d\n",
+					__func__, ret);
+				wdata->pnfs_error = ret;
+				goto out;
+			}
+		}
+		bio = bl_add_page_to_bio(bio, wdata->npages - i, WRITE,
+					 isect, pages[i], be,
+					 bl_end_io_write, par);
+		if (IS_ERR(bio)) {
+			wdata->pnfs_error = PTR_ERR(bio);
+			goto out;
+		}
+		isect += PAGE_CACHE_SECTORS;
+		last_isect = isect;
+		extent_length -= PAGE_CACHE_SECTORS;
+	}
+
+	/* Last page inside INVALID extent */
+	if (be->be_state == PNFS_BLOCK_INVALID_DATA) {
+		bio = bl_submit_bio(WRITE, bio);
+		temp = last_isect >> PAGE_CACHE_SECTOR_SHIFT;
+		npg_zero = npg_per_block - do_div(temp, npg_per_block);
+		if (npg_zero < npg_per_block) {
+			last = 1;
+			goto fill_invalid_ext;
+		}
+	}
+
+write_done:
+	wdata->res.count = (last_isect << SECTOR_SHIFT) - (offset);
+	if (count < wdata->res.count) {
+		wdata->res.count = count;
+	}
+out:
+	bl_put_extent(be);
+	bl_submit_bio(WRITE, bio);
+	put_parallel(par);
+	return PNFS_ATTEMPTED;
+}
+
+/* FIXME - range ignored */
+static void
+release_extents(struct pnfs_block_layout *bl, struct pnfs_layout_range *range)
+{
+	int i;
+	struct pnfs_block_extent *be;
+
+	spin_lock(&bl->bl_ext_lock);
+	for (i = 0; i < EXTENT_LISTS; i++) {
+		while (!list_empty(&bl->bl_extents[i])) {
+			be = list_first_entry(&bl->bl_extents[i],
+					      struct pnfs_block_extent,
+					      be_node);
+			list_del(&be->be_node);
+			bl_put_extent(be);
+		}
+	}
+	spin_unlock(&bl->bl_ext_lock);
+}
+
+static void
+release_inval_marks(struct pnfs_inval_markings *marks)
+{
+	struct pnfs_inval_tracking *pos, *temp;
+
+	list_for_each_entry_safe(pos, temp, &marks->im_tree.mtt_stub, it_link) {
+		list_del(&pos->it_link);
+		kfree(pos);
+	}
+	return;
+}
+
+static void bl_free_layout_hdr(struct pnfs_layout_hdr *lo)
+{
+	struct pnfs_block_layout *bl = BLK_LO2EXT(lo);
+
+	dprintk("%s enter\n", __func__);
+	release_extents(bl, NULL);
+	release_inval_marks(&bl->bl_inval);
+	kfree(bl);
+}
+
+static struct pnfs_layout_hdr *bl_alloc_layout_hdr(struct inode *inode,
+						   gfp_t gfp_flags)
+{
+	struct pnfs_block_layout *bl;
+
+	dprintk("%s enter\n", __func__);
+	bl = kzalloc(sizeof(*bl), gfp_flags);
+	if (!bl)
+		return NULL;
+	spin_lock_init(&bl->bl_ext_lock);
+	INIT_LIST_HEAD(&bl->bl_extents[0]);
+	INIT_LIST_HEAD(&bl->bl_extents[1]);
+	INIT_LIST_HEAD(&bl->bl_commit);
+	INIT_LIST_HEAD(&bl->bl_committing);
+	bl->bl_count = 0;
+	bl->bl_blocksize = NFS_SERVER(inode)->pnfs_blksize >> SECTOR_SHIFT;
+	BL_INIT_INVAL_MARKS(&bl->bl_inval, bl->bl_blocksize);
+	return &bl->bl_layout;
+}
+
+static void bl_free_lseg(struct pnfs_layout_segment *lseg)
+{
+	dprintk("%s enter\n", __func__);
+	kfree(lseg);
+}
+
+/* We pretty much ignore lseg, and store all data layout wide, so we
+ * can correctly merge.
+ */
+static struct pnfs_layout_segment *bl_alloc_lseg(struct pnfs_layout_hdr *lo,
+						 struct nfs4_layoutget_res *lgr,
+						 gfp_t gfp_flags)
+{
+	struct pnfs_layout_segment *lseg;
+	int status;
+
+	dprintk("%s enter\n", __func__);
+	lseg = kzalloc(sizeof(*lseg), gfp_flags);
+	if (!lseg)
+		return ERR_PTR(-ENOMEM);
+	status = nfs4_blk_process_layoutget(lo, lgr, gfp_flags);
+	if (status) {
+		/* We don't want to call the full-blown bl_free_lseg,
+		 * since on error extents were not touched.
+		 */
+		kfree(lseg);
+		return ERR_PTR(status);
+	}
+	return lseg;
+}
+
+static void
+bl_encode_layoutcommit(struct pnfs_layout_hdr *lo, struct xdr_stream *xdr,
+		       const struct nfs4_layoutcommit_args *arg)
+{
+	dprintk("%s enter\n", __func__);
+	encode_pnfs_block_layoutupdate(BLK_LO2EXT(lo), xdr, arg);
+}
+
+static void
+bl_cleanup_layoutcommit(struct nfs4_layoutcommit_data *lcdata)
+{
+	struct pnfs_layout_hdr *lo = NFS_I(lcdata->args.inode)->layout;
+
+	dprintk("%s enter\n", __func__);
+	clean_pnfs_block_layoutupdate(BLK_LO2EXT(lo), &lcdata->args, lcdata->res.status);
+}
+
+static void free_blk_mountid(struct block_mount_id *mid)
+{
+	if (mid) {
+		struct pnfs_block_dev *dev;
+		spin_lock(&mid->bm_lock);
+		while (!list_empty(&mid->bm_devlist)) {
+			dev = list_first_entry(&mid->bm_devlist,
+					       struct pnfs_block_dev,
+					       bm_node);
+			list_del(&dev->bm_node);
+			bl_free_block_dev(dev);
+		}
+		spin_unlock(&mid->bm_lock);
+		kfree(mid);
+	}
+}
+
+/* This is mostly copied from the filelayout's get_device_info function.
+ * It seems much of this should be at the generic pnfs level.
+ */
+static struct pnfs_block_dev *
+nfs4_blk_get_deviceinfo(struct nfs_server *server, const struct nfs_fh *fh,
+			struct nfs4_deviceid *d_id)
+{
+	struct pnfs_device *dev;
+	struct pnfs_block_dev *rv = NULL;
+	u32 max_resp_sz;
+	int max_pages;
+	struct page **pages = NULL;
+	int i, rc;
+
+	/*
+	 * Use the session max response size as the basis for setting
+	 * GETDEVICEINFO's maxcount
+	 */
+	max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
+	max_pages = max_resp_sz >> PAGE_SHIFT;
+	dprintk("%s max_resp_sz %u max_pages %d\n",
+		__func__, max_resp_sz, max_pages);
+
+	dev = kmalloc(sizeof(*dev), GFP_NOFS);
+	if (!dev) {
+		dprintk("%s kmalloc failed\n", __func__);
+		return NULL;
+	}
+
+	pages = kzalloc(max_pages * sizeof(struct page *), GFP_NOFS);
+	if (pages == NULL) {
+		kfree(dev);
+		return NULL;
+	}
+	for (i = 0; i < max_pages; i++) {
+		pages[i] = alloc_page(GFP_NOFS);
+		if (!pages[i])
+			goto out_free;
+	}
+
+	memcpy(&dev->dev_id, d_id, sizeof(*d_id));
+	dev->layout_type = LAYOUT_BLOCK_VOLUME;
+	dev->pages = pages;
+	dev->pgbase = 0;
+	dev->pglen = PAGE_SIZE * max_pages;
+	dev->mincount = 0;
+
+	dprintk("%s: dev_id: %s\n", __func__, dev->dev_id.data);
+	rc = nfs4_proc_getdeviceinfo(server, dev);
+	dprintk("%s getdevice info returns %d\n", __func__, rc);
+	if (rc)
+		goto out_free;
+
+	rv = nfs4_blk_decode_device(server, dev);
+ out_free:
+	for (i = 0; i < max_pages; i++)
+		__free_page(pages[i]);
+	kfree(pages);
+	kfree(dev);
+	return rv;
+}
+
+static int
+bl_set_layoutdriver(struct nfs_server *server, const struct nfs_fh *fh)
+{
+	struct block_mount_id *b_mt_id = NULL;
+	struct pnfs_devicelist *dlist = NULL;
+	struct pnfs_block_dev *bdev;
+	LIST_HEAD(block_disklist);
+	int status = 0, i;
+
+	dprintk("%s enter\n", __func__);
+
+	if (server->pnfs_blksize == 0) {
+		dprintk("%s Server did not return blksize\n", __func__);
+		return -EINVAL;
+	}
+	b_mt_id = kzalloc(sizeof(struct block_mount_id), GFP_NOFS);
+	if (!b_mt_id) {
+		status = -ENOMEM;
+		goto out_error;
+	}
+	/* Initialize nfs4 block layout mount id */
+	spin_lock_init(&b_mt_id->bm_lock);
+	INIT_LIST_HEAD(&b_mt_id->bm_devlist);
+
+	dlist = kmalloc(sizeof(struct pnfs_devicelist), GFP_NOFS);
+	if (!dlist) {
+		status = -ENOMEM;
+		goto out_error;
+	}
+	dlist->eof = 0;
+	while (!dlist->eof) {
+		status = nfs4_proc_getdevicelist(server, fh, dlist);
+		if (status)
+			goto out_error;
+		dprintk("%s GETDEVICELIST numdevs=%i, eof=%i\n",
+			__func__, dlist->num_devs, dlist->eof);
+		for (i = 0; i < dlist->num_devs; i++) {
+			bdev = nfs4_blk_get_deviceinfo(server, fh,
+						       &dlist->dev_id[i]);
+			if (!bdev) {
+				status = -ENODEV;
+				goto out_error;
+			}
+			spin_lock(&b_mt_id->bm_lock);
+			list_add(&bdev->bm_node, &b_mt_id->bm_devlist);
+			spin_unlock(&b_mt_id->bm_lock);
+		}
+	}
+	dprintk("%s SUCCESS\n", __func__);
+	server->pnfs_ld_data = b_mt_id;
+
+ out_return:
+	kfree(dlist);
+	return status;
+
+ out_error:
+	free_blk_mountid(b_mt_id);
+	goto out_return;
+}
+
+static int
+bl_clear_layoutdriver(struct nfs_server *server)
+{
+	struct block_mount_id *b_mt_id = server->pnfs_ld_data;
+
+	dprintk("%s enter\n", __func__);
+	free_blk_mountid(b_mt_id);
+	dprintk("%s RETURNS\n", __func__);
+	return 0;
+}
+
+static const struct nfs_pageio_ops bl_pg_read_ops = {
+	.pg_init = pnfs_generic_pg_init_read,
+	.pg_test = pnfs_generic_pg_test,
+	.pg_doio = pnfs_generic_pg_readpages,
+};
+
+static const struct nfs_pageio_ops bl_pg_write_ops = {
+	.pg_init = pnfs_generic_pg_init_write,
+	.pg_test = pnfs_generic_pg_test,
+	.pg_doio = pnfs_generic_pg_writepages,
+};
+
+static struct pnfs_layoutdriver_type blocklayout_type = {
+	.id				= LAYOUT_BLOCK_VOLUME,
+	.name				= "LAYOUT_BLOCK_VOLUME",
+	.read_pagelist			= bl_read_pagelist,
+	.write_pagelist			= bl_write_pagelist,
+	.alloc_layout_hdr		= bl_alloc_layout_hdr,
+	.free_layout_hdr		= bl_free_layout_hdr,
+	.alloc_lseg			= bl_alloc_lseg,
+	.free_lseg			= bl_free_lseg,
+	.encode_layoutcommit		= bl_encode_layoutcommit,
+	.cleanup_layoutcommit		= bl_cleanup_layoutcommit,
+	.set_layoutdriver		= bl_set_layoutdriver,
+	.clear_layoutdriver		= bl_clear_layoutdriver,
+	.pg_read_ops			= &bl_pg_read_ops,
+	.pg_write_ops			= &bl_pg_write_ops,
+};
+
+static const struct rpc_pipe_ops bl_upcall_ops = {
+	.upcall		= bl_pipe_upcall,
+	.downcall	= bl_pipe_downcall,
+	.destroy_msg	= bl_pipe_destroy_msg,
+};
+
+static int __init nfs4blocklayout_init(void)
+{
+	struct vfsmount *mnt;
+	struct path path;
+	int ret;
+
+	dprintk("%s: NFSv4 Block Layout Driver Registering...\n", __func__);
+
+	ret = pnfs_register_layoutdriver(&blocklayout_type);
+	if (ret)
+		goto out;
+
+	init_waitqueue_head(&bl_wq);
+
+	mnt = rpc_get_mount();
+	if (IS_ERR(mnt)) {
+		ret = PTR_ERR(mnt);
+		goto out_remove;
+	}
+
+	ret = vfs_path_lookup(mnt->mnt_root,
+			      mnt,
+			      NFS_PIPE_DIRNAME, 0, &path);
+	if (ret)
+		goto out_remove;
+
+	bl_device_pipe = rpc_mkpipe(path.dentry, "blocklayout", NULL,
+				    &bl_upcall_ops, 0);
+	if (IS_ERR(bl_device_pipe)) {
+		ret = PTR_ERR(bl_device_pipe);
+		goto out_remove;
+	}
+out:
+	return ret;
+
+out_remove:
+	pnfs_unregister_layoutdriver(&blocklayout_type);
+	return ret;
+}
+
+static void __exit nfs4blocklayout_exit(void)
+{
+	dprintk("%s: NFSv4 Block Layout Driver Unregistering...\n",
+	       __func__);
+
+	pnfs_unregister_layoutdriver(&blocklayout_type);
+	rpc_unlink(bl_device_pipe);
+}
+
+MODULE_ALIAS("nfs-layouttype4-3");
+
+module_init(nfs4blocklayout_init);
+module_exit(nfs4blocklayout_exit);

fs/nfs/blocklayout/blocklayout.h

···
+/*
+ *  linux/fs/nfs/blocklayout/blocklayout.h
+ *
+ *  Module for the NFSv4.1 pNFS block layout driver.
+ *
+ *  Copyright (c) 2006 The Regents of the University of Michigan.
+ *  All rights reserved.
+ *
+ *  Andy Adamson <[e-mail]>
+ *  Fred Isaman <[e-mail]>
+ *
+ * permission is granted to use, copy, create derivative works and
+ * redistribute this software and such derivative works for any purpose,
+ * so long as the name of the university of michigan is not used in
+ * any advertising or publicity pertaining to the use or distribution
+ * of this software without specific, written prior authorization.  if
+ * the above copyright notice or any other identification of the
+ * university of michigan is included in any copy of any portion of
+ * this software, then the disclaimer below must also be included.
+ *
+ * this software is provided as is, without representation from the
+ * university of michigan as to its fitness for any purpose, and without
+ * warranty by the university of michigan of any kind, either express
+ * or implied, including without limitation the implied warranties of
+ * merchantability and fitness for a particular purpose.  the regents
+ * of the university of michigan shall not be liable for any damages,
+ * including special, indirect, incidental, or consequential damages,
+ * with respect to any claim arising out or in connection with the use
+ * of the software, even if it has been or is hereafter advised of the
+ * possibility of such damages.
+ */
+#ifndef FS_NFS_NFS4BLOCKLAYOUT_H
+#define FS_NFS_NFS4BLOCKLAYOUT_H
+
+#include <linux/device-mapper.h>
+#include <linux/nfs_fs.h>
+#include <linux/sunrpc/rpc_pipe_fs.h>
+
+#include "../pnfs.h"
+
+#define PAGE_CACHE_SECTORS (PAGE_CACHE_SIZE >> SECTOR_SHIFT)
+#define PAGE_CACHE_SECTOR_SHIFT (PAGE_CACHE_SHIFT - SECTOR_SHIFT)
+
+struct block_mount_id {
+	spinlock_t			bm_lock;    /* protects list */
+	struct list_head		bm_devlist; /* holds pnfs_block_dev */
+};
+
+struct pnfs_block_dev {
+	struct list_head		bm_node;
+	struct nfs4_deviceid		bm_mdevid;    /* associated devid */
+	struct block_device		*bm_mdev;     /* meta device itself */
+};
+
+enum exstate4 {
+	PNFS_BLOCK_READWRITE_DATA	= 0,
+	PNFS_BLOCK_READ_DATA		= 1,
+	PNFS_BLOCK_INVALID_DATA		= 2, /* mapped, but data is invalid */
+	PNFS_BLOCK_NONE_DATA		= 3  /* unmapped, it's a hole */
+};
+
+#define MY_MAX_TAGS (15) /* tag bitnums used must be less than this */
+
+struct my_tree {
+	sector_t		mtt_step_size;	/* Internal sector alignment */
+	struct list_head	mtt_stub; /* Should be a radix tree */
+};
+
+struct pnfs_inval_markings {
+	spinlock_t	im_lock;
+	struct my_tree	im_tree;	/* Sectors that need LAYOUTCOMMIT */
+	sector_t	im_block_size;	/* Server blocksize in sectors */
+};
+
+struct pnfs_inval_tracking {
+	struct list_head it_link;
+	int		 it_sector;
+	int		 it_tags;
+};
+
+/* sector_t fields are all in 512-byte sectors */
+struct pnfs_block_extent {
+	struct kref	be_refcnt;
+	struct list_head be_node;	/* link into lseg list */
+	struct nfs4_deviceid be_devid;  /* FIXME: could use device cache instead */
+	struct block_device *be_mdev;
+	sector_t	be_f_offset;	/* the starting offset in the file */
+	sector_t	be_length;	/* the size of the extent */
+	sector_t	be_v_offset;	/* the starting offset in the volume */
+	enum exstate4	be_state;	/* the state of this extent */
+	struct pnfs_inval_markings *be_inval; /* tracks INVAL->RW transition */
+};
+
+/* Shortened extent used by LAYOUTCOMMIT */
+struct pnfs_block_short_extent {
+	struct list_head bse_node;
+	struct nfs4_deviceid bse_devid;
+	struct block_device *bse_mdev;
+	sector_t	bse_f_offset;	/* the starting offset in the file */
+	sector_t	bse_length;	/* the size of the extent */
+};
+
+static inline void
+BL_INIT_INVAL_MARKS(struct pnfs_inval_markings *marks, sector_t blocksize)
+{
+	spin_lock_init(&marks->im_lock);
+	INIT_LIST_HEAD(&marks->im_tree.mtt_stub);
+	marks->im_block_size = blocksize;
+	marks->im_tree.mtt_step_size = min((sector_t)PAGE_CACHE_SECTORS,
+					   blocksize);
+}
+
+enum extentclass4 {
+	RW_EXTENT       = 0, /* READWRTE and INVAL */
+	RO_EXTENT       = 1, /* READ and NONE */
+	EXTENT_LISTS    = 2,
+};
+
+static inline int bl_choose_list(enum exstate4 state)
+{
+	if (state == PNFS_BLOCK_READ_DATA || state == PNFS_BLOCK_NONE_DATA)
+		return RO_EXTENT;
+	else
+		return RW_EXTENT;
+}
+
+struct pnfs_block_layout {
+	struct pnfs_layout_hdr bl_layout;
+	struct pnfs_inval_markings bl_inval; /* tracks INVAL->RW transition */
+	spinlock_t		bl_ext_lock;   /* Protects list manipulation */
+	struct list_head	bl_extents[EXTENT_LISTS]; /* R and RW extents */
+	struct list_head	bl_commit;	/* Needs layout commit */
+	struct list_head	bl_committing;	/* Layout committing */
+	unsigned int		bl_count;	/* entries in bl_commit */
+	sector_t		bl_blocksize;  /* Server blocksize in sectors */
+};
+
+#define BLK_ID(lo) ((struct block_mount_id *)(NFS_SERVER(lo->plh_inode)->pnfs_ld_data))
+
+static inline struct pnfs_block_layout *
+BLK_LO2EXT(struct pnfs_layout_hdr *lo)
+{
+	return container_of(lo, struct pnfs_block_layout, bl_layout);
+}
+
+static inline struct pnfs_block_layout *
+BLK_LSEG2EXT(struct pnfs_layout_segment *lseg)
+{
+	return BLK_LO2EXT(lseg->pls_layout);
+}
+
+struct bl_dev_msg {
+	int status;
+	uint32_t major, minor;
+};
+
+struct bl_msg_hdr {
+	u8  type;
+	u16 totallen; /* length of entire message, including hdr itself */
+};
+
+extern struct dentry *bl_device_pipe;
+extern wait_queue_head_t bl_wq;
+
+#define BL_DEVICE_UMOUNT               0x0 /* Umount--delete devices */
+#define BL_DEVICE_MOUNT                0x1 /* Mount--create devices*/
+#define BL_DEVICE_REQUEST_INIT         0x0 /* Start request */
+#define BL_DEVICE_REQUEST_PROC         0x1 /* User level process succeeds */
+#define BL_DEVICE_REQUEST_ERR          0x2 /* User level process fails */
+
+/* blocklayoutdev.c */
+ssize_t bl_pipe_upcall(struct file *, struct rpc_pipe_msg *,
+		       char __user *, size_t);
+ssize_t bl_pipe_downcall(struct file *, const char __user *, size_t);
+void bl_pipe_destroy_msg(struct rpc_pipe_msg *);
+struct block_device *nfs4_blkdev_get(dev_t dev);
+int nfs4_blkdev_put(struct block_device *bdev);
+struct pnfs_block_dev *nfs4_blk_decode_device(struct nfs_server *server,
+						struct pnfs_device *dev);
+int nfs4_blk_process_layoutget(struct pnfs_layout_hdr *lo,
+				struct nfs4_layoutget_res *lgr, gfp_t gfp_flags);
+
+/* blocklayoutdm.c */
+void bl_free_block_dev(struct pnfs_block_dev *bdev);
+
+/* extents.c */
+struct pnfs_block_extent *
+bl_find_get_extent(struct pnfs_block_layout *bl, sector_t isect,
+		struct pnfs_block_extent **cow_read);
+int bl_mark_sectors_init(struct pnfs_inval_markings *marks,
+			     sector_t offset, sector_t length,
+			     sector_t **pages);
+void bl_put_extent(struct pnfs_block_extent *be);
+struct pnfs_block_extent *bl_alloc_extent(void);
+int bl_is_sector_init(struct pnfs_inval_markings *marks, sector_t isect);
+int encode_pnfs_block_layoutupdate(struct pnfs_block_layout *bl,
+				   struct xdr_stream *xdr,
+				   const struct nfs4_layoutcommit_args *arg);
+void clean_pnfs_block_layoutupdate(struct pnfs_block_layout *bl,
+				   const struct nfs4_layoutcommit_args *arg,
+				   int status);
+int bl_add_merge_extent(struct pnfs_block_layout *bl,
+			 struct pnfs_block_extent *new);
+int bl_mark_for_commit(struct pnfs_block_extent *be,
+			sector_t offset, sector_t length);
+
+#endif /* FS_NFS_NFS4BLOCKLAYOUT_H */

fs/nfs/blocklayout/blocklayoutdev.c

···
+/*
+ *  linux/fs/nfs/blocklayout/blocklayoutdev.c
+ *
+ *  Device operations for the pnfs nfs4 file layout driver.
+ *
+ *  Copyright (c) 2006 The Regents of the University of Michigan.
+ *  All rights reserved.
+ *
+ *  Andy Adamson <[e-mail]>
+ *  Fred Isaman <[e-mail]>
+ *
+ * permission is granted to use, copy, create derivative works and
+ * redistribute this software and such derivative works for any purpose,
+ * so long as the name of the university of michigan is not used in
+ * any advertising or publicity pertaining to the use or distribution
+ * of this software without specific, written prior authorization.  if
+ * the above copyright notice or any other identification of the
+ * university of michigan is included in any copy of any portion of
+ * this software, then the disclaimer below must also be included.
+ *
+ * this software is provided as is, without representation from the
+ * university of michigan as to its fitness for any purpose, and without
+ * warranty by the university of michigan of any kind, either express
+ * or implied, including without limitation the implied warranties of
+ * merchantability and fitness for a particular purpose.  the regents
+ * of the university of michigan shall not be liable for any damages,
+ * including special, indirect, incidental, or consequential damages,
+ * with respect to any claim arising out or in connection with the use
+ * of the software, even if it has been or is hereafter advised of the
+ * possibility of such damages.
+ */
+#include <linux/module.h>
+#include <linux/buffer_head.h> /* __bread */
+
+#include <linux/genhd.h>
+#include <linux/blkdev.h>
+#include <linux/hash.h>
+
+#include "blocklayout.h"
+
+#define NFSDBG_FACILITY         NFSDBG_PNFS_LD
+
+static int decode_sector_number(__be32 **rp, sector_t *sp)
+{
+	uint64_t s;
+
+	*rp = xdr_decode_hyper(*rp, &s);
+	if (s & 0x1ff) {
+		printk(KERN_WARNING "%s: sector not aligned\n", __func__);
+		return -1;
+	}
+	*sp = s >> SECTOR_SHIFT;
+	return 0;
+}
+
+/* Open a block_device by device number. */
+struct block_device *nfs4_blkdev_get(dev_t dev)
+{
+	struct block_device *bd;
+
+	dprintk("%s enter\n", __func__);
+	bd = blkdev_get_by_dev(dev, FMODE_READ, NULL);
+	if (IS_ERR(bd))
+		goto fail;
+	return bd;
+fail:
+	dprintk("%s failed to open device : %ld\n",
+			__func__, PTR_ERR(bd));
+	return NULL;
+}
+
+/*
+ * Release the block device
+ */
+int nfs4_blkdev_put(struct block_device *bdev)
+{
+	dprintk("%s for device %d:%d\n", __func__, MAJOR(bdev->bd_dev),
+			MINOR(bdev->bd_dev));
+	return blkdev_put(bdev, FMODE_READ);
+}
+
+/*
+ * Shouldn't there be a rpc_generic_upcall() to do this for us?
+ */
+ssize_t bl_pipe_upcall(struct file *filp, struct rpc_pipe_msg *msg,
+		       char __user *dst, size_t buflen)
+{
+	char *data = (char *)msg->data + msg->copied;
+	size_t mlen = min(msg->len - msg->copied, buflen);
+	unsigned long left;
+
+	left = copy_to_user(dst, data, mlen);
+	if (left == mlen) {
+		msg->errno = -EFAULT;
+		return -EFAULT;
+	}
+
+	mlen -= left;
+	msg->copied += mlen;
+	msg->errno = 0;
+	return mlen;
+}
+
+static struct bl_dev_msg bl_mount_reply;
+
+ssize_t bl_pipe_downcall(struct file *filp, const char __user *src,
+			 size_t mlen)
+{
+	if (mlen != sizeof (struct bl_dev_msg))
+		return -EINVAL;
+
+	if (copy_from_user(&bl_mount_reply, src, mlen) != 0)
+		return -EFAULT;
+
+	wake_up(&bl_wq);
+
+	return mlen;
+}
+
+void bl_pipe_destroy_msg(struct rpc_pipe_msg *msg)
+{
+	if (msg->errno >= 0)
+		return;
+	wake_up(&bl_wq);
+}
+
+/*
+ * Decodes pnfs_block_deviceaddr4 which is XDR encoded in dev->dev_addr_buf.
+ */
+struct pnfs_block_dev *
+nfs4_blk_decode_device(struct nfs_server *server,
+		       struct pnfs_device *dev)
+{
+	struct pnfs_block_dev *rv = NULL;
+	struct block_device *bd = NULL;
+	struct rpc_pipe_msg msg;
+	struct bl_msg_hdr bl_msg = {
+		.type = BL_DEVICE_MOUNT,
+		.totallen = dev->mincount,
+	};
+	uint8_t *dataptr;
+	DECLARE_WAITQUEUE(wq, current);
+	struct bl_dev_msg *reply = &bl_mount_reply;
+	int offset, len, i;
+
+	dprintk("%s CREATING PIPEFS MESSAGE\n", __func__);
+	dprintk("%s: deviceid: %s, mincount: %d\n", __func__, dev->dev_id.data,
+		dev->mincount);
+
+	memset(&msg, 0, sizeof(msg));
+	msg.data = kzalloc(sizeof(bl_msg) + dev->mincount, GFP_NOFS);
+	if (!msg.data) {
+		rv = ERR_PTR(-ENOMEM);
+		goto out;
+	}
+
+	memcpy(msg.data, &bl_msg, sizeof(bl_msg));
+	dataptr = (uint8_t *) msg.data;
+	len = dev->mincount;
+	offset = sizeof(bl_msg);
+	for (i = 0; len > 0; i++) {
+		memcpy(&dataptr[offset], page_address(dev->pages[i]),
+				len < PAGE_CACHE_SIZE ? len : PAGE_CACHE_SIZE);
+		len -= PAGE_CACHE_SIZE;
+		offset += PAGE_CACHE_SIZE;
+	}
+	msg.len = sizeof(bl_msg) + dev->mincount;
+
+	dprintk("%s CALLING USERSPACE DAEMON\n", __func__);
+	add_wait_queue(&bl_wq, &wq);
+	if (rpc_queue_upcall(bl_device_pipe->d_inode, &msg) < 0) {
+		remove_wait_queue(&bl_wq, &wq);
+		goto out;
+	}
+
+	set_current_state(TASK_UNINTERRUPTIBLE);
+	schedule();
+	__set_current_state(TASK_RUNNING);
+	remove_wait_queue(&bl_wq, &wq);
+
+	if (reply->status != BL_DEVICE_REQUEST_PROC) {
+		dprintk("%s failed to open device: %d\n",
+			__func__, reply->status);
+		rv = ERR_PTR(-EINVAL);
+		goto out;
+	}
+
+	bd = nfs4_blkdev_get(MKDEV(reply->major, reply->minor));
+	if (IS_ERR(bd)) {
+		dprintk("%s failed to open device : %ld\n",
+			__func__, PTR_ERR(bd));
+		goto out;
+	}
+
+	rv = kzalloc(sizeof(*rv), GFP_NOFS);
+	if (!rv) {
+		rv = ERR_PTR(-ENOMEM);
+		goto out;
+	}
+
+	rv->bm_mdev = bd;
+	memcpy(&rv->bm_mdevid, &dev->dev_id, sizeof(struct nfs4_deviceid));
+	dprintk("%s Created device %s with bd_block_size %u\n",
+		__func__,
+		bd->bd_disk->disk_name,
+		bd->bd_block_size);
+
+out:
+	kfree(msg.data);
+	return rv;
+}
+
+/* Map deviceid returned by the server to constructed block_device */
+static struct block_device *translate_devid(struct pnfs_layout_hdr *lo,
+					    struct nfs4_deviceid *id)
+{
+	struct block_device *rv = NULL;
+	struct block_mount_id *mid;
+	struct pnfs_block_dev *dev;
+
+	dprintk("%s enter, lo=%p, id=%p\n", __func__, lo, id);
+	mid = BLK_ID(lo);
+	spin_lock(&mid->bm_lock);
+	list_for_each_entry(dev, &mid->bm_devlist, bm_node) {
+		if (memcmp(id->data, dev->bm_mdevid.data,
+			   NFS4_DEVICEID4_SIZE) == 0) {
+			rv = dev->bm_mdev;
+			goto out;
+		}
+	}
+ out:
+	spin_unlock(&mid->bm_lock);
+	dprintk("%s returning %p\n", __func__, rv);
+	return rv;
+}
+
+/* Tracks info needed to ensure extents in layout obey constraints of spec */
+struct layout_verification {
+	u32 mode;	/* R or RW */
+	u64 start;	/* Expected start of next non-COW extent */
+	u64 inval;	/* Start of INVAL coverage */
+	u64 cowread;	/* End of COW read coverage */
+};
+
+/* Verify the extent meets the layout requirements of the pnfs-block draft,
+ * section 2.3.1.
+ */
+static int verify_extent(struct pnfs_block_extent *be,
+			 struct layout_verification *lv)
+{
+	if (lv->mode == IOMODE_READ) {
+		if (be->be_state == PNFS_BLOCK_READWRITE_DATA ||
+		    be->be_state == PNFS_BLOCK_INVALID_DATA)
+			return -EIO;
+		if (be->be_f_offset != lv->start)
+			return -EIO;
+		lv->start += be->be_length;
+		return 0;
+	}
+	/* lv->mode == IOMODE_RW */
+	if (be->be_state == PNFS_BLOCK_READWRITE_DATA) {
+		if (be->be_f_offset != lv->start)
+			return -EIO;
+		if (lv->cowread > lv->start)
+			return -EIO;
+		lv->start += be->be_length;
+		lv->inval = lv->start;
+		return 0;
+	} else if (be->be_state == PNFS_BLOCK_INVALID_DATA) {
+		if (be->be_f_offset != lv->start)
+			return -EIO;
+		lv->start += be->be_length;
+		return 0;
+	} else if (be->be_state == PNFS_BLOCK_READ_DATA) {
+		if (be->be_f_offset > lv->start)
+			return -EIO;
+		if (be->be_f_offset < lv->inval)
+			return -EIO;
+		if (be->be_f_offset < lv->cowread)
+			return -EIO;
+		/* It looks like you might want to min this with lv->start,
+		 * but you really don't.
+		 */
+		lv->inval = lv->inval + be->be_length;
+		lv->cowread = be->be_f_offset + be->be_length;
+		return 0;
+	} else
+		return -EIO;
+}
+
+/* XDR decode pnfs_block_layout4 structure */
+int
+nfs4_blk_process_layoutget(struct pnfs_layout_hdr *lo,
+			   struct nfs4_layoutget_res *lgr, gfp_t gfp_flags)
+{
+	struct pnfs_block_layout *bl = BLK_LO2EXT(lo);
+	int i, status = -EIO;
+	uint32_t count;
+	struct pnfs_block_extent *be = NULL, *save;
+	struct xdr_stream stream;
+	struct xdr_buf buf;
+	struct page *scratch;
+	__be32 *p;
+	struct layout_verification lv = {
+		.mode = lgr->range.iomode,
+		.start = lgr->range.offset >> SECTOR_SHIFT,
+		.inval = lgr->range.offset >> SECTOR_SHIFT,
+		.cowread = lgr->range.offset >> SECTOR_SHIFT,
+	};
+	LIST_HEAD(extents);
+
+	dprintk("---> %s\n", __func__);
+
+	scratch = alloc_page(gfp_flags);
+	if (!scratch)
+		return -ENOMEM;
+
+	xdr_init_decode_pages(&stream, &buf, lgr->layoutp->pages, lgr->layoutp->len);
+	xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE);
+
+	p = xdr_inline_decode(&stream, 4);
+	if (unlikely(!p))
+		goto out_err;
+
+	count = be32_to_cpup(p++);
+
+	dprintk("%s enter, number of extents %i\n", __func__, count);
+	p = xdr_inline_decode(&stream, (28 + NFS4_DEVICEID4_SIZE) * count);
+	if (unlikely(!p))
+		goto out_err;
+
+	/* Decode individual extents, putting them in temporary
+	 * staging area until whole layout is decoded to make error
+	 * recovery easier.
+	 */
+	for (i = 0; i < count; i++) {
+		be = bl_alloc_extent();
+		if (!be) {
+			status = -ENOMEM;
+			goto out_err;
+		}
+		memcpy(&be->be_devid, p, NFS4_DEVICEID4_SIZE);
+		p += XDR_QUADLEN(NFS4_DEVICEID4_SIZE);
+		be->be_mdev = translate_devid(lo, &be->be_devid);
+		if (!be->be_mdev)
+			goto out_err;
+
+		/* The next three values are read in as bytes,
+		 * but stored as 512-byte sector lengths
+		 */
+		if (decode_sector_number(&p, &be->be_f_offset) < 0)
+			goto out_err;
+		if (decode_sector_number(&p, &be->be_length) < 0)
+			goto out_err;
+		if (decode_sector_number(&p, &be->be_v_offset) < 0)
+			goto out_err;
+		be->be_state = be32_to_cpup(p++);
+		if (be->be_state == PNFS_BLOCK_INVALID_DATA)
+			be->be_inval = &bl->bl_inval;
+		if (verify_extent(be, &lv)) {
+			dprintk("%s verify failed\n", __func__);
+			goto out_err;
+		}
+		list_add_tail(&be->be_node, &extents);
+	}
+	if (lgr->range.offset + lgr->range.length !=
+			lv.start << SECTOR_SHIFT) {
+		dprintk("%s Final length mismatch\n", __func__);
+		be = NULL;
+		goto out_err;
+	}
+	if (lv.start < lv.cowread) {
+		dprintk("%s Final uncovered COW extent\n", __func__);
+		be = NULL;
+		goto out_err;
+	}
+	/* Extents decoded properly, now try to merge them in to
+	 * existing layout extents.
+	 */
+	spin_lock(&bl->bl_ext_lock);
+	list_for_each_entry_safe(be, save, &extents, be_node) {
+		list_del(&be->be_node);
+		status = bl_add_merge_extent(bl, be);
+		if (status) {
+			spin_unlock(&bl->bl_ext_lock);
+			/* This is a fairly catastrophic error, as the
+			 * entire layout extent lists are now corrupted.
+			 * We should have some way to distinguish this.
+			 */
+			be = NULL;
+			goto out_err;
+		}
+	}
+	spin_unlock(&bl->bl_ext_lock);
+	status = 0;
+ out:
+	__free_page(scratch);
+	dprintk("%s returns %i\n", __func__, status);
+	return status;
+
+ out_err:
+	bl_put_extent(be);
+	while (!list_empty(&extents)) {
+		be = list_first_entry(&extents, struct pnfs_block_extent,
+				      be_node);
+		list_del(&be->be_node);
+		bl_put_extent(be);
+	}
+	goto out;
+}

fs/nfs/blocklayout/blocklayoutdm.c

···
+/*
+ *  linux/fs/nfs/blocklayout/blocklayoutdm.c
+ *
+ *  Module for the NFSv4.1 pNFS block layout driver.
+ *
+ *  Copyright (c) 2007 The Regents of the University of Michigan.
+ *  All rights reserved.
+ *
+ *  Fred Isaman <[e-mail]>
+ *  Andy Adamson <[e-mail]>
+ *
+ * permission is granted to use, copy, create derivative works and
+ * redistribute this software and such derivative works for any purpose,
+ * so long as the name of the university of michigan is not used in
+ * any advertising or publicity pertaining to the use or distribution
+ * of this software without specific, written prior authorization.  if
+ * the above copyright notice or any other identification of the
+ * university of michigan is included in any copy of any portion of
+ * this software, then the disclaimer below must also be included.
+ *
+ * this software is provided as is, without representation from the
+ * university of michigan as to its fitness for any purpose, and without
+ * warranty by the university of michigan of any kind, either express
+ * or implied, including without limitation the implied warranties of
+ * merchantability and fitness for a particular purpose.  the regents
+ * of the university of michigan shall not be liable for any damages,
+ * including special, indirect, incidental, or consequential damages,
+ * with respect to any claim arising out or in connection with the use
+ * of the software, even if it has been or is hereafter advised of the
+ * possibility of such damages.
+ */
+
+#include <linux/genhd.h> /* gendisk - used in a dprintk*/
+#include <linux/sched.h>
+#include <linux/hash.h>
+
+#include "blocklayout.h"
+
+#define NFSDBG_FACILITY         NFSDBG_PNFS_LD
+
+static void dev_remove(dev_t dev)
+{
+	struct rpc_pipe_msg msg;
+	struct bl_dev_msg bl_umount_request;
+	struct bl_msg_hdr bl_msg = {
+		.type = BL_DEVICE_UMOUNT,
+		.totallen = sizeof(bl_umount_request),
+	};
+	uint8_t *dataptr;
+	DECLARE_WAITQUEUE(wq, current);
+
+	dprintk("Entering %s\n", __func__);
+
+	memset(&msg, 0, sizeof(msg));
+	msg.data = kzalloc(1 + sizeof(bl_umount_request), GFP_NOFS);
+	if (!msg.data)
+		goto out;
+
+	memset(&bl_umount_request, 0, sizeof(bl_umount_request));
+	bl_umount_request.major = MAJOR(dev);
+	bl_umount_request.minor = MINOR(dev);
+
+	memcpy(msg.data, &bl_msg, sizeof(bl_msg));
+	dataptr = (uint8_t *) msg.data;
+	memcpy(&dataptr[sizeof(bl_msg)], &bl_umount_request, sizeof(bl_umount_request));
+	msg.len = sizeof(bl_msg) + bl_msg.totallen;
+
+	add_wait_queue(&bl_wq, &wq);
+	if (rpc_queue_upcall(bl_device_pipe->d_inode, &msg) < 0) {
+		remove_wait_queue(&bl_wq, &wq);
+		goto out;
+	}
+
+	set_current_state(TASK_UNINTERRUPTIBLE);
+	schedule();
+	__set_current_state(TASK_RUNNING);
+	remove_wait_queue(&bl_wq, &wq);
+
+out:
+	kfree(msg.data);
+}
+
+/*
+ * Release meta device
+ */
+static void nfs4_blk_metadev_release(struct pnfs_block_dev *bdev)
+{
+	int rv;
+
+	dprintk("%s Releasing\n", __func__);
+	rv = nfs4_blkdev_put(bdev->bm_mdev);
+	if (rv)
+		printk(KERN_ERR "%s nfs4_blkdev_put returns %d\n",
+				__func__, rv);
+
+	dev_remove(bdev->bm_mdev->bd_dev);
+}
+
+void bl_free_block_dev(struct pnfs_block_dev *bdev)
+{
+	if (bdev) {
+		if (bdev->bm_mdev) {
+			dprintk("%s Removing DM device: %d:%d\n",
+				__func__,
+				MAJOR(bdev->bm_mdev->bd_dev),
+				MINOR(bdev->bm_mdev->bd_dev));
+			nfs4_blk_metadev_release(bdev);
+		}
+		kfree(bdev);
+	}
+}

fs/nfs/blocklayout/extents.c

···
+/*
+ *  linux/fs/nfs/blocklayout/blocklayout.h
+ *
+ *  Module for the NFSv4.1 pNFS block layout driver.
+ *
+ *  Copyright (c) 2006 The Regents of the University of Michigan.
+ *  All rights reserved.
+ *
+ *  Andy Adamson <[e-mail]>
+ *  Fred Isaman <[e-mail]>
+ *
+ * permission is granted to use, copy, create derivative works and
+ * redistribute this software and such derivative works for any purpose,
+ * so long as the name of the university of michigan is not used in
+ * any advertising or publicity pertaining to the use or distribution
+ * of this software without specific, written prior authorization.  if
+ * the above copyright notice or any other identification of the
+ * university of michigan is included in any copy of any portion of
+ * this software, then the disclaimer below must also be included.
+ *
+ * this software is provided as is, without representation from the
+ * university of michigan as to its fitness for any purpose, and without
+ * warranty by the university of michigan of any kind, either express
+ * or implied, including without limitation the implied warranties of
+ * merchantability and fitness for a particular purpose.  the regents
+ * of the university of michigan shall not be liable for any damages,
+ * including special, indirect, incidental, or consequential damages,
+ * with respect to any claim arising out or in connection with the use
+ * of the software, even if it has been or is hereafter advised of the
+ * possibility of such damages.
+ */
+
+#include "blocklayout.h"
+#define NFSDBG_FACILITY         NFSDBG_PNFS_LD
+
+/* Bit numbers */
+#define EXTENT_INITIALIZED 0
+#define EXTENT_WRITTEN     1
+#define EXTENT_IN_COMMIT   2
+#define INTERNAL_EXISTS    MY_MAX_TAGS
+#define INTERNAL_MASK      ((1 << INTERNAL_EXISTS) - 1)
+
+/* Returns largest t<=s s.t. t%base==0 */
+static inline sector_t normalize(sector_t s, int base)
+{
+	sector_t tmp = s; /* Since do_div modifies its argument */
+	return s - do_div(tmp, base);
+}
+
+static inline sector_t normalize_up(sector_t s, int base)
+{
+	return normalize(s + base - 1, base);
+}
+
+/* Complete stub using list while determine API wanted */
+
+/* Returns tags, or negative */
+static int32_t _find_entry(struct my_tree *tree, u64 s)
+{
+	struct pnfs_inval_tracking *pos;
+
+	dprintk("%s(%llu) enter\n", __func__, s);
+	list_for_each_entry_reverse(pos, &tree->mtt_stub, it_link) {
+		if (pos->it_sector > s)
+			continue;
+		else if (pos->it_sector == s)
+			return pos->it_tags & INTERNAL_MASK;
+		else
+			break;
+	}
+	return -ENOENT;
+}
+
+static inline
+int _has_tag(struct my_tree *tree, u64 s, int32_t tag)
+{
+	int32_t tags;
+
+	dprintk("%s(%llu, %i) enter\n", __func__, s, tag);
+	s = normalize(s, tree->mtt_step_size);
+	tags = _find_entry(tree, s);
+	if ((tags < 0) || !(tags & (1 << tag)))
+		return 0;
+	else
+		return 1;
+}
+
+/* Creates entry with tag, or if entry already exists, unions tag to it.
+ * If storage is not NULL, newly created entry will use it.
+ * Returns number of entries added, or negative on error.
+ */
+static int _add_entry(struct my_tree *tree, u64 s, int32_t tag,
+		      struct pnfs_inval_tracking *storage)
+{
+	int found = 0;
+	struct pnfs_inval_tracking *pos;
+
+	dprintk("%s(%llu, %i, %p) enter\n", __func__, s, tag, storage);
+	list_for_each_entry_reverse(pos, &tree->mtt_stub, it_link) {
+		if (pos->it_sector > s)
+			continue;
+		else if (pos->it_sector == s) {
+			found = 1;
+			break;
+		} else
+			break;
+	}
+	if (found) {
+		pos->it_tags |= (1 << tag);
+		return 0;
+	} else {
+		struct pnfs_inval_tracking *new;
+		if (storage)
+			new = storage;
+		else {
+			new = kmalloc(sizeof(*new), GFP_NOFS);
+			if (!new)
+				return -ENOMEM;
+		}
+		new->it_sector = s;
+		new->it_tags = (1 << tag);
+		list_add(&new->it_link, &pos->it_link);
+		return 1;
+	}
+}
+
+/* XXXX Really want option to not create */
+/* Over range, unions tag with existing entries, else creates entry with tag */
+static int _set_range(struct my_tree *tree, int32_t tag, u64 s, u64 length)
+{
+	u64 i;
+
+	dprintk("%s(%i, %llu, %llu) enter\n", __func__, tag, s, length);
+	for (i = normalize(s, tree->mtt_step_size); i < s + length;
+	     i += tree->mtt_step_size)
+		if (_add_entry(tree, i, tag, NULL))
+			return -ENOMEM;
+	return 0;
+}
+
+/* Ensure that future operations on given range of tree will not malloc */
+static int _preload_range(struct my_tree *tree, u64 offset, u64 length)
+{
+	u64 start, end, s;
+	int count, i, used = 0, status = -ENOMEM;
+	struct pnfs_inval_tracking **storage;
+
+	dprintk("%s(%llu, %llu) enter\n", __func__, offset, length);
+	start = normalize(offset, tree->mtt_step_size);
+	end = normalize_up(offset + length, tree->mtt_step_size);
+	count = (int)(end - start) / (int)tree->mtt_step_size;
+
+	/* Pre-malloc what memory we might need */
+	storage = kmalloc(sizeof(*storage) * count, GFP_NOFS);
+	if (!storage)
+		return -ENOMEM;
+	for (i = 0; i < count; i++) {
+		storage[i] = kmalloc(sizeof(struct pnfs_inval_tracking),
+				     GFP_NOFS);
+		if (!storage[i])
+			goto out_cleanup;
+	}
+
+	/* Now need lock - HOW??? */
+
+	for (s = start; s < end; s += tree->mtt_step_size)
+		used += _add_entry(tree, s, INTERNAL_EXISTS, storage[used]);
+
+	/* Unlock - HOW??? */
+	status = 0;
+
+ out_cleanup:
+	for (i = used; i < count; i++) {
+		if (!storage[i])
+			break;
+		kfree(storage[i]);
+	}
+	kfree(storage);
+	return status;
+}
+
+static void set_needs_init(sector_t *array, sector_t offset)
+{
+	sector_t *p = array;
+
+	dprintk("%s enter\n", __func__);
+	if (!p)
+		return;
+	while (*p < offset)
+		p++;
+	if (*p == offset)
+		return;
+	else if (*p == ~0) {
+		*p++ = offset;
+		*p = ~0;
+		return;
+	} else {
+		sector_t *save = p;
+		dprintk("%s Adding %llu\n", __func__, (u64)offset);
+		while (*p != ~0)
+			p++;
+		p++;
+		memmove(save + 1, save, (char *)p - (char *)save);
+		*save = offset;
+		return;
+	}
+}
+
+/* We are relying on page lock to serialize this */
+int bl_is_sector_init(struct pnfs_inval_markings *marks, sector_t isect)
+{
+	int rv;
+
+	spin_lock(&marks->im_lock);
+	rv = _has_tag(&marks->im_tree, isect, EXTENT_INITIALIZED);
+	spin_unlock(&marks->im_lock);
+	return rv;
+}
+
+/* Assume start, end already sector aligned */
+static int
+_range_has_tag(struct my_tree *tree, u64 start, u64 end, int32_t tag)
+{
+	struct pnfs_inval_tracking *pos;
+	u64 expect = 0;
+
+	dprintk("%s(%llu, %llu, %i) enter\n", __func__, start, end, tag);
+	list_for_each_entry_reverse(pos, &tree->mtt_stub, it_link) {
+		if (pos->it_sector >= end)
+			continue;
+		if (!expect) {
+			if ((pos->it_sector == end - tree->mtt_step_size) &&
+			    (pos->it_tags & (1 << tag))) {
+				expect = pos->it_sector - tree->mtt_step_size;
+				if (pos->it_sector < tree->mtt_step_size || expect < start)
+					return 1;
+				continue;
+			} else {
+				return 0;
+			}
+		}
+		if (pos->it_sector != expect || !(pos->it_tags & (1 << tag)))
+			return 0;
+		expect -= tree->mtt_step_size;
+		if (expect < start)
+			return 1;
+	}
+	return 0;
+}
+
+static int is_range_written(struct pnfs_inval_markings *marks,
+			    sector_t start, sector_t end)
+{
+	int rv;
+
+	spin_lock(&marks->im_lock);
+	rv = _range_has_tag(&marks->im_tree, start, end, EXTENT_WRITTEN);
+	spin_unlock(&marks->im_lock);
+	return rv;
+}
+
+/* Marks sectors in [offest, offset_length) as having been initialized.
+ * All lengths are step-aligned, where step is min(pagesize, blocksize).
+ * Notes where partial block is initialized, and helps prepare it for
+ * complete initialization later.
+ */
+/* Currently assumes offset is page-aligned */
+int bl_mark_sectors_init(struct pnfs_inval_markings *marks,
+			     sector_t offset, sector_t length,
+			     sector_t **pages)
+{
+	sector_t s, start, end;
+	sector_t *array = NULL; /* Pages to mark */
+
+	dprintk("%s(offset=%llu,len=%llu) enter\n",
+		__func__, (u64)offset, (u64)length);
+	s = max((sector_t) 3,
+		2 * (marks->im_block_size / (PAGE_CACHE_SECTORS)));
+	dprintk("%s set max=%llu\n", __func__, (u64)s);
+	if (pages) {
+		array = kmalloc(s * sizeof(sector_t), GFP_NOFS);
+		if (!array)
+			goto outerr;
+		array[0] = ~0;
+	}
+
+	start = normalize(offset, marks->im_block_size);
+	end = normalize_up(offset + length, marks->im_block_size);
+	if (_preload_range(&marks->im_tree, start, end - start))
+		goto outerr;
+
+	spin_lock(&marks->im_lock);
+
+	for (s = normalize_up(start, PAGE_CACHE_SECTORS);
+	     s < offset; s += PAGE_CACHE_SECTORS) {
+		dprintk("%s pre-area pages\n", __func__);
+		/* Portion of used block is not initialized */
+		if (!_has_tag(&marks->im_tree, s, EXTENT_INITIALIZED))
+			set_needs_init(array, s);
+	}
+	if (_set_range(&marks->im_tree, EXTENT_INITIALIZED, offset, length))
+		goto out_unlock;
+	for (s = normalize_up(offset + length, PAGE_CACHE_SECTORS);
+	     s < end; s += PAGE_CACHE_SECTORS) {
+		dprintk("%s post-area pages\n", __func__);
+		if (!_has_tag(&marks->im_tree, s, EXTENT_INITIALIZED))
+			set_needs_init(array, s);
+	}
+
+	spin_unlock(&marks->im_lock);
+
+	if (pages) {
+		if (array[0] == ~0) {
+			kfree(array);
+			*pages = NULL;
+		} else
+			*pages = array;
+	}
+	return 0;
+
+ out_unlock:
+	spin_unlock(&marks->im_lock);
+ outerr:
+	if (pages) {
+		kfree(array);
+		*pages = NULL;
+	}
+	return -ENOMEM;
+}
+
+/* Marks sectors in [offest, offset+length) as having been written to disk.
+ * All lengths should be block aligned.
+ */
+static int mark_written_sectors(struct pnfs_inval_markings *marks,
+				sector_t offset, sector_t length)
+{
+	int status;
+
+	dprintk("%s(offset=%llu,len=%llu) enter\n", __func__,
+		(u64)offset, (u64)length);
+	spin_lock(&marks->im_lock);
+	status = _set_range(&marks->im_tree, EXTENT_WRITTEN, offset, length);
+	spin_unlock(&marks->im_lock);
+	return status;
+}
+
+static void print_short_extent(struct pnfs_block_short_extent *be)
+{
+	dprintk("PRINT SHORT EXTENT extent %p\n", be);
+	if (be) {
+		dprintk("        be_f_offset %llu\n", (u64)be->bse_f_offset);
+		dprintk("        be_length   %llu\n", (u64)be->bse_length);
+	}
+}
+
+static void print_clist(struct list_head *list, unsigned int count)
+{
+	struct pnfs_block_short_extent *be;
+	unsigned int i = 0;
+
+	ifdebug(FACILITY) {
+		printk(KERN_DEBUG "****************\n");
+		printk(KERN_DEBUG "Extent list looks like:\n");
+		list_for_each_entry(be, list, bse_node) {
+			i++;
+			print_short_extent(be);
+		}
+		if (i != count)
+			printk(KERN_DEBUG "\n\nExpected %u entries\n\n\n", count);
+		printk(KERN_DEBUG "****************\n");
+	}
+}
+
+/* Note: In theory, we should do more checking that devid's match between
+ * old and new, but if they don't, the lists are too corrupt to salvage anyway.
+ */
+/* Note this is very similar to bl_add_merge_extent */
+static void add_to_commitlist(struct pnfs_block_layout *bl,
+			      struct pnfs_block_short_extent *new)
+{
+	struct list_head *clist = &bl->bl_commit;
+	struct pnfs_block_short_extent *old, *save;
+	sector_t end = new->bse_f_offset + new->bse_length;
+
+	dprintk("%s enter\n", __func__);
+	print_short_extent(new);
+	print_clist(clist, bl->bl_count);
+	bl->bl_count++;
+	/* Scan for proper place to insert, extending new to the left
+	 * as much as possible.
+	 */
+	list_for_each_entry_safe(old, save, clist, bse_node) {
+		if (new->bse_f_offset < old->bse_f_offset)
+			break;
+		if (end <= old->bse_f_offset + old->bse_length) {
+			/* Range is already in list */
+			bl->bl_count--;
+			kfree(new);
+			return;
+		} else if (new->bse_f_offset <=
+				old->bse_f_offset + old->bse_length) {
+			/* new overlaps or abuts existing be */
+			if (new->bse_mdev == old->bse_mdev) {
+				/* extend new to fully replace old */
+				new->bse_length += new->bse_f_offset -
+						old->bse_f_offset;
+				new->bse_f_offset = old->bse_f_offset;
+				list_del(&old->bse_node);
+				bl->bl_count--;
+				kfree(old);
+			}
+		}
+	}
+	/* Note that if we never hit the above break, old will not point to a
+	 * valid extent.  However, in that case &old->bse_node==list.
+	 */
+	list_add_tail(&new->bse_node, &old->bse_node);
+	/* Scan forward for overlaps.  If we find any, extend new and
+	 * remove the overlapped extent.
+	 */
+	old = list_prepare_entry(new, clist, bse_node);
+	list_for_each_entry_safe_continue(old, save, clist, bse_node) {
+		if (end < old->bse_f_offset)
+			break;
+		/* new overlaps or abuts old */
+		if (new->bse_mdev == old->bse_mdev) {
+			if (end < old->bse_f_offset + old->bse_length) {
+				/* extend new to fully cover old */
+				end = old->bse_f_offset + old->bse_length;
+				new->bse_length = end - new->bse_f_offset;
+			}
+			list_del(&old->bse_node);
+			bl->bl_count--;
+			kfree(old);
+		}
+	}
+	dprintk("%s: after merging\n", __func__);
+	print_clist(clist, bl->bl_count);
+}
+
+/* Note the range described by offset, length is guaranteed to be contained
+ * within be.
+ */
+int bl_mark_for_commit(struct pnfs_block_extent *be,
+		    sector_t offset, sector_t length)
+{
+	sector_t new_end, end = offset + length;
+	struct pnfs_block_short_extent *new;
+	struct pnfs_block_layout *bl = container_of(be->be_inval,
+						    struct pnfs_block_layout,
+						    bl_inval);
+
+	new = kmalloc(sizeof(*new), GFP_NOFS);
+	if (!new)
+		return -ENOMEM;
+
+	mark_written_sectors(be->be_inval, offset, length);
+	/* We want to add the range to commit list, but it must be
+	 * block-normalized, and verified that the normalized range has
+	 * been entirely written to disk.
+	 */
+	new->bse_f_offset = offset;
+	offset = normalize(offset, bl->bl_blocksize);
+	if (offset < new->bse_f_offset) {
+		if (is_range_written(be->be_inval, offset, new->bse_f_offset))
+			new->bse_f_offset = offset;
+		else
+			new->bse_f_offset = offset + bl->bl_blocksize;
+	}
+	new_end = normalize_up(end, bl->bl_blocksize);
+	if (end < new_end) {
+		if (is_range_written(be->be_inval, end, new_end))
+			end = new_end;
+		else
+			end = new_end - bl->bl_blocksize;
+	}
+	if (end <= new->bse_f_offset) {
+		kfree(new);
+		return 0;
+	}
+	new->bse_length = end - new->bse_f_offset;
+	new->bse_devid = be->be_devid;
+	new->bse_mdev = be->be_mdev;
+
+	spin_lock(&bl->bl_ext_lock);
+	/* new will be freed, either by add_to_commitlist if it decides not
+	 * to use it, or after LAYOUTCOMMIT uses it in the commitlist.
+	 */
+	add_to_commitlist(bl, new);
+	spin_unlock(&bl->bl_ext_lock);
+	return 0;
+}
+
+static void print_bl_extent(struct pnfs_block_extent *be)
+{
+	dprintk("PRINT EXTENT extent %p\n", be);
+	if (be) {
+		dprintk("        be_f_offset %llu\n", (u64)be->be_f_offset);
+		dprintk("        be_length   %llu\n", (u64)be->be_length);
+		dprintk("        be_v_offset %llu\n", (u64)be->be_v_offset);
+		dprintk("        be_state    %d\n", be->be_state);
+	}
+}
+
+static void
+destroy_extent(struct kref *kref)
+{
+	struct pnfs_block_extent *be;
+
+	be = container_of(kref, struct pnfs_block_extent, be_refcnt);
+	dprintk("%s be=%p\n", __func__, be);
+	kfree(be);
+}
+
+void
+bl_put_extent(struct pnfs_block_extent *be)
+{
+	if (be) {
+		dprintk("%s enter %p (%i)\n", __func__, be,
+			atomic_read(&be->be_refcnt.refcount));
+		kref_put(&be->be_refcnt, destroy_extent);
+	}
+}
+
+struct pnfs_block_extent *bl_alloc_extent(void)
+{
+	struct pnfs_block_extent *be;
+
+	be = kmalloc(sizeof(struct pnfs_block_extent), GFP_NOFS);
+	if (!be)
+		return NULL;
+	INIT_LIST_HEAD(&be->be_node);
+	kref_init(&be->be_refcnt);
+	be->be_inval = NULL;
+	return be;
+}
+
+static void print_elist(struct list_head *list)
+{
+	struct pnfs_block_extent *be;
+	dprintk("****************\n");
+	dprintk("Extent list looks like:\n");
+	list_for_each_entry(be, list, be_node) {
+		print_bl_extent(be);
+	}
+	dprintk("****************\n");
+}
+
+static inline int
+extents_consistent(struct pnfs_block_extent *old, struct pnfs_block_extent *new)
+{
+	/* Note this assumes new->be_f_offset >= old->be_f_offset */
+	return (new->be_state == old->be_state) &&
+		((new->be_state == PNFS_BLOCK_NONE_DATA) ||
+		 ((new->be_v_offset - old->be_v_offset ==
+		   new->be_f_offset - old->be_f_offset) &&
+		  new->be_mdev == old->be_mdev));
+}
+
+/* Adds new to appropriate list in bl, modifying new and removing existing
+ * extents as appropriate to deal with overlaps.
+ *
+ * See bl_find_get_extent for list constraints.
+ *
+ * Refcount on new is already set.  If end up not using it, or error out,
+ * need to put the reference.
+ *
+ * bl->bl_ext_lock is held by caller.
+ */
+int
+bl_add_merge_extent(struct pnfs_block_layout *bl,
+		     struct pnfs_block_extent *new)
+{
+	struct pnfs_block_extent *be, *tmp;
+	sector_t end = new->be_f_offset + new->be_length;
+	struct list_head *list;
+
+	dprintk("%s enter with be=%p\n", __func__, new);
+	print_bl_extent(new);
+	list = &bl->bl_extents[bl_choose_list(new->be_state)];
+	print_elist(list);
+
+	/* Scan for proper place to insert, extending new to the left
+	 * as much as possible.
+	 */
+	list_for_each_entry_safe_reverse(be, tmp, list, be_node) {
+		if (new->be_f_offset >= be->be_f_offset + be->be_length)
+			break;
+		if (new->be_f_offset >= be->be_f_offset) {
+			if (end <= be->be_f_offset + be->be_length) {
+				/* new is a subset of existing be*/
+				if (extents_consistent(be, new)) {
+					dprintk("%s: new is subset, ignoring\n",
+						__func__);
+					bl_put_extent(new);
+					return 0;
+				} else {
+					goto out_err;
+				}
+			} else {
+				/* |<--   be   -->|
+				 *          |<--   new   -->| */
+				if (extents_consistent(be, new)) {
+					/* extend new to fully replace be */
+					new->be_length += new->be_f_offset -
+						be->be_f_offset;
+					new->be_f_offset = be->be_f_offset;
+					new->be_v_offset = be->be_v_offset;
+					dprintk("%s: removing %p\n", __func__, be);
+					list_del(&be->be_node);
+					bl_put_extent(be);
+				} else {
+					goto out_err;
+				}
+			}
+		} else if (end >= be->be_f_offset + be->be_length) {
+			/* new extent overlap existing be */
+			if (extents_consistent(be, new)) {
+				/* extend new to fully replace be */
+				dprintk("%s: removing %p\n", __func__, be);
+				list_del(&be->be_node);
+				bl_put_extent(be);
+			} else {
+				goto out_err;
+			}
+		} else if (end > be->be_f_offset) {
+			/*           |<--   be   -->|
+			 *|<--   new   -->| */
+			if (extents_consistent(new, be)) {
+				/* extend new to fully replace be */
+				new->be_length += be->be_f_offset + be->be_length -
+					new->be_f_offset - new->be_length;
+				dprintk("%s: removing %p\n", __func__, be);
+				list_del(&be->be_node);
+				bl_put_extent(be);
+			} else {
+				goto out_err;
+			}
+		}
+	}
+	/* Note that if we never hit the above break, be will not point to a
+	 * valid extent.  However, in that case &be->be_node==list.
+	 */
+	list_add(&new->be_node, &be->be_node);
+	dprintk("%s: inserting new\n", __func__);
+	print_elist(list);
+	/* FIXME - The per-list consistency checks have all been done,
+	 * should now check cross-list consistency.
+	 */
+	return 0;
+
+ out_err:
+	bl_put_extent(new);
+	return -EIO;
+}
+
+/* Returns extent, or NULL.  If a second READ extent exists, it is returned
+ * in cow_read, if given.
+ *
+ * The extents are kept in two seperate ordered lists, one for READ and NONE,
+ * one for READWRITE and INVALID.  Within each list, we assume:
+ * 1. Extents are ordered by file offset.
+ * 2. For any given isect, there is at most one extents that matches.
+ */
+struct pnfs_block_extent *
+bl_find_get_extent(struct pnfs_block_layout *bl, sector_t isect,
+	    struct pnfs_block_extent **cow_read)
+{
+	struct pnfs_block_extent *be, *cow, *ret;
+	int i;
+
+	dprintk("%s enter with isect %llu\n", __func__, (u64)isect);
+	cow = ret = NULL;
+	spin_lock(&bl->bl_ext_lock);
+	for (i = 0; i < EXTENT_LISTS; i++) {
+		list_for_each_entry_reverse(be, &bl->bl_extents[i], be_node) {
+			if (isect >= be->be_f_offset + be->be_length)
+				break;
+			if (isect >= be->be_f_offset) {
+				/* We have found an extent */
+				dprintk("%s Get %p (%i)\n", __func__, be,
+					atomic_read(&be->be_refcnt.refcount));
+				kref_get(&be->be_refcnt);
+				if (!ret)
+					ret = be;
+				else if (be->be_state != PNFS_BLOCK_READ_DATA)
+					bl_put_extent(be);
+				else
+					cow = be;
+				break;
+			}
+		}
+		if (ret &&
+		    (!cow_read || ret->be_state != PNFS_BLOCK_INVALID_DATA))
+			break;
+	}
+	spin_unlock(&bl->bl_ext_lock);
+	if (cow_read)
+		*cow_read = cow;
+	print_bl_extent(ret);
+	return ret;
+}
+
+/* Similar to bl_find_get_extent, but called with lock held, and ignores cow */
+static struct pnfs_block_extent *
+bl_find_get_extent_locked(struct pnfs_block_layout *bl, sector_t isect)
+{
+	struct pnfs_block_extent *be, *ret = NULL;
+	int i;
+
+	dprintk("%s enter with isect %llu\n", __func__, (u64)isect);
+	for (i = 0; i < EXTENT_LISTS; i++) {
+		if (ret)
+			break;
+		list_for_each_entry_reverse(be, &bl->bl_extents[i], be_node) {
+			if (isect >= be->be_f_offset + be->be_length)
+				break;
+			if (isect >= be->be_f_offset) {
+				/* We have found an extent */
+				dprintk("%s Get %p (%i)\n", __func__, be,
+					atomic_read(&be->be_refcnt.refcount));
+				kref_get(&be->be_refcnt);
+				ret = be;
+				break;
+			}
+		}
+	}
+	print_bl_extent(ret);
+	return ret;
+}
+
+int
+encode_pnfs_block_layoutupdate(struct pnfs_block_layout *bl,
+			       struct xdr_stream *xdr,
+			       const struct nfs4_layoutcommit_args *arg)
+{
+	struct pnfs_block_short_extent *lce, *save;
+	unsigned int count = 0;
+	__be32 *p, *xdr_start;
+
+	dprintk("%s enter\n", __func__);
+	/* BUG - creation of bl_commit is buggy - need to wait for
+	 * entire block to be marked WRITTEN before it can be added.
+	 */
+	spin_lock(&bl->bl_ext_lock);
+	/* Want to adjust for possible truncate */
+	/* We now want to adjust argument range */
+
+	/* XDR encode the ranges found */
+	xdr_start = xdr_reserve_space(xdr, 8);
+	if (!xdr_start)
+		goto out;
+	list_for_each_entry_safe(lce, save, &bl->bl_commit, bse_node) {
+		p = xdr_reserve_space(xdr, 7 * 4 + sizeof(lce->bse_devid.data));
+		if (!p)
+			break;
+		p = xdr_encode_opaque_fixed(p, lce->bse_devid.data, NFS4_DEVICEID4_SIZE);
+		p = xdr_encode_hyper(p, lce->bse_f_offset << SECTOR_SHIFT);
+		p = xdr_encode_hyper(p, lce->bse_length << SECTOR_SHIFT);
+		p = xdr_encode_hyper(p, 0LL);
+		*p++ = cpu_to_be32(PNFS_BLOCK_READWRITE_DATA);
+		list_del(&lce->bse_node);
+		list_add_tail(&lce->bse_node, &bl->bl_committing);
+		bl->bl_count--;
+		count++;
+	}
+	xdr_start[0] = cpu_to_be32((xdr->p - xdr_start - 1) * 4);
+	xdr_start[1] = cpu_to_be32(count);
+out:
+	spin_unlock(&bl->bl_ext_lock);
+	dprintk("%s found %i ranges\n", __func__, count);
+	return 0;
+}
+
+/* Helper function to set_to_rw that initialize a new extent */
+static void
+_prep_new_extent(struct pnfs_block_extent *new,
+		 struct pnfs_block_extent *orig,
+		 sector_t offset, sector_t length, int state)
+{
+	kref_init(&new->be_refcnt);
+	/* don't need to INIT_LIST_HEAD(&new->be_node) */
+	memcpy(&new->be_devid, &orig->be_devid, sizeof(struct nfs4_deviceid));
+	new->be_mdev = orig->be_mdev;
+	new->be_f_offset = offset;
+	new->be_length = length;
+	new->be_v_offset = orig->be_v_offset - orig->be_f_offset + offset;
+	new->be_state = state;
+	new->be_inval = orig->be_inval;
+}
+
+/* Tries to merge be with extent in front of it in list.
+ * Frees storage if not used.
+ */
+static struct pnfs_block_extent *
+_front_merge(struct pnfs_block_extent *be, struct list_head *head,
+	     struct pnfs_block_extent *storage)
+{
+	struct pnfs_block_extent *prev;
+
+	if (!storage)
+		goto no_merge;
+	if (&be->be_node == head || be->be_node.prev == head)
+		goto no_merge;
+	prev = list_entry(be->be_node.prev, struct pnfs_block_extent, be_node);
+	if ((prev->be_f_offset + prev->be_length != be->be_f_offset) ||
+	    !extents_consistent(prev, be))
+		goto no_merge;
+	_prep_new_extent(storage, prev, prev->be_f_offset,
+			 prev->be_length + be->be_length, prev->be_state);
+	list_replace(&prev->be_node, &storage->be_node);
+	bl_put_extent(prev);
+	list_del(&be->be_node);
+	bl_put_extent(be);
+	return storage;
+
+ no_merge:
+	kfree(storage);
+	return be;
+}
+
+static u64
+set_to_rw(struct pnfs_block_layout *bl, u64 offset, u64 length)
+{
+	u64 rv = offset + length;
+	struct pnfs_block_extent *be, *e1, *e2, *e3, *new, *old;
+	struct pnfs_block_extent *children[3];
+	struct pnfs_block_extent *merge1 = NULL, *merge2 = NULL;
+	int i = 0, j;
+
+	dprintk("%s(%llu, %llu)\n", __func__, offset, length);
+	/* Create storage for up to three new extents e1, e2, e3 */
+	e1 = kmalloc(sizeof(*e1), GFP_ATOMIC);
+	e2 = kmalloc(sizeof(*e2), GFP_ATOMIC);
+	e3 = kmalloc(sizeof(*e3), GFP_ATOMIC);
+	/* BUG - we are ignoring any failure */
+	if (!e1 || !e2 || !e3)
+		goto out_nosplit;
+
+	spin_lock(&bl->bl_ext_lock);
+	be = bl_find_get_extent_locked(bl, offset);
+	rv = be->be_f_offset + be->be_length;
+	if (be->be_state != PNFS_BLOCK_INVALID_DATA) {
+		spin_unlock(&bl->bl_ext_lock);
+		goto out_nosplit;
+	}
+	/* Add e* to children, bumping e*'s krefs */
+	if (be->be_f_offset != offset) {
+		_prep_new_extent(e1, be, be->be_f_offset,
+				 offset - be->be_f_offset,
+				 PNFS_BLOCK_INVALID_DATA);
+		children[i++] = e1;
+		print_bl_extent(e1);
+	} else
+		merge1 = e1;
+	_prep_new_extent(e2, be, offset,
+			 min(length, be->be_f_offset + be->be_length - offset),
+			 PNFS_BLOCK_READWRITE_DATA);
+	children[i++] = e2;
+	print_bl_extent(e2);
+	if (offset + length < be->be_f_offset + be->be_length) {
+		_prep_new_extent(e3, be, e2->be_f_offset + e2->be_length,
+				 be->be_f_offset + be->be_length -
+				 offset - length,
+				 PNFS_BLOCK_INVALID_DATA);
+		children[i++] = e3;
+		print_bl_extent(e3);
+	} else
+		merge2 = e3;
+
+	/* Remove be from list, and insert the e* */
+	/* We don't get refs on e*, since this list is the base reference
+	 * set when init'ed.
+	 */
+	if (i < 3)
+		children[i] = NULL;
+	new = children[0];
+	list_replace(&be->be_node, &new->be_node);
+	bl_put_extent(be);
+	new = _front_merge(new, &bl->bl_extents[RW_EXTENT], merge1);
+	for (j = 1; j < i; j++) {
+		old = new;
+		new = children[j];
+		list_add(&new->be_node, &old->be_node);
+	}
+	if (merge2) {
+		/* This is a HACK, should just create a _back_merge function */
+		new = list_entry(new->be_node.next,
+				 struct pnfs_block_extent, be_node);
+		new = _front_merge(new, &bl->bl_extents[RW_EXTENT], merge2);
+	}
+	spin_unlock(&bl->bl_ext_lock);
+
+	/* Since we removed the base reference above, be is now scheduled for
+	 * destruction.
+	 */
+	bl_put_extent(be);
+	dprintk("%s returns %llu after split\n", __func__, rv);
+	return rv;
+
+ out_nosplit:
+	kfree(e1);
+	kfree(e2);
+	kfree(e3);
+	dprintk("%s returns %llu without splitting\n", __func__, rv);
+	return rv;
+}
+
+void
+clean_pnfs_block_layoutupdate(struct pnfs_block_layout *bl,
+			      const struct nfs4_layoutcommit_args *arg,
+			      int status)
+{
+	struct pnfs_block_short_extent *lce, *save;
+
+	dprintk("%s status %d\n", __func__, status);
+	list_for_each_entry_safe(lce, save, &bl->bl_committing, bse_node) {
+		if (likely(!status)) {
+			u64 offset = lce->bse_f_offset;
+			u64 end = offset + lce->bse_length;
+
+			do {
+				offset = set_to_rw(bl, offset, end - offset);
+			} while (offset < end);
+			list_del(&lce->bse_node);
+
+			kfree(lce);
+		} else {
+			list_del(&lce->bse_node);
+			spin_lock(&bl->bl_ext_lock);
+			add_to_commitlist(bl, lce);
+			spin_unlock(&bl->bl_ext_lock);
+		}
+	}
+}

fs/nfs/client.c

···
 	.nrvers			= ARRAY_SIZE(nfs_version),
 	.version		= nfs_version,
 	.stats			= &nfs_rpcstat,
-	.pipe_dir_name		= "/nfs",
+	.pipe_dir_name		= NFS_PIPE_DIRNAME,
 };
 
 struct rpc_stat nfs_rpcstat = {
···
 /*
  * Load up the server record from information gained in an fsinfo record
  */
-static void nfs_server_set_fsinfo(struct nfs_server *server, struct nfs_fsinfo *fsinfo)
+static void nfs_server_set_fsinfo(struct nfs_server *server,
+				  struct nfs_fh *mntfh,
+				  struct nfs_fsinfo *fsinfo)
 {
 	unsigned long max_rpc_payload;
 
···
 	if (server->wsize > NFS_MAX_FILE_IO_SIZE)
 		server->wsize = NFS_MAX_FILE_IO_SIZE;
 	server->wpages = (server->wsize + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
-	set_pnfs_layoutdriver(server, fsinfo->layouttype);
+	server->pnfs_blksize = fsinfo->blksize;
+	set_pnfs_layoutdriver(server, mntfh, fsinfo->layouttype);
 
 	server->wtmult = nfs_block_bits(fsinfo->wtmult, NULL);
 
···
 	if (error < 0)
 		goto out_error;
 
-	nfs_server_set_fsinfo(server, &fsinfo);
+	nfs_server_set_fsinfo(server, mntfh, &fsinfo);
 
 	/* Get some general file system info */
 	if (server->namelen == 0) {

fs/nfs/dir.c

···
 
 #endif /* CONFIG_NFS_V4 */
 
-static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct rpc_cred *cred)
+static struct nfs_open_dir_context *alloc_nfs_open_dir_context(struct inode *dir, struct rpc_cred *cred)
 {
 	struct nfs_open_dir_context *ctx;
 	ctx = kmalloc(sizeof(*ctx), GFP_KERNEL);
 	if (ctx != NULL) {
 		ctx->duped = 0;
+		ctx->attr_gencount = NFS_I(dir)->attr_gencount;
 		ctx->dir_cookie = 0;
 		ctx->dup_cookie = 0;
 		ctx->cred = get_rpccred(cred);
-	} else
-		ctx = ERR_PTR(-ENOMEM);
-	return ctx;
+		return ctx;
+	}
+	return  ERR_PTR(-ENOMEM);
 }
 
 static void put_nfs_open_dir_context(struct nfs_open_dir_context *ctx)
···
 	cred = rpc_lookup_cred();
 	if (IS_ERR(cred))
 		return PTR_ERR(cred);
-	ctx = alloc_nfs_open_dir_context(cred);
+	ctx = alloc_nfs_open_dir_context(inode, cred);
 	if (IS_ERR(ctx)) {
 		res = PTR_ERR(ctx);
 		goto out;
···
 {
 	loff_t diff = desc->file->f_pos - desc->current_index;
 	unsigned int index;
-	struct nfs_open_dir_context *ctx = desc->file->private_data;
 
 	if (diff < 0)
 		goto out_eof;
···
 	index = (unsigned int)diff;
 	*desc->dir_cookie = array->array[index].cookie;
 	desc->cache_entry_index = index;
-	ctx->duped = 0;
 	return 0;
 out_eof:
 	desc->eof = 1;
···
 	int i;
 	loff_t new_pos;
 	int status = -EAGAIN;
-	struct nfs_open_dir_context *ctx = desc->file->private_data;
 
 	for (i = 0; i < array->size; i++) {
 		if (array->array[i].cookie == *desc->dir_cookie) {
+			struct nfs_inode *nfsi = NFS_I(desc->file->f_path.dentry->d_inode);
+			struct nfs_open_dir_context *ctx = desc->file->private_data;
+
 			new_pos = desc->current_index + i;
-			if (new_pos < desc->file->f_pos) {
+			if (ctx->attr_gencount != nfsi->attr_gencount
+			    || (nfsi->cache_validity & (NFS_INO_INVALID_ATTR|NFS_INO_INVALID_DATA))) {
+				ctx->duped = 0;
+				ctx->attr_gencount = nfsi->attr_gencount;
+			} else if (new_pos < desc->file->f_pos) {
+				if (ctx->duped > 0
+				    && ctx->dup_cookie == *desc->dir_cookie) {
+					if (printk_ratelimit()) {
+						pr_notice("NFS: directory %s/%s contains a readdir loop."
+								"Please contact your server vendor.  "
+								"The file: %s has duplicate cookie %llu\n",
+								desc->file->f_dentry->d_parent->d_name.name,
+								desc->file->f_dentry->d_name.name,
+								array->array[i].string.name,
+								*desc->dir_cookie);
+					}
+					status = -ELOOP;
+					goto out;
+				}
 				ctx->dup_cookie = *desc->dir_cookie;
-				ctx->duped = 1;
+				ctx->duped = -1;
 			}
 			desc->file->f_pos = new_pos;
 			desc->cache_entry_index = i;
···
 		if (*desc->dir_cookie == array->last_cookie)
 			desc->eof = 1;
 	}
+out:
 	return status;
 }
 
···
 	struct nfs_cache_array *array = NULL;
 	struct nfs_open_dir_context *ctx = file->private_data;
 
-	if (ctx->duped != 0 && ctx->dup_cookie == *desc->dir_cookie) {
-		if (printk_ratelimit()) {
-			pr_notice("NFS: directory %s/%s contains a readdir loop.  "
-				"Please contact your server vendor.  "
-				"Offending cookie: %llu\n",
-				file->f_dentry->d_parent->d_name.name,
-				file->f_dentry->d_name.name,
-				*desc->dir_cookie);
-		}
-		res = -ELOOP;
-		goto out;
-	}
-
 	array = nfs_readdir_get_array(desc->page);
 	if (IS_ERR(array)) {
 		res = PTR_ERR(array);
···
 			*desc->dir_cookie = array->array[i+1].cookie;
 		else
 			*desc->dir_cookie = array->last_cookie;
+		if (ctx->duped != 0)
+			ctx->duped = 1;
 	}
 	if (array->eof_index >= 0)
 		desc->eof = 1;
···
 	struct page	*page = NULL;
 	int		status;
 	struct inode *inode = desc->file->f_path.dentry->d_inode;
+	struct nfs_open_dir_context *ctx = desc->file->private_data;
 
 	dfprintk(DIRCACHE, "NFS: uncached_readdir() searching for cookie %Lu\n",
 			(unsigned long long)*desc->dir_cookie);
···
 	desc->page_index = 0;
 	desc->last_cookie = *desc->dir_cookie;
 	desc->page = page;
+	ctx->duped = 0;
 
 	status = nfs_readdir_xdr_to_array(desc, page, inode);
 	if (status < 0)

fs/nfs/nfs4_fs.h

···
 extern const u32 nfs4_fattr_bitmap[2];
 extern const u32 nfs4_statfs_bitmap[2];
 extern const u32 nfs4_pathconf_bitmap[2];
-extern const u32 nfs4_fsinfo_bitmap[2];
+extern const u32 nfs4_fsinfo_bitmap[3];
 extern const u32 nfs4_fs_locations_bitmap[2];
 
 /* nfs4renewd.c */

fs/nfs/nfs4filelayout.c

···
 
 	pnfs_set_layoutcommit(wdata);
 	dprintk("%s ionde %lu pls_end_pos %lu\n", __func__, wdata->inode->i_ino,
-		(unsigned long) wdata->lseg->pls_end_pos);
+		(unsigned long) NFS_I(wdata->inode)->layout->plh_lwb);
 }
 
 /*

fs/nfs/nfs4proc.c

···
 	0
 };
 
-const u32 nfs4_fsinfo_bitmap[2] = { FATTR4_WORD0_MAXFILESIZE
+const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
 			| FATTR4_WORD0_MAXREAD
 			| FATTR4_WORD0_MAXWRITE
 			| FATTR4_WORD0_LEASE_TIME,
 			FATTR4_WORD1_TIME_DELTA
-			| FATTR4_WORD1_FS_LAYOUT_TYPES
+			| FATTR4_WORD1_FS_LAYOUT_TYPES,
+			FATTR4_WORD2_LAYOUT_BLKSIZE
 };
 
 const u32 nfs4_fs_locations_bitmap[2] = {
···
 	return status;
 }
 
+/*
+ * Retrieve the list of Data Server devices from the MDS.
+ */
+static int _nfs4_getdevicelist(struct nfs_server *server,
+				    const struct nfs_fh *fh,
+				    struct pnfs_devicelist *devlist)
+{
+	struct nfs4_getdevicelist_args args = {
+		.fh = fh,
+		.layoutclass = server->pnfs_curr_ld->id,
+	};
+	struct nfs4_getdevicelist_res res = {
+		.devlist = devlist,
+	};
+	struct rpc_message msg = {
+		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICELIST],
+		.rpc_argp = &args,
+		.rpc_resp = &res,
+	};
+	int status;
+
+	dprintk("--> %s\n", __func__);
+	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
+				&res.seq_res, 0);
+	dprintk("<-- %s status=%d\n", __func__, status);
+	return status;
+}
+
+int nfs4_proc_getdevicelist(struct nfs_server *server,
+			    const struct nfs_fh *fh,
+			    struct pnfs_devicelist *devlist)
+{
+	struct nfs4_exception exception = { };
+	int err;
+
+	do {
+		err = nfs4_handle_exception(server,
+				_nfs4_getdevicelist(server, fh, devlist),
+				&exception);
+	} while (exception.retry);
+
+	dprintk("%s: err=%d, num_devs=%u\n", __func__,
+		err, devlist->num_devs);
+
+	return err;
+}
+EXPORT_SYMBOL_GPL(nfs4_proc_getdevicelist);
+
 static int
 _nfs4_proc_getdeviceinfo(struct nfs_server *server, struct pnfs_device *pdev)
 {
···
 static void nfs4_layoutcommit_release(void *calldata)
 {
 	struct nfs4_layoutcommit_data *data = calldata;
+	struct pnfs_layout_segment *lseg, *tmp;
 
+	pnfs_cleanup_layoutcommit(data);
 	/* Matched by references in pnfs_set_layoutcommit */
-	put_lseg(data->lseg);
+	list_for_each_entry_safe(lseg, tmp, &data->lseg_list, pls_lc_list) {
+		list_del_init(&lseg->pls_lc_list);
+		if (test_and_clear_bit(NFS_LSEG_LAYOUTCOMMIT,
+				       &lseg->pls_flags))
+			put_lseg(lseg);
+	}
 	put_rpccred(data->cred);
 	kfree(data);
 }

fs/nfs/nfs4xdr.c

···
 #define encode_restorefh_maxsz  (op_encode_hdr_maxsz)
 #define decode_restorefh_maxsz  (op_decode_hdr_maxsz)
 #define encode_fsinfo_maxsz	(encode_getattr_maxsz)
-#define decode_fsinfo_maxsz	(op_decode_hdr_maxsz + 15)
+/* The 5 accounts for the PNFS attributes, and assumes that at most three
+ * layout types will be returned.
+ */
+#define decode_fsinfo_maxsz	(op_decode_hdr_maxsz + \
+				 nfs4_fattr_bitmap_maxsz + 4 + 8 + 5)
 #define encode_renew_maxsz	(op_encode_hdr_maxsz + 3)
 #define decode_renew_maxsz	(op_decode_hdr_maxsz)
 #define encode_setclientid_maxsz \
···
 				XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + 5)
 #define encode_reclaim_complete_maxsz	(op_encode_hdr_maxsz + 4)
 #define decode_reclaim_complete_maxsz	(op_decode_hdr_maxsz + 4)
+#define encode_getdevicelist_maxsz (op_encode_hdr_maxsz + 4 + \
+				encode_verifier_maxsz)
+#define decode_getdevicelist_maxsz (op_decode_hdr_maxsz + \
+				2 /* nfs_cookie4 gdlr_cookie */ + \
+				decode_verifier_maxsz \
+				  /* verifier4 gdlr_verifier */ + \
+				1 /* gdlr_deviceid_list count */ + \
+				XDR_QUADLEN(NFS4_PNFS_GETDEVLIST_MAXNUM * \
+					    NFS4_DEVICEID4_SIZE) \
+				  /* gdlr_deviceid_list */ + \
+				1 /* bool gdlr_eof */)
 #define encode_getdeviceinfo_maxsz (op_encode_hdr_maxsz + 4 + \
 				XDR_QUADLEN(NFS4_DEVICEID4_SIZE))
 #define decode_getdeviceinfo_maxsz (op_decode_hdr_maxsz + \
···
 #define NFS4_dec_reclaim_complete_sz	(compound_decode_hdr_maxsz + \
 					 decode_sequence_maxsz + \
 					 decode_reclaim_complete_maxsz)
+#define NFS4_enc_getdevicelist_sz (compound_encode_hdr_maxsz + \
+				encode_sequence_maxsz + \
+				encode_putfh_maxsz + \
+				encode_getdevicelist_maxsz)
+#define NFS4_dec_getdevicelist_sz (compound_decode_hdr_maxsz + \
+				decode_sequence_maxsz + \
+				decode_putfh_maxsz + \
+				decode_getdevicelist_maxsz)
 #define NFS4_enc_getdeviceinfo_sz (compound_encode_hdr_maxsz +    \
 				encode_sequence_maxsz +\
 				encode_getdeviceinfo_maxsz)
···
 	hdr->replen += decode_getattr_maxsz;
 }
 
+static void
+encode_getattr_three(struct xdr_stream *xdr,
+		     uint32_t bm0, uint32_t bm1, uint32_t bm2,
+		     struct compound_hdr *hdr)
+{
+	__be32 *p;
+
+	p = reserve_space(xdr, 4);
+	*p = cpu_to_be32(OP_GETATTR);
+	if (bm2) {
+		p = reserve_space(xdr, 16);
+		*p++ = cpu_to_be32(3);
+		*p++ = cpu_to_be32(bm0);
+		*p++ = cpu_to_be32(bm1);
+		*p = cpu_to_be32(bm2);
+	} else if (bm1) {
+		p = reserve_space(xdr, 12);
+		*p++ = cpu_to_be32(2);
+		*p++ = cpu_to_be32(bm0);
+		*p = cpu_to_be32(bm1);
+	} else {
+		p = reserve_space(xdr, 8);
+		*p++ = cpu_to_be32(1);
+		*p = cpu_to_be32(bm0);
+	}
+	hdr->nops++;
+	hdr->replen += decode_getattr_maxsz;
+}
+
 static void encode_getfattr(struct xdr_stream *xdr, const u32* bitmask, struct compound_hdr *hdr)
 {
 	encode_getattr_two(xdr, bitmask[0] & nfs4_fattr_bitmap[0],
···
 
 static void encode_fsinfo(struct xdr_stream *xdr, const u32* bitmask, struct compound_hdr *hdr)
 {
-	encode_getattr_two(xdr, bitmask[0] & nfs4_fsinfo_bitmap[0],
-			   bitmask[1] & nfs4_fsinfo_bitmap[1], hdr);
+	encode_getattr_three(xdr,
+			     bitmask[0] & nfs4_fsinfo_bitmap[0],
+			     bitmask[1] & nfs4_fsinfo_bitmap[1],
+			     bitmask[2] & nfs4_fsinfo_bitmap[2],
+			     hdr);
 }
 
 static void encode_fs_locations(struct xdr_stream *xdr, const u32* bitmask, struct compound_hdr *hdr)
···
 
 #ifdef CONFIG_NFS_V4_1
 static void
+encode_getdevicelist(struct xdr_stream *xdr,
+		     const struct nfs4_getdevicelist_args *args,
+		     struct compound_hdr *hdr)
+{
+	__be32 *p;
+	nfs4_verifier dummy = {
+		.data = "dummmmmy",
+	};
+
+	p = reserve_space(xdr, 20);
+	*p++ = cpu_to_be32(OP_GETDEVICELIST);
+	*p++ = cpu_to_be32(args->layoutclass);
+	*p++ = cpu_to_be32(NFS4_PNFS_GETDEVLIST_MAXNUM);
+	xdr_encode_hyper(p, 0ULL);                          /* cookie */
+	encode_nfs4_verifier(xdr, &dummy);
+	hdr->nops++;
+	hdr->replen += decode_getdevicelist_maxsz;
+}
+
+static void
 encode_getdeviceinfo(struct xdr_stream *xdr,
 		     const struct nfs4_getdeviceinfo_args *args,
 		     struct compound_hdr *hdr)
···
 	*p++ = cpu_to_be32(OP_LAYOUTCOMMIT);
 	/* Only whole file layouts */
 	p = xdr_encode_hyper(p, 0); /* offset */
-	p = xdr_encode_hyper(p, NFS4_MAX_UINT64); /* length */
+	p = xdr_encode_hyper(p, args->lastbytewritten + 1);	/* length */
 	*p++ = cpu_to_be32(0); /* reclaim */
 	p = xdr_encode_opaque_fixed(p, args->stateid.data, NFS4_STATEID_SIZE);
 	*p++ = cpu_to_be32(1); /* newoffset = TRUE */
···
 	struct compound_hdr hdr = {
 		.nops	= 0,
 	};
-	const u32 lease_bitmap[2] = { FATTR4_WORD0_LEASE_TIME, 0 };
+	const u32 lease_bitmap[3] = { FATTR4_WORD0_LEASE_TIME };
 
 	encode_compound_hdr(xdr, req, &hdr);
 	encode_setclientid_confirm(xdr, arg, &hdr);
···
 	struct compound_hdr hdr = {
 		.minorversion = nfs4_xdr_minorversion(&args->la_seq_args),
 	};
-	const u32 lease_bitmap[2] = { FATTR4_WORD0_LEASE_TIME, 0 };
+	const u32 lease_bitmap[3] = { FATTR4_WORD0_LEASE_TIME };
 
 	encode_compound_hdr(xdr, req, &hdr);
 	encode_sequence(xdr, &args->la_seq_args, &hdr);
···
 	encode_compound_hdr(xdr, req, &hdr);
 	encode_sequence(xdr, &args->seq_args, &hdr);
 	encode_reclaim_complete(xdr, args, &hdr);
+	encode_nops(&hdr);
+}
+
+/*
+ * Encode GETDEVICELIST request
+ */
+static void nfs4_xdr_enc_getdevicelist(struct rpc_rqst *req,
+				       struct xdr_stream *xdr,
+				       struct nfs4_getdevicelist_args *args)
+{
+	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);
+	encode_getdevicelist(xdr, args, &hdr);
 	encode_nops(&hdr);
 }
 
···
 		goto out_overflow;
 	bmlen = be32_to_cpup(p);
 
-	bitmap[0] = bitmap[1] = 0;
+	bitmap[0] = bitmap[1] = bitmap[2] = 0;
 	p = xdr_inline_decode(xdr, (bmlen << 2));
 	if (unlikely(!p))
 		goto out_overflow;
 	if (bmlen > 0) {
 		bitmap[0] = be32_to_cpup(p++);
-		if (bmlen > 1)
-			bitmap[1] = be32_to_cpup(p);
+		if (bmlen > 1) {
+			bitmap[1] = be32_to_cpup(p++);
+			if (bmlen > 2)
+				bitmap[2] = be32_to_cpup(p);
+		}
 	}
 	return 0;
 out_overflow:
···
 			return ret;
 		bitmap[0] &= ~FATTR4_WORD0_SUPPORTED_ATTRS;
 	} else
-		bitmask[0] = bitmask[1] = 0;
-	dprintk("%s: bitmask=%08x:%08x\n", __func__, bitmask[0], bitmask[1]);
+		bitmask[0] = bitmask[1] = bitmask[2] = 0;
+	dprintk("%s: bitmask=%08x:%08x:%08x\n", __func__,
+		bitmask[0], bitmask[1], bitmask[2]);
 	return 0;
 }
 
···
 static int decode_server_caps(struct xdr_stream *xdr, struct nfs4_server_caps_res *res)
 {
 	__be32 *savep;
-	uint32_t attrlen, bitmap[2] = {0};
+	uint32_t attrlen, bitmap[3] = {0};
 	int status;
 
 	if ((status = decode_op_hdr(xdr, OP_GETATTR)) != 0)
···
 static int decode_statfs(struct xdr_stream *xdr, struct nfs_fsstat *fsstat)
 {
 	__be32 *savep;
-	uint32_t attrlen, bitmap[2] = {0};
+	uint32_t attrlen, bitmap[3] = {0};
 	int status;
 
 	if ((status = decode_op_hdr(xdr, OP_GETATTR)) != 0)
···
 static int decode_pathconf(struct xdr_stream *xdr, struct nfs_pathconf *pathconf)
 {
 	__be32 *savep;
-	uint32_t attrlen, bitmap[2] = {0};
+	uint32_t attrlen, bitmap[3] = {0};
 	int status;
 
 	if ((status = decode_op_hdr(xdr, OP_GETATTR)) != 0)
···
 {
 	__be32 *savep;
 	uint32_t attrlen,
-		 bitmap[2] = {0};
+		 bitmap[3] = {0};
 	int status;
 
 	status = decode_op_hdr(xdr, OP_GETATTR);
···
 	return status;
 }
 
+/*
+ * The prefered block size for layout directed io
+ */
+static int decode_attr_layout_blksize(struct xdr_stream *xdr, uint32_t *bitmap,
+				      uint32_t *res)
+{
+	__be32 *p;
+
+	dprintk("%s: bitmap is %x\n", __func__, bitmap[2]);
+	*res = 0;
+	if (bitmap[2] & FATTR4_WORD2_LAYOUT_BLKSIZE) {
+		p = xdr_inline_decode(xdr, 4);
+		if (unlikely(!p)) {
+			print_overflow_msg(__func__, xdr);
+			return -EIO;
+		}
+		*res = be32_to_cpup(p);
+		bitmap[2] &= ~FATTR4_WORD2_LAYOUT_BLKSIZE;
+	}
+	return 0;
+}
+
 static int decode_fsinfo(struct xdr_stream *xdr, struct nfs_fsinfo *fsinfo)
 {
 	__be32 *savep;
-	uint32_t attrlen, bitmap[2];
+	uint32_t attrlen, bitmap[3];
 	int status;
 
 	if ((status = decode_op_hdr(xdr, OP_GETATTR)) != 0)
···
 		goto xdr_error;
 	status = decode_attr_pnfstype(xdr, bitmap, &fsinfo->layouttype);
 	if (status != 0)
+		goto xdr_error;
+	status = decode_attr_layout_blksize(xdr, bitmap, &fsinfo->blksize);
+	if (status)
 		goto xdr_error;
 
 	status = verify_attr_len(xdr, savep, attrlen);
···
 {
 	__be32 *savep;
 	uint32_t attrlen,
-		 bitmap[2] = {0};
+		 bitmap[3] = {0};
 	struct kvec *iov = req->rq_rcv_buf.head;
 	int status;
 
···
 }
 
 #if defined(CONFIG_NFS_V4_1)
+/*
+ * TODO: Need to handle case when EOF != true;
+ */
+static int decode_getdevicelist(struct xdr_stream *xdr,
+				struct pnfs_devicelist *res)
+{
+	__be32 *p;
+	int status, i;
+	struct nfs_writeverf verftemp;
+
+	status = decode_op_hdr(xdr, OP_GETDEVICELIST);
+	if (status)
+		return status;
+
+	p = xdr_inline_decode(xdr, 8 + 8 + 4);
+	if (unlikely(!p))
+		goto out_overflow;
+
+	/* TODO: Skip cookie for now */
+	p += 2;
+
+	/* Read verifier */
+	p = xdr_decode_opaque_fixed(p, verftemp.verifier, 8);
+
+	res->num_devs = be32_to_cpup(p);
+
+	dprintk("%s: num_dev %d\n", __func__, res->num_devs);
+
+	if (res->num_devs > NFS4_PNFS_GETDEVLIST_MAXNUM) {
+		printk(KERN_ERR "%s too many result dev_num %u\n",
+				__func__, res->num_devs);
+		return -EIO;
+	}
+
+	p = xdr_inline_decode(xdr,
+			      res->num_devs * NFS4_DEVICEID4_SIZE + 4);
+	if (unlikely(!p))
+		goto out_overflow;
+	for (i = 0; i < res->num_devs; i++)
+		p = xdr_decode_opaque_fixed(p, res->dev_id[i].data,
+					    NFS4_DEVICEID4_SIZE);
+	res->eof = be32_to_cpup(p);
+	return 0;
+out_overflow:
+	print_overflow_msg(__func__, xdr);
+	return -EIO;
+}
 
 static int decode_getdeviceinfo(struct xdr_stream *xdr,
 				struct pnfs_device *pdev)
···
 	int status;
 
 	status = decode_op_hdr(xdr, OP_LAYOUTCOMMIT);
+	res->status = status;
 	if (status)
 		return status;
 
···
 }
 
 /*
+ * Decode GETDEVICELIST response
+ */
+static int nfs4_xdr_dec_getdevicelist(struct rpc_rqst *rqstp,
+				      struct xdr_stream *xdr,
+				      struct nfs4_getdevicelist_res *res)
+{
+	struct compound_hdr hdr;
+	int status;
+
+	dprintk("encoding getdevicelist!\n");
+
+	status = decode_compound_hdr(xdr, &hdr);
+	if (status != 0)
+		goto out;
+	status = decode_sequence(xdr, &res->seq_res, rqstp);
+	if (status != 0)
+		goto out;
+	status = decode_putfh(xdr);
+	if (status != 0)
+		goto out;
+	status = decode_getdevicelist(xdr, res->devlist);
+out:
+	return status;
+}
+
+/*
  * Decode GETDEVINFO response
  */
 static int nfs4_xdr_dec_getdeviceinfo(struct rpc_rqst *rqstp,
···
 int nfs4_decode_dirent(struct xdr_stream *xdr, struct nfs_entry *entry,
 		       int plus)
 {
-	uint32_t bitmap[2] = {0};
+	uint32_t bitmap[3] = {0};
 	uint32_t len;
 	__be32 *p = xdr_inline_decode(xdr, 4);
 	if (unlikely(!p))
···
 	PROC(SECINFO_NO_NAME,	enc_secinfo_no_name,	dec_secinfo_no_name),
 	PROC(TEST_STATEID,	enc_test_stateid,	dec_test_stateid),
 	PROC(FREE_STATEID,	enc_free_stateid,	dec_free_stateid),
+	PROC(GETDEVICELIST,	enc_getdevicelist,	dec_getdevicelist),
 #endif /* CONFIG_NFS_V4_1 */
 };
 

fs/nfs/pnfs.c

···
 void
 unset_pnfs_layoutdriver(struct nfs_server *nfss)
 {
-	if (nfss->pnfs_curr_ld)
+	if (nfss->pnfs_curr_ld) {
+		if (nfss->pnfs_curr_ld->clear_layoutdriver)
+			nfss->pnfs_curr_ld->clear_layoutdriver(nfss);
 		module_put(nfss->pnfs_curr_ld->owner);
+	}
 	nfss->pnfs_curr_ld = NULL;
 }
 
···
  * @id layout type. Zero (illegal layout type) indicates pNFS not in use.
  */
 void
-set_pnfs_layoutdriver(struct nfs_server *server, u32 id)
+set_pnfs_layoutdriver(struct nfs_server *server, const struct nfs_fh *mntfh,
+		      u32 id)
 {
 	struct pnfs_layoutdriver_type *ld_type = NULL;
 
···
 		goto out_no_driver;
 	}
 	server->pnfs_curr_ld = ld_type;
+	if (ld_type->set_layoutdriver
+	    && ld_type->set_layoutdriver(server, mntfh)) {
+		printk(KERN_ERR "%s: Error initializing pNFS layout driver %u.\n",
+				__func__, id);
+		module_put(ld_type->owner);
+		goto out_no_driver;
+	}
 
 	dprintk("%s: pNFS module for %u set\n", __func__, id);
 	return;
···
 pnfs_free_layout_hdr(struct pnfs_layout_hdr *lo)
 {
 	struct pnfs_layoutdriver_type *ld = NFS_SERVER(lo->plh_inode)->pnfs_curr_ld;
+	put_rpccred(lo->plh_lc_cred);
 	return ld->alloc_layout_hdr ? ld->free_layout_hdr(lo) : kfree(lo);
 }
 
···
 init_lseg(struct pnfs_layout_hdr *lo, struct pnfs_layout_segment *lseg)
 {
 	INIT_LIST_HEAD(&lseg->pls_list);
+	INIT_LIST_HEAD(&lseg->pls_lc_list);
 	atomic_set(&lseg->pls_refcount, 1);
 	smp_mb();
 	set_bit(NFS_LSEG_VALID, &lseg->pls_flags);
···
 }
 
 static struct pnfs_layout_hdr *
-alloc_init_layout_hdr(struct inode *ino, gfp_t gfp_flags)
+alloc_init_layout_hdr(struct inode *ino,
+		      struct nfs_open_context *ctx,
+		      gfp_t gfp_flags)
 {
 	struct pnfs_layout_hdr *lo;
 
···
 	INIT_LIST_HEAD(&lo->plh_segs);
 	INIT_LIST_HEAD(&lo->plh_bulk_recall);
 	lo->plh_inode = ino;
+	lo->plh_lc_cred = get_rpccred(ctx->state->owner->so_cred);
 	return lo;
 }
 
 static struct pnfs_layout_hdr *
-pnfs_find_alloc_layout(struct inode *ino, gfp_t gfp_flags)
+pnfs_find_alloc_layout(struct inode *ino,
+		       struct nfs_open_context *ctx,
+		       gfp_t gfp_flags)
 {
 	struct nfs_inode *nfsi = NFS_I(ino);
 	struct pnfs_layout_hdr *new = NULL;
···
 			return nfsi->layout;
 	}
 	spin_unlock(&ino->i_lock);
-	new = alloc_init_layout_hdr(ino, gfp_flags);
+	new = alloc_init_layout_hdr(ino, ctx, gfp_flags);
 	spin_lock(&ino->i_lock);
 
 	if (likely(nfsi->layout == NULL))	/* Won the race? */
···
 	if (!pnfs_enabled_sb(NFS_SERVER(ino)))
 		return NULL;
 	spin_lock(&ino->i_lock);
-	lo = pnfs_find_alloc_layout(ino, gfp_flags);
+	lo = pnfs_find_alloc_layout(ino, ctx, gfp_flags);
 	if (lo == NULL) {
 		dprintk("%s ERROR: can't get pnfs_layout_hdr\n", __func__);
 		goto out_unlock;
···
 EXPORT_SYMBOL_GPL(pnfs_generic_pg_readpages);
 
 /*
- * Currently there is only one (whole file) write lseg.
+ * There can be multiple RW segments.
  */
-static struct pnfs_layout_segment *pnfs_list_write_lseg(struct inode *inode)
+static void pnfs_list_write_lseg(struct inode *inode, struct list_head *listp)
 {
-	struct pnfs_layout_segment *lseg, *rv = NULL;
+	struct pnfs_layout_segment *lseg;
 
-	list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list)
-		if (lseg->pls_range.iomode == IOMODE_RW)
-			rv = lseg;
-	return rv;
+	list_for_each_entry(lseg, &NFS_I(inode)->layout->plh_segs, pls_list) {
+		if (lseg->pls_range.iomode == IOMODE_RW &&
+		    test_bit(NFS_LSEG_LAYOUTCOMMIT, &lseg->pls_flags))
+			list_add(&lseg->pls_lc_list, listp);
+	}
 }
 
 void
···
 
 	spin_lock(&nfsi->vfs_inode.i_lock);
 	if (!test_and_set_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) {
-		/* references matched in nfs4_layoutcommit_release */
-		get_lseg(wdata->lseg);
-		wdata->lseg->pls_lc_cred =
-			get_rpccred(wdata->args.context->state->owner->so_cred);
 		mark_as_dirty = true;
 		dprintk("%s: Set layoutcommit for inode %lu ",
 			__func__, wdata->inode->i_ino);
 	}
-	if (end_pos > wdata->lseg->pls_end_pos)
-		wdata->lseg->pls_end_pos = end_pos;
+	if (!test_and_set_bit(NFS_LSEG_LAYOUTCOMMIT, &wdata->lseg->pls_flags)) {
+		/* references matched in nfs4_layoutcommit_release */
+		get_lseg(wdata->lseg);
+	}
+	if (end_pos > nfsi->layout->plh_lwb)
+		nfsi->layout->plh_lwb = end_pos;
 	spin_unlock(&nfsi->vfs_inode.i_lock);
+	dprintk("%s: lseg %p end_pos %llu\n",
+		__func__, wdata->lseg, nfsi->layout->plh_lwb);
 
 	/* if pnfs_layoutcommit_inode() runs between inode locks, the next one
 	 * will be a noop because NFS_INO_LAYOUTCOMMIT will not be set */
···
 		mark_inode_dirty_sync(wdata->inode);
 }
 EXPORT_SYMBOL_GPL(pnfs_set_layoutcommit);
+
+void pnfs_cleanup_layoutcommit(struct nfs4_layoutcommit_data *data)
+{
+	struct nfs_server *nfss = NFS_SERVER(data->args.inode);
+
+	if (nfss->pnfs_curr_ld->cleanup_layoutcommit)
+		nfss->pnfs_curr_ld->cleanup_layoutcommit(data);
+}
 
 /*
  * For the LAYOUT4_NFSV4_1_FILES layout type, NFS_DATA_SYNC WRITEs and
···
 {
 	struct nfs4_layoutcommit_data *data;
 	struct nfs_inode *nfsi = NFS_I(inode);
-	struct pnfs_layout_segment *lseg;
-	struct rpc_cred *cred;
 	loff_t end_pos;
 	int status = 0;
 
···
 		goto out;
 	}
 
+	INIT_LIST_HEAD(&data->lseg_list);
 	spin_lock(&inode->i_lock);
 	if (!test_and_clear_bit(NFS_INO_LAYOUTCOMMIT, &nfsi->flags)) {
 		spin_unlock(&inode->i_lock);
 		kfree(data);
 		goto out;
 	}
-	/*
-	 * Currently only one (whole file) write lseg which is referenced
-	 * in pnfs_set_layoutcommit and will be found.
-	 */
-	lseg = pnfs_list_write_lseg(inode);
 
-	end_pos = lseg->pls_end_pos;
-	cred = lseg->pls_lc_cred;
-	lseg->pls_end_pos = 0;
-	lseg->pls_lc_cred = NULL;
+	pnfs_list_write_lseg(inode, &data->lseg_list);
+
+	end_pos = nfsi->layout->plh_lwb;
+	nfsi->layout->plh_lwb = 0;
 
 	memcpy(&data->args.stateid.data, nfsi->layout->plh_stateid.data,
 		sizeof(nfsi->layout->plh_stateid.data));
 	spin_unlock(&inode->i_lock);
 
 	data->args.inode = inode;
-	data->lseg = lseg;
-	data->cred = cred;
+	data->cred = get_rpccred(nfsi->layout->plh_lc_cred);
 	nfs_fattr_init(&data->fattr);
 	data->args.bitmask = NFS_SERVER(inode)->cache_consistency_bitmask;
 	data->res.fattr = &data->fattr;

fs/nfs/pnfs.h

···
 enum {
 	NFS_LSEG_VALID = 0,	/* cleared when lseg is recalled/returned */
 	NFS_LSEG_ROC,		/* roc bit received from server */
+	NFS_LSEG_LAYOUTCOMMIT,	/* layoutcommit bit set for layoutcommit */
 };
 
 struct pnfs_layout_segment {
 	struct list_head pls_list;
+	struct list_head pls_lc_list;
 	struct pnfs_layout_range pls_range;
 	atomic_t pls_refcount;
 	unsigned long pls_flags;
 	struct pnfs_layout_hdr *pls_layout;
-	struct rpc_cred	*pls_lc_cred; /* LAYOUTCOMMIT credential */
-	loff_t pls_end_pos; /* LAYOUTCOMMIT write end */
 };
 
 enum pnfs_try_status {
···
 	struct module *owner;
 	unsigned flags;
 
+	int (*set_layoutdriver) (struct nfs_server *, const struct nfs_fh *);
+	int (*clear_layoutdriver) (struct nfs_server *);
+
 	struct pnfs_layout_hdr * (*alloc_layout_hdr) (struct inode *inode, gfp_t gfp_flags);
 	void (*free_layout_hdr) (struct pnfs_layout_hdr *);
 
···
 				     struct xdr_stream *xdr,
 				     const struct nfs4_layoutreturn_args *args);
 
+	void (*cleanup_layoutcommit) (struct nfs4_layoutcommit_data *data);
+
 	void (*encode_layoutcommit) (struct pnfs_layout_hdr *layoutid,
 				     struct xdr_stream *xdr,
 				     const struct nfs4_layoutcommit_args *args);
···
 	unsigned long		plh_block_lgets; /* block LAYOUTGET if >0 */
 	u32			plh_barrier; /* ignore lower seqids */
 	unsigned long		plh_flags;
+	loff_t			plh_lwb; /* last write byte for layoutcommit */
+	struct rpc_cred		*plh_lc_cred; /* layoutcommit cred */
 	struct inode		*plh_inode;
 };
 
···
 	unsigned int  pglen;
 };
 
+#define NFS4_PNFS_GETDEVLIST_MAXNUM 16
+
+struct pnfs_devicelist {
+	unsigned int		eof;
+	unsigned int		num_devs;
+	struct nfs4_deviceid	dev_id[NFS4_PNFS_GETDEVLIST_MAXNUM];
+};
+
 extern int pnfs_register_layoutdriver(struct pnfs_layoutdriver_type *);
 extern void pnfs_unregister_layoutdriver(struct pnfs_layoutdriver_type *);
 
 /* nfs4proc.c */
+extern int nfs4_proc_getdevicelist(struct nfs_server *server,
+				   const struct nfs_fh *fh,
+				   struct pnfs_devicelist *devlist);
 extern int nfs4_proc_getdeviceinfo(struct nfs_server *server,
 				   struct pnfs_device *dev);
 extern int nfs4_proc_layoutget(struct nfs4_layoutget *lgp);
···
 bool pnfs_pageio_init_read(struct nfs_pageio_descriptor *, struct inode *);
 bool pnfs_pageio_init_write(struct nfs_pageio_descriptor *, struct inode *, int);
 
-void set_pnfs_layoutdriver(struct nfs_server *, u32 id);
+void set_pnfs_layoutdriver(struct nfs_server *, const struct nfs_fh *, u32);
 void unset_pnfs_layoutdriver(struct nfs_server *);
 void pnfs_generic_pg_init_read(struct nfs_pageio_descriptor *, struct nfs_page *);
 int pnfs_generic_pg_readpages(struct nfs_pageio_descriptor *desc);
···
 void pnfs_roc_set_barrier(struct inode *ino, u32 barrier);
 bool pnfs_roc_drain(struct inode *ino, u32 *barrier);
 void pnfs_set_layoutcommit(struct nfs_write_data *wdata);
+void pnfs_cleanup_layoutcommit(struct nfs4_layoutcommit_data *data);
 int pnfs_layoutcommit_inode(struct inode *inode, bool sync);
 int _pnfs_return_layout(struct inode *);
 int pnfs_ld_write_done(struct nfs_write_data *);
···
 	return false;
 }
 
-static inline void set_pnfs_layoutdriver(struct nfs_server *s, u32 id)
+static inline void set_pnfs_layoutdriver(struct nfs_server *s,
+					 const struct nfs_fh *mntfh, u32 id);
 {
 }
 

include/linux/nfs.h

···
 #define NFS_MNT_VERSION		1
 #define NFS_MNT3_VERSION	3
 
+#define NFS_PIPE_DIRNAME "/nfs"
+
 /*
  * NFS stats. The good thing with these values is that NFSv3 errors are
  * a superset of NFSv2 errors (with the exception of NFSERR_WFLUSH which

include/linux/nfs4.h

···
 	NFSPROC4_CLNT_SECINFO_NO_NAME,
 	NFSPROC4_CLNT_TEST_STATEID,
 	NFSPROC4_CLNT_FREE_STATEID,
+	NFSPROC4_CLNT_GETDEVICELIST,
 };
 
 /* nfs41 types */

include/linux/nfs_fs.h

···
 
 struct nfs_open_dir_context {
 	struct rpc_cred *cred;
+	unsigned long attr_gencount;
 	__u64 dir_cookie;
 	__u64 dup_cookie;
-	int duped;
+	signed char duped;
 };
 
 /*

include/linux/nfs_fs_sb.h

···
 #endif
 
 #ifdef CONFIG_NFS_V4
-	u32			attr_bitmask[2];/* V4 bitmask representing the set
+	u32			attr_bitmask[3];/* V4 bitmask representing the set
 						   of attributes supported on this
 						   filesystem */
 	u32			cache_consistency_bitmask[2];
···
 						   filesystem */
 	struct pnfs_layoutdriver_type  *pnfs_curr_ld; /* Active layout driver */
 	struct rpc_wait_queue	roc_rpcwaitq;
+	u32			pnfs_blksize;	/* layout_blksize attr */
+	void			*pnfs_ld_data;	/* per mount point data */
 
 	/* the following fields are protected by nfs_client->cl_lock */
 	struct rb_root		state_owners;

include/linux/nfs_xdr.h

···
 	struct timespec		time_delta; /* server time granularity */
 	__u32			lease_time; /* in seconds */
 	__u32			layouttype; /* supported pnfs layout driver */
+	__u32			blksize; /* preferred pnfs io block size */
 };
 
 struct nfs_fsstat {
···
 	gfp_t gfp_flags;
 };
 
+struct nfs4_getdevicelist_args {
+	const struct nfs_fh *fh;
+	u32 layoutclass;
+	struct nfs4_sequence_args seq_args;
+};
+
+struct nfs4_getdevicelist_res {
+	struct pnfs_devicelist *devlist;
+	struct nfs4_sequence_res seq_res;
+};
+
 struct nfs4_getdeviceinfo_args {
 	struct pnfs_device *pdev;
 	struct nfs4_sequence_args seq_args;
···
 	struct nfs_fattr *fattr;
 	const struct nfs_server *server;
 	struct nfs4_sequence_res seq_res;
+	int status;
 };
 
 struct nfs4_layoutcommit_data {
 	struct rpc_task task;
 	struct nfs_fattr fattr;
-	struct pnfs_layout_segment *lseg;
+	struct list_head lseg_list;
 	struct rpc_cred *cred;
 	struct nfs4_layoutcommit_args args;
 	struct nfs4_layoutcommit_res res;
···
 };
 
 struct nfs4_server_caps_res {
-	u32				attr_bitmask[2];
+	u32				attr_bitmask[3];
 	u32				acl_bitmask;
 	u32				has_links;
 	u32				has_symlinks;