diff options
Diffstat (limited to 'fs/nfs/blocklayout')
-rw-r--r-- | fs/nfs/blocklayout/Makefile | 5 | ||||
-rw-r--r-- | fs/nfs/blocklayout/blocklayout.c | 1019 | ||||
-rw-r--r-- | fs/nfs/blocklayout/blocklayout.h | 207 | ||||
-rw-r--r-- | fs/nfs/blocklayout/blocklayoutdev.c | 410 | ||||
-rw-r--r-- | fs/nfs/blocklayout/blocklayoutdm.c | 111 | ||||
-rw-r--r-- | fs/nfs/blocklayout/extents.c | 935 |
6 files changed, 2687 insertions, 0 deletions
diff --git a/fs/nfs/blocklayout/Makefile b/fs/nfs/blocklayout/Makefile new file mode 100644 index 0000000..d581550 --- /dev/null +++ b/fs/nfs/blocklayout/Makefile @@ -0,0 +1,5 @@ +# +# 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 diff --git a/fs/nfs/blocklayout/blocklayout.c b/fs/nfs/blocklayout/blocklayout.c new file mode 100644 index 0000000..e56564d --- /dev/null +++ b/fs/nfs/blocklayout/blocklayout.c @@ -0,0 +1,1019 @@ +/* + * 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 <andros@citi.umich.edu> + * Fred Isaman <iisaman@umich.edu> + * + * 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 <andros@citi.umich.edu>"); +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); diff --git a/fs/nfs/blocklayout/blocklayout.h b/fs/nfs/blocklayout/blocklayout.h new file mode 100644 index 0000000..f27d827 --- /dev/null +++ b/fs/nfs/blocklayout/blocklayout.h @@ -0,0 +1,207 @@ +/* + * 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 <andros@citi.umich.edu> + * Fred Isaman <iisaman@umich.edu> + * + * 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 */ diff --git a/fs/nfs/blocklayout/blocklayoutdev.c b/fs/nfs/blocklayout/blocklayoutdev.c new file mode 100644 index 0000000..a83b393 --- /dev/null +++ b/fs/nfs/blocklayout/blocklayoutdev.c @@ -0,0 +1,410 @@ +/* + * 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 <andros@citi.umich.edu> + * Fred Isaman <iisaman@umich.edu> + * + * 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; +} diff --git a/fs/nfs/blocklayout/blocklayoutdm.c b/fs/nfs/blocklayout/blocklayoutdm.c new file mode 100644 index 0000000..d055c75 --- /dev/null +++ b/fs/nfs/blocklayout/blocklayoutdm.c @@ -0,0 +1,111 @@ +/* + * 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 <iisaman@umich.edu> + * Andy Adamson <andros@citi.umich.edu> + * + * 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); + } +} diff --git a/fs/nfs/blocklayout/extents.c b/fs/nfs/blocklayout/extents.c new file mode 100644 index 0000000..19fa7b0 --- /dev/null +++ b/fs/nfs/blocklayout/extents.c @@ -0,0 +1,935 @@ +/* + * 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 <andros@citi.umich.edu> + * Fred Isaman <iisaman@umich.edu> + * + * 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); + } + } +} |