diff options
Diffstat (limited to 'fs/exofs/inode.c')
| -rw-r--r-- | fs/exofs/inode.c | 827 |
1 files changed, 520 insertions, 307 deletions
diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c index 2afbcebeda7..3f9cafd7393 100644 --- a/fs/exofs/inode.c +++ b/fs/exofs/inode.c @@ -31,29 +31,36 @@ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ -#include <linux/writeback.h> -#include <linux/buffer_head.h> -#include <scsi/scsi_device.h> +#include <linux/slab.h> #include "exofs.h" #define EXOFS_DBGMSG2(M...) do {} while (0) -enum { BIO_MAX_PAGES_KMALLOC = - (PAGE_SIZE - sizeof(struct bio)) / sizeof(struct bio_vec), -}; +unsigned exofs_max_io_pages(struct ore_layout *layout, + unsigned expected_pages) +{ + unsigned pages = min_t(unsigned, expected_pages, + layout->max_io_length / PAGE_SIZE); + + return pages; +} struct page_collect { struct exofs_sb_info *sbi; - struct request_queue *req_q; struct inode *inode; unsigned expected_pages; - struct exofs_io_state *ios; + struct ore_io_state *ios; - struct bio *bio; + struct page **pages; + unsigned alloc_pages; unsigned nr_pages; unsigned long length; loff_t pg_first; /* keep 64bit also in 32-arches */ + bool read_4_write; /* This means two things: that the read is sync + * And the pages should not be unlocked. + */ + struct page *that_locked_page; }; static void _pcol_init(struct page_collect *pcol, unsigned expected_pages, @@ -62,69 +69,67 @@ static void _pcol_init(struct page_collect *pcol, unsigned expected_pages, struct exofs_sb_info *sbi = inode->i_sb->s_fs_info; pcol->sbi = sbi; - /* Create master bios on first Q, later on cloning, each clone will be - * allocated on it's destination Q - */ - pcol->req_q = osd_request_queue(sbi->s_ods[0]); pcol->inode = inode; pcol->expected_pages = expected_pages; pcol->ios = NULL; - pcol->bio = NULL; + pcol->pages = NULL; + pcol->alloc_pages = 0; pcol->nr_pages = 0; pcol->length = 0; pcol->pg_first = -1; + pcol->read_4_write = false; + pcol->that_locked_page = NULL; } static void _pcol_reset(struct page_collect *pcol) { pcol->expected_pages -= min(pcol->nr_pages, pcol->expected_pages); - pcol->bio = NULL; + pcol->pages = NULL; + pcol->alloc_pages = 0; pcol->nr_pages = 0; pcol->length = 0; pcol->pg_first = -1; pcol->ios = NULL; + pcol->that_locked_page = NULL; /* this is probably the end of the loop but in writes * it might not end here. don't be left with nothing */ if (!pcol->expected_pages) - pcol->expected_pages = BIO_MAX_PAGES_KMALLOC; + pcol->expected_pages = + exofs_max_io_pages(&pcol->sbi->layout, ~0); } static int pcol_try_alloc(struct page_collect *pcol) { - int pages = min_t(unsigned, pcol->expected_pages, - BIO_MAX_PAGES_KMALLOC); - - if (!pcol->ios) { /* First time allocate io_state */ - int ret = exofs_get_io_state(pcol->sbi, &pcol->ios); + unsigned pages; - if (ret) - return ret; - } + /* TODO: easily support bio chaining */ + pages = exofs_max_io_pages(&pcol->sbi->layout, pcol->expected_pages); for (; pages; pages >>= 1) { - pcol->bio = bio_kmalloc(GFP_KERNEL, pages); - if (likely(pcol->bio)) + pcol->pages = kmalloc(pages * sizeof(struct page *), + GFP_KERNEL); + if (likely(pcol->pages)) { + pcol->alloc_pages = pages; return 0; + } } - EXOFS_ERR("Failed to bio_kmalloc expected_pages=%u\n", + EXOFS_ERR("Failed to kmalloc expected_pages=%u\n", pcol->expected_pages); return -ENOMEM; } static void pcol_free(struct page_collect *pcol) { - if (pcol->bio) { - bio_put(pcol->bio); - pcol->bio = NULL; - } + kfree(pcol->pages); + pcol->pages = NULL; if (pcol->ios) { - exofs_put_io_state(pcol->ios); + ore_put_io_state(pcol->ios); pcol->ios = NULL; } } @@ -132,23 +137,25 @@ static void pcol_free(struct page_collect *pcol) static int pcol_add_page(struct page_collect *pcol, struct page *page, unsigned len) { - int added_len = bio_add_pc_page(pcol->req_q, pcol->bio, page, len, 0); - if (unlikely(len != added_len)) + if (unlikely(pcol->nr_pages >= pcol->alloc_pages)) return -ENOMEM; - ++pcol->nr_pages; + pcol->pages[pcol->nr_pages++] = page; pcol->length += len; return 0; } +enum {PAGE_WAS_NOT_IN_IO = 17}; static int update_read_page(struct page *page, int ret) { - if (ret == 0) { + switch (ret) { + case 0: /* Everything is OK */ SetPageUptodate(page); if (PageError(page)) ClearPageError(page); - } else if (ret == -EFAULT) { + break; + case -EFAULT: /* In this case we were trying to read something that wasn't on * disk yet - return a page full of zeroes. This should be OK, * because the object should be empty (if there was a write @@ -159,16 +166,22 @@ static int update_read_page(struct page *page, int ret) SetPageUptodate(page); if (PageError(page)) ClearPageError(page); - ret = 0; /* recovered error */ EXOFS_DBGMSG("recovered read error\n"); - } else /* Error */ + /* fall through */ + case PAGE_WAS_NOT_IN_IO: + ret = 0; /* recovered error */ + break; + default: SetPageError(page); - + } return ret; } static void update_write_page(struct page *page, int ret) { + if (unlikely(ret == PAGE_WAS_NOT_IN_IO)) + return; /* don't pass start don't collect $200 */ + if (ret) { mapping_set_error(page->mapping, ret); SetPageError(page); @@ -179,27 +192,27 @@ static void update_write_page(struct page *page, int ret) /* Called at the end of reads, to optionally unlock pages and update their * status. */ -static int __readpages_done(struct page_collect *pcol, bool do_unlock) +static int __readpages_done(struct page_collect *pcol) { - struct bio_vec *bvec; int i; - u64 resid; u64 good_bytes; u64 length = 0; - int ret = exofs_check_io(pcol->ios, &resid); + int ret = ore_check_io(pcol->ios, NULL); - if (likely(!ret)) + if (likely(!ret)) { good_bytes = pcol->length; - else - good_bytes = pcol->length - resid; + ret = PAGE_WAS_NOT_IN_IO; + } else { + good_bytes = 0; + } - EXOFS_DBGMSG("readpages_done(0x%lx) good_bytes=0x%llx" + EXOFS_DBGMSG2("readpages_done(0x%lx) good_bytes=0x%llx" " length=0x%lx nr_pages=%u\n", pcol->inode->i_ino, _LLU(good_bytes), pcol->length, pcol->nr_pages); - __bio_for_each_segment(bvec, pcol->bio, i, 0) { - struct page *page = bvec->bv_page; + for (i = 0; i < pcol->nr_pages; i++) { + struct page *page = pcol->pages[i]; struct inode *inode = page->mapping->host; int page_stat; @@ -216,33 +229,32 @@ static int __readpages_done(struct page_collect *pcol, bool do_unlock) page_stat ? "bad_bytes" : "good_bytes"); ret = update_read_page(page, page_stat); - if (do_unlock) + if (!pcol->read_4_write) unlock_page(page); - length += bvec->bv_len; + length += PAGE_SIZE; } pcol_free(pcol); - EXOFS_DBGMSG("readpages_done END\n"); + EXOFS_DBGMSG2("readpages_done END\n"); return ret; } /* callback of async reads */ -static void readpages_done(struct exofs_io_state *ios, void *p) +static void readpages_done(struct ore_io_state *ios, void *p) { struct page_collect *pcol = p; - __readpages_done(pcol, true); + __readpages_done(pcol); atomic_dec(&pcol->sbi->s_curr_pending); kfree(pcol); } static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw) { - struct bio_vec *bvec; int i; - __bio_for_each_segment(bvec, pcol->bio, i, 0) { - struct page *page = bvec->bv_page; + for (i = 0; i < pcol->nr_pages; i++) { + struct page *page = pcol->pages[i]; if (rw == READ) update_read_page(page, ret); @@ -253,26 +265,71 @@ static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw) } } -static int read_exec(struct page_collect *pcol, bool is_sync) +static int _maybe_not_all_in_one_io(struct ore_io_state *ios, + struct page_collect *pcol_src, struct page_collect *pcol) +{ + /* length was wrong or offset was not page aligned */ + BUG_ON(pcol_src->nr_pages < ios->nr_pages); + + if (pcol_src->nr_pages > ios->nr_pages) { + struct page **src_page; + unsigned pages_less = pcol_src->nr_pages - ios->nr_pages; + unsigned long len_less = pcol_src->length - ios->length; + unsigned i; + int ret; + + /* This IO was trimmed */ + pcol_src->nr_pages = ios->nr_pages; + pcol_src->length = ios->length; + + /* Left over pages are passed to the next io */ + pcol->expected_pages += pages_less; + pcol->nr_pages = pages_less; + pcol->length = len_less; + src_page = pcol_src->pages + pcol_src->nr_pages; + pcol->pg_first = (*src_page)->index; + + ret = pcol_try_alloc(pcol); + if (unlikely(ret)) + return ret; + + for (i = 0; i < pages_less; ++i) + pcol->pages[i] = *src_page++; + + EXOFS_DBGMSG("Length was adjusted nr_pages=0x%x " + "pages_less=0x%x expected_pages=0x%x " + "next_offset=0x%llx next_len=0x%lx\n", + pcol_src->nr_pages, pages_less, pcol->expected_pages, + pcol->pg_first * PAGE_SIZE, pcol->length); + } + return 0; +} + +static int read_exec(struct page_collect *pcol) { struct exofs_i_info *oi = exofs_i(pcol->inode); - struct exofs_io_state *ios = pcol->ios; + struct ore_io_state *ios; struct page_collect *pcol_copy = NULL; int ret; - if (!pcol->bio) + if (!pcol->pages) return 0; - /* see comment in _readpage() about sync reads */ - WARN_ON(is_sync && (pcol->nr_pages != 1)); + if (!pcol->ios) { + int ret = ore_get_rw_state(&pcol->sbi->layout, &oi->oc, true, + pcol->pg_first << PAGE_CACHE_SHIFT, + pcol->length, &pcol->ios); + + if (ret) + return ret; + } - ios->bio = pcol->bio; - ios->length = pcol->length; - ios->offset = pcol->pg_first << PAGE_CACHE_SHIFT; + ios = pcol->ios; + ios->pages = pcol->pages; - if (is_sync) { - exofs_oi_read(oi, pcol->ios); - return __readpages_done(pcol, false); + if (pcol->read_4_write) { + ore_read(pcol->ios); + return __readpages_done(pcol); } pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL); @@ -284,26 +341,32 @@ static int read_exec(struct page_collect *pcol, bool is_sync) *pcol_copy = *pcol; ios->done = readpages_done; ios->private = pcol_copy; - ret = exofs_oi_read(oi, ios); + + /* pages ownership was passed to pcol_copy */ + _pcol_reset(pcol); + + ret = _maybe_not_all_in_one_io(ios, pcol_copy, pcol); if (unlikely(ret)) goto err; - atomic_inc(&pcol->sbi->s_curr_pending); + EXOFS_DBGMSG2("read_exec(0x%lx) offset=0x%llx length=0x%llx\n", + pcol->inode->i_ino, _LLU(ios->offset), _LLU(ios->length)); - EXOFS_DBGMSG("read_exec obj=0x%llx start=0x%llx length=0x%lx\n", - ios->obj.id, _LLU(ios->offset), pcol->length); + ret = ore_read(ios); + if (unlikely(ret)) + goto err; + + atomic_inc(&pcol->sbi->s_curr_pending); - /* pages ownership was passed to pcol_copy */ - _pcol_reset(pcol); return 0; err: - if (!is_sync) - _unlock_pcol_pages(pcol, ret, READ); - - pcol_free(pcol); - + if (!pcol_copy) /* Failed before ownership transfer */ + pcol_copy = pcol; + _unlock_pcol_pages(pcol_copy, ret, READ); + pcol_free(pcol_copy); kfree(pcol_copy); + return ret; } @@ -324,11 +387,15 @@ static int readpage_strip(void *data, struct page *page) size_t len; int ret; + BUG_ON(!PageLocked(page)); + /* FIXME: Just for debugging, will be removed */ if (PageUptodate(page)) EXOFS_ERR("PageUptodate(0x%lx, 0x%lx)\n", pcol->inode->i_ino, page->index); + pcol->that_locked_page = page; + if (page->index < end_index) len = PAGE_CACHE_SIZE; else if (page->index == end_index) @@ -346,11 +413,14 @@ static int readpage_strip(void *data, struct page *page) if (PageError(page)) ClearPageError(page); - unlock_page(page); - EXOFS_DBGMSG("readpage_strip(0x%lx, 0x%lx) empty page," - " splitting\n", inode->i_ino, page->index); + if (!pcol->read_4_write) + unlock_page(page); + EXOFS_DBGMSG("readpage_strip(0x%lx) empty page len=%zx " + "read_4_write=%d index=0x%lx end_index=0x%lx " + "splitting\n", inode->i_ino, len, + pcol->read_4_write, page->index, end_index); - return read_exec(pcol, false); + return read_exec(pcol); } try_again: @@ -360,13 +430,13 @@ try_again: } else if (unlikely((pcol->pg_first + pcol->nr_pages) != page->index)) { /* Discontinuity detected, split the request */ - ret = read_exec(pcol, false); + ret = read_exec(pcol); if (unlikely(ret)) goto fail; goto try_again; } - if (!pcol->bio) { + if (!pcol->pages) { ret = pcol_try_alloc(pcol); if (unlikely(ret)) goto fail; @@ -385,7 +455,7 @@ try_again: page, len, pcol->nr_pages, pcol->length); /* split the request, and start again with current page */ - ret = read_exec(pcol, false); + ret = read_exec(pcol); if (unlikely(ret)) goto fail; @@ -414,26 +484,28 @@ static int exofs_readpages(struct file *file, struct address_space *mapping, return ret; } - return read_exec(&pcol, false); + ret = read_exec(&pcol); + if (unlikely(ret)) + return ret; + + return read_exec(&pcol); } -static int _readpage(struct page *page, bool is_sync) +static int _readpage(struct page *page, bool read_4_write) { struct page_collect pcol; int ret; _pcol_init(&pcol, 1, page->mapping->host); - /* readpage_strip might call read_exec(,is_sync==false) at several - * places but not if we have a single page. - */ + pcol.read_4_write = read_4_write; ret = readpage_strip(&pcol, page); if (ret) { EXOFS_ERR("_readpage => %d\n", ret); return ret; } - return read_exec(&pcol, is_sync); + return read_exec(&pcol); } /* @@ -445,30 +517,30 @@ static int exofs_readpage(struct file *file, struct page *page) } /* Callback for osd_write. All writes are asynchronous */ -static void writepages_done(struct exofs_io_state *ios, void *p) +static void writepages_done(struct ore_io_state *ios, void *p) { struct page_collect *pcol = p; - struct bio_vec *bvec; int i; - u64 resid; u64 good_bytes; u64 length = 0; - int ret = exofs_check_io(ios, &resid); + int ret = ore_check_io(ios, NULL); atomic_dec(&pcol->sbi->s_curr_pending); - if (likely(!ret)) + if (likely(!ret)) { good_bytes = pcol->length; - else - good_bytes = pcol->length - resid; + ret = PAGE_WAS_NOT_IN_IO; + } else { + good_bytes = 0; + } - EXOFS_DBGMSG("writepages_done(0x%lx) good_bytes=0x%llx" + EXOFS_DBGMSG2("writepages_done(0x%lx) good_bytes=0x%llx" " length=0x%lx nr_pages=%u\n", pcol->inode->i_ino, _LLU(good_bytes), pcol->length, pcol->nr_pages); - __bio_for_each_segment(bvec, pcol->bio, i, 0) { - struct page *page = bvec->bv_page; + for (i = 0; i < pcol->nr_pages; i++) { + struct page *page = pcol->pages[i]; struct inode *inode = page->mapping->host; int page_stat; @@ -485,58 +557,129 @@ static void writepages_done(struct exofs_io_state *ios, void *p) EXOFS_DBGMSG2(" writepages_done(0x%lx, 0x%lx) status=%d\n", inode->i_ino, page->index, page_stat); - length += bvec->bv_len; + length += PAGE_SIZE; } pcol_free(pcol); kfree(pcol); - EXOFS_DBGMSG("writepages_done END\n"); + EXOFS_DBGMSG2("writepages_done END\n"); } +static struct page *__r4w_get_page(void *priv, u64 offset, bool *uptodate) +{ + struct page_collect *pcol = priv; + pgoff_t index = offset / PAGE_SIZE; + + if (!pcol->that_locked_page || + (pcol->that_locked_page->index != index)) { + struct page *page; + loff_t i_size = i_size_read(pcol->inode); + + if (offset >= i_size) { + *uptodate = true; + EXOFS_DBGMSG2("offset >= i_size index=0x%lx\n", index); + return ZERO_PAGE(0); + } + + page = find_get_page(pcol->inode->i_mapping, index); + if (!page) { + page = find_or_create_page(pcol->inode->i_mapping, + index, GFP_NOFS); + if (unlikely(!page)) { + EXOFS_DBGMSG("grab_cache_page Failed " + "index=0x%llx\n", _LLU(index)); + return NULL; + } + unlock_page(page); + } + if (PageDirty(page) || PageWriteback(page)) + *uptodate = true; + else + *uptodate = PageUptodate(page); + EXOFS_DBGMSG2("index=0x%lx uptodate=%d\n", index, *uptodate); + return page; + } else { + EXOFS_DBGMSG2("YES that_locked_page index=0x%lx\n", + pcol->that_locked_page->index); + *uptodate = true; + return pcol->that_locked_page; + } +} + +static void __r4w_put_page(void *priv, struct page *page) +{ + struct page_collect *pcol = priv; + + if ((pcol->that_locked_page != page) && (ZERO_PAGE(0) != page)) { + EXOFS_DBGMSG2("index=0x%lx\n", page->index); + page_cache_release(page); + return; + } + EXOFS_DBGMSG2("that_locked_page index=0x%lx\n", + ZERO_PAGE(0) == page ? -1 : page->index); +} + +static const struct _ore_r4w_op _r4w_op = { + .get_page = &__r4w_get_page, + .put_page = &__r4w_put_page, +}; + static int write_exec(struct page_collect *pcol) { struct exofs_i_info *oi = exofs_i(pcol->inode); - struct exofs_io_state *ios = pcol->ios; + struct ore_io_state *ios; struct page_collect *pcol_copy = NULL; int ret; - if (!pcol->bio) + if (!pcol->pages) return 0; + BUG_ON(pcol->ios); + ret = ore_get_rw_state(&pcol->sbi->layout, &oi->oc, false, + pcol->pg_first << PAGE_CACHE_SHIFT, + pcol->length, &pcol->ios); + if (unlikely(ret)) + goto err; + pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL); if (!pcol_copy) { - EXOFS_ERR("write_exec: Faild to kmalloc(pcol)\n"); + EXOFS_ERR("write_exec: Failed to kmalloc(pcol)\n"); ret = -ENOMEM; goto err; } *pcol_copy = *pcol; - pcol_copy->bio->bi_rw |= (1 << BIO_RW); /* FIXME: bio_set_dir() */ - - ios->bio = pcol_copy->bio; - ios->offset = pcol_copy->pg_first << PAGE_CACHE_SHIFT; - ios->length = pcol_copy->length; + ios = pcol->ios; + ios->pages = pcol_copy->pages; ios->done = writepages_done; + ios->r4w = &_r4w_op; ios->private = pcol_copy; - ret = exofs_oi_write(oi, ios); + /* pages ownership was passed to pcol_copy */ + _pcol_reset(pcol); + + ret = _maybe_not_all_in_one_io(ios, pcol_copy, pcol); + if (unlikely(ret)) + goto err; + + EXOFS_DBGMSG2("write_exec(0x%lx) offset=0x%llx length=0x%llx\n", + pcol->inode->i_ino, _LLU(ios->offset), _LLU(ios->length)); + + ret = ore_write(ios); if (unlikely(ret)) { - EXOFS_ERR("write_exec: exofs_oi_write() Faild\n"); + EXOFS_ERR("write_exec: ore_write() Failed\n"); goto err; } atomic_inc(&pcol->sbi->s_curr_pending); - EXOFS_DBGMSG("write_exec(0x%lx, 0x%llx) start=0x%llx length=0x%lx\n", - pcol->inode->i_ino, pcol->pg_first, _LLU(ios->offset), - pcol->length); - /* pages ownership was passed to pcol_copy */ - _pcol_reset(pcol); return 0; err: - _unlock_pcol_pages(pcol, ret, WRITE); - pcol_free(pcol); + if (!pcol_copy) /* Failed before ownership transfer */ + pcol_copy = pcol; + _unlock_pcol_pages(pcol_copy, ret, WRITE); + pcol_free(pcol_copy); kfree(pcol_copy); return ret; @@ -605,7 +748,7 @@ try_again: goto try_again; } - if (!pcol->bio) { + if (!pcol->pages) { ret = pcol_try_alloc(pcol); if (unlikely(ret)) goto fail; @@ -616,14 +759,14 @@ try_again: ret = pcol_add_page(pcol, page, len); if (unlikely(ret)) { - EXOFS_DBGMSG("Failed pcol_add_page " + EXOFS_DBGMSG2("Failed pcol_add_page " "nr_pages=%u total_length=0x%lx\n", pcol->nr_pages, pcol->length); /* split the request, next loop will start again */ ret = write_exec(pcol); if (unlikely(ret)) { - EXOFS_DBGMSG("write_exec faild => %d", ret); + EXOFS_DBGMSG("write_exec failed => %d", ret); goto fail; } @@ -663,7 +806,7 @@ static int exofs_writepages(struct address_space *mapping, if (expected_pages < 32L) expected_pages = 32L; - EXOFS_DBGMSG("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx " + EXOFS_DBGMSG2("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx " "nrpages=%lu start=0x%lx end=0x%lx expected_pages=%ld\n", mapping->host->i_ino, wbc->range_start, wbc->range_end, mapping->nrpages, start, end, expected_pages); @@ -671,14 +814,33 @@ static int exofs_writepages(struct address_space *mapping, _pcol_init(&pcol, expected_pages, mapping->host); ret = write_cache_pages(mapping, wbc, writepage_strip, &pcol); - if (ret) { + if (unlikely(ret)) { EXOFS_ERR("write_cache_pages => %d\n", ret); return ret; } - return write_exec(&pcol); + ret = write_exec(&pcol); + if (unlikely(ret)) + return ret; + + if (wbc->sync_mode == WB_SYNC_ALL) { + return write_exec(&pcol); /* pump the last reminder */ + } else if (pcol.nr_pages) { + /* not SYNC let the reminder join the next writeout */ + unsigned i; + + for (i = 0; i < pcol.nr_pages; i++) { + struct page *page = pcol.pages[i]; + + end_page_writeback(page); + set_page_dirty(page); + unlock_page(page); + } + } + return 0; } +/* static int exofs_writepage(struct page *page, struct writeback_control *wbc) { struct page_collect pcol; @@ -694,6 +856,13 @@ static int exofs_writepage(struct page *page, struct writeback_control *wbc) return write_exec(&pcol); } +*/ +/* i_mutex held using inode->i_size directly */ +static void _write_failed(struct inode *inode, loff_t to) +{ + if (to > inode->i_size) + truncate_pagecache(inode, inode->i_size); +} int exofs_write_begin(struct file *file, struct address_space *mapping, loff_t pos, unsigned len, unsigned flags, @@ -707,8 +876,8 @@ int exofs_write_begin(struct file *file, struct address_space *mapping, ret = simple_write_begin(file, mapping, pos, len, flags, pagep, fsdata); if (ret) { - EXOFS_DBGMSG("simple_write_begin faild\n"); - return ret; + EXOFS_DBGMSG("simple_write_begin failed\n"); + goto out; } page = *pagep; @@ -716,13 +885,33 @@ int exofs_write_begin(struct file *file, struct address_space *mapping, /* read modify write */ if (!PageUptodate(page) && (len != PAGE_CACHE_SIZE)) { + loff_t i_size = i_size_read(mapping->host); + pgoff_t end_index = i_size >> PAGE_CACHE_SHIFT; + size_t rlen; + + if (page->index < end_index) + rlen = PAGE_CACHE_SIZE; + else if (page->index == end_index) + rlen = i_size & ~PAGE_CACHE_MASK; + else + rlen = 0; + + if (!rlen) { + clear_highpage(page); + SetPageUptodate(page); + goto out; + } + ret = _readpage(page, true); if (ret) { /*SetPageError was done by _readpage. Is it ok?*/ unlock_page(page); - EXOFS_DBGMSG("__readpage_filler faild\n"); + EXOFS_DBGMSG("__readpage failed\n"); } } +out: + if (unlikely(ret)) + _write_failed(mapping->host, pos + len); return ret; } @@ -748,18 +937,59 @@ static int exofs_write_end(struct file *file, struct address_space *mapping, int ret; ret = simple_write_end(file, mapping,pos, len, copied, page, fsdata); + if (unlikely(ret)) + _write_failed(inode, pos + len); + + /* TODO: once simple_write_end marks inode dirty remove */ if (i_size != inode->i_size) mark_inode_dirty(inode); return ret; } +static int exofs_releasepage(struct page *page, gfp_t gfp) +{ + EXOFS_DBGMSG("page 0x%lx\n", page->index); + WARN_ON(1); + return 0; +} + +static void exofs_invalidatepage(struct page *page, unsigned int offset, + unsigned int length) +{ + EXOFS_DBGMSG("page 0x%lx offset 0x%x length 0x%x\n", + page->index, offset, length); + WARN_ON(1); +} + + + /* TODO: Should be easy enough to do proprly */ +static ssize_t exofs_direct_IO(int rw, struct kiocb *iocb, + struct iov_iter *iter, loff_t offset) +{ + return 0; +} + const struct address_space_operations exofs_aops = { .readpage = exofs_readpage, .readpages = exofs_readpages, - .writepage = exofs_writepage, + .writepage = NULL, .writepages = exofs_writepages, .write_begin = exofs_write_begin_export, .write_end = exofs_write_end, + .releasepage = exofs_releasepage, + .set_page_dirty = __set_page_dirty_nobuffers, + .invalidatepage = exofs_invalidatepage, + + /* Not implemented Yet */ + .bmap = NULL, /* TODO: use osd's OSD_ACT_READ_MAP */ + .direct_IO = exofs_direct_IO, + + /* With these NULL has special meaning or default is not exported */ + .get_xip_mem = NULL, + .migratepage = NULL, + .launder_page = NULL, + .is_partially_uptodate = NULL, + .error_remove_page = NULL, }; /****************************************************************************** @@ -776,124 +1006,109 @@ static inline int exofs_inode_is_fast_symlink(struct inode *inode) return S_ISLNK(inode->i_mode) && (oi->i_data[0] != 0); } -/* - * get_block_t - Fill in a buffer_head - * An OSD takes care of block allocation so we just fake an allocation by - * putting in the inode's sector_t in the buffer_head. - * TODO: What about the case of create==0 and @iblock does not exist in the - * object? - */ -static int exofs_get_block(struct inode *inode, sector_t iblock, - struct buffer_head *bh_result, int create) -{ - map_bh(bh_result, inode->i_sb, iblock); - return 0; -} - -const struct osd_attr g_attr_logical_length = ATTR_DEF( - OSD_APAGE_OBJECT_INFORMATION, OSD_ATTR_OI_LOGICAL_LENGTH, 8); - -static int _do_truncate(struct inode *inode) +static int _do_truncate(struct inode *inode, loff_t newsize) { struct exofs_i_info *oi = exofs_i(inode); - loff_t isize = i_size_read(inode); + struct exofs_sb_info *sbi = inode->i_sb->s_fs_info; int ret; inode->i_mtime = inode->i_ctime = CURRENT_TIME; - nobh_truncate_page(inode->i_mapping, isize, exofs_get_block); + ret = ore_truncate(&sbi->layout, &oi->oc, (u64)newsize); + if (likely(!ret)) + truncate_setsize(inode, newsize); - ret = exofs_oi_truncate(oi, (u64)isize); - EXOFS_DBGMSG("(0x%lx) size=0x%llx\n", inode->i_ino, isize); + EXOFS_DBGMSG2("(0x%lx) size=0x%llx ret=>%d\n", + inode->i_ino, newsize, ret); return ret; } /* - * Truncate a file to the specified size - all we have to do is set the size - * attribute. We make sure the object exists first. - */ -void exofs_truncate(struct inode *inode) -{ - struct exofs_i_info *oi = exofs_i(inode); - int ret; - - if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) - || S_ISLNK(inode->i_mode))) - return; - if (exofs_inode_is_fast_symlink(inode)) - return; - if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) - return; - - /* if we are about to truncate an object, and it hasn't been - * created yet, wait - */ - if (unlikely(wait_obj_created(oi))) - goto fail; - - ret = _do_truncate(inode); - if (ret) - goto fail; - -out: - mark_inode_dirty(inode); - return; -fail: - make_bad_inode(inode); - goto out; -} - -/* - * Set inode attributes - just call generic functions. + * Set inode attributes - update size attribute on OSD if needed, + * otherwise just call generic functions. */ int exofs_setattr(struct dentry *dentry, struct iattr *iattr) { struct inode *inode = dentry->d_inode; int error; + /* if we are about to modify an object, and it hasn't been + * created yet, wait + */ + error = wait_obj_created(exofs_i(inode)); + if (unlikely(error)) + return error; + error = inode_change_ok(inode, iattr); - if (error) + if (unlikely(error)) return error; - error = inode_setattr(inode, iattr); - return error; + if ((iattr->ia_valid & ATTR_SIZE) && + iattr->ia_size != i_size_read(inode)) { + error = _do_truncate(inode, iattr->ia_size); + if (unlikely(error)) + return error; + } + + setattr_copy(inode, iattr); + mark_inode_dirty(inode); + return 0; } +static const struct osd_attr g_attr_inode_file_layout = ATTR_DEF( + EXOFS_APAGE_FS_DATA, + EXOFS_ATTR_INODE_FILE_LAYOUT, + 0); +static const struct osd_attr g_attr_inode_dir_layout = ATTR_DEF( + EXOFS_APAGE_FS_DATA, + EXOFS_ATTR_INODE_DIR_LAYOUT, + 0); + /* - * Read an inode from the OSD, and return it as is. We also return the size - * attribute in the 'obj_size' argument. + * Read the Linux inode info from the OSD, and return it as is. In exofs the + * inode info is in an application specific page/attribute of the osd-object. */ static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi, - struct exofs_fcb *inode, uint64_t *obj_size) + struct exofs_fcb *inode) { struct exofs_sb_info *sbi = sb->s_fs_info; - struct osd_attr attrs[2]; - struct exofs_io_state *ios; + struct osd_attr attrs[] = { + [0] = g_attr_inode_data, + [1] = g_attr_inode_file_layout, + [2] = g_attr_inode_dir_layout, + }; + struct ore_io_state *ios; + struct exofs_on_disk_inode_layout *layout; int ret; - *obj_size = ~0; - ret = exofs_get_io_state(sbi, &ios); + ret = ore_get_io_state(&sbi->layout, &oi->oc, &ios); if (unlikely(ret)) { - EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__); + EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__); return ret; } - ios->obj.id = exofs_oi_objno(oi); - exofs_make_credential(oi->i_cred, &ios->obj); - ios->cred = oi->i_cred; + attrs[1].len = exofs_on_disk_inode_layout_size(sbi->oc.numdevs); + attrs[2].len = exofs_on_disk_inode_layout_size(sbi->oc.numdevs); - attrs[0] = g_attr_inode_data; - attrs[1] = g_attr_logical_length; ios->in_attr = attrs; ios->in_attr_len = ARRAY_SIZE(attrs); - ret = exofs_sbi_read(ios); - if (ret) + ret = ore_read(ios); + if (unlikely(ret)) { + EXOFS_ERR("object(0x%llx) corrupted, return empty file=>%d\n", + _LLU(oi->one_comp.obj.id), ret); + memset(inode, 0, sizeof(*inode)); + inode->i_mode = 0040000 | (0777 & ~022); + /* If object is lost on target we might as well enable it's + * delete. + */ + ret = 0; goto out; + } ret = extract_attr_from_ios(ios, &attrs[0]); if (ret) { - EXOFS_ERR("%s: extract_attr of inode_data failed\n", __func__); + EXOFS_ERR("%s: extract_attr 0 of inode failed\n", __func__); goto out; } WARN_ON(attrs[0].len != EXOFS_INO_ATTR_SIZE); @@ -901,14 +1116,36 @@ static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi, ret = extract_attr_from_ios(ios, &attrs[1]); if (ret) { - EXOFS_ERR("%s: extract_attr of logical_length failed\n", - __func__); + EXOFS_ERR("%s: extract_attr 1 of inode failed\n", __func__); goto out; } - *obj_size = get_unaligned_be64(attrs[1].val_ptr); + if (attrs[1].len) { + layout = attrs[1].val_ptr; + if (layout->gen_func != cpu_to_le16(LAYOUT_MOVING_WINDOW)) { + EXOFS_ERR("%s: unsupported files layout %d\n", + __func__, layout->gen_func); + ret = -ENOTSUPP; + goto out; + } + } + + ret = extract_attr_from_ios(ios, &attrs[2]); + if (ret) { + EXOFS_ERR("%s: extract_attr 2 of inode failed\n", __func__); + goto out; + } + if (attrs[2].len) { + layout = attrs[2].val_ptr; + if (layout->gen_func != cpu_to_le16(LAYOUT_MOVING_WINDOW)) { + EXOFS_ERR("%s: unsupported meta-data layout %d\n", + __func__, layout->gen_func); + ret = -ENOTSUPP; + goto out; + } + } out: - exofs_put_io_state(ios); + ore_put_io_state(ios); return ret; } @@ -925,7 +1162,6 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) struct exofs_i_info *oi; struct exofs_fcb fcb; struct inode *inode; - uint64_t obj_size; int ret; inode = iget_locked(sb, ino); @@ -935,9 +1171,11 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) return inode; oi = exofs_i(inode); __oi_init(oi); + exofs_init_comps(&oi->oc, &oi->one_comp, sb->s_fs_info, + exofs_oi_objno(oi)); /* read the inode from the osd */ - ret = exofs_get_inode(sb, oi, &fcb, &obj_size); + ret = exofs_get_inode(sb, oi, &fcb); if (ret) goto bad_inode; @@ -945,9 +1183,9 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) /* copy stuff from on-disk struct to in-memory struct */ inode->i_mode = le16_to_cpu(fcb.i_mode); - inode->i_uid = le32_to_cpu(fcb.i_uid); - inode->i_gid = le32_to_cpu(fcb.i_gid); - inode->i_nlink = le16_to_cpu(fcb.i_links_count); + i_uid_write(inode, le32_to_cpu(fcb.i_uid)); + i_gid_write(inode, le32_to_cpu(fcb.i_gid)); + set_nlink(inode, le16_to_cpu(fcb.i_links_count)); inode->i_ctime.tv_sec = (signed)le32_to_cpu(fcb.i_ctime); inode->i_atime.tv_sec = (signed)le32_to_cpu(fcb.i_atime); inode->i_mtime.tv_sec = (signed)le32_to_cpu(fcb.i_mtime); @@ -958,13 +1196,6 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) inode->i_blkbits = EXOFS_BLKSHIFT; inode->i_generation = le32_to_cpu(fcb.i_generation); - if ((inode->i_size != obj_size) && - (!exofs_inode_is_fast_symlink(inode))) { - EXOFS_ERR("WARNING: Size of inode=%llu != object=%llu\n", - inode->i_size, _LLU(obj_size)); - /* FIXME: call exofs_inode_recovery() */ - } - oi->i_dir_start_lookup = 0; if ((inode->i_nlink == 0) && (inode->i_mode == 0)) { @@ -983,6 +1214,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) memcpy(oi->i_data, fcb.i_data, sizeof(fcb.i_data)); } + inode->i_mapping->backing_dev_info = sb->s_bdi; if (S_ISREG(inode->i_mode)) { inode->i_op = &exofs_file_inode_operations; inode->i_fop = &exofs_file_operations; @@ -1019,31 +1251,35 @@ bad_inode: int __exofs_wait_obj_created(struct exofs_i_info *oi) { if (!obj_created(oi)) { + EXOFS_DBGMSG("!obj_created\n"); BUG_ON(!obj_2bcreated(oi)); wait_event(oi->i_wq, obj_created(oi)); + EXOFS_DBGMSG("wait_event done\n"); } return unlikely(is_bad_inode(&oi->vfs_inode)) ? -EIO : 0; } + /* * Callback function from exofs_new_inode(). The important thing is that we * set the obj_created flag so that other methods know that the object exists on * the OSD. */ -static void create_done(struct exofs_io_state *ios, void *p) +static void create_done(struct ore_io_state *ios, void *p) { struct inode *inode = p; struct exofs_i_info *oi = exofs_i(inode); struct exofs_sb_info *sbi = inode->i_sb->s_fs_info; int ret; - ret = exofs_check_io(ios, NULL); - exofs_put_io_state(ios); + ret = ore_check_io(ios, NULL); + ore_put_io_state(ios); atomic_dec(&sbi->s_curr_pending); if (unlikely(ret)) { - EXOFS_ERR("object=0x%llx creation faild in pid=0x%llx", - _LLU(exofs_oi_objno(oi)), _LLU(sbi->s_pid)); + EXOFS_ERR("object=0x%llx creation failed in pid=0x%llx", + _LLU(exofs_oi_objno(oi)), + _LLU(oi->one_comp.obj.partition)); /*TODO: When FS is corrupted creation can fail, object already * exist. Get rid of this asynchronous creation, if exist * increment the obj counter and try the next object. Until we @@ -1054,23 +1290,21 @@ static void create_done(struct exofs_io_state *ios, void *p) set_obj_created(oi); - atomic_dec(&inode->i_count); wake_up(&oi->i_wq); } /* * Set up a new inode and create an object for it on the OSD */ -struct inode *exofs_new_inode(struct inode *dir, int mode) +struct inode *exofs_new_inode(struct inode *dir, umode_t mode) { - struct super_block *sb; + struct super_block *sb = dir->i_sb; + struct exofs_sb_info *sbi = sb->s_fs_info; struct inode *inode; struct exofs_i_info *oi; - struct exofs_sb_info *sbi; - struct exofs_io_state *ios; + struct ore_io_state *ios; int ret; - sb = dir->i_sb; inode = new_inode(sb); if (!inode) return ERR_PTR(-ENOMEM); @@ -1080,19 +1314,8 @@ struct inode *exofs_new_inode(struct inode *dir, int mode) set_obj_2bcreated(oi); - sbi = sb->s_fs_info; - - sb->s_dirt = 1; - inode->i_uid = current->cred->fsuid; - if (dir->i_mode & S_ISGID) { - inode->i_gid = dir->i_gid; - if (S_ISDIR(mode)) - mode |= S_ISGID; - } else { - inode->i_gid = current->cred->fsgid; - } - inode->i_mode = mode; - + inode->i_mapping->backing_dev_info = sb->s_bdi; + inode_init_owner(inode, dir, mode); inode->i_ino = sbi->s_nextid++; inode->i_blkbits = EXOFS_BLKSHIFT; inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; @@ -1102,29 +1325,24 @@ struct inode *exofs_new_inode(struct inode *dir, int mode) spin_unlock(&sbi->s_next_gen_lock); insert_inode_hash(inode); + exofs_init_comps(&oi->oc, &oi->one_comp, sb->s_fs_info, + exofs_oi_objno(oi)); + exofs_sbi_write_stats(sbi); /* Make sure new sbi->s_nextid is on disk */ + mark_inode_dirty(inode); - ret = exofs_get_io_state(sbi, &ios); + ret = ore_get_io_state(&sbi->layout, &oi->oc, &ios); if (unlikely(ret)) { - EXOFS_ERR("exofs_new_inode: exofs_get_io_state failed\n"); + EXOFS_ERR("exofs_new_inode: ore_get_io_state failed\n"); return ERR_PTR(ret); } - ios->obj.id = exofs_oi_objno(oi); - exofs_make_credential(oi->i_cred, &ios->obj); - - /* increment the refcount so that the inode will still be around when we - * reach the callback - */ - atomic_inc(&inode->i_count); - ios->done = create_done; ios->private = inode; - ios->cred = oi->i_cred; - ret = exofs_sbi_create(ios); + + ret = ore_create(ios); if (ret) { - atomic_dec(&inode->i_count); - exofs_put_io_state(ios); + ore_put_io_state(ios); return ERR_PTR(ret); } atomic_inc(&sbi->s_curr_pending); @@ -1143,11 +1361,11 @@ struct updatei_args { /* * Callback function from exofs_update_inode(). */ -static void updatei_done(struct exofs_io_state *ios, void *p) +static void updatei_done(struct ore_io_state *ios, void *p) { struct updatei_args *args = p; - exofs_put_io_state(ios); + ore_put_io_state(ios); atomic_dec(&args->sbi->s_curr_pending); @@ -1163,21 +1381,23 @@ static int exofs_update_inode(struct inode *inode, int do_sync) struct exofs_i_info *oi = exofs_i(inode); struct super_block *sb = inode->i_sb; struct exofs_sb_info *sbi = sb->s_fs_info; - struct exofs_io_state *ios; + struct ore_io_state *ios; struct osd_attr attr; struct exofs_fcb *fcb; struct updatei_args *args; int ret; args = kzalloc(sizeof(*args), GFP_KERNEL); - if (!args) + if (!args) { + EXOFS_DBGMSG("Failed kzalloc of args\n"); return -ENOMEM; + } fcb = &args->fcb; fcb->i_mode = cpu_to_le16(inode->i_mode); - fcb->i_uid = cpu_to_le32(inode->i_uid); - fcb->i_gid = cpu_to_le32(inode->i_gid); + fcb->i_uid = cpu_to_le32(i_uid_read(inode)); + fcb->i_gid = cpu_to_le32(i_gid_read(inode)); fcb->i_links_count = cpu_to_le16(inode->i_nlink); fcb->i_ctime = cpu_to_le32(inode->i_ctime.tv_sec); fcb->i_atime = cpu_to_le32(inode->i_atime.tv_sec); @@ -1200,9 +1420,9 @@ static int exofs_update_inode(struct inode *inode, int do_sync) } else memcpy(fcb->i_data, oi->i_data, sizeof(fcb->i_data)); - ret = exofs_get_io_state(sbi, &ios); + ret = ore_get_io_state(&sbi->layout, &oi->oc, &ios); if (unlikely(ret)) { - EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__); + EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__); goto free_args; } @@ -1211,12 +1431,7 @@ static int exofs_update_inode(struct inode *inode, int do_sync) ios->out_attr_len = 1; ios->out_attr = &attr; - if (!obj_created(oi)) { - EXOFS_DBGMSG("!obj_created\n"); - BUG_ON(!obj_2bcreated(oi)); - wait_event(oi->i_wq, obj_created(oi)); - EXOFS_DBGMSG("wait_event done\n"); - } + wait_obj_created(oi); if (!do_sync) { args->sbi = sbi; @@ -1224,34 +1439,36 @@ static int exofs_update_inode(struct inode *inode, int do_sync) ios->private = args; } - ret = exofs_oi_write(oi, ios); + ret = ore_write(ios); if (!do_sync && !ret) { atomic_inc(&sbi->s_curr_pending); goto out; /* deallocation in updatei_done */ } - exofs_put_io_state(ios); + ore_put_io_state(ios); free_args: kfree(args); out: - EXOFS_DBGMSG("ret=>%d\n", ret); + EXOFS_DBGMSG("(0x%lx) do_sync=%d ret=>%d\n", + inode->i_ino, do_sync, ret); return ret; } -int exofs_write_inode(struct inode *inode, int wait) +int exofs_write_inode(struct inode *inode, struct writeback_control *wbc) { - return exofs_update_inode(inode, wait); + /* FIXME: fix fsync and use wbc->sync_mode == WB_SYNC_ALL */ + return exofs_update_inode(inode, 1); } /* * Callback function from exofs_delete_inode() - don't have much cleaning up to * do. */ -static void delete_done(struct exofs_io_state *ios, void *p) +static void delete_done(struct ore_io_state *ios, void *p) { struct exofs_sb_info *sbi = p; - exofs_put_io_state(ios); + ore_put_io_state(ios); atomic_dec(&sbi->s_curr_pending); } @@ -1261,48 +1478,44 @@ static void delete_done(struct exofs_io_state *ios, void *p) * from the OSD here. We make sure the object was created before we try and * delete it. */ -void exofs_delete_inode(struct inode *inode) +void exofs_evict_inode(struct inode *inode) { struct exofs_i_info *oi = exofs_i(inode); struct super_block *sb = inode->i_sb; struct exofs_sb_info *sbi = sb->s_fs_info; - struct exofs_io_state *ios; + struct ore_io_state *ios; int ret; - truncate_inode_pages(&inode->i_data, 0); + truncate_inode_pages_final(&inode->i_data); - if (is_bad_inode(inode)) + /* TODO: should do better here */ + if (inode->i_nlink || is_bad_inode(inode)) goto no_delete; - mark_inode_dirty(inode); - exofs_update_inode(inode, inode_needs_sync(inode)); - inode->i_size = 0; - if (inode->i_blocks) - exofs_truncate(inode); - clear_inode(inode); - ret = exofs_get_io_state(sbi, &ios); + /* if we are deleting an obj that hasn't been created yet, wait. + * This also makes sure that create_done cannot be called with an + * already evicted inode. + */ + wait_obj_created(oi); + /* ignore the error, attempt a remove anyway */ + + /* Now Remove the OSD objects */ + ret = ore_get_io_state(&sbi->layout, &oi->oc, &ios); if (unlikely(ret)) { - EXOFS_ERR("%s: exofs_get_io_state failed\n", __func__); + EXOFS_ERR("%s: ore_get_io_state failed\n", __func__); return; } - /* if we are deleting an obj that hasn't been created yet, wait */ - if (!obj_created(oi)) { - BUG_ON(!obj_2bcreated(oi)); - wait_event(oi->i_wq, obj_created(oi)); - } - - ios->obj.id = exofs_oi_objno(oi); ios->done = delete_done; ios->private = sbi; - ios->cred = oi->i_cred; - ret = exofs_sbi_remove(ios); + + ret = ore_remove(ios); if (ret) { - EXOFS_ERR("%s: exofs_sbi_remove failed\n", __func__); - exofs_put_io_state(ios); + EXOFS_ERR("%s: ore_remove failed\n", __func__); + ore_put_io_state(ios); return; } atomic_inc(&sbi->s_curr_pending); |