From 9f249162fbf1aea1335e13c57fd5355d00e07a47 Mon Sep 17 00:00:00 2001 From: Thadeu Lima de Souza Cascardo Date: Mon, 27 Jul 2009 13:26:32 -0300 Subject: [PATCH 1/9] trivial: some small fixes in exofs documentation Add exofs.txt to filesystems Documentation index and fix some typos, identation and grammar. Signed-off-by: Thadeu Lima de Souza Cascardo Signed-off-by: Boaz Harrosh --- Documentation/filesystems/00-INDEX | 2 ++ Documentation/filesystems/exofs.txt | 23 ++++++++++++----------- 2 files changed, 14 insertions(+), 11 deletions(-) diff --git a/Documentation/filesystems/00-INDEX b/Documentation/filesystems/00-INDEX index f15621ee5599..7001782ab932 100644 --- a/Documentation/filesystems/00-INDEX +++ b/Documentation/filesystems/00-INDEX @@ -36,6 +36,8 @@ dnotify.txt - info about directory notification in Linux. ecryptfs.txt - docs on eCryptfs: stacked cryptographic filesystem for Linux. +exofs.txt + - info, usage, mount options, design about EXOFS. ext2.txt - info, mount options and specifications for the Ext2 filesystem. ext3.txt diff --git a/Documentation/filesystems/exofs.txt b/Documentation/filesystems/exofs.txt index 0ced74c2f73c..abd2a9b5b787 100644 --- a/Documentation/filesystems/exofs.txt +++ b/Documentation/filesystems/exofs.txt @@ -60,13 +60,13 @@ USAGE mkfs.exofs --pid=65536 --format /dev/osd0 - The --format is optional if not specified no OSD_FORMAT will be - preformed and a clean file system will be created in the specified pid, + The --format is optional. If not specified, no OSD_FORMAT will be + performed and a clean file system will be created in the specified pid, in the available space of the target. (Use --format=size_in_meg to limit the total LUN space available) - If pid already exist it will be deleted and a new one will be created in it's - place. Be careful. + If pid already exists, it will be deleted and a new one will be created in + its place. Be careful. An exofs lives inside a single OSD partition. You can create multiple exofs filesystems on the same device using multiple pids. @@ -81,7 +81,7 @@ USAGE 7. For reference (See do-exofs example script): do-exofs start - an example of how to perform the above steps. - do-exofs stop - an example of how to unmount the file system. + do-exofs stop - an example of how to unmount the file system. do-exofs format - an example of how to format and mkfs a new exofs. 8. Extra compilation flags (uncomment in fs/exofs/Kbuild): @@ -104,8 +104,8 @@ Where: exofs specific options: Options are separated by commas (,) pid= - The partition number to mount/create as container of the filesystem. - This option is mandatory - to= - Timeout in ticks for a single command + This option is mandatory. + to= - Timeout in ticks for a single command. default is (60 * HZ) [for debugging only] =============================================================================== @@ -116,7 +116,7 @@ DESIGN with a special ID (defined in common.h). Information included in the file system control block is used to fill the in-memory superblock structure at mount time. This object is created before - the file system is used by mkexofs.c It contains information such as: + the file system is used by mkexofs.c. It contains information such as: - The file system's magic number - The next inode number to be allocated @@ -134,8 +134,8 @@ DESIGN attributes. This applies to both regular files and other types (directories, device files, symlinks, etc.). -* Credentials are generated per object (inode and superblock) when they is - created in memory (read off disk or created). The credential works for all +* Credentials are generated per object (inode and superblock) when they are + created in memory (read from disk or created). The credential works for all operations and is used as long as the object remains in memory. * Async OSD operations are used whenever possible, but the target may execute @@ -145,7 +145,8 @@ DESIGN from executing in reverse order: - The following are handled with the OBJ_CREATED and OBJ_2BCREATED flags. OBJ_CREATED is set when we know the object exists on the OSD - - in create's callback function, and when we successfully do a read_inode. + in create's callback function, and when we successfully do a + read_inode. OBJ_2BCREATED is set in the beginning of the create function, so we know that we should wait. - create/delete: delete should wait until the object is created From 58311c43dfc3997a1f7b5883f827443f34108f8f Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Tue, 14 Jul 2009 11:06:08 +0300 Subject: [PATCH 2/9] exofs: More sane debug print debug prints should be somewhat useful without actually reading the source code Signed-off-by: Boaz Harrosh --- fs/exofs/inode.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c index 6c10f7476699..44748613be07 100644 --- a/fs/exofs/inode.c +++ b/fs/exofs/inode.c @@ -950,8 +950,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) #ifdef EXOFS_DEBUG_OBJ_ISIZE if ((inode->i_size != sanity) && (!exofs_inode_is_fast_symlink(inode))) { - EXOFS_ERR("WARNING: Size of object from inode and " - "attributes differ (%lld != %llu)\n", + EXOFS_ERR("WARNING: Size of inode=%llu != object=%llu\n", inode->i_size, _LLU(sanity)); } #endif From fe33cc1ee170c0e3b47ab9cbac07083b3446961c Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Sun, 1 Nov 2009 18:28:14 +0200 Subject: [PATCH 3/9] exofs: dbg-print less Iner-loops printing is converted to EXOFS_DBG2 which is #defined to nothing. It is now almost bareable to just leave debug-on. Every operation is printed once, with most relevant info (I hope). Signed-off-by: Boaz Harrosh --- fs/exofs/inode.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c index 44748613be07..01fa798e8fdb 100644 --- a/fs/exofs/inode.c +++ b/fs/exofs/inode.c @@ -41,6 +41,8 @@ # define EXOFS_DEBUG_OBJ_ISIZE 1 #endif +#define EXOFS_DBGMSG2(M...) do {} while (0) + struct page_collect { struct exofs_sb_info *sbi; struct request_queue *req_q; @@ -198,7 +200,7 @@ static int __readpages_done(struct osd_request *or, struct page_collect *pcol, else page_stat = ret; - EXOFS_DBGMSG(" readpages_done(0x%lx, 0x%lx) %s\n", + EXOFS_DBGMSG2(" readpages_done(0x%lx, 0x%lx) %s\n", inode->i_ino, page->index, page_stat ? "bad_bytes" : "good_bytes"); @@ -370,12 +372,12 @@ try_again: if (len != PAGE_CACHE_SIZE) zero_user(page, len, PAGE_CACHE_SIZE - len); - EXOFS_DBGMSG(" readpage_strip(0x%lx, 0x%lx) len=0x%zx\n", + EXOFS_DBGMSG2(" readpage_strip(0x%lx, 0x%lx) len=0x%zx\n", inode->i_ino, page->index, len); ret = pcol_add_page(pcol, page, len); if (ret) { - EXOFS_DBGMSG("Failed pcol_add_page pages[i]=%p " + EXOFS_DBGMSG2("Failed pcol_add_page pages[i]=%p " "this_len=0x%zx nr_pages=%u length=0x%lx\n", page, len, pcol->nr_pages, pcol->length); @@ -482,7 +484,7 @@ static void writepages_done(struct osd_request *or, void *p) update_write_page(page, page_stat); unlock_page(page); - EXOFS_DBGMSG(" writepages_done(0x%lx, 0x%lx) status=%d\n", + EXOFS_DBGMSG2(" writepages_done(0x%lx, 0x%lx) status=%d\n", inode->i_ino, page->index, page_stat); length += bvec->bv_len; @@ -609,7 +611,7 @@ try_again: goto fail; } - EXOFS_DBGMSG(" writepage_strip(0x%lx, 0x%lx) len=0x%zx\n", + EXOFS_DBGMSG2(" writepage_strip(0x%lx, 0x%lx) len=0x%zx\n", inode->i_ino, page->index, len); ret = pcol_add_page(pcol, page, len); From 9cfdc7aa9f1b59627029ad00a58c3f59eb2cc383 Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Tue, 4 Aug 2009 20:40:29 +0300 Subject: [PATCH 4/9] exofs: refactor exofs_i_info initialization into common helper There are two places that initialize inodes: exofs_iget() and exofs_new_inode() As more members of exofs_i_info that need initialization are added this code will grow. (soon) Signed-off-by: Boaz Harrosh --- fs/exofs/inode.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c index 01fa798e8fdb..7bc71a7d30a8 100644 --- a/fs/exofs/inode.c +++ b/fs/exofs/inode.c @@ -908,6 +908,12 @@ out: return ret; } + +static void __oi_init(struct exofs_i_info *oi) +{ + init_waitqueue_head(&oi->i_wq); + oi->i_flags = 0; +} /* * Fill in an inode read from the OSD and set it up for use */ @@ -925,13 +931,13 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) if (!(inode->i_state & I_NEW)) return inode; oi = exofs_i(inode); + __oi_init(oi); /* read the inode from the osd */ ret = exofs_get_inode(sb, oi, &fcb, &sanity); if (ret) goto bad_inode; - init_waitqueue_head(&oi->i_wq); set_obj_created(oi); /* copy stuff from on-disk struct to in-memory struct */ @@ -1062,8 +1068,8 @@ struct inode *exofs_new_inode(struct inode *dir, int mode) return ERR_PTR(-ENOMEM); oi = exofs_i(inode); + __oi_init(oi); - init_waitqueue_head(&oi->i_wq); set_obj_2bcreated(oi); sbi = sb->s_fs_info; From 19fe294f2eaee33574ac1fdcf3cc26880de820ea Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Thu, 3 Sep 2009 20:38:02 +0300 Subject: [PATCH 5/9] exofs: Prints on mount and unmout It is important to print in the logs when a filesystem was mounted and eventually unmounted. Print the osd-device's osd_name and pid the FS was mounted/unmounted on. TODO: How to also print the namespace path the filesystem was mounted on? Signed-off-by: Boaz Harrosh --- fs/exofs/super.c | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/fs/exofs/super.c b/fs/exofs/super.c index 9f500dec3b59..920f0165edf3 100644 --- a/fs/exofs/super.c +++ b/fs/exofs/super.c @@ -257,6 +257,15 @@ static void exofs_write_super(struct super_block *sb) sb->s_dirt = 0; } +static void _exofs_print_device(const char *msg, const char *dev_path, + struct osd_dev *od, u64 pid) +{ + const struct osd_dev_info *odi = osduld_device_info(od); + + printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n", + msg, dev_path ?: "", odi->osdname, _LLU(pid)); +} + /* * This function is called when the vfs is freeing the superblock. We just * need to free our own part. @@ -279,6 +288,7 @@ static void exofs_put_super(struct super_block *sb) msecs_to_jiffies(100)); } + _exofs_print_device("Unmounting", NULL, sbi->s_dev, sbi->s_pid); osduld_put_device(sbi->s_dev); kfree(sb->s_fs_info); sb->s_fs_info = NULL; @@ -395,6 +405,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) goto free_sbi; } + _exofs_print_device("Mounting", opts->dev_name, sbi->s_dev, sbi->s_pid); ret = 0; out: if (or) From cae012d8532879544326fff5fa2ae22a6dfe8e23 Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Mon, 2 Nov 2009 18:19:24 +0200 Subject: [PATCH 6/9] exofs: statfs blocks is sectors not FS blocks Even though exofs has a 4k block size, statfs blocks is in sectors (512 bytes). Also if target returns 0 for capacity then make it ULLONG_MAX. df does not like zero-size filesystems Signed-off-by: Boaz Harrosh --- fs/exofs/super.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) diff --git a/fs/exofs/super.c b/fs/exofs/super.c index 920f0165edf3..28add3eac0a4 100644 --- a/fs/exofs/super.c +++ b/fs/exofs/super.c @@ -473,9 +473,11 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) goto out; ret = extract_attr_from_req(or, &attrs[0]); - if (likely(!ret)) + if (likely(!ret)) { capacity = get_unaligned_be64(attrs[0].val_ptr); - else + if (unlikely(!capacity)) + capacity = ULLONG_MAX; + } else EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n"); ret = extract_attr_from_req(or, &attrs[1]); @@ -487,8 +489,8 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) /* fill in the stats buffer */ buf->f_type = EXOFS_SUPER_MAGIC; buf->f_bsize = EXOFS_BLKSIZE; - buf->f_blocks = (capacity >> EXOFS_BLKSHIFT); - buf->f_bfree = ((capacity - used) >> EXOFS_BLKSHIFT); + buf->f_blocks = capacity >> 9; + buf->f_bfree = (capacity - used) >> 9; buf->f_bavail = buf->f_bfree; buf->f_files = sbi->s_numfiles; buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles; From 8ce9bdd1fbe962933736d7977e972972cd5d754c Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Tue, 3 Nov 2009 16:46:00 +0200 Subject: [PATCH 7/9] exofs: move osd.c to ios.c If I do a "git mv" together with a massive code change and commit in one patch, git looses the rename and records a delete/new instead. This is bad because I want a rename recorded so later rebased/cherry-picked patches to the old name will work. Also the --follow is lost. Signed-off-by: Boaz Harrosh --- fs/exofs/Kbuild | 2 +- fs/exofs/{osd.c => ios.c} | 0 2 files changed, 1 insertion(+), 1 deletion(-) rename fs/exofs/{osd.c => ios.c} (100%) diff --git a/fs/exofs/Kbuild b/fs/exofs/Kbuild index cc2d22db119c..2d0f757fda3e 100644 --- a/fs/exofs/Kbuild +++ b/fs/exofs/Kbuild @@ -12,5 +12,5 @@ # Kbuild - Gets included from the Kernels Makefile and build system # -exofs-y := osd.o inode.o file.o symlink.o namei.o dir.o super.o +exofs-y := ios.o inode.o file.o symlink.o namei.o dir.o super.o obj-$(CONFIG_EXOFS_FS) += exofs.o diff --git a/fs/exofs/osd.c b/fs/exofs/ios.c similarity index 100% rename from fs/exofs/osd.c rename to fs/exofs/ios.c From 06886a5a3dc5a5abe0a4d257c26317bde7047be8 Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Sun, 8 Nov 2009 14:54:08 +0200 Subject: [PATCH 8/9] exofs: Move all operations to an io_engine In anticipation for multi-device operations, we separate osd operations into an abstract I/O API. Currently only one device is used but later when adding more devices, we will drive all devices in parallel according to a "data_map" that describes how data is arranged on multiple devices. The file system level operates, like before, as if there is one object (inode-number) and an i_size. The io engine will split this to the same object-number but on multiple device. At first we introduce Mirror (raid 1) layout. But at the final outcome we intend to fully implement the pNFS-Objects data-map, including raid 0,4,5,6 over mirrored devices, over multiple device-groups. And more. See: http://tools.ietf.org/html/draft-ietf-nfsv4-pnfs-obj-12 * Define an io_state based API for accessing osd storage devices in an abstract way. Usage: First a caller allocates an io state with: exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** ios); Then calles one of: exofs_sbi_create(struct exofs_io_state *ios); exofs_sbi_remove(struct exofs_io_state *ios); exofs_sbi_write(struct exofs_io_state *ios); exofs_sbi_read(struct exofs_io_state *ios); exofs_oi_truncate(struct exofs_i_info *oi, u64 new_len); And when done exofs_put_io_state(struct exofs_io_state *ios); * Convert all source files to use this new API * Convert from bio_alloc to bio_kmalloc * In io engine we make use of the now fixed osd_req_decode_sense There are no functional changes or on disk additions after this patch. Signed-off-by: Boaz Harrosh --- fs/exofs/common.h | 18 --- fs/exofs/exofs.h | 87 ++++++++++- fs/exofs/inode.c | 383 ++++++++++++++++++++++----------------------- fs/exofs/ios.c | 390 +++++++++++++++++++++++++++++++++++++++------- fs/exofs/super.c | 120 +++++--------- 5 files changed, 646 insertions(+), 352 deletions(-) diff --git a/fs/exofs/common.h b/fs/exofs/common.h index c6718e4817fe..ce1c71692599 100644 --- a/fs/exofs/common.h +++ b/fs/exofs/common.h @@ -155,22 +155,4 @@ enum { (((name_len) + offsetof(struct exofs_dir_entry, name) + \ EXOFS_DIR_ROUND) & ~EXOFS_DIR_ROUND) -/************************* - * function declarations * - *************************/ -/* osd.c */ -void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], - const struct osd_obj_id *obj); - -int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid); -static inline int exofs_check_ok(struct osd_request *or) -{ - return exofs_check_ok_resid(or, NULL, NULL); -} -int exofs_sync_op(struct osd_request *or, int timeout, u8 *cred); -int exofs_async_op(struct osd_request *or, - osd_req_done_fn *async_done, void *caller_context, u8 *cred); - -int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr); - #endif /*ifndef __EXOFS_COM_H__*/ diff --git a/fs/exofs/exofs.h b/fs/exofs/exofs.h index 5ec72e020b22..2e08859a89e8 100644 --- a/fs/exofs/exofs.h +++ b/fs/exofs/exofs.h @@ -30,14 +30,13 @@ * along with exofs; if not, write to the Free Software * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA */ +#ifndef __EXOFS_H__ +#define __EXOFS_H__ #include #include #include "common.h" -#ifndef __EXOFS_H__ -#define __EXOFS_H__ - #define EXOFS_ERR(fmt, a...) printk(KERN_ERR "exofs: " fmt, ##a) #ifdef CONFIG_EXOFS_DEBUG @@ -56,6 +55,7 @@ */ struct exofs_sb_info { struct osd_dev *s_dev; /* returned by get_osd_dev */ + struct exofs_fscb s_fscb; /* Written often, pre-allocate*/ osd_id s_pid; /* partition ID of file system*/ int s_timeout; /* timeout for OSD operations */ uint64_t s_nextid; /* highest object ID used */ @@ -79,6 +79,50 @@ struct exofs_i_info { struct inode vfs_inode; /* normal in-memory inode */ }; +static inline osd_id exofs_oi_objno(struct exofs_i_info *oi) +{ + return oi->vfs_inode.i_ino + EXOFS_OBJ_OFF; +} + +struct exofs_io_state; +typedef void (*exofs_io_done_fn)(struct exofs_io_state *or, void *private); + +struct exofs_io_state { + struct kref kref; + + void *private; + exofs_io_done_fn done; + + struct exofs_sb_info *sbi; + struct osd_obj_id obj; + u8 *cred; + + /* Global read/write IO*/ + loff_t offset; + unsigned long length; + void *kern_buff; + struct bio *bio; + + /* Attributes */ + unsigned in_attr_len; + struct osd_attr *in_attr; + unsigned out_attr_len; + struct osd_attr *out_attr; + + /* Variable array of size numdevs */ + unsigned numdevs; + struct exofs_per_dev_state { + struct osd_request *or; + struct bio *bio; + } per_dev[]; +}; + +static inline unsigned exofs_io_state_size(unsigned numdevs) +{ + return sizeof(struct exofs_io_state) + + sizeof(struct exofs_per_dev_state) * numdevs; +} + /* * our inode flags */ @@ -130,6 +174,42 @@ static inline struct exofs_i_info *exofs_i(struct inode *inode) /************************* * function declarations * *************************/ + +/* ios.c */ +void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], + const struct osd_obj_id *obj); +int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj, + u64 offset, void *p, unsigned length); + +int exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** ios); +void exofs_put_io_state(struct exofs_io_state *ios); + +int exofs_check_io(struct exofs_io_state *ios, u64 *resid); + +int exofs_sbi_create(struct exofs_io_state *ios); +int exofs_sbi_remove(struct exofs_io_state *ios); +int exofs_sbi_write(struct exofs_io_state *ios); +int exofs_sbi_read(struct exofs_io_state *ios); + +int extract_attr_from_ios(struct exofs_io_state *ios, struct osd_attr *attr); + +int exofs_oi_truncate(struct exofs_i_info *oi, u64 new_len); +static inline int exofs_oi_write(struct exofs_i_info *oi, + struct exofs_io_state *ios) +{ + ios->obj.id = exofs_oi_objno(oi); + ios->cred = oi->i_cred; + return exofs_sbi_write(ios); +} + +static inline int exofs_oi_read(struct exofs_i_info *oi, + struct exofs_io_state *ios) +{ + ios->obj.id = exofs_oi_objno(oi); + ios->cred = oi->i_cred; + return exofs_sbi_read(ios); +} + /* inode.c */ void exofs_truncate(struct inode *inode); int exofs_setattr(struct dentry *, struct iattr *); @@ -169,6 +249,7 @@ extern const struct file_operations exofs_file_operations; /* inode.c */ extern const struct address_space_operations exofs_aops; +extern const struct osd_attr g_attr_logical_length; /* namei.c */ extern const struct inode_operations exofs_dir_inode_operations; diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c index 7bc71a7d30a8..7578950fd135 100644 --- a/fs/exofs/inode.c +++ b/fs/exofs/inode.c @@ -37,17 +37,18 @@ #include "exofs.h" -#ifdef CONFIG_EXOFS_DEBUG -# define EXOFS_DEBUG_OBJ_ISIZE 1 -#endif - #define EXOFS_DBGMSG2(M...) do {} while (0) +enum { BIO_MAX_PAGES_KMALLOC = + (PAGE_SIZE - sizeof(struct bio)) / sizeof(struct bio_vec), +}; + 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 bio *bio; unsigned nr_pages; @@ -56,7 +57,7 @@ struct page_collect { }; static void _pcol_init(struct page_collect *pcol, unsigned expected_pages, - struct inode *inode) + struct inode *inode) { struct exofs_sb_info *sbi = inode->i_sb->s_fs_info; @@ -65,13 +66,11 @@ static void _pcol_init(struct page_collect *pcol, unsigned expected_pages, pcol->inode = inode; pcol->expected_pages = expected_pages; + pcol->ios = NULL; pcol->bio = NULL; pcol->nr_pages = 0; pcol->length = 0; pcol->pg_first = -1; - - EXOFS_DBGMSG("_pcol_init ino=0x%lx expected_pages=%u\n", inode->i_ino, - expected_pages); } static void _pcol_reset(struct page_collect *pcol) @@ -82,35 +81,49 @@ static void _pcol_reset(struct page_collect *pcol) pcol->nr_pages = 0; pcol->length = 0; pcol->pg_first = -1; - EXOFS_DBGMSG("_pcol_reset ino=0x%lx expected_pages=%u\n", - pcol->inode->i_ino, pcol->expected_pages); + pcol->ios = 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 = 128; + pcol->expected_pages = BIO_MAX_PAGES_KMALLOC; } static int pcol_try_alloc(struct page_collect *pcol) { - int pages = min_t(unsigned, pcol->expected_pages, BIO_MAX_PAGES); + 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); + + if (ret) + return ret; + } for (; pages; pages >>= 1) { - pcol->bio = bio_alloc(GFP_KERNEL, pages); + pcol->bio = bio_kmalloc(GFP_KERNEL, pages); if (likely(pcol->bio)) return 0; } - EXOFS_ERR("Failed to kcalloc expected_pages=%u\n", + EXOFS_ERR("Failed to bio_kmalloc expected_pages=%u\n", pcol->expected_pages); return -ENOMEM; } static void pcol_free(struct page_collect *pcol) { - bio_put(pcol->bio); - pcol->bio = NULL; + if (pcol->bio) { + bio_put(pcol->bio); + pcol->bio = NULL; + } + + if (pcol->ios) { + exofs_put_io_state(pcol->ios); + pcol->ios = NULL; + } } static int pcol_add_page(struct page_collect *pcol, struct page *page, @@ -163,22 +176,17 @@ 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 osd_request *or, struct page_collect *pcol, - bool do_unlock) +static int __readpages_done(struct page_collect *pcol, bool do_unlock) { struct bio_vec *bvec; int i; u64 resid; u64 good_bytes; u64 length = 0; - int ret = exofs_check_ok_resid(or, &resid, NULL); - - osd_end_request(or); + int ret = exofs_check_io(pcol->ios, &resid); if (likely(!ret)) good_bytes = pcol->length; - else if (!resid) - good_bytes = 0; else good_bytes = pcol->length - resid; @@ -216,13 +224,13 @@ static int __readpages_done(struct osd_request *or, struct page_collect *pcol, } /* callback of async reads */ -static void readpages_done(struct osd_request *or, void *p) +static void readpages_done(struct exofs_io_state *ios, void *p) { struct page_collect *pcol = p; - __readpages_done(or, pcol, true); + __readpages_done(pcol, true); atomic_dec(&pcol->sbi->s_curr_pending); - kfree(p); + kfree(pcol); } static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw) @@ -240,17 +248,13 @@ static void _unlock_pcol_pages(struct page_collect *pcol, int ret, int rw) unlock_page(page); } - pcol_free(pcol); } static int read_exec(struct page_collect *pcol, bool is_sync) { struct exofs_i_info *oi = exofs_i(pcol->inode); - struct osd_obj_id obj = {pcol->sbi->s_pid, - pcol->inode->i_ino + EXOFS_OBJ_OFF}; - struct osd_request *or = NULL; + struct exofs_io_state *ios = pcol->ios; struct page_collect *pcol_copy = NULL; - loff_t i_start = pcol->pg_first << PAGE_CACHE_SHIFT; int ret; if (!pcol->bio) @@ -259,17 +263,13 @@ static int read_exec(struct page_collect *pcol, bool is_sync) /* see comment in _readpage() about sync reads */ WARN_ON(is_sync && (pcol->nr_pages != 1)); - or = osd_start_request(pcol->sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - ret = -ENOMEM; - goto err; - } - - osd_req_read(or, &obj, i_start, pcol->bio, pcol->length); + ios->bio = pcol->bio; + ios->length = pcol->length; + ios->offset = pcol->pg_first << PAGE_CACHE_SHIFT; if (is_sync) { - exofs_sync_op(or, pcol->sbi->s_timeout, oi->i_cred); - return __readpages_done(or, pcol, false); + exofs_oi_read(oi, pcol->ios); + return __readpages_done(pcol, false); } pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL); @@ -279,14 +279,16 @@ static int read_exec(struct page_collect *pcol, bool is_sync) } *pcol_copy = *pcol; - ret = exofs_async_op(or, readpages_done, pcol_copy, oi->i_cred); + ios->done = readpages_done; + ios->private = pcol_copy; + ret = exofs_oi_read(oi, ios); if (unlikely(ret)) goto err; atomic_inc(&pcol->sbi->s_curr_pending); EXOFS_DBGMSG("read_exec obj=0x%llx start=0x%llx length=0x%lx\n", - obj.id, _LLU(i_start), pcol->length); + ios->obj.id, _LLU(ios->offset), pcol->length); /* pages ownership was passed to pcol_copy */ _pcol_reset(pcol); @@ -295,12 +297,10 @@ static int read_exec(struct page_collect *pcol, bool is_sync) err: if (!is_sync) _unlock_pcol_pages(pcol, ret, READ); - else /* Pages unlocked by caller in sync mode only free bio */ - pcol_free(pcol); + + pcol_free(pcol); kfree(pcol_copy); - if (or) - osd_end_request(or); return ret; } @@ -421,9 +421,8 @@ static int _readpage(struct page *page, bool is_sync) _pcol_init(&pcol, 1, page->mapping->host); - /* readpage_strip might call read_exec(,async) inside at several places - * but this is safe for is_async=0 since read_exec will not do anything - * when we have a single page. + /* readpage_strip might call read_exec(,is_sync==false) at several + * places but not if we have a single page. */ ret = readpage_strip(&pcol, page); if (ret) { @@ -442,8 +441,8 @@ static int exofs_readpage(struct file *file, struct page *page) return _readpage(page, false); } -/* Callback for osd_write. All writes are asynchronouse */ -static void writepages_done(struct osd_request *or, void *p) +/* Callback for osd_write. All writes are asynchronous */ +static void writepages_done(struct exofs_io_state *ios, void *p) { struct page_collect *pcol = p; struct bio_vec *bvec; @@ -451,16 +450,12 @@ static void writepages_done(struct osd_request *or, void *p) u64 resid; u64 good_bytes; u64 length = 0; + int ret = exofs_check_io(ios, &resid); - int ret = exofs_check_ok_resid(or, NULL, &resid); - - osd_end_request(or); atomic_dec(&pcol->sbi->s_curr_pending); if (likely(!ret)) good_bytes = pcol->length; - else if (!resid) - good_bytes = 0; else good_bytes = pcol->length - resid; @@ -498,23 +493,13 @@ static void writepages_done(struct osd_request *or, void *p) static int write_exec(struct page_collect *pcol) { struct exofs_i_info *oi = exofs_i(pcol->inode); - struct osd_obj_id obj = {pcol->sbi->s_pid, - pcol->inode->i_ino + EXOFS_OBJ_OFF}; - struct osd_request *or = NULL; + struct exofs_io_state *ios = pcol->ios; struct page_collect *pcol_copy = NULL; - loff_t i_start = pcol->pg_first << PAGE_CACHE_SHIFT; int ret; if (!pcol->bio) return 0; - or = osd_start_request(pcol->sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("write_exec: Faild to osd_start_request()\n"); - ret = -ENOMEM; - goto err; - } - pcol_copy = kmalloc(sizeof(*pcol_copy), GFP_KERNEL); if (!pcol_copy) { EXOFS_ERR("write_exec: Faild to kmalloc(pcol)\n"); @@ -525,16 +510,22 @@ static int write_exec(struct page_collect *pcol) *pcol_copy = *pcol; pcol_copy->bio->bi_rw |= (1 << BIO_RW); /* FIXME: bio_set_dir() */ - osd_req_write(or, &obj, i_start, pcol_copy->bio, pcol_copy->length); - ret = exofs_async_op(or, writepages_done, pcol_copy, oi->i_cred); + + ios->bio = pcol_copy->bio; + ios->offset = pcol_copy->pg_first << PAGE_CACHE_SHIFT; + ios->length = pcol_copy->length; + ios->done = writepages_done; + ios->private = pcol_copy; + + ret = exofs_oi_write(oi, ios); if (unlikely(ret)) { - EXOFS_ERR("write_exec: exofs_async_op() Faild\n"); + EXOFS_ERR("write_exec: exofs_oi_write() Faild\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(i_start), + pcol->inode->i_ino, pcol->pg_first, _LLU(ios->offset), pcol->length); /* pages ownership was passed to pcol_copy */ _pcol_reset(pcol); @@ -542,9 +533,9 @@ static int write_exec(struct page_collect *pcol) err: _unlock_pcol_pages(pcol, ret, WRITE); + pcol_free(pcol); kfree(pcol_copy); - if (or) - osd_end_request(or); + return ret; } @@ -588,6 +579,9 @@ static int writepage_strip(struct page *page, if (PageError(page)) ClearPageError(page); unlock_page(page); + EXOFS_DBGMSG("writepage_strip(0x%lx, 0x%lx) " + "outside the limits\n", + inode->i_ino, page->index); return 0; } } @@ -602,6 +596,9 @@ try_again: ret = write_exec(pcol); if (unlikely(ret)) goto fail; + + EXOFS_DBGMSG("writepage_strip(0x%lx, 0x%lx) Discontinuity\n", + inode->i_ino, page->index); goto try_again; } @@ -636,6 +633,8 @@ try_again: return 0; fail: + EXOFS_DBGMSG("Error: writepage_strip(0x%lx, 0x%lx)=>%d\n", + inode->i_ino, page->index, ret); set_bit(AS_EIO, &page->mapping->flags); unlock_page(page); return ret; @@ -654,14 +653,17 @@ static int exofs_writepages(struct address_space *mapping, wbc->range_end >> PAGE_CACHE_SHIFT; if (start || end) - expected_pages = min(end - start + 1, 32L); + expected_pages = end - start + 1; else expected_pages = mapping->nrpages; - EXOFS_DBGMSG("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx" - " m->nrpages=%lu start=0x%lx end=0x%lx\n", + if (expected_pages < 32L) + expected_pages = 32L; + + EXOFS_DBGMSG("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); + mapping->nrpages, start, end, expected_pages); _pcol_init(&pcol, expected_pages, mapping->host); @@ -773,19 +775,28 @@ static int exofs_get_block(struct inode *inode, sector_t iblock, 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) +{ + struct exofs_i_info *oi = exofs_i(inode); + loff_t isize = i_size_read(inode); + int ret; + + inode->i_mtime = inode->i_ctime = CURRENT_TIME; + + nobh_truncate_page(inode->i_mapping, isize, exofs_get_block); + + ret = exofs_oi_truncate(oi, (u64)isize); + EXOFS_DBGMSG("(0x%lx) size=0x%llx\n", inode->i_ino, isize); + 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_sb_info *sbi = inode->i_sb->s_fs_info; struct exofs_i_info *oi = exofs_i(inode); - struct osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF}; - struct osd_request *or; - struct osd_attr attr; - loff_t isize = i_size_read(inode); - __be64 newsize; int ret; if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) @@ -795,22 +806,6 @@ void exofs_truncate(struct inode *inode) return; if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) return; - inode->i_mtime = inode->i_ctime = CURRENT_TIME; - - nobh_truncate_page(inode->i_mapping, isize, exofs_get_block); - - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("ERROR: exofs_truncate: osd_start_request failed\n"); - goto fail; - } - - osd_req_set_attributes(or, &obj); - - newsize = cpu_to_be64((u64)isize); - attr = g_attr_logical_length; - attr.val_ptr = &newsize; - osd_req_add_set_attr_list(or, &attr, 1); /* if we are about to truncate an object, and it hasn't been * created yet, wait @@ -818,8 +813,7 @@ void exofs_truncate(struct inode *inode) if (unlikely(wait_obj_created(oi))) goto fail; - ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred); - osd_end_request(or); + ret = _do_truncate(inode); if (ret) goto fail; @@ -849,66 +843,57 @@ int exofs_setattr(struct dentry *dentry, struct iattr *iattr) /* * Read an inode from the OSD, and return it as is. We also return the size - * attribute in the 'sanity' argument if we got compiled with debugging turned - * on. + * attribute in the 'obj_size' argument. */ static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi, - struct exofs_fcb *inode, uint64_t *sanity) + struct exofs_fcb *inode, uint64_t *obj_size) { struct exofs_sb_info *sbi = sb->s_fs_info; - struct osd_request *or; - struct osd_attr attr; - struct osd_obj_id obj = {sbi->s_pid, - oi->vfs_inode.i_ino + EXOFS_OBJ_OFF}; + struct osd_attr attrs[2]; + struct exofs_io_state *ios; int ret; - exofs_make_credential(oi->i_cred, &obj); - - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("exofs_get_inode: osd_start_request failed.\n"); - return -ENOMEM; + *obj_size = ~0; + ret = exofs_get_io_state(sbi, &ios); + if (unlikely(ret)) { + EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__); + return ret; } - osd_req_get_attributes(or, &obj); - /* we need the inode attribute */ - osd_req_add_get_attr_list(or, &g_attr_inode_data, 1); + ios->obj.id = exofs_oi_objno(oi); + exofs_make_credential(oi->i_cred, &ios->obj); + ios->cred = oi->i_cred; -#ifdef EXOFS_DEBUG_OBJ_ISIZE - /* we get the size attributes to do a sanity check */ - osd_req_add_get_attr_list(or, &g_attr_logical_length, 1); -#endif + 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_sync_op(or, sbi->s_timeout, oi->i_cred); + ret = exofs_sbi_read(ios); if (ret) goto out; - attr = g_attr_inode_data; - ret = extract_attr_from_req(or, &attr); + ret = extract_attr_from_ios(ios, &attrs[0]); if (ret) { - EXOFS_ERR("exofs_get_inode: extract_attr_from_req failed\n"); + EXOFS_ERR("%s: extract_attr of inode_data failed\n", __func__); goto out; } + WARN_ON(attrs[0].len != EXOFS_INO_ATTR_SIZE); + memcpy(inode, attrs[0].val_ptr, EXOFS_INO_ATTR_SIZE); - WARN_ON(attr.len != EXOFS_INO_ATTR_SIZE); - memcpy(inode, attr.val_ptr, EXOFS_INO_ATTR_SIZE); - -#ifdef EXOFS_DEBUG_OBJ_ISIZE - attr = g_attr_logical_length; - ret = extract_attr_from_req(or, &attr); + ret = extract_attr_from_ios(ios, &attrs[1]); if (ret) { - EXOFS_ERR("ERROR: extract attr from or failed\n"); + EXOFS_ERR("%s: extract_attr of logical_length failed\n", + __func__); goto out; } - *sanity = get_unaligned_be64(attr.val_ptr); -#endif + *obj_size = get_unaligned_be64(attrs[1].val_ptr); out: - osd_end_request(or); + exofs_put_io_state(ios); return ret; } - static void __oi_init(struct exofs_i_info *oi) { init_waitqueue_head(&oi->i_wq); @@ -922,7 +907,7 @@ 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 uninitialized_var(sanity); + uint64_t obj_size; int ret; inode = iget_locked(sb, ino); @@ -934,7 +919,7 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino) __oi_init(oi); /* read the inode from the osd */ - ret = exofs_get_inode(sb, oi, &fcb, &sanity); + ret = exofs_get_inode(sb, oi, &fcb, &obj_size); if (ret) goto bad_inode; @@ -955,13 +940,12 @@ 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); -#ifdef EXOFS_DEBUG_OBJ_ISIZE - if ((inode->i_size != sanity) && + 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(sanity)); + inode->i_size, _LLU(obj_size)); + /* FIXME: call exofs_inode_recovery() */ } -#endif oi->i_dir_start_lookup = 0; @@ -1027,23 +1011,30 @@ int __exofs_wait_obj_created(struct exofs_i_info *oi) * set the obj_created flag so that other methods know that the object exists on * the OSD. */ -static void create_done(struct osd_request *or, void *p) +static void create_done(struct exofs_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_ok(or); - osd_end_request(or); + ret = exofs_check_io(ios, NULL); + exofs_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(sbi->s_pid), _LLU(inode->i_ino + EXOFS_OBJ_OFF)); - make_bad_inode(inode); - } else - set_obj_created(oi); + _LLU(exofs_oi_objno(oi)), _LLU(sbi->s_pid)); + /*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 + * succeed. All these dangling objects will be made into lost + * files by chkfs.exofs + */ + } + + set_obj_created(oi); atomic_dec(&inode->i_count); wake_up(&oi->i_wq); @@ -1058,8 +1049,7 @@ struct inode *exofs_new_inode(struct inode *dir, int mode) struct inode *inode; struct exofs_i_info *oi; struct exofs_sb_info *sbi; - struct osd_request *or; - struct osd_obj_id obj; + struct exofs_io_state *ios; int ret; sb = dir->i_sb; @@ -1096,28 +1086,28 @@ struct inode *exofs_new_inode(struct inode *dir, int mode) mark_inode_dirty(inode); - obj.partition = sbi->s_pid; - obj.id = inode->i_ino + EXOFS_OBJ_OFF; - exofs_make_credential(oi->i_cred, &obj); - - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("exofs_new_inode: osd_start_request failed\n"); - return ERR_PTR(-ENOMEM); + ret = exofs_get_io_state(sbi, &ios); + if (unlikely(ret)) { + EXOFS_ERR("exofs_new_inode: exofs_get_io_state failed\n"); + return ERR_PTR(ret); } - osd_req_create_object(or, &obj); + 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); - ret = exofs_async_op(or, create_done, inode, oi->i_cred); + ios->done = create_done; + ios->private = inode; + ios->cred = oi->i_cred; + ret = exofs_sbi_create(ios); if (ret) { atomic_dec(&inode->i_count); - osd_end_request(or); - return ERR_PTR(-EIO); + exofs_put_io_state(ios); + return ERR_PTR(ret); } atomic_inc(&sbi->s_curr_pending); @@ -1135,11 +1125,11 @@ struct updatei_args { /* * Callback function from exofs_update_inode(). */ -static void updatei_done(struct osd_request *or, void *p) +static void updatei_done(struct exofs_io_state *ios, void *p) { struct updatei_args *args = p; - osd_end_request(or); + exofs_put_io_state(ios); atomic_dec(&args->sbi->s_curr_pending); @@ -1155,8 +1145,7 @@ 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 osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF}; - struct osd_request *or; + struct exofs_io_state *ios; struct osd_attr attr; struct exofs_fcb *fcb; struct updatei_args *args; @@ -1193,18 +1182,16 @@ static int exofs_update_inode(struct inode *inode, int do_sync) } else memcpy(fcb->i_data, oi->i_data, sizeof(fcb->i_data)); - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("exofs_update_inode: osd_start_request failed.\n"); - ret = -ENOMEM; + ret = exofs_get_io_state(sbi, &ios); + if (unlikely(ret)) { + EXOFS_ERR("%s: exofs_get_io_state failed.\n", __func__); goto free_args; } - osd_req_set_attributes(or, &obj); - attr = g_attr_inode_data; attr.val_ptr = fcb; - osd_req_add_set_attr_list(or, &attr, 1); + ios->out_attr_len = 1; + ios->out_attr = &attr; if (!obj_created(oi)) { EXOFS_DBGMSG("!obj_created\n"); @@ -1213,22 +1200,19 @@ static int exofs_update_inode(struct inode *inode, int do_sync) EXOFS_DBGMSG("wait_event done\n"); } - if (do_sync) { - ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred); - osd_end_request(or); - goto free_args; - } else { + if (!do_sync) { args->sbi = sbi; + ios->done = updatei_done; + ios->private = args; + } - ret = exofs_async_op(or, updatei_done, args, oi->i_cred); - if (ret) { - osd_end_request(or); - goto free_args; - } + ret = exofs_oi_write(oi, ios); + if (!do_sync && !ret) { atomic_inc(&sbi->s_curr_pending); goto out; /* deallocation in updatei_done */ } + exofs_put_io_state(ios); free_args: kfree(args); out: @@ -1245,11 +1229,12 @@ int exofs_write_inode(struct inode *inode, int wait) * Callback function from exofs_delete_inode() - don't have much cleaning up to * do. */ -static void delete_done(struct osd_request *or, void *p) +static void delete_done(struct exofs_io_state *ios, void *p) { - struct exofs_sb_info *sbi; - osd_end_request(or); - sbi = p; + struct exofs_sb_info *sbi = p; + + exofs_put_io_state(ios); + atomic_dec(&sbi->s_curr_pending); } @@ -1263,8 +1248,7 @@ void exofs_delete_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 osd_obj_id obj = {sbi->s_pid, inode->i_ino + EXOFS_OBJ_OFF}; - struct osd_request *or; + struct exofs_io_state *ios; int ret; truncate_inode_pages(&inode->i_data, 0); @@ -1281,25 +1265,26 @@ void exofs_delete_inode(struct inode *inode) clear_inode(inode); - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("exofs_delete_inode: osd_start_request failed\n"); + ret = exofs_get_io_state(sbi, &ios); + if (unlikely(ret)) { + EXOFS_ERR("%s: exofs_get_io_state failed\n", __func__); return; } - osd_req_remove_object(or, &obj); - /* 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)); } - ret = exofs_async_op(or, delete_done, sbi, oi->i_cred); + ios->obj.id = exofs_oi_objno(oi); + ios->done = delete_done; + ios->private = sbi; + ios->cred = oi->i_cred; + ret = exofs_sbi_remove(ios); if (ret) { - EXOFS_ERR( - "ERROR: @exofs_delete_inode exofs_async_op failed\n"); - osd_end_request(or); + EXOFS_ERR("%s: exofs_sbi_remove failed\n", __func__); + exofs_put_io_state(ios); return; } atomic_inc(&sbi->s_curr_pending); diff --git a/fs/exofs/ios.c b/fs/exofs/ios.c index 4372542df284..bb2f9d341fdf 100644 --- a/fs/exofs/ios.c +++ b/fs/exofs/ios.c @@ -23,88 +23,327 @@ */ #include -#include #include "exofs.h" -int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid) -{ - struct osd_sense_info osi; - int ret = osd_req_decode_sense(or, &osi); - - if (ret) { /* translate to Linux codes */ - if (osi.additional_code == scsi_invalid_field_in_cdb) { - if (osi.cdb_field_offset == OSD_CFO_STARTING_BYTE) - ret = -EFAULT; - if (osi.cdb_field_offset == OSD_CFO_OBJECT_ID) - ret = -ENOENT; - else - ret = -EINVAL; - } else if (osi.additional_code == osd_quota_error) - ret = -ENOSPC; - else - ret = -EIO; - } - - /* FIXME: should be include in osd_sense_info */ - if (in_resid) - *in_resid = or->in.req ? or->in.req->resid_len : 0; - - if (out_resid) - *out_resid = or->out.req ? or->out.req->resid_len : 0; - - return ret; -} - void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj) { osd_sec_init_nosec_doall_caps(cred_a, obj, false, true); } -/* - * Perform a synchronous OSD operation. - */ -int exofs_sync_op(struct osd_request *or, int timeout, uint8_t *credential) +int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj, + u64 offset, void *p, unsigned length) { + struct osd_request *or = osd_start_request(od, GFP_KERNEL); +/* struct osd_sense_info osi = {.key = 0};*/ int ret; - or->timeout = timeout; - ret = osd_finalize_request(or, 0, credential, NULL); - if (ret) { + if (unlikely(!or)) { + EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__); + return -ENOMEM; + } + ret = osd_req_read_kern(or, obj, offset, p, length); + if (unlikely(ret)) { + EXOFS_DBGMSG("%s: osd_req_read_kern failed.\n", __func__); + goto out; + } + + ret = osd_finalize_request(or, 0, cred, NULL); + if (unlikely(ret)) { EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret); - return ret; + goto out; } ret = osd_execute_request(or); - - if (ret) + if (unlikely(ret)) EXOFS_DBGMSG("osd_execute_request() => %d\n", ret); /* osd_req_decode_sense(or, ret); */ + +out: + osd_end_request(or); return ret; } -/* - * Perform an asynchronous OSD operation. - */ -int exofs_async_op(struct osd_request *or, osd_req_done_fn *async_done, - void *caller_context, u8 *cred) +int exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** pios) { - int ret; + struct exofs_io_state *ios; - ret = osd_finalize_request(or, 0, cred, NULL); - if (ret) { - EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret); - return ret; + /*TODO: Maybe use kmem_cach per sbi of size + * exofs_io_state_size(sbi->s_numdevs) + */ + ios = kzalloc(exofs_io_state_size(1), GFP_KERNEL); + if (unlikely(!ios)) { + *pios = NULL; + return -ENOMEM; } - ret = osd_execute_request_async(or, async_done, caller_context); + ios->sbi = sbi; + ios->obj.partition = sbi->s_pid; + *pios = ios; + return 0; +} - if (ret) - EXOFS_DBGMSG("osd_execute_request_async() => %d\n", ret); +void exofs_put_io_state(struct exofs_io_state *ios) +{ + if (ios) { + unsigned i; + + for (i = 0; i < ios->numdevs; i++) { + struct exofs_per_dev_state *per_dev = &ios->per_dev[i]; + + if (per_dev->or) + osd_end_request(per_dev->or); + if (per_dev->bio) + bio_put(per_dev->bio); + } + + kfree(ios); + } +} + +static void _sync_done(struct exofs_io_state *ios, void *p) +{ + struct completion *waiting = p; + + complete(waiting); +} + +static void _last_io(struct kref *kref) +{ + struct exofs_io_state *ios = container_of( + kref, struct exofs_io_state, kref); + + ios->done(ios, ios->private); +} + +static void _done_io(struct osd_request *or, void *p) +{ + struct exofs_io_state *ios = p; + + kref_put(&ios->kref, _last_io); +} + +static int exofs_io_execute(struct exofs_io_state *ios) +{ + DECLARE_COMPLETION_ONSTACK(wait); + bool sync = (ios->done == NULL); + int i, ret; + + if (sync) { + ios->done = _sync_done; + ios->private = &wait; + } + + for (i = 0; i < ios->numdevs; i++) { + struct osd_request *or = ios->per_dev[i].or; + if (unlikely(!or)) + continue; + + ret = osd_finalize_request(or, 0, ios->cred, NULL); + if (unlikely(ret)) { + EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", + ret); + return ret; + } + } + + kref_init(&ios->kref); + + for (i = 0; i < ios->numdevs; i++) { + struct osd_request *or = ios->per_dev[i].or; + if (unlikely(!or)) + continue; + + kref_get(&ios->kref); + osd_execute_request_async(or, _done_io, ios); + } + + kref_put(&ios->kref, _last_io); + ret = 0; + + if (sync) { + wait_for_completion(&wait); + ret = exofs_check_io(ios, NULL); + } return ret; } -int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr) +int exofs_check_io(struct exofs_io_state *ios, u64 *resid) +{ + enum osd_err_priority acumulated_osd_err = 0; + int acumulated_lin_err = 0; + int i; + + for (i = 0; i < ios->numdevs; i++) { + struct osd_sense_info osi; + int ret = osd_req_decode_sense(ios->per_dev[i].or, &osi); + + if (likely(!ret)) + continue; + + if (unlikely(ret == -EFAULT)) { + EXOFS_DBGMSG("%s: EFAULT Need page clear\n", __func__); + /*FIXME: All the pages in this device range should: + * clear_highpage(page); + */ + } + + if (osi.osd_err_pri >= acumulated_osd_err) { + acumulated_osd_err = osi.osd_err_pri; + acumulated_lin_err = ret; + } + } + + /* TODO: raid specific residual calculations */ + if (resid) { + if (likely(!acumulated_lin_err)) + *resid = 0; + else + *resid = ios->length; + } + + return acumulated_lin_err; +} + +int exofs_sbi_create(struct exofs_io_state *ios) +{ + int i, ret; + + for (i = 0; i < 1; i++) { + struct osd_request *or; + + or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + if (unlikely(!or)) { + EXOFS_ERR("%s: osd_start_request failed\n", __func__); + ret = -ENOMEM; + goto out; + } + ios->per_dev[i].or = or; + ios->numdevs++; + + osd_req_create_object(or, &ios->obj); + } + ret = exofs_io_execute(ios); + +out: + return ret; +} + +int exofs_sbi_remove(struct exofs_io_state *ios) +{ + int i, ret; + + for (i = 0; i < 1; i++) { + struct osd_request *or; + + or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + if (unlikely(!or)) { + EXOFS_ERR("%s: osd_start_request failed\n", __func__); + ret = -ENOMEM; + goto out; + } + ios->per_dev[i].or = or; + ios->numdevs++; + + osd_req_remove_object(or, &ios->obj); + } + ret = exofs_io_execute(ios); + +out: + return ret; +} + +int exofs_sbi_write(struct exofs_io_state *ios) +{ + int i, ret; + + for (i = 0; i < 1; i++) { + struct osd_request *or; + + or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + if (unlikely(!or)) { + EXOFS_ERR("%s: osd_start_request failed\n", __func__); + ret = -ENOMEM; + goto out; + } + ios->per_dev[i].or = or; + ios->numdevs++; + + if (ios->bio) { + struct bio *bio; + + bio = ios->bio; + + osd_req_write(or, &ios->obj, ios->offset, bio, + ios->length); +/* EXOFS_DBGMSG("write sync=%d\n", sync);*/ + } else if (ios->kern_buff) { + osd_req_write_kern(or, &ios->obj, ios->offset, + ios->kern_buff, ios->length); +/* EXOFS_DBGMSG("write_kern sync=%d\n", sync);*/ + } else { + osd_req_set_attributes(or, &ios->obj); +/* EXOFS_DBGMSG("set_attributes sync=%d\n", sync);*/ + } + + if (ios->out_attr) + osd_req_add_set_attr_list(or, ios->out_attr, + ios->out_attr_len); + + if (ios->in_attr) + osd_req_add_get_attr_list(or, ios->in_attr, + ios->in_attr_len); + } + ret = exofs_io_execute(ios); + +out: + return ret; +} + +int exofs_sbi_read(struct exofs_io_state *ios) +{ + int i, ret; + + for (i = 0; i < 1; i++) { + struct osd_request *or; + + or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + if (unlikely(!or)) { + EXOFS_ERR("%s: osd_start_request failed\n", __func__); + ret = -ENOMEM; + goto out; + } + ios->per_dev[i].or = or; + ios->numdevs++; + + if (ios->bio) { + osd_req_read(or, &ios->obj, ios->offset, ios->bio, + ios->length); +/* EXOFS_DBGMSG("read sync=%d\n", sync);*/ + } else if (ios->kern_buff) { + osd_req_read_kern(or, &ios->obj, ios->offset, + ios->kern_buff, ios->length); +/* EXOFS_DBGMSG("read_kern sync=%d\n", sync);*/ + } else { + osd_req_get_attributes(or, &ios->obj); +/* EXOFS_DBGMSG("get_attributes sync=%d\n", sync);*/ + } + + if (ios->out_attr) + osd_req_add_set_attr_list(or, ios->out_attr, + ios->out_attr_len); + + if (ios->in_attr) + osd_req_add_get_attr_list(or, ios->in_attr, + ios->in_attr_len); + } + ret = exofs_io_execute(ios); + +out: + return ret; +} + +int extract_attr_from_ios(struct exofs_io_state *ios, struct osd_attr *attr) { struct osd_attr cur_attr = {.attr_page = 0}; /* start with zeros */ void *iter = NULL; @@ -112,7 +351,8 @@ int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr) do { nelem = 1; - osd_req_decode_get_attr_list(or, &cur_attr, &nelem, &iter); + osd_req_decode_get_attr_list(ios->per_dev[0].or, + &cur_attr, &nelem, &iter); if ((cur_attr.attr_page == attr->attr_page) && (cur_attr.attr_id == attr->attr_id)) { attr->len = cur_attr.len; @@ -123,3 +363,43 @@ int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr) return -EIO; } + +int exofs_oi_truncate(struct exofs_i_info *oi, u64 size) +{ + struct exofs_sb_info *sbi = oi->vfs_inode.i_sb->s_fs_info; + struct exofs_io_state *ios; + struct osd_attr attr; + __be64 newsize; + int i, ret; + + if (exofs_get_io_state(sbi, &ios)) + return -ENOMEM; + + ios->obj.id = exofs_oi_objno(oi); + ios->cred = oi->i_cred; + + newsize = cpu_to_be64(size); + attr = g_attr_logical_length; + attr.val_ptr = &newsize; + + for (i = 0; i < 1; i++) { + struct osd_request *or; + + or = osd_start_request(sbi->s_dev, GFP_KERNEL); + if (unlikely(!or)) { + EXOFS_ERR("%s: osd_start_request failed\n", __func__); + ret = -ENOMEM; + goto out; + } + ios->per_dev[i].or = or; + ios->numdevs++; + + osd_req_set_attributes(or, &ios->obj); + osd_req_add_set_attr_list(or, &attr, 1); + } + ret = exofs_io_execute(ios); + +out: + exofs_put_io_state(ios); + return ret; +} diff --git a/fs/exofs/super.c b/fs/exofs/super.c index 28add3eac0a4..4cd97f526d49 100644 --- a/fs/exofs/super.c +++ b/fs/exofs/super.c @@ -203,49 +203,40 @@ int exofs_sync_fs(struct super_block *sb, int wait) { struct exofs_sb_info *sbi; struct exofs_fscb *fscb; - struct osd_request *or; - struct osd_obj_id obj; + struct exofs_io_state *ios; int ret = -ENOMEM; - fscb = kzalloc(sizeof(struct exofs_fscb), GFP_KERNEL); - if (!fscb) { - EXOFS_ERR("exofs_write_super: memory allocation failed.\n"); - return -ENOMEM; - } - lock_super(sb); sbi = sb->s_fs_info; + fscb = &sbi->s_fscb; + + ret = exofs_get_io_state(sbi, &ios); + if (ret) + goto out; + + ios->length = sizeof(*fscb); + memset(fscb, 0, ios->length); fscb->s_nextid = cpu_to_le64(sbi->s_nextid); fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles); fscb->s_magic = cpu_to_le16(sb->s_magic); fscb->s_newfs = 0; - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_ERR("exofs_write_super: osd_start_request failed.\n"); - goto out; - } + ios->obj.id = EXOFS_SUPER_ID; + ios->offset = 0; + ios->kern_buff = fscb; + ios->cred = sbi->s_cred; - obj.partition = sbi->s_pid; - obj.id = EXOFS_SUPER_ID; - ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb)); + ret = exofs_sbi_write(ios); if (unlikely(ret)) { - EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n"); - goto out; - } - - ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred); - if (unlikely(ret)) { - EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n"); + EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__); goto out; } sb->s_dirt = 0; out: - if (or) - osd_end_request(or); + EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret); + exofs_put_io_state(ios); unlock_super(sb); - kfree(fscb); return ret; } @@ -302,24 +293,23 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) struct inode *root; struct exofs_mountopt *opts = data; struct exofs_sb_info *sbi; /*extended info */ + struct osd_dev *od; /* Master device */ struct exofs_fscb fscb; /*on-disk superblock info */ - struct osd_request *or = NULL; struct osd_obj_id obj; int ret; sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); if (!sbi) return -ENOMEM; - sb->s_fs_info = sbi; /* use mount options to fill superblock */ - sbi->s_dev = osduld_path_lookup(opts->dev_name); - if (IS_ERR(sbi->s_dev)) { - ret = PTR_ERR(sbi->s_dev); - sbi->s_dev = NULL; + od = osduld_path_lookup(opts->dev_name); + if (IS_ERR(od)) { + ret = PTR_ERR(od); goto free_sbi; } + sbi->s_dev = od; sbi->s_pid = opts->pid; sbi->s_timeout = opts->timeout; @@ -333,35 +323,13 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) sb->s_bdev = NULL; sb->s_dev = 0; - /* read data from on-disk superblock object */ obj.partition = sbi->s_pid; obj.id = EXOFS_SUPER_ID; exofs_make_credential(sbi->s_cred, &obj); - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - if (!silent) - EXOFS_ERR( - "exofs_fill_super: osd_start_request failed.\n"); - ret = -ENOMEM; + ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb)); + if (unlikely(ret)) goto free_sbi; - } - ret = osd_req_read_kern(or, &obj, 0, &fscb, sizeof(fscb)); - if (unlikely(ret)) { - if (!silent) - EXOFS_ERR( - "exofs_fill_super: osd_req_read_kern failed.\n"); - ret = -ENOMEM; - goto free_sbi; - } - - ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred); - if (unlikely(ret)) { - if (!silent) - EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n"); - ret = -EIO; - goto free_sbi; - } sb->s_magic = le16_to_cpu(fscb.s_magic); sbi->s_nextid = le64_to_cpu(fscb.s_nextid); @@ -380,6 +348,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) spin_lock_init(&sbi->s_next_gen_lock); /* set up operation vectors */ + sb->s_fs_info = sbi; sb->s_op = &exofs_sops; sb->s_export_op = &exofs_export_ops; root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF); @@ -406,16 +375,14 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) } _exofs_print_device("Mounting", opts->dev_name, sbi->s_dev, sbi->s_pid); - ret = 0; -out: - if (or) - osd_end_request(or); - return ret; + return 0; free_sbi: + EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n", + opts->dev_name, sbi->s_pid, ret); osduld_put_device(sbi->s_dev); /* NULL safe */ kfree(sbi); - goto out; + return ret; } /* @@ -444,7 +411,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) { struct super_block *sb = dentry->d_sb; struct exofs_sb_info *sbi = sb->s_fs_info; - struct osd_obj_id obj = {sbi->s_pid, 0}; + struct exofs_io_state *ios; struct osd_attr attrs[] = { ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS, OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)), @@ -453,26 +420,25 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) }; uint64_t capacity = ULLONG_MAX; uint64_t used = ULLONG_MAX; - struct osd_request *or; uint8_t cred_a[OSD_CAP_LEN]; int ret; - /* get used/capacity attributes */ - exofs_make_credential(cred_a, &obj); - - or = osd_start_request(sbi->s_dev, GFP_KERNEL); - if (unlikely(!or)) { - EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n"); - return -ENOMEM; + ret = exofs_get_io_state(sbi, &ios); + if (ret) { + EXOFS_DBGMSG("exofs_get_io_state failed.\n"); + return ret; } - osd_req_get_attributes(or, &obj); - osd_req_add_get_attr_list(or, attrs, ARRAY_SIZE(attrs)); - ret = exofs_sync_op(or, sbi->s_timeout, cred_a); + exofs_make_credential(cred_a, &ios->obj); + ios->cred = sbi->s_cred; + ios->in_attr = attrs; + ios->in_attr_len = ARRAY_SIZE(attrs); + + ret = exofs_sbi_read(ios); if (unlikely(ret)) goto out; - ret = extract_attr_from_req(or, &attrs[0]); + ret = extract_attr_from_ios(ios, &attrs[0]); if (likely(!ret)) { capacity = get_unaligned_be64(attrs[0].val_ptr); if (unlikely(!capacity)) @@ -480,7 +446,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) } else EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n"); - ret = extract_attr_from_req(or, &attrs[1]); + ret = extract_attr_from_ios(ios, &attrs[1]); if (likely(!ret)) used = get_unaligned_be64(attrs[1].val_ptr); else @@ -497,7 +463,7 @@ static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) buf->f_namelen = EXOFS_NAME_LEN; out: - osd_end_request(or); + exofs_put_io_state(ios); return ret; } From 04dc1e88ad9c9f9639019e9646a89ce0ebf706bb Mon Sep 17 00:00:00 2001 From: Boaz Harrosh Date: Mon, 16 Nov 2009 16:03:05 +0200 Subject: [PATCH 9/9] exofs: Multi-device mirror support This patch changes on-disk format, it is accompanied with a parallel patch to mkfs.exofs that enables multi-device capabilities. After this patch, old exofs will refuse to mount a new formatted FS and new exofs will refuse an old format. This is done by moving the magic field offset inside the FSCB. A new FSCB *version* field was added. In the future, exofs will refuse to mount unmatched FSCB version. To up-grade or down-grade an exofs one must use mkfs.exofs --upgrade option before mounting. Introduced, a new object that contains a *device-table*. This object contains the default *data-map* and a linear array of devices information, which identifies the devices used in the filesystem. This object is only written to offline by mkfs.exofs. This is why it is kept separate from the FSCB, since the later is written to while mounted. Same partition number, same object number is used on all devices only the device varies. * define the new format, then load the device table on mount time make sure every thing is supported. * Change I/O engine to now support Mirror IO, .i.e write same data to multiple devices, read from a random device to spread the read-load from multiple clients (TODO: stripe read) Implementation notes: A few points introduced in previous patch should be mentioned here: * Special care was made so absolutlly all operation that have any chance of failing are done before any osd-request is executed. This is to minimize the need for a data consistency recovery, to only real IO errors. * Each IO state has a kref. It starts at 1, any osd-request executed will increment the kref, finally when all are executed the first ref is dropped. At IO-done, each request completion decrements the kref, the last one to return executes the internal _last_io() routine. _last_io() will call the registered io_state_done. On sync mode a caller does not supply a done method, indicating a synchronous request, the caller is put to sleep and a special io_state_done is registered that will awaken the caller. Though also in sync mode all operations are executed in parallel. Signed-off-by: Boaz Harrosh --- fs/exofs/common.h | 63 +++++++++++-- fs/exofs/exofs.h | 12 ++- fs/exofs/inode.c | 5 +- fs/exofs/ios.c | 38 +++++--- fs/exofs/pnfs.h | 51 +++++++++++ fs/exofs/super.c | 220 ++++++++++++++++++++++++++++++++++++++++++++-- 6 files changed, 361 insertions(+), 28 deletions(-) create mode 100644 fs/exofs/pnfs.h diff --git a/fs/exofs/common.h b/fs/exofs/common.h index ce1c71692599..b1b178e61718 100644 --- a/fs/exofs/common.h +++ b/fs/exofs/common.h @@ -49,6 +49,7 @@ #define EXOFS_MIN_PID 0x10000 /* Smallest partition ID */ #define EXOFS_OBJ_OFF 0x10000 /* offset for objects */ #define EXOFS_SUPER_ID 0x10000 /* object ID for on-disk superblock */ +#define EXOFS_DEVTABLE_ID 0x10001 /* object ID for on-disk device table */ #define EXOFS_ROOT_ID 0x10002 /* object ID for root directory */ /* exofs Application specific page/attribute */ @@ -78,17 +79,67 @@ enum { #define EXOFS_SUPER_MAGIC 0x5DF5 /* - * The file system control block - stored in an object's data (mainly, the one - * with ID EXOFS_SUPER_ID). This is where the in-memory superblock is stored - * on disk. Right now it just has a magic value, which is basically a sanity - * check on our ability to communicate with the object store. + * The file system control block - stored in object EXOFS_SUPER_ID's data. + * This is where the in-memory superblock is stored on disk. */ +enum {EXOFS_FSCB_VER = 1, EXOFS_DT_VER = 1}; struct exofs_fscb { __le64 s_nextid; /* Highest object ID used */ - __le32 s_numfiles; /* Number of files on fs */ + __le64 s_numfiles; /* Number of files on fs */ + __le32 s_version; /* == EXOFS_FSCB_VER */ __le16 s_magic; /* Magic signature */ __le16 s_newfs; /* Non-zero if this is a new fs */ -}; + + /* From here on it's a static part, only written by mkexofs */ + __le64 s_dev_table_oid; /* Resurved, not used */ + __le64 s_dev_table_count; /* == 0 means no dev_table */ +} __packed; + +/* + * Describes the raid used in the FS. It is part of the device table. + * This here is taken from the pNFS-objects definition. In exofs we + * use one raid policy through-out the filesystem. (NOTE: the funny + * alignment at begining. We take care of it at exofs_device_table. + */ +struct exofs_dt_data_map { + __le32 cb_num_comps; + __le64 cb_stripe_unit; + __le32 cb_group_width; + __le32 cb_group_depth; + __le32 cb_mirror_cnt; + __le32 cb_raid_algorithm; +} __packed; + +/* + * This is an osd device information descriptor. It is a single entry in + * the exofs device table. It describes an osd target lun which + * contains data belonging to this FS. (Same partition_id on all devices) + */ +struct exofs_dt_device_info { + __le32 systemid_len; + u8 systemid[OSD_SYSTEMID_LEN]; + __le64 long_name_offset; /* If !0 then offset-in-file */ + __le32 osdname_len; /* */ + u8 osdname[44]; /* Embbeded, Ususally an asci uuid */ +} __packed; + +/* + * The EXOFS device table - stored in object EXOFS_DEVTABLE_ID's data. + * It contains the raid used for this multy-device FS and an array of + * participating devices. + */ +struct exofs_device_table { + __le32 dt_version; /* == EXOFS_DT_VER */ + struct exofs_dt_data_map dt_data_map; /* Raid policy to use */ + + /* Resurved space For future use. Total includeing this: + * (8 * sizeof(le64)) + */ + __le64 __Resurved[4]; + + __le64 dt_num_devices; /* Array size */ + struct exofs_dt_device_info dt_dev_table[]; /* Array of devices */ +} __packed; /**************************************************************************** * inode-related things diff --git a/fs/exofs/exofs.h b/fs/exofs/exofs.h index 2e08859a89e8..c35fd4623986 100644 --- a/fs/exofs/exofs.h +++ b/fs/exofs/exofs.h @@ -37,6 +37,11 @@ #include #include "common.h" +/* FIXME: Remove once pnfs hits mainline + * #include + */ +#include "pnfs.h" + #define EXOFS_ERR(fmt, a...) printk(KERN_ERR "exofs: " fmt, ##a) #ifdef CONFIG_EXOFS_DEBUG @@ -54,7 +59,6 @@ * our extension to the in-memory superblock */ struct exofs_sb_info { - struct osd_dev *s_dev; /* returned by get_osd_dev */ struct exofs_fscb s_fscb; /* Written often, pre-allocate*/ osd_id s_pid; /* partition ID of file system*/ int s_timeout; /* timeout for OSD operations */ @@ -63,7 +67,11 @@ struct exofs_sb_info { spinlock_t s_next_gen_lock; /* spinlock for gen # update */ u32 s_next_generation; /* next gen # to use */ atomic_t s_curr_pending; /* number of pending commands */ - uint8_t s_cred[OSD_CAP_LEN]; /* all-powerful credential */ + uint8_t s_cred[OSD_CAP_LEN]; /* credential for the fscb */ + + struct pnfs_osd_data_map data_map; /* Default raid to use */ + unsigned s_numdevs; /* Num of devices in array */ + struct osd_dev *s_ods[1]; /* Variable length, minimum 1 */ }; /* diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c index 7578950fd135..698a8636d39c 100644 --- a/fs/exofs/inode.c +++ b/fs/exofs/inode.c @@ -62,7 +62,10 @@ 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; - pcol->req_q = osd_request_queue(sbi->s_dev); + /* 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; diff --git a/fs/exofs/ios.c b/fs/exofs/ios.c index bb2f9d341fdf..5bad01fa1f9f 100644 --- a/fs/exofs/ios.c +++ b/fs/exofs/ios.c @@ -71,7 +71,7 @@ int exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** pios) /*TODO: Maybe use kmem_cach per sbi of size * exofs_io_state_size(sbi->s_numdevs) */ - ios = kzalloc(exofs_io_state_size(1), GFP_KERNEL); + ios = kzalloc(exofs_io_state_size(sbi->s_numdevs), GFP_KERNEL); if (unlikely(!ios)) { *pios = NULL; return -ENOMEM; @@ -209,10 +209,10 @@ int exofs_sbi_create(struct exofs_io_state *ios) { int i, ret; - for (i = 0; i < 1; i++) { + for (i = 0; i < ios->sbi->s_numdevs; i++) { struct osd_request *or; - or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + or = osd_start_request(ios->sbi->s_ods[i], GFP_KERNEL); if (unlikely(!or)) { EXOFS_ERR("%s: osd_start_request failed\n", __func__); ret = -ENOMEM; @@ -233,10 +233,10 @@ int exofs_sbi_remove(struct exofs_io_state *ios) { int i, ret; - for (i = 0; i < 1; i++) { + for (i = 0; i < ios->sbi->s_numdevs; i++) { struct osd_request *or; - or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + or = osd_start_request(ios->sbi->s_ods[i], GFP_KERNEL); if (unlikely(!or)) { EXOFS_ERR("%s: osd_start_request failed\n", __func__); ret = -ENOMEM; @@ -257,10 +257,10 @@ int exofs_sbi_write(struct exofs_io_state *ios) { int i, ret; - for (i = 0; i < 1; i++) { + for (i = 0; i < ios->sbi->s_numdevs; i++) { struct osd_request *or; - or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + or = osd_start_request(ios->sbi->s_ods[i], GFP_KERNEL); if (unlikely(!or)) { EXOFS_ERR("%s: osd_start_request failed\n", __func__); ret = -ENOMEM; @@ -272,7 +272,21 @@ int exofs_sbi_write(struct exofs_io_state *ios) if (ios->bio) { struct bio *bio; - bio = ios->bio; + if (i != 0) { + bio = bio_kmalloc(GFP_KERNEL, + ios->bio->bi_max_vecs); + if (unlikely(!bio)) { + ret = -ENOMEM; + goto out; + } + + __bio_clone(bio, ios->bio); + bio->bi_bdev = NULL; + bio->bi_next = NULL; + ios->per_dev[i].bio = bio; + } else { + bio = ios->bio; + } osd_req_write(or, &ios->obj, ios->offset, bio, ios->length); @@ -306,8 +320,10 @@ int exofs_sbi_read(struct exofs_io_state *ios) for (i = 0; i < 1; i++) { struct osd_request *or; + unsigned first_dev = (unsigned)ios->obj.id; - or = osd_start_request(ios->sbi->s_dev, GFP_KERNEL); + first_dev %= ios->sbi->s_numdevs; + or = osd_start_request(ios->sbi->s_ods[first_dev], GFP_KERNEL); if (unlikely(!or)) { EXOFS_ERR("%s: osd_start_request failed\n", __func__); ret = -ENOMEM; @@ -382,10 +398,10 @@ int exofs_oi_truncate(struct exofs_i_info *oi, u64 size) attr = g_attr_logical_length; attr.val_ptr = &newsize; - for (i = 0; i < 1; i++) { + for (i = 0; i < sbi->s_numdevs; i++) { struct osd_request *or; - or = osd_start_request(sbi->s_dev, GFP_KERNEL); + or = osd_start_request(sbi->s_ods[i], GFP_KERNEL); if (unlikely(!or)) { EXOFS_ERR("%s: osd_start_request failed\n", __func__); ret = -ENOMEM; diff --git a/fs/exofs/pnfs.h b/fs/exofs/pnfs.h new file mode 100644 index 000000000000..423033addd1f --- /dev/null +++ b/fs/exofs/pnfs.h @@ -0,0 +1,51 @@ +/* + * Copyright (C) 2008, 2009 + * Boaz Harrosh + * + * This file is part of exofs. + * + * exofs is free software; you can redistribute it and/or modify it under the + * terms of the GNU General Public License version 2 as published by the Free + * Software Foundation. + * + */ + +/* FIXME: Remove this file once pnfs hits mainline */ + +#ifndef __EXOFS_PNFS_H__ +#define __EXOFS_PNFS_H__ + +#if defined(CONFIG_PNFS) + + +/* FIXME: move this file to: linux/exportfs/pnfs_osd_xdr.h */ +#include "../nfs/objlayout/pnfs_osd_xdr.h" + +#else /* defined(CONFIG_PNFS) */ + +enum pnfs_iomode { + IOMODE_READ = 1, + IOMODE_RW = 2, + IOMODE_ANY = 3, +}; + +/* Layout Structure */ +enum pnfs_osd_raid_algorithm4 { + PNFS_OSD_RAID_0 = 1, + PNFS_OSD_RAID_4 = 2, + PNFS_OSD_RAID_5 = 3, + PNFS_OSD_RAID_PQ = 4 /* Reed-Solomon P+Q */ +}; + +struct pnfs_osd_data_map { + u32 odm_num_comps; + u64 odm_stripe_unit; + u32 odm_group_width; + u32 odm_group_depth; + u32 odm_mirror_cnt; + u32 odm_raid_algorithm; +}; + +#endif /* else defined(CONFIG_PNFS) */ + +#endif /* __EXOFS_PNFS_H__ */ diff --git a/fs/exofs/super.c b/fs/exofs/super.c index 4cd97f526d49..a1d1e77b12eb 100644 --- a/fs/exofs/super.c +++ b/fs/exofs/super.c @@ -214,12 +214,17 @@ int exofs_sync_fs(struct super_block *sb, int wait) if (ret) goto out; - ios->length = sizeof(*fscb); + /* Note: We only write the changing part of the fscb. .i.e upto the + * the fscb->s_dev_table_oid member. There is no read-modify-write + * here. + */ + ios->length = offsetof(struct exofs_fscb, s_dev_table_oid); memset(fscb, 0, ios->length); fscb->s_nextid = cpu_to_le64(sbi->s_nextid); fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles); fscb->s_magic = cpu_to_le16(sb->s_magic); fscb->s_newfs = 0; + fscb->s_version = EXOFS_FSCB_VER; ios->obj.id = EXOFS_SUPER_ID; ios->offset = 0; @@ -257,6 +262,20 @@ static void _exofs_print_device(const char *msg, const char *dev_path, msg, dev_path ?: "", odi->osdname, _LLU(pid)); } +void exofs_free_sbi(struct exofs_sb_info *sbi) +{ + while (sbi->s_numdevs) { + int i = --sbi->s_numdevs; + struct osd_dev *od = sbi->s_ods[i]; + + if (od) { + sbi->s_ods[i] = NULL; + osduld_put_device(od); + } + } + kfree(sbi); +} + /* * This function is called when the vfs is freeing the superblock. We just * need to free our own part. @@ -279,12 +298,182 @@ static void exofs_put_super(struct super_block *sb) msecs_to_jiffies(100)); } - _exofs_print_device("Unmounting", NULL, sbi->s_dev, sbi->s_pid); - osduld_put_device(sbi->s_dev); - kfree(sb->s_fs_info); + _exofs_print_device("Unmounting", NULL, sbi->s_ods[0], sbi->s_pid); + + exofs_free_sbi(sbi); sb->s_fs_info = NULL; } +static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs, + struct exofs_device_table *dt) +{ + sbi->data_map.odm_num_comps = + le32_to_cpu(dt->dt_data_map.cb_num_comps); + sbi->data_map.odm_stripe_unit = + le64_to_cpu(dt->dt_data_map.cb_stripe_unit); + sbi->data_map.odm_group_width = + le32_to_cpu(dt->dt_data_map.cb_group_width); + sbi->data_map.odm_group_depth = + le32_to_cpu(dt->dt_data_map.cb_group_depth); + sbi->data_map.odm_mirror_cnt = + le32_to_cpu(dt->dt_data_map.cb_mirror_cnt); + sbi->data_map.odm_raid_algorithm = + le32_to_cpu(dt->dt_data_map.cb_raid_algorithm); + +/* FIXME: Hard coded mirror only for now. if not so do not mount */ + if ((sbi->data_map.odm_num_comps != numdevs) || + (sbi->data_map.odm_stripe_unit != EXOFS_BLKSIZE) || + (sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) || + (sbi->data_map.odm_mirror_cnt != (numdevs - 1))) + return -EINVAL; + else + return 0; +} + +/* @odi is valid only as long as @fscb_dev is valid */ +static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev, + struct osd_dev_info *odi) +{ + odi->systemid_len = le32_to_cpu(dt_dev->systemid_len); + memcpy(odi->systemid, dt_dev->systemid, odi->systemid_len); + + odi->osdname_len = le32_to_cpu(dt_dev->osdname_len); + odi->osdname = dt_dev->osdname; + + /* FIXME support long names. Will need a _put function */ + if (dt_dev->long_name_offset) + return -EINVAL; + + /* Make sure osdname is printable! + * mkexofs should give us space for a null-terminator else the + * device-table is invalid. + */ + if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname))) + odi->osdname_len = sizeof(dt_dev->osdname) - 1; + dt_dev->osdname[odi->osdname_len] = 0; + + /* If it's all zeros something is bad we read past end-of-obj */ + return !(odi->systemid_len || odi->osdname_len); +} + +static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi, + unsigned table_count) +{ + struct exofs_sb_info *sbi = *psbi; + struct osd_dev *fscb_od; + struct osd_obj_id obj = {.partition = sbi->s_pid, + .id = EXOFS_DEVTABLE_ID}; + struct exofs_device_table *dt; + unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) + + sizeof(*dt); + unsigned numdevs, i; + int ret; + + dt = kmalloc(table_bytes, GFP_KERNEL); + if (unlikely(!dt)) { + EXOFS_ERR("ERROR: allocating %x bytes for device table\n", + table_bytes); + return -ENOMEM; + } + + fscb_od = sbi->s_ods[0]; + sbi->s_ods[0] = NULL; + sbi->s_numdevs = 0; + ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes); + if (unlikely(ret)) { + EXOFS_ERR("ERROR: reading device table\n"); + goto out; + } + + numdevs = le64_to_cpu(dt->dt_num_devices); + if (unlikely(!numdevs)) { + ret = -EINVAL; + goto out; + } + WARN_ON(table_count != numdevs); + + ret = _read_and_match_data_map(sbi, numdevs, dt); + if (unlikely(ret)) + goto out; + + if (likely(numdevs > 1)) { + unsigned size = numdevs * sizeof(sbi->s_ods[0]); + + sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL); + if (unlikely(!sbi)) { + ret = -ENOMEM; + goto out; + } + memset(&sbi->s_ods[1], 0, size - sizeof(sbi->s_ods[0])); + *psbi = sbi; + } + + for (i = 0; i < numdevs; i++) { + struct exofs_fscb fscb; + struct osd_dev_info odi; + struct osd_dev *od; + + if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) { + EXOFS_ERR("ERROR: Read all-zeros device entry\n"); + ret = -EINVAL; + goto out; + } + + printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n", + i, odi.osdname); + + /* On all devices the device table is identical. The user can + * specify any one of the participating devices on the command + * line. We always keep them in device-table order. + */ + if (fscb_od && osduld_device_same(fscb_od, &odi)) { + sbi->s_ods[i] = fscb_od; + ++sbi->s_numdevs; + fscb_od = NULL; + continue; + } + + od = osduld_info_lookup(&odi); + if (unlikely(IS_ERR(od))) { + ret = PTR_ERR(od); + EXOFS_ERR("ERROR: device requested is not found " + "osd_name-%s =>%d\n", odi.osdname, ret); + goto out; + } + + sbi->s_ods[i] = od; + ++sbi->s_numdevs; + + /* Read the fscb of the other devices to make sure the FS + * partition is there. + */ + ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, + sizeof(fscb)); + if (unlikely(ret)) { + EXOFS_ERR("ERROR: Malformed participating device " + "error reading fscb osd_name-%s\n", + odi.osdname); + goto out; + } + + /* TODO: verify other information is correct and FS-uuid + * matches. Benny what did you say about device table + * generation and old devices? + */ + } + +out: + kfree(dt); + if (unlikely(!ret && fscb_od)) { + EXOFS_ERR( + "ERROR: Bad device-table container device not present\n"); + osduld_put_device(fscb_od); + ret = -EINVAL; + } + + return ret; +} + /* * Read the superblock from the OSD and fill in the fields */ @@ -296,6 +485,7 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) struct osd_dev *od; /* Master device */ struct exofs_fscb fscb; /*on-disk superblock info */ struct osd_obj_id obj; + unsigned table_count; int ret; sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); @@ -309,7 +499,8 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) goto free_sbi; } - sbi->s_dev = od; + sbi->s_ods[0] = od; + sbi->s_numdevs = 1; sbi->s_pid = opts->pid; sbi->s_timeout = opts->timeout; @@ -342,11 +533,24 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) ret = -EINVAL; goto free_sbi; } + if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) { + EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n", + EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version)); + ret = -EINVAL; + goto free_sbi; + } /* start generation numbers from a random point */ get_random_bytes(&sbi->s_next_generation, sizeof(u32)); spin_lock_init(&sbi->s_next_gen_lock); + table_count = le64_to_cpu(fscb.s_dev_table_count); + if (table_count) { + ret = exofs_read_lookup_dev_table(&sbi, table_count); + if (unlikely(ret)) + goto free_sbi; + } + /* set up operation vectors */ sb->s_fs_info = sbi; sb->s_op = &exofs_sops; @@ -374,14 +578,14 @@ static int exofs_fill_super(struct super_block *sb, void *data, int silent) goto free_sbi; } - _exofs_print_device("Mounting", opts->dev_name, sbi->s_dev, sbi->s_pid); + _exofs_print_device("Mounting", opts->dev_name, sbi->s_ods[0], + sbi->s_pid); return 0; free_sbi: EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n", opts->dev_name, sbi->s_pid, ret); - osduld_put_device(sbi->s_dev); /* NULL safe */ - kfree(sbi); + exofs_free_sbi(sbi); return ret; }