exofs: Move all operations to an io_engine
Boaz Harrosh [Sun, 8 Nov 2009 12:54:08 +0000 (14:54 +0200)]
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 <bharrosh@panasas.com>

fs/exofs/common.h
fs/exofs/exofs.h
fs/exofs/inode.c
fs/exofs/ios.c
fs/exofs/super.c

index c6718e4..ce1c716 100644 (file)
@@ -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__*/
index 5ec72e0..2e08859 100644 (file)
  * 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 <linux/fs.h>
 #include <linux/time.h>
 #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;
index 7bc71a7..7578950 100644 (file)
 
 #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);
index 4372542..bb2f9d3 100644 (file)
  */
 
 #include <scsi/scsi_device.h>
-#include <scsi/osd_sense.h>
 
 #include "exofs.h"
 
-int exofs_check_ok_resid(struct osd_request *or, u64 *in_resid, u64 *out_resid)
+void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj)
 {
-       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;
+       osd_sec_init_nosec_doall_caps(cred_a, obj, false, true);
+}
+
+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;
+
+       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;
        }
 
-       /* FIXME: should be include in osd_sense_info */
-       if (in_resid)
-               *in_resid = or->in.req ? or->in.req->resid_len : 0;
+       ret = osd_finalize_request(or, 0, cred, NULL);
+       if (unlikely(ret)) {
+               EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret);
+               goto out;
+       }
 
-       if (out_resid)
-               *out_resid = or->out.req ? or->out.req->resid_len : 0;
+       ret = osd_execute_request(or);
+       if (unlikely(ret))
+               EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
+       /* osd_req_decode_sense(or, ret); */
 
+out:
+       osd_end_request(or);
        return ret;
 }
 
-void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj)
+int exofs_get_io_state(struct exofs_sb_info *sbi, struct exofs_io_state** pios)
 {
-       osd_sec_init_nosec_doall_caps(cred_a, obj, false, true);
+       struct exofs_io_state *ios;
+
+       /*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;
+       }
+
+       ios->sbi = sbi;
+       ios->obj.partition = sbi->s_pid;
+       *pios = ios;
+       return 0;
 }
 
-/*
- * Perform a synchronous OSD operation.
- */
-int exofs_sync_op(struct osd_request *or, int timeout, uint8_t *credential)
+void exofs_put_io_state(struct exofs_io_state *ios)
 {
-       int ret;
+       if (ios) {
+               unsigned i;
 
-       or->timeout = timeout;
-       ret = osd_finalize_request(or, 0, credential, NULL);
-       if (ret) {
-               EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret);
-               return ret;
+               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);
        }
+}
 
-       ret = osd_execute_request(or);
+static void _sync_done(struct exofs_io_state *ios, void *p)
+{
+       struct completion *waiting = p;
 
-       if (ret)
-               EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
-       /* osd_req_decode_sense(or, ret); */
+       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;
 }
 
-/*
- * 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_check_io(struct exofs_io_state *ios, u64 *resid)
 {
-       int ret;
+       enum osd_err_priority acumulated_osd_err = 0;
+       int acumulated_lin_err = 0;
+       int i;
 
-       ret = osd_finalize_request(or, 0, cred, NULL);
-       if (ret) {
-               EXOFS_DBGMSG("Faild to osd_finalize_request() => %d\n", ret);
-               return ret;
+       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);
 
-       ret = osd_execute_request_async(or, async_done, caller_context);
+               if (ios->in_attr)
+                       osd_req_add_get_attr_list(or, ios->in_attr,
+                                                 ios->in_attr_len);
+       }
+       ret = exofs_io_execute(ios);
 
-       if (ret)
-               EXOFS_DBGMSG("osd_execute_request_async() => %d\n", ret);
+out:
        return ret;
 }
 
-int extract_attr_from_req(struct osd_request *or, struct osd_attr *attr)
+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;
+}
index 28add3e..4cd97f5 100644 (file)
@@ -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;
-       }
-
-       obj.partition = sbi->s_pid;
-       obj.id = EXOFS_SUPER_ID;
-       ret = osd_req_write_kern(or, &obj, 0, fscb, sizeof(*fscb));
-       if (unlikely(ret)) {
-               EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
-               goto out;
-       }
+       ios->obj.id = EXOFS_SUPER_ID;
+       ios->offset = 0;
+       ios->kern_buff = fscb;
+       ios->cred = sbi->s_cred;
 
-       ret = exofs_sync_op(or, sbi->s_timeout, sbi->s_cred);
+       ret = exofs_sbi_write(ios);
        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;
-               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;
+       ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb));
+       if (unlikely(ret))
                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;
 }