1 files changed, 300 insertions, 247 deletions

diff --git a/fs/exofs/inode.c b/fs/exofs/inode.c
index 6c10f7476699..76d2a79ef93e 100644
--- a/fs/exofs/inode.c
+++ b/fs/exofs/inode.c
@@ -31,94 +31,117 @@
  * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
  */
 
+#include <linux/slab.h>
 #include <linux/writeback.h>
 #include <linux/buffer_head.h>
 #include <scsi/scsi_device.h>
 
 #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),
+        MAX_PAGES_KMALLOC =
+                PAGE_SIZE / sizeof(struct page *),
+};
 
 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;
+        struct page **pages;
+        unsigned alloc_pages;
         unsigned nr_pages;
         unsigned long length;
         loff_t pg_first; /* keep 64bit also in 32-arches */
 };
 
 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;
 
         pcol->sbi = sbi;
-        pcol->req_q = osd_request_queue(sbi->s_dev);
         pcol->inode = inode;
         pcol->expected_pages = expected_pages;
 
-        pcol->bio = NULL;
+        pcol->ios = NULL;
+        pcol->pages = NULL;
+        pcol->alloc_pages = 0;
         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)
 {
         pcol->expected_pages -= min(pcol->nr_pages, pcol->expected_pages);
 
-        pcol->bio = NULL;
+        pcol->pages = NULL;
+        pcol->alloc_pages = 0;
         pcol->nr_pages = 0;
         pcol->length = 0;
         pcol->pg_first = -1;
-        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 = MAX_PAGES_KMALLOC;
 }
 
 static int pcol_try_alloc(struct page_collect *pcol)
 {
-        int pages = min_t(unsigned, pcol->expected_pages, BIO_MAX_PAGES);
+        unsigned pages = min_t(unsigned, pcol->expected_pages,
+                          MAX_PAGES_KMALLOC);
+
+        if (!pcol->ios) { /* First time allocate io_state */
+                int ret = exofs_get_io_state(&pcol->sbi->layout, &pcol->ios);
+
+                if (ret)
+                        return ret;
+        }
+
+        /* TODO: easily support bio chaining */
+        pages =  min_t(unsigned, pages,
+                       pcol->sbi->layout.group_width * BIO_MAX_PAGES_KMALLOC);
 
         for (; pages; pages >>= 1) {
-                pcol->bio = bio_alloc(GFP_KERNEL, pages);
-                if (likely(pcol->bio))
+                pcol->pages = kmalloc(pages * sizeof(struct page *),
+                                      GFP_KERNEL);
+                if (likely(pcol->pages)) {
+                        pcol->alloc_pages = pages;
                         return 0;
+                }
         }
 
-        EXOFS_ERR("Failed to kcalloc expected_pages=%u\n",
+        EXOFS_ERR("Failed to kmalloc expected_pages=%u\n",
                   pcol->expected_pages);
         return -ENOMEM;
 }
 
 static void pcol_free(struct page_collect *pcol)
 {
-        bio_put(pcol->bio);
-        pcol->bio = NULL;
+        kfree(pcol->pages);
+        pcol->pages = 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,
                          unsigned len)
 {
-        int added_len = bio_add_pc_page(pcol->req_q, pcol->bio, page, len, 0);
-        if (unlikely(len != added_len))
+        if (unlikely(pcol->nr_pages >= pcol->alloc_pages))
                 return -ENOMEM;
 
-        ++pcol->nr_pages;
+        pcol->pages[pcol->nr_pages++] = page;
         pcol->length += len;
         return 0;
 }
@@ -161,32 +184,26 @@ 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;
 
-        EXOFS_DBGMSG("readpages_done(0x%lx) good_bytes=0x%llx"
+        EXOFS_DBGMSG2("readpages_done(0x%lx) good_bytes=0x%llx"
                      " length=0x%lx nr_pages=%u\n",
                      pcol->inode->i_ino, _LLU(good_bytes), pcol->length,
                      pcol->nr_pages);
 
-        __bio_for_each_segment(bvec, pcol->bio, i, 0) {
-                struct page *page = bvec->bv_page;
+        for (i = 0; i < pcol->nr_pages; i++) {
+                struct page *page = pcol->pages[i];
                 struct inode *inode = page->mapping->host;
                 int page_stat;
 
@@ -198,38 +215,37 @@ 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");
 
                 ret = update_read_page(page, page_stat);
                 if (do_unlock)
                         unlock_page(page);
-                length += bvec->bv_len;
+                length += PAGE_SIZE;
         }
 
         pcol_free(pcol);
-        EXOFS_DBGMSG("readpages_done END\n");
+        EXOFS_DBGMSG2("readpages_done END\n");
         return ret;
 }
 
 /* callback of async reads */
-static void readpages_done(struct 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)
 {
-        struct bio_vec *bvec;
         int i;
 
-        __bio_for_each_segment(bvec, pcol->bio, i, 0) {
-                struct page *page = bvec->bv_page;
+        for (i = 0; i < pcol->nr_pages; i++) {
+                struct page *page = pcol->pages[i];
 
                 if (rw == READ)
                         update_read_page(page, ret);
@@ -238,36 +254,29 @@ 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)
+        if (!pcol->pages)
                 return 0;
 
         /* 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->pages = pcol->pages;
+        ios->nr_pages = pcol->nr_pages;
+        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);
@@ -277,14 +286,16 @@ 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);
+        EXOFS_DBGMSG2("read_exec obj=0x%llx start=0x%llx length=0x%lx\n",
+                  ios->obj.id, _LLU(ios->offset), pcol->length);
 
         /* pages ownership was passed to pcol_copy */
         _pcol_reset(pcol);
@@ -293,12 +304,10 @@ 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;
 }
 
@@ -361,7 +370,7 @@ try_again:
                 goto try_again;
         }
 
-        if (!pcol->bio) {
+        if (!pcol->pages) {
                 ret = pcol_try_alloc(pcol);
                 if (unlikely(ret))
                         goto fail;
@@ -370,12 +379,12 @@ 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);
 
@@ -419,9 +428,8 @@ 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) {
@@ -440,35 +448,30 @@ 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;
         int i;
         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;
 
-        EXOFS_DBGMSG("writepages_done(0x%lx) good_bytes=0x%llx"
+        EXOFS_DBGMSG2("writepages_done(0x%lx) good_bytes=0x%llx"
                      " length=0x%lx nr_pages=%u\n",
                      pcol->inode->i_ino, _LLU(good_bytes), pcol->length,
                      pcol->nr_pages);
 
-        __bio_for_each_segment(bvec, pcol->bio, i, 0) {
-                struct page *page = bvec->bv_page;
+        for (i = 0; i < pcol->nr_pages; i++) {
+                struct page *page = pcol->pages[i];
                 struct inode *inode = page->mapping->host;
                 int page_stat;
 
@@ -482,37 +485,27 @@ 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;
+                length += PAGE_SIZE;
         }
 
         pcol_free(pcol);
         kfree(pcol);
-        EXOFS_DBGMSG("writepages_done END\n");
+        EXOFS_DBGMSG2("writepages_done END\n");
 }
 
 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)
+        if (!pcol->pages)
                 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");
@@ -522,17 +515,22 @@ 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->pages = pcol_copy->pages;
+        ios->nr_pages = pcol_copy->nr_pages;
+        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),
+        EXOFS_DBGMSG2("write_exec(0x%lx, 0x%llx) start=0x%llx length=0x%lx\n",
+                  pcol->inode->i_ino, pcol->pg_first, _LLU(ios->offset),
                   pcol->length);
         /* pages ownership was passed to pcol_copy */
         _pcol_reset(pcol);
@@ -540,9 +538,9 @@ 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;
 }
 
@@ -586,6 +584,9 @@ 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;
                 }
         }
@@ -600,21 +601,24 @@ 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;
         }
 
-        if (!pcol->bio) {
+        if (!pcol->pages) {
                 ret = pcol_try_alloc(pcol);
                 if (unlikely(ret))
                         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);
         if (unlikely(ret)) {
-                EXOFS_DBGMSG("Failed pcol_add_page "
+                EXOFS_DBGMSG2("Failed pcol_add_page "
                              "nr_pages=%u total_length=0x%lx\n",
                              pcol->nr_pages, pcol->length);
 
@@ -634,6 +638,8 @@ 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;
@@ -652,14 +658,17 @@ 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_DBGMSG2("inode(0x%lx) wbc->start=0x%llx wbc->end=0x%llx "
+                     "nrpages=%lu start=0x%lx end=0x%lx expected_pages=%ld\n",
                      mapping->host->i_ino, wbc->range_start, wbc->range_end,
-                     mapping->nrpages, start, end);
+                     mapping->nrpages, start, end, expected_pages);
 
         _pcol_init(&pcol, expected_pages, mapping->host);
 
@@ -731,13 +740,28 @@ static int exofs_write_begin_export(struct file *file,
                                         fsdata);
 }
 
+static int exofs_write_end(struct file *file, struct address_space *mapping,
+                        loff_t pos, unsigned len, unsigned copied,
+                        struct page *page, void *fsdata)
+{
+        struct inode *inode = mapping->host;
+        /* According to comment in simple_write_end i_mutex is held */
+        loff_t i_size = inode->i_size;
+        int ret;
+
+        ret = simple_write_end(file, mapping,pos, len, copied, page, fsdata);
+        if (i_size != inode->i_size)
+                mark_inode_dirty(inode);
+        return ret;
+}
+
 const struct address_space_operations exofs_aops = {
         .readpage       = exofs_readpage,
         .readpages      = exofs_readpages,
         .writepage      = exofs_writepage,
         .writepages     = exofs_writepages,
         .write_begin    = exofs_write_begin_export,
-        .write_end      = simple_write_end,
+        .write_end      = exofs_write_end,
 };
 
 /******************************************************************************
@@ -771,19 +795,28 @@ 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)
@@ -793,22 +826,6 @@ 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
@@ -816,8 +833,7 @@ 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;
 
@@ -845,67 +861,110 @@ int exofs_setattr(struct dentry *dentry, struct iattr *iattr)
         return error;
 }
 
+static const struct osd_attr g_attr_inode_file_layout = ATTR_DEF(
+        EXOFS_APAGE_FS_DATA,
+        EXOFS_ATTR_INODE_FILE_LAYOUT,
+        0);
+static const struct osd_attr g_attr_inode_dir_layout = ATTR_DEF(
+        EXOFS_APAGE_FS_DATA,
+        EXOFS_ATTR_INODE_DIR_LAYOUT,
+        0);
+
 /*
- * Read an inode from the OSD, and return it as is.  We also return the size
- * attribute in the 'sanity' argument if we got compiled with debugging turned
- * on.
+ * Read the Linux inode info from the OSD, and return it as is. In exofs the
+ * inode info is in an application specific page/attribute of the osd-object.
  */
 static int exofs_get_inode(struct super_block *sb, struct exofs_i_info *oi,
-                    struct exofs_fcb *inode, uint64_t *sanity)
+                    struct exofs_fcb *inode)
 {
         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[] = {
+                [0] = g_attr_inode_data,
+                [1] = g_attr_inode_file_layout,
+                [2] = g_attr_inode_dir_layout,
+        };
+        struct exofs_io_state *ios;
+        struct exofs_on_disk_inode_layout *layout;
         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;
+        ret = exofs_get_io_state(&sbi->layout, &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[1].len = exofs_on_disk_inode_layout_size(sbi->layout.s_numdevs);
+        attrs[2].len = exofs_on_disk_inode_layout_size(sbi->layout.s_numdevs);
 
-        ret = exofs_sync_op(or, sbi->s_timeout, oi->i_cred);
-        if (ret)
+        ios->in_attr = attrs;
+        ios->in_attr_len = ARRAY_SIZE(attrs);
+
+        ret = exofs_sbi_read(ios);
+        if (unlikely(ret)) {
+                EXOFS_ERR("object(0x%llx) corrupted, return empty file=>%d\n",
+                          _LLU(ios->obj.id), ret);
+                memset(inode, 0, sizeof(*inode));
+                inode->i_mode = 0040000 | (0777 & ~022);
+                /* If object is lost on target we might as well enable it's
+                 * delete.
+                 */
+                if ((ret == -ENOENT) || (ret == -EINVAL))
+                        ret = 0;
                 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);
+        ret = extract_attr_from_ios(ios, &attrs[1]);
+        if (ret) {
+                EXOFS_ERR("%s: extract_attr of inode_data failed\n", __func__);
+                goto out;
+        }
+        if (attrs[1].len) {
+                layout = attrs[1].val_ptr;
+                if (layout->gen_func != cpu_to_le16(LAYOUT_MOVING_WINDOW)) {
+                        EXOFS_ERR("%s: unsupported files layout %d\n",
+                                __func__, layout->gen_func);
+                        ret = -ENOTSUPP;
+                        goto out;
+                }
+        }
 
-#ifdef EXOFS_DEBUG_OBJ_ISIZE
-        attr = g_attr_logical_length;
-        ret = extract_attr_from_req(or, &attr);
+        ret = extract_attr_from_ios(ios, &attrs[2]);
         if (ret) {
-                EXOFS_ERR("ERROR: extract attr from or failed\n");
+                EXOFS_ERR("%s: extract_attr of inode_data failed\n", __func__);
                 goto out;
         }
-        *sanity = get_unaligned_be64(attr.val_ptr);
-#endif
+        if (attrs[2].len) {
+                layout = attrs[2].val_ptr;
+                if (layout->gen_func != cpu_to_le16(LAYOUT_MOVING_WINDOW)) {
+                        EXOFS_ERR("%s: unsupported meta-data layout %d\n",
+                                __func__, layout->gen_func);
+                        ret = -ENOTSUPP;
+                        goto out;
+                }
+        }
 
 out:
-        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);
+        oi->i_flags = 0;
+}
 /*
  * Fill in an inode read from the OSD and set it up for use
  */
@@ -914,7 +973,6 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
         struct exofs_i_info *oi;
         struct exofs_fcb fcb;
         struct inode *inode;
-        uint64_t uninitialized_var(sanity);
         int ret;
 
         inode = iget_locked(sb, ino);
@@ -923,13 +981,13 @@ 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);
+        ret = exofs_get_inode(sb, oi, &fcb);
         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 */
@@ -947,15 +1005,6 @@ struct inode *exofs_iget(struct super_block *sb, unsigned long ino)
         inode->i_blkbits = EXOFS_BLKSHIFT;
         inode->i_generation = le32_to_cpu(fcb.i_generation);
 
-#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",
-                          inode->i_size, _LLU(sanity));
-        }
-#endif
-
         oi->i_dir_start_lookup = 0;
 
         if ((inode->i_nlink == 0) && (inode->i_mode == 0)) {
@@ -1020,23 +1069,30 @@ 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->layout.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);
@@ -1051,8 +1107,7 @@ 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;
@@ -1061,8 +1116,8 @@ 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;
@@ -1089,28 +1144,28 @@ 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->layout, &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);
 
@@ -1128,11 +1183,11 @@ 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);
 
@@ -1148,16 +1203,17 @@ 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;
         int ret;
 
         args = kzalloc(sizeof(*args), GFP_KERNEL);
-        if (!args)
+        if (!args) {
+                EXOFS_DBGMSG("Faild kzalloc of args\n");
                 return -ENOMEM;
+        }
 
         fcb = &args->fcb;
 
@@ -1186,18 +1242,16 @@ 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->layout, &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");
@@ -1206,43 +1260,42 @@ 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:
-        EXOFS_DBGMSG("ret=>%d\n", ret);
+        EXOFS_DBGMSG("(0x%lx) do_sync=%d ret=>%d\n",
+                     inode->i_ino, do_sync, ret);
         return ret;
 }
 
-int exofs_write_inode(struct inode *inode, int wait)
+int exofs_write_inode(struct inode *inode, struct writeback_control *wbc)
 {
-        return exofs_update_inode(inode, wait);
+        return exofs_update_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
 }
 
 /*
  * 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);
 }
 
@@ -1256,8 +1309,7 @@ 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);
@@ -1274,25 +1326,26 @@ 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->layout, &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);