Btrfs: Fix page count calculation
[cascardo/linux.git] / fs / btrfs / file.c
index e354c33..b0ff34b 100644 (file)
@@ -48,30 +48,34 @@ static noinline int btrfs_copy_from_user(loff_t pos, int num_pages,
                                         struct page **prepared_pages,
                                         struct iov_iter *i)
 {
-       size_t copied;
+       size_t copied = 0;
        int pg = 0;
        int offset = pos & (PAGE_CACHE_SIZE - 1);
+       int total_copied = 0;
 
        while (write_bytes > 0) {
                size_t count = min_t(size_t,
                                     PAGE_CACHE_SIZE - offset, write_bytes);
                struct page *page = prepared_pages[pg];
-again:
-               if (unlikely(iov_iter_fault_in_readable(i, count)))
-                       return -EFAULT;
-
-               /* Copy data from userspace to the current page */
-               copied = iov_iter_copy_from_user(page, i, offset, count);
+               /*
+                * Copy data from userspace to the current page
+                *
+                * Disable pagefault to avoid recursive lock since
+                * the pages are already locked
+                */
+               pagefault_disable();
+               copied = iov_iter_copy_from_user_atomic(page, i, offset, count);
+               pagefault_enable();
 
                /* Flush processor's dcache for this page */
                flush_dcache_page(page);
                iov_iter_advance(i, copied);
                write_bytes -= copied;
+               total_copied += copied;
 
+               /* Return to btrfs_file_aio_write to fault page */
                if (unlikely(copied == 0)) {
-                       count = min_t(size_t, PAGE_CACHE_SIZE - offset,
-                                     iov_iter_single_seg_count(i));
-                       goto again;
+                       break;
                }
 
                if (unlikely(copied < PAGE_CACHE_SIZE - offset)) {
@@ -81,7 +85,7 @@ again:
                        offset = 0;
                }
        }
-       return 0;
+       return total_copied;
 }
 
 /*
@@ -220,6 +224,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
 
                        split->bdev = em->bdev;
                        split->flags = flags;
+                       split->compress_type = em->compress_type;
                        ret = add_extent_mapping(em_tree, split);
                        BUG_ON(ret);
                        free_extent_map(split);
@@ -234,6 +239,7 @@ int btrfs_drop_extent_cache(struct inode *inode, u64 start, u64 end,
                        split->len = em->start + em->len - (start + len);
                        split->bdev = em->bdev;
                        split->flags = flags;
+                       split->compress_type = em->compress_type;
 
                        if (compressed) {
                                split->block_len = em->block_len;
@@ -786,8 +792,12 @@ again:
        for (i = 0; i < num_pages; i++) {
                pages[i] = grab_cache_page(inode->i_mapping, index + i);
                if (!pages[i]) {
-                       err = -ENOMEM;
-                       BUG_ON(1);
+                       int c;
+                       for (c = i - 1; c >= 0; c--) {
+                               unlock_page(pages[c]);
+                               page_cache_release(pages[c]);
+                       }
+                       return -ENOMEM;
                }
                wait_on_page_writeback(pages[i]);
        }
@@ -854,6 +864,8 @@ static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
        unsigned long last_index;
        int will_write;
        int buffered = 0;
+       int copied = 0;
+       int dirty_pages = 0;
 
        will_write = ((file->f_flags & O_DSYNC) || IS_SYNC(inode) ||
                      (file->f_flags & O_DIRECT));
@@ -884,6 +896,17 @@ static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
        if (err)
                goto out;
 
+       /*
+        * If BTRFS flips readonly due to some impossible error
+        * (fs_info->fs_state now has BTRFS_SUPER_FLAG_ERROR),
+        * although we have opened a file as writable, we have
+        * to stop this write operation to ensure FS consistency.
+        */
+       if (root->fs_info->fs_state & BTRFS_SUPER_FLAG_ERROR) {
+               err = -EROFS;
+               goto out;
+       }
+
        file_update_time(file);
        BTRFS_I(inode)->sequence++;
 
@@ -926,6 +949,10 @@ static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
                     PAGE_CACHE_SIZE, PAGE_CACHE_SIZE /
                     (sizeof(struct page *)));
        pages = kmalloc(nrptrs * sizeof(struct page *), GFP_KERNEL);
+       if (!pages) {
+               ret = -ENOMEM;
+               goto out;
+       }
 
        /* generic_write_checks can change our pos */
        start_pos = pos;
@@ -964,13 +991,23 @@ static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
                size_t write_bytes = min(iov_iter_count(&i),
                                         nrptrs * (size_t)PAGE_CACHE_SIZE -
                                         offset);
-               size_t num_pages = (write_bytes + PAGE_CACHE_SIZE - 1) >>
-                                       PAGE_CACHE_SHIFT;
+               size_t num_pages = (write_bytes + offset +
+                                   PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
 
                WARN_ON(num_pages > nrptrs);
                memset(pages, 0, sizeof(struct page *) * nrptrs);
 
-               ret = btrfs_delalloc_reserve_space(inode, write_bytes);
+               /*
+                * Fault pages before locking them in prepare_pages
+                * to avoid recursive lock
+                */
+               if (unlikely(iov_iter_fault_in_readable(&i, write_bytes))) {
+                       ret = -EFAULT;
+                       goto out;
+               }
+
+               ret = btrfs_delalloc_reserve_space(inode,
+                                       num_pages << PAGE_CACHE_SHIFT);
                if (ret)
                        goto out;
 
@@ -978,37 +1015,49 @@ static ssize_t btrfs_file_aio_write(struct kiocb *iocb,
                                    pos, first_index, last_index,
                                    write_bytes);
                if (ret) {
-                       btrfs_delalloc_release_space(inode, write_bytes);
+                       btrfs_delalloc_release_space(inode,
+                                       num_pages << PAGE_CACHE_SHIFT);
                        goto out;
                }
 
-               ret = btrfs_copy_from_user(pos, num_pages,
+               copied = btrfs_copy_from_user(pos, num_pages,
                                           write_bytes, pages, &i);
-               if (ret == 0) {
+               dirty_pages = (copied + offset + PAGE_CACHE_SIZE - 1) >>
+                               PAGE_CACHE_SHIFT;
+
+               if (num_pages > dirty_pages) {
+                       if (copied > 0)
+                               atomic_inc(
+                                       &BTRFS_I(inode)->outstanding_extents);
+                       btrfs_delalloc_release_space(inode,
+                                       (num_pages - dirty_pages) <<
+                                       PAGE_CACHE_SHIFT);
+               }
+
+               if (copied > 0) {
                        dirty_and_release_pages(NULL, root, file, pages,
-                                               num_pages, pos, write_bytes);
+                                               dirty_pages, pos, copied);
                }
 
                btrfs_drop_pages(pages, num_pages);
-               if (ret) {
-                       btrfs_delalloc_release_space(inode, write_bytes);
-                       goto out;
-               }
 
-               if (will_write) {
-                       filemap_fdatawrite_range(inode->i_mapping, pos,
-                                                pos + write_bytes - 1);
-               } else {
-                       balance_dirty_pages_ratelimited_nr(inode->i_mapping,
-                                                          num_pages);
-                       if (num_pages <
-                           (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
-                               btrfs_btree_balance_dirty(root, 1);
-                       btrfs_throttle(root);
+               if (copied > 0) {
+                       if (will_write) {
+                               filemap_fdatawrite_range(inode->i_mapping, pos,
+                                                        pos + copied - 1);
+                       } else {
+                               balance_dirty_pages_ratelimited_nr(
+                                                       inode->i_mapping,
+                                                       dirty_pages);
+                               if (dirty_pages <
+                               (root->leafsize >> PAGE_CACHE_SHIFT) + 1)
+                                       btrfs_btree_balance_dirty(root, 1);
+                               btrfs_throttle(root);
+                       }
                }
 
-               pos += write_bytes;
-               num_written += write_bytes;
+               pos += copied;
+               num_written += copied;
 
                cond_resched();
        }
@@ -1047,8 +1096,14 @@ out:
 
                if ((file->f_flags & O_DSYNC) || IS_SYNC(inode)) {
                        trans = btrfs_start_transaction(root, 0);
+                       if (IS_ERR(trans)) {
+                               num_written = PTR_ERR(trans);
+                               goto done;
+                       }
+                       mutex_lock(&inode->i_mutex);
                        ret = btrfs_log_dentry_safe(trans, root,
                                                    file->f_dentry);
+                       mutex_unlock(&inode->i_mutex);
                        if (ret == 0) {
                                ret = btrfs_sync_log(trans, root);
                                if (ret == 0)
@@ -1067,6 +1122,7 @@ out:
                             (start_pos + num_written - 1) >> PAGE_CACHE_SHIFT);
                }
        }
+done:
        current->backing_dev_info = NULL;
        return num_written ? num_written : err;
 }