These changes are the raw update to linux-4.4.6-rt14. Kernel sources
[kvmfornfv.git] / kernel / fs / f2fs / file.c
index 2b52e48..a197215 100644 (file)
 #include <linux/uaccess.h>
 #include <linux/mount.h>
 #include <linux/pagevec.h>
+#include <linux/random.h>
 
 #include "f2fs.h"
 #include "node.h"
 #include "segment.h"
 #include "xattr.h"
 #include "acl.h"
+#include "gc.h"
 #include "trace.h"
 #include <trace/events/f2fs.h>
 
@@ -72,7 +74,8 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
                goto mapped;
 
        /* page is wholly or partially inside EOF */
-       if (((page->index + 1) << PAGE_CACHE_SHIFT) > i_size_read(inode)) {
+       if (((loff_t)(page->index + 1) << PAGE_CACHE_SHIFT) >
+                                               i_size_read(inode)) {
                unsigned offset;
                offset = i_size_read(inode) & ~PAGE_CACHE_MASK;
                zero_user_segment(page, offset, PAGE_CACHE_SIZE);
@@ -84,6 +87,13 @@ static int f2fs_vm_page_mkwrite(struct vm_area_struct *vma,
 mapped:
        /* fill the page */
        f2fs_wait_on_page_writeback(page, DATA);
+
+       /* wait for GCed encrypted page writeback */
+       if (f2fs_encrypted_inode(inode) && S_ISREG(inode->i_mode))
+               f2fs_wait_on_encrypted_page_writeback(sbi, dn.data_blkaddr);
+
+       /* if gced page is attached, don't write to cold segment */
+       clear_cold_data(page);
 out:
        sb_end_pagefault(inode->i_sb);
        return block_page_mkwrite_return(err);
@@ -105,7 +115,7 @@ static int get_parent_ino(struct inode *inode, nid_t *pino)
        if (!dentry)
                return 0;
 
-       if (update_dent_inode(inode, &dentry->d_name)) {
+       if (update_dent_inode(inode, inode, &dentry->d_name)) {
                dput(dentry);
                return 0;
        }
@@ -122,6 +132,8 @@ static inline bool need_do_checkpoint(struct inode *inode)
 
        if (!S_ISREG(inode->i_mode) || inode->i_nlink != 1)
                need_cp = true;
+       else if (file_enc_name(inode) && need_dentry_mark(sbi, inode->i_ino))
+               need_cp = true;
        else if (file_wrong_pino(inode))
                need_cp = true;
        else if (!space_for_roll_forward(sbi))
@@ -200,8 +212,8 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
        }
 
        /* if the inode is dirty, let's recover all the time */
-       if (!datasync && is_inode_flag_set(fi, FI_DIRTY_INODE)) {
-               update_inode_page(inode);
+       if (!datasync) {
+               f2fs_write_inode(inode, NULL);
                goto go_write;
        }
 
@@ -271,7 +283,7 @@ flush_out:
        ret = f2fs_issue_flush(sbi);
 out:
        trace_f2fs_sync_file_exit(inode, need_cp, datasync, ret);
-       f2fs_trace_ios(NULL, NULL, 1);
+       f2fs_trace_ios(NULL, 1);
        return ret;
 }
 
@@ -337,7 +349,7 @@ static loff_t f2fs_seek_block(struct file *file, loff_t offset, int whence)
 
        dirty = __get_first_dirty_index(inode->i_mapping, pgofs, whence);
 
-       for (; data_ofs < isize; data_ofs = pgofs << PAGE_CACHE_SHIFT) {
+       for (; data_ofs < isize; data_ofs = (loff_t)pgofs << PAGE_CACHE_SHIFT) {
                set_new_dnode(&dn, inode, NULL, NULL, 0);
                err = get_dnode_of_data(&dn, pgofs, LOOKUP_NODE_RA);
                if (err && err != -ENOENT) {
@@ -407,6 +419,12 @@ static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
 {
        struct inode *inode = file_inode(file);
 
+       if (f2fs_encrypted_inode(inode)) {
+               int err = f2fs_get_encryption_info(inode);
+               if (err)
+                       return 0;
+       }
+
        /* we don't need to use inline_data strictly */
        if (f2fs_has_inline_data(inode)) {
                int err = f2fs_convert_inline_inode(inode);
@@ -419,11 +437,23 @@ static int f2fs_file_mmap(struct file *file, struct vm_area_struct *vma)
        return 0;
 }
 
+static int f2fs_file_open(struct inode *inode, struct file *filp)
+{
+       int ret = generic_file_open(inode, filp);
+
+       if (!ret && f2fs_encrypted_inode(inode)) {
+               ret = f2fs_get_encryption_info(inode);
+               if (ret)
+                       ret = -EACCES;
+       }
+       return ret;
+}
+
 int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 {
-       int nr_free = 0, ofs = dn->ofs_in_node;
        struct f2fs_sb_info *sbi = F2FS_I_SB(dn->inode);
        struct f2fs_node *raw_node;
+       int nr_free = 0, ofs = dn->ofs_in_node, len = count;
        __le32 *addr;
 
        raw_node = F2FS_NODE(dn->node_page);
@@ -436,14 +466,22 @@ int truncate_data_blocks_range(struct dnode_of_data *dn, int count)
 
                dn->data_blkaddr = NULL_ADDR;
                set_data_blkaddr(dn);
-               f2fs_update_extent_cache(dn);
                invalidate_blocks(sbi, blkaddr);
                if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page))
                        clear_inode_flag(F2FS_I(dn->inode),
                                                FI_FIRST_BLOCK_WRITTEN);
                nr_free++;
        }
+
        if (nr_free) {
+               pgoff_t fofs;
+               /*
+                * once we invalidate valid blkaddr in range [ofs, ofs + count],
+                * we will invalidate all blkaddr in the whole range.
+                */
+               fofs = start_bidx_of_node(ofs_of_node(dn->node_page),
+                                               F2FS_I(dn->inode)) + ofs;
+               f2fs_update_extent_cache_range(dn, fofs, 0, len);
                dec_valid_block_count(sbi, dn->inode, nr_free);
                set_page_dirty(dn->node_page);
                sync_inode_page(dn);
@@ -461,28 +499,32 @@ void truncate_data_blocks(struct dnode_of_data *dn)
 }
 
 static int truncate_partial_data_page(struct inode *inode, u64 from,
-                                                               bool force)
+                                                               bool cache_only)
 {
        unsigned offset = from & (PAGE_CACHE_SIZE - 1);
+       pgoff_t index = from >> PAGE_CACHE_SHIFT;
+       struct address_space *mapping = inode->i_mapping;
        struct page *page;
 
-       if (!offset && !force)
+       if (!offset && !cache_only)
                return 0;
 
-       page = find_data_page(inode, from >> PAGE_CACHE_SHIFT, force);
-       if (IS_ERR(page))
+       if (cache_only) {
+               page = f2fs_grab_cache_page(mapping, index, false);
+               if (page && PageUptodate(page))
+                       goto truncate_out;
+               f2fs_put_page(page, 1);
                return 0;
+       }
 
-       lock_page(page);
-       if (unlikely(!PageUptodate(page) ||
-                       page->mapping != inode->i_mapping))
-               goto out;
-
+       page = get_lock_data_page(inode, index, true);
+       if (IS_ERR(page))
+               return 0;
+truncate_out:
        f2fs_wait_on_page_writeback(page, DATA);
        zero_user(page, offset, PAGE_CACHE_SIZE - offset);
-       if (!force)
+       if (!cache_only || !f2fs_encrypted_inode(inode) || !S_ISREG(inode->i_mode))
                set_page_dirty(page);
-out:
        f2fs_put_page(page, 1);
        return 0;
 }
@@ -551,24 +593,30 @@ out:
        return err;
 }
 
-void f2fs_truncate(struct inode *inode)
+int f2fs_truncate(struct inode *inode, bool lock)
 {
+       int err;
+
        if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
                                S_ISLNK(inode->i_mode)))
-               return;
+               return 0;
 
        trace_f2fs_truncate(inode);
 
        /* we should check inline_data size */
-       if (f2fs_has_inline_data(inode) && !f2fs_may_inline(inode)) {
-               if (f2fs_convert_inline_inode(inode))
-                       return;
+       if (f2fs_has_inline_data(inode) && !f2fs_may_inline_data(inode)) {
+               err = f2fs_convert_inline_inode(inode);
+               if (err)
+                       return err;
        }
 
-       if (!truncate_blocks(inode, i_size_read(inode), true)) {
-               inode->i_mtime = inode->i_ctime = CURRENT_TIME;
-               mark_inode_dirty(inode);
-       }
+       err = truncate_blocks(inode, i_size_read(inode), lock);
+       if (err)
+               return err;
+
+       inode->i_mtime = inode->i_ctime = CURRENT_TIME;
+       mark_inode_dirty(inode);
+       return 0;
 }
 
 int f2fs_getattr(struct vfsmount *mnt,
@@ -622,16 +670,23 @@ int f2fs_setattr(struct dentry *dentry, struct iattr *attr)
                return err;
 
        if (attr->ia_valid & ATTR_SIZE) {
-               if (attr->ia_size != i_size_read(inode)) {
+               if (f2fs_encrypted_inode(inode) &&
+                               f2fs_get_encryption_info(inode))
+                       return -EACCES;
+
+               if (attr->ia_size <= i_size_read(inode)) {
                        truncate_setsize(inode, attr->ia_size);
-                       f2fs_truncate(inode);
+                       err = f2fs_truncate(inode, true);
+                       if (err)
+                               return err;
                        f2fs_balance_fs(F2FS_I_SB(inode));
                } else {
                        /*
-                        * giving a chance to truncate blocks past EOF which
-                        * are fallocated with FALLOC_FL_KEEP_SIZE.
+                        * do not trim all blocks after i_size if target size is
+                        * larger than i_size.
                         */
-                       f2fs_truncate(inode);
+                       truncate_setsize(inode, attr->ia_size);
+                       inode->i_mtime = inode->i_ctime = CURRENT_TIME;
                }
        }
 
@@ -663,14 +718,14 @@ const struct inode_operations f2fs_file_inode_operations = {
        .fiemap         = f2fs_fiemap,
 };
 
-static void fill_zero(struct inode *inode, pgoff_t index,
+static int fill_zero(struct inode *inode, pgoff_t index,
                                        loff_t start, loff_t len)
 {
        struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
        struct page *page;
 
        if (!len)
-               return;
+               return 0;
 
        f2fs_balance_fs(sbi);
 
@@ -678,33 +733,43 @@ static void fill_zero(struct inode *inode, pgoff_t index,
        page = get_new_data_page(inode, NULL, index, false);
        f2fs_unlock_op(sbi);
 
-       if (!IS_ERR(page)) {
-               f2fs_wait_on_page_writeback(page, DATA);
-               zero_user(page, start, len);
-               set_page_dirty(page);
-               f2fs_put_page(page, 1);
-       }
+       if (IS_ERR(page))
+               return PTR_ERR(page);
+
+       f2fs_wait_on_page_writeback(page, DATA);
+       zero_user(page, start, len);
+       set_page_dirty(page);
+       f2fs_put_page(page, 1);
+       return 0;
 }
 
 int truncate_hole(struct inode *inode, pgoff_t pg_start, pgoff_t pg_end)
 {
-       pgoff_t index;
        int err;
 
-       for (index = pg_start; index < pg_end; index++) {
+       while (pg_start < pg_end) {
                struct dnode_of_data dn;
+               pgoff_t end_offset, count;
 
                set_new_dnode(&dn, inode, NULL, NULL, 0);
-               err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
+               err = get_dnode_of_data(&dn, pg_start, LOOKUP_NODE);
                if (err) {
-                       if (err == -ENOENT)
+                       if (err == -ENOENT) {
+                               pg_start++;
                                continue;
+                       }
                        return err;
                }
 
-               if (dn.data_blkaddr != NULL_ADDR)
-                       truncate_data_blocks_range(&dn, 1);
+               end_offset = ADDRS_PER_PAGE(dn.node_page, F2FS_I(inode));
+               count = min(end_offset - dn.ofs_in_node, pg_end - pg_start);
+
+               f2fs_bug_on(F2FS_I_SB(inode), count == 0 || count > end_offset);
+
+               truncate_data_blocks_range(&dn, count);
                f2fs_put_dnode(&dn);
+
+               pg_start += count;
        }
        return 0;
 }
@@ -715,13 +780,6 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
        loff_t off_start, off_end;
        int ret = 0;
 
-       if (!S_ISREG(inode->i_mode))
-               return -EOPNOTSUPP;
-
-       /* skip punching hole beyond i_size */
-       if (offset >= inode->i_size)
-               return ret;
-
        if (f2fs_has_inline_data(inode)) {
                ret = f2fs_convert_inline_inode(inode);
                if (ret)
@@ -735,14 +793,22 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
        off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
 
        if (pg_start == pg_end) {
-               fill_zero(inode, pg_start, off_start,
+               ret = fill_zero(inode, pg_start, off_start,
                                                off_end - off_start);
+               if (ret)
+                       return ret;
        } else {
-               if (off_start)
-                       fill_zero(inode, pg_start++, off_start,
-                                       PAGE_CACHE_SIZE - off_start);
-               if (off_end)
-                       fill_zero(inode, pg_end, 0, off_end);
+               if (off_start) {
+                       ret = fill_zero(inode, pg_start++, off_start,
+                                               PAGE_CACHE_SIZE - off_start);
+                       if (ret)
+                               return ret;
+               }
+               if (off_end) {
+                       ret = fill_zero(inode, pg_end, 0, off_end);
+                       if (ret)
+                               return ret;
+               }
 
                if (pg_start < pg_end) {
                        struct address_space *mapping = inode->i_mapping;
@@ -751,8 +817,8 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
 
                        f2fs_balance_fs(sbi);
 
-                       blk_start = pg_start << PAGE_CACHE_SHIFT;
-                       blk_end = pg_end << PAGE_CACHE_SHIFT;
+                       blk_start = (loff_t)pg_start << PAGE_CACHE_SHIFT;
+                       blk_end = (loff_t)pg_end << PAGE_CACHE_SHIFT;
                        truncate_inode_pages_range(mapping, blk_start,
                                        blk_end - 1);
 
@@ -765,6 +831,320 @@ static int punch_hole(struct inode *inode, loff_t offset, loff_t len)
        return ret;
 }
 
+static int __exchange_data_block(struct inode *inode, pgoff_t src,
+                                       pgoff_t dst, bool full)
+{
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       struct dnode_of_data dn;
+       block_t new_addr;
+       bool do_replace = false;
+       int ret;
+
+       set_new_dnode(&dn, inode, NULL, NULL, 0);
+       ret = get_dnode_of_data(&dn, src, LOOKUP_NODE_RA);
+       if (ret && ret != -ENOENT) {
+               return ret;
+       } else if (ret == -ENOENT) {
+               new_addr = NULL_ADDR;
+       } else {
+               new_addr = dn.data_blkaddr;
+               if (!is_checkpointed_data(sbi, new_addr)) {
+                       dn.data_blkaddr = NULL_ADDR;
+                       /* do not invalidate this block address */
+                       set_data_blkaddr(&dn);
+                       f2fs_update_extent_cache(&dn);
+                       do_replace = true;
+               }
+               f2fs_put_dnode(&dn);
+       }
+
+       if (new_addr == NULL_ADDR)
+               return full ? truncate_hole(inode, dst, dst + 1) : 0;
+
+       if (do_replace) {
+               struct page *ipage = get_node_page(sbi, inode->i_ino);
+               struct node_info ni;
+
+               if (IS_ERR(ipage)) {
+                       ret = PTR_ERR(ipage);
+                       goto err_out;
+               }
+
+               set_new_dnode(&dn, inode, ipage, NULL, 0);
+               ret = f2fs_reserve_block(&dn, dst);
+               if (ret)
+                       goto err_out;
+
+               truncate_data_blocks_range(&dn, 1);
+
+               get_node_info(sbi, dn.nid, &ni);
+               f2fs_replace_block(sbi, &dn, dn.data_blkaddr, new_addr,
+                               ni.version, true);
+               f2fs_put_dnode(&dn);
+       } else {
+               struct page *psrc, *pdst;
+
+               psrc = get_lock_data_page(inode, src, true);
+               if (IS_ERR(psrc))
+                       return PTR_ERR(psrc);
+               pdst = get_new_data_page(inode, NULL, dst, false);
+               if (IS_ERR(pdst)) {
+                       f2fs_put_page(psrc, 1);
+                       return PTR_ERR(pdst);
+               }
+               f2fs_copy_page(psrc, pdst);
+               set_page_dirty(pdst);
+               f2fs_put_page(pdst, 1);
+               f2fs_put_page(psrc, 1);
+
+               return truncate_hole(inode, src, src + 1);
+       }
+       return 0;
+
+err_out:
+       if (!get_dnode_of_data(&dn, src, LOOKUP_NODE)) {
+               dn.data_blkaddr = new_addr;
+               set_data_blkaddr(&dn);
+               f2fs_update_extent_cache(&dn);
+               f2fs_put_dnode(&dn);
+       }
+       return ret;
+}
+
+static int f2fs_do_collapse(struct inode *inode, pgoff_t start, pgoff_t end)
+{
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       pgoff_t nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
+       int ret = 0;
+
+       for (; end < nrpages; start++, end++) {
+               f2fs_balance_fs(sbi);
+               f2fs_lock_op(sbi);
+               ret = __exchange_data_block(inode, end, start, true);
+               f2fs_unlock_op(sbi);
+               if (ret)
+                       break;
+       }
+       return ret;
+}
+
+static int f2fs_collapse_range(struct inode *inode, loff_t offset, loff_t len)
+{
+       pgoff_t pg_start, pg_end;
+       loff_t new_size;
+       int ret;
+
+       if (offset + len >= i_size_read(inode))
+               return -EINVAL;
+
+       /* collapse range should be aligned to block size of f2fs. */
+       if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
+               return -EINVAL;
+
+       f2fs_balance_fs(F2FS_I_SB(inode));
+
+       if (f2fs_has_inline_data(inode)) {
+               ret = f2fs_convert_inline_inode(inode);
+               if (ret)
+                       return ret;
+       }
+
+       pg_start = offset >> PAGE_CACHE_SHIFT;
+       pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
+
+       /* write out all dirty pages from offset */
+       ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
+       if (ret)
+               return ret;
+
+       truncate_pagecache(inode, offset);
+
+       ret = f2fs_do_collapse(inode, pg_start, pg_end);
+       if (ret)
+               return ret;
+
+       /* write out all moved pages, if possible */
+       filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
+       truncate_pagecache(inode, offset);
+
+       new_size = i_size_read(inode) - len;
+       truncate_pagecache(inode, new_size);
+
+       ret = truncate_blocks(inode, new_size, true);
+       if (!ret)
+               i_size_write(inode, new_size);
+
+       return ret;
+}
+
+static int f2fs_zero_range(struct inode *inode, loff_t offset, loff_t len,
+                                                               int mode)
+{
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       struct address_space *mapping = inode->i_mapping;
+       pgoff_t index, pg_start, pg_end;
+       loff_t new_size = i_size_read(inode);
+       loff_t off_start, off_end;
+       int ret = 0;
+
+       ret = inode_newsize_ok(inode, (len + offset));
+       if (ret)
+               return ret;
+
+       f2fs_balance_fs(sbi);
+
+       if (f2fs_has_inline_data(inode)) {
+               ret = f2fs_convert_inline_inode(inode);
+               if (ret)
+                       return ret;
+       }
+
+       ret = filemap_write_and_wait_range(mapping, offset, offset + len - 1);
+       if (ret)
+               return ret;
+
+       truncate_pagecache_range(inode, offset, offset + len - 1);
+
+       pg_start = ((unsigned long long) offset) >> PAGE_CACHE_SHIFT;
+       pg_end = ((unsigned long long) offset + len) >> PAGE_CACHE_SHIFT;
+
+       off_start = offset & (PAGE_CACHE_SIZE - 1);
+       off_end = (offset + len) & (PAGE_CACHE_SIZE - 1);
+
+       if (pg_start == pg_end) {
+               ret = fill_zero(inode, pg_start, off_start,
+                                               off_end - off_start);
+               if (ret)
+                       return ret;
+
+               if (offset + len > new_size)
+                       new_size = offset + len;
+               new_size = max_t(loff_t, new_size, offset + len);
+       } else {
+               if (off_start) {
+                       ret = fill_zero(inode, pg_start++, off_start,
+                                               PAGE_CACHE_SIZE - off_start);
+                       if (ret)
+                               return ret;
+
+                       new_size = max_t(loff_t, new_size,
+                                       (loff_t)pg_start << PAGE_CACHE_SHIFT);
+               }
+
+               for (index = pg_start; index < pg_end; index++) {
+                       struct dnode_of_data dn;
+                       struct page *ipage;
+
+                       f2fs_lock_op(sbi);
+
+                       ipage = get_node_page(sbi, inode->i_ino);
+                       if (IS_ERR(ipage)) {
+                               ret = PTR_ERR(ipage);
+                               f2fs_unlock_op(sbi);
+                               goto out;
+                       }
+
+                       set_new_dnode(&dn, inode, ipage, NULL, 0);
+                       ret = f2fs_reserve_block(&dn, index);
+                       if (ret) {
+                               f2fs_unlock_op(sbi);
+                               goto out;
+                       }
+
+                       if (dn.data_blkaddr != NEW_ADDR) {
+                               invalidate_blocks(sbi, dn.data_blkaddr);
+
+                               dn.data_blkaddr = NEW_ADDR;
+                               set_data_blkaddr(&dn);
+
+                               dn.data_blkaddr = NULL_ADDR;
+                               f2fs_update_extent_cache(&dn);
+                       }
+                       f2fs_put_dnode(&dn);
+                       f2fs_unlock_op(sbi);
+
+                       new_size = max_t(loff_t, new_size,
+                               (loff_t)(index + 1) << PAGE_CACHE_SHIFT);
+               }
+
+               if (off_end) {
+                       ret = fill_zero(inode, pg_end, 0, off_end);
+                       if (ret)
+                               goto out;
+
+                       new_size = max_t(loff_t, new_size, offset + len);
+               }
+       }
+
+out:
+       if (!(mode & FALLOC_FL_KEEP_SIZE) && i_size_read(inode) < new_size) {
+               i_size_write(inode, new_size);
+               mark_inode_dirty(inode);
+               update_inode_page(inode);
+       }
+
+       return ret;
+}
+
+static int f2fs_insert_range(struct inode *inode, loff_t offset, loff_t len)
+{
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       pgoff_t pg_start, pg_end, delta, nrpages, idx;
+       loff_t new_size;
+       int ret = 0;
+
+       new_size = i_size_read(inode) + len;
+       if (new_size > inode->i_sb->s_maxbytes)
+               return -EFBIG;
+
+       if (offset >= i_size_read(inode))
+               return -EINVAL;
+
+       /* insert range should be aligned to block size of f2fs. */
+       if (offset & (F2FS_BLKSIZE - 1) || len & (F2FS_BLKSIZE - 1))
+               return -EINVAL;
+
+       f2fs_balance_fs(sbi);
+
+       if (f2fs_has_inline_data(inode)) {
+               ret = f2fs_convert_inline_inode(inode);
+               if (ret)
+                       return ret;
+       }
+
+       ret = truncate_blocks(inode, i_size_read(inode), true);
+       if (ret)
+               return ret;
+
+       /* write out all dirty pages from offset */
+       ret = filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
+       if (ret)
+               return ret;
+
+       truncate_pagecache(inode, offset);
+
+       pg_start = offset >> PAGE_CACHE_SHIFT;
+       pg_end = (offset + len) >> PAGE_CACHE_SHIFT;
+       delta = pg_end - pg_start;
+       nrpages = (i_size_read(inode) + PAGE_SIZE - 1) / PAGE_SIZE;
+
+       for (idx = nrpages - 1; idx >= pg_start && idx != -1; idx--) {
+               f2fs_lock_op(sbi);
+               ret = __exchange_data_block(inode, idx, idx + delta, false);
+               f2fs_unlock_op(sbi);
+               if (ret)
+                       break;
+       }
+
+       /* write out all moved pages, if possible */
+       filemap_write_and_wait_range(inode->i_mapping, offset, LLONG_MAX);
+       truncate_pagecache(inode, offset);
+
+       if (!ret)
+               i_size_write(inode, new_size);
+       return ret;
+}
+
 static int expand_inode_data(struct inode *inode, loff_t offset,
                                        loff_t len, int mode)
 {
@@ -808,9 +1188,10 @@ noalloc:
                if (pg_start == pg_end)
                        new_size = offset + len;
                else if (index == pg_start && off_start)
-                       new_size = (index + 1) << PAGE_CACHE_SHIFT;
+                       new_size = (loff_t)(index + 1) << PAGE_CACHE_SHIFT;
                else if (index == pg_end)
-                       new_size = (index << PAGE_CACHE_SHIFT) + off_end;
+                       new_size = ((loff_t)index << PAGE_CACHE_SHIFT) +
+                                                               off_end;
                else
                        new_size += PAGE_CACHE_SIZE;
        }
@@ -830,23 +1211,44 @@ static long f2fs_fallocate(struct file *file, int mode,
                                loff_t offset, loff_t len)
 {
        struct inode *inode = file_inode(file);
-       long ret;
+       long ret = 0;
+
+       /* f2fs only support ->fallocate for regular file */
+       if (!S_ISREG(inode->i_mode))
+               return -EINVAL;
 
-       if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
+       if (f2fs_encrypted_inode(inode) &&
+               (mode & (FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_INSERT_RANGE)))
+               return -EOPNOTSUPP;
+
+       if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE |
+                       FALLOC_FL_COLLAPSE_RANGE | FALLOC_FL_ZERO_RANGE |
+                       FALLOC_FL_INSERT_RANGE))
                return -EOPNOTSUPP;
 
        mutex_lock(&inode->i_mutex);
 
-       if (mode & FALLOC_FL_PUNCH_HOLE)
+       if (mode & FALLOC_FL_PUNCH_HOLE) {
+               if (offset >= inode->i_size)
+                       goto out;
+
                ret = punch_hole(inode, offset, len);
-       else
+       } else if (mode & FALLOC_FL_COLLAPSE_RANGE) {
+               ret = f2fs_collapse_range(inode, offset, len);
+       } else if (mode & FALLOC_FL_ZERO_RANGE) {
+               ret = f2fs_zero_range(inode, offset, len, mode);
+       } else if (mode & FALLOC_FL_INSERT_RANGE) {
+               ret = f2fs_insert_range(inode, offset, len);
+       } else {
                ret = expand_inode_data(inode, offset, len, mode);
+       }
 
        if (!ret) {
                inode->i_mtime = inode->i_ctime = CURRENT_TIME;
                mark_inode_dirty(inode);
        }
 
+out:
        mutex_unlock(&inode->i_mutex);
 
        trace_f2fs_fallocate(inode, mode, offset, len, ret);
@@ -946,6 +1348,7 @@ static int f2fs_ioc_getversion(struct file *filp, unsigned long arg)
 static int f2fs_ioc_start_atomic_write(struct file *filp)
 {
        struct inode *inode = file_inode(filp);
+       int ret;
 
        if (!inode_owner_or_capable(inode))
                return -EACCES;
@@ -955,9 +1358,12 @@ static int f2fs_ioc_start_atomic_write(struct file *filp)
        if (f2fs_is_atomic_file(inode))
                return 0;
 
-       set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
+       ret = f2fs_convert_inline_inode(inode);
+       if (ret)
+               return ret;
 
-       return f2fs_convert_inline_inode(inode);
+       set_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
+       return 0;
 }
 
 static int f2fs_ioc_commit_atomic_write(struct file *filp)
@@ -975,18 +1381,23 @@ static int f2fs_ioc_commit_atomic_write(struct file *filp)
        if (ret)
                return ret;
 
-       if (f2fs_is_atomic_file(inode))
-               commit_inmem_pages(inode, false);
+       if (f2fs_is_atomic_file(inode)) {
+               clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
+               ret = commit_inmem_pages(inode, false);
+               if (ret)
+                       goto err_out;
+       }
 
-       ret = f2fs_sync_file(filp, 0, LONG_MAX, 0);
+       ret = f2fs_sync_file(filp, 0, LLONG_MAX, 0);
+err_out:
        mnt_drop_write_file(filp);
-       clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
        return ret;
 }
 
 static int f2fs_ioc_start_volatile_write(struct file *filp)
 {
        struct inode *inode = file_inode(filp);
+       int ret;
 
        if (!inode_owner_or_capable(inode))
                return -EACCES;
@@ -994,9 +1405,12 @@ static int f2fs_ioc_start_volatile_write(struct file *filp)
        if (f2fs_is_volatile_file(inode))
                return 0;
 
-       set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
+       ret = f2fs_convert_inline_inode(inode);
+       if (ret)
+               return ret;
 
-       return f2fs_convert_inline_inode(inode);
+       set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
+       return 0;
 }
 
 static int f2fs_ioc_release_volatile_write(struct file *filp)
@@ -1012,8 +1426,7 @@ static int f2fs_ioc_release_volatile_write(struct file *filp)
        if (!f2fs_is_first_block_written(inode))
                return truncate_partial_data_page(inode, 0, true);
 
-       punch_hole(inode, 0, F2FS_BLKSIZE);
-       return 0;
+       return punch_hole(inode, 0, F2FS_BLKSIZE);
 }
 
 static int f2fs_ioc_abort_volatile_write(struct file *filp)
@@ -1030,16 +1443,10 @@ static int f2fs_ioc_abort_volatile_write(struct file *filp)
 
        f2fs_balance_fs(F2FS_I_SB(inode));
 
-       if (f2fs_is_atomic_file(inode)) {
-               commit_inmem_pages(inode, false);
-               clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
-       }
+       clear_inode_flag(F2FS_I(inode), FI_ATOMIC_FILE);
+       clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
+       commit_inmem_pages(inode, true);
 
-       if (f2fs_is_volatile_file(inode)) {
-               clear_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
-               filemap_fdatawrite(inode->i_mapping);
-               set_inode_flag(F2FS_I(inode), FI_VOLATILE_FILE);
-       }
        mnt_drop_write_file(filp);
        return ret;
 }
@@ -1073,6 +1480,10 @@ static int f2fs_ioc_shutdown(struct file *filp, unsigned long arg)
        case F2FS_GOING_DOWN_NOSYNC:
                f2fs_stop_checkpoint(sbi);
                break;
+       case F2FS_GOING_DOWN_METAFLUSH:
+               sync_meta_pages(sbi, META, LONG_MAX);
+               f2fs_stop_checkpoint(sbi);
+               break;
        default:
                return -EINVAL;
        }
@@ -1109,6 +1520,132 @@ static int f2fs_ioc_fitrim(struct file *filp, unsigned long arg)
        return 0;
 }
 
+static bool uuid_is_nonzero(__u8 u[16])
+{
+       int i;
+
+       for (i = 0; i < 16; i++)
+               if (u[i])
+                       return true;
+       return false;
+}
+
+static int f2fs_ioc_set_encryption_policy(struct file *filp, unsigned long arg)
+{
+#ifdef CONFIG_F2FS_FS_ENCRYPTION
+       struct f2fs_encryption_policy policy;
+       struct inode *inode = file_inode(filp);
+
+       if (copy_from_user(&policy, (struct f2fs_encryption_policy __user *)arg,
+                               sizeof(policy)))
+               return -EFAULT;
+
+       return f2fs_process_policy(&policy, inode);
+#else
+       return -EOPNOTSUPP;
+#endif
+}
+
+static int f2fs_ioc_get_encryption_policy(struct file *filp, unsigned long arg)
+{
+#ifdef CONFIG_F2FS_FS_ENCRYPTION
+       struct f2fs_encryption_policy policy;
+       struct inode *inode = file_inode(filp);
+       int err;
+
+       err = f2fs_get_policy(inode, &policy);
+       if (err)
+               return err;
+
+       if (copy_to_user((struct f2fs_encryption_policy __user *)arg, &policy,
+                                                       sizeof(policy)))
+               return -EFAULT;
+       return 0;
+#else
+       return -EOPNOTSUPP;
+#endif
+}
+
+static int f2fs_ioc_get_encryption_pwsalt(struct file *filp, unsigned long arg)
+{
+       struct inode *inode = file_inode(filp);
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       int err;
+
+       if (!f2fs_sb_has_crypto(inode->i_sb))
+               return -EOPNOTSUPP;
+
+       if (uuid_is_nonzero(sbi->raw_super->encrypt_pw_salt))
+               goto got_it;
+
+       err = mnt_want_write_file(filp);
+       if (err)
+               return err;
+
+       /* update superblock with uuid */
+       generate_random_uuid(sbi->raw_super->encrypt_pw_salt);
+
+       err = f2fs_commit_super(sbi, false);
+
+       mnt_drop_write_file(filp);
+       if (err) {
+               /* undo new data */
+               memset(sbi->raw_super->encrypt_pw_salt, 0, 16);
+               return err;
+       }
+got_it:
+       if (copy_to_user((__u8 __user *)arg, sbi->raw_super->encrypt_pw_salt,
+                                                                       16))
+               return -EFAULT;
+       return 0;
+}
+
+static int f2fs_ioc_gc(struct file *filp, unsigned long arg)
+{
+       struct inode *inode = file_inode(filp);
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       __u32 sync;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       if (get_user(sync, (__u32 __user *)arg))
+               return -EFAULT;
+
+       if (f2fs_readonly(sbi->sb))
+               return -EROFS;
+
+       if (!sync) {
+               if (!mutex_trylock(&sbi->gc_mutex))
+                       return -EBUSY;
+       } else {
+               mutex_lock(&sbi->gc_mutex);
+       }
+
+       return f2fs_gc(sbi, sync);
+}
+
+static int f2fs_ioc_write_checkpoint(struct file *filp, unsigned long arg)
+{
+       struct inode *inode = file_inode(filp);
+       struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+       struct cp_control cpc;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EPERM;
+
+       if (f2fs_readonly(sbi->sb))
+               return -EROFS;
+
+       cpc.reason = __get_cp_reason(sbi);
+
+       mutex_lock(&sbi->gc_mutex);
+       write_checkpoint(sbi, &cpc);
+       mutex_unlock(&sbi->gc_mutex);
+
+       return 0;
+}
+
 long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 {
        switch (cmd) {
@@ -1132,11 +1669,33 @@ long f2fs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
                return f2fs_ioc_shutdown(filp, arg);
        case FITRIM:
                return f2fs_ioc_fitrim(filp, arg);
+       case F2FS_IOC_SET_ENCRYPTION_POLICY:
+               return f2fs_ioc_set_encryption_policy(filp, arg);
+       case F2FS_IOC_GET_ENCRYPTION_POLICY:
+               return f2fs_ioc_get_encryption_policy(filp, arg);
+       case F2FS_IOC_GET_ENCRYPTION_PWSALT:
+               return f2fs_ioc_get_encryption_pwsalt(filp, arg);
+       case F2FS_IOC_GARBAGE_COLLECT:
+               return f2fs_ioc_gc(filp, arg);
+       case F2FS_IOC_WRITE_CHECKPOINT:
+               return f2fs_ioc_write_checkpoint(filp, arg);
        default:
                return -ENOTTY;
        }
 }
 
+static ssize_t f2fs_file_write_iter(struct kiocb *iocb, struct iov_iter *from)
+{
+       struct inode *inode = file_inode(iocb->ki_filp);
+
+       if (f2fs_encrypted_inode(inode) &&
+                               !f2fs_has_encryption_key(inode) &&
+                               f2fs_get_encryption_info(inode))
+               return -EACCES;
+
+       return generic_file_write_iter(iocb, from);
+}
+
 #ifdef CONFIG_COMPAT
 long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 {
@@ -1157,8 +1716,8 @@ long f2fs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
 const struct file_operations f2fs_file_operations = {
        .llseek         = f2fs_llseek,
        .read_iter      = generic_file_read_iter,
-       .write_iter     = generic_file_write_iter,
-       .open           = generic_file_open,
+       .write_iter     = f2fs_file_write_iter,
+       .open           = f2fs_file_open,
        .release        = f2fs_release_file,
        .mmap           = f2fs_file_mmap,
        .fsync          = f2fs_sync_file,