Re: [PATCH 06/10] ext4: punch_hole should wait for DIO writers V2

[Jan Kara <jack@suse.cz>]

On Mon 24-09-12 15:44:16, Dmitry Monakhov wrote:
> punch_hole are the places where we have to wait for all existing writers
> (writeback, aio, dio), but currently we simply flush pended end_io request
> which is not sufficient. Even more i_mutex is not holded while punch_hole
> which obviously result in dangerous data corruption due to
> write-after-free.
> 
> This patch performs following changes:
> - Guard punch_hole with i_mutex
> - Recheck inode flags under i_mutex
> - Block all new dio readers in order to prevent information leak caused by
>   read-after-free pattern.
> - punch_hole now wait for all writers in flight
>   NOTE: XXX write-after-free race is still possible because
>   truncate_pagecache_range()  is not completely reliable and where
>   is no easy way to stop writeback while punch_hole is in progress.
  The race I know about is writing via mmap after
truncate_pagecache_range() but before buffers are really truncated. Do you
know about something else?

> Changes from V1:
>   Add flag checks once we hold i_mutex
  Looks good. You can add:
Reviewed-by: Jan Kara <jack@suse.cz>

								Honza

> 
> Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
> ---
>  fs/ext4/extents.c |   48 ++++++++++++++++++++++++++++++++----------------
>  1 files changed, 32 insertions(+), 16 deletions(-)
> 
> diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
> index 6326874..2d58f4c 100644
> --- a/fs/ext4/extents.c
> +++ b/fs/ext4/extents.c
> @@ -4821,9 +4821,29 @@ int ext4_ext_punch_hole(struct file *file, loff_t offset, loff_t length)
>  	loff_t first_page_offset, last_page_offset;
>  	int credits, err = 0;
>  
> +	/*
> +	 * Write out all dirty pages to avoid race conditions
> +	 * Then release them.
> +	 */
> +	if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
> +		err = filemap_write_and_wait_range(mapping,
> +			offset, offset + length - 1);
> +
> +		if (err)
> +			return err;
> +	}
> +
> +	mutex_lock(&inode->i_mutex);
> +	/* Need recheck file flags under mutex */
> +	/* It's not possible punch hole on append only file */
> +	if (IS_APPEND(inode) || IS_IMMUTABLE(inode))
> +		return -EPERM;
> +	if (IS_SWAPFILE(inode))
> +		return -ETXTBSY;
> +
>  	/* No need to punch hole beyond i_size */
>  	if (offset >= inode->i_size)
> -		return 0;
> +		goto out_mutex;
>  
>  	/*
>  	 * If the hole extends beyond i_size, set the hole
> @@ -4841,31 +4861,23 @@ int ext4_ext_punch_hole(struct file *file, loff_t offset, loff_t length)
>  	first_page_offset = first_page << PAGE_CACHE_SHIFT;
>  	last_page_offset = last_page << PAGE_CACHE_SHIFT;
>  
> -	/*
> -	 * Write out all dirty pages to avoid race conditions
> -	 * Then release them.
> -	 */
> -	if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
> -		err = filemap_write_and_wait_range(mapping,
> -			offset, offset + length - 1);
> -
> -		if (err)
> -			return err;
> -	}
> -
>  	/* Now release the pages */
>  	if (last_page_offset > first_page_offset) {
>  		truncate_pagecache_range(inode, first_page_offset,
>  					 last_page_offset - 1);
>  	}
>  
> -	/* finish any pending end_io work */
> +	/* Wait all existing dio workers, newcomers will block on i_mutex */
> +	ext4_inode_block_unlocked_dio(inode);
> +	inode_dio_wait(inode);
>  	ext4_unwritten_wait(inode);
>  
>  	credits = ext4_writepage_trans_blocks(inode);
>  	handle = ext4_journal_start(inode, credits);
> -	if (IS_ERR(handle))
> -		return PTR_ERR(handle);
> +	if (IS_ERR(handle)) {
> +		err = PTR_ERR(handle);
> +		goto out_dio;
> +	}
>  
>  
>  	/*
> @@ -4955,6 +4967,10 @@ out:
>  	inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
>  	ext4_mark_inode_dirty(handle, inode);
>  	ext4_journal_stop(handle);
> +out_dio:
> +	ext4_inode_resume_unlocked_dio(inode);
> +out_mutex:
> +	mutex_unlock(&inode->i_mutex);
>  	return err;
>  }
>  int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
> -- 
> 1.7.7.6
> 
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR