Received-SPF: pass (google.com: domain of linux-kernel+bounces-63561-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Date: Tue, 13 Feb 2024 14:07:13 +0100
From: Jan Kara <jack@suse.cz>
To: Christoph Hellwig <hch@lst.de>
Cc: linux-mm@kvack.org, Matthew Wilcox <willy@infradead.org>,
	Jan Kara <jack@suse.com>, David Howells <dhowells@redhat.com>,
	Brian Foster <bfoster@redhat.com>,
	Christian Brauner <brauner@kernel.org>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 01/14] writeback: don't call mapping_set_error in
 writepage_cb
Message-ID: <20240213130713.ysuxaqcwizqwjke2@quack3>
References: <20240212071348.1369918-1-hch@lst.de>
 <20240212071348.1369918-2-hch@lst.de>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240212071348.1369918-2-hch@lst.de>

On Mon 12-02-24 08:13:35, Christoph Hellwig wrote:
> writepage_cb is the iterator callback for write_cache_pages, which
> already tracks all errors and returns them to the caller.  There is
> no need to additionally cal mapping_set_error which is intended
                          ^^^ call

> for contexts where the error can't be directly returned (e.g. the
> I/O completion handlers).
> 
> Remove the mapping_set_error call in writepage_cb which is not only
> superfluous but also buggy as it can be called with the error argument
> set to AOP_WRITEPAGE_ACTIVATE, which is not actually an error but a
> magic return value asking the caller to unlock the page.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>

Our error handling in writeback has always been ... spotty. E.g.
block_write_full_page() and iomap_writepage_map() call mapping_set_error()
as well so this seems to be a common way to do things, OTOH ext4 calls
mapping_set_error() only on IO completion. I guess the question is how
an error in ->writepages from background writeback should propagate to
eventual fsync(2) caller? Because currently such error propagates all the
way up to writeback_sb_inodes() where it is silently dropped...

								Honza

> ---
>  mm/page-writeback.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> index 3f255534986a2f..62901fa905f01e 100644
> --- a/mm/page-writeback.c
> +++ b/mm/page-writeback.c
> @@ -2534,9 +2534,8 @@ static int writepage_cb(struct folio *folio, struct writeback_control *wbc,
>  		void *data)
>  {
>  	struct address_space *mapping = data;
> -	int ret = mapping->a_ops->writepage(&folio->page, wbc);
> -	mapping_set_error(mapping, ret);
> -	return ret;
> +
> +	return mapping->a_ops->writepage(&folio->page, wbc);
>  }
>  
>  int do_writepages(struct address_space *mapping, struct writeback_control *wbc)
> -- 
> 2.39.2
> 
-- 
Jan Kara <jack@suse.com>
SUSE Labs, CR