Hi Dan,
On Wed, Aug 12, 2015 at 01:40:10PM +0300, Dan Carpenter wrote:
> Hello Jaegeuk Kim,
>
> The patch 86531d6b84bc: "f2fs: callers take care of the page from bio
> error" from Jul 15, 2015, leads to the following static checker
> warning:
>
> fs/f2fs/node.c:1042 ra_node_page()
> warn: 'err' can be either negative or positive
>
> fs/f2fs/node.c
> 1025 void ra_node_page(struct f2fs_sb_info *sbi, nid_t nid)
> 1026 {
> 1027 struct page *apage;
> 1028 int err;
> 1029
> 1030 apage = find_get_page(NODE_MAPPING(sbi), nid);
> 1031 if (apage && PageUptodate(apage)) {
> 1032 f2fs_put_page(apage, 0);
> 1033 return;
> 1034 }
> 1035 f2fs_put_page(apage, 0);
> 1036
> 1037 apage = grab_cache_page(NODE_MAPPING(sbi), nid);
> 1038 if (!apage)
> 1039 return;
> 1040
> 1041 err = read_node_page(apage, READA);
> 1042 f2fs_put_page(apage, err ? 1 : 0);
>
> In the old code we took errors into consideration but now we treat them
> as LOCKED_PAGE. Is that intentional? A lot of the other callers in
> that patch still check for errors...
This is just for read-ahead node page function.
So, yes, if it returns LOCKED_PAGE or any error, the page must be unlocked
via f2fs_put_page(1).
Otherwise, end_io will unlock the page.
Hmm, any workaround to avoid that warning?
Thanks,
>
> 1043 }
>
> regards,
> dan carpenter
------------------------------------------------------------------------------
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
