[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20110330.204832.186542149.ryusuke@osrg.net>
Date: Wed, 30 Mar 2011 20:48:32 +0900 (JST)
From: Ryusuke Konishi <konishi.ryusuke@....ntt.co.jp>
To: jaxboe@...ionio.com
Cc: konishi.ryusuke@....ntt.co.jp, linux-nilfs@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] nilfs2: fix oops due to a bad aops initialization
On Wed, 30 Mar 2011 13:17:18 +0200, Jens Axboe wrote:
> On 2011-03-30 13:07, Jens Axboe wrote:
> > On 2011-03-30 06:25, Ryusuke Konishi wrote:
> >> diff --git a/fs/nilfs2/page.c b/fs/nilfs2/page.c
> >> index 4d2a1ee..9d2dc6b 100644
> >> --- a/fs/nilfs2/page.c
> >> +++ b/fs/nilfs2/page.c
> >> @@ -495,12 +495,14 @@ unsigned nilfs_page_count_clean_buffers(struct page *page,
> >> void nilfs_mapping_init(struct address_space *mapping,
> >> struct backing_dev_info *bdi)
> >> {
> >> + static const struct address_space_operations empty_aops;
> >> +
> >> mapping->host = NULL;
> >> mapping->flags = 0;
> >> mapping_set_gfp_mask(mapping, GFP_NOFS);
> >> mapping->assoc_mapping = NULL;
> >> mapping->backing_dev_info = bdi;
> >> - mapping->a_ops = NULL;
> >> + mapping->a_ops = &empty_aops;
> >> }
> >
> > Hmm wait, inode init should set the mapping aops to an empty type
> > already. Does the OOPS go away if you just remove the mapping->a_ops =
> > NULL assignment?
>
> In any case, I think this is a better fix.
>
> diff --git a/fs/inode.c b/fs/inode.c
> index 5f4e11a..b818730 100644
> --- a/fs/inode.c
> +++ b/fs/inode.c
> @@ -125,6 +125,13 @@ __cacheline_aligned_in_smp DEFINE_SPINLOCK(inode_wb_list_lock);
> static DECLARE_RWSEM(iprune_sem);
>
> /*
> + * Empty aops. Can be used for the cases where the user does not
> + * define any of the address_space operations.
> + */
> +const struct address_space_operations empty_aops = {
> +};
> +
> +/*
> * Statistics gathering..
> */
> struct inodes_stat_t inodes_stat;
> @@ -176,7 +183,6 @@ int proc_nr_inodes(ctl_table *table, int write,
> */
> int inode_init_always(struct super_block *sb, struct inode *inode)
> {
> - static const struct address_space_operations empty_aops;
> static const struct inode_operations empty_iops;
> static const struct file_operations empty_fops;
> struct address_space *const mapping = &inode->i_data;
> diff --git a/fs/nilfs2/page.c b/fs/nilfs2/page.c
> index 4d2a1ee..1168059 100644
> --- a/fs/nilfs2/page.c
> +++ b/fs/nilfs2/page.c
> @@ -500,7 +500,7 @@ void nilfs_mapping_init(struct address_space *mapping,
> mapping_set_gfp_mask(mapping, GFP_NOFS);
> mapping->assoc_mapping = NULL;
> mapping->backing_dev_info = bdi;
> - mapping->a_ops = NULL;
> + mapping->a_ops = &empty_aops;
> }
>
> /*
> diff --git a/fs/ubifs/xattr.c b/fs/ubifs/xattr.c
> index c74400f..403c192 100644
> --- a/fs/ubifs/xattr.c
> +++ b/fs/ubifs/xattr.c
> @@ -80,7 +80,6 @@ enum {
> };
>
> static const struct inode_operations none_inode_operations;
> -static const struct address_space_operations none_address_operations;
> static const struct file_operations none_file_operations;
>
> /**
> @@ -130,7 +129,7 @@ static int create_xattr(struct ubifs_info *c, struct inode *host,
> }
>
> /* Re-define all operations to be "nothing" */
> - inode->i_mapping->a_ops = &none_address_operations;
> + inode->i_mapping->a_ops = &empty_aops;
> inode->i_op = &none_inode_operations;
> inode->i_fop = &none_file_operations;
>
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 52f283c..1b95af3 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -613,6 +613,8 @@ struct address_space_operations {
> int (*error_remove_page)(struct address_space *, struct page *);
> };
>
> +extern const struct address_space_operations empty_aops;
> +
> /*
> * pagecache_write_begin/pagecache_write_end must be used by general code
> * to write into the pagecache.
>
> --
> Jens Axboe
Sorry but I've already sent a pull request to Linus with your
Acked-by. And, this seems to be doing two things. One is for fixing
the current kerenl oops and another is for sharing the empty address
space operations.
I have no problem with both patches, but for the purpose of the bug
fix, the first one looks simpler.
Thanks,
Ryusuke Konishi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists