Re: [PATCH v4] f2fs: fix potential .flags overflow on 32bit architecture

[Joe Perches]

On Mon, 2020-03-23 at 10:41 +0800, Chao Yu wrote:
> f2fs_inode_info.flags is unsigned long variable, it has 32 bits
> in 32bit architecture, since we introduced FI_MMAP_FILE flag
> when we support data compression, we may access memory cross
> the border of .flags field, corrupting .i_sem field, result in
> below deadlock.
[]
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
[]
> @@ -682,6 +682,47 @@ enum {
[]
> +/* used for f2fs_inode_info->flags */
> +enum {
[]
> +	FI_MAX,			/* max flag, never be used */
> +};
> +
> +/* f2fs_inode_info.flags array size */
> +#define FI_ARRAY_SIZE		(BITS_TO_LONGS(FI_MAX))

Perhaps FI_ARRAY_SIZE isn't necessary.

> +
>  struct f2fs_inode_info {
>  	struct inode vfs_inode;		/* serve a vfs inode */
>  	unsigned long i_flags;		/* keep an inode flags for ioctl */
> @@ -694,7 +735,7 @@ struct f2fs_inode_info {
>  	umode_t i_acl_mode;		/* keep file acl mode temporarily */
>  
>  	/* Use below internally in f2fs*/
> -	unsigned long flags;		/* use to pass per-file flags */
> +	unsigned long flags[FI_ARRAY_SIZE];	/* use to pass per-file flags */

and BITS_TO_LONGS should be used here.

> diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c
[]
> @@ -362,7 +363,8 @@ static int do_read_inode(struct inode *inode)
>  	fi->i_flags = le32_to_cpu(ri->i_flags);
>  	if (S_ISREG(inode->i_mode))
>  		fi->i_flags &= ~F2FS_PROJINHERIT_FL;
> -	fi->flags = 0;
> +	for (i = 0; i < FI_ARRAY_SIZE; i++)
> +		fi->flags[i] = 0;

And this could become

	bitmap_zero(fi->flags, BITS_TO_LONG(FI_MAX));

Is FI_ARRAY_SIZE used anywhere else?