Re: [PATCH 4/4] udf: Replace bkl with a mutex for protectudf_sb_info struct

[Christoph Hellwig]

> +		mutex_lock(&sbi->lock);
>  		result = udf_relocate_blocks(inode->i_sb,
>  						old_block, &new_block);
> +		mutex_unlock(&sbi->lock);

Moving the locking inside udf_relocate_blocks would be cleaner.

> @@ -570,7 +569,7 @@ static int udf_remount_fs(struct super_block *sb, int *flags, char *options)
>  	if (!udf_parse_options(options, &uopt, true))
>  		return -EINVAL;
>  
> -	lock_kernel();
> +	mutex_lock(&sbi->lock);

What are you protecting against here?  Concurrent remount calls are
protects against by the VFS.  Is there any reader that takes sbi->lock
to get a consistent view of the various options?  It doesn't seem you
introduce one in this series, so it seems we could do fine without any
locking.

> -		unlock_kernel();
>  		return -ENOMEM;
>  	}
>  
> +	mutex_init(&sbi->lock);
> +	mutex_lock(&sbi->lock);

What are you locking against here?  I can't fine anything that puts the
superblock on a global list in fill_super, and I can't find any code
that would look a superblock up that's not fully set up.  I don't think
synchronization here in fill_super is needed at all.

> +	/* Serialize writer access, replace the old bkl */
> +	struct mutex lock;

Same comment as for the per-inode mutex applies here, too.

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/