Re: [PATCH/RFC] s390: Hypervisor File System

[Pekka Enberg <penberg@cs.helsinki.fi>]

On Mon, 2006-04-24 at 19:19 +0200, Michael Holzheu wrote:
> +static int hypfs_create_cpu_files(struct super_block *sb,
> +				  struct dentry *cpus_dir, void *cpu_info)
> +{
> +	struct dentry *cpu_dir;
> +	char buffer[TMP_SIZE];

Holy cow! That's 1 KB allocated on the stack! Please use kmalloc()
instead.

> +static int hypfs_create_phys_cpu_files(struct super_block *sb,
> +				       struct dentry *cpus_dir, void *cpu_info)
> +{
> +	struct dentry *cpu_dir;
> +	char buffer[TMP_SIZE];

Ditto.

> +static ssize_t hypfs_aio_write(struct kiocb *iocb, const char __user *buf,
> +			       size_t count, loff_t pos)
> +{
> +	int rc;
> +
> +	mutex_lock(&hypfs_lock);
> +	if (last_update_time == get_seconds()) {
> +		rc = -EBUSY;
> +		goto out;
> +	}
> +	hypfs_delete_tree(hypfs_sblk->s_root);

To state what I said earlier: the use of a global hypfs_sblk is
problematic because now we can only have the filesystem mounted once. So
I would really like to see some other way of updating. How do you feel
about the s_ops->fs_remount thing?

				Pekka