Re: [PATCH v2 07/18] nitro_enclaves: Init misc device providing the ioctl interface

[Paraschiv, Andra-Irina]

On 22/05/2020 10:07, Greg KH wrote:
> On Fri, May 22, 2020 at 09:29:35AM +0300, Andra Paraschiv wrote:
> > +static char *ne_cpus;
> > +module_param(ne_cpus, charp, 0644);
> > +MODULE_PARM_DESC(ne_cpus, "<cpu-list> - CPU pool used for Nitro Enclaves");
> This is not the 1990's, don't use module parameters if you can help it.
> Why is this needed, and where is it documented?

This is a CPU pool that can be set by the root user and that includes 
CPUs set aside to be used for the enclave(s) setup; these CPUs are 
offlined. From this CPU pool, the kernel logic chooses the CPUs that are 
set for the created enclave(s).

The cpu-list format is matching the same that is documented here:

https://www.kernel.org/doc/html/latest/admin-guide/kernel-parameters.html

I've also thought of having a sysfs entry for the setup of this enclave 
CPU pool.

> > +/* CPU pool used for Nitro Enclaves. */
> > +struct ne_cpu_pool {
> > +	/* Available CPUs in the pool. */
> > +	cpumask_var_t avail;
> > +	struct mutex mutex;
> > +};
> > +
> > +static struct ne_cpu_pool ne_cpu_pool;
> > +
> > +static int ne_open(struct inode *node, struct file *file)
> > +{
> > +	return 0;
> > +}
> If open does nothing, just don't even provide it.

I removed this and other file ops occurrences that do nothing for now.

> > +
> > +static long ne_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
> > +{
> > +	switch (cmd) {
> > +
> > +	default:
> > +		return -ENOTTY;
> > +	}
> > +
> > +	return 0;
> > +}
> Same for ioctl.

This logic is completed in the next patch in the series.

> > +
> > +static int ne_release(struct inode *inode, struct file *file)
> > +{
> > +	return 0;
> > +}
> Same for release.

Done, I removed it for now.

> > +
> > +static const struct file_operations ne_fops = {
> > +	.owner		= THIS_MODULE,
> > +	.llseek		= noop_llseek,
> > +	.unlocked_ioctl	= ne_ioctl,
> > +	.open		= ne_open,
> > +	.release	= ne_release,
> > +};
> > +
> > +struct miscdevice ne_miscdevice = {
> > +	.minor	= MISC_DYNAMIC_MINOR,
> > +	.name	= NE_DEV_NAME,
> > +	.fops	= &ne_fops,
> > +	.mode	= 0660,
> > +};
> > +
> > +static int __init ne_init(void)
> > +{
> > +	unsigned int cpu = 0;
> > +	unsigned int cpu_sibling = 0;
> > +	int rc = -EINVAL;
> > +
> > +	memset(&ne_cpu_pool, 0, sizeof(ne_cpu_pool));
> Why did you just set a structure to 0 that was already initialized by
> the system to 0?  Are you sure about this?

True, this is not needed. Removed the memset() call.

> > +
> > +	if (!zalloc_cpumask_var(&ne_cpu_pool.avail, GFP_KERNEL))
> > +		return -ENOMEM;
> > +
> > +	mutex_init(&ne_cpu_pool.mutex);
> > +
> > +	rc = cpulist_parse(ne_cpus, ne_cpu_pool.avail);
> > +	if (rc < 0) {
> > +		pr_err_ratelimited(NE "Error in cpulist parse [rc=%d]\n", rc);
> Again, drop all ratelimited stuff please.

Updated to pr_err().

Thank you.

Andra



Amazon Development Center (Romania) S.R.L. registered office: 27A Sf. Lazar Street, UBC5, floor 2, Iasi, Iasi County, 700045, Romania. Registered in Romania. Registration number J22/2621/2005.