Re: [RFC 1/3] mm: Add VM pressure notifications

On Wed, 7 Nov 2012 03:01:28 -0800 Anton Vorontsov anton.vorontsov@linaro.org wrote:

This patch introduces vmpressure_fd() system call. The system call creates a new file descriptor that can be used to monitor Linux' virtual memory management pressure.

I noticed a couple of quick things as I was looking this over...

+static ssize_t vmpressure_read(struct file *file, char __user *buf,

• 	       size_t count, loff_t *ppos)
  

+{

• struct vmpressure_watch *watch = file->private_data;
• struct vmpressure_event event;
• int ret;
• if (count < sizeof(event))

• return -EINVAL;
  

• ret = wait_event_interruptible(watch->waitq,

• 		       atomic_read(&watch->pending));
  

Would it make sense to support non-blocking reads? Perhaps a process would like to simply know that current pressure level?

+SYSCALL_DEFINE1(vmpressure_fd, struct vmpressure_config __user *, config) +{

• struct vmpressure_watch *watch;
• struct file *file;
• int ret;
• int fd;
• watch = kzalloc(sizeof(*watch), GFP_KERNEL);
• if (!watch)

• return -ENOMEM;
  

• ret = copy_from_user(&watch->config, config, sizeof(*config));
• if (ret)

• goto err_free;
  

This is wrong - you'll return the number of uncopied bytes to user space. You'll need a "ret = -EFAULT;" in there somewhere.

jon