Re: [PATCH 5/5] tracing: add binary buffer files for use with splice

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 03 Mar 2009 21:49:26 -0500 Steven Rostedt <rostedt@goodmis.org> wrote:

> +static ssize_t
> +tracing_buffers_read(struct file *filp, char __user *ubuf,
> +		     size_t count, loff_t *ppos)
> +{
> +	struct ftrace_buffer_info *info = filp->private_data;
> +	unsigned int pos;
> +	ssize_t ret;
> +	size_t size;
> +
> +	/* Do we have previous read data to read? */
> +	if (info->read < PAGE_SIZE)
> +		goto read;
> +
> +	info->read = 0;
> +
> +	ret = ring_buffer_read_page(info->tr->buffer,
> +				    &info->spare,
> +				    count,
> +				    info->cpu, 0);
> +	if (ret < 0)
> +		return 0;
> +
> +	pos = ring_buffer_page_len(info->spare);
> +
> +	if (pos < PAGE_SIZE)
> +		memset(info->spare + pos, 0, PAGE_SIZE - pos);
> +
> +read:
> +	size = PAGE_SIZE - info->read;
> +	if (size > count)
> +		size = count;
> +
> +	ret = copy_to_user(ubuf, info->spare + info->read, size);
> +	if (ret)
> +		return -EFAULT;

Conventionally a read() system call will return the number of bytes
copied, and will only return -EFOO if the number of bytes copied was
zero.

Lots of parts of the kernel break this, but it's usually device drivers
and scruffy pseudo files, in which case a partial file read doesn't
make much sense.  This doesn't make the broken behaviour right, but at
least we have a bit of a weaselly excuse in that case.

> +	*ppos += size;
> +	info->read += size;
> +
> +	return size;
> +}