Re: [PATCH 2/2] tracing: Add test for user space strings when filtering on string pointers

[Sven Schnelle <svens@linux.ibm.com>]

Hi Steve,

Steven Rostedt <rostedt@goodmis.org> writes:

> From: Steven Rostedt <rostedt@goodmis.org>
>
> Pingfan reported that the following causes a fault:
>
>   echo "filename ~ \"cpu\"" > events/syscalls/sys_enter_openat/filter
>   echo 1 > events/syscalls/sys_enter_at/enable
>

[..]

> +static __always_inline char *test_string(char *str)
> +{
> +	struct ustring_buffer *ubuf;
> +	char __user *ustr;
> +	char *kstr;
> +
> +	if (!ustring_per_cpu)
> +		return NULL;
> +
> +	ubuf = this_cpu_ptr(ustring_per_cpu);
> +	kstr = ubuf->buffer;
> +
> +	if (likely((unsigned long)str >= TASK_SIZE)) {

I think that would not work on architectures where addresses for kernel
and user space could overlap, i.e. with different address spaces?

> +		/* For safety, do not trust the string pointer */
> +		if (!strncpy_from_kernel_nofault(kstr, str, USTRING_BUF_SIZE))
> +			return NULL;
> +	} else {
> +		/* user space address? */
> +		ustr = str;
> +		if (!strncpy_from_user_nofault(kstr, ustr, USTRING_BUF_SIZE))
> +			return NULL;
> +	}
> +	return kstr;
> +}