On Mon, Feb 22, 2021 at 11:11:33AM +0100, Pavel Machek wrote:

Hi!

quoted

quoted

+* COUNTER_ENABLE_EVENTS_IOCTL:
+  Enables monitoring the events specified by the Counter watches that
+  were queued by ``COUNTER_ADD_WATCH_IOCTL``. If events are already
+  enabled, the new set of watches replaces the old one. Calling this
+  ioctl also has the effect of clearing the queue of watches added by
+  ``COUNTER_ADD_WATCH_IOCTL``.
+
+* COUNTER_DISABLE_EVENTS_IOCTL:
+  Stops monitoring the previously enabled events.

Is there a way to remove a watch? 

Is there a way to eat all kernel memory and crash the system by adding
too many watches?

There can only ever be as many watches as there are Counter components
for the respective Counter device. This is enforced by
counter_set_event_node() which checks whether a particular watch has
been created before and returns an EINVAL if it so has.

quoted

quoted

+For example, the following userspace code opens ``/dev/counter0``,
+configures the ``COUNTER_EVENT_INDEX`` event channel 0 to gather Count 0
+and Count 1, and prints out the data as it becomes available on the
+character device node::
+

Consider adding an example program under tools/

quoted

+        #include <fcntl.h>
+        #include <linux/counter.h>
+        #include <stdio.h>
+        #include <string.h>
+        #include <sys/ioctl.h>
+        #include <unistd.h>
+
+        struct counter_watch watches[2] = {
+                {
+                        .component.type = COUNTER_COMPONENT_COUNT,
+                        .component.scope = COUNTER_SCOPE_COUNT,
+                        .component.parent = 0,

Good to add comments on what these elements actually are?

quoted

quoted

+                fd = open("/dev/counter0", O_RDWR);
+
+                ioctl(fd, COUNTER_ADD_WATCH_IOCTL, watches);
+                ioctl(fd, COUNTER_ADD_WATCH_IOCTL, watches + 1);
+                ioctl(fd, COUNTER_ENABLE_EVENTS_IOCTL);

quoted

quoted

+                for (;;) {
+                        read(fd, event_data, sizeof(event_data));

If this goes to tools it really should have error handling and
handling of short read.

Best regards,
							Pavel

-- 
http://www.livejournal.com/~pavelmachek

Ack. I'll improve the error handling for this.

William Breathitt Gray