On Mon, Apr 9, 2012 at 2:38 AM, Anton Vorontsov
<anton.vorontsov@xxxxxxxxxx> wrote:
> This patch implements a new event type, it will trigger whenever a
> value crosses a user-specified threshold. It works two-way, i.e. when
> a value crosses the threshold from a lesser values side to a greater
> values side, and vice versa.
>
> We use the event type in an userspace low-memory killer: we get a
> notification when memory becomes low, so we start freeing memory by
> killing unneeded processes, and we get notification when memory hits
> the threshold from another side, so we know that we freed enough of
> memory.
>
> Signed-off-by: Anton Vorontsov <anton.vorontsov@xxxxxxxxxx>
> ---
> include/linux/vmevent.h | 9 +++++++++
> mm/vmevent.c | 21 +++++++++++++++++++++
> tools/testing/vmevent/vmevent-test.c | 15 ++++++++++-----
> 3 files changed, 40 insertions(+), 5 deletions(-)
>
> diff --git a/include/linux/vmevent.h b/include/linux/vmevent.h
> index 64357e4..00cc04f 100644
> --- a/include/linux/vmevent.h
> +++ b/include/linux/vmevent.h
> @@ -22,6 +22,15 @@ enum {
> * Sample value is less than user-specified value
> */
> VMEVENT_ATTR_STATE_VALUE_LT = (1UL << 0),
> + /*
> + * Sample value crossed user-specified value
> + */
> + VMEVENT_ATTR_STATE_VALUE_CROSS = (1UL << 2),
> +
> + /* Last saved state, used internally by the kernel. */
> + __VMEVENT_ATTR_STATE_LAST = (1UL << 30),
> + /* Not first sample, used internally by the kernel. */
> + __VMEVENT_ATTR_STATE_NFIRST = (1UL << 31),
> };
>
> struct vmevent_attr {
> diff --git a/mm/vmevent.c b/mm/vmevent.c
> index a56174f..f8fd2d6 100644
> --- a/mm/vmevent.c
> +++ b/mm/vmevent.c
> @@ -1,5 +1,6 @@
> #include <linux/anon_inodes.h>
> #include <linux/atomic.h>
> +#include <linux/compiler.h>
> #include <linux/vmevent.h>
> #include <linux/syscalls.h>
> #include <linux/timer.h>
> @@ -94,6 +95,26 @@ static bool vmevent_match(struct vmevent_watch *watch)
> if (attr->state & VMEVENT_ATTR_STATE_VALUE_LT) {
> if (value < attr->value)
> return true;
> + } else if (attr->state & VMEVENT_ATTR_STATE_VALUE_CROSS) {
> + bool fst = !(attr->state & __VMEVENT_ATTR_STATE_NFIRST);
> + bool old = attr->state & __VMEVENT_ATTR_STATE_LAST;
> + bool new = value < attr->value;
> + bool chg = old ^ new;
> + bool ret = chg;
> +
> + /*
> + * This is not 'lt' or 'gt' match, so on the first
> + * sample assume we crossed the threshold.
> + */
> + if (unlikely(fst)) {
> + attr->state |= __VMEVENT_ATTR_STATE_NFIRST;
> + ret = true;
> + }
> +
> + attr->state &= ~__VMEVENT_ATTR_STATE_LAST;
> + attr->state |= new ? __VMEVENT_ATTR_STATE_LAST : 0;
> +
> + return ret;
> }
> }
>
Can't we implement this by specifying both VMEVENT_ATTR_STATE_VALUE_LT
and VMEVENT_ATTR_STATE_VALUE_GT in userspace? I assume the problem
with current approach is that you get more than one notifications,
right? We can implement a "single-shot" flag to deal with that.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href