From: Lin Ming on
Add group scheduling transactional APIs to struct pmu.
These APIs will be implemented in arch code, based on Peter's idea as
below.

> the idea behind hw_perf_group_sched_in() is to not perform
> schedulability tests on each event in the group, but to add the group
as
> a whole and then perform one test.
>
> Of course, when that test fails, you'll have to roll-back the whole
> group again.
>
> So start_txn (or a better name) would simply toggle a flag in the pmu
> implementation that will make pmu::enable() not perform the
> schedulablilty test.
>
> Then commit_txn() will perform the schedulability test (so note the
> method has to have a !void return value, my mistake in the earlier
> email).
>
> This will allow us to use the regular
> kernel/perf_event.c::group_sched_in() and all the rollback code.
> Currently each hw_perf_group_sched_in() implementation duplicates all
> the rolllback code (with various bugs).

*_txn was renamed to *_group_trans for better readability.

->start_group_trans:
Start group events scheduling transaction, set a flag to make
pmu::enable() not perform the schedulability test, it will be performed
at commit time.

->commit_group_trans:
Commit group events scheduling transaction, perform the group
schedulability as a whole

->stop_group_trans:
Stop group events scheduling transaction, clear the flag so
pmu::enable() will perform the schedulability test.

Reviewed-by: Stephane Eranian <eranian(a)google.com>
Reviewed-by: Peter Zijlstra <peterz(a)infradead.org>
Reviewed-by: Frederic Weisbecker <fweisbec(a)gmail.com>
Signed-off-by: Lin Ming <ming.m.lin(a)intel.com>

---
include/linux/perf_event.h | 20 +++++++++++++++++---
kernel/perf_event.c | 33 ++++++++++++++++++++-------------
2 files changed, 37 insertions(+), 16 deletions(-)

diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index ace31fb..f514825 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -532,6 +532,8 @@ struct hw_perf_event {

struct perf_event;

+#define PERF_EVENT_TRAN_STARTED 1
+
/**
* struct pmu - generic performance monitoring unit
*/
@@ -542,6 +544,21 @@ struct pmu {
void (*stop) (struct perf_event *event);
void (*read) (struct perf_event *event);
void (*unthrottle) (struct perf_event *event);
+
+ /*
+ * group events scheduling is treated as a transaction,
+ * add group events as a whole and perform one schedulability test.
+ * If test fails, roll back the whole group
+ */
+
+ /* start group events transaction */
+ void (*start_group_trans) (const struct pmu *pmu);
+
+ /* stop group events transaction */
+ void (*stop_group_trans) (const struct pmu *pmu);
+
+ /* commit group events transaction */
+ int (*commit_group_trans) (const struct pmu *pmu);
};

/**
@@ -807,9 +824,6 @@ extern void perf_disable(void);
extern void perf_enable(void);
extern int perf_event_task_disable(void);
extern int perf_event_task_enable(void);
-extern int hw_perf_group_sched_in(struct perf_event *group_leader,
- struct perf_cpu_context *cpuctx,
- struct perf_event_context *ctx);
extern void perf_event_update_userpage(struct perf_event *event);
extern int perf_event_release_kernel(struct perf_event *event);
extern struct perf_event *
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 9dbe8cd..eac6fe7 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -83,14 +83,6 @@ extern __weak const struct pmu *hw_perf_event_init(struct perf_event *event)
void __weak hw_perf_disable(void) { barrier(); }
void __weak hw_perf_enable(void) { barrier(); }

-int __weak
-hw_perf_group_sched_in(struct perf_event *group_leader,
- struct perf_cpu_context *cpuctx,
- struct perf_event_context *ctx)
-{
- return 0;
-}
-
void __weak perf_event_print_debug(void) { }

static DEFINE_PER_CPU(int, perf_disable_count);
@@ -641,15 +633,20 @@ group_sched_in(struct perf_event *group_event,
struct perf_cpu_context *cpuctx,
struct perf_event_context *ctx)
{
- struct perf_event *event, *partial_group;
+ struct perf_event *event, *partial_group = NULL;
+ const struct pmu *pmu = group_event->pmu;
+ bool trans = false;
int ret;

if (group_event->state == PERF_EVENT_STATE_OFF)
return 0;

- ret = hw_perf_group_sched_in(group_event, cpuctx, ctx);
- if (ret)
- return ret < 0 ? ret : 0;
+ /* Check if group transaction availabe */
+ if (pmu->start_group_trans)
+ trans = true;
+
+ if (trans)
+ pmu->start_group_trans(pmu);

if (event_sched_in(group_event, cpuctx, ctx))
return -EAGAIN;
@@ -664,9 +661,19 @@ group_sched_in(struct perf_event *group_event,
}
}

- return 0;
+ if (trans) {
+ ret = pmu->commit_group_trans(pmu);
+ if (!ret) {
+ pmu->stop_group_trans(pmu);
+
+ return 0;
+ }
+ }

group_error:
+ if (trans)
+ pmu->stop_group_trans(pmu);
+
/*
* Groups can be scheduled in as one unit only, so undo any
* partial group before returning:



--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/