kvm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test
@ 2023-02-03 16:23 Alexandru Elisei
  2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 1/2] arm/psci: Test that CPU 1 has been successfully brought online Alexandru Elisei
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Alexandru Elisei @ 2023-02-03 16:23 UTC (permalink / raw)
  To: andrew.jones, kvm, kvmarm

The series adds a test for the PSCI function CPU_OFF. The test is
implemented in patch #2, "arm/psci: Add PSCI CPU_OFF test case".  Patch #1,
"arm/psci: Test that CPU 1 has been successfully brought online" is a
preparatory patch to allow the CPU_OFF test to run after the CPU_ON test.
Executing the CPU_OFF test after the CPU_ON test makes the most sense,
since CPU_OFF requires all the CPUs to be brought online before the test
can execute, and this is exactly what the CPU_ON test does.

I believe that proving that a CPU has been successfully offlined is an
undecidable problem - it might just be that the CPU is stuck at a higher
exception level doing something else entirely, and if the test were to wait
long enough, the CPU would return from the CPU_OFF call and start executing
code, thus failing to be offlined. Right now, the test waits for 100ms
after checking that all the other CPUs are offline. I thought this was a
good balance between making the test fast and being reasonably sure that
the offline succeeded. I'm open to suggestions here if anyone thinks
otherwise.

Tested for both the arm and arm64 architectures: on an x86 machine (qemu
with TCG), and a rockpro64 (qemu and kvmtool).

And Nikita is not longer at Arm, and I don't have a new email address where
she can be reached, so I didn't CC her.

Changelog:

v5 -> v6:
- Added Reviewed-by tag to #1 (Drew).
- Fixed the unintended change that added the argc/argv arguments to main()
  in #2 (Drew).
- Fixed bailing early when all secondaries are offline in #2 (Drew).
- Run the cpu-off test only if some of the secondaries are not idling
  instead if cpu-on has failed (cpu-on can fail for other reasons too)
  (Drew).

v4 -> v5:
- Open code smp_prepare_secondary in psci.c instead of exposing it as a
  library function in the cpu_on test (Drew).
- Don't return early if a CPU failed to online CPU1 in the cpu_on test, and
  check the return values for the CPUs that succeeded to online CPU 1
  (Drew).
- Drop cpu_off_success in the cpu_off test in favour of checking
  AFFINITY_INFO and the cpu idle mask (Drew).
- Hopefully made the cpu_off test faster (Drew).

v3 -> v4:
- Moved ownership of the series to Alexandru Elisei. All bugs are mine.
- Moved the timeout for the CPU_ON test to patch #1.
- Changed the include order for arm/psic.c in patch #1 to order the headers
  alphabetically.
- Run the CPU_OFF test only if CPU_ON succeeds, because CPU_OFF expects all
  CPUs to be online, the test would otherwise hang forever.
- Minor style changes here and there.

v2 -> v3:
- Add timeout so that test does not hang if CPU1 fails to come online
- Remove unnecessary call of on_cpus() in the condition where target CPU is not online.

v1 -> v2:
- Modify PSCI CPU_ON test to ensure CPU 1 remains online after the execution of the test.
- Addition of PSCI CPU_OFF test and calling it after PSCI CPU_ON test has been executed.

Alexandru Elisei (1):
  arm/psci: Test that CPU 1 has been successfully brought online

Nikita Venkatesh (1):
  arm/psci: Add PSCI CPU_OFF test case

 arm/psci.c        | 138 +++++++++++++++++++++++++++++++++++++++++-----
 lib/arm/asm/smp.h |   9 ++-
 lib/arm/smp.c     |   4 --
 lib/errata.h      |   2 +
 4 files changed, 134 insertions(+), 19 deletions(-)


base-commit: 2480430a36102f8ea276b3bfb1d64d5dacc23b8f
-- 
2.39.1


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [kvm-unit-tests PATCH v6 1/2] arm/psci: Test that CPU 1 has been successfully brought online
  2023-02-03 16:23 [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Alexandru Elisei
@ 2023-02-03 16:23 ` Alexandru Elisei
  2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 2/2] arm/psci: Add PSCI CPU_OFF test case Alexandru Elisei
  2023-02-14 20:18 ` [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Andrew Jones
  2 siblings, 0 replies; 4+ messages in thread
From: Alexandru Elisei @ 2023-02-03 16:23 UTC (permalink / raw)
  To: andrew.jones, kvm, kvmarm

For the PSCI CPU_ON function test, all other CPUs perform a CPU_ON call
that target CPU 1. The test is considered a success if CPU_ON returns PSCI
SUCCESS exactly once, and for the rest of the calls PSCI ALREADY_ON.

Enhance the test by checking that CPU 1 is actually online and able to
execute code. Also make the test more robust by checking that the CPU_ON
call returns, instead of assuming that it will always succeed and
hanging indefinitely if it doesn't.

Since the CPU 1 thread is now being set up properly by kvm-unit-tests
when being brought online, it becomes possible to add other tests in the
future that require all CPUs.

The include header order in arm/psci.c has been changed to be in
alphabetic order. This means moving the errata.h include before
libcflat.h, which causes compilation to fail because of missing includes
in errata.h. Fix that also by including the needed header in errata.h.

Reviewed-by: Andrew Jones <andrew.jones@linux.dev>
Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
---
 arm/psci.c        | 67 ++++++++++++++++++++++++++++++++++++++---------
 lib/arm/asm/smp.h |  9 ++++++-
 lib/arm/smp.c     |  4 ---
 lib/errata.h      |  2 ++
 4 files changed, 64 insertions(+), 18 deletions(-)

diff --git a/arm/psci.c b/arm/psci.c
index efa0722c0566..f7238f8e0bbd 100644
--- a/arm/psci.c
+++ b/arm/psci.c
@@ -7,11 +7,14 @@
  *
  * This work is licensed under the terms of the GNU LGPL, version 2.
  */
-#include <libcflat.h>
 #include <errata.h>
+#include <libcflat.h>
+
+#include <asm/delay.h>
+#include <asm/mmu.h>
 #include <asm/processor.h>
-#include <asm/smp.h>
 #include <asm/psci.h>
+#include <asm/smp.h>
 
 static bool invalid_function_exception;
 
@@ -72,46 +75,84 @@ static int cpu_on_ret[NR_CPUS];
 static cpumask_t cpu_on_ready, cpu_on_done;
 static volatile int cpu_on_start;
 
-static void cpu_on_secondary_entry(void)
+extern void secondary_entry(void);
+static void cpu_on_do_wake_target(void)
 {
 	int cpu = smp_processor_id();
 
 	cpumask_set_cpu(cpu, &cpu_on_ready);
 	while (!cpu_on_start)
 		cpu_relax();
-	cpu_on_ret[cpu] = psci_cpu_on(cpus[1], __pa(halt));
+	cpu_on_ret[cpu] = psci_cpu_on(cpus[1], __pa(secondary_entry));
 	cpumask_set_cpu(cpu, &cpu_on_done);
 }
 
+static void cpu_on_target(void)
+{
+	int cpu = smp_processor_id();
+
+	cpumask_set_cpu(cpu, &cpu_on_done);
+}
+
+extern struct secondary_data secondary_data;
+
+/* Open code the setup part from smp_boot_secondary(). */
+static void psci_cpu_on_prepare_secondary(int cpu, secondary_entry_fn entry)
+{
+	secondary_data.stack = thread_stack_alloc();
+	secondary_data.entry = entry;
+	mmu_mark_disabled(cpu);
+}
+
 static bool psci_cpu_on_test(void)
 {
 	bool failed = false;
 	int ret_success = 0;
-	int cpu;
-
-	cpumask_set_cpu(1, &cpu_on_ready);
-	cpumask_set_cpu(1, &cpu_on_done);
+	int i, cpu;
 
 	for_each_present_cpu(cpu) {
 		if (cpu < 2)
 			continue;
-		smp_boot_secondary(cpu, cpu_on_secondary_entry);
+		smp_boot_secondary(cpu, cpu_on_do_wake_target);
 	}
 
 	cpumask_set_cpu(0, &cpu_on_ready);
+	cpumask_set_cpu(1, &cpu_on_ready);
 	while (!cpumask_full(&cpu_on_ready))
 		cpu_relax();
 
+	/*
+	 * Configure CPU 1 after all secondaries are online to avoid
+	 * secondary_data being overwritten.
+	 */
+	psci_cpu_on_prepare_secondary(1, cpu_on_target);
+
 	cpu_on_start = 1;
 	smp_mb();
 
-	cpu_on_ret[0] = psci_cpu_on(cpus[1], __pa(halt));
+	cpu_on_ret[0] = psci_cpu_on(cpus[1], __pa(secondary_entry));
 	cpumask_set_cpu(0, &cpu_on_done);
 
-	while (!cpumask_full(&cpu_on_done))
-		cpu_relax();
+	report_info("waiting for CPU1 to come online...");
+	for (i = 0; i < 100; i++) {
+		mdelay(10);
+		if (cpumask_full(&cpu_on_done))
+			break;
+	}
 
-	for_each_present_cpu(cpu) {
+	if (!cpumask_full(&cpu_on_done)) {
+		for_each_present_cpu(cpu) {
+			if (!cpumask_test_cpu(cpu, &cpu_on_done)) {
+				if (cpu == 1)
+					report_info("CPU1 failed to come online");
+				else
+					report_info("CPU%d failed to online CPU1", cpu);
+			}
+		}
+		failed = true;
+	}
+
+	for_each_cpu(cpu, &cpu_on_done) {
 		if (cpu == 1)
 			continue;
 		if (cpu_on_ret[cpu] == PSCI_RET_SUCCESS) {
diff --git a/lib/arm/asm/smp.h b/lib/arm/asm/smp.h
index 077afde85520..dee4c1a883e7 100644
--- a/lib/arm/asm/smp.h
+++ b/lib/arm/asm/smp.h
@@ -10,6 +10,14 @@
 
 #define smp_processor_id()		(current_thread_info()->cpu)
 
+typedef void (*secondary_entry_fn)(void);
+
+struct secondary_data {
+	void *stack;            /* must be first member of struct */
+	secondary_entry_fn entry;
+};
+extern struct secondary_data secondary_data;
+
 extern bool cpu0_calls_idle;
 
 extern void halt(void);
@@ -48,7 +56,6 @@ static inline void set_cpu_idle(int cpu, bool idle)
 		cpumask_clear_cpu(cpu, &cpu_idle_mask);
 }
 
-typedef void (*secondary_entry_fn)(void);
 extern void smp_boot_secondary(int cpu, secondary_entry_fn entry);
 extern void on_cpu_async(int cpu, void (*func)(void *data), void *data);
 extern void on_cpu(int cpu, void (*func)(void *data), void *data);
diff --git a/lib/arm/smp.c b/lib/arm/smp.c
index 98a5054e039b..1d470d1aab45 100644
--- a/lib/arm/smp.c
+++ b/lib/arm/smp.c
@@ -21,10 +21,6 @@ cpumask_t cpu_present_mask;
 cpumask_t cpu_online_mask;
 cpumask_t cpu_idle_mask;
 
-struct secondary_data {
-	void *stack;            /* must be first member of struct */
-	secondary_entry_fn entry;
-};
 struct secondary_data secondary_data;
 static struct spinlock lock;
 
diff --git a/lib/errata.h b/lib/errata.h
index 5af0eb3bf8e2..de8205d8b370 100644
--- a/lib/errata.h
+++ b/lib/errata.h
@@ -6,6 +6,8 @@
  */
 #ifndef _ERRATA_H_
 #define _ERRATA_H_
+#include <libcflat.h>
+
 #include "config.h"
 
 #ifndef CONFIG_ERRATA_FORCE
-- 
2.39.1


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [kvm-unit-tests PATCH v6 2/2] arm/psci: Add PSCI CPU_OFF test case
  2023-02-03 16:23 [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Alexandru Elisei
  2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 1/2] arm/psci: Test that CPU 1 has been successfully brought online Alexandru Elisei
@ 2023-02-03 16:23 ` Alexandru Elisei
  2023-02-14 20:18 ` [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Andrew Jones
  2 siblings, 0 replies; 4+ messages in thread
From: Alexandru Elisei @ 2023-02-03 16:23 UTC (permalink / raw)
  To: andrew.jones, kvm, kvmarm; +Cc: Nikita Venkatesh

From: Nikita Venkatesh <Nikita.Venkatesh@arm.com>

The test uses the following method.

The primary CPU brings up all the secondary CPUs, which are held in a wait
loop. Once the primary releases the CPUs, each of the secondary CPUs
proceed to issue CPU_OFF.

The primary CPU then checks for the status of the individual CPU_OFF
request. There is a chance that some CPUs might return from the CPU_OFF
function call after the primary CPU has finished the scan. There is no
foolproof method to handle this, but the test tries its best to
eliminate these false positives by introducing an extra delay if all the
CPUs are reported offline after the initial scan.

Signed-off-by: Nikita Venkatesh <Nikita.Venkatesh@arm.com>
[ Alex E: Skip CPU_OFF test if cpu onlining failed, drop cpu_off_success in
          favour of checking AFFINITY_INFO, commit message tweaking ]
Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
---
 arm/psci.c | 71 +++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 70 insertions(+), 1 deletion(-)

diff --git a/arm/psci.c b/arm/psci.c
index f7238f8e0bbd..55308c8f4766 100644
--- a/arm/psci.c
+++ b/arm/psci.c
@@ -72,8 +72,9 @@ static bool psci_affinity_info_off(void)
 }
 
 static int cpu_on_ret[NR_CPUS];
-static cpumask_t cpu_on_ready, cpu_on_done;
+static cpumask_t cpu_on_ready, cpu_on_done, cpu_off_done;
 static volatile int cpu_on_start;
+static volatile int cpu_off_start;
 
 extern void secondary_entry(void);
 static void cpu_on_do_wake_target(void)
@@ -171,6 +172,67 @@ static bool psci_cpu_on_test(void)
 	return !failed;
 }
 
+static void cpu_off_secondary_entry(void *data)
+{
+	int cpu = smp_processor_id();
+
+	while (!cpu_off_start)
+		cpu_relax();
+	cpumask_set_cpu(cpu, &cpu_off_done);
+	cpu_psci_cpu_die();
+}
+
+static bool psci_cpu_off_test(void)
+{
+	bool failed = false;
+	int i, count, cpu;
+
+	for_each_present_cpu(cpu) {
+		if (cpu == 0)
+			continue;
+		on_cpu_async(cpu, cpu_off_secondary_entry, NULL);
+	}
+
+	cpumask_set_cpu(0, &cpu_off_done);
+
+	cpu_off_start = 1;
+	report_info("waiting for the CPUs to be offlined...");
+	while (!cpumask_full(&cpu_off_done))
+		cpu_relax();
+
+	/* Allow all the other CPUs to complete the operation */
+	for (i = 0; i < 100; i++) {
+		mdelay(10);
+
+		count = 0;
+		for_each_present_cpu(cpu) {
+			if (cpu == 0)
+				continue;
+			if (psci_affinity_info(cpus[cpu], 0) != PSCI_0_2_AFFINITY_LEVEL_OFF)
+				count++;
+		}
+		if (count == 0)
+			break;
+	}
+
+	/* Try to catch CPUs that return from CPU_OFF. */
+	if (count == 0)
+		mdelay(100);
+
+	for_each_present_cpu(cpu) {
+		if (cpu == 0)
+			continue;
+		if (cpu_idle(cpu)) {
+			report_info("CPU%d failed to be offlined", cpu);
+			if (psci_affinity_info(cpus[cpu], 0) == PSCI_0_2_AFFINITY_LEVEL_OFF)
+				report_info("AFFINITY_INFO incorrectly reports CPU%d as offline", cpu);
+			failed = true;
+		}
+	}
+
+	return !failed;
+}
+
 int main(void)
 {
 	int ver = psci_invoke(PSCI_0_2_FN_PSCI_VERSION, 0, 0, 0);
@@ -193,6 +255,13 @@ int main(void)
 	else
 		report_skip("Skipping unsafe cpu-on test. Set ERRATA_6c7a5dce22b3=y to enable.");
 
+	assert(!cpu_idle(0));
+
+	if (!ERRATA(6c7a5dce22b3) || cpumask_weight(&cpu_idle_mask) == nr_cpus - 1)
+		report(psci_cpu_off_test(), "cpu-off");
+	else
+		report_skip("Skipping cpu-off test because the cpu-on test failed");
+
 done:
 #if 0
 	report_summary();
-- 
2.39.1


^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test
  2023-02-03 16:23 [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Alexandru Elisei
  2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 1/2] arm/psci: Test that CPU 1 has been successfully brought online Alexandru Elisei
  2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 2/2] arm/psci: Add PSCI CPU_OFF test case Alexandru Elisei
@ 2023-02-14 20:18 ` Andrew Jones
  2 siblings, 0 replies; 4+ messages in thread
From: Andrew Jones @ 2023-02-14 20:18 UTC (permalink / raw)
  To: Alexandru Elisei; +Cc: kvm, kvmarm

On Fri, Feb 03, 2023 at 04:23:51PM +0000, Alexandru Elisei wrote:
> The series adds a test for the PSCI function CPU_OFF. The test is
> implemented in patch #2, "arm/psci: Add PSCI CPU_OFF test case".  Patch #1,
> "arm/psci: Test that CPU 1 has been successfully brought online" is a
> preparatory patch to allow the CPU_OFF test to run after the CPU_ON test.
> Executing the CPU_OFF test after the CPU_ON test makes the most sense,
> since CPU_OFF requires all the CPUs to be brought online before the test
> can execute, and this is exactly what the CPU_ON test does.
> 
> I believe that proving that a CPU has been successfully offlined is an
> undecidable problem - it might just be that the CPU is stuck at a higher
> exception level doing something else entirely, and if the test were to wait
> long enough, the CPU would return from the CPU_OFF call and start executing
> code, thus failing to be offlined. Right now, the test waits for 100ms
> after checking that all the other CPUs are offline. I thought this was a
> good balance between making the test fast and being reasonably sure that
> the offline succeeded. I'm open to suggestions here if anyone thinks
> otherwise.
> 
> Tested for both the arm and arm64 architectures: on an x86 machine (qemu
> with TCG), and a rockpro64 (qemu and kvmtool).
> 
> And Nikita is not longer at Arm, and I don't have a new email address where
> she can be reached, so I didn't CC her.
> 
> Changelog:
> 
> v5 -> v6:
> - Added Reviewed-by tag to #1 (Drew).
> - Fixed the unintended change that added the argc/argv arguments to main()
>   in #2 (Drew).
> - Fixed bailing early when all secondaries are offline in #2 (Drew).
> - Run the cpu-off test only if some of the secondaries are not idling
>   instead if cpu-on has failed (cpu-on can fail for other reasons too)
>   (Drew).
> 
> v4 -> v5:
> - Open code smp_prepare_secondary in psci.c instead of exposing it as a
>   library function in the cpu_on test (Drew).
> - Don't return early if a CPU failed to online CPU1 in the cpu_on test, and
>   check the return values for the CPUs that succeeded to online CPU 1
>   (Drew).
> - Drop cpu_off_success in the cpu_off test in favour of checking
>   AFFINITY_INFO and the cpu idle mask (Drew).
> - Hopefully made the cpu_off test faster (Drew).
> 
> v3 -> v4:
> - Moved ownership of the series to Alexandru Elisei. All bugs are mine.
> - Moved the timeout for the CPU_ON test to patch #1.
> - Changed the include order for arm/psic.c in patch #1 to order the headers
>   alphabetically.
> - Run the CPU_OFF test only if CPU_ON succeeds, because CPU_OFF expects all
>   CPUs to be online, the test would otherwise hang forever.
> - Minor style changes here and there.
> 
> v2 -> v3:
> - Add timeout so that test does not hang if CPU1 fails to come online
> - Remove unnecessary call of on_cpus() in the condition where target CPU is not online.
> 
> v1 -> v2:
> - Modify PSCI CPU_ON test to ensure CPU 1 remains online after the execution of the test.
> - Addition of PSCI CPU_OFF test and calling it after PSCI CPU_ON test has been executed.
> 
> Alexandru Elisei (1):
>   arm/psci: Test that CPU 1 has been successfully brought online
> 
> Nikita Venkatesh (1):
>   arm/psci: Add PSCI CPU_OFF test case
> 
>  arm/psci.c        | 138 +++++++++++++++++++++++++++++++++++++++++-----
>  lib/arm/asm/smp.h |   9 ++-
>  lib/arm/smp.c     |   4 --
>  lib/errata.h      |   2 +
>  4 files changed, 134 insertions(+), 19 deletions(-)
> 
> 
> base-commit: 2480430a36102f8ea276b3bfb1d64d5dacc23b8f
> -- 
> 2.39.1
>

Applied, thanks

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2023-02-14 20:18 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-03 16:23 [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Alexandru Elisei
2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 1/2] arm/psci: Test that CPU 1 has been successfully brought online Alexandru Elisei
2023-02-03 16:23 ` [kvm-unit-tests PATCH v6 2/2] arm/psci: Add PSCI CPU_OFF test case Alexandru Elisei
2023-02-14 20:18 ` [kvm-unit-tests PATCH v6 0/2] arm: Add PSCI CPU_OFF test Andrew Jones

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).