[PATCH v2 0/8] riscv: Unaligned access speed probing fixes and skipping

[PATCH v2 0/8] riscv: Unaligned access speed probing fixes and skipping

[Andrew Jones]

The first six patches of this series are fixes and cleanups of the
unaligned access speed probing code. The next patch introduces a
kernel command line option that allows the probing to be skipped.
This command line option is a different approach than Jesse's [1].
[1] takes a cpu-list for a particular speed, supporting heterogeneous
platforms. With this approach, the kernel command line should only
be used for homogeneous platforms. [1] also only allowed 'fast' and
'slow' to be selected. This parameter also supports 'unsupported',
which could be useful for testing code paths gated on that. The final
patch adds the documentation.

(I'd be happy to split the fixes from the new skip support if we want to
discuss the skip support independently, but I want to base on the fixes
and I'm not sure if patchwork supports Based-on: $MESSAGE_ID/$LORE_URL
or not at the moment, so I'm just posting together for now in order to
be able to check for my patchwork green lights!)

[1] https://lore.kernel.org/linux-riscv/20240805173816.3722002-1-jesse@rivosinc.com/

Thanks,
drew


Andrew Jones (8):
  riscv: Annotate unaligned access init functions
  riscv: Fix riscv_online_cpu_vec
  riscv: Fix check_unaligned_access_all_cpus
  riscv: Change check_unaligned_access_speed_all_cpus to void
  riscv: Fix set up of cpu hotplug callbacks
  riscv: Fix set up of vector cpu hotplug callback
  riscv: Add parameter for skipping access speed tests
  Documentation/kernel-parameters: Add riscv unaligned speed parameters

 .../admin-guide/kernel-parameters.txt         |  16 ++
 arch/riscv/include/asm/cpufeature.h           |   4 +-
 arch/riscv/kernel/traps_misaligned.c          |  14 +-
 arch/riscv/kernel/unaligned_access_speed.c    | 212 +++++++++++-------
 4 files changed, 154 insertions(+), 92 deletions(-)

-- 
2.48.1

[PATCH v2 1/8] riscv: Annotate unaligned access init functions

[Andrew Jones]

Several functions used in unaligned access probing are only run at
init time. Annotate them appropriately.

Fixes: f413aae96cda ("riscv: Set unaligned access speed at compile time")
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/include/asm/cpufeature.h        |  4 ++--
 arch/riscv/kernel/traps_misaligned.c       |  8 ++++----
 arch/riscv/kernel/unaligned_access_speed.c | 14 +++++++-------
 3 files changed, 13 insertions(+), 13 deletions(-)

diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h
index 569140d6e639..19defdc2002d 100644
--- a/arch/riscv/include/asm/cpufeature.h
+++ b/arch/riscv/include/asm/cpufeature.h
@@ -63,7 +63,7 @@ void __init riscv_user_isa_enable(void);
 #define __RISCV_ISA_EXT_SUPERSET_VALIDATE(_name, _id, _sub_exts, _validate) \
 	_RISCV_ISA_EXT_DATA(_name, _id, _sub_exts, ARRAY_SIZE(_sub_exts), _validate)
 
-bool check_unaligned_access_emulated_all_cpus(void);
+bool __init check_unaligned_access_emulated_all_cpus(void);
 #if defined(CONFIG_RISCV_SCALAR_MISALIGNED)
 void check_unaligned_access_emulated(struct work_struct *work __always_unused);
 void unaligned_emulation_finish(void);
@@ -76,7 +76,7 @@ static inline bool unaligned_ctl_available(void)
 }
 #endif
 
-bool check_vector_unaligned_access_emulated_all_cpus(void);
+bool __init check_vector_unaligned_access_emulated_all_cpus(void);
 #if defined(CONFIG_RISCV_VECTOR_MISALIGNED)
 void check_vector_unaligned_access_emulated(struct work_struct *work __always_unused);
 DECLARE_PER_CPU(long, vector_misaligned_access);
diff --git a/arch/riscv/kernel/traps_misaligned.c b/arch/riscv/kernel/traps_misaligned.c
index 7cc108aed74e..aacbd9d7196e 100644
--- a/arch/riscv/kernel/traps_misaligned.c
+++ b/arch/riscv/kernel/traps_misaligned.c
@@ -605,7 +605,7 @@ void check_vector_unaligned_access_emulated(struct work_struct *work __always_un
 	kernel_vector_end();
 }
 
-bool check_vector_unaligned_access_emulated_all_cpus(void)
+bool __init check_vector_unaligned_access_emulated_all_cpus(void)
 {
 	int cpu;
 
@@ -625,7 +625,7 @@ bool check_vector_unaligned_access_emulated_all_cpus(void)
 	return true;
 }
 #else
-bool check_vector_unaligned_access_emulated_all_cpus(void)
+bool __init check_vector_unaligned_access_emulated_all_cpus(void)
 {
 	return false;
 }
@@ -659,7 +659,7 @@ void check_unaligned_access_emulated(struct work_struct *work __always_unused)
 	}
 }
 
-bool check_unaligned_access_emulated_all_cpus(void)
+bool __init check_unaligned_access_emulated_all_cpus(void)
 {
 	int cpu;
 
@@ -684,7 +684,7 @@ bool unaligned_ctl_available(void)
 	return unaligned_ctl;
 }
 #else
-bool check_unaligned_access_emulated_all_cpus(void)
+bool __init check_unaligned_access_emulated_all_cpus(void)
 {
 	return false;
 }
diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index 91f189cf1611..b7a8ff7ba6df 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -121,7 +121,7 @@ static int check_unaligned_access(void *param)
 	return 0;
 }
 
-static void check_unaligned_access_nonboot_cpu(void *param)
+static void __init check_unaligned_access_nonboot_cpu(void *param)
 {
 	unsigned int cpu = smp_processor_id();
 	struct page **pages = param;
@@ -175,7 +175,7 @@ static void set_unaligned_access_static_branches(void)
 	modify_unaligned_access_branches(&fast_and_online, num_online_cpus());
 }
 
-static int lock_and_set_unaligned_access_static_branch(void)
+static int __init lock_and_set_unaligned_access_static_branch(void)
 {
 	cpus_read_lock();
 	set_unaligned_access_static_branches();
@@ -218,7 +218,7 @@ static int riscv_offline_cpu(unsigned int cpu)
 }
 
 /* Measure unaligned access speed on all CPUs present at boot in parallel. */
-static int check_unaligned_access_speed_all_cpus(void)
+static int __init check_unaligned_access_speed_all_cpus(void)
 {
 	unsigned int cpu;
 	unsigned int cpu_count = num_possible_cpus();
@@ -264,7 +264,7 @@ static int check_unaligned_access_speed_all_cpus(void)
 	return 0;
 }
 #else /* CONFIG_RISCV_PROBE_UNALIGNED_ACCESS */
-static int check_unaligned_access_speed_all_cpus(void)
+static int __init check_unaligned_access_speed_all_cpus(void)
 {
 	return 0;
 }
@@ -379,7 +379,7 @@ static int riscv_online_cpu_vec(unsigned int cpu)
 }
 
 /* Measure unaligned access speed on all CPUs present at boot in parallel. */
-static int vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
+static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
 {
 	schedule_on_each_cpu(check_vector_unaligned_access);
 
@@ -393,13 +393,13 @@ static int vec_check_unaligned_access_speed_all_cpus(void *unused __always_unuse
 	return 0;
 }
 #else /* CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS */
-static int vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
+static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
 {
 	return 0;
 }
 #endif
 
-static int check_unaligned_access_all_cpus(void)
+static int __init check_unaligned_access_all_cpus(void)
 {
 	bool all_cpus_emulated, all_cpus_vec_unsupported;
 
-- 
2.48.1

[PATCH v2 2/8] riscv: Fix riscv_online_cpu_vec

[Andrew Jones]

We shouldn't probe when we already know vector is unsupported and
we should probe when we see we don't yet know whether it's supported.
Furthermore, we should ensure we've set the access type to
unsupported when we don't have vector at all.

Fixes: e7c9d66e313b ("RISC-V: Report vector unaligned access speed hwprobe")
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/kernel/unaligned_access_speed.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index b7a8ff7ba6df..161964cf2abc 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -367,10 +367,12 @@ static void check_vector_unaligned_access(struct work_struct *work __always_unus
 
 static int riscv_online_cpu_vec(unsigned int cpu)
 {
-	if (!has_vector())
+	if (!has_vector()) {
+		per_cpu(vector_misaligned_access, cpu) = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
 		return 0;
+	}
 
-	if (per_cpu(vector_misaligned_access, cpu) != RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED)
+	if (per_cpu(vector_misaligned_access, cpu) != RISCV_HWPROBE_MISALIGNED_VECTOR_UNKNOWN)
 		return 0;
 
 	check_vector_unaligned_access_emulated(NULL);
-- 
2.48.1

[PATCH v2 3/8] riscv: Fix check_unaligned_access_all_cpus

[Andrew Jones]

check_vector_unaligned_access_emulated_all_cpus(), like its name
suggests, will return true when all cpus emulate unaligned vector
accesses. If the function returned false it may have been because
vector isn't supported at all (!has_vector()) or because at least
one cpu doesn't emulate unaligned vector accesses. Since false may
be returned for two cases, checking for it isn't sufficient when
attempting to determine if we should proceed with the vector speed
check. Move the !has_vector() functionality to
check_unaligned_access_all_cpus() in order for
check_vector_unaligned_access_emulated_all_cpus() to return false
for a single case.

Fixes: e7c9d66e313b ("RISC-V: Report vector unaligned access speed hwprobe")
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/kernel/traps_misaligned.c       |  6 ------
 arch/riscv/kernel/unaligned_access_speed.c | 11 +++++++----
 2 files changed, 7 insertions(+), 10 deletions(-)

diff --git a/arch/riscv/kernel/traps_misaligned.c b/arch/riscv/kernel/traps_misaligned.c
index aacbd9d7196e..4354c87c0376 100644
--- a/arch/riscv/kernel/traps_misaligned.c
+++ b/arch/riscv/kernel/traps_misaligned.c
@@ -609,12 +609,6 @@ bool __init check_vector_unaligned_access_emulated_all_cpus(void)
 {
 	int cpu;
 
-	if (!has_vector()) {
-		for_each_online_cpu(cpu)
-			per_cpu(vector_misaligned_access, cpu) = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
-		return false;
-	}
-
 	schedule_on_each_cpu(check_vector_unaligned_access_emulated);
 
 	for_each_online_cpu(cpu)
diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index 161964cf2abc..02b485dc4bc4 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -403,13 +403,16 @@ static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __alway
 
 static int __init check_unaligned_access_all_cpus(void)
 {
-	bool all_cpus_emulated, all_cpus_vec_unsupported;
+	bool all_cpus_emulated;
+	int cpu;
 
 	all_cpus_emulated = check_unaligned_access_emulated_all_cpus();
-	all_cpus_vec_unsupported = check_vector_unaligned_access_emulated_all_cpus();
 
-	if (!all_cpus_vec_unsupported &&
-	    IS_ENABLED(CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS)) {
+	if (!has_vector()) {
+		for_each_online_cpu(cpu)
+			per_cpu(vector_misaligned_access, cpu) = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
+	} else if (!check_vector_unaligned_access_emulated_all_cpus() &&
+		   IS_ENABLED(CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS)) {
 		kthread_run(vec_check_unaligned_access_speed_all_cpus,
 			    NULL, "vec_check_unaligned_access_speed_all_cpus");
 	}
-- 
2.48.1

[PATCH v2 4/8] riscv: Change check_unaligned_access_speed_all_cpus to void

[Andrew Jones]

The return value of check_unaligned_access_speed_all_cpus() is always
zero, so make the function void so we don't need to concern ourselves
with it. The change also allows us to tidy up
check_unaligned_access_all_cpus() a bit.

Reviewed-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/kernel/unaligned_access_speed.c | 15 +++++----------
 1 file changed, 5 insertions(+), 10 deletions(-)

diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index 02b485dc4bc4..780f1c5f512a 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -218,7 +218,7 @@ static int riscv_offline_cpu(unsigned int cpu)
 }
 
 /* Measure unaligned access speed on all CPUs present at boot in parallel. */
-static int __init check_unaligned_access_speed_all_cpus(void)
+static void __init check_unaligned_access_speed_all_cpus(void)
 {
 	unsigned int cpu;
 	unsigned int cpu_count = num_possible_cpus();
@@ -226,7 +226,7 @@ static int __init check_unaligned_access_speed_all_cpus(void)
 
 	if (!bufs) {
 		pr_warn("Allocation failure, not measuring misaligned performance\n");
-		return 0;
+		return;
 	}
 
 	/*
@@ -261,12 +261,10 @@ static int __init check_unaligned_access_speed_all_cpus(void)
 	}
 
 	kfree(bufs);
-	return 0;
 }
 #else /* CONFIG_RISCV_PROBE_UNALIGNED_ACCESS */
-static int __init check_unaligned_access_speed_all_cpus(void)
+static void __init check_unaligned_access_speed_all_cpus(void)
 {
-	return 0;
 }
 #endif
 
@@ -403,10 +401,10 @@ static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __alway
 
 static int __init check_unaligned_access_all_cpus(void)
 {
-	bool all_cpus_emulated;
 	int cpu;
 
-	all_cpus_emulated = check_unaligned_access_emulated_all_cpus();
+	if (!check_unaligned_access_emulated_all_cpus())
+		check_unaligned_access_speed_all_cpus();
 
 	if (!has_vector()) {
 		for_each_online_cpu(cpu)
@@ -417,9 +415,6 @@ static int __init check_unaligned_access_all_cpus(void)
 			    NULL, "vec_check_unaligned_access_speed_all_cpus");
 	}
 
-	if (!all_cpus_emulated)
-		return check_unaligned_access_speed_all_cpus();
-
 	return 0;
 }
 
-- 
2.48.1

[PATCH v2 5/8] riscv: Fix set up of cpu hotplug callbacks

[Andrew Jones]

CPU hotplug callbacks should be set up even if we detected all
current cpus emulate misaligned accesses, since we want to
ensure our expectations of all cpus emulating is maintained.

Fixes: 6e5ce7f2eae3 ("riscv: Decouple emulated unaligned accesses from access speed")
Fixes: e7c9d66e313b ("RISC-V: Report vector unaligned access speed hwprobe")
Reviewed-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/kernel/unaligned_access_speed.c | 27 +++++++++++-----------
 1 file changed, 13 insertions(+), 14 deletions(-)

diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index 780f1c5f512a..c9d3237649bb 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -247,13 +247,6 @@ static void __init check_unaligned_access_speed_all_cpus(void)
 	/* Check core 0. */
 	smp_call_on_cpu(0, check_unaligned_access, bufs[0], true);
 
-	/*
-	 * Setup hotplug callbacks for any new CPUs that come online or go
-	 * offline.
-	 */
-	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
-				  riscv_online_cpu, riscv_offline_cpu);
-
 out:
 	for_each_cpu(cpu, cpu_online_mask) {
 		if (bufs[cpu])
@@ -383,13 +376,6 @@ static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __alway
 {
 	schedule_on_each_cpu(check_vector_unaligned_access);
 
-	/*
-	 * Setup hotplug callbacks for any new CPUs that come online or go
-	 * offline.
-	 */
-	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
-				  riscv_online_cpu_vec, NULL);
-
 	return 0;
 }
 #else /* CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS */
@@ -415,6 +401,19 @@ static int __init check_unaligned_access_all_cpus(void)
 			    NULL, "vec_check_unaligned_access_speed_all_cpus");
 	}
 
+	/*
+	 * Setup hotplug callbacks for any new CPUs that come online or go
+	 * offline.
+	 */
+#ifdef CONFIG_RISCV_PROBE_UNALIGNED_ACCESS
+	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
+				  riscv_online_cpu, riscv_offline_cpu);
+#endif
+#ifdef CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS
+	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
+				  riscv_online_cpu_vec, NULL);
+#endif
+
 	return 0;
 }
 
-- 
2.48.1

[PATCH v2 6/8] riscv: Fix set up of vector cpu hotplug callback

[Andrew Jones]

Whether or not we have RISCV_PROBE_VECTOR_UNALIGNED_ACCESS we need to
set up a cpu hotplug callback to check if we have vector at all,
since, when we don't have vector, we need to set
vector_misaligned_access to unsupported rather than leave it the
default of unknown.

Fixes: e7c9d66e313b ("RISC-V: Report vector unaligned access speed hwprobe")
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/kernel/unaligned_access_speed.c | 31 +++++++++++-----------
 1 file changed, 16 insertions(+), 15 deletions(-)

diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index c9d3237649bb..d9d4ca1fadc7 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -356,6 +356,20 @@ static void check_vector_unaligned_access(struct work_struct *work __always_unus
 	per_cpu(vector_misaligned_access, cpu) = speed;
 }
 
+/* Measure unaligned access speed on all CPUs present at boot in parallel. */
+static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
+{
+	schedule_on_each_cpu(check_vector_unaligned_access);
+
+	return 0;
+}
+#else /* CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS */
+static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
+{
+	return 0;
+}
+#endif
+
 static int riscv_online_cpu_vec(unsigned int cpu)
 {
 	if (!has_vector()) {
@@ -363,27 +377,16 @@ static int riscv_online_cpu_vec(unsigned int cpu)
 		return 0;
 	}
 
+#ifdef CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS
 	if (per_cpu(vector_misaligned_access, cpu) != RISCV_HWPROBE_MISALIGNED_VECTOR_UNKNOWN)
 		return 0;
 
 	check_vector_unaligned_access_emulated(NULL);
 	check_vector_unaligned_access(NULL);
-	return 0;
-}
-
-/* Measure unaligned access speed on all CPUs present at boot in parallel. */
-static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
-{
-	schedule_on_each_cpu(check_vector_unaligned_access);
+#endif
 
 	return 0;
 }
-#else /* CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS */
-static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __always_unused)
-{
-	return 0;
-}
-#endif
 
 static int __init check_unaligned_access_all_cpus(void)
 {
@@ -409,10 +412,8 @@ static int __init check_unaligned_access_all_cpus(void)
 	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
 				  riscv_online_cpu, riscv_offline_cpu);
 #endif
-#ifdef CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS
 	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
 				  riscv_online_cpu_vec, NULL);
-#endif
 
 	return 0;
 }
-- 
2.48.1

[PATCH v2 7/8] riscv: Add parameter for skipping access speed tests

[Andrew Jones]

Allow skipping scalar and vector unaligned access speed tests. This
is useful for testing alternative code paths and to skip the tests in
environments where they run too slowly. All CPUs must have the same
unaligned access speed.

The code movement is because we now need the scalar cpu hotplug
callback to always run, so we need to bring it and its supporting
functions out of CONFIG_RISCV_PROBE_UNALIGNED_ACCESS.

Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 arch/riscv/kernel/unaligned_access_speed.c | 162 ++++++++++++++-------
 1 file changed, 107 insertions(+), 55 deletions(-)

diff --git a/arch/riscv/kernel/unaligned_access_speed.c b/arch/riscv/kernel/unaligned_access_speed.c
index d9d4ca1fadc7..241d47004f37 100644
--- a/arch/riscv/kernel/unaligned_access_speed.c
+++ b/arch/riscv/kernel/unaligned_access_speed.c
@@ -24,6 +24,9 @@
 DEFINE_PER_CPU(long, misaligned_access_speed) = RISCV_HWPROBE_MISALIGNED_SCALAR_UNKNOWN;
 DEFINE_PER_CPU(long, vector_misaligned_access) = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
 
+static long unaligned_scalar_speed_param = RISCV_HWPROBE_MISALIGNED_SCALAR_UNKNOWN;
+static long unaligned_vector_speed_param = RISCV_HWPROBE_MISALIGNED_VECTOR_UNKNOWN;
+
 #ifdef CONFIG_RISCV_PROBE_UNALIGNED_ACCESS
 static cpumask_t fast_misaligned_access;
 static int check_unaligned_access(void *param)
@@ -130,6 +133,50 @@ static void __init check_unaligned_access_nonboot_cpu(void *param)
 		check_unaligned_access(pages[cpu]);
 }
 
+/* Measure unaligned access speed on all CPUs present at boot in parallel. */
+static void __init check_unaligned_access_speed_all_cpus(void)
+{
+	unsigned int cpu;
+	unsigned int cpu_count = num_possible_cpus();
+	struct page **bufs = kcalloc(cpu_count, sizeof(*bufs), GFP_KERNEL);
+
+	if (!bufs) {
+		pr_warn("Allocation failure, not measuring misaligned performance\n");
+		return;
+	}
+
+	/*
+	 * Allocate separate buffers for each CPU so there's no fighting over
+	 * cache lines.
+	 */
+	for_each_cpu(cpu, cpu_online_mask) {
+		bufs[cpu] = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
+		if (!bufs[cpu]) {
+			pr_warn("Allocation failure, not measuring misaligned performance\n");
+			goto out;
+		}
+	}
+
+	/* Check everybody except 0, who stays behind to tend jiffies. */
+	on_each_cpu(check_unaligned_access_nonboot_cpu, bufs, 1);
+
+	/* Check core 0. */
+	smp_call_on_cpu(0, check_unaligned_access, bufs[0], true);
+
+out:
+	for_each_cpu(cpu, cpu_online_mask) {
+		if (bufs[cpu])
+			__free_pages(bufs[cpu], MISALIGNED_BUFFER_ORDER);
+	}
+
+	kfree(bufs);
+}
+#else /* CONFIG_RISCV_PROBE_UNALIGNED_ACCESS */
+static void __init check_unaligned_access_speed_all_cpus(void)
+{
+}
+#endif
+
 DEFINE_STATIC_KEY_FALSE(fast_unaligned_access_speed_key);
 
 static void modify_unaligned_access_branches(cpumask_t *mask, int weight)
@@ -191,9 +238,14 @@ static int riscv_online_cpu(unsigned int cpu)
 	static struct page *buf;
 
 	/* We are already set since the last check */
-	if (per_cpu(misaligned_access_speed, cpu) != RISCV_HWPROBE_MISALIGNED_SCALAR_UNKNOWN)
+	if (per_cpu(misaligned_access_speed, cpu) != RISCV_HWPROBE_MISALIGNED_SCALAR_UNKNOWN) {
 		goto exit;
+	} else if (unaligned_scalar_speed_param != RISCV_HWPROBE_MISALIGNED_SCALAR_UNKNOWN) {
+		per_cpu(misaligned_access_speed, cpu) = unaligned_scalar_speed_param;
+		goto exit;
+	}
 
+#ifdef CONFIG_RISCV_PROBE_UNALIGNED_ACCESS
 	check_unaligned_access_emulated(NULL);
 	buf = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
 	if (!buf) {
@@ -203,6 +255,7 @@ static int riscv_online_cpu(unsigned int cpu)
 
 	check_unaligned_access(buf);
 	__free_pages(buf, MISALIGNED_BUFFER_ORDER);
+#endif
 
 exit:
 	set_unaligned_access_static_branches();
@@ -217,50 +270,6 @@ static int riscv_offline_cpu(unsigned int cpu)
 	return 0;
 }
 
-/* Measure unaligned access speed on all CPUs present at boot in parallel. */
-static void __init check_unaligned_access_speed_all_cpus(void)
-{
-	unsigned int cpu;
-	unsigned int cpu_count = num_possible_cpus();
-	struct page **bufs = kcalloc(cpu_count, sizeof(*bufs), GFP_KERNEL);
-
-	if (!bufs) {
-		pr_warn("Allocation failure, not measuring misaligned performance\n");
-		return;
-	}
-
-	/*
-	 * Allocate separate buffers for each CPU so there's no fighting over
-	 * cache lines.
-	 */
-	for_each_cpu(cpu, cpu_online_mask) {
-		bufs[cpu] = alloc_pages(GFP_KERNEL, MISALIGNED_BUFFER_ORDER);
-		if (!bufs[cpu]) {
-			pr_warn("Allocation failure, not measuring misaligned performance\n");
-			goto out;
-		}
-	}
-
-	/* Check everybody except 0, who stays behind to tend jiffies. */
-	on_each_cpu(check_unaligned_access_nonboot_cpu, bufs, 1);
-
-	/* Check core 0. */
-	smp_call_on_cpu(0, check_unaligned_access, bufs[0], true);
-
-out:
-	for_each_cpu(cpu, cpu_online_mask) {
-		if (bufs[cpu])
-			__free_pages(bufs[cpu], MISALIGNED_BUFFER_ORDER);
-	}
-
-	kfree(bufs);
-}
-#else /* CONFIG_RISCV_PROBE_UNALIGNED_ACCESS */
-static void __init check_unaligned_access_speed_all_cpus(void)
-{
-}
-#endif
-
 #ifdef CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS
 static void check_vector_unaligned_access(struct work_struct *work __always_unused)
 {
@@ -372,8 +381,8 @@ static int __init vec_check_unaligned_access_speed_all_cpus(void *unused __alway
 
 static int riscv_online_cpu_vec(unsigned int cpu)
 {
-	if (!has_vector()) {
-		per_cpu(vector_misaligned_access, cpu) = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
+	if (unaligned_vector_speed_param != RISCV_HWPROBE_MISALIGNED_VECTOR_UNKNOWN) {
+		per_cpu(vector_misaligned_access, cpu) = unaligned_vector_speed_param;
 		return 0;
 	}
 
@@ -388,30 +397,73 @@ static int riscv_online_cpu_vec(unsigned int cpu)
 	return 0;
 }
 
+static const char * const speed_str[] __initconst = { NULL, NULL, "slow", "fast", "unsupported" };
+
+static int __init set_unaligned_scalar_speed_param(char *str)
+{
+	if (!strcmp(str, speed_str[RISCV_HWPROBE_MISALIGNED_SCALAR_SLOW]))
+		unaligned_scalar_speed_param = RISCV_HWPROBE_MISALIGNED_SCALAR_SLOW;
+	else if (!strcmp(str, speed_str[RISCV_HWPROBE_MISALIGNED_SCALAR_FAST]))
+		unaligned_scalar_speed_param = RISCV_HWPROBE_MISALIGNED_SCALAR_FAST;
+	else if (!strcmp(str, speed_str[RISCV_HWPROBE_MISALIGNED_SCALAR_UNSUPPORTED]))
+		unaligned_scalar_speed_param = RISCV_HWPROBE_MISALIGNED_SCALAR_UNSUPPORTED;
+	else
+		return -EINVAL;
+
+	return 1;
+}
+__setup("unaligned_scalar_speed=", set_unaligned_scalar_speed_param);
+
+static int __init set_unaligned_vector_speed_param(char *str)
+{
+	if (!strcmp(str, speed_str[RISCV_HWPROBE_MISALIGNED_VECTOR_SLOW]))
+		unaligned_vector_speed_param = RISCV_HWPROBE_MISALIGNED_VECTOR_SLOW;
+	else if (!strcmp(str, speed_str[RISCV_HWPROBE_MISALIGNED_VECTOR_FAST]))
+		unaligned_vector_speed_param = RISCV_HWPROBE_MISALIGNED_VECTOR_FAST;
+	else if (!strcmp(str, speed_str[RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED]))
+		unaligned_vector_speed_param = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
+	else
+		return -EINVAL;
+
+	return 1;
+}
+__setup("unaligned_vector_speed=", set_unaligned_vector_speed_param);
+
 static int __init check_unaligned_access_all_cpus(void)
 {
 	int cpu;
 
-	if (!check_unaligned_access_emulated_all_cpus())
+	if (unaligned_scalar_speed_param == RISCV_HWPROBE_MISALIGNED_SCALAR_UNKNOWN &&
+	    !check_unaligned_access_emulated_all_cpus()) {
 		check_unaligned_access_speed_all_cpus();
-
-	if (!has_vector()) {
+	} else {
+		pr_info("scalar unaligned access speed set to '%s' by command line\n",
+			speed_str[unaligned_scalar_speed_param]);
 		for_each_online_cpu(cpu)
-			per_cpu(vector_misaligned_access, cpu) = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
-	} else if (!check_vector_unaligned_access_emulated_all_cpus() &&
-		   IS_ENABLED(CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS)) {
+			per_cpu(misaligned_access_speed, cpu) = unaligned_scalar_speed_param;
+	}
+
+	if (!has_vector())
+		unaligned_vector_speed_param = RISCV_HWPROBE_MISALIGNED_VECTOR_UNSUPPORTED;
+
+	if (unaligned_vector_speed_param == RISCV_HWPROBE_MISALIGNED_VECTOR_UNKNOWN &&
+	    !check_vector_unaligned_access_emulated_all_cpus() &&
+	    IS_ENABLED(CONFIG_RISCV_PROBE_VECTOR_UNALIGNED_ACCESS)) {
 		kthread_run(vec_check_unaligned_access_speed_all_cpus,
 			    NULL, "vec_check_unaligned_access_speed_all_cpus");
+	} else {
+		pr_info("vector unaligned access speed set to '%s' by command line\n",
+			speed_str[unaligned_vector_speed_param]);
+		for_each_online_cpu(cpu)
+			per_cpu(vector_misaligned_access, cpu) = unaligned_vector_speed_param;
 	}
 
 	/*
 	 * Setup hotplug callbacks for any new CPUs that come online or go
 	 * offline.
 	 */
-#ifdef CONFIG_RISCV_PROBE_UNALIGNED_ACCESS
 	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
 				  riscv_online_cpu, riscv_offline_cpu);
-#endif
 	cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN, "riscv:online",
 				  riscv_online_cpu_vec, NULL);
 
-- 
2.48.1

[PATCH v2 8/8] Documentation/kernel-parameters: Add riscv unaligned speed parameters

[Andrew Jones]

Document riscv parameters used to select scalar and vector unaligned
access speeds.

Signed-off-by: Andrew Jones <ajones@ventanamicro.com>
---
 Documentation/admin-guide/kernel-parameters.txt | 16 ++++++++++++++++
 1 file changed, 16 insertions(+)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index fb8752b42ec8..9e3c5fecfa52 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -7477,6 +7477,22 @@
 			Note that genuine overcurrent events won't be
 			reported either.
 
+	unaligned_scalar_speed=
+			[RISCV]
+			Format: {slow | fast | unsupported}
+			Allow skipping scalar unaligned access speed tests. This
+			is useful for testing alternative code paths and to skip
+			the tests in environments where they run too slowly. All
+			CPUs must have the same scalar unaligned access speed.
+
+	unaligned_vector_speed=
+			[RISCV]
+			Format: {slow | fast | unsupported}
+			Allow skipping vector unaligned access speed tests. This
+			is useful for testing alternative code paths and to skip
+			the tests in environments where they run too slowly. All
+			CPUs must have the same vector unaligned access speed.
+
 	unknown_nmi_panic
 			[X86] Cause panic on unknown NMI.
 
-- 
2.48.1