[PATCH v3 0/3] Randomize memory access of dirty_log_perf_test

[PATCH v3 0/3] Randomize memory access of dirty_log_perf_test

[Colton Lewis]

This patch adds the ability to randomize parts of dirty_log_perf_test,
specifically the order pages are accessed and whether pages are read
or written.

This version scraps the use of arrays to remove the memory management
concerns and implementes a well-known random number generator, the
Park-Miller Linear Congruential Generator, for the guest code to use
directly. From comments on v2, several others agree this is the right
approach. I came to the same conclusion and was already working on it.

Though I had this idea and ignored it previously over concerns about
introducing mysterious code, doing some research convinced me
Park-Miller is both extremely easy to understand and high enough
quality for this purpose. I also believe this approach better
preserves the integrity of the test as a couple math instructions are
definitely faster than fetching from memory.

Compiled and run on x86_64 and arm64 to test the new flags.
Based on v5.19 as tests on kvm/queue do not build

Colton Lewis (3):
  KVM: selftests: Implement random number generation for guest code.
  KVM: selftests: Randomize which pages are written vs read.
  KVM: selftests: Randomize page access order.

 .../selftests/kvm/access_tracking_perf_test.c |  2 +-
 .../selftests/kvm/dirty_log_perf_test.c       | 48 +++++++++++++------
 .../selftests/kvm/include/perf_test_util.h    |  8 +++-
 .../selftests/kvm/lib/perf_test_util.c        | 40 +++++++++++++---
 4 files changed, 74 insertions(+), 24 deletions(-)

--
2.37.2.672.g94769d06f0-goog

[PATCH v3 1/3] KVM: selftests: Implement random number generation for guest code.

[Colton Lewis]

Implement random number generation for guest code to randomize parts
of the test, making it less predictable and a more accurate reflection
of reality.

Create a -r argument to specify a random seed. If no argument is
provided, the seed defaults to the current Unix timestamp.

The random number generator chosen is the Park-Miller Linear
Congruential Generator, a fancy name for a basic and well-understood
random number generator entirely sufficient for this purpose. Each
vCPU calculates its own seed by adding its index to the seed provided.

Signed-off-by: Colton Lewis <coltonlewis@google.com>
---
 .../selftests/kvm/dirty_log_perf_test.c        | 12 ++++++++++--
 .../selftests/kvm/include/perf_test_util.h     |  2 ++
 .../testing/selftests/kvm/lib/perf_test_util.c | 18 +++++++++++++++++-
 3 files changed, 29 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
index d60a34cdfaee..2f91acd94130 100644
--- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
@@ -126,6 +126,7 @@ struct test_params {
 	bool partition_vcpu_memory_access;
 	enum vm_mem_backing_src_type backing_src;
 	int slots;
+	uint32_t random_seed;
 };
 
 static void toggle_dirty_logging(struct kvm_vm *vm, int slots, bool enable)
@@ -220,6 +221,8 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 				 p->slots, p->backing_src,
 				 p->partition_vcpu_memory_access);
 
+	pr_info("Random seed: %u\n", p->random_seed);
+	perf_test_set_random_seed(vm, p->random_seed);
 	perf_test_set_wr_fract(vm, p->wr_fract);
 
 	guest_num_pages = (nr_vcpus * guest_percpu_mem_size) >> vm_get_page_shift(vm);
@@ -337,7 +340,7 @@ static void help(char *name)
 {
 	puts("");
 	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
-	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-s mem type]"
+	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
 	       "[-x memslots]\n", name);
 	puts("");
 	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
@@ -362,6 +365,7 @@ static void help(char *name)
 	printf(" -v: specify the number of vCPUs to run.\n");
 	printf(" -o: Overlap guest memory accesses instead of partitioning\n"
 	       "     them into a separate region of memory for each vCPU.\n");
+	printf(" -r: specify the starting random seed.\n");
 	backing_src_help("-s");
 	printf(" -x: Split the memory region into this number of memslots.\n"
 	       "     (default: 1)\n");
@@ -378,6 +382,7 @@ int main(int argc, char *argv[])
 		.partition_vcpu_memory_access = true,
 		.backing_src = DEFAULT_VM_MEM_SRC,
 		.slots = 1,
+		.random_seed = time(NULL),
 	};
 	int opt;
 
@@ -388,7 +393,7 @@ int main(int argc, char *argv[])
 
 	guest_modes_append_default();
 
-	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:os:x:")) != -1) {
+	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:or:s:x:")) != -1) {
 		switch (opt) {
 		case 'g':
 			dirty_log_manual_caps = 0;
@@ -421,6 +426,9 @@ int main(int argc, char *argv[])
 		case 'o':
 			p.partition_vcpu_memory_access = false;
 			break;
+		case 'r':
+			p.random_seed = atoi(optarg);
+			break;
 		case 's':
 			p.backing_src = parse_backing_src_type(optarg);
 			break;
diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
index d822cb670f1c..f18530984b42 100644
--- a/tools/testing/selftests/kvm/include/perf_test_util.h
+++ b/tools/testing/selftests/kvm/include/perf_test_util.h
@@ -34,6 +34,7 @@ struct perf_test_args {
 	uint64_t gpa;
 	uint64_t size;
 	uint64_t guest_page_size;
+	uint32_t random_seed;
 	int wr_fract;
 
 	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
@@ -51,6 +52,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
 void perf_test_destroy_vm(struct kvm_vm *vm);
 
 void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract);
+void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
 
 void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
 void perf_test_join_vcpu_threads(int vcpus);
diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
index f989ff91f022..1292ed7d1193 100644
--- a/tools/testing/selftests/kvm/lib/perf_test_util.c
+++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
@@ -36,6 +36,13 @@ static void (*vcpu_thread_fn)(struct perf_test_vcpu_args *);
 /* Set to true once all vCPU threads are up and running. */
 static bool all_vcpu_threads_running;
 
+
+/* Park-Miller LCG using standard constants */
+static uint32_t perf_test_random(uint32_t seed)
+{
+	return (uint64_t)seed * 48271 % ((uint32_t)(1 << 31) - 1);
+}
+
 /*
  * Continuously write to the first 8 bytes of each page in the
  * specified region.
@@ -47,6 +54,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
 	uint64_t gva;
 	uint64_t pages;
 	int i;
+	uint32_t rand = pta->random_seed + vcpu_id;
 
 	/* Make sure vCPU args data structure is not corrupt. */
 	GUEST_ASSERT(vcpu_args->vcpu_id == vcpu_id);
@@ -56,6 +64,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
 
 	while (true) {
 		for (i = 0; i < pages; i++) {
+			rand = perf_test_random(rand);
 			uint64_t addr = gva + (i * pta->guest_page_size);
 
 			if (i % pta->wr_fract == 0)
@@ -115,8 +124,9 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
 
 	pr_info("Testing guest mode: %s\n", vm_guest_mode_string(mode));
 
-	/* By default vCPUs will write to memory. */
+	/* Set perf_test_args defaults. */
 	pta->wr_fract = 1;
+	pta->random_seed = time(NULL);
 
 	/*
 	 * Snapshot the non-huge page size.  This is used by the guest code to
@@ -224,6 +234,12 @@ void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract)
 	sync_global_to_guest(vm, perf_test_args);
 }
 
+void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed)
+{
+	perf_test_args.random_seed = random_seed;
+	sync_global_to_guest(vm, perf_test_args);
+}
+
 uint64_t __weak perf_test_nested_pages(int nr_vcpus)
 {
 	return 0;
-- 
2.37.2.789.g6183377224-goog

Re: [PATCH v3 1/3] KVM: selftests: Implement random number generation for guest code.

[Ricardo Koller]

On Thu, Sep 01, 2022 at 07:52:35PM +0000, Colton Lewis wrote:
> Implement random number generation for guest code to randomize parts
> of the test, making it less predictable and a more accurate reflection
> of reality.
> 
> Create a -r argument to specify a random seed. If no argument is
> provided, the seed defaults to the current Unix timestamp.
> 
> The random number generator chosen is the Park-Miller Linear
> Congruential Generator, a fancy name for a basic and well-understood
> random number generator entirely sufficient for this purpose. Each
> vCPU calculates its own seed by adding its index to the seed provided.
> 
> Signed-off-by: Colton Lewis <coltonlewis@google.com>
> ---
>  .../selftests/kvm/dirty_log_perf_test.c        | 12 ++++++++++--
>  .../selftests/kvm/include/perf_test_util.h     |  2 ++
>  .../testing/selftests/kvm/lib/perf_test_util.c | 18 +++++++++++++++++-
>  3 files changed, 29 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> index d60a34cdfaee..2f91acd94130 100644
> --- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
> +++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> @@ -126,6 +126,7 @@ struct test_params {
>  	bool partition_vcpu_memory_access;
>  	enum vm_mem_backing_src_type backing_src;
>  	int slots;
> +	uint32_t random_seed;
>  };
>  
>  static void toggle_dirty_logging(struct kvm_vm *vm, int slots, bool enable)
> @@ -220,6 +221,8 @@ static void run_test(enum vm_guest_mode mode, void *arg)
>  				 p->slots, p->backing_src,
>  				 p->partition_vcpu_memory_access);
>  
> +	pr_info("Random seed: %u\n", p->random_seed);
> +	perf_test_set_random_seed(vm, p->random_seed);
>  	perf_test_set_wr_fract(vm, p->wr_fract);
>  
>  	guest_num_pages = (nr_vcpus * guest_percpu_mem_size) >> vm_get_page_shift(vm);
> @@ -337,7 +340,7 @@ static void help(char *name)
>  {
>  	puts("");
>  	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
> -	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-s mem type]"
> +	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
>  	       "[-x memslots]\n", name);
>  	puts("");
>  	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
> @@ -362,6 +365,7 @@ static void help(char *name)
>  	printf(" -v: specify the number of vCPUs to run.\n");
>  	printf(" -o: Overlap guest memory accesses instead of partitioning\n"
>  	       "     them into a separate region of memory for each vCPU.\n");
> +	printf(" -r: specify the starting random seed.\n");
>  	backing_src_help("-s");
>  	printf(" -x: Split the memory region into this number of memslots.\n"
>  	       "     (default: 1)\n");
> @@ -378,6 +382,7 @@ int main(int argc, char *argv[])
>  		.partition_vcpu_memory_access = true,
>  		.backing_src = DEFAULT_VM_MEM_SRC,
>  		.slots = 1,
> +		.random_seed = time(NULL),
>  	};
>  	int opt;
>  
> @@ -388,7 +393,7 @@ int main(int argc, char *argv[])
>  
>  	guest_modes_append_default();
>  
> -	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:os:x:")) != -1) {
> +	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:or:s:x:")) != -1) {
>  		switch (opt) {
>  		case 'g':
>  			dirty_log_manual_caps = 0;
> @@ -421,6 +426,9 @@ int main(int argc, char *argv[])
>  		case 'o':
>  			p.partition_vcpu_memory_access = false;
>  			break;
> +		case 'r':
> +			p.random_seed = atoi(optarg);
> +			break;
>  		case 's':
>  			p.backing_src = parse_backing_src_type(optarg);
>  			break;
> diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
> index d822cb670f1c..f18530984b42 100644
> --- a/tools/testing/selftests/kvm/include/perf_test_util.h
> +++ b/tools/testing/selftests/kvm/include/perf_test_util.h
> @@ -34,6 +34,7 @@ struct perf_test_args {
>  	uint64_t gpa;
>  	uint64_t size;
>  	uint64_t guest_page_size;
> +	uint32_t random_seed;
>  	int wr_fract;
>  
>  	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
> @@ -51,6 +52,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
>  void perf_test_destroy_vm(struct kvm_vm *vm);
>  
>  void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract);
> +void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
>  
>  void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
>  void perf_test_join_vcpu_threads(int vcpus);
> diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
> index f989ff91f022..1292ed7d1193 100644
> --- a/tools/testing/selftests/kvm/lib/perf_test_util.c
> +++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
> @@ -36,6 +36,13 @@ static void (*vcpu_thread_fn)(struct perf_test_vcpu_args *);
>  /* Set to true once all vCPU threads are up and running. */
>  static bool all_vcpu_threads_running;
>  
> +
> +/* Park-Miller LCG using standard constants */
> +static uint32_t perf_test_random(uint32_t seed)
> +{
> +	return (uint64_t)seed * 48271 % ((uint32_t)(1 << 31) - 1);
> +}

Nit: I would prefer moving this to include/kvm_test_util.h, maybe
something like: get_next_random(seed). There could be other users of
this.

> +
>  /*
>   * Continuously write to the first 8 bytes of each page in the
>   * specified region.
> @@ -47,6 +54,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  	uint64_t gva;
>  	uint64_t pages;
>  	int i;
> +	uint32_t rand = pta->random_seed + vcpu_id;
>  
>  	/* Make sure vCPU args data structure is not corrupt. */
>  	GUEST_ASSERT(vcpu_args->vcpu_id == vcpu_id);
> @@ -56,6 +64,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  
>  	while (true) {
>  		for (i = 0; i < pages; i++) {
> +			rand = perf_test_random(rand);
>  			uint64_t addr = gva + (i * pta->guest_page_size);
>  
>  			if (i % pta->wr_fract == 0)
> @@ -115,8 +124,9 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
>  
>  	pr_info("Testing guest mode: %s\n", vm_guest_mode_string(mode));
>  
> -	/* By default vCPUs will write to memory. */
> +	/* Set perf_test_args defaults. */
>  	pta->wr_fract = 1;
> +	pta->random_seed = time(NULL);
>  
>  	/*
>  	 * Snapshot the non-huge page size.  This is used by the guest code to
> @@ -224,6 +234,12 @@ void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract)
>  	sync_global_to_guest(vm, perf_test_args);
>  }
>  
> +void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed)
> +{
> +	perf_test_args.random_seed = random_seed;
> +	sync_global_to_guest(vm, perf_test_args);
> +}
> +
>  uint64_t __weak perf_test_nested_pages(int nr_vcpus)
>  {
>  	return 0;
> -- 
> 2.37.2.789.g6183377224-goog
> 

Re: [PATCH v3 1/3] KVM: selftests: Implement random number generation for guest code.

[Colton Lewis]

Ricardo Koller <ricarkol@google.com> writes:

> On Thu, Sep 01, 2022 at 07:52:35PM +0000, Colton Lewis wrote:
>> +
>> +/* Park-Miller LCG using standard constants */
>> +static uint32_t perf_test_random(uint32_t seed)
>> +{
>> +	return (uint64_t)seed * 48271 % ((uint32_t)(1 << 31) - 1);
>> +}

> Nit: I would prefer moving this to include/kvm_test_util.h, maybe
> something like: get_next_random(seed). There could be other users of
> this.


Agree this might be useful elsewhere. Whoever needs it can move it to
the common header then.

Re: [PATCH v3 1/3] KVM: selftests: Implement random number generation for guest code.

[Ricardo Koller]

On Thu, Sep 01, 2022 at 07:52:35PM +0000, Colton Lewis wrote:
> Implement random number generation for guest code to randomize parts
> of the test, making it less predictable and a more accurate reflection
> of reality.
> 
> Create a -r argument to specify a random seed. If no argument is
> provided, the seed defaults to the current Unix timestamp.
> 
> The random number generator chosen is the Park-Miller Linear
> Congruential Generator, a fancy name for a basic and well-understood
> random number generator entirely sufficient for this purpose. Each
> vCPU calculates its own seed by adding its index to the seed provided.
> 

Reviewed-by: Ricardo Koller <ricarkol@google.com>

> Signed-off-by: Colton Lewis <coltonlewis@google.com>
> ---
>  .../selftests/kvm/dirty_log_perf_test.c        | 12 ++++++++++--
>  .../selftests/kvm/include/perf_test_util.h     |  2 ++
>  .../testing/selftests/kvm/lib/perf_test_util.c | 18 +++++++++++++++++-
>  3 files changed, 29 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> index d60a34cdfaee..2f91acd94130 100644
> --- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
> +++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> @@ -126,6 +126,7 @@ struct test_params {
>  	bool partition_vcpu_memory_access;
>  	enum vm_mem_backing_src_type backing_src;
>  	int slots;
> +	uint32_t random_seed;
>  };
>  
>  static void toggle_dirty_logging(struct kvm_vm *vm, int slots, bool enable)
> @@ -220,6 +221,8 @@ static void run_test(enum vm_guest_mode mode, void *arg)
>  				 p->slots, p->backing_src,
>  				 p->partition_vcpu_memory_access);
>  
> +	pr_info("Random seed: %u\n", p->random_seed);
> +	perf_test_set_random_seed(vm, p->random_seed);
>  	perf_test_set_wr_fract(vm, p->wr_fract);
>  
>  	guest_num_pages = (nr_vcpus * guest_percpu_mem_size) >> vm_get_page_shift(vm);
> @@ -337,7 +340,7 @@ static void help(char *name)
>  {
>  	puts("");
>  	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
> -	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-s mem type]"
> +	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
>  	       "[-x memslots]\n", name);
>  	puts("");
>  	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
> @@ -362,6 +365,7 @@ static void help(char *name)
>  	printf(" -v: specify the number of vCPUs to run.\n");
>  	printf(" -o: Overlap guest memory accesses instead of partitioning\n"
>  	       "     them into a separate region of memory for each vCPU.\n");
> +	printf(" -r: specify the starting random seed.\n");
>  	backing_src_help("-s");
>  	printf(" -x: Split the memory region into this number of memslots.\n"
>  	       "     (default: 1)\n");
> @@ -378,6 +382,7 @@ int main(int argc, char *argv[])
>  		.partition_vcpu_memory_access = true,
>  		.backing_src = DEFAULT_VM_MEM_SRC,
>  		.slots = 1,
> +		.random_seed = time(NULL),
>  	};
>  	int opt;
>  
> @@ -388,7 +393,7 @@ int main(int argc, char *argv[])
>  
>  	guest_modes_append_default();
>  
> -	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:os:x:")) != -1) {
> +	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:or:s:x:")) != -1) {
>  		switch (opt) {
>  		case 'g':
>  			dirty_log_manual_caps = 0;
> @@ -421,6 +426,9 @@ int main(int argc, char *argv[])
>  		case 'o':
>  			p.partition_vcpu_memory_access = false;
>  			break;
> +		case 'r':
> +			p.random_seed = atoi(optarg);
> +			break;
>  		case 's':
>  			p.backing_src = parse_backing_src_type(optarg);
>  			break;
> diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
> index d822cb670f1c..f18530984b42 100644
> --- a/tools/testing/selftests/kvm/include/perf_test_util.h
> +++ b/tools/testing/selftests/kvm/include/perf_test_util.h
> @@ -34,6 +34,7 @@ struct perf_test_args {
>  	uint64_t gpa;
>  	uint64_t size;
>  	uint64_t guest_page_size;
> +	uint32_t random_seed;
>  	int wr_fract;
>  
>  	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
> @@ -51,6 +52,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
>  void perf_test_destroy_vm(struct kvm_vm *vm);
>  
>  void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract);
> +void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
>  
>  void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
>  void perf_test_join_vcpu_threads(int vcpus);
> diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
> index f989ff91f022..1292ed7d1193 100644
> --- a/tools/testing/selftests/kvm/lib/perf_test_util.c
> +++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
> @@ -36,6 +36,13 @@ static void (*vcpu_thread_fn)(struct perf_test_vcpu_args *);
>  /* Set to true once all vCPU threads are up and running. */
>  static bool all_vcpu_threads_running;
>  
> +
> +/* Park-Miller LCG using standard constants */
> +static uint32_t perf_test_random(uint32_t seed)
> +{
> +	return (uint64_t)seed * 48271 % ((uint32_t)(1 << 31) - 1);
> +}
> +
>  /*
>   * Continuously write to the first 8 bytes of each page in the
>   * specified region.
> @@ -47,6 +54,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  	uint64_t gva;
>  	uint64_t pages;
>  	int i;
> +	uint32_t rand = pta->random_seed + vcpu_id;
>  
>  	/* Make sure vCPU args data structure is not corrupt. */
>  	GUEST_ASSERT(vcpu_args->vcpu_id == vcpu_id);
> @@ -56,6 +64,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  
>  	while (true) {
>  		for (i = 0; i < pages; i++) {
> +			rand = perf_test_random(rand);
>  			uint64_t addr = gva + (i * pta->guest_page_size);
>  
>  			if (i % pta->wr_fract == 0)
> @@ -115,8 +124,9 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
>  
>  	pr_info("Testing guest mode: %s\n", vm_guest_mode_string(mode));
>  
> -	/* By default vCPUs will write to memory. */
> +	/* Set perf_test_args defaults. */
>  	pta->wr_fract = 1;
> +	pta->random_seed = time(NULL);
>  
>  	/*
>  	 * Snapshot the non-huge page size.  This is used by the guest code to
> @@ -224,6 +234,12 @@ void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract)
>  	sync_global_to_guest(vm, perf_test_args);
>  }
>  
> +void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed)
> +{
> +	perf_test_args.random_seed = random_seed;
> +	sync_global_to_guest(vm, perf_test_args);
> +}
> +
>  uint64_t __weak perf_test_nested_pages(int nr_vcpus)
>  {
>  	return 0;
> -- 
> 2.37.2.789.g6183377224-goog
> 

[PATCH v3 2/3] KVM: selftests: Randomize which pages are written vs read.

[Colton Lewis]

Randomize which pages are written vs read using the random number
generator.

Change the variable wr_fract and associated function calls to
write_percent that now operates as a percentage from 0 to 100 where X
means each page has an X% chance of being written. Change the -f
argument to -w to reflect the new variable semantics. Keep the same
default of 100 percent writes.

Signed-off-by: Colton Lewis <coltonlewis@google.com>
---
 .../selftests/kvm/access_tracking_perf_test.c |  2 +-
 .../selftests/kvm/dirty_log_perf_test.c       | 28 ++++++++++---------
 .../selftests/kvm/include/perf_test_util.h    |  4 +--
 .../selftests/kvm/lib/perf_test_util.c        | 10 +++----
 4 files changed, 23 insertions(+), 21 deletions(-)

diff --git a/tools/testing/selftests/kvm/access_tracking_perf_test.c b/tools/testing/selftests/kvm/access_tracking_perf_test.c
index d8909032317a..d86046ef3a0b 100644
--- a/tools/testing/selftests/kvm/access_tracking_perf_test.c
+++ b/tools/testing/selftests/kvm/access_tracking_perf_test.c
@@ -274,7 +274,7 @@ static void run_iteration(struct kvm_vm *vm, int vcpus, const char *description)
 static void access_memory(struct kvm_vm *vm, int vcpus, enum access_type access,
 			  const char *description)
 {
-	perf_test_set_wr_fract(vm, (access == ACCESS_READ) ? INT_MAX : 1);
+	perf_test_set_write_percent(vm, (access == ACCESS_READ) ? 0 : 100);
 	iteration_work = ITERATION_ACCESS_MEMORY;
 	run_iteration(vm, vcpus, description);
 }
diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
index 2f91acd94130..c9441f8354be 100644
--- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
@@ -122,10 +122,10 @@ static void vcpu_worker(struct perf_test_vcpu_args *vcpu_args)
 struct test_params {
 	unsigned long iterations;
 	uint64_t phys_offset;
-	int wr_fract;
 	bool partition_vcpu_memory_access;
 	enum vm_mem_backing_src_type backing_src;
 	int slots;
+	uint32_t write_percent;
 	uint32_t random_seed;
 };
 
@@ -223,7 +223,7 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 
 	pr_info("Random seed: %u\n", p->random_seed);
 	perf_test_set_random_seed(vm, p->random_seed);
-	perf_test_set_wr_fract(vm, p->wr_fract);
+	perf_test_set_write_percent(vm, p->write_percent);
 
 	guest_num_pages = (nr_vcpus * guest_percpu_mem_size) >> vm_get_page_shift(vm);
 	guest_num_pages = vm_adjust_num_guest_pages(mode, guest_num_pages);
@@ -341,7 +341,7 @@ static void help(char *name)
 	puts("");
 	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
 	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
-	       "[-x memslots]\n", name);
+	       "[-x memslots] [-w percentage]\n", name);
 	puts("");
 	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
 	       TEST_HOST_LOOP_N);
@@ -358,10 +358,6 @@ static void help(char *name)
 	printf(" -b: specify the size of the memory region which should be\n"
 	       "     dirtied by each vCPU. e.g. 10M or 3G.\n"
 	       "     (default: 1G)\n");
-	printf(" -f: specify the fraction of pages which should be written to\n"
-	       "     as opposed to simply read, in the form\n"
-	       "     1/<fraction of pages to write>.\n"
-	       "     (default: 1 i.e. all pages are written to.)\n");
 	printf(" -v: specify the number of vCPUs to run.\n");
 	printf(" -o: Overlap guest memory accesses instead of partitioning\n"
 	       "     them into a separate region of memory for each vCPU.\n");
@@ -369,6 +365,11 @@ static void help(char *name)
 	backing_src_help("-s");
 	printf(" -x: Split the memory region into this number of memslots.\n"
 	       "     (default: 1)\n");
+	printf(" -w: specify the percentage of pages which should be written to\n"
+	       "     as an integer from 0-100 inclusive. This is probabalistic,\n"
+	       "     so -w X means each page has an X%% chance of writing\n"
+	       "     and a (100-X)%% chance of reading.\n"
+	       "     (default: 100 i.e. all pages are written to.)\n");
 	puts("");
 	exit(0);
 }
@@ -378,10 +379,10 @@ int main(int argc, char *argv[])
 	int max_vcpus = kvm_check_cap(KVM_CAP_MAX_VCPUS);
 	struct test_params p = {
 		.iterations = TEST_HOST_LOOP_N,
-		.wr_fract = 1,
 		.partition_vcpu_memory_access = true,
 		.backing_src = DEFAULT_VM_MEM_SRC,
 		.slots = 1,
+		.write_percent = 100,
 		.random_seed = time(NULL),
 	};
 	int opt;
@@ -393,7 +394,7 @@ int main(int argc, char *argv[])
 
 	guest_modes_append_default();
 
-	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:or:s:x:")) != -1) {
+	while ((opt = getopt(argc, argv, "ghi:p:m:nb:v:or:s:x:w:")) != -1) {
 		switch (opt) {
 		case 'g':
 			dirty_log_manual_caps = 0;
@@ -413,10 +414,11 @@ int main(int argc, char *argv[])
 		case 'b':
 			guest_percpu_mem_size = parse_size(optarg);
 			break;
-		case 'f':
-			p.wr_fract = atoi(optarg);
-			TEST_ASSERT(p.wr_fract >= 1,
-				    "Write fraction cannot be less than one");
+		case 'w':
+			perf_test_args.write_percent = atoi(optarg);
+			TEST_ASSERT(perf_test_args.write_percent >= 0
+				    && perf_test_args.write_percent <= 100,
+				    "Write percentage must be between 0 and 100");
 			break;
 		case 'v':
 			nr_vcpus = atoi(optarg);
diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
index f18530984b42..52c450eb6b9b 100644
--- a/tools/testing/selftests/kvm/include/perf_test_util.h
+++ b/tools/testing/selftests/kvm/include/perf_test_util.h
@@ -35,7 +35,7 @@ struct perf_test_args {
 	uint64_t size;
 	uint64_t guest_page_size;
 	uint32_t random_seed;
-	int wr_fract;
+	int write_percent;
 
 	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
 	bool nested;
@@ -51,7 +51,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
 				   bool partition_vcpu_memory_access);
 void perf_test_destroy_vm(struct kvm_vm *vm);
 
-void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract);
+void perf_test_set_write_percent(struct kvm_vm *vm, uint32_t write_percent);
 void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
 
 void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
index 1292ed7d1193..be6176faaf8e 100644
--- a/tools/testing/selftests/kvm/lib/perf_test_util.c
+++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
@@ -64,10 +64,10 @@ void perf_test_guest_code(uint32_t vcpu_id)
 
 	while (true) {
 		for (i = 0; i < pages; i++) {
-			rand = perf_test_random(rand);
 			uint64_t addr = gva + (i * pta->guest_page_size);
+			rand = perf_test_random(rand);
 
-			if (i % pta->wr_fract == 0)
+			if (rand % 100 < pta->write_percent)
 				*(uint64_t *)addr = 0x0123456789ABCDEF;
 			else
 				READ_ONCE(*(uint64_t *)addr);
@@ -125,7 +125,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
 	pr_info("Testing guest mode: %s\n", vm_guest_mode_string(mode));
 
 	/* Set perf_test_args defaults. */
-	pta->wr_fract = 1;
+	pta->write_percent = 100;
 	pta->random_seed = time(NULL);
 
 	/*
@@ -228,9 +228,9 @@ void perf_test_destroy_vm(struct kvm_vm *vm)
 	kvm_vm_free(vm);
 }
 
-void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract)
+void perf_test_set_write_percent(struct kvm_vm *vm, uint32_t write_percent)
 {
-	perf_test_args.wr_fract = wr_fract;
+	perf_test_args.write_percent = write_percent;
 	sync_global_to_guest(vm, perf_test_args);
 }
 
-- 
2.37.2.789.g6183377224-goog

Re: [PATCH v3 2/3] KVM: selftests: Randomize which pages are written vs read.

[Ricardo Koller]

On Thu, Sep 01, 2022 at 07:52:36PM +0000, Colton Lewis wrote:
> Randomize which pages are written vs read using the random number
> generator.
nit:       ^ I haven't seen this style before (the period at the end)

> 
> Change the variable wr_fract and associated function calls to
> write_percent that now operates as a percentage from 0 to 100 where X
> means each page has an X% chance of being written. Change the -f
> argument to -w to reflect the new variable semantics. Keep the same
> default of 100 percent writes.
> 
> Signed-off-by: Colton Lewis <coltonlewis@google.com>
> ---
>  .../selftests/kvm/access_tracking_perf_test.c |  2 +-
>  .../selftests/kvm/dirty_log_perf_test.c       | 28 ++++++++++---------
>  .../selftests/kvm/include/perf_test_util.h    |  4 +--
>  .../selftests/kvm/lib/perf_test_util.c        | 10 +++----
>  4 files changed, 23 insertions(+), 21 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/access_tracking_perf_test.c b/tools/testing/selftests/kvm/access_tracking_perf_test.c
> index d8909032317a..d86046ef3a0b 100644
> --- a/tools/testing/selftests/kvm/access_tracking_perf_test.c
> +++ b/tools/testing/selftests/kvm/access_tracking_perf_test.c
> @@ -274,7 +274,7 @@ static void run_iteration(struct kvm_vm *vm, int vcpus, const char *description)
>  static void access_memory(struct kvm_vm *vm, int vcpus, enum access_type access,
>  			  const char *description)
>  {
> -	perf_test_set_wr_fract(vm, (access == ACCESS_READ) ? INT_MAX : 1);
> +	perf_test_set_write_percent(vm, (access == ACCESS_READ) ? 0 : 100);
>  	iteration_work = ITERATION_ACCESS_MEMORY;
>  	run_iteration(vm, vcpus, description);
>  }
> diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> index 2f91acd94130..c9441f8354be 100644
> --- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
> +++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> @@ -122,10 +122,10 @@ static void vcpu_worker(struct perf_test_vcpu_args *vcpu_args)
>  struct test_params {
>  	unsigned long iterations;
>  	uint64_t phys_offset;
> -	int wr_fract;
>  	bool partition_vcpu_memory_access;
>  	enum vm_mem_backing_src_type backing_src;
>  	int slots;
> +	uint32_t write_percent;

nit: make it an int to match perf_test_args.write_percent

>  	uint32_t random_seed;
>  };
>  
> @@ -223,7 +223,7 @@ static void run_test(enum vm_guest_mode mode, void *arg)
>  
>  	pr_info("Random seed: %u\n", p->random_seed);
>  	perf_test_set_random_seed(vm, p->random_seed);
> -	perf_test_set_wr_fract(vm, p->wr_fract);
> +	perf_test_set_write_percent(vm, p->write_percent);
>  
>  	guest_num_pages = (nr_vcpus * guest_percpu_mem_size) >> vm_get_page_shift(vm);
>  	guest_num_pages = vm_adjust_num_guest_pages(mode, guest_num_pages);
> @@ -341,7 +341,7 @@ static void help(char *name)
>  	puts("");
>  	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
>  	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
> -	       "[-x memslots]\n", name);
> +	       "[-x memslots] [-w percentage]\n", name);
>  	puts("");
>  	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
>  	       TEST_HOST_LOOP_N);
> @@ -358,10 +358,6 @@ static void help(char *name)
>  	printf(" -b: specify the size of the memory region which should be\n"
>  	       "     dirtied by each vCPU. e.g. 10M or 3G.\n"
>  	       "     (default: 1G)\n");
> -	printf(" -f: specify the fraction of pages which should be written to\n"
> -	       "     as opposed to simply read, in the form\n"
> -	       "     1/<fraction of pages to write>.\n"
> -	       "     (default: 1 i.e. all pages are written to.)\n");
>  	printf(" -v: specify the number of vCPUs to run.\n");
>  	printf(" -o: Overlap guest memory accesses instead of partitioning\n"
>  	       "     them into a separate region of memory for each vCPU.\n");
> @@ -369,6 +365,11 @@ static void help(char *name)
>  	backing_src_help("-s");
>  	printf(" -x: Split the memory region into this number of memslots.\n"
>  	       "     (default: 1)\n");
> +	printf(" -w: specify the percentage of pages which should be written to\n"
> +	       "     as an integer from 0-100 inclusive. This is probabalistic,\n"
> +	       "     so -w X means each page has an X%% chance of writing\n"
> +	       "     and a (100-X)%% chance of reading.\n"
> +	       "     (default: 100 i.e. all pages are written to.)\n");
>  	puts("");
>  	exit(0);
>  }
> @@ -378,10 +379,10 @@ int main(int argc, char *argv[])
>  	int max_vcpus = kvm_check_cap(KVM_CAP_MAX_VCPUS);
>  	struct test_params p = {
>  		.iterations = TEST_HOST_LOOP_N,
> -		.wr_fract = 1,
>  		.partition_vcpu_memory_access = true,
>  		.backing_src = DEFAULT_VM_MEM_SRC,
>  		.slots = 1,
> +		.write_percent = 100,
>  		.random_seed = time(NULL),
>  	};
>  	int opt;
> @@ -393,7 +394,7 @@ int main(int argc, char *argv[])
>  
>  	guest_modes_append_default();
>  
> -	while ((opt = getopt(argc, argv, "ghi:p:m:nb:f:v:or:s:x:")) != -1) {
> +	while ((opt = getopt(argc, argv, "ghi:p:m:nb:v:or:s:x:w:")) != -1) {
>  		switch (opt) {
>  		case 'g':
>  			dirty_log_manual_caps = 0;
> @@ -413,10 +414,11 @@ int main(int argc, char *argv[])
>  		case 'b':
>  			guest_percpu_mem_size = parse_size(optarg);
>  			break;
> -		case 'f':
> -			p.wr_fract = atoi(optarg);
> -			TEST_ASSERT(p.wr_fract >= 1,
> -				    "Write fraction cannot be less than one");
> +		case 'w':
> +			perf_test_args.write_percent = atoi(optarg);

I'm a bit confused, where is p.write_percent being set? I later see

	perf_test_set_write_percent(vm, p->write_percent);

that rewrites perf_test_args.write_percent with whatever was in
p->write_percent.

> +			TEST_ASSERT(perf_test_args.write_percent >= 0
> +				    && perf_test_args.write_percent <= 100,
> +				    "Write percentage must be between 0 and 100");
>  			break;
>  		case 'v':
>  			nr_vcpus = atoi(optarg);
> diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
> index f18530984b42..52c450eb6b9b 100644
> --- a/tools/testing/selftests/kvm/include/perf_test_util.h
> +++ b/tools/testing/selftests/kvm/include/perf_test_util.h
> @@ -35,7 +35,7 @@ struct perf_test_args {
>  	uint64_t size;
>  	uint64_t guest_page_size;
>  	uint32_t random_seed;
> -	int wr_fract;
> +	int write_percent;
>  
>  	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
>  	bool nested;
> @@ -51,7 +51,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
>  				   bool partition_vcpu_memory_access);
>  void perf_test_destroy_vm(struct kvm_vm *vm);
>  
> -void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract);
> +void perf_test_set_write_percent(struct kvm_vm *vm, uint32_t write_percent);
>  void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
>  
>  void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
> diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
> index 1292ed7d1193..be6176faaf8e 100644
> --- a/tools/testing/selftests/kvm/lib/perf_test_util.c
> +++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
> @@ -64,10 +64,10 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  
>  	while (true) {
>  		for (i = 0; i < pages; i++) {
> -			rand = perf_test_random(rand);
>  			uint64_t addr = gva + (i * pta->guest_page_size);
> +			rand = perf_test_random(rand);
>  
> -			if (i % pta->wr_fract == 0)
> +			if (rand % 100 < pta->write_percent)
>  				*(uint64_t *)addr = 0x0123456789ABCDEF;
>  			else
>  				READ_ONCE(*(uint64_t *)addr);
> @@ -125,7 +125,7 @@ struct kvm_vm *perf_test_create_vm(enum vm_guest_mode mode, int vcpus,
>  	pr_info("Testing guest mode: %s\n", vm_guest_mode_string(mode));
>  
>  	/* Set perf_test_args defaults. */
> -	pta->wr_fract = 1;
> +	pta->write_percent = 100;
>  	pta->random_seed = time(NULL);
>  
>  	/*
> @@ -228,9 +228,9 @@ void perf_test_destroy_vm(struct kvm_vm *vm)
>  	kvm_vm_free(vm);
>  }
>  
> -void perf_test_set_wr_fract(struct kvm_vm *vm, int wr_fract)
> +void perf_test_set_write_percent(struct kvm_vm *vm, uint32_t write_percent)
>  {
> -	perf_test_args.wr_fract = wr_fract;
> +	perf_test_args.write_percent = write_percent;
>  	sync_global_to_guest(vm, perf_test_args);
>  }
>  
> -- 
> 2.37.2.789.g6183377224-goog
> 

Re: [PATCH v3 2/3] KVM: selftests: Randomize which pages are written vs read.

[Colton Lewis]

Ricardo Koller <ricarkol@google.com> writes:

> On Thu, Sep 01, 2022 at 07:52:36PM +0000, Colton Lewis wrote:
>> Randomize which pages are written vs read using the random number
>> generator.
> nit:       ^ I haven't seen this style before (the period at the end)


I will change style in future patches, but checkpatch doesn't complain
and there are many examples in git history.

$ git log --oneline | grep '\.$' | wc -l
   43812

>> diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c  
>> b/tools/testing/selftests/kvm/dirty_log_perf_test.c
>> index 2f91acd94130..c9441f8354be 100644
>> --- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
>> +++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
>> @@ -122,10 +122,10 @@ static void vcpu_worker(struct perf_test_vcpu_args  
>> *vcpu_args)
>>   struct test_params {
>>   	unsigned long iterations;
>>   	uint64_t phys_offset;
>> -	int wr_fract;
>>   	bool partition_vcpu_memory_access;
>>   	enum vm_mem_backing_src_type backing_src;
>>   	int slots;
>> +	uint32_t write_percent;

> nit: make it an int to match perf_test_args.write_percent


That inconsistency bothers me. I'll make perf_test_args.write_percent a
uint32 instead since that's what most of the code uses. Will take care
of your other nits while I'm at it.

>> @@ -413,10 +414,11 @@ int main(int argc, char *argv[])
>>   		case 'b':
>>   			guest_percpu_mem_size = parse_size(optarg);
>>   			break;
>> -		case 'f':
>> -			p.wr_fract = atoi(optarg);
>> -			TEST_ASSERT(p.wr_fract >= 1,
>> -				    "Write fraction cannot be less than one");
>> +		case 'w':
>> +			perf_test_args.write_percent = atoi(optarg);

> I'm a bit confused, where is p.write_percent being set? I later see

> 	perf_test_set_write_percent(vm, p->write_percent);

> that rewrites perf_test_args.write_percent with whatever was in
> p->write_percent.


Hmm. Thought I fixed that mistake before. Should be p.write_percent
there.

[PATCH v3 3/3] KVM: selftests: Randomize page access order.

[Colton Lewis]

Create the ability to randomize page access order with the -a
argument, including the possibility that the same pages may be hit
multiple times during an iteration or not at all.

Signed-off-by: Colton Lewis <coltonlewis@google.com>
---
 .../testing/selftests/kvm/dirty_log_perf_test.c  | 12 ++++++++++--
 .../selftests/kvm/include/perf_test_util.h       |  2 ++
 tools/testing/selftests/kvm/lib/perf_test_util.c | 16 +++++++++++++++-
 3 files changed, 27 insertions(+), 3 deletions(-)

diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
index c9441f8354be..631b3883ed12 100644
--- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
+++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
@@ -127,6 +127,7 @@ struct test_params {
 	int slots;
 	uint32_t write_percent;
 	uint32_t random_seed;
+	bool random_access;
 };
 
 static void toggle_dirty_logging(struct kvm_vm *vm, int slots, bool enable)
@@ -269,6 +270,9 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 	pr_info("Enabling dirty logging time: %ld.%.9lds\n\n",
 		ts_diff.tv_sec, ts_diff.tv_nsec);
 
+	/* Set random access here, after population phase. */
+	perf_test_set_random_access(vm, p->random_access);
+
 	while (iteration < p->iterations) {
 		/*
 		 * Incrementing the iteration number will start the vCPUs
@@ -339,10 +343,11 @@ static void run_test(enum vm_guest_mode mode, void *arg)
 static void help(char *name)
 {
 	puts("");
-	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
+	printf("usage: %s [-h] [-a] [-i iterations] [-p offset] [-g] "
 	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
 	       "[-x memslots] [-w percentage]\n", name);
 	puts("");
+	printf(" -a: access memory randomly rather than in order.\n");
 	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
 	       TEST_HOST_LOOP_N);
 	printf(" -g: Do not enable KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2. This\n"
@@ -394,8 +399,11 @@ int main(int argc, char *argv[])
 
 	guest_modes_append_default();
 
-	while ((opt = getopt(argc, argv, "ghi:p:m:nb:v:or:s:x:w:")) != -1) {
+	while ((opt = getopt(argc, argv, "aghi:p:m:nb:v:or:s:x:w:")) != -1) {
 		switch (opt) {
+		case 'a':
+			p.random_access = true;
+			break;
 		case 'g':
 			dirty_log_manual_caps = 0;
 			break;
diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
index 52c450eb6b9b..380c31375b60 100644
--- a/tools/testing/selftests/kvm/include/perf_test_util.h
+++ b/tools/testing/selftests/kvm/include/perf_test_util.h
@@ -39,6 +39,7 @@ struct perf_test_args {
 
 	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
 	bool nested;
+	bool random_access;
 
 	struct perf_test_vcpu_args vcpu_args[KVM_MAX_VCPUS];
 };
@@ -53,6 +54,7 @@ void perf_test_destroy_vm(struct kvm_vm *vm);
 
 void perf_test_set_write_percent(struct kvm_vm *vm, uint32_t write_percent);
 void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
+void perf_test_set_random_access(struct kvm_vm *vm, bool random_access);
 
 void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
 void perf_test_join_vcpu_threads(int vcpus);
diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
index be6176faaf8e..c6123b75d5e3 100644
--- a/tools/testing/selftests/kvm/lib/perf_test_util.c
+++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
@@ -53,6 +53,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
 	struct perf_test_vcpu_args *vcpu_args = &pta->vcpu_args[vcpu_id];
 	uint64_t gva;
 	uint64_t pages;
+	uint64_t addr;
 	int i;
 	uint32_t rand = pta->random_seed + vcpu_id;
 
@@ -64,7 +65,14 @@ void perf_test_guest_code(uint32_t vcpu_id)
 
 	while (true) {
 		for (i = 0; i < pages; i++) {
-			uint64_t addr = gva + (i * pta->guest_page_size);
+			rand = perf_test_random(rand);
+
+			if (pta->random_access)
+				addr = gva +
+					((rand % pages) * pta->guest_page_size);
+			else
+				addr = gva + (i * pta->guest_page_size);
+
 			rand = perf_test_random(rand);
 
 			if (rand % 100 < pta->write_percent)
@@ -240,6 +248,12 @@ void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed)
 	sync_global_to_guest(vm, perf_test_args);
 }
 
+void perf_test_set_random_access(struct kvm_vm *vm, bool random_access)
+{
+	perf_test_args.random_access = random_access;
+	sync_global_to_guest(vm, perf_test_args);
+}
+
 uint64_t __weak perf_test_nested_pages(int nr_vcpus)
 {
 	return 0;
-- 
2.37.2.789.g6183377224-goog

Re: [PATCH v3 3/3] KVM: selftests: Randomize page access order.

[Ricardo Koller]

On Thu, Sep 01, 2022 at 07:52:37PM +0000, Colton Lewis wrote:
> Create the ability to randomize page access order with the -a
> argument, including the possibility that the same pages may be hit
> multiple times during an iteration or not at all.
> 
> Signed-off-by: Colton Lewis <coltonlewis@google.com>
> ---
>  .../testing/selftests/kvm/dirty_log_perf_test.c  | 12 ++++++++++--
>  .../selftests/kvm/include/perf_test_util.h       |  2 ++
>  tools/testing/selftests/kvm/lib/perf_test_util.c | 16 +++++++++++++++-
>  3 files changed, 27 insertions(+), 3 deletions(-)
> 
> diff --git a/tools/testing/selftests/kvm/dirty_log_perf_test.c b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> index c9441f8354be..631b3883ed12 100644
> --- a/tools/testing/selftests/kvm/dirty_log_perf_test.c
> +++ b/tools/testing/selftests/kvm/dirty_log_perf_test.c
> @@ -127,6 +127,7 @@ struct test_params {
>  	int slots;
>  	uint32_t write_percent;
>  	uint32_t random_seed;
> +	bool random_access;
>  };
>  
>  static void toggle_dirty_logging(struct kvm_vm *vm, int slots, bool enable)
> @@ -269,6 +270,9 @@ static void run_test(enum vm_guest_mode mode, void *arg)
>  	pr_info("Enabling dirty logging time: %ld.%.9lds\n\n",
>  		ts_diff.tv_sec, ts_diff.tv_nsec);
>  
> +	/* Set random access here, after population phase. */
> +	perf_test_set_random_access(vm, p->random_access);
> +
>  	while (iteration < p->iterations) {
>  		/*
>  		 * Incrementing the iteration number will start the vCPUs
> @@ -339,10 +343,11 @@ static void run_test(enum vm_guest_mode mode, void *arg)
>  static void help(char *name)
>  {
>  	puts("");
> -	printf("usage: %s [-h] [-i iterations] [-p offset] [-g] "
> +	printf("usage: %s [-h] [-a] [-i iterations] [-p offset] [-g] "
>  	       "[-m mode] [-n] [-b vcpu bytes] [-v vcpus] [-o] [-r random seed ] [-s mem type]"
>  	       "[-x memslots] [-w percentage]\n", name);
>  	puts("");
> +	printf(" -a: access memory randomly rather than in order.\n");
>  	printf(" -i: specify iteration counts (default: %"PRIu64")\n",
>  	       TEST_HOST_LOOP_N);
>  	printf(" -g: Do not enable KVM_CAP_MANUAL_DIRTY_LOG_PROTECT2. This\n"
> @@ -394,8 +399,11 @@ int main(int argc, char *argv[])
>  
>  	guest_modes_append_default();
>  
> -	while ((opt = getopt(argc, argv, "ghi:p:m:nb:v:or:s:x:w:")) != -1) {
> +	while ((opt = getopt(argc, argv, "aghi:p:m:nb:v:or:s:x:w:")) != -1) {
>  		switch (opt) {
> +		case 'a':
> +			p.random_access = true;
> +			break;
>  		case 'g':
>  			dirty_log_manual_caps = 0;
>  			break;
> diff --git a/tools/testing/selftests/kvm/include/perf_test_util.h b/tools/testing/selftests/kvm/include/perf_test_util.h
> index 52c450eb6b9b..380c31375b60 100644
> --- a/tools/testing/selftests/kvm/include/perf_test_util.h
> +++ b/tools/testing/selftests/kvm/include/perf_test_util.h
> @@ -39,6 +39,7 @@ struct perf_test_args {
>  
>  	/* Run vCPUs in L2 instead of L1, if the architecture supports it. */
>  	bool nested;
> +	bool random_access;
>  
>  	struct perf_test_vcpu_args vcpu_args[KVM_MAX_VCPUS];
>  };
> @@ -53,6 +54,7 @@ void perf_test_destroy_vm(struct kvm_vm *vm);
>  
>  void perf_test_set_write_percent(struct kvm_vm *vm, uint32_t write_percent);
>  void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed);
> +void perf_test_set_random_access(struct kvm_vm *vm, bool random_access);
>  
>  void perf_test_start_vcpu_threads(int vcpus, void (*vcpu_fn)(struct perf_test_vcpu_args *));
>  void perf_test_join_vcpu_threads(int vcpus);
> diff --git a/tools/testing/selftests/kvm/lib/perf_test_util.c b/tools/testing/selftests/kvm/lib/perf_test_util.c
> index be6176faaf8e..c6123b75d5e3 100644
> --- a/tools/testing/selftests/kvm/lib/perf_test_util.c
> +++ b/tools/testing/selftests/kvm/lib/perf_test_util.c
> @@ -53,6 +53,7 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  	struct perf_test_vcpu_args *vcpu_args = &pta->vcpu_args[vcpu_id];
>  	uint64_t gva;
>  	uint64_t pages;
> +	uint64_t addr;
>  	int i;
>  	uint32_t rand = pta->random_seed + vcpu_id;
>  
> @@ -64,7 +65,14 @@ void perf_test_guest_code(uint32_t vcpu_id)
>  
>  	while (true) {
>  		for (i = 0; i < pages; i++) {
> -			uint64_t addr = gva + (i * pta->guest_page_size);
> +			rand = perf_test_random(rand);
> +
> +			if (pta->random_access)
> +				addr = gva +
> +					((rand % pages) * pta->guest_page_size);
> +			else
> +				addr = gva + (i * pta->guest_page_size);
> +
>  			rand = perf_test_random(rand);
>  
>  			if (rand % 100 < pta->write_percent)
> @@ -240,6 +248,12 @@ void perf_test_set_random_seed(struct kvm_vm *vm, uint32_t random_seed)
>  	sync_global_to_guest(vm, perf_test_args);
>  }
>  
> +void perf_test_set_random_access(struct kvm_vm *vm, bool random_access)
> +{
> +	perf_test_args.random_access = random_access;
> +	sync_global_to_guest(vm, perf_test_args);
> +}

This can be merged onto the sync_global_to_guest() done for the seed.

> +
>  uint64_t __weak perf_test_nested_pages(int nr_vcpus)
>  {
>  	return 0;
> -- 
> 2.37.2.789.g6183377224-goog
> 

Re: [PATCH v3 3/3] KVM: selftests: Randomize page access order.

[Colton Lewis]

Ricardo Koller <ricarkol@google.com> writes:

>> +void perf_test_set_random_access(struct kvm_vm *vm, bool random_access)
>> +{
>> +	perf_test_args.random_access = random_access;
>> +	sync_global_to_guest(vm, perf_test_args);
>> +}

> This can be merged onto the sync_global_to_guest() done for the seed.


That seems like a micro optimization. I'd prefer every perf_test_set*
function have consistent behavior.

If needed, sync_global_to_guest should work on the individual field.