[PATCH i-g-t v3 11/14] tests/xe_exec_sip: Extend SIP interaction testing

Zbigniew Kempczyński zbigniew.kempczynski at intel.com
Wed Aug 21 09:49:53 UTC 2024


On Fri, Aug 09, 2024 at 02:38:10PM +0200, Christoph Manszewski wrote:
> Extend xe_exec_sip test by adding subtests that check SIP interaction
> sanity with regard to resets and hardware debugging capabilities like
> breakpoints, and invalid instruction exceptions.
> 
> Signed-off-by: Mika Kuoppala <mika.kuoppala at linux.intel.com>
> Signed-off-by: Dominik Grzegorzek <dominik.grzegorzek at intel.com>
> Signed-off-by: Christoph Manszewski <christoph.manszewski at intel.com>
> Signed-off-by: Dominik Karol Piątkowski <dominik.karol.piatkowski at intel.com>
> Signed-off-by: Karolina Stolarek <karolina.stolarek at intel.com>
> Signed-off-by: Andrzej Hajda <andrzej.hajda at intel.com>
> ---
>  tests/intel/xe_exec_sip.c | 332 +++++++++++++++++++++++++++++++++++---
>  1 file changed, 310 insertions(+), 22 deletions(-)
> 
> diff --git a/tests/intel/xe_exec_sip.c b/tests/intel/xe_exec_sip.c
> index 4b599e7f6..0db0dc4b8 100644
> --- a/tests/intel/xe_exec_sip.c
> +++ b/tests/intel/xe_exec_sip.c
> @@ -21,6 +21,7 @@
>  #include "gpgpu_shader.h"
>  #include "igt.h"
>  #include "igt_sysfs.h"
> +#include "xe/xe_eudebug.h"
>  #include "xe/xe_ioctl.h"
>  #include "xe/xe_query.h"
>  
> @@ -30,9 +31,29 @@
>  #define COLOR_C4 0xc4
>  
>  #define SHADER_CANARY 0x01010101
> +#define SIP_CANARY 0x02020202
>  
>  #define NSEC_PER_MSEC (1000 * 1000ull)
>  
> +#define SHADER_BREAKPOINT 0
> +#define SHADER_WRITE 1
> +#define SHADER_WAIT 2
> +#define SHADER_INV_INSTR_DISABLED 3
> +#define SHADER_INV_INSTR_THREAD_ENABLED 4
> +#define SHADER_INV_INSTR_WALKER_ENABLED 5
> +#define SHADER_HANG 6
> +#define SIP_WRITE 7
> +#define SIP_NULL 8
> +#define SIP_WAIT 9
> +#define SIP_HEAVY 10
> +#define SIP_INV_INSTR 11
> +
> +#define F_SUBMIT_TWICE	(1 << 0)
> +
> +/* Control Register cr0.1 bits for exception handling */
> +#define ILLEGAL_OPCODE_ENABLE BIT(12)
> +#define ILLEGAL_OPCODE_STATUS BIT(28)
> +
>  static struct intel_buf *
>  create_fill_buf(int fd, int width, int height, uint8_t color)
>  {
> @@ -52,24 +73,109 @@ create_fill_buf(int fd, int width, int height, uint8_t color)
>  	return buf;
>  }
>  
> -static struct gpgpu_shader *get_shader(int fd)
> +static struct gpgpu_shader *get_shader(int fd, const int shadertype)
>  {
>  	static struct gpgpu_shader *shader;
> +	uint32_t bad;
>  
>  	shader = gpgpu_shader_create(fd);
> +	if (shadertype == SHADER_INV_INSTR_WALKER_ENABLED)
> +		shader->illegal_opcode_exception_enable = true;
> +
>  	gpgpu_shader__write_dword(shader, SHADER_CANARY, 0);
> +
> +	switch (shadertype) {
> +	case SHADER_HANG:
> +		gpgpu_shader__label(shader, 0);
> +		gpgpu_shader__nop(shader);
> +		gpgpu_shader__jump(shader, 0);
> +		break;
> +	case SHADER_WAIT:
> +		gpgpu_shader__wait(shader);
> +		break;
> +	case SHADER_WRITE:
> +		break;
> +	case SHADER_BREAKPOINT:
> +		gpgpu_shader__nop(shader);
> +		gpgpu_shader__breakpoint(shader);
> +		break;
> +	case SHADER_INV_INSTR_THREAD_ENABLED:
> +		gpgpu_shader__set_exception(shader, ILLEGAL_OPCODE_ENABLE);
> +		/* fall through */
> +	case SHADER_INV_INSTR_DISABLED:
> +	case SHADER_INV_INSTR_WALKER_ENABLED:
> +		bad = (shadertype == SHADER_INV_INSTR_DISABLED) ? ILLEGAL_OPCODE_ENABLE : 0;
> +		gpgpu_shader__write_on_exception(shader, 1, 0, ILLEGAL_OPCODE_ENABLE, bad);
> +		gpgpu_shader__nop(shader);
> +		gpgpu_shader__nop(shader);
> +		/* modify second nop, set only opcode bits[6:0] */
> +		shader->instr[shader->size/4 - 1][0] = 0x7f;
> +		/* SIP should clear exception bit */
> +		bad = ILLEGAL_OPCODE_STATUS;
> +		gpgpu_shader__write_on_exception(shader, 2, 0, ILLEGAL_OPCODE_STATUS, bad);
> +		break;
> +	}
> +
>  	gpgpu_shader__eot(shader);
>  	return shader;
>  }
>  
> -static uint32_t gpgpu_shader(int fd, struct intel_bb *ibb, unsigned int threads,
> -			     unsigned int width, unsigned int height)
> +static struct gpgpu_shader *get_sip(int fd, const int siptype,
> +				    const int shadertype, unsigned int y_offset)
> +{
> +	static struct gpgpu_shader *sip;
> +
> +	if (siptype == SIP_NULL)
> +		return NULL;
> +
> +	sip = gpgpu_shader_create(fd);
> +	gpgpu_shader__write_dword(sip, SIP_CANARY, y_offset);
> +
> +	switch (siptype) {
> +	case SIP_WRITE:
> +		break;
> +	case SIP_WAIT:
> +		gpgpu_shader__wait(sip);
> +		break;
> +	case SIP_HEAVY:
> +		/* Depending on the generation, the production sip
> +		 * executes between 145 to 157 instructions.
> +		 * It performs at most 45 data port writes and 5 data port reads.
> +		 * Make sure our heavy sip is at least twice heavy as production one.
> +		 */
> +		gpgpu_shader__loop_begin(sip, 0);
> +		gpgpu_shader__write_dword(sip, 0xdeadbeef, y_offset);
> +		gpgpu_shader__write_dword(sip, SIP_CANARY, y_offset);
> +		gpgpu_shader__loop_end(sip, 0, 45);
> +
> +		gpgpu_shader__loop_begin(sip, 1);
> +		gpgpu_shader__jump_neq(sip, 1, y_offset, SIP_CANARY);
> +		gpgpu_shader__loop_end(sip, 1, 10);
> +
> +		gpgpu_shader__wait(sip);
> +		break;
> +	case SIP_INV_INSTR:
> +		gpgpu_shader__write_on_exception(sip, 1, y_offset, ILLEGAL_OPCODE_STATUS, 0);
> +		break;
> +	}
> +
> +	gpgpu_shader__end_system_routine(sip, shadertype == SHADER_BREAKPOINT);
> +	return sip;
> +}
> +
> +static uint32_t gpgpu_shader(int fd, struct intel_bb *ibb, const int shadertype, const int siptype,
> +			     unsigned int threads, unsigned int width, unsigned int height)
>  {
>  	struct intel_buf *buf = create_fill_buf(fd, width, height, COLOR_C4);
> -	struct gpgpu_shader *shader = get_shader(fd);
> +	struct gpgpu_shader *sip = get_sip(fd, siptype, shadertype, height / 2);
> +	struct gpgpu_shader *shader = get_shader(fd, shadertype);
>  
> -	gpgpu_shader_exec(ibb, buf, 1, threads, shader, NULL, 0, 0);
> +	gpgpu_shader_exec(ibb, buf, 1, threads, shader, sip, 0, 0);
> +
> +	if (sip)
> +		gpgpu_shader_destroy(sip);
>  	gpgpu_shader_destroy(shader);
> +
>  	return buf->handle;
>  }
>  
> @@ -83,11 +189,11 @@ static void check_fill_buf(uint8_t *ptr, const int width, const int x,
>  		     color, val, x, y);
>  }
>  
> -static void check_buf(int fd, uint32_t handle, int width, int height,
> -		      uint8_t poison_c)
> +static int check_buf(int fd, uint32_t handle, int width, int height,
> +		      int shadertype, int siptype, uint8_t poison_c)
>  {
>  	unsigned int sz = ALIGN(width * height, 4096);
> -	int thread_count = 0;
> +	int thread_count = 0, sip_count = 0;
>  	uint32_t *ptr;
>  	int i, j;
>  
> @@ -105,9 +211,87 @@ static void check_buf(int fd, uint32_t handle, int width, int height,
>  		i = 0;
>  	}
>  
> +	for (i = 0, j = height / 2; j < height; ++j) {
> +		if (ptr[j * width / 4] == SIP_CANARY) {
> +			++sip_count;
> +			i = 4;
> +		}
> +
> +		for (; i < width; i++)
> +			check_fill_buf((uint8_t *)ptr, width, i, j, poison_c);
> +
> +		i = 0;
> +	}
> +
>  	igt_assert(thread_count);
> +	if (shadertype == SHADER_INV_INSTR_DISABLED)
> +		igt_assert(!sip_count);
> +	else if ((siptype != SIP_NULL && xe_eudebug_debugger_available(fd)) ||
> +		 (siptype == SIP_INV_INSTR && shadertype != SHADER_INV_INSTR_DISABLED))
> +		igt_assert_f(thread_count == sip_count,
> +			     "Thread and SIP count mismatch, %d != %d\n",
> +			     thread_count, sip_count);
> +	else
> +		igt_assert(sip_count == 0);
>  
>  	munmap(ptr, sz);
> +
> +	return sip_count;
> +}
> +
> +#define USERCOREDUMP_FORMAT "usercoredumps/%d/%d"
> +static char *get_latest_usercoredump(int dir)
> +{
> +	char tmp[256];
> +	int i = 1;
> +
> +	do {
> +		snprintf(tmp, sizeof(tmp), USERCOREDUMP_FORMAT, getpid(), i++);
> +	} while (igt_sysfs_has_attr(dir, tmp));
> +
> +	snprintf(tmp, sizeof(tmp), USERCOREDUMP_FORMAT, getpid(), i-2);
> +	return igt_sysfs_get(dir, tmp);
> +}
> +
> +static void check_usercoredump(int fd, int sip, int dispatched)
> +{
> +	int dir = igt_debugfs_dir(fd);
> +	char *usercoredump, *str;
> +	unsigned int before, after;
> +	char match[256];
> +
> +	if (sip != SIP_WAIT && sip != SIP_HEAVY)
> +		return;
> +
> +	/* XXX reinstate when offline coredumps are implemented */
> +#ifndef XXX_ATTENTIONS_THROUGH_COREDUMPS
> +	return;
> +#endif
> +	usercoredump = get_latest_usercoredump(dir);
> +	igt_assert(usercoredump);
> +	igt_debug("%s\n", usercoredump);
> +
> +	snprintf(match, sizeof(match), "PID: %d", getpid());
> +	str = strstr(usercoredump, match);
> +	igt_assert(str);
> +
> +	snprintf(match, sizeof(match), "Comm: %s", igt_test_name());
> +	str = strstr(str, match);
> +	igt_assert(str);
> +
> +	str = strstr(str, "TD_ATT");
> +	igt_assert(str);
> +	igt_assert_eq(sscanf(str, "TD_ATT before (%d):", &before), 1);
> +	str = strstr(str + 1, "TD_ATT");
> +	igt_assert_eq(sscanf(str, "TD_ATT after (%d):", &after), 1);
> +
> +	igt_info("attentions %d before, %d after\n", before, after);
> +
> +	igt_assert_eq(before, dispatched);
> +	igt_assert_eq(after, dispatched);
> +
> +	free(usercoredump);
> +	close(dir);
>  }
>  
>  static uint64_t
> @@ -128,17 +312,58 @@ xe_sysfs_get_job_timeout_ms(int fd, struct drm_xe_engine_class_instance *eci)
>   * Description: check basic shader with write operation
>   * Run type: BAT
>   *
> + * SUBTEST: sanity-after-timeout
> + * Description: check basic shader execution after job timeout
> + *
> + * SUBTEST: wait-writesip-nodebug
> + * Description: verify that we don't enter SIP after wait with debugging disabled.
> + *
> + * SUBTEST: invalidinstr-disabled
> + * Description: Verify that we don't enter SIP after running into an invalid
> + *		instruction when exception is not enabled.
> + *
> + * SUBTEST: invalidinstr-thread-enabled
> + * Description: Verify that we enter SIP after running into an invalid instruction
> + *              when exception is enabled from thread.
> + *
> + * SUBTEST: invalidinstr-walker-enabled
> + * Description: Verify that we enter SIP after running into an invalid instruction
> + *              when exception is enabled from COMPUTE_WALKER.
> + *
> + * SUBTEST: breakpoint-writesip-nodebug
> + * Description: verify that we don't enter SIP after hitting breakpoint in shader
> + *		when debugging is disabled.
> + *
> + * SUBTEST: breakpoint-writesip
> + * Description: Test that we enter SIP after hitting breakpoint in shader.
> + *
> + * SUBTEST: breakpoint-writesip-twice
> + * Description: Test twice that we enter SIP after hitting breakpoint in shader.
> + *
> + * SUBTEST: breakpoint-waitsip
> + * Description: Test that we reset after seeing the attention without the debugger.
> + *
> + * SUBTEST: breakpoint-waitsip-heavy
> + * Description:
> + *	Test that we reset after seeing the attention from heavy SIP, that resembles
> + *	the production one, without the debugger.
>   */
> -static void test_sip(struct drm_xe_engine_class_instance *eci, uint32_t flags)
> +static void test_sip(int shader, int sip, struct drm_xe_engine_class_instance *eci, uint32_t flags)
>  {
>  	unsigned int threads = 512;
>  	unsigned int height = max_t(threads, HEIGHT, threads * 2);
> -	uint32_t exec_queue_id, handle, vm_id;
>  	unsigned int width = WIDTH;
> +	struct drm_xe_ext_set_property ext = {
> +		.base.name = DRM_XE_EXEC_QUEUE_EXTENSION_SET_PROPERTY,
> +		.property = DRM_XE_EXEC_QUEUE_SET_PROPERTY_EUDEBUG,
> +		.value = DRM_XE_EXEC_QUEUE_EUDEBUG_FLAG_ENABLE,
> +	};
>  	struct timespec ts = { };
> -	uint64_t timeout;
> +	int done = 0;
> +	uint32_t exec_queue_id, handle, vm_id;
>  	struct intel_bb *ibb;
> -	int fd;
> +	uint64_t timeout;
> +	int dispatched, fd;
>  
>  	igt_debug("Using %s\n", xe_engine_class_string(eci->engine_class));
>  
> @@ -152,19 +377,24 @@ static void test_sip(struct drm_xe_engine_class_instance *eci, uint32_t flags)
>  	timeout *= NSEC_PER_MSEC;
>  	timeout *= igt_run_in_simulation() ? 10 : 1;
>  
> -	exec_queue_id = xe_exec_queue_create(fd, vm_id, eci, 0);
> -	ibb = intel_bb_create_with_context(fd, exec_queue_id, vm_id, NULL, 4096);
> +	exec_queue_id = xe_exec_queue_create(fd, vm_id, eci,
> +					     xe_eudebug_debugger_available(fd) ?
> +					     to_user_pointer(&ext) : 0);
> +	do {
> +		ibb = intel_bb_create_with_context(fd, exec_queue_id, vm_id, NULL, 4096);
>  
> -	igt_nsec_elapsed(&ts);
> -	handle = gpgpu_shader(fd, ibb, threads, width, height);
> +		igt_nsec_elapsed(&ts);
> +		handle = gpgpu_shader(fd, ibb, shader, sip, threads, width, height);
>  
> -	intel_bb_sync(ibb);
> -	igt_assert_lt_u64(igt_nsec_elapsed(&ts), timeout);
> +		intel_bb_sync(ibb);
> +		igt_assert_lt_u64(igt_nsec_elapsed(&ts), timeout);
>  
> -	check_buf(fd, handle, width, height, COLOR_C4);
> +		dispatched = check_buf(fd, handle, width, height, shader, sip, COLOR_C4);
> +		check_usercoredump(fd, sip, dispatched);
>  
> -	gem_close(fd, handle);
> -	intel_bb_destroy(ibb);
> +		gem_close(fd, handle);
> +		intel_bb_destroy(ibb);
> +	} while (!done++ && (flags & F_SUBMIT_TWICE));
>  
>  	xe_exec_queue_destroy(fd, exec_queue_id);
>  	xe_vm_destroy(fd, vm_id);
> @@ -183,13 +413,71 @@ static void test_sip(struct drm_xe_engine_class_instance *eci, uint32_t flags)
>  igt_main
>  {
>  	struct drm_xe_engine_class_instance *eci;
> +	bool was_enabled;
>  	int fd;
>  
>  	igt_fixture
>  		fd = drm_open_driver(DRIVER_XE);
>  
>  	test_render_and_compute("sanity", fd, eci)
> -		test_sip(eci, 0);
> +		test_sip(SHADER_WRITE, SIP_NULL, eci, 0);
> +
> +	test_render_and_compute("sanity-after-timeout", fd, eci) {
> +		test_sip(SHADER_HANG, SIP_NULL, eci, 0);
> +
> +		xe_for_each_engine(fd, eci)
> +			if (eci->engine_class == DRM_XE_ENGINE_CLASS_RENDER ||
> +			    eci->engine_class == DRM_XE_ENGINE_CLASS_COMPUTE)
> +				test_sip(SHADER_WRITE, SIP_NULL, eci, 0);
> +	}
> +
> +	/* Debugger disabled (TD_CTL not set) */
> +	igt_subtest_group {
> +		igt_fixture {
> +			was_enabled = xe_eudebug_enable(fd, false);
> +			igt_require(!xe_eudebug_debugger_available(fd));
> +		}
> +
> +		test_render_and_compute("wait-writesip-nodebug", fd, eci)
> +			test_sip(SHADER_WAIT, SIP_WRITE, eci, 0);
> +
> +		test_render_and_compute("invalidinstr-disabled", fd, eci)
> +			test_sip(SHADER_INV_INSTR_DISABLED, SIP_INV_INSTR, eci, 0);
> +
> +		test_render_and_compute("invalidinstr-thread-enabled", fd, eci)
> +			test_sip(SHADER_INV_INSTR_THREAD_ENABLED, SIP_INV_INSTR, eci, 0);
> +
> +		test_render_and_compute("invalidinstr-walker-enabled", fd, eci)
> +			test_sip(SHADER_INV_INSTR_WALKER_ENABLED, SIP_INV_INSTR, eci, 0);
> +
> +		test_render_and_compute("breakpoint-writesip-nodebug", fd, eci)
> +			test_sip(SHADER_BREAKPOINT, SIP_WRITE, eci, 0);
> +
> +		igt_fixture
> +			xe_eudebug_enable(fd, was_enabled);
> +	}
> +
> +	/* Debugger enabled (TD_CTL set) */
> +	igt_subtest_group {
> +		igt_fixture {
> +			was_enabled = xe_eudebug_enable(fd, true);

Shouldn't this require the debugger to be available?

--
Zbigniew

> +		}
> +
> +		test_render_and_compute("breakpoint-writesip", fd, eci)
> +			test_sip(SHADER_BREAKPOINT, SIP_WRITE, eci, 0);
> +
> +		test_render_and_compute("breakpoint-writesip-twice", fd, eci)
> +			test_sip(SHADER_BREAKPOINT, SIP_WRITE, eci, F_SUBMIT_TWICE);
> +
> +		test_render_and_compute("breakpoint-waitsip", fd, eci)
> +			test_sip(SHADER_BREAKPOINT, SIP_WAIT, eci, 0);
> +
> +		test_render_and_compute("breakpoint-waitsip-heavy", fd, eci)
> +			test_sip(SHADER_BREAKPOINT, SIP_HEAVY, eci, 0);
> +
> +		igt_fixture
> +			xe_eudebug_enable(fd, was_enabled);
> +	}
>  
>  	igt_fixture
>  		drm_close_driver(fd);
> -- 
> 2.34.1
> 


More information about the igt-dev mailing list