/* * Copyright (c) 2013 Intel Corporation * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice (including the next * paragraph) shall be included in all copies or substantial portions of the * Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS * IN THE SOFTWARE. * * Authors: * Mika Kuoppala * */ #include "igt.h" #include "igt_sysfs.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #define RS_NO_ERROR 0 #define RS_BATCH_ACTIVE (1 << 0) #define RS_BATCH_PENDING (1 << 1) #define RS_UNKNOWN (1 << 2) static uint32_t devid; struct local_drm_i915_reset_stats { __u32 ctx_id; __u32 flags; __u32 reset_count; __u32 batch_active; __u32 batch_pending; __u32 pad; }; #define MAX_FD 32 #define GET_RESET_STATS_IOCTL DRM_IOWR(DRM_COMMAND_BASE + 0x32, struct local_drm_i915_reset_stats) #define LOCAL_I915_EXEC_VEBOX (4 << 0) static void sync_gpu(void) { int fd = drm_open_driver(DRIVER_INTEL); gem_quiescent_gpu(fd); close(fd); } static int noop(int fd, uint32_t ctx, const struct intel_execution_engine *e) { const uint32_t bbe = MI_BATCH_BUFFER_END; struct drm_i915_gem_execbuffer2 eb; struct drm_i915_gem_exec_object2 exec; int ret; memset(&exec, 0, sizeof(exec)); exec.handle = gem_create(fd, 4096); igt_assert((int)exec.handle > 0); gem_write(fd, exec.handle, 0, &bbe, sizeof(bbe)); memset(&eb, 0, sizeof(eb)); eb.buffers_ptr = to_user_pointer(&exec); eb.buffer_count = 1; eb.flags = eb_ring(e); i915_execbuffer2_set_context_id(eb, ctx); ret = __gem_execbuf(fd, &eb); if (ret < 0) { gem_close(fd, exec.handle); return ret; } return exec.handle; } static int has_engine(int fd, uint32_t ctx, const struct intel_execution_engine *e) { int handle = noop(fd, ctx, e); if (handle < 0) return 0; gem_close(fd, handle); return 1; } static void check_context(const struct intel_execution_engine *e) { int fd = drm_open_driver(DRIVER_INTEL); gem_require_contexts(fd); igt_require(has_engine(fd, gem_context_create(fd), e)); close(fd); } static int gem_reset_stats(int fd, int ctx_id, struct local_drm_i915_reset_stats *rs) { memset(rs, 0, sizeof(*rs)); rs->ctx_id = ctx_id; rs->reset_count = -1; if (drmIoctl(fd, GET_RESET_STATS_IOCTL, rs)) return -errno; igt_assert(rs->reset_count != -1); return 0; } static int gem_reset_status(int fd, int ctx_id) { struct local_drm_i915_reset_stats rs; int ret; ret = gem_reset_stats(fd, ctx_id, &rs); if (ret) return ret; if (rs.batch_active) return RS_BATCH_ACTIVE; if (rs.batch_pending) return RS_BATCH_PENDING; return RS_NO_ERROR; } static struct timespec ts_injected; #define BAN HANG_ALLOW_BAN #define ASYNC 2 static void inject_hang(int fd, uint32_t ctx, const struct intel_execution_engine *e, unsigned flags) { igt_hang_t hang; clock_gettime(CLOCK_MONOTONIC, &ts_injected); hang = igt_hang_ctx(fd, ctx, eb_ring(e), flags & BAN); if ((flags & ASYNC) == 0) igt_post_hang_ring(fd, hang); } static const char *status_to_string(int x) { const char *strings[] = { "No error", "Guilty", "Pending", }; if (x >= ARRAY_SIZE(strings)) return "Unknown"; return strings[x]; } static int _assert_reset_status(int idx, int fd, int ctx, int status) { int rs; rs = gem_reset_status(fd, ctx); if (rs < 0) { igt_info("reset status for %d ctx %d returned %d\n", idx, ctx, rs); return rs; } if (rs != status) { igt_info("%d:%d expected '%s' [%d], found '%s' [%d]\n", idx, ctx, status_to_string(status), status, status_to_string(rs), rs); return 1; } return 0; } #define assert_reset_status(idx, fd, ctx, status) \ igt_assert(_assert_reset_status(idx, fd, ctx, status) == 0) static void test_rs(const struct intel_execution_engine *e, int num_fds, int hang_index, int rs_assumed_no_hang) { int fd[MAX_FD]; int i; igt_assert_lte(num_fds, MAX_FD); igt_assert_lt(hang_index, MAX_FD); igt_debug("num fds=%d, hang index=%d\n", num_fds, hang_index); for (i = 0; i < num_fds; i++) { fd[i] = drm_open_driver(DRIVER_INTEL); assert_reset_status(i, fd[i], 0, RS_NO_ERROR); } sync_gpu(); for (i = 0; i < num_fds; i++) { if (i == hang_index) inject_hang(fd[i], 0, e, ASYNC); else igt_assert(noop(fd[i], 0, e) > 0); } sync_gpu(); for (i = 0; i < num_fds; i++) { if (hang_index < 0) { assert_reset_status(i, fd[i], 0, rs_assumed_no_hang); continue; } if (i < hang_index) assert_reset_status(i, fd[i], 0, RS_NO_ERROR); if (i == hang_index) assert_reset_status(i, fd[i], 0, RS_BATCH_ACTIVE); if (i > hang_index) assert_reset_status(i, fd[i], 0, RS_NO_ERROR); } igt_assert(igt_seconds_elapsed(&ts_injected) <= 30); for (i = 0; i < num_fds; i++) close(fd[i]); } #define MAX_CTX 100 static void test_rs_ctx(const struct intel_execution_engine *e, int num_fds, int num_ctx, int hang_index, int hang_context) { int i, j; int fd[MAX_FD]; int ctx[MAX_FD][MAX_CTX]; igt_assert_lte(num_fds, MAX_FD); igt_assert_lt(hang_index, MAX_FD); igt_assert_lte(num_ctx, MAX_CTX); igt_assert_lt(hang_context, MAX_CTX); test_rs(e, num_fds, -1, RS_NO_ERROR); for (i = 0; i < num_fds; i++) { fd[i] = drm_open_driver(DRIVER_INTEL); igt_assert(fd[i]); assert_reset_status(i, fd[i], 0, RS_NO_ERROR); for (j = 0; j < num_ctx; j++) { ctx[i][j] = gem_context_create(fd[i]); } assert_reset_status(i, fd[i], 0, RS_NO_ERROR); } for (i = 0; i < num_fds; i++) { assert_reset_status(i, fd[i], 0, RS_NO_ERROR); for (j = 0; j < num_ctx; j++) assert_reset_status(i, fd[i], ctx[i][j], RS_NO_ERROR); assert_reset_status(i, fd[i], 0, RS_NO_ERROR); } for (i = 0; i < num_fds; i++) { for (j = 0; j < num_ctx; j++) { if (i == hang_index && j == hang_context) inject_hang(fd[i], ctx[i][j], e, ASYNC); else igt_assert(noop(fd[i], ctx[i][j], e) > 0); } } sync_gpu(); igt_assert(igt_seconds_elapsed(&ts_injected) <= 30); for (i = 0; i < num_fds; i++) assert_reset_status(i, fd[i], 0, RS_NO_ERROR); for (i = 0; i < num_fds; i++) { for (j = 0; j < num_ctx; j++) { if (i < hang_index) assert_reset_status(i, fd[i], ctx[i][j], RS_NO_ERROR); if (i == hang_index && j < hang_context) assert_reset_status(i, fd[i], ctx[i][j], RS_NO_ERROR); if (i == hang_index && j == hang_context) assert_reset_status(i, fd[i], ctx[i][j], RS_BATCH_ACTIVE); if (i == hang_index && j > hang_context) assert_reset_status(i, fd[i], ctx[i][j], RS_NO_ERROR); if (i > hang_index) assert_reset_status(i, fd[i], ctx[i][j], RS_NO_ERROR); } } for (i = 0; i < num_fds; i++) { assert_reset_status(i, fd[i], 0, RS_NO_ERROR); close(fd[i]); } } static void test_ban(const struct intel_execution_engine *e) { struct local_drm_i915_reset_stats rs_bad, rs_good; int fd_bad, fd_good; int ban, retry = 10; int active_count = 0; fd_bad = drm_open_driver(DRIVER_INTEL); fd_good = drm_open_driver(DRIVER_INTEL); assert_reset_status(fd_bad, fd_bad, 0, RS_NO_ERROR); assert_reset_status(fd_good, fd_good, 0, RS_NO_ERROR); noop(fd_bad, 0, e); noop(fd_good, 0, e); assert_reset_status(fd_bad, fd_bad, 0, RS_NO_ERROR); assert_reset_status(fd_good, fd_good, 0, RS_NO_ERROR); inject_hang(fd_bad, 0, e, BAN | ASYNC); active_count++; noop(fd_good, 0, e); noop(fd_good, 0, e); while (retry--) { inject_hang(fd_bad, 0, e, BAN); active_count++; ban = noop(fd_bad, 0, e); if (ban == -EIO) break; /* Should not happen often but sometimes hang is declared too * slow due to our way of faking hang using loop */ gem_close(fd_bad, ban); igt_info("retrying for ban (%d)\n", retry); } igt_assert_eq(ban, -EIO); igt_assert_lt(0, noop(fd_good, 0, e)); assert_reset_status(fd_bad, fd_bad, 0, RS_BATCH_ACTIVE); igt_assert_eq(gem_reset_stats(fd_bad, 0, &rs_bad), 0); igt_assert_eq(rs_bad.batch_active, active_count); assert_reset_status(fd_good, fd_good, 0, RS_NO_ERROR); igt_assert_eq(gem_reset_stats(fd_good, 0, &rs_good), 0); igt_assert_eq(rs_good.batch_active, 0); close(fd_bad); close(fd_good); } static void test_ban_ctx(const struct intel_execution_engine *e) { struct local_drm_i915_reset_stats rs_bad, rs_good; int fd, ban, retry = 10; uint32_t ctx_good, ctx_bad; int active_count = 0; fd = drm_open_driver(DRIVER_INTEL); assert_reset_status(fd, fd, 0, RS_NO_ERROR); ctx_good = gem_context_create(fd); ctx_bad = gem_context_create(fd); assert_reset_status(fd, fd, 0, RS_NO_ERROR); assert_reset_status(fd, fd, ctx_good, RS_NO_ERROR); assert_reset_status(fd, fd, ctx_bad, RS_NO_ERROR); noop(fd, ctx_bad, e); noop(fd, ctx_good, e); assert_reset_status(fd, fd, ctx_good, RS_NO_ERROR); assert_reset_status(fd, fd, ctx_bad, RS_NO_ERROR); inject_hang(fd, ctx_bad, e, BAN | ASYNC); active_count++; noop(fd, ctx_good, e); noop(fd, ctx_good, e); while (retry--) { inject_hang(fd, ctx_bad, e, BAN); active_count++; ban = noop(fd, ctx_bad, e); if (ban == -EIO) break; /* Should not happen often but sometimes hang is declared too * slow due to our way of faking hang using loop */ gem_close(fd, ban); igt_info("retrying for ban (%d)\n", retry); } igt_assert_eq(ban, -EIO); igt_assert_lt(0, noop(fd, ctx_good, e)); assert_reset_status(fd, fd, ctx_bad, RS_BATCH_ACTIVE); igt_assert_eq(gem_reset_stats(fd, ctx_bad, &rs_bad), 0); igt_assert_eq(rs_bad.batch_active, active_count); assert_reset_status(fd, fd, ctx_good, RS_NO_ERROR); igt_assert_eq(gem_reset_stats(fd, ctx_good, &rs_good), 0); igt_assert_eq(rs_good.batch_active, 0); close(fd); } static void test_unrelated_ctx(const struct intel_execution_engine *e) { int fd1,fd2; int ctx_guilty, ctx_unrelated; fd1 = drm_open_driver(DRIVER_INTEL); fd2 = drm_open_driver(DRIVER_INTEL); assert_reset_status(0, fd1, 0, RS_NO_ERROR); assert_reset_status(1, fd2, 0, RS_NO_ERROR); ctx_guilty = gem_context_create(fd1); ctx_unrelated = gem_context_create(fd2); assert_reset_status(0, fd1, ctx_guilty, RS_NO_ERROR); assert_reset_status(1, fd2, ctx_unrelated, RS_NO_ERROR); inject_hang(fd1, ctx_guilty, e, 0); assert_reset_status(0, fd1, ctx_guilty, RS_BATCH_ACTIVE); assert_reset_status(1, fd2, ctx_unrelated, RS_NO_ERROR); gem_sync(fd2, noop(fd2, ctx_unrelated, e)); assert_reset_status(0, fd1, ctx_guilty, RS_BATCH_ACTIVE); assert_reset_status(1, fd2, ctx_unrelated, RS_NO_ERROR); close(fd1); close(fd2); } static int get_reset_count(int fd, int ctx) { int ret; struct local_drm_i915_reset_stats rs; ret = gem_reset_stats(fd, ctx, &rs); if (ret) return ret; return rs.reset_count; } static void test_close_pending_ctx(const struct intel_execution_engine *e) { int fd = drm_open_driver(DRIVER_INTEL); uint32_t ctx = gem_context_create(fd); assert_reset_status(fd, fd, ctx, RS_NO_ERROR); inject_hang(fd, ctx, e, 0); gem_context_destroy(fd, ctx); igt_assert_eq(__gem_context_destroy(fd, ctx), -ENOENT); close(fd); } static void test_close_pending(const struct intel_execution_engine *e) { int fd = drm_open_driver(DRIVER_INTEL); assert_reset_status(fd, fd, 0, RS_NO_ERROR); inject_hang(fd, 0, e, 0); close(fd); } static void noop_on_each_ring(int fd, const bool reverse) { const uint32_t bbe = MI_BATCH_BUFFER_END; struct drm_i915_gem_execbuffer2 eb; struct drm_i915_gem_exec_object2 obj; const struct intel_execution_engine *e; memset(&obj, 0, sizeof(obj)); obj.handle = gem_create(fd, 4096); gem_write(fd, obj.handle, 0, &bbe, sizeof(bbe)); memset(&eb, 0, sizeof(eb)); eb.buffers_ptr = to_user_pointer(&obj); eb.buffer_count = 1; if (reverse) { for (e = intel_execution_engines; e->name; e++) ; while (--e >= intel_execution_engines) { eb.flags = eb_ring(e); __gem_execbuf(fd, &eb); } } else { for (e = intel_execution_engines; e->name; e++) { eb.flags = eb_ring(e); __gem_execbuf(fd, &eb); } } gem_sync(fd, obj.handle); gem_close(fd, obj.handle); } static void test_close_pending_fork(const struct intel_execution_engine *e, const bool reverse) { int fd = drm_open_driver(DRIVER_INTEL); igt_hang_t hang; int pid; assert_reset_status(fd, fd, 0, RS_NO_ERROR); hang = igt_hang_ctx(fd, 0, eb_ring(e), 0); sleep(1); /* Avoid helpers as we need to kill the child * without any extra signal handling on behalf of * lib/drmtest.c */ pid = fork(); if (pid == 0) { const int fd2 = drm_open_driver(DRIVER_INTEL); igt_assert_lte(0, fd2); /* The crucial component is that we schedule the same noop batch * on each ring. This exercises batch_obj reference counting, * when gpu is reset and ring lists are cleared. */ noop_on_each_ring(fd2, reverse); close(fd2); pause(); exit(0); } else { igt_assert_lt(0, pid); sleep(1); /* Kill the child to reduce refcounts on batch_objs */ kill(pid, SIGKILL); } igt_post_hang_ring(fd, hang); close(fd); } static void test_reset_count(const struct intel_execution_engine *e, const bool create_ctx) { int fd = drm_open_driver(DRIVER_INTEL); int ctx; long c1, c2; if (create_ctx) ctx = gem_context_create(fd); else ctx = 0; assert_reset_status(fd, fd, ctx, RS_NO_ERROR); c1 = get_reset_count(fd, ctx); igt_assert(c1 >= 0); inject_hang(fd, ctx, e, 0); assert_reset_status(fd, fd, ctx, RS_BATCH_ACTIVE); c2 = get_reset_count(fd, ctx); igt_assert(c2 >= 0); igt_assert(c2 == (c1 + 1)); igt_fork(child, 1) { igt_drop_root(); c2 = get_reset_count(fd, ctx); igt_assert(c2 == 0); } igt_waitchildren(); if (create_ctx) gem_context_destroy(fd, ctx); close(fd); } static int _test_params(int fd, int ctx, uint32_t flags, uint32_t pad) { struct local_drm_i915_reset_stats rs; memset(&rs, 0, sizeof(rs)); rs.ctx_id = ctx; rs.flags = flags; rs.reset_count = rand(); rs.batch_active = rand(); rs.batch_pending = rand(); rs.pad = pad; if (drmIoctl(fd, GET_RESET_STATS_IOCTL, &rs)) return -errno; return 0; } typedef enum { root = 0, user } cap_t; static void _check_param_ctx(const int fd, const int ctx, const cap_t cap) { const uint32_t bad = rand() + 1; if (ctx == 0) { igt_assert_eq(_test_params(fd, ctx, 0, 0), 0); if (cap != root) { igt_assert(get_reset_count(fd, ctx) == 0); } } igt_assert_eq(_test_params(fd, ctx, 0, bad), -EINVAL); igt_assert_eq(_test_params(fd, ctx, bad, 0), -EINVAL); igt_assert_eq(_test_params(fd, ctx, bad, bad), -EINVAL); } static void check_params(const int fd, const int ctx, cap_t cap) { igt_assert(ioctl(fd, GET_RESET_STATS_IOCTL, 0) == -1); igt_assert_eq(_test_params(fd, 0xbadbad, 0, 0), -ENOENT); _check_param_ctx(fd, ctx, cap); } static void _test_param(const int fd, const int ctx) { check_params(fd, ctx, root); igt_fork(child, 1) { check_params(fd, ctx, root); igt_drop_root(); check_params(fd, ctx, user); } check_params(fd, ctx, root); igt_waitchildren(); } static void test_params_ctx(void) { int fd; fd = drm_open_driver(DRIVER_INTEL); _test_param(fd, gem_context_create(fd)); close(fd); } static void test_params(void) { int fd; fd = drm_open_driver(DRIVER_INTEL); _test_param(fd, 0); close(fd); } static const struct intel_execution_engine * next_engine(int fd, const struct intel_execution_engine *e) { do { e++; if (e->name == NULL) e = intel_execution_engines; if (e->exec_id == 0) e++; } while (!has_engine(fd, 0, e)); return e; } static void defer_hangcheck(const struct intel_execution_engine *engine) { const struct intel_execution_engine *next; int fd, count_start, count_end; int seconds = 30; fd = drm_open_driver(DRIVER_INTEL); next = next_engine(fd, engine); igt_skip_on(next == engine); count_start = get_reset_count(fd, 0); igt_assert_lte(0, count_start); inject_hang(fd, 0, engine, 0); while (--seconds) { noop(fd, 0, next); count_end = get_reset_count(fd, 0); igt_assert_lte(0, count_end); if (count_end > count_start) break; sleep(1); } igt_assert_lt(count_start, count_end); close(fd); } static bool gem_has_reset_stats(int fd) { struct local_drm_i915_reset_stats rs; int ret; /* Carefully set flags and pad to zero, otherwise we get -EINVAL */ memset(&rs, 0, sizeof(rs)); ret = drmIoctl(fd, GET_RESET_STATS_IOCTL, &rs); if (ret == 0) return true; /* If we get EPERM, we have support but did not have CAP_SYSADM */ if (ret == -1 && errno == EPERM) return true; return false; } #define RUN_TEST(...) do { sync_gpu(); __VA_ARGS__; sync_gpu(); } while (0) #define RUN_CTX_TEST(...) do { check_context(e); RUN_TEST(__VA_ARGS__); } while (0) igt_main { const struct intel_execution_engine *e; igt_fixture { int fd; bool has_reset_stats; bool using_full_reset; fd = drm_open_driver(DRIVER_INTEL); devid = intel_get_drm_devid(fd); has_reset_stats = gem_has_reset_stats(fd); igt_assert(igt_sysfs_set_parameter (fd, "reset", "%d", 1 /* only global reset */)); using_full_reset = !gem_engine_reset_enabled(fd) && gem_gpu_reset_enabled(fd); close(fd); igt_require_f(has_reset_stats, "No reset stats ioctl support. Too old kernel?\n"); igt_require_f(using_full_reset, "Full GPU reset is not enabled. Is enable_hangcheck set?\n"); } igt_subtest("params") test_params(); igt_subtest_f("params-ctx") RUN_TEST(test_params_ctx()); for (e = intel_execution_engines; e->name; e++) { igt_subtest_f("reset-stats-%s", e->name) RUN_TEST(test_rs(e, 4, 1, 0)); igt_subtest_f("reset-stats-ctx-%s", e->name) RUN_CTX_TEST(test_rs_ctx(e, 4, 4, 1, 2)); igt_subtest_f("ban-%s", e->name) RUN_TEST(test_ban(e)); igt_subtest_f("ban-ctx-%s", e->name) RUN_CTX_TEST(test_ban_ctx(e)); igt_subtest_f("reset-count-%s", e->name) RUN_TEST(test_reset_count(e, false)); igt_subtest_f("reset-count-ctx-%s", e->name) RUN_CTX_TEST(test_reset_count(e, true)); igt_subtest_f("unrelated-ctx-%s", e->name) RUN_CTX_TEST(test_unrelated_ctx(e)); igt_subtest_f("close-pending-%s", e->name) RUN_TEST(test_close_pending(e)); igt_subtest_f("close-pending-ctx-%s", e->name) RUN_CTX_TEST(test_close_pending_ctx(e)); igt_subtest_f("close-pending-fork-%s", e->name) RUN_TEST(test_close_pending_fork(e, false)); igt_subtest_f("close-pending-fork-reverse-%s", e->name) RUN_TEST(test_close_pending_fork(e, true)); igt_subtest_f("defer-hangcheck-%s", e->name) RUN_TEST(defer_hangcheck(e)); } igt_fixture { int fd; fd = drm_open_driver(DRIVER_INTEL); igt_assert(igt_sysfs_set_parameter (fd, "reset", "%d", INT_MAX /* any reset method */)); close(fd); } }