The GPU is trashing the low pages of its reserved memory upon reset. If we are using this memory for ringbuffers, then we will dutiful resubmit the trashed rings after the reset causing further resets, and worse. We must exclude this range from our own use. The value of 128KiB was found by empirical measurement on gen9.
Signed-off-by: Chris Wilson chris@chris-wilson.co.uk Cc: stable@vger.kernel.org --- v2 comes with a selftest to see how widespread the issue is --- drivers/gpu/drm/i915/Kconfig.debug | 1 + drivers/gpu/drm/i915/gem/i915_gem_stolen.c | 5 +- drivers/gpu/drm/i915/gt/selftest_reset.c | 141 +++++++++++++++++++++ 3 files changed, 145 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/Kconfig.debug b/drivers/gpu/drm/i915/Kconfig.debug index 206882e154bc..0fb7fd0ef717 100644 --- a/drivers/gpu/drm/i915/Kconfig.debug +++ b/drivers/gpu/drm/i915/Kconfig.debug @@ -162,6 +162,7 @@ config DRM_I915_SELFTEST select DRM_EXPORT_FOR_TESTS if m select FAULT_INJECTION select PRIME_NUMBERS + select CRC32 help Choose this option to allow the driver to perform selftests upon loading; also requires the i915.selftest=1 module parameter. To diff --git a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c index 0be5e8683337..c0cc2a972a11 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_stolen.c @@ -53,8 +53,9 @@ int i915_gem_stolen_insert_node(struct drm_i915_private *i915, struct drm_mm_node *node, u64 size, unsigned alignment) { - return i915_gem_stolen_insert_node_in_range(i915, node, size, - alignment, 0, U64_MAX); + return i915_gem_stolen_insert_node_in_range(i915, node, + size, alignment, + SZ_128K, U64_MAX); }
void i915_gem_stolen_remove_node(struct drm_i915_private *i915, diff --git a/drivers/gpu/drm/i915/gt/selftest_reset.c b/drivers/gpu/drm/i915/gt/selftest_reset.c index 35406ecdf0b2..f73f132e57c4 100644 --- a/drivers/gpu/drm/i915/gt/selftest_reset.c +++ b/drivers/gpu/drm/i915/gt/selftest_reset.c @@ -3,9 +3,149 @@ * Copyright © 2018 Intel Corporation */
+#include <linux/crc32.h> + +#include "i915_memcpy.h" #include "i915_selftest.h" #include "selftests/igt_reset.h" #include "selftests/igt_atomic.h" +#include "selftests/igt_spinner.h" + +static int igt_reset_stolen(void *arg) +{ + struct intel_gt *gt = arg; + struct i915_ggtt *ggtt = >->i915->ggtt; + struct resource *dsm = >->i915->dsm; + resource_size_t num_pages, page, max, count; + struct intel_engine_cs *engine; + intel_wakeref_t wakeref; + enum intel_engine_id id; + struct igt_spinner spin; + u32 seed, *crc; + void *tmp; + int err; + + if (!drm_mm_node_allocated(&ggtt->error_capture)) + return 0; + + num_pages = resource_size(dsm) >> PAGE_SHIFT; + if (!num_pages) + return 0; + + crc = kmalloc_array(num_pages, sizeof(u32), GFP_KERNEL); + if (!crc) + return -ENOMEM; + + tmp = kmalloc(PAGE_SIZE, GFP_KERNEL); + if (!tmp) { + err = -ENOMEM; + goto err_crc; + } + + igt_global_reset_lock(gt); + wakeref = intel_runtime_pm_get(gt->uncore->rpm); + + err = igt_spinner_init(&spin, gt); + if (err) + goto err_lock; + + for_each_engine(engine, gt, id) { + struct i915_request *rq; + + intel_engine_pm_get(engine); + + rq = igt_spinner_create_request(&spin, + engine->kernel_context, + MI_ARB_CHECK); + i915_request_add(rq); + + intel_engine_pm_put(engine); + } + + seed = 0; + for (page = 0; page < num_pages; page++) { + dma_addr_t dma; + void __iomem *s; + void *in; + + dma = (dma_addr_t)dsm->start + (page << PAGE_SHIFT); + + ggtt->vm.insert_page(&ggtt->vm, dma, + ggtt->error_capture.start, + I915_CACHE_NONE, 0); + mb(); + + s = io_mapping_map_wc(&ggtt->iomap, + ggtt->error_capture.start, + PAGE_SIZE); + + in = s; + if (i915_memcpy_from_wc(tmp, s, PAGE_SIZE)) + in = tmp; + seed = crc32_le(seed, in, PAGE_SIZE); + + io_mapping_unmap(s); + + crc[page] = seed; + } + mb(); + ggtt->vm.clear_range(&ggtt->vm, ggtt->error_capture.start, PAGE_SIZE); + + intel_gt_reset(gt, ALL_ENGINES, NULL); + + max = 0; + seed = 0; + count = 0; + for (page = 0; page < num_pages; page++) { + dma_addr_t dma; + void __iomem *s; + void *in; + + dma = (dma_addr_t)dsm->start + (page << PAGE_SHIFT); + + ggtt->vm.insert_page(&ggtt->vm, dma, + ggtt->error_capture.start, + I915_CACHE_NONE, 0); + mb(); + + s = io_mapping_map_wc(&ggtt->iomap, + ggtt->error_capture.start, + PAGE_SIZE); + + in = s; + if (i915_memcpy_from_wc(tmp, s, PAGE_SIZE)) + in = tmp; + seed = crc32_le(seed, in, PAGE_SIZE); + + io_mapping_unmap(s); + + if (seed != crc[page]) { + pr_info("stolen page %pa modified by GPU reset\n", + &page); + seed = crc[page]; + max = page; + count++; + } + } + mb(); + ggtt->vm.clear_range(&ggtt->vm, ggtt->error_capture.start, PAGE_SIZE); + + if (count > 0) { + pr_err("Reset clobbered %pa pages of stolen, last clobber at page %pa\n", &count, &max); + err = -EINVAL; + } + + igt_spinner_fini(&spin); + +err_lock: + intel_runtime_pm_put(gt->uncore->rpm, wakeref); + igt_global_reset_unlock(gt); + + kfree(tmp); +err_crc: + kfree(crc); + return err; +}
static int igt_global_reset(void *arg) { @@ -164,6 +304,7 @@ int intel_reset_live_selftests(struct drm_i915_private *i915) { static const struct i915_subtest tests[] = { SUBTEST(igt_global_reset), /* attempt to recover GPU first */ + SUBTEST(igt_reset_stolen), SUBTEST(igt_wedged_reset), SUBTEST(igt_atomic_reset), SUBTEST(igt_atomic_engine_reset),