Intel ID: PSIRT-TA-201910-001 CVEID: CVE-2019-14615
Summary of Vulnerability ------------------------ Insufficient control flow in certain data structures for some Intel(R) Processors with Intel Processor Graphics may allow an unauthenticated user to potentially enable information disclosure via local access
Products affected: ------------------ Intel CPU’s with Gen7, Gen7.5 and Gen9 Graphics.
Mitigation Summary ------------------ This patch provides mitigation for Gen7 and Gen7.5 hardware only. Patch for Gen9 devices have been provided and merged to Linux mainline, and backported to stable kernels. Note that Gen8 is not impacted due to a previously implemented workaround.
The mitigation involves submitting a custom EU kernel prior to every context restore, in order to forcibly clear down residual EU and URB resources.
This security mitigation change does not trigger any known performance regression. Performance is on par with current mainline/drm-tip.
Note on Address Space Isolation (Full PPGTT) --------------------------------------------
Isolation of EU kernel assets should be considered complementary to the existing support for address space isolation (aka Full PPGTT), since without address space isolation there is minimal value in preventing leakage between EU contexts. Full PPGTT has long been supported on Gen Gfx devices since Gen8, and protection against EU residual leakage is a welcome addition for these newer platforms.
By contrast, Gen7 and Gen7.5 device introduced Full PPGTT support only as a hardware development feature for anticipated Gen8 productization. Support was never intended for, or provided to the Linux kernels for these platforms. Recent work (still ongoing) to the mainline kernel is retroactively providing this support, but due to the level of complexity it is not practical to attempt to backport this to earlier stable kernels. Since without Full PPGTT, EU residuals protection has questionable benefit, *there are no plans to provide stable kernel backports for this patch series.*
Mika Kuoppala (1): drm/i915: Add mechanism to submit a context WA on ring submission
Prathap Kumar Valsan (1): drm/i915/gen7: Clear all EU/L3 residual contexts
drivers/gpu/drm/i915/Makefile | 1 + drivers/gpu/drm/i915/gt/gen7_renderclear.c | 535 ++++++++++++++++++ drivers/gpu/drm/i915/gt/gen7_renderclear.h | 15 + drivers/gpu/drm/i915/gt/intel_gpu_commands.h | 17 +- .../gpu/drm/i915/gt/intel_ring_submission.c | 133 ++++- drivers/gpu/drm/i915/i915_utils.h | 5 + 6 files changed, 700 insertions(+), 6 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/gen7_renderclear.c create mode 100644 drivers/gpu/drm/i915/gt/gen7_renderclear.h
From: Mika Kuoppala mika.kuoppala@linux.intel.com
This patch adds framework to submit an arbitrary batchbuffer on each context switch to clear residual state for render engine on Gen7/7.5 devices.
The idea of always emitting the context and vm setup around each request is primary to make reset recovery easy, and not require rewriting the ringbuffer. As each request would set up its own context, leaving it to the HW to notice and elide no-op context switches, we could restart the ring at any point, and reorder the requests freely.
However, to avoid emitting clear_residuals() between consecutive requests in the ringbuffer of the same context, we do want to track the current context in the ring. In doing so, we need to be careful to only record a context switch when we are sure the next request will be emitted.
This security mitigation change does not trigger any performance regression. Performance is on par with current mainline/drm-tip.
Signed-off-by: Mika Kuoppala mika.kuoppala@linux.intel.com Signed-off-by: Prathap Kumar Valsan prathap.kumar.valsan@intel.com Signed-off-by: Akeem G Abodunrin akeem.g.abodunrin@intel.com Cc: Chris Wilson chris@chris-wilson.co.uk Cc: Balestrieri Francesco francesco.balestrieri@intel.com Cc: Bloomfield Jon jon.bloomfield@intel.com Cc: Dutt Sudeep sudeep.dutt@intel.com --- .../gpu/drm/i915/gt/intel_ring_submission.c | 132 +++++++++++++++++- 1 file changed, 129 insertions(+), 3 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_ring_submission.c b/drivers/gpu/drm/i915/gt/intel_ring_submission.c index 9aa86ba15ce7..9f3bfe499446 100644 --- a/drivers/gpu/drm/i915/gt/intel_ring_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_ring_submission.c @@ -1385,7 +1385,9 @@ static int load_pd_dir(struct i915_request *rq, return rq->engine->emit_flush(rq, EMIT_FLUSH); }
-static inline int mi_set_context(struct i915_request *rq, u32 flags) +static inline int mi_set_context(struct i915_request *rq, + struct intel_context *ce, + u32 flags) { struct drm_i915_private *i915 = rq->i915; struct intel_engine_cs *engine = rq->engine; @@ -1460,7 +1462,7 @@ static inline int mi_set_context(struct i915_request *rq, u32 flags)
*cs++ = MI_NOOP; *cs++ = MI_SET_CONTEXT; - *cs++ = i915_ggtt_offset(rq->context->state) | flags; + *cs++ = i915_ggtt_offset(ce->state) | flags; /* * w/a: MI_SET_CONTEXT must always be followed by MI_NOOP * WaMiSetContext_Hang:snb,ivb,vlv @@ -1575,13 +1577,56 @@ static int switch_mm(struct i915_request *rq, struct i915_address_space *vm) return rq->engine->emit_flush(rq, EMIT_INVALIDATE); }
+static int clear_residuals(struct i915_request *rq) +{ + struct intel_engine_cs *engine = rq->engine; + int ret; + + GEM_BUG_ON(!engine->kernel_context->state); + + ret = switch_mm(rq, vm_alias(engine->kernel_context)); + if (ret) + return ret; + + ret = mi_set_context(rq, + engine->kernel_context, + MI_MM_SPACE_GTT | MI_RESTORE_INHIBIT); + if (ret) + return ret; + + ret = engine->emit_bb_start(rq, + engine->wa_ctx.vma->node.start, 0, + 0); + if (ret) + return ret; + + ret = engine->emit_flush(rq, EMIT_FLUSH); + if (ret) + return ret; + + /* Always invalidate before the next switch_mm() */ + return engine->emit_flush(rq, EMIT_INVALIDATE); +} + static int switch_context(struct i915_request *rq) { + struct intel_engine_cs *engine = rq->engine; struct intel_context *ce = rq->context; + void **residuals = NULL; int ret;
GEM_BUG_ON(HAS_EXECLISTS(rq->i915));
+ if (engine->wa_ctx.vma && ce != engine->kernel_context) { + if (engine->wa_ctx.vma->private != ce) { + ret = clear_residuals(rq); + if (ret) + return ret; + + residuals = &engine->wa_ctx.vma->private; + } + } + ret = switch_mm(rq, vm_alias(ce)); if (ret) return ret; @@ -1601,7 +1646,7 @@ static int switch_context(struct i915_request *rq) else flags |= MI_RESTORE_INHIBIT;
- ret = mi_set_context(rq, flags); + ret = mi_set_context(rq, ce, flags); if (ret) return ret; } @@ -1610,6 +1655,20 @@ static int switch_context(struct i915_request *rq) if (ret) return ret;
+ /* + * Now past the point of no return, this request _will_ be emitted. + * + * Or at least this preamble will be emitted, the request may be + * interrupted prior to submitting the user payload. If so, we + * still submit the "empty" request in order to preserve global + * state tracking such as this, our tracking of the current + * dirty context. + */ + if (residuals) { + intel_context_put(*residuals); + *residuals = intel_context_get(ce); + } + return 0; }
@@ -1794,6 +1853,11 @@ static void ring_release(struct intel_engine_cs *engine)
intel_engine_cleanup_common(engine);
+ if (engine->wa_ctx.vma) { + intel_context_put(engine->wa_ctx.vma->private); + i915_vma_unpin_and_release(&engine->wa_ctx.vma, 0); + } + intel_ring_unpin(engine->legacy.ring); intel_ring_put(engine->legacy.ring);
@@ -1941,6 +2005,60 @@ static void setup_vecs(struct intel_engine_cs *engine) engine->emit_fini_breadcrumb = gen7_xcs_emit_breadcrumb; }
+static int gen7_ctx_switch_bb_setup(struct intel_engine_cs * const engine, + struct i915_vma * const vma) +{ + return 0; +} + +static int gen7_ctx_switch_bb_init(struct intel_engine_cs *engine) +{ + struct drm_i915_gem_object *obj; + struct i915_vma *vma; + int size; + int err; + + size = gen7_ctx_switch_bb_setup(engine, NULL /* probe size */); + if (size <= 0) + return size; + + size = ALIGN(size, PAGE_SIZE); + obj = i915_gem_object_create_internal(engine->i915, size); + if (IS_ERR(obj)) + return PTR_ERR(obj); + + vma = i915_vma_instance(obj, engine->gt->vm, NULL); + if (IS_ERR(vma)) { + err = PTR_ERR(vma); + goto err_obj; + } + + vma->private = intel_context_create(engine); /* dummy residuals */ + if (IS_ERR(vma->private)) { + err = PTR_ERR(vma->private); + goto err_obj; + } + + err = i915_vma_pin(vma, 0, 0, PIN_USER | PIN_HIGH); + if (err) + goto err_private; + + err = gen7_ctx_switch_bb_setup(engine, vma); + if (err) + goto err_unpin; + + engine->wa_ctx.vma = vma; + return 0; + +err_unpin: + i915_vma_unpin(vma); +err_private: + intel_context_put(vma->private); +err_obj: + i915_gem_object_put(obj); + return err; +} + int intel_ring_submission_setup(struct intel_engine_cs *engine) { struct intel_timeline *timeline; @@ -1994,11 +2112,19 @@ int intel_ring_submission_setup(struct intel_engine_cs *engine)
GEM_BUG_ON(timeline->hwsp_ggtt != engine->status_page.vma);
+ if (IS_GEN(engine->i915, 7) && engine->class == RENDER_CLASS) { + err = gen7_ctx_switch_bb_init(engine); + if (err) + goto err_ring_unpin; + } + /* Finally, take ownership and responsibility for cleanup! */ engine->release = ring_release;
return 0;
+err_ring_unpin: + intel_ring_unpin(ring); err_ring: intel_ring_put(ring); err_timeline_unpin:
From: Prathap Kumar Valsan prathap.kumar.valsan@intel.com
On gen7 and gen7.5 devices, there could be leftover data residuals in EU/L3 from the retiring context. This patch introduces workaround to clear that residual contexts, by submitting a batch buffer with dedicated HW context to the GPU with ring allocation for each context switching.
This security mitigation change does not trigger any performance regression. Performance is on par with current mainline/drm-tip.
Signed-off-by: Mika Kuoppala mika.kuoppala@linux.intel.com Signed-off-by: Prathap Kumar Valsan prathap.kumar.valsan@intel.com Signed-off-by: Akeem G Abodunrin akeem.g.abodunrin@intel.com Cc: Chris Wilson chris.p.wilson@intel.com Cc: Balestrieri Francesco francesco.balestrieri@intel.com Cc: Bloomfield Jon jon.bloomfield@intel.com Cc: Dutt Sudeep sudeep.dutt@intel.com --- drivers/gpu/drm/i915/Makefile | 1 + drivers/gpu/drm/i915/gt/gen7_renderclear.c | 535 ++++++++++++++++++ drivers/gpu/drm/i915/gt/gen7_renderclear.h | 15 + drivers/gpu/drm/i915/gt/intel_gpu_commands.h | 17 +- .../gpu/drm/i915/gt/intel_ring_submission.c | 3 +- drivers/gpu/drm/i915/i915_utils.h | 5 + 6 files changed, 572 insertions(+), 4 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/gen7_renderclear.c create mode 100644 drivers/gpu/drm/i915/gt/gen7_renderclear.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 3c88d7d8c764..f96bae664a03 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -78,6 +78,7 @@ gt-y += \ gt/debugfs_gt.o \ gt/debugfs_gt_pm.o \ gt/gen6_ppgtt.o \ + gt/gen7_renderclear.o \ gt/gen8_ppgtt.o \ gt/intel_breadcrumbs.o \ gt/intel_context.o \ diff --git a/drivers/gpu/drm/i915/gt/gen7_renderclear.c b/drivers/gpu/drm/i915/gt/gen7_renderclear.c new file mode 100644 index 000000000000..a6f5f1602e33 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/gen7_renderclear.c @@ -0,0 +1,535 @@ +// SPDX-License-Identifier: MIT +/* + * Copyright © 2019 Intel Corporation + */ + +#include "gen7_renderclear.h" +#include "i915_drv.h" +#include "i915_utils.h" +#include "intel_gpu_commands.h" + +#define MAX_URB_ENTRIES 64 +#define STATE_SIZE (4 * 1024) +#define GT3_INLINE_DATA_DELAYS 0x1E00 +#define batch_advance(Y, CS) GEM_BUG_ON((Y)->end != (CS)) + +/* + * Media CB Kernel for gen7 devices + * TODO: Add comments to kernel, indicating what each array of hex does or + * include header file, which has assembly source and support in igt to be + * able to generate kernel in this same format + */ +static const u32 cb7_kernel[][4] = { + { 0x00000001, 0x26020128, 0x00000024, 0x00000000 }, + { 0x00000040, 0x20280c21, 0x00000028, 0x00000001 }, + { 0x01000010, 0x20000c20, 0x0000002c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000002c }, + { 0x00600001, 0x20600061, 0x00000000, 0x00000000 }, + { 0x00000008, 0x20601c85, 0x00000e00, 0x0000000c }, + { 0x00000005, 0x20601ca5, 0x00000060, 0x00000001 }, + { 0x00000008, 0x20641c85, 0x00000e00, 0x0000000d }, + { 0x00000005, 0x20641ca5, 0x00000064, 0x00000003 }, + { 0x00000041, 0x207424a5, 0x00000064, 0x00000034 }, + { 0x00000040, 0x206014a5, 0x00000060, 0x00000074 }, + { 0x00000008, 0x20681c85, 0x00000e00, 0x00000008 }, + { 0x00000005, 0x20681ca5, 0x00000068, 0x0000000f }, + { 0x00000041, 0x20701ca5, 0x00000060, 0x00000010 }, + { 0x00000040, 0x206814a5, 0x00000068, 0x00000070 }, + { 0x00600001, 0x20a00061, 0x00000000, 0x00000000 }, + { 0x00000005, 0x206c1c85, 0x00000e00, 0x00000007 }, + { 0x00000041, 0x206c1ca5, 0x0000006c, 0x00000004 }, + { 0x00600001, 0x20800021, 0x008d0000, 0x00000000 }, + { 0x00000001, 0x20800021, 0x0000006c, 0x00000000 }, + { 0x00000001, 0x20840021, 0x00000068, 0x00000000 }, + { 0x00000001, 0x20880061, 0x00000000, 0x00000003 }, + { 0x00000005, 0x208c0d21, 0x00000086, 0xffffffff }, + { 0x05600032, 0x20a01fa1, 0x008d0080, 0x02190001 }, + { 0x00000040, 0x20a01ca5, 0x000000a0, 0x00000001 }, + { 0x05600032, 0x20a01fa1, 0x008d0080, 0x040a8001 }, + { 0x02000040, 0x20281c21, 0x00000028, 0xffffffff }, + { 0x00010220, 0x34001c00, 0x00001400, 0xfffffffc }, + { 0x00000001, 0x26020128, 0x00000024, 0x00000000 }, + { 0x00000001, 0x220000e4, 0x00000000, 0x00000000 }, + { 0x00000001, 0x220801ec, 0x00000000, 0x007f007f }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00600001, 0x2fe00021, 0x008d0000, 0x00000000 }, + { 0x00200001, 0x20400121, 0x00450020, 0x00000000 }, + { 0x00000001, 0x20480061, 0x00000000, 0x000f000f }, + { 0x00000005, 0x204c0d21, 0x00000046, 0xffffffef }, + { 0x00800001, 0x20600061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20800061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20a00061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20c00061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20e00061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x21000061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x21200061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x21400061, 0x00000000, 0x00000000 }, + { 0x05600032, 0x20001fa0, 0x008d0040, 0x120a8000 }, + { 0x00000040, 0x20402d21, 0x00000020, 0x00100010 }, + { 0x05600032, 0x20001fa0, 0x008d0040, 0x120a8000 }, + { 0x02000040, 0x22083d8c, 0x00000208, 0xffffffff }, + { 0x00800001, 0xa0000109, 0x00000602, 0x00000000 }, + { 0x00000040, 0x22001c84, 0x00000200, 0x00000020 }, + { 0x00010220, 0x34001c00, 0x00001400, 0xfffffff8 }, + { 0x07600032, 0x20001fa0, 0x008d0fe0, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, +}; + +/* + * Media CB Kernel for gen7.5 devices + * TODO: Add comments to kernel, indicating what each array of hex does or + * include header file, which has assembly source and support in igt to be + * able to generate kernel in this same format + */ +static const u32 cb75_kernel[][4] = { + { 0x00000001, 0x26020128, 0x00000024, 0x00000000 }, + { 0x00000040, 0x20280c21, 0x00000028, 0x00000001 }, + { 0x01000010, 0x20000c20, 0x0000002c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000160 }, + { 0x00600001, 0x20600061, 0x00000000, 0x00000000 }, + { 0x00000008, 0x20601c85, 0x00000e00, 0x0000000c }, + { 0x00000005, 0x20601ca5, 0x00000060, 0x00000001 }, + { 0x00000008, 0x20641c85, 0x00000e00, 0x0000000d }, + { 0x00000005, 0x20641ca5, 0x00000064, 0x00000003 }, + { 0x00000041, 0x207424a5, 0x00000064, 0x00000034 }, + { 0x00000040, 0x206014a5, 0x00000060, 0x00000074 }, + { 0x00000008, 0x20681c85, 0x00000e00, 0x00000008 }, + { 0x00000005, 0x20681ca5, 0x00000068, 0x0000000f }, + { 0x00000041, 0x20701ca5, 0x00000060, 0x00000010 }, + { 0x00000040, 0x206814a5, 0x00000068, 0x00000070 }, + { 0x00600001, 0x20a00061, 0x00000000, 0x00000000 }, + { 0x00000005, 0x206c1c85, 0x00000e00, 0x00000007 }, + { 0x00000041, 0x206c1ca5, 0x0000006c, 0x00000004 }, + { 0x00600001, 0x20800021, 0x008d0000, 0x00000000 }, + { 0x00000001, 0x20800021, 0x0000006c, 0x00000000 }, + { 0x00000001, 0x20840021, 0x00000068, 0x00000000 }, + { 0x00000001, 0x20880061, 0x00000000, 0x00000003 }, + { 0x00000005, 0x208c0d21, 0x00000086, 0xffffffff }, + { 0x05600032, 0x20a01fa1, 0x008d0080, 0x02190001 }, + { 0x00000040, 0x20a01ca5, 0x000000a0, 0x00000001 }, + { 0x05600032, 0x20a01fa1, 0x008d0080, 0x040a8001 }, + { 0x02000040, 0x20281c21, 0x00000028, 0xffffffff }, + { 0x00010220, 0x34001c00, 0x00001400, 0xffffffe0 }, + { 0x00000001, 0x26020128, 0x00000024, 0x00000000 }, + { 0x00000001, 0x220000e4, 0x00000000, 0x00000000 }, + { 0x00000001, 0x220801ec, 0x00000000, 0x007f007f }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00600001, 0x2fe00021, 0x008d0000, 0x00000000 }, + { 0x00200001, 0x20400121, 0x00450020, 0x00000000 }, + { 0x00000001, 0x20480061, 0x00000000, 0x000f000f }, + { 0x00000005, 0x204c0d21, 0x00000046, 0xffffffef }, + { 0x00800001, 0x20600061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20800061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20a00061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20c00061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x20e00061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x21000061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x21200061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x21400061, 0x00000000, 0x00000000 }, + { 0x05600032, 0x20001fa0, 0x008d0040, 0x120a8000 }, + { 0x00000040, 0x20402d21, 0x00000020, 0x00100010 }, + { 0x05600032, 0x20001fa0, 0x008d0040, 0x120a8000 }, + { 0x02000040, 0x22083d8c, 0x00000208, 0xffffffff }, + { 0x00800001, 0xa0000109, 0x00000602, 0x00000000 }, + { 0x00000040, 0x22001c84, 0x00000200, 0x00000020 }, + { 0x00010220, 0x34001c00, 0x00001400, 0xffffffc0 }, + { 0x07600032, 0x20001fa0, 0x008d0fe0, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, +}; + +struct cb_kernel { + const void *data; + u32 size; +}; + +#define CB_KERNEL(name) { .data = (name), .size = sizeof(name) } + +static const struct cb_kernel cb_kernel_gen7 = CB_KERNEL(cb7_kernel); +static const struct cb_kernel cb_kernel_hsw = CB_KERNEL(cb75_kernel); + +struct batch_chunk { + struct i915_vma *vma; + u32 offset; + u32 *start; + u32 *end; + u32 max_items; +}; + +struct batch_vals { + u32 max_primitives; + u32 max_urb_entries; + u32 cmd_size; + u32 state_size; + u32 state_start; + u32 batch_size; + u32 surface_height; + u32 surface_width; + u32 scratch_size; + u32 max_size; +}; + +static void +batch_get_defaults(struct drm_i915_private *i915, struct batch_vals *bv) +{ + if (IS_HASWELL(i915)) { + bv->max_primitives = 280; + bv->max_urb_entries = MAX_URB_ENTRIES; + bv->surface_height = 16 * 16; + bv->surface_width = 32 * 2 * 16; + } else { + bv->max_primitives = 128; + bv->max_urb_entries = MAX_URB_ENTRIES / 2; + bv->surface_height = 16 * 8; + bv->surface_width = 32 * 16; + } + bv->cmd_size = bv->max_primitives * 4096; + bv->state_size = STATE_SIZE; + bv->state_start = bv->cmd_size; + bv->batch_size = bv->cmd_size + bv->state_size; + bv->scratch_size = bv->surface_height * bv->surface_width; + bv->max_size = bv->batch_size + bv->scratch_size; +} + +static void batch_init(struct batch_chunk *bc, + struct i915_vma *vma, + u32 *start, u32 offset, u32 max_bytes) +{ + bc->vma = vma; + bc->offset = offset; + bc->start = start + bc->offset / sizeof(*bc->start); + bc->end = bc->start; + bc->max_items = max_bytes / sizeof(*bc->start); +} + +static u32 batch_offset(const struct batch_chunk *bc, u32 *cs) +{ + return (cs - bc->start) * sizeof(*bc->start) + bc->offset; +} + +static u32 batch_addr(const struct batch_chunk *bc) +{ + return bc->vma->node.start; +} + +static void batch_add(struct batch_chunk *bc, const u32 d) +{ + GEM_DEBUG_WARN_ON((bc->end - bc->start) >= bc->max_items); + *bc->end++ = d; +} + +static u32 *batch_alloc_items(struct batch_chunk *bc, u32 align, u32 items) +{ + u32 *map; + + if (align) { + u32 *end = ptr_align(bc->end, align); + + memset32(bc->end, 0, end - bc->end); + bc->end = end; + } + + map = bc->end; + bc->end += items; + + return map; +} + +static u32 *batch_alloc_bytes(struct batch_chunk *bc, u32 align, u32 bytes) +{ + GEM_BUG_ON(!IS_ALIGNED(bytes, sizeof(*bc->start))); + return batch_alloc_items(bc, align, bytes / sizeof(*bc->start)); +} + +static u32 +gen7_fill_surface_state(struct batch_chunk *state, + const u32 dst_offset, + const struct batch_vals *bv) +{ + u32 surface_h = bv->surface_height; + u32 surface_w = bv->surface_width; + u32 *cs = batch_alloc_items(state, 32, 8); + u32 offset = batch_offset(state, cs); + +#define SURFACE_2D 1 +#define SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 +#define RENDER_CACHE_READ_WRITE 1 + + *cs++ = SURFACE_2D << 29 | + (SURFACEFORMAT_B8G8R8A8_UNORM << 18) | + (RENDER_CACHE_READ_WRITE << 8); + + *cs++ = batch_addr(state) + dst_offset; + + *cs++ = ((surface_h / 4 - 1) << 16) | (surface_w / 4 - 1); + *cs++ = surface_w; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; +#define SHADER_CHANNELS(r, g, b, a) \ + (((r) << 25) | ((g) << 22) | ((b) << 19) | ((a) << 16)) + *cs++ = SHADER_CHANNELS(4, 5, 6, 7); + batch_advance(state, cs); + + return offset; +} + +static u32 +gen7_fill_binding_table(struct batch_chunk *state, + const struct batch_vals *bv) +{ + u32 *cs = batch_alloc_items(state, 32, 8); + u32 offset = batch_offset(state, cs); + u32 surface_start; + + surface_start = gen7_fill_surface_state(state, bv->batch_size, bv); + *cs++ = surface_start - state->offset; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + batch_advance(state, cs); + + return offset; +} + +static u32 +gen7_fill_kernel_data(struct batch_chunk *state, + const u32 *data, + const u32 size) +{ + return batch_offset(state, + memcpy(batch_alloc_bytes(state, 64, size), + data, size)); +} + +static u32 +gen7_fill_interface_descriptor(struct batch_chunk *state, + const struct batch_vals *bv, + const struct cb_kernel *kernel, + unsigned int count) +{ + u32 *cs = batch_alloc_items(state, 32, 8 * count); + u32 offset = batch_offset(state, cs); + + *cs++ = gen7_fill_kernel_data(state, kernel->data, kernel->size); + *cs++ = (1 << 7) | (1 << 13); + *cs++ = 0; + *cs++ = (gen7_fill_binding_table(state, bv) - state->offset) | 1; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + batch_advance(state, cs); + + /* 1 - 63dummy idds */ + memset32(cs, 0x00, (count - 1) * 8); + + return offset; +} + +static void +gen7_emit_state_base_address(struct batch_chunk *batch, + u32 surface_state_base) +{ + u32 *cs = batch_alloc_items(batch, 0, 12); + + *cs++ = STATE_BASE_ADDRESS | (12 - 2); + /* general */ + *cs++ = batch_addr(batch) | BASE_ADDRESS_MODIFY; + /* surface */ + *cs++ = batch_addr(batch) | surface_state_base | BASE_ADDRESS_MODIFY; + /* dynamic */ + *cs++ = batch_addr(batch) | BASE_ADDRESS_MODIFY; + /* indirect */ + *cs++ = batch_addr(batch) | BASE_ADDRESS_MODIFY; + /* instruction */ + *cs++ = batch_addr(batch) | BASE_ADDRESS_MODIFY; + + /* general/dynamic/indirect/instruction access Bound */ + *cs++ = 0; + *cs++ = BASE_ADDRESS_MODIFY; + *cs++ = 0; + *cs++ = BASE_ADDRESS_MODIFY; + *cs++ = 0; + *cs++ = 0; + batch_advance(batch, cs); +} + +static void +gen7_emit_vfe_state(struct batch_chunk *batch, + const struct batch_vals *bv, + u32 urb_size, u32 curbe_size, + u32 mode) +{ + u32 urb_entries = bv->max_urb_entries; + u32 threads = bv->max_primitives - 1; + u32 *cs = batch_alloc_items(batch, 32, 8); + + *cs++ = MEDIA_VFE_STATE | (8 - 2); + + /* scratch buffer */ + *cs++ = 0; + + /* number of threads & urb entries for GPGPU vs Media Mode */ + *cs++ = threads << 16 | urb_entries << 8 | mode << 2; + + *cs++ = 0; + + /* urb entry size & curbe size in 256 bits unit */ + *cs++ = urb_size << 16 | curbe_size; + + /* scoreboard */ + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + batch_advance(batch, cs); +} + +static void +gen7_emit_interface_descriptor_load(struct batch_chunk *batch, + const u32 interface_descriptor, + unsigned int count) +{ + u32 *cs = batch_alloc_items(batch, 8, 4); + + *cs++ = MEDIA_INTERFACE_DESCRIPTOR_LOAD | (4 - 2); + *cs++ = 0; + *cs++ = count * 8 * sizeof(*cs); + + /* + * interface descriptor address - it is relative to the dynamics base + * address + */ + *cs++ = interface_descriptor; + batch_advance(batch, cs); +} + +static void +gen7_emit_media_object(struct batch_chunk *batch, + unsigned int media_object_index) +{ + unsigned int x_offset = (media_object_index % 16) * 64; + unsigned int y_offset = (media_object_index / 16) * 16; + unsigned int inline_data_size; + unsigned int media_batch_size; + unsigned int i; + u32 *cs; + + inline_data_size = 112 * 8; + media_batch_size = inline_data_size + 6; + + cs = batch_alloc_items(batch, 8, media_batch_size); + + *cs++ = MEDIA_OBJECT | (media_batch_size - 2); + + /* interface descriptor offset */ + *cs++ = 0; + + /* without indirect data */ + *cs++ = 0; + *cs++ = 0; + + /* scoreboard */ + *cs++ = 0; + *cs++ = 0; + + /* inline */ + *cs++ = (y_offset << 16) | (x_offset); + *cs++ = 0; + *cs++ = GT3_INLINE_DATA_DELAYS; + for (i = 3; i < inline_data_size; i++) + *cs++ = 0; + + batch_advance(batch, cs); +} + +static void gen7_emit_pipeline_flush(struct batch_chunk *batch) +{ + u32 *cs = batch_alloc_items(batch, 0, 5); + + *cs++ = GFX_OP_PIPE_CONTROL(5); + *cs++ = PIPE_CONTROL_STATE_CACHE_INVALIDATE | + PIPE_CONTROL_GLOBAL_GTT_IVB; + *cs++ = 0; + *cs++ = 0; + *cs++ = 0; + batch_advance(batch, cs); +} + +static void emit_batch(struct i915_vma * const vma, + u32 *start, + const struct batch_vals *bv) +{ + struct drm_i915_private *i915 = vma->vm->i915; + unsigned int desc_count = 64; + const u32 urb_size = 112; + struct batch_chunk cmds, state; + u32 interface_descriptor; + unsigned int i; + + batch_init(&cmds, vma, start, 0, bv->cmd_size); + batch_init(&state, vma, start, bv->state_start, bv->state_size); + + interface_descriptor = + gen7_fill_interface_descriptor(&state, bv, + IS_HASWELL(i915) ? + &cb_kernel_hsw : &cb_kernel_gen7, + desc_count); + gen7_emit_pipeline_flush(&cmds); + batch_add(&cmds, PIPELINE_SELECT | PIPELINE_SELECT_MEDIA); + batch_add(&cmds, MI_NOOP); + gen7_emit_state_base_address(&cmds, interface_descriptor); + gen7_emit_pipeline_flush(&cmds); + + gen7_emit_vfe_state(&cmds, bv, urb_size - 1, 0, 0); + + gen7_emit_interface_descriptor_load(&cmds, + interface_descriptor, + desc_count); + + for (i = 0; i < bv->max_primitives; i++) + gen7_emit_media_object(&cmds, i); + + batch_add(&cmds, MI_BATCH_BUFFER_END); +} + +int gen7_setup_clear_gpr_bb(struct intel_engine_cs * const engine, + struct i915_vma * const vma) +{ + struct batch_vals bv; + u32 *batch; + + batch_get_defaults(engine->i915, &bv); + if (!vma) + return bv.max_size; + + GEM_BUG_ON(vma->obj->base.size < bv.max_size); + + batch = i915_gem_object_pin_map(vma->obj, I915_MAP_WC); + if (IS_ERR(batch)) + return PTR_ERR(batch); + + emit_batch(vma, memset(batch, 0, bv.max_size), &bv); + + i915_gem_object_flush_map(vma->obj); + i915_gem_object_unpin_map(vma->obj); + + return 0; +} diff --git a/drivers/gpu/drm/i915/gt/gen7_renderclear.h b/drivers/gpu/drm/i915/gt/gen7_renderclear.h new file mode 100644 index 000000000000..bb100748e2c6 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/gen7_renderclear.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: MIT */ +/* + * Copyright © 2019 Intel Corporation + */ + +#ifndef __GEN7_RENDERCLEAR_H__ +#define __GEN7_RENDERCLEAR_H__ + +struct intel_engine_cs; +struct i915_vma; + +int gen7_setup_clear_gpr_bb(struct intel_engine_cs * const engine, + struct i915_vma * const vma); + +#endif /* __GEN7_RENDERCLEAR_H__ */ diff --git a/drivers/gpu/drm/i915/gt/intel_gpu_commands.h b/drivers/gpu/drm/i915/gt/intel_gpu_commands.h index 51b8718513bc..f04214a54f75 100644 --- a/drivers/gpu/drm/i915/gt/intel_gpu_commands.h +++ b/drivers/gpu/drm/i915/gt/intel_gpu_commands.h @@ -292,10 +292,21 @@ #define MI_STORE_URB_MEM MI_INSTR(0x2D, 0) #define MI_CONDITIONAL_BATCH_BUFFER_END MI_INSTR(0x36, 0)
-#define PIPELINE_SELECT ((0x3<<29)|(0x1<<27)|(0x1<<24)|(0x4<<16)) -#define GFX_OP_3DSTATE_VF_STATISTICS ((0x3<<29)|(0x1<<27)|(0x0<<24)|(0xB<<16)) -#define MEDIA_VFE_STATE ((0x3<<29)|(0x2<<27)|(0x0<<24)|(0x0<<16)) +#define STATE_BASE_ADDRESS \ + ((0x3 << 29) | (0x0 << 27) | (0x1 << 24) | (0x1 << 16)) +#define BASE_ADDRESS_MODIFY REG_BIT(0) +#define PIPELINE_SELECT \ + ((0x3 << 29) | (0x1 << 27) | (0x1 << 24) | (0x4 << 16)) +#define PIPELINE_SELECT_MEDIA REG_BIT(0) +#define GFX_OP_3DSTATE_VF_STATISTICS \ + ((0x3 << 29) | (0x1 << 27) | (0x0 << 24) | (0xB << 16)) +#define MEDIA_VFE_STATE \ + ((0x3 << 29) | (0x2 << 27) | (0x0 << 24) | (0x0 << 16)) #define MEDIA_VFE_STATE_MMIO_ACCESS_MASK (0x18) +#define MEDIA_INTERFACE_DESCRIPTOR_LOAD \ + ((0x3 << 29) | (0x2 << 27) | (0x0 << 24) | (0x2 << 16)) +#define MEDIA_OBJECT \ + ((0x3 << 29) | (0x2 << 27) | (0x1 << 24) | (0x0 << 16)) #define GPGPU_OBJECT ((0x3<<29)|(0x2<<27)|(0x1<<24)|(0x4<<16)) #define GPGPU_WALKER ((0x3<<29)|(0x2<<27)|(0x1<<24)|(0x5<<16)) #define GFX_OP_3DSTATE_DX9_CONSTANTF_VS \ diff --git a/drivers/gpu/drm/i915/gt/intel_ring_submission.c b/drivers/gpu/drm/i915/gt/intel_ring_submission.c index 9f3bfe499446..dc0632ea9fde 100644 --- a/drivers/gpu/drm/i915/gt/intel_ring_submission.c +++ b/drivers/gpu/drm/i915/gt/intel_ring_submission.c @@ -34,6 +34,7 @@ #include "gem/i915_gem_context.h"
#include "gen6_ppgtt.h" +#include "gen7_renderclear.h" #include "i915_drv.h" #include "i915_trace.h" #include "intel_context.h" @@ -2008,7 +2009,7 @@ static void setup_vecs(struct intel_engine_cs *engine) static int gen7_ctx_switch_bb_setup(struct intel_engine_cs * const engine, struct i915_vma * const vma) { - return 0; + return gen7_setup_clear_gpr_bb(engine, vma); }
static int gen7_ctx_switch_bb_init(struct intel_engine_cs *engine) diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h index b0ade76bec90..7ac5b3565845 100644 --- a/drivers/gpu/drm/i915/i915_utils.h +++ b/drivers/gpu/drm/i915/i915_utils.h @@ -172,6 +172,11 @@ __check_struct_size(size_t base, size_t arr, size_t count, size_t *size) (typeof(ptr))(__v + 1); \ })
+#define ptr_align(ptr, align) ({ \ + unsigned long __v = (unsigned long)(ptr); \ + (typeof(ptr))round_up(__v, (align)); \ +}) + #define page_mask_bits(ptr) ptr_mask_bits(ptr, PAGE_SHIFT) #define page_unmask_bits(ptr) ptr_unmask_bits(ptr, PAGE_SHIFT) #define page_pack_bits(ptr, bits) ptr_pack_bits(ptr, bits, PAGE_SHIFT)
On Thu, 30 Jan 2020, Akeem G Abodunrin akeem.g.abodunrin@intel.com wrote:
diff --git a/drivers/gpu/drm/i915/i915_utils.h b/drivers/gpu/drm/i915/i915_utils.h index b0ade76bec90..7ac5b3565845 100644 --- a/drivers/gpu/drm/i915/i915_utils.h +++ b/drivers/gpu/drm/i915/i915_utils.h @@ -172,6 +172,11 @@ __check_struct_size(size_t base, size_t arr, size_t count, size_t *size) (typeof(ptr))(__v + 1); \ })
+#define ptr_align(ptr, align) ({ \
- unsigned long __v = (unsigned long)(ptr); \
- (typeof(ptr))round_up(__v, (align)); \
+})
There's PTR_ALIGN() in include/kernel.h.
BR, Jani.
Quoting Akeem G Abodunrin (2020-01-30 18:57:21)
From: Prathap Kumar Valsan prathap.kumar.valsan@intel.com
On gen7 and gen7.5 devices, there could be leftover data residuals in EU/L3 from the retiring context. This patch introduces workaround to clear that residual contexts, by submitting a batch buffer with dedicated HW context to the GPU with ring allocation for each context switching.
This security mitigation change does not trigger any performance regression. Performance is on par with current mainline/drm-tip.
Signed-off-by: Mika Kuoppala mika.kuoppala@linux.intel.com Signed-off-by: Prathap Kumar Valsan prathap.kumar.valsan@intel.com Signed-off-by: Akeem G Abodunrin akeem.g.abodunrin@intel.com Cc: Chris Wilson chris.p.wilson@intel.com Cc: Balestrieri Francesco francesco.balestrieri@intel.com Cc: Bloomfield Jon jon.bloomfield@intel.com Cc: Dutt Sudeep sudeep.dutt@intel.com
drivers/gpu/drm/i915/Makefile | 1 + drivers/gpu/drm/i915/gt/gen7_renderclear.c | 535 ++++++++++++++++++ drivers/gpu/drm/i915/gt/gen7_renderclear.h | 15 + drivers/gpu/drm/i915/gt/intel_gpu_commands.h | 17 +- .../gpu/drm/i915/gt/intel_ring_submission.c | 3 +- drivers/gpu/drm/i915/i915_utils.h | 5 + 6 files changed, 572 insertions(+), 4 deletions(-) create mode 100644 drivers/gpu/drm/i915/gt/gen7_renderclear.c create mode 100644 drivers/gpu/drm/i915/gt/gen7_renderclear.h
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile index 3c88d7d8c764..f96bae664a03 100644 --- a/drivers/gpu/drm/i915/Makefile +++ b/drivers/gpu/drm/i915/Makefile @@ -78,6 +78,7 @@ gt-y += \ gt/debugfs_gt.o \ gt/debugfs_gt_pm.o \ gt/gen6_ppgtt.o \
gt/gen7_renderclear.o \ gt/gen8_ppgtt.o \ gt/intel_breadcrumbs.o \ gt/intel_context.o \diff --git a/drivers/gpu/drm/i915/gt/gen7_renderclear.c b/drivers/gpu/drm/i915/gt/gen7_renderclear.c new file mode 100644 index 000000000000..a6f5f1602e33 --- /dev/null +++ b/drivers/gpu/drm/i915/gt/gen7_renderclear.c @@ -0,0 +1,535 @@ +// SPDX-License-Identifier: MIT +/*
- Copyright © 2019 Intel Corporation
- */
+#include "gen7_renderclear.h" +#include "i915_drv.h" +#include "i915_utils.h" +#include "intel_gpu_commands.h"
+#define MAX_URB_ENTRIES 64 +#define STATE_SIZE (4 * 1024) +#define GT3_INLINE_DATA_DELAYS 0x1E00 +#define batch_advance(Y, CS) GEM_BUG_ON((Y)->end != (CS))
+/*
- Media CB Kernel for gen7 devices
- TODO: Add comments to kernel, indicating what each array of hex does or
- include header file, which has assembly source and support in igt to be
- able to generate kernel in this same format
- */
Having the original source code for the kernels in IGT is the best way to proceed. The kernels should also be split into separate files which can be generated from IGT and copied over as-is for easy verification.
Regards, Joonas
dri-devel@lists.freedesktop.org
-
Akeem G Abodunrin -
Jani Nikula -
Joonas Lahtinen