linux/drivers/gpu/drm/i915/selftests/intel_workarounds.c

902 lines
19 KiB
C
Raw Normal View History

/*
* SPDX-License-Identifier: MIT
*
* Copyright © 2018 Intel Corporation
*/
#include "../i915_selftest.h"
#include "../i915_reset.h"
#include "igt_flush_test.h"
#include "igt_reset.h"
#include "igt_spinner.h"
#include "igt_wedge_me.h"
#include "mock_context.h"
#include "mock_drm.h"
static const struct wo_register {
enum intel_platform platform;
u32 reg;
} wo_registers[] = {
{ INTEL_GEMINILAKE, 0x731c }
};
#define REF_NAME_MAX (INTEL_ENGINE_CS_MAX_NAME + 4)
struct wa_lists {
struct i915_wa_list gt_wa_list;
struct {
char name[REF_NAME_MAX];
struct i915_wa_list wa_list;
} engine[I915_NUM_ENGINES];
};
static void
reference_lists_init(struct drm_i915_private *i915, struct wa_lists *lists)
{
struct intel_engine_cs *engine;
enum intel_engine_id id;
memset(lists, 0, sizeof(*lists));
wa_init_start(&lists->gt_wa_list, "GT_REF");
gt_init_workarounds(i915, &lists->gt_wa_list);
wa_init_finish(&lists->gt_wa_list);
for_each_engine(engine, i915, id) {
struct i915_wa_list *wal = &lists->engine[id].wa_list;
char *name = lists->engine[id].name;
snprintf(name, REF_NAME_MAX, "%s_REF", engine->name);
wa_init_start(wal, name);
engine_init_workarounds(engine, wal);
wa_init_finish(wal);
}
}
static void
reference_lists_fini(struct drm_i915_private *i915, struct wa_lists *lists)
{
struct intel_engine_cs *engine;
enum intel_engine_id id;
for_each_engine(engine, i915, id)
intel_wa_list_free(&lists->engine[id].wa_list);
intel_wa_list_free(&lists->gt_wa_list);
}
static struct drm_i915_gem_object *
read_nonprivs(struct i915_gem_context *ctx, struct intel_engine_cs *engine)
{
const u32 base = engine->mmio_base;
struct drm_i915_gem_object *result;
intel_wakeref_t wakeref;
struct i915_request *rq;
struct i915_vma *vma;
u32 srm, *cs;
int err;
int i;
result = i915_gem_object_create_internal(engine->i915, PAGE_SIZE);
if (IS_ERR(result))
return result;
i915_gem_object_set_cache_coherency(result, I915_CACHE_LLC);
cs = i915_gem_object_pin_map(result, I915_MAP_WB);
if (IS_ERR(cs)) {
err = PTR_ERR(cs);
goto err_obj;
}
memset(cs, 0xc5, PAGE_SIZE);
drm/i915: Flush pages on acquisition When we return pages to the system, we ensure that they are marked as being in the CPU domain since any external access is uncontrolled and we must assume the worst. This means that we need to always flush the pages on acquisition if we need to use them on the GPU, and from the beginning have used set-domain. Set-domain is overkill for the purpose as it is a general synchronisation barrier, but our intent is to only flush the pages being swapped in. If we move that flush into the pages acquisition phase, we know then that when we have obj->mm.pages, they are coherent with the GPU and need only maintain that status without resorting to heavy handed use of set-domain. The principle knock-on effect for userspace is through mmap-gtt pagefaulting. Our uAPI has always implied that the GTT mmap was async (especially as when any pagefault occurs is unpredicatable to userspace) and so userspace had to apply explicit domain control itself (set-domain). However, swapping is transparent to the kernel, and so on first fault we need to acquire the pages and make them coherent for access through the GTT. Our use of set-domain here leaks into the uABI that the first pagefault was synchronous. This is unintentional and baring a few igt should be unoticed, nevertheless we bump the uABI version for mmap-gtt to reflect the change in behaviour. Another implication of the change is that gem_create() is presumed to create an object that is coherent with the CPU and is in the CPU write domain, so a set-domain(CPU) following a gem_create() would be a minor operation that merely checked whether we could allocate all pages for the object. On applying this change, a set-domain(CPU) causes a clflush as we acquire the pages. This will have a small impact on mesa as we move the clflush here on !llc from execbuf time to create, but that should have minimal performance impact as the same clflush exists but is now done early and because of the clflush issue, userspace recycles bo and so should resist allocating fresh objects. Internally, the presumption that objects are created in the CPU write-domain and remain so through writes to obj->mm.mapping is more prevalent than I expected; but easy enough to catch and apply a manual flush. For the future, we should push the page flush from the central set_pages() into the callers so that we can more finely control when it is applied, but for now doing it one location is easier to validate, at the cost of sometimes flushing when there is no need. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.william.auld@gmail.com> Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Antonio Argenziano <antonio.argenziano@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190321161908.8007-1-chris@chris-wilson.co.uk
2019-03-21 16:19:07 +00:00
i915_gem_object_flush_map(result);
i915_gem_object_unpin_map(result);
vma = i915_vma_instance(result, &engine->i915->ggtt.vm, NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err_obj;
}
err = i915_vma_pin(vma, 0, 0, PIN_GLOBAL);
if (err)
goto err_obj;
rq = ERR_PTR(-ENODEV);
with_intel_runtime_pm(engine->i915, wakeref)
rq = i915_request_alloc(engine, ctx);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
goto err_pin;
}
err = i915_vma_move_to_active(vma, rq, EXEC_OBJECT_WRITE);
if (err)
goto err_req;
srm = MI_STORE_REGISTER_MEM | MI_SRM_LRM_GLOBAL_GTT;
if (INTEL_GEN(ctx->i915) >= 8)
srm++;
cs = intel_ring_begin(rq, 4 * RING_MAX_NONPRIV_SLOTS);
if (IS_ERR(cs)) {
err = PTR_ERR(cs);
goto err_req;
}
for (i = 0; i < RING_MAX_NONPRIV_SLOTS; i++) {
*cs++ = srm;
*cs++ = i915_mmio_reg_offset(RING_FORCE_TO_NONPRIV(base, i));
*cs++ = i915_ggtt_offset(vma) + sizeof(u32) * i;
*cs++ = 0;
}
intel_ring_advance(rq, cs);
i915_gem_object_get(result);
i915_gem_object_set_active_reference(result);
i915_request_add(rq);
i915_vma_unpin(vma);
return result;
err_req:
i915_request_add(rq);
err_pin:
i915_vma_unpin(vma);
err_obj:
i915_gem_object_put(result);
return ERR_PTR(err);
}
static u32
get_whitelist_reg(const struct intel_engine_cs *engine, unsigned int i)
{
i915_reg_t reg = i < engine->whitelist.count ?
engine->whitelist.list[i].reg :
RING_NOPID(engine->mmio_base);
return i915_mmio_reg_offset(reg);
}
static void
print_results(const struct intel_engine_cs *engine, const u32 *results)
{
unsigned int i;
for (i = 0; i < RING_MAX_NONPRIV_SLOTS; i++) {
u32 expected = get_whitelist_reg(engine, i);
u32 actual = results[i];
pr_info("RING_NONPRIV[%d]: expected 0x%08x, found 0x%08x\n",
i, expected, actual);
}
}
static int check_whitelist(struct i915_gem_context *ctx,
struct intel_engine_cs *engine)
{
struct drm_i915_gem_object *results;
struct igt_wedge_me wedge;
u32 *vaddr;
int err;
int i;
results = read_nonprivs(ctx, engine);
if (IS_ERR(results))
return PTR_ERR(results);
err = 0;
igt_wedge_on_timeout(&wedge, ctx->i915, HZ / 5) /* a safety net! */
err = i915_gem_object_set_to_cpu_domain(results, false);
if (i915_terminally_wedged(ctx->i915))
err = -EIO;
if (err)
goto out_put;
vaddr = i915_gem_object_pin_map(results, I915_MAP_WB);
if (IS_ERR(vaddr)) {
err = PTR_ERR(vaddr);
goto out_put;
}
for (i = 0; i < RING_MAX_NONPRIV_SLOTS; i++) {
u32 expected = get_whitelist_reg(engine, i);
u32 actual = vaddr[i];
if (expected != actual) {
print_results(engine, vaddr);
pr_err("Invalid RING_NONPRIV[%d], expected 0x%08x, found 0x%08x\n",
i, expected, actual);
err = -EINVAL;
break;
}
}
i915_gem_object_unpin_map(results);
out_put:
i915_gem_object_put(results);
return err;
}
static int do_device_reset(struct intel_engine_cs *engine)
{
i915_reset(engine->i915, engine->mask, "live_workarounds");
return 0;
}
static int do_engine_reset(struct intel_engine_cs *engine)
{
return i915_reset_engine(engine, "live_workarounds");
}
static int
switch_to_scratch_context(struct intel_engine_cs *engine,
struct igt_spinner *spin)
{
struct i915_gem_context *ctx;
struct i915_request *rq;
intel_wakeref_t wakeref;
int err = 0;
ctx = kernel_context(engine->i915);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
GEM_BUG_ON(i915_gem_context_is_bannable(ctx));
rq = ERR_PTR(-ENODEV);
with_intel_runtime_pm(engine->i915, wakeref)
rq = igt_spinner_create_request(spin, ctx, engine, MI_NOOP);
kernel_context_close(ctx);
if (IS_ERR(rq)) {
spin = NULL;
err = PTR_ERR(rq);
goto err;
}
i915_request_add(rq);
if (spin && !igt_wait_for_spinner(spin, rq)) {
pr_err("Spinner failed to start\n");
err = -ETIMEDOUT;
}
err:
if (err && spin)
igt_spinner_end(spin);
return err;
}
static int check_whitelist_across_reset(struct intel_engine_cs *engine,
int (*reset)(struct intel_engine_cs *),
const char *name)
{
struct drm_i915_private *i915 = engine->i915;
struct i915_gem_context *ctx;
struct igt_spinner spin;
intel_wakeref_t wakeref;
int err;
pr_info("Checking %d whitelisted registers (RING_NONPRIV) [%s]\n",
engine->whitelist.count, name);
err = igt_spinner_init(&spin, i915);
if (err)
return err;
ctx = kernel_context(i915);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
err = check_whitelist(ctx, engine);
if (err) {
pr_err("Invalid whitelist *before* %s reset!\n", name);
goto out;
}
err = switch_to_scratch_context(engine, &spin);
if (err)
goto out;
with_intel_runtime_pm(i915, wakeref)
err = reset(engine);
igt_spinner_end(&spin);
igt_spinner_fini(&spin);
if (err) {
pr_err("%s reset failed\n", name);
goto out;
}
err = check_whitelist(ctx, engine);
if (err) {
pr_err("Whitelist not preserved in context across %s reset!\n",
name);
goto out;
}
kernel_context_close(ctx);
ctx = kernel_context(i915);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
err = check_whitelist(ctx, engine);
if (err) {
pr_err("Invalid whitelist *after* %s reset in fresh context!\n",
name);
goto out;
}
out:
kernel_context_close(ctx);
return err;
}
static struct i915_vma *create_scratch(struct i915_gem_context *ctx)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
void *ptr;
int err;
obj = i915_gem_object_create_internal(ctx->i915, PAGE_SIZE);
if (IS_ERR(obj))
return ERR_CAST(obj);
i915_gem_object_set_cache_coherency(obj, I915_CACHE_LLC);
ptr = i915_gem_object_pin_map(obj, I915_MAP_WB);
if (IS_ERR(ptr)) {
err = PTR_ERR(ptr);
goto err_obj;
}
memset(ptr, 0xc5, PAGE_SIZE);
drm/i915: Flush pages on acquisition When we return pages to the system, we ensure that they are marked as being in the CPU domain since any external access is uncontrolled and we must assume the worst. This means that we need to always flush the pages on acquisition if we need to use them on the GPU, and from the beginning have used set-domain. Set-domain is overkill for the purpose as it is a general synchronisation barrier, but our intent is to only flush the pages being swapped in. If we move that flush into the pages acquisition phase, we know then that when we have obj->mm.pages, they are coherent with the GPU and need only maintain that status without resorting to heavy handed use of set-domain. The principle knock-on effect for userspace is through mmap-gtt pagefaulting. Our uAPI has always implied that the GTT mmap was async (especially as when any pagefault occurs is unpredicatable to userspace) and so userspace had to apply explicit domain control itself (set-domain). However, swapping is transparent to the kernel, and so on first fault we need to acquire the pages and make them coherent for access through the GTT. Our use of set-domain here leaks into the uABI that the first pagefault was synchronous. This is unintentional and baring a few igt should be unoticed, nevertheless we bump the uABI version for mmap-gtt to reflect the change in behaviour. Another implication of the change is that gem_create() is presumed to create an object that is coherent with the CPU and is in the CPU write domain, so a set-domain(CPU) following a gem_create() would be a minor operation that merely checked whether we could allocate all pages for the object. On applying this change, a set-domain(CPU) causes a clflush as we acquire the pages. This will have a small impact on mesa as we move the clflush here on !llc from execbuf time to create, but that should have minimal performance impact as the same clflush exists but is now done early and because of the clflush issue, userspace recycles bo and so should resist allocating fresh objects. Internally, the presumption that objects are created in the CPU write-domain and remain so through writes to obj->mm.mapping is more prevalent than I expected; but easy enough to catch and apply a manual flush. For the future, we should push the page flush from the central set_pages() into the callers so that we can more finely control when it is applied, but for now doing it one location is easier to validate, at the cost of sometimes flushing when there is no need. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.william.auld@gmail.com> Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Antonio Argenziano <antonio.argenziano@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190321161908.8007-1-chris@chris-wilson.co.uk
2019-03-21 16:19:07 +00:00
i915_gem_object_flush_map(obj);
i915_gem_object_unpin_map(obj);
vma = i915_vma_instance(obj, &ctx->ppgtt->vm, NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err_obj;
}
err = i915_vma_pin(vma, 0, 0, PIN_USER);
if (err)
goto err_obj;
err = i915_gem_object_set_to_cpu_domain(obj, false);
if (err)
goto err_obj;
return vma;
err_obj:
i915_gem_object_put(obj);
return ERR_PTR(err);
}
static struct i915_vma *create_batch(struct i915_gem_context *ctx)
{
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
int err;
obj = i915_gem_object_create_internal(ctx->i915, 16 * PAGE_SIZE);
if (IS_ERR(obj))
return ERR_CAST(obj);
vma = i915_vma_instance(obj, &ctx->ppgtt->vm, NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err_obj;
}
err = i915_vma_pin(vma, 0, 0, PIN_USER);
if (err)
goto err_obj;
err = i915_gem_object_set_to_wc_domain(obj, true);
if (err)
goto err_obj;
return vma;
err_obj:
i915_gem_object_put(obj);
return ERR_PTR(err);
}
static u32 reg_write(u32 old, u32 new, u32 rsvd)
{
if (rsvd == 0x0000ffff) {
old &= ~(new >> 16);
old |= new & (new >> 16);
} else {
old &= ~rsvd;
old |= new & rsvd;
}
return old;
}
static bool wo_register(struct intel_engine_cs *engine, u32 reg)
{
enum intel_platform platform = INTEL_INFO(engine->i915)->platform;
int i;
for (i = 0; i < ARRAY_SIZE(wo_registers); i++) {
if (wo_registers[i].platform == platform &&
wo_registers[i].reg == reg)
return true;
}
return false;
}
static int check_dirty_whitelist(struct i915_gem_context *ctx,
struct intel_engine_cs *engine)
{
const u32 values[] = {
0x00000000,
0x01010101,
0x10100101,
0x03030303,
0x30300303,
0x05050505,
0x50500505,
0x0f0f0f0f,
0xf00ff00f,
0x10101010,
0xf0f01010,
0x30303030,
0xa0a03030,
0x50505050,
0xc0c05050,
0xf0f0f0f0,
0x11111111,
0x33333333,
0x55555555,
0x0000ffff,
0x00ff00ff,
0xff0000ff,
0xffff00ff,
0xffffffff,
};
struct i915_vma *scratch;
struct i915_vma *batch;
int err = 0, i, v;
u32 *cs, *results;
scratch = create_scratch(ctx);
if (IS_ERR(scratch))
return PTR_ERR(scratch);
batch = create_batch(ctx);
if (IS_ERR(batch)) {
err = PTR_ERR(batch);
goto out_scratch;
}
for (i = 0; i < engine->whitelist.count; i++) {
u32 reg = i915_mmio_reg_offset(engine->whitelist.list[i].reg);
u64 addr = scratch->node.start;
struct i915_request *rq;
u32 srm, lrm, rsvd;
u32 expect;
int idx;
if (wo_register(engine, reg))
continue;
srm = MI_STORE_REGISTER_MEM;
lrm = MI_LOAD_REGISTER_MEM;
if (INTEL_GEN(ctx->i915) >= 8)
lrm++, srm++;
pr_debug("%s: Writing garbage to %x\n",
engine->name, reg);
cs = i915_gem_object_pin_map(batch->obj, I915_MAP_WC);
if (IS_ERR(cs)) {
err = PTR_ERR(cs);
goto out_batch;
}
/* SRM original */
*cs++ = srm;
*cs++ = reg;
*cs++ = lower_32_bits(addr);
*cs++ = upper_32_bits(addr);
idx = 1;
for (v = 0; v < ARRAY_SIZE(values); v++) {
/* LRI garbage */
*cs++ = MI_LOAD_REGISTER_IMM(1);
*cs++ = reg;
*cs++ = values[v];
/* SRM result */
*cs++ = srm;
*cs++ = reg;
*cs++ = lower_32_bits(addr + sizeof(u32) * idx);
*cs++ = upper_32_bits(addr + sizeof(u32) * idx);
idx++;
}
for (v = 0; v < ARRAY_SIZE(values); v++) {
/* LRI garbage */
*cs++ = MI_LOAD_REGISTER_IMM(1);
*cs++ = reg;
*cs++ = ~values[v];
/* SRM result */
*cs++ = srm;
*cs++ = reg;
*cs++ = lower_32_bits(addr + sizeof(u32) * idx);
*cs++ = upper_32_bits(addr + sizeof(u32) * idx);
idx++;
}
GEM_BUG_ON(idx * sizeof(u32) > scratch->size);
/* LRM original -- don't leave garbage in the context! */
*cs++ = lrm;
*cs++ = reg;
*cs++ = lower_32_bits(addr);
*cs++ = upper_32_bits(addr);
*cs++ = MI_BATCH_BUFFER_END;
drm/i915: Flush pages on acquisition When we return pages to the system, we ensure that they are marked as being in the CPU domain since any external access is uncontrolled and we must assume the worst. This means that we need to always flush the pages on acquisition if we need to use them on the GPU, and from the beginning have used set-domain. Set-domain is overkill for the purpose as it is a general synchronisation barrier, but our intent is to only flush the pages being swapped in. If we move that flush into the pages acquisition phase, we know then that when we have obj->mm.pages, they are coherent with the GPU and need only maintain that status without resorting to heavy handed use of set-domain. The principle knock-on effect for userspace is through mmap-gtt pagefaulting. Our uAPI has always implied that the GTT mmap was async (especially as when any pagefault occurs is unpredicatable to userspace) and so userspace had to apply explicit domain control itself (set-domain). However, swapping is transparent to the kernel, and so on first fault we need to acquire the pages and make them coherent for access through the GTT. Our use of set-domain here leaks into the uABI that the first pagefault was synchronous. This is unintentional and baring a few igt should be unoticed, nevertheless we bump the uABI version for mmap-gtt to reflect the change in behaviour. Another implication of the change is that gem_create() is presumed to create an object that is coherent with the CPU and is in the CPU write domain, so a set-domain(CPU) following a gem_create() would be a minor operation that merely checked whether we could allocate all pages for the object. On applying this change, a set-domain(CPU) causes a clflush as we acquire the pages. This will have a small impact on mesa as we move the clflush here on !llc from execbuf time to create, but that should have minimal performance impact as the same clflush exists but is now done early and because of the clflush issue, userspace recycles bo and so should resist allocating fresh objects. Internally, the presumption that objects are created in the CPU write-domain and remain so through writes to obj->mm.mapping is more prevalent than I expected; but easy enough to catch and apply a manual flush. For the future, we should push the page flush from the central set_pages() into the callers so that we can more finely control when it is applied, but for now doing it one location is easier to validate, at the cost of sometimes flushing when there is no need. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> Cc: Matthew Auld <matthew.william.auld@gmail.com> Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com> Cc: Antonio Argenziano <antonio.argenziano@intel.com> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com> Reviewed-by: Matthew Auld <matthew.william.auld@gmail.com> Link: https://patchwork.freedesktop.org/patch/msgid/20190321161908.8007-1-chris@chris-wilson.co.uk
2019-03-21 16:19:07 +00:00
i915_gem_object_flush_map(batch->obj);
i915_gem_object_unpin_map(batch->obj);
i915_gem_chipset_flush(ctx->i915);
rq = i915_request_alloc(engine, ctx);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
goto out_batch;
}
if (engine->emit_init_breadcrumb) { /* Be nice if we hang */
err = engine->emit_init_breadcrumb(rq);
if (err)
goto err_request;
}
err = engine->emit_bb_start(rq,
batch->node.start, PAGE_SIZE,
0);
if (err)
goto err_request;
err_request:
i915_request_add(rq);
if (err)
goto out_batch;
if (i915_request_wait(rq, I915_WAIT_LOCKED, HZ / 5) < 0) {
pr_err("%s: Futzing %x timedout; cancelling test\n",
engine->name, reg);
i915_gem_set_wedged(ctx->i915);
err = -EIO;
goto out_batch;
}
results = i915_gem_object_pin_map(scratch->obj, I915_MAP_WB);
if (IS_ERR(results)) {
err = PTR_ERR(results);
goto out_batch;
}
GEM_BUG_ON(values[ARRAY_SIZE(values) - 1] != 0xffffffff);
rsvd = results[ARRAY_SIZE(values)]; /* detect write masking */
if (!rsvd) {
pr_err("%s: Unable to write to whitelisted register %x\n",
engine->name, reg);
err = -EINVAL;
goto out_unpin;
}
expect = results[0];
idx = 1;
for (v = 0; v < ARRAY_SIZE(values); v++) {
expect = reg_write(expect, values[v], rsvd);
if (results[idx] != expect)
err++;
idx++;
}
for (v = 0; v < ARRAY_SIZE(values); v++) {
expect = reg_write(expect, ~values[v], rsvd);
if (results[idx] != expect)
err++;
idx++;
}
if (err) {
pr_err("%s: %d mismatch between values written to whitelisted register [%x], and values read back!\n",
engine->name, err, reg);
pr_info("%s: Whitelisted register: %x, original value %08x, rsvd %08x\n",
engine->name, reg, results[0], rsvd);
expect = results[0];
idx = 1;
for (v = 0; v < ARRAY_SIZE(values); v++) {
u32 w = values[v];
expect = reg_write(expect, w, rsvd);
pr_info("Wrote %08x, read %08x, expect %08x\n",
w, results[idx], expect);
idx++;
}
for (v = 0; v < ARRAY_SIZE(values); v++) {
u32 w = ~values[v];
expect = reg_write(expect, w, rsvd);
pr_info("Wrote %08x, read %08x, expect %08x\n",
w, results[idx], expect);
idx++;
}
err = -EINVAL;
}
out_unpin:
i915_gem_object_unpin_map(scratch->obj);
if (err)
break;
}
if (igt_flush_test(ctx->i915, I915_WAIT_LOCKED))
err = -EIO;
out_batch:
i915_vma_unpin_and_release(&batch, 0);
out_scratch:
i915_vma_unpin_and_release(&scratch, 0);
return err;
}
static int live_dirty_whitelist(void *arg)
{
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine;
struct i915_gem_context *ctx;
enum intel_engine_id id;
intel_wakeref_t wakeref;
struct drm_file *file;
int err = 0;
/* Can the user write to the whitelisted registers? */
if (INTEL_GEN(i915) < 7) /* minimum requirement for LRI, SRM, LRM */
return 0;
wakeref = intel_runtime_pm_get(i915);
mutex_unlock(&i915->drm.struct_mutex);
file = mock_file(i915);
mutex_lock(&i915->drm.struct_mutex);
if (IS_ERR(file)) {
err = PTR_ERR(file);
goto out_rpm;
}
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out_file;
}
for_each_engine(engine, i915, id) {
if (engine->whitelist.count == 0)
continue;
err = check_dirty_whitelist(ctx, engine);
if (err)
goto out_file;
}
out_file:
mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
mutex_lock(&i915->drm.struct_mutex);
out_rpm:
intel_runtime_pm_put(i915, wakeref);
return err;
}
static int live_reset_whitelist(void *arg)
{
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine = i915->engine[RCS0];
int err = 0;
/* If we reset the gpu, we should not lose the RING_NONPRIV */
if (!engine || engine->whitelist.count == 0)
return 0;
igt_global_reset_lock(i915);
if (intel_has_reset_engine(i915)) {
err = check_whitelist_across_reset(engine,
do_engine_reset,
"engine");
if (err)
goto out;
}
if (intel_has_gpu_reset(i915)) {
err = check_whitelist_across_reset(engine,
do_device_reset,
"device");
if (err)
goto out;
}
out:
igt_global_reset_unlock(i915);
return err;
}
static bool verify_gt_engine_wa(struct drm_i915_private *i915,
struct wa_lists *lists, const char *str)
{
struct intel_engine_cs *engine;
enum intel_engine_id id;
bool ok = true;
ok &= wa_list_verify(&i915->uncore, &lists->gt_wa_list, str);
for_each_engine(engine, i915, id)
ok &= wa_list_verify(engine->uncore,
&lists->engine[id].wa_list, str);
return ok;
}
static int
live_gpu_reset_gt_engine_workarounds(void *arg)
{
struct drm_i915_private *i915 = arg;
intel_wakeref_t wakeref;
struct wa_lists lists;
bool ok;
if (!intel_has_gpu_reset(i915))
return 0;
pr_info("Verifying after GPU reset...\n");
igt_global_reset_lock(i915);
wakeref = intel_runtime_pm_get(i915);
reference_lists_init(i915, &lists);
ok = verify_gt_engine_wa(i915, &lists, "before reset");
if (!ok)
goto out;
i915_reset(i915, ALL_ENGINES, "live_workarounds");
ok = verify_gt_engine_wa(i915, &lists, "after reset");
out:
reference_lists_fini(i915, &lists);
intel_runtime_pm_put(i915, wakeref);
igt_global_reset_unlock(i915);
return ok ? 0 : -ESRCH;
}
static int
live_engine_reset_gt_engine_workarounds(void *arg)
{
struct drm_i915_private *i915 = arg;
struct intel_engine_cs *engine;
struct i915_gem_context *ctx;
struct igt_spinner spin;
enum intel_engine_id id;
struct i915_request *rq;
intel_wakeref_t wakeref;
struct wa_lists lists;
int ret = 0;
if (!intel_has_reset_engine(i915))
return 0;
ctx = kernel_context(i915);
if (IS_ERR(ctx))
return PTR_ERR(ctx);
igt_global_reset_lock(i915);
wakeref = intel_runtime_pm_get(i915);
reference_lists_init(i915, &lists);
for_each_engine(engine, i915, id) {
bool ok;
pr_info("Verifying after %s reset...\n", engine->name);
ok = verify_gt_engine_wa(i915, &lists, "before reset");
if (!ok) {
ret = -ESRCH;
goto err;
}
i915_reset_engine(engine, "live_workarounds");
ok = verify_gt_engine_wa(i915, &lists, "after idle reset");
if (!ok) {
ret = -ESRCH;
goto err;
}
ret = igt_spinner_init(&spin, i915);
if (ret)
goto err;
rq = igt_spinner_create_request(&spin, ctx, engine, MI_NOOP);
if (IS_ERR(rq)) {
ret = PTR_ERR(rq);
igt_spinner_fini(&spin);
goto err;
}
i915_request_add(rq);
if (!igt_wait_for_spinner(&spin, rq)) {
pr_err("Spinner failed to start\n");
igt_spinner_fini(&spin);
ret = -ETIMEDOUT;
goto err;
}
i915_reset_engine(engine, "live_workarounds");
igt_spinner_end(&spin);
igt_spinner_fini(&spin);
ok = verify_gt_engine_wa(i915, &lists, "after busy reset");
if (!ok) {
ret = -ESRCH;
goto err;
}
}
err:
reference_lists_fini(i915, &lists);
intel_runtime_pm_put(i915, wakeref);
igt_global_reset_unlock(i915);
kernel_context_close(ctx);
igt_flush_test(i915, I915_WAIT_LOCKED);
return ret;
}
int intel_workarounds_live_selftests(struct drm_i915_private *i915)
{
static const struct i915_subtest tests[] = {
SUBTEST(live_dirty_whitelist),
SUBTEST(live_reset_whitelist),
SUBTEST(live_gpu_reset_gt_engine_workarounds),
SUBTEST(live_engine_reset_gt_engine_workarounds),
};
int err;
if (i915_terminally_wedged(i915))
return 0;
mutex_lock(&i915->drm.struct_mutex);
err = i915_subtests(tests, i915);
mutex_unlock(&i915->drm.struct_mutex);
return err;
}