From patchwork Mon Oct 21 11:01:37 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 11201927 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 3DA0A139A for ; Mon, 21 Oct 2019 11:01:52 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 260782084C for ; Mon, 21 Oct 2019 11:01:52 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 260782084C Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=chris-wilson.co.uk Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=intel-gfx-bounces@lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 0FF7B89E59; Mon, 21 Oct 2019 11:01:51 +0000 (UTC) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from fireflyinternet.com (mail.fireflyinternet.com [109.228.58.192]) by gabe.freedesktop.org (Postfix) with ESMTPS id CFA3D89D7D; Mon, 21 Oct 2019 11:01:47 +0000 (UTC) X-Default-Received-SPF: pass (skip=forwardok (res=PASS)) x-ip-name=78.156.65.138; Received: from haswell.alporthouse.com (unverified [78.156.65.138]) by fireflyinternet.com (Firefly Internet (M1)) with ESMTP id 18911183-1500050 for multiple; Mon, 21 Oct 2019 12:01:41 +0100 From: Chris Wilson To: intel-gfx@lists.freedesktop.org Date: Mon, 21 Oct 2019 12:01:37 +0100 Message-Id: <20191021110138.16731-1-chris@chris-wilson.co.uk> X-Mailer: git-send-email 2.24.0.rc0 MIME-Version: 1.0 Subject: [Intel-gfx] [PATCH i-g-t 1/2] i915: Start putting the mmio_base to wider use X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: igt-dev@lists.freedesktop.org Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" Several tests depend upon the implicit engine->mmio_base but have no means of determining the physical layout. Since the kernel has started providing this information, start putting it to use. Signed-off-by: Chris Wilson --- lib/i915/gem_engine_topology.c | 84 ++++++++++++++++++++++++++++++++++ lib/i915/gem_engine_topology.h | 5 ++ tests/i915/gem_ctx_shared.c | 38 +++++---------- tests/i915/gem_exec_latency.c | 17 ++++--- 4 files changed, 111 insertions(+), 33 deletions(-) diff --git a/lib/i915/gem_engine_topology.c b/lib/i915/gem_engine_topology.c index 790d455ff..bd200a4b9 100644 --- a/lib/i915/gem_engine_topology.c +++ b/lib/i915/gem_engine_topology.c @@ -21,7 +21,12 @@ * IN THE SOFTWARE. */ +#include +#include + #include "drmtest.h" +#include "igt_sysfs.h" +#include "intel_chipset.h" #include "ioctl_wrappers.h" #include "i915/gem_engine_topology.h" @@ -337,3 +342,82 @@ bool gem_engine_is_equal(const struct intel_execution_engine2 *e1, { return e1->class == e2->class && e1->instance == e2->instance; } + +static int descend(int dir, const char *path) +{ + int fd; + + fd = openat(dir, path, O_RDONLY); + close(dir); + + return fd; +} + +int gem_engine_property_scanf(int i915, const char *engine, const char *attr, + const char *fmt, ...) +{ + FILE *file; + va_list ap; + int ret; + int fd; + + fd = igt_sysfs_open(i915); + if (fd < 0) + return fd; + + fd = descend(fd, "engine"); + if (fd < 0) + return fd; + + fd = descend(fd, engine); + if (fd < 0) + return fd; + + fd = descend(fd, attr); + if (fd < 0) + return fd; + + file = fdopen(fd, "r"); + if (!file) { + close(fd); + return -1; + } + + va_start(ap, fmt); + ret = vfscanf(file, fmt, ap); + va_end(ap); + + fclose(file); + return ret; +} + +uint32_t gem_engine_mmio_base(int i915, const char *engine) +{ + unsigned int mmio = 0; + + if (gem_engine_property_scanf(i915, engine, "mmio_base", + "%x", &mmio) < 0) { + int gen = intel_gen(intel_get_drm_devid(i915)); + + /* The layout of xcs1+ is unreliable -- hence the property! */ + if (!strcmp(engine, "rcs0")) { + mmio = 0x2000; + } else if (!strcmp(engine, "bcs0")) { + mmio = 0x22000; + } else if (!strcmp(engine, "vcs0")) { + if (gen < 6) + mmio = 0x4000; + else if (gen < 11) + mmio = 0x12000; + else + mmio = 0x1c0000; + } else if (!strcmp(engine, "vecs0")) { + if (gen < 11) + mmio = 0x1a000; + else + mmio = 0x1c8000; + } + } + + return mmio; +} diff --git a/lib/i915/gem_engine_topology.h b/lib/i915/gem_engine_topology.h index d98773e06..e728ebd93 100644 --- a/lib/i915/gem_engine_topology.h +++ b/lib/i915/gem_engine_topology.h @@ -74,4 +74,9 @@ struct intel_execution_engine2 gem_eb_flags_to_engine(unsigned int flags); ((e__) = intel_get_current_physical_engine(&i__)); \ intel_next_engine(&i__)) +__attribute__((format(scanf, 4, 5))) +int gem_engine_property_scanf(int i915, const char *engine, const char *attr, + const char *fmt, ...); +uint32_t gem_engine_mmio_base(int i915, const char *engine); + #endif /* GEM_ENGINE_TOPOLOGY_H */ diff --git a/tests/i915/gem_ctx_shared.c b/tests/i915/gem_ctx_shared.c index f78524822..594468738 100644 --- a/tests/i915/gem_ctx_shared.c +++ b/tests/i915/gem_ctx_shared.c @@ -38,6 +38,7 @@ #include +#include "i915/gem_engine_topology.h" #include "igt_rand.h" #include "igt_vgem.h" #include "sync_file.h" @@ -558,6 +559,14 @@ static uint32_t store_timestamp(int i915, return obj.handle; } +static uint32_t ring_base(int i915, unsigned ring) +{ + if (ring == I915_EXEC_DEFAULT) + ring = I915_EXEC_RENDER; /* XXX */ + + return gem_engine_mmio_base(i915, gem_eb_flags_to_engine(ring).name); +} + static void independent(int i915, unsigned ring, unsigned flags) { const int TIMESTAMP = 1023; @@ -565,33 +574,8 @@ static void independent(int i915, unsigned ring, unsigned flags) igt_spin_t *spin[MAX_ELSP_QLEN]; unsigned int mmio_base; - /* XXX i915_query()! */ - switch (ring) { - case I915_EXEC_DEFAULT: - case I915_EXEC_RENDER: - mmio_base = 0x2000; - break; -#if 0 - case I915_EXEC_BSD: - mmio_base = 0x12000; - break; -#endif - case I915_EXEC_BLT: - mmio_base = 0x22000; - break; - -#define GEN11_VECS0_BASE 0x1c8000 -#define GEN11_VECS1_BASE 0x1d8000 - case I915_EXEC_VEBOX: - if (intel_gen(intel_get_drm_devid(i915)) >= 11) - mmio_base = GEN11_VECS0_BASE; - else - mmio_base = 0x1a000; - break; - - default: - igt_skip("mmio base not known\n"); - } + mmio_base = ring_base(i915, ring); + igt_require_f(mmio_base, "mmio base not known\n"); for (int n = 0; n < ARRAY_SIZE(spin); n++) { const struct igt_spin_factory opts = { diff --git a/tests/i915/gem_exec_latency.c b/tests/i915/gem_exec_latency.c index 9ddb348c0..ea1eee0aa 100644 --- a/tests/i915/gem_exec_latency.c +++ b/tests/i915/gem_exec_latency.c @@ -109,7 +109,7 @@ poll_ring(int fd, unsigned ring, const char *name) igt_spin_free(fd, spin[0]); } -#define RCS_TIMESTAMP (0x2000 + 0x358) +#define TIMESTAMP (0x358) static void latency_on_ring(int fd, unsigned ring, const char *name, unsigned flags) @@ -119,6 +119,7 @@ static void latency_on_ring(int fd, struct drm_i915_gem_exec_object2 obj[3]; struct drm_i915_gem_relocation_entry reloc; struct drm_i915_gem_execbuffer2 execbuf; + const uint32_t mmio_base = gem_engine_mmio_base(fd, name); igt_spin_t *spin = NULL; IGT_CORK_HANDLE(c); volatile uint32_t *reg; @@ -128,7 +129,8 @@ static void latency_on_ring(int fd, double gpu_latency; int i, j; - reg = (volatile uint32_t *)((volatile char *)igt_global_mmio + RCS_TIMESTAMP); + igt_require(mmio_base); + reg = (volatile uint32_t *)((volatile char *)igt_global_mmio + mmio_base + TIMESTAMP); memset(&execbuf, 0, sizeof(execbuf)); execbuf.buffers_ptr = to_user_pointer(&obj[1]); @@ -176,7 +178,7 @@ static void latency_on_ring(int fd, map[i++] = 0x24 << 23 | 1; if (has_64bit_reloc) map[i-1]++; - map[i++] = RCS_TIMESTAMP; /* ring local! */ + map[i++] = mmio_base + TIMESTAMP; map[i++] = offset; if (has_64bit_reloc) map[i++] = offset >> 32; @@ -266,12 +268,15 @@ static void latency_from_ring(int fd, struct drm_i915_gem_exec_object2 obj[3]; struct drm_i915_gem_relocation_entry reloc; struct drm_i915_gem_execbuffer2 execbuf; + const uint32_t mmio_base = gem_engine_mmio_base(fd, name); const unsigned int repeats = ring_size / 2; unsigned int other; uint32_t *map, *results; uint32_t ctx[2] = {}; int i, j; + igt_require(mmio_base); + if (flags & PREEMPT) { ctx[0] = gem_context_create(fd); gem_context_set_priority(fd, ctx[0], -1023); @@ -352,7 +357,7 @@ static void latency_from_ring(int fd, map[i++] = 0x24 << 23 | 1; if (has_64bit_reloc) map[i-1]++; - map[i++] = RCS_TIMESTAMP; /* ring local! */ + map[i++] = mmio_base + TIMESTAMP; map[i++] = offset; if (has_64bit_reloc) map[i++] = offset >> 32; @@ -377,7 +382,7 @@ static void latency_from_ring(int fd, map[i++] = 0x24 << 23 | 1; if (has_64bit_reloc) map[i-1]++; - map[i++] = RCS_TIMESTAMP; /* ring local! */ + map[i++] = mmio_base + TIMESTAMP; map[i++] = offset; if (has_64bit_reloc) map[i++] = offset >> 32; @@ -670,7 +675,7 @@ igt_main ring_size = 1024; intel_register_access_init(&mmio_data, intel_get_pci_device(), false, device); - rcs_clock = clockrate(device, RCS_TIMESTAMP); + rcs_clock = clockrate(device, 0x2000 + TIMESTAMP); igt_info("RCS timestamp clock: %.0fKHz, %.1fns\n", rcs_clock / 1e3, 1e9 / rcs_clock); rcs_clock = 1e9 / rcs_clock; From patchwork Mon Oct 21 11:01:38 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Chris Wilson X-Patchwork-Id: 11201925 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 0AFCA139A for ; Mon, 21 Oct 2019 11:01:51 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id E73F92084C for ; Mon, 21 Oct 2019 11:01:50 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org E73F92084C Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=chris-wilson.co.uk Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=intel-gfx-bounces@lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 6BE4D89D7D; Mon, 21 Oct 2019 11:01:49 +0000 (UTC) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from fireflyinternet.com (mail.fireflyinternet.com [109.228.58.192]) by gabe.freedesktop.org (Postfix) with ESMTPS id CF9F189A16; Mon, 21 Oct 2019 11:01:47 +0000 (UTC) X-Default-Received-SPF: pass (skip=forwardok (res=PASS)) x-ip-name=78.156.65.138; Received: from haswell.alporthouse.com (unverified [78.156.65.138]) by fireflyinternet.com (Firefly Internet (M1)) with ESMTP id 18911184-1500050 for multiple; Mon, 21 Oct 2019 12:01:41 +0100 From: Chris Wilson To: intel-gfx@lists.freedesktop.org Date: Mon, 21 Oct 2019 12:01:38 +0100 Message-Id: <20191021110138.16731-2-chris@chris-wilson.co.uk> X-Mailer: git-send-email 2.24.0.rc0 In-Reply-To: <20191021110138.16731-1-chris@chris-wilson.co.uk> References: <20191021110138.16731-1-chris@chris-wilson.co.uk> MIME-Version: 1.0 Subject: [Intel-gfx] [PATCH i-g-t 2/2] i915/gem_ctx_isolation: Check engine relative registers X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: igt-dev@lists.freedesktop.org Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" Some of the non-privileged registers are at the same offset on each engine. We can improve our coverage for unknown HW layout by using the reported engine->mmio_base for relative offsets. Signed-off-by: Chris Wilson --- tests/i915/gem_ctx_isolation.c | 160 ++++++++++++++++++++------------- 1 file changed, 99 insertions(+), 61 deletions(-) diff --git a/tests/i915/gem_ctx_isolation.c b/tests/i915/gem_ctx_isolation.c index 6aa27133c..2ed71dd34 100644 --- a/tests/i915/gem_ctx_isolation.c +++ b/tests/i915/gem_ctx_isolation.c @@ -70,6 +70,7 @@ static const struct named_register { uint32_t ignore_bits; uint32_t write_mask; /* some registers bits do not exist */ bool masked; + bool relative; } nonpriv_registers[] = { { "NOPID", NOCTX, RCS0, 0x2094 }, { "MI_PREDICATE_RESULT_2", NOCTX, RCS0, 0x23bc }, @@ -150,67 +151,45 @@ static const struct named_register { { "HALF_SLICE_CHICKEN7", GEN_RANGE(11, 11), RCS0, 0xe194, .masked = true }, { "SAMPLER_MODE", GEN_RANGE(11, 11), RCS0, 0xe18c, .masked = true }, - { "BCS_GPR", GEN9, BCS0, 0x22600, 32 }, { "BCS_SWCTRL", GEN8, BCS0, 0x22200, .write_mask = 0x3, .masked = true }, { "MFC_VDBOX1", NOCTX, VCS0, 0x12800, 64 }, { "MFC_VDBOX2", NOCTX, VCS1, 0x1c800, 64 }, - { "VCS0_GPR", GEN_RANGE(9, 10), VCS0, 0x12600, 32 }, - { "VCS1_GPR", GEN_RANGE(9, 10), VCS1, 0x1c600, 32 }, - { "VECS_GPR", GEN_RANGE(9, 10), VECS0, 0x1a600, 32 }, - - { "VCS0_GPR", GEN11, VCS0, 0x1c0600, 32 }, - { "VCS1_GPR", GEN11, VCS1, 0x1c4600, 32 }, - { "VCS2_GPR", GEN11, VCS2, 0x1d0600, 32 }, - { "VCS3_GPR", GEN11, VCS3, 0x1d4600, 32 }, - { "VECS_GPR", GEN11, VECS0, 0x1c8600, 32 }, + { "xCS_GPR", GEN9, ALL, 0x600, 32, .relative = true }, {} }, ignore_registers[] = { { "RCS timestamp", GEN6, ~0u, 0x2358 }, { "BCS timestamp", GEN7, ~0u, 0x22358 }, - { "VCS0 timestamp", GEN_RANGE(7, 10), ~0u, 0x12358 }, - { "VCS1 timestamp", GEN_RANGE(7, 10), ~0u, 0x1c358 }, - { "VECS timestamp", GEN_RANGE(8, 10), ~0u, 0x1a358 }, - - { "VCS0 timestamp", GEN11, ~0u, 0x1c0358 }, - { "VCS1 timestamp", GEN11, ~0u, 0x1c4358 }, - { "VCS2 timestamp", GEN11, ~0u, 0x1d0358 }, - { "VCS3 timestamp", GEN11, ~0u, 0x1d4358 }, - { "VECS timestamp", GEN11, ~0u, 0x1c8358 }, + { "xCS timestamp", GEN8, ALL, 0x358, .relative = true }, /* huc read only */ - { "BSD0 0x2000", GEN11, ~0u, 0x1c0000 + 0x2000 }, - { "BSD0 0x2000", GEN11, ~0u, 0x1c0000 + 0x2014 }, - { "BSD0 0x2000", GEN11, ~0u, 0x1c0000 + 0x23b0 }, - - { "BSD1 0x2000", GEN11, ~0u, 0x1c4000 + 0x2000 }, - { "BSD1 0x2000", GEN11, ~0u, 0x1c4000 + 0x2014 }, - { "BSD1 0x2000", GEN11, ~0u, 0x1c4000 + 0x23b0 }, - - { "BSD2 0x2000", GEN11, ~0u, 0x1d0000 + 0x2000 }, - { "BSD2 0x2000", GEN11, ~0u, 0x1d0000 + 0x2014 }, - { "BSD2 0x2000", GEN11, ~0u, 0x1d0000 + 0x23b0 }, - - { "BSD3 0x2000", GEN11, ~0u, 0x1d4000 + 0x2000 }, - { "BSD3 0x2000", GEN11, ~0u, 0x1d4000 + 0x2014 }, - { "BSD3 0x2000", GEN11, ~0u, 0x1d4000 + 0x23b0 }, + { "BSD 0x2000", GEN11, ALL, 0x2000, .relative = true }, + { "BSD 0x2014", GEN11, ALL, 0x2014, .relative = true }, + { "BSD 0x23b0", GEN11, ALL, 0x23b0, .relative = true}, {} }; -static const char *register_name(uint32_t offset, char *buf, size_t len) +static const char * +register_name(uint32_t offset, uint32_t mmio_base, char *buf, size_t len) { for (const struct named_register *r = nonpriv_registers; r->name; r++) { unsigned int width = r->count ? 4*r->count : 4; - if (offset >= r->offset && offset < r->offset + width) { + uint32_t base; + + base = r->offset; + if (r->relative) + base += mmio_base; + + if (offset >= base && offset < base + width) { if (r->count <= 1) return r->name; snprintf(buf, len, "%s[%d]", - r->name, (offset - r->offset)/4); + r->name, (offset - base) / 4); return buf; } } @@ -218,22 +197,35 @@ static const char *register_name(uint32_t offset, char *buf, size_t len) return "unknown"; } -static const struct named_register *lookup_register(uint32_t offset) +static const struct named_register * +lookup_register(uint32_t offset, uint32_t mmio_base) { for (const struct named_register *r = nonpriv_registers; r->name; r++) { unsigned int width = r->count ? 4*r->count : 4; - if (offset >= r->offset && offset < r->offset + width) + uint32_t base; + + base = r->offset; + if (r->relative) + base += mmio_base; + + if (offset >= base && offset < base + width) return r; } return NULL; } -static bool ignore_register(uint32_t offset) +static bool ignore_register(uint32_t offset, uint32_t mmio_base) { for (const struct named_register *r = ignore_registers; r->name; r++) { unsigned int width = r->count ? 4*r->count : 4; - if (offset >= r->offset && offset < r->offset + width) + uint32_t base; + + base = r->offset; + if (r->relative) + base += mmio_base; + + if (offset >= base && offset < base + width) return true; } @@ -248,6 +240,7 @@ static void tmpl_regs(int fd, { const unsigned int gen_bit = 1 << intel_gen(intel_get_drm_devid(fd)); const unsigned int engine_bit = ENGINE(e->class, e->instance); + const uint32_t mmio_base = gem_engine_mmio_base(fd, e->name); unsigned int regs_size; uint32_t *regs; @@ -259,12 +252,20 @@ static void tmpl_regs(int fd, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU); for (const struct named_register *r = nonpriv_registers; r->name; r++) { + uint32_t offset; + if (!(r->engine_mask & engine_bit)) continue; if (!(r->gen_mask & gen_bit)) continue; - for (unsigned count = r->count ?: 1, offset = r->offset; - count--; offset += 4) { + if (r->relative && !mmio_base) + continue; + + offset = r->offset; + if (r->relative) + offset += mmio_base; + + for (unsigned count = r->count ?: 1; count--; offset += 4) { uint32_t x = value; if (r->write_mask) x &= r->write_mask; @@ -284,6 +285,7 @@ static uint32_t read_regs(int fd, const unsigned int gen = intel_gen(intel_get_drm_devid(fd)); const unsigned int gen_bit = 1 << gen; const unsigned int engine_bit = ENGINE(e->class, e->instance); + const uint32_t mmio_base = gem_engine_mmio_base(fd, e->name); const bool r64b = gen >= 8; struct drm_i915_gem_exec_object2 obj[2]; struct drm_i915_gem_relocation_entry *reloc; @@ -311,13 +313,20 @@ static uint32_t read_regs(int fd, n = 0; for (const struct named_register *r = nonpriv_registers; r->name; r++) { + uint32_t offset; + if (!(r->engine_mask & engine_bit)) continue; if (!(r->gen_mask & gen_bit)) continue; + if (r->relative && !mmio_base) + continue; + + offset = r->offset; + if (r->relative) + offset += mmio_base; - for (unsigned count = r->count ?: 1, offset = r->offset; - count--; offset += 4) { + for (unsigned count = r->count ?: 1; count--; offset += 4) { *b++ = 0x24 << 23 | (1 + r64b); /* SRM */ *b++ = offset; reloc[n].target_handle = obj[0].handle; @@ -357,6 +366,7 @@ static void write_regs(int fd, { const unsigned int gen_bit = 1 << intel_gen(intel_get_drm_devid(fd)); const unsigned int engine_bit = ENGINE(e->class, e->instance); + const uint32_t mmio_base = gem_engine_mmio_base(fd, e->name); struct drm_i915_gem_exec_object2 obj; struct drm_i915_gem_execbuffer2 execbuf; unsigned int batch_size; @@ -372,12 +382,20 @@ static void write_regs(int fd, gem_set_domain(fd, obj.handle, I915_GEM_DOMAIN_CPU, I915_GEM_DOMAIN_CPU); for (const struct named_register *r = nonpriv_registers; r->name; r++) { + uint32_t offset; + if (!(r->engine_mask & engine_bit)) continue; if (!(r->gen_mask & gen_bit)) continue; - for (unsigned count = r->count ?: 1, offset = r->offset; - count--; offset += 4) { + if (r->relative && !mmio_base) + continue; + + offset = r->offset; + if (r->relative) + offset += mmio_base; + + for (unsigned count = r->count ?: 1; count--; offset += 4) { uint32_t x = value; if (r->write_mask) x &= r->write_mask; @@ -410,6 +428,7 @@ static void restore_regs(int fd, const unsigned int gen = intel_gen(intel_get_drm_devid(fd)); const unsigned int gen_bit = 1 << gen; const unsigned int engine_bit = ENGINE(e->class, e->instance); + const uint32_t mmio_base = gem_engine_mmio_base(fd, e->name); const bool r64b = gen >= 8; struct drm_i915_gem_exec_object2 obj[2]; struct drm_i915_gem_execbuffer2 execbuf; @@ -437,13 +456,20 @@ static void restore_regs(int fd, n = 0; for (const struct named_register *r = nonpriv_registers; r->name; r++) { + uint32_t offset; + if (!(r->engine_mask & engine_bit)) continue; if (!(r->gen_mask & gen_bit)) continue; + if (r->relative && !mmio_base) + continue; + + offset = r->offset; + if (r->relative) + offset += mmio_base; - for (unsigned count = r->count ?: 1, offset = r->offset; - count--; offset += 4) { + for (unsigned count = r->count ?: 1; count--; offset += 4) { *b++ = 0x29 << 23 | (1 + r64b); /* LRM */ *b++ = offset; reloc[n].target_handle = obj[0].handle; @@ -479,6 +505,7 @@ static void dump_regs(int fd, const int gen = intel_gen(intel_get_drm_devid(fd)); const unsigned int gen_bit = 1 << gen; const unsigned int engine_bit = ENGINE(e->class, e->instance); + const uint32_t mmio_base = gem_engine_mmio_base(fd, e->name); unsigned int regs_size; uint32_t *out; @@ -489,26 +516,36 @@ static void dump_regs(int fd, gem_set_domain(fd, regs, I915_GEM_DOMAIN_CPU, 0); for (const struct named_register *r = nonpriv_registers; r->name; r++) { + uint32_t offset; + if (!(r->engine_mask & engine_bit)) continue; if (!(r->gen_mask & gen_bit)) continue; + if (r->relative && !mmio_base) + continue; + + offset = r->offset; + if (r->relative) + offset += mmio_base; if (r->count <= 1) { igt_debug("0x%04x (%s): 0x%08x\n", - r->offset, r->name, out[r->offset/4]); + offset, r->name, out[offset / 4]); } else { for (unsigned x = 0; x < r->count; x++) igt_debug("0x%04x (%s[%d]): 0x%08x\n", - r->offset+4*x, r->name, x, - out[r->offset/4 + x]); + offset + 4 * x, r->name, x, + out[offset / 4 + x]); } } munmap(out, regs_size); } -static void compare_regs(int fd, uint32_t A, uint32_t B, const char *who) +static void compare_regs(int fd, const struct intel_execution_engine2 *e, + uint32_t A, uint32_t B, const char *who) { + const uint32_t mmio_base = gem_engine_mmio_base(fd, e->name); unsigned int num_errors; unsigned int regs_size; uint32_t *a, *b; @@ -532,11 +569,11 @@ static void compare_regs(int fd, uint32_t A, uint32_t B, const char *who) if (a[n] == b[n]) continue; - if (ignore_register(offset)) + if (ignore_register(offset, mmio_base)) continue; mask = ~0u; - r = lookup_register(offset); + r = lookup_register(offset, mmio_base); if (r && r->masked) mask >>= 16; if (r && r->ignore_bits) @@ -547,7 +584,7 @@ static void compare_regs(int fd, uint32_t A, uint32_t B, const char *who) igt_warn("Register 0x%04x (%s): A=%08x B=%08x\n", offset, - register_name(offset, buf, sizeof(buf)), + register_name(offset, mmio_base, buf, sizeof(buf)), a[n] & mask, b[n] & mask); num_errors++; } @@ -638,7 +675,7 @@ static void nonpriv(int fd, igt_spin_free(fd, spin); - compare_regs(fd, tmpl, regs[1], "nonpriv read/writes"); + compare_regs(fd, e, tmpl, regs[1], "nonpriv read/writes"); for (int n = 0; n < ARRAY_SIZE(regs); n++) gem_close(fd, regs[n]); @@ -708,8 +745,9 @@ static void isolation(int fd, igt_spin_free(fd, spin); if (!(flags & DIRTY1)) - compare_regs(fd, regs[0], tmp, "two reads of the same ctx"); - compare_regs(fd, regs[0], regs[1], "two virgin contexts"); + compare_regs(fd, e, regs[0], tmp, + "two reads of the same ctx"); + compare_regs(fd, e, regs[0], regs[1], "two virgin contexts"); for (int n = 0; n < ARRAY_SIZE(ctx); n++) { gem_close(fd, regs[n]); @@ -829,13 +867,13 @@ static void preservation(int fd, char buf[80]; snprintf(buf, sizeof(buf), "dirty %x context\n", values[v]); - compare_regs(fd, regs[v][0], regs[v][1], buf); + compare_regs(fd, e, regs[v][0], regs[v][1], buf); gem_close(fd, regs[v][0]); gem_close(fd, regs[v][1]); gem_context_destroy(fd, ctx[v]); } - compare_regs(fd, regs[num_values][0], regs[num_values][1], "clean"); + compare_regs(fd, e, regs[num_values][0], regs[num_values][1], "clean"); gem_context_destroy(fd, ctx[num_values]); }