1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
|
From 8bbabd0adfc498df205067032efc08d91df6305b Mon Sep 17 00:00:00 2001
From: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Date: Tue, 19 Jun 2018 10:27:53 -0400
Subject: [PATCH 4751/5725] drm/amdgpu: Add parsing SQ_EDC_INFO to SQ IH v3.
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Access to SQ_EDC_INFO requires selecting register instance and
hence mutex lock when accessing GRBM_GFX_INDEX for which a work
is schedueled from IH. But SQ interrupt can be raised on many instances
at once which means queuing work will usually succeed for the first one
but fail for the rest since the work takes time to process. To avoid
losing info about other interrupt instances call the parsing function
directly from high IRQ when current work hasn't finished and avoid
accessing SQ_EDC_INFO in that case.
v2:
Simplify high IRQ and BH handlers synchronization using work_pending.
Remove {READ,WRITE}_ONCE notations since smp_{r,w}mb are implicit
compiler barriers.
v3:
Remove exlicit memory barriers as scedule_work has r/w barriers.
Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 7 +++
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 82 +++++++++++++++++++++++++++++------
2 files changed, 76 insertions(+), 13 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index c55e675..3b456ce 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -951,6 +951,11 @@ struct amdgpu_ngg {
bool init;
};
+struct sq_work {
+ struct work_struct work;
+ unsigned ih_data;
+};
+
struct amdgpu_gfx {
struct mutex gpu_clock_mutex;
struct amdgpu_gfx_config config;
@@ -991,6 +996,8 @@ struct amdgpu_gfx {
struct amdgpu_irq_src priv_inst_irq;
struct amdgpu_irq_src cp_ecc_error_irq;
struct amdgpu_irq_src sq_irq;
+ struct sq_work sq_work;
+
/* gfx status */
uint32_t gfx_current_status;
/* ce ram size*/
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index ad28dd9..294fa59 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -704,6 +704,17 @@ static const u32 stoney_mgcg_cgcg_init[] =
mmCGTS_SM_CTRL_REG, 0xffffffff, 0x96940200,
};
+
+static const char * const sq_edc_source_names[] = {
+ "SQ_EDC_INFO_SOURCE_INVALID: No EDC error has occurred",
+ "SQ_EDC_INFO_SOURCE_INST: EDC source is Instruction Fetch",
+ "SQ_EDC_INFO_SOURCE_SGPR: EDC source is SGPR or SQC data return",
+ "SQ_EDC_INFO_SOURCE_VGPR: EDC source is VGPR",
+ "SQ_EDC_INFO_SOURCE_LDS: EDC source is LDS",
+ "SQ_EDC_INFO_SOURCE_GDS: EDC source is GDS",
+ "SQ_EDC_INFO_SOURCE_TA: EDC source is TA",
+};
+
static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev);
static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev);
static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev);
@@ -2005,6 +2016,8 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id,
return 0;
}
+static void gfx_v8_0_sq_irq_work_func(struct work_struct *work);
+
static int gfx_v8_0_sw_init(void *handle)
{
int i, j, k, r, ring_id;
@@ -2068,6 +2081,8 @@ static int gfx_v8_0_sw_init(void *handle)
return r;
}
+ INIT_WORK(&adev->gfx.sq_work.work, gfx_v8_0_sq_irq_work_func);
+
adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE;
gfx_v8_0_scratch_init(adev);
@@ -6967,14 +6982,11 @@ static int gfx_v8_0_cp_ecc_error_irq(struct amdgpu_device *adev,
return 0;
}
-static int gfx_v8_0_sq_irq(struct amdgpu_device *adev,
- struct amdgpu_irq_src *source,
- struct amdgpu_iv_entry *entry)
+static void gfx_v8_0_parse_sq_irq(struct amdgpu_device *adev, unsigned ih_data)
{
- u8 enc, se_id;
+ u32 enc, se_id, sh_id, cu_id;
char type[20];
- unsigned ih_data = entry->src_data[0];
-
+ int sq_edc_source = -1;
enc = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_CMN, ENCODING);
se_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_CMN, SE_ID);
@@ -7000,6 +7012,24 @@ static int gfx_v8_0_sq_irq(struct amdgpu_device *adev,
case 1:
case 2:
+ cu_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, CU_ID);
+ sh_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SH_ID);
+
+ /*
+ * This function can be called either directly from ISR
+ * or from BH in which case we can access SQ_EDC_INFO
+ * instance
+ */
+ if (in_task()) {
+ mutex_lock(&adev->grbm_idx_mutex);
+ gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id);
+
+ sq_edc_source = REG_GET_FIELD(RREG32(mmSQ_EDC_INFO), SQ_EDC_INFO, SOURCE);
+
+ gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff);
+ mutex_unlock(&adev->grbm_idx_mutex);
+ }
+
if (enc == 1)
sprintf(type, "instruction intr");
else
@@ -7007,20 +7037,46 @@ static int gfx_v8_0_sq_irq(struct amdgpu_device *adev,
DRM_INFO(
"SQ %s detected: "
- "se_id %d, cu_id %d, simd_id %d, wave_id %d, vm_id %d\n"
- "trap %s, sh_id %d. ",
- type, se_id,
- REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, CU_ID),
+ "se_id %d, sh_id %d, cu_id %d, simd_id %d, wave_id %d, vm_id %d "
+ "trap %s, sq_ed_info.source %s.\n",
+ type, se_id, sh_id, cu_id,
REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SIMD_ID),
REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, WAVE_ID),
REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, VM_ID),
REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, PRIV) ? "true" : "false",
- REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SH_ID)
- );
+ (sq_edc_source != -1) ? sq_edc_source_names[sq_edc_source] : "unavailable"
+ );
break;
default:
DRM_ERROR("SQ invalid encoding type\n.");
- return -EINVAL;
+ }
+}
+
+static void gfx_v8_0_sq_irq_work_func(struct work_struct *work)
+{
+
+ struct amdgpu_device *adev = container_of(work, struct amdgpu_device, gfx.sq_work.work);
+ struct sq_work *sq_work = container_of(work, struct sq_work, work);
+
+ gfx_v8_0_parse_sq_irq(adev, sq_work->ih_data);
+}
+
+static int gfx_v8_0_sq_irq(struct amdgpu_device *adev,
+ struct amdgpu_irq_src *source,
+ struct amdgpu_iv_entry *entry)
+{
+ unsigned ih_data = entry->src_data[0];
+
+ /*
+ * Try to submit work so SQ_EDC_INFO can be accessed from
+ * BH. If previous work submission hasn't finished yet
+ * just print whatever info is possible directly from the ISR.
+ */
+ if (work_pending(&adev->gfx.sq_work.work)) {
+ gfx_v8_0_parse_sq_irq(adev, ih_data);
+ } else {
+ adev->gfx.sq_work.ih_data = ih_data;
+ schedule_work(&adev->gfx.sq_work.work);
}
return 0;
--
2.7.4
|
