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
|
commit 034994cfffbb2371b720e3f49378031ebc12645e
Author: Eric Anholt <eric@anholt.net>
Date: Thu Oct 2 12:24:47 2008 -0700
drm: Clean up many sparse warnings in i915.
Signed-off-by: Eric Anholt <eric@anholt.net>
diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c
index dbd3f49..814cc12 100644
--- a/drivers/gpu/drm/i915/i915_dma.c
+++ b/drivers/gpu/drm/i915/i915_dma.c
@@ -76,7 +76,7 @@ int i915_wait_ring(struct drm_device * dev, int n, const char *caller)
* Sets up the hardware status page for devices that need a physical address
* in the register.
*/
-int i915_init_phys_hws(struct drm_device *dev)
+static int i915_init_phys_hws(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
/* Program Hardware Status Page */
@@ -101,7 +101,7 @@ int i915_init_phys_hws(struct drm_device *dev)
* Frees the hardware status page, whether it's a physical address or a virtual
* address set up by the X Server.
*/
-void i915_free_hws(struct drm_device *dev)
+static void i915_free_hws(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
if (dev_priv->status_page_dmah) {
@@ -145,8 +145,8 @@ static int i915_dma_cleanup(struct drm_device * dev)
if (dev_priv->ring.virtual_start) {
drm_core_ioremapfree(&dev_priv->ring.map, dev);
- dev_priv->ring.virtual_start = 0;
- dev_priv->ring.map.handle = 0;
+ dev_priv->ring.virtual_start = NULL;
+ dev_priv->ring.map.handle = NULL;
dev_priv->ring.map.size = 0;
}
@@ -827,9 +827,7 @@ int i915_driver_load(struct drm_device *dev, unsigned long flags)
base = drm_get_resource_start(dev, mmio_bar);
size = drm_get_resource_len(dev, mmio_bar);
- ret = drm_addmap(dev, base, size, _DRM_REGISTERS,
- _DRM_KERNEL | _DRM_DRIVER,
- &dev_priv->mmio_map);
+ dev_priv->regs = ioremap(base, size);
i915_gem_load(dev);
@@ -867,8 +865,8 @@ int i915_driver_unload(struct drm_device *dev)
i915_free_hws(dev);
- if (dev_priv->mmio_map)
- drm_rmmap(dev, dev_priv->mmio_map);
+ if (dev_priv->regs != NULL)
+ iounmap(dev_priv->regs);
intel_opregion_free(dev);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 8547f0a..b184d54 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -110,8 +110,8 @@ struct intel_opregion {
typedef struct drm_i915_private {
struct drm_device *dev;
+ void __iomem *regs;
drm_local_map_t *sarea;
- drm_local_map_t *mmio_map;
drm_i915_sarea_t *sarea_priv;
drm_i915_ring_buffer_t ring;
@@ -553,12 +553,12 @@ extern void opregion_enable_asle(struct drm_device *dev);
LOCK_TEST_WITH_RETURN(dev, file_priv); \
} while (0)
-#define I915_READ(reg) DRM_READ32(dev_priv->mmio_map, (reg))
-#define I915_WRITE(reg,val) DRM_WRITE32(dev_priv->mmio_map, (reg), (val))
-#define I915_READ16(reg) DRM_READ16(dev_priv->mmio_map, (reg))
-#define I915_WRITE16(reg,val) DRM_WRITE16(dev_priv->mmio_map, (reg), (val))
-#define I915_READ8(reg) DRM_READ8(dev_priv->mmio_map, (reg))
-#define I915_WRITE8(reg,val) DRM_WRITE8(dev_priv->mmio_map, (reg), (val))
+#define I915_READ(reg) readl(dev_priv->regs + (reg))
+#define I915_WRITE(reg,val) writel(val, dev_priv->regs + (reg))
+#define I915_READ16(reg) readw(dev_priv->regs + (reg))
+#define I915_WRITE16(reg,val) writel(val, dev_priv->regs + (reg))
+#define I915_READ8(reg) readb(dev_priv->regs + (reg))
+#define I915_WRITE8(reg,val) writeb(val, dev_priv->regs + (reg))
#define I915_VERBOSE 0
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 6ecfd10..6a89449 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -176,7 +176,8 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj,
ssize_t remain;
loff_t offset;
char __user *user_data;
- char *vaddr;
+ char __iomem *vaddr;
+ char *vaddr_atomic;
int i, o, l;
int ret = 0;
unsigned long pfn;
@@ -219,16 +220,20 @@ i915_gem_gtt_pwrite(struct drm_device *dev, struct drm_gem_object *obj,
pfn = (dev->agp->base >> PAGE_SHIFT) + i;
#ifdef CONFIG_HIGHMEM
- /* kmap_atomic can't map IO pages on non-HIGHMEM kernels
+ /* This is a workaround for the low performance of iounmap
+ * (approximate 10% cpu cost on normal 3D workloads).
+ * kmap_atomic on HIGHMEM kernels happens to let us map card
+ * memory without taking IPIs. When the vmap rework lands
+ * we should be able to dump this hack.
*/
- vaddr = kmap_atomic_pfn(pfn, KM_USER0);
+ vaddr_atomic = kmap_atomic_pfn(pfn, KM_USER0);
#if WATCH_PWRITE
DRM_INFO("pwrite i %d o %d l %d pfn %ld vaddr %p\n",
- i, o, l, pfn, vaddr);
+ i, o, l, pfn, vaddr_atomic);
#endif
- unwritten = __copy_from_user_inatomic_nocache(vaddr + o,
+ unwritten = __copy_from_user_inatomic_nocache(vaddr_atomic + o,
user_data, l);
- kunmap_atomic(vaddr, KM_USER0);
+ kunmap_atomic(vaddr_atomic, KM_USER0);
if (unwritten)
#endif /* CONFIG_HIGHMEM */
@@ -271,7 +276,7 @@ fail:
return ret;
}
-int
+static int
i915_gem_shmem_pwrite(struct drm_device *dev, struct drm_gem_object *obj,
struct drm_i915_gem_pwrite *args,
struct drm_file *file_priv)
@@ -587,7 +592,7 @@ i915_add_request(struct drm_device *dev, uint32_t flush_domains)
* Ensures that all commands in the ring are finished
* before signalling the CPU
*/
-uint32_t
+static uint32_t
i915_retire_commands(struct drm_device *dev)
{
drm_i915_private_t *dev_priv = dev->dev_private;
@@ -734,7 +739,7 @@ i915_gem_retire_work_handler(struct work_struct *work)
* Waits for a sequence number to be signaled, and cleans up the
* request and object lists appropriately for that event.
*/
-int
+static int
i915_wait_request(struct drm_device *dev, uint32_t seqno)
{
drm_i915_private_t *dev_priv = dev->dev_private;
@@ -1483,7 +1488,7 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
struct drm_i915_gem_object *obj_priv = obj->driver_private;
int i, ret;
uint32_t last_reloc_offset = -1;
- void *reloc_page = NULL;
+ void __iomem *reloc_page = NULL;
/* Choose the GTT offset for our buffer and put it there. */
ret = i915_gem_object_pin(obj, (uint32_t) entry->alignment);
@@ -1500,8 +1505,8 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
for (i = 0; i < entry->relocation_count; i++) {
struct drm_gem_object *target_obj;
struct drm_i915_gem_object *target_obj_priv;
- uint32_t reloc_val, reloc_offset, *reloc_entry;
- int ret;
+ uint32_t reloc_val, reloc_offset;
+ uint32_t __iomem *reloc_entry;
ret = copy_from_user(&reloc, relocs + i, sizeof(reloc));
if (ret != 0) {
@@ -1624,7 +1629,7 @@ i915_gem_object_pin_and_relocate(struct drm_gem_object *obj,
}
}
- reloc_entry = (uint32_t *)((char *)reloc_page +
+ reloc_entry = (uint32_t __iomem *)(reloc_page +
(reloc_offset & (PAGE_SIZE - 1)));
reloc_val = target_obj_priv->gtt_offset + reloc.delta;
|
