* Next batch of almost reviewed patches, up to the VMA leak fix
@ 2016-08-03 13:42 Chris Wilson
2016-08-03 13:42 ` [PATCH 01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Chris Wilson
` (29 more replies)
0 siblings, 30 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
I was running on an old Pineview box and for some reason decided to
check UXA performance. That lead to the discovery that we were not
removing a conflicting uncached mapping of the global GTT and
required rearranging the global GTT initialisation to fix the
regression. Everything but
[PATCH 02/28] drm/i915: Split GGTT initialisation between probing and
[PATCH 03/28] drm/i915: Keep ggtt->probe() as a local
[PATCH 04/28] drm/i915: Update GGTT initialisation functions to take
[PATCH 05/28] drm/i915: Split early global GTT initialisation
is reviewed and ready to be merged.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* [PATCH 01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
@ 2016-08-03 13:42 ` Chris Wilson
2016-08-03 13:42 ` [PATCH 02/28] drm/i915: Split GGTT initialisation between probing and setup Chris Wilson
` (28 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
As we can now have multiple VMA inside the global GTT (with partial
mappings, rotations, etc), it is no longer true that there may just be a
single GGTT entry and so we should walk the full vma_list to count up
the actual usage. In addition to unifying the two walkers, switch from
multiplying the object size for each vma to summing the bound vma sizes.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 42 +++++++++++++------------------------
1 file changed, 15 insertions(+), 27 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 410dc35a7a8a..e7724be9f4db 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -335,6 +335,7 @@ static int per_file_stats(int id, void *ptr, void *data)
struct drm_i915_gem_object *obj = ptr;
struct file_stats *stats = data;
struct i915_vma *vma;
+ int bound = 0;
stats->count++;
stats->total += obj->base.size;
@@ -342,41 +343,28 @@ static int per_file_stats(int id, void *ptr, void *data)
if (obj->base.name || obj->base.dma_buf)
stats->shared += obj->base.size;
- if (USES_FULL_PPGTT(obj->base.dev)) {
- list_for_each_entry(vma, &obj->vma_list, obj_link) {
- struct i915_hw_ppgtt *ppgtt;
+ list_for_each_entry(vma, &obj->vma_list, obj_link) {
+ if (!drm_mm_node_allocated(&vma->node))
+ continue;
- if (!drm_mm_node_allocated(&vma->node))
- continue;
+ bound++;
- if (vma->is_ggtt) {
- stats->global += obj->base.size;
- continue;
- }
+ if (vma->is_ggtt) {
+ stats->global += vma->node.size;
+ } else {
+ struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vma->vm);
- ppgtt = container_of(vma->vm, struct i915_hw_ppgtt, base);
if (ppgtt->file_priv != stats->file_priv)
continue;
-
- if (obj->active) /* XXX per-vma statistic */
- stats->active += obj->base.size;
- else
- stats->inactive += obj->base.size;
-
- return 0;
- }
- } else {
- if (i915_gem_obj_ggtt_bound(obj)) {
- stats->global += obj->base.size;
- if (obj->active)
- stats->active += obj->base.size;
- else
- stats->inactive += obj->base.size;
- return 0;
}
+
+ if (obj->active) /* XXX per-vma statistic */
+ stats->active += vma->node.size;
+ else
+ stats->inactive += vma->node.size;
}
- if (!list_empty(&obj->global_list))
+ if (!bound)
stats->unbound += obj->base.size;
return 0;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 02/28] drm/i915: Split GGTT initialisation between probing and setup
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
2016-08-03 13:42 ` [PATCH 01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Chris Wilson
@ 2016-08-03 13:42 ` Chris Wilson
2016-08-03 13:54 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local Chris Wilson
` (27 subsequent siblings)
29 siblings, 1 reply; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
In order to handle conflicting drivers (i.e. vgacon) having a different
setup of hardware, we have to remove those other drivers before we try
to setup our own mappings. This requires us to split GGTT initialisation
between probing for the hardware location (part of the PCI BAR) and
later establishing the kernel resources for it.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_drv.c | 18 +++++++++++-------
drivers/gpu/drm/i915/i915_gem_gtt.c | 34 +++++++++++++++++++++++-----------
drivers/gpu/drm/i915/i915_gem_gtt.h | 1 +
3 files changed, 35 insertions(+), 18 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 83afdd0597b5..8996c2a6cd07 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1004,16 +1004,10 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
intel_sanitize_options(dev_priv);
- ret = i915_ggtt_init_hw(dev);
+ ret = i915_ggtt_probe_hw(dev);
if (ret)
return ret;
- ret = i915_ggtt_enable_hw(dev);
- if (ret) {
- DRM_ERROR("failed to enable GGTT\n");
- goto out_ggtt;
- }
-
/* WARNING: Apparently we must kick fbdev drivers before vgacon,
* otherwise the vga fbdev driver falls over. */
ret = i915_kick_out_firmware_fb(dev_priv);
@@ -1028,6 +1022,16 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
goto out_ggtt;
}
+ ret = i915_ggtt_init_hw(dev);
+ if (ret)
+ return ret;
+
+ ret = i915_ggtt_enable_hw(dev);
+ if (ret) {
+ DRM_ERROR("failed to enable GGTT\n");
+ goto out_ggtt;
+ }
+
pci_set_master(dev->pdev);
/* overlay on gen2 is broken and can't address above 1G */
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 671b1cab5e54..3fc77776d89b 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3208,10 +3208,10 @@ static void i915_gmch_remove(struct i915_address_space *vm)
}
/**
- * i915_ggtt_init_hw - Initialize GGTT hardware
+ * i915_ggtt_probe_hw - Probe GGTT hardware location
* @dev: DRM device
*/
-int i915_ggtt_init_hw(struct drm_device *dev)
+int i915_ggtt_probe_hw(struct drm_device *dev)
{
struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
@@ -3254,14 +3254,6 @@ int i915_ggtt_init_hw(struct drm_device *dev)
ggtt->mappable_end = min(ggtt->mappable_end, ggtt->base.total);
}
- /*
- * Initialise stolen early so that we may reserve preallocated
- * objects for the BIOS to KMS transition.
- */
- ret = i915_gem_init_stolen(dev);
- if (ret)
- goto out_gtt_cleanup;
-
/* GMADR is the PCI mmio aperture into the global GTT. */
DRM_INFO("Memory usable by graphics device = %lluM\n",
ggtt->base.total >> 20);
@@ -3273,10 +3265,30 @@ int i915_ggtt_init_hw(struct drm_device *dev)
#endif
return 0;
+}
+
+/**
+ * i915_ggtt_init_hw - Initialize GGTT hardware
+ * @dev: DRM device
+ */
+int i915_ggtt_init_hw(struct drm_device *dev)
+{
+ struct drm_i915_private *dev_priv = to_i915(dev);
+ struct i915_ggtt *ggtt = &dev_priv->ggtt;
+ int ret;
+
+ /*
+ * Initialise stolen early so that we may reserve preallocated
+ * objects for the BIOS to KMS transition.
+ */
+ ret = i915_gem_init_stolen(dev);
+ if (ret)
+ goto out_gtt_cleanup;
+
+ return 0;
out_gtt_cleanup:
ggtt->base.cleanup(&ggtt->base);
-
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index c4a657973cc6..5b6744a5c944 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -521,6 +521,7 @@ i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
px_dma(ppgtt->base.scratch_pd);
}
+int i915_ggtt_probe_hw(struct drm_device *dev);
int i915_ggtt_init_hw(struct drm_device *dev);
int i915_ggtt_enable_hw(struct drm_device *dev);
void i915_gem_init_ggtt(struct drm_device *dev);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
2016-08-03 13:42 ` [PATCH 01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Chris Wilson
2016-08-03 13:42 ` [PATCH 02/28] drm/i915: Split GGTT initialisation between probing and setup Chris Wilson
@ 2016-08-03 13:42 ` Chris Wilson
2016-08-03 14:00 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 04/28] " Chris Wilson
` (26 subsequent siblings)
29 siblings, 1 reply; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
The ggtt->probe() vfunc is only called once during GGTT initialisation.
We can keep it in a local for the function, sparing a few bytes in our
i915_ggtt persistent struct.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_gem_gtt.c | 15 ++++++++-------
drivers/gpu/drm/i915/i915_gem_gtt.h | 2 --
2 files changed, 8 insertions(+), 9 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 3fc77776d89b..70178a77d0c8 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3215,13 +3215,17 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
{
struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
+ int (*probe)(struct i915_ggtt *ggtt);
int ret;
+ ggtt->base.dev = dev;
+ ggtt->base.is_ggtt = true;
+
if (INTEL_INFO(dev)->gen <= 5) {
- ggtt->probe = i915_gmch_probe;
+ probe = i915_gmch_probe;
ggtt->base.cleanup = i915_gmch_remove;
} else if (INTEL_INFO(dev)->gen < 8) {
- ggtt->probe = gen6_gmch_probe;
+ probe = gen6_gmch_probe;
ggtt->base.cleanup = gen6_gmch_remove;
if (HAS_EDRAM(dev))
@@ -3235,14 +3239,11 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
else
ggtt->base.pte_encode = snb_pte_encode;
} else {
- ggtt->probe = gen8_gmch_probe;
+ probe = gen8_gmch_probe;
ggtt->base.cleanup = gen6_gmch_remove;
}
- ggtt->base.dev = dev;
- ggtt->base.is_ggtt = true;
-
- ret = ggtt->probe(ggtt);
+ ret = probe(ggtt);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 5b6744a5c944..defb274f3ba5 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -365,8 +365,6 @@ struct i915_ggtt {
bool do_idle_maps;
int mtrr;
-
- int (*probe)(struct i915_ggtt *ggtt);
};
struct i915_hw_ppgtt {
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 04/28] drm/i915: Update GGTT initialisation functions to take drm_i915_private
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (2 preceding siblings ...)
2016-08-03 13:42 ` [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local Chris Wilson
@ 2016-08-03 13:42 ` Chris Wilson
2016-08-03 14:05 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 05/28] drm/i915: Split early global GTT initialisation Chris Wilson
` (25 subsequent siblings)
29 siblings, 1 reply; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
Since these are internal functions they operate on drm_i915_private and
not the drm_device being passed in. So pass in the drm_i915_private
instead, and remove one layer of dancing. No space wins here, just
conforming to the norm in function parameters.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_drv.c | 12 ++++-----
drivers/gpu/drm/i915/i915_gem.c | 2 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 54 +++++++++++++++++--------------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 10 +++----
4 files changed, 37 insertions(+), 41 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 8996c2a6cd07..ed52f61d8560 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1004,7 +1004,7 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
intel_sanitize_options(dev_priv);
- ret = i915_ggtt_probe_hw(dev);
+ ret = i915_ggtt_probe_hw(dev_priv);
if (ret)
return ret;
@@ -1022,11 +1022,11 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
goto out_ggtt;
}
- ret = i915_ggtt_init_hw(dev);
+ ret = i915_ggtt_init_hw(dev_priv);
if (ret)
return ret;
- ret = i915_ggtt_enable_hw(dev);
+ ret = i915_ggtt_enable_hw(dev_priv);
if (ret) {
DRM_ERROR("failed to enable GGTT\n");
goto out_ggtt;
@@ -1104,7 +1104,7 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
return 0;
out_ggtt:
- i915_ggtt_cleanup_hw(dev);
+ i915_ggtt_cleanup_hw(dev_priv);
return ret;
}
@@ -1124,7 +1124,7 @@ static void i915_driver_cleanup_hw(struct drm_i915_private *dev_priv)
pm_qos_remove_request(&dev_priv->pm_qos);
arch_phys_wc_del(ggtt->mtrr);
io_mapping_free(ggtt->mappable);
- i915_ggtt_cleanup_hw(dev);
+ i915_ggtt_cleanup_hw(dev_priv);
}
/**
@@ -1570,7 +1570,7 @@ static int i915_drm_resume(struct drm_device *dev)
disable_rpm_wakeref_asserts(dev_priv);
- ret = i915_ggtt_enable_hw(dev);
+ ret = i915_ggtt_enable_hw(dev_priv);
if (ret)
DRM_ERROR("failed to re-enable GGTT\n");
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index bc871449751a..653197811d43 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4543,7 +4543,7 @@ int i915_gem_init(struct drm_device *dev)
intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
i915_gem_init_userptr(dev_priv);
- i915_gem_init_ggtt(dev);
+ i915_gem_init_ggtt(dev_priv);
ret = i915_gem_context_init(dev);
if (ret)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 70178a77d0c8..23a4960deb64 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2746,7 +2746,7 @@ static void i915_gtt_color_adjust(struct drm_mm_node *node,
*end -= 4096;
}
-static int i915_gem_setup_global_gtt(struct drm_device *dev,
+static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
u64 start,
u64 mappable_end,
u64 end)
@@ -2760,7 +2760,6 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
* aperture. One page should be enough to keep any prefetching inside
* of the aperture.
*/
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
struct drm_mm_node *entry;
struct drm_i915_gem_object *obj;
@@ -2781,7 +2780,7 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
if (ret)
return ret;
- if (!HAS_LLC(dev))
+ if (!HAS_LLC(dev_priv))
ggtt->base.mm.color_adjust = i915_gtt_color_adjust;
/* Mark any preallocated objects as occupied */
@@ -2813,14 +2812,14 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
/* And finally clear the reserved guard page */
ggtt->base.clear_range(&ggtt->base, end - PAGE_SIZE, PAGE_SIZE, true);
- if (USES_PPGTT(dev) && !USES_FULL_PPGTT(dev)) {
+ if (USES_PPGTT(dev_priv) && !USES_FULL_PPGTT(dev_priv)) {
struct i915_hw_ppgtt *ppgtt;
ppgtt = kzalloc(sizeof(*ppgtt), GFP_KERNEL);
if (!ppgtt)
return -ENOMEM;
- ret = __hw_ppgtt_init(dev, ppgtt);
+ ret = __hw_ppgtt_init(&dev_priv->drm, ppgtt);
if (ret) {
ppgtt->base.cleanup(&ppgtt->base);
kfree(ppgtt);
@@ -2851,23 +2850,22 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
/**
* i915_gem_init_ggtt - Initialize GEM for Global GTT
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-void i915_gem_init_ggtt(struct drm_device *dev)
+void i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
- i915_gem_setup_global_gtt(dev, 0, ggtt->mappable_end, ggtt->base.total);
+ i915_gem_setup_global_gtt(dev_priv,
+ 0, ggtt->mappable_end, ggtt->base.total);
}
/**
* i915_ggtt_cleanup_hw - Clean up GGTT hardware initialization
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-void i915_ggtt_cleanup_hw(struct drm_device *dev)
+void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
if (dev_priv->mm.aliasing_ppgtt) {
@@ -2876,7 +2874,7 @@ void i915_ggtt_cleanup_hw(struct drm_device *dev)
ppgtt->base.cleanup(&ppgtt->base);
}
- i915_gem_cleanup_stolen(dev);
+ i915_gem_cleanup_stolen(&dev_priv->drm);
if (drm_mm_initialized(&ggtt->base.mm)) {
intel_vgt_deballoon(dev_priv);
@@ -3209,32 +3207,31 @@ static void i915_gmch_remove(struct i915_address_space *vm)
/**
* i915_ggtt_probe_hw - Probe GGTT hardware location
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-int i915_ggtt_probe_hw(struct drm_device *dev)
+int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int (*probe)(struct i915_ggtt *ggtt);
int ret;
- ggtt->base.dev = dev;
+ ggtt->base.dev = &dev_priv->drm;
ggtt->base.is_ggtt = true;
- if (INTEL_INFO(dev)->gen <= 5) {
+ if (INTEL_GEN(dev_priv) <= 5) {
probe = i915_gmch_probe;
ggtt->base.cleanup = i915_gmch_remove;
- } else if (INTEL_INFO(dev)->gen < 8) {
+ } else if (INTEL_GEN(dev_priv) < 8) {
probe = gen6_gmch_probe;
ggtt->base.cleanup = gen6_gmch_remove;
- if (HAS_EDRAM(dev))
+ if (HAS_EDRAM(dev_priv))
ggtt->base.pte_encode = iris_pte_encode;
- else if (IS_HASWELL(dev))
+ else if (IS_HASWELL(dev_priv))
ggtt->base.pte_encode = hsw_pte_encode;
- else if (IS_VALLEYVIEW(dev))
+ else if (IS_VALLEYVIEW(dev_priv))
ggtt->base.pte_encode = byt_pte_encode;
- else if (INTEL_INFO(dev)->gen >= 7)
+ else if (INTEL_GEN(dev_priv) >= 7)
ggtt->base.pte_encode = ivb_pte_encode;
else
ggtt->base.pte_encode = snb_pte_encode;
@@ -3270,11 +3267,10 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
/**
* i915_ggtt_init_hw - Initialize GGTT hardware
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-int i915_ggtt_init_hw(struct drm_device *dev)
+int i915_ggtt_init_hw(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int ret;
@@ -3282,7 +3278,7 @@ int i915_ggtt_init_hw(struct drm_device *dev)
* Initialise stolen early so that we may reserve preallocated
* objects for the BIOS to KMS transition.
*/
- ret = i915_gem_init_stolen(dev);
+ ret = i915_gem_init_stolen(&dev_priv->drm);
if (ret)
goto out_gtt_cleanup;
@@ -3293,9 +3289,9 @@ out_gtt_cleanup:
return ret;
}
-int i915_ggtt_enable_hw(struct drm_device *dev)
+int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv)
{
- if (INTEL_INFO(dev)->gen < 6 && !intel_enable_gtt())
+ if (INTEL_GEN(dev_priv) < 6 && !intel_enable_gtt())
return -EIO;
return 0;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index defb274f3ba5..7db8555aa254 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -519,11 +519,11 @@ i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
px_dma(ppgtt->base.scratch_pd);
}
-int i915_ggtt_probe_hw(struct drm_device *dev);
-int i915_ggtt_init_hw(struct drm_device *dev);
-int i915_ggtt_enable_hw(struct drm_device *dev);
-void i915_gem_init_ggtt(struct drm_device *dev);
-void i915_ggtt_cleanup_hw(struct drm_device *dev);
+int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
+int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
+int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
+void i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
+void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
int i915_ppgtt_init_hw(struct drm_device *dev);
void i915_ppgtt_release(struct kref *kref);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 05/28] drm/i915: Split early global GTT initialisation
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (3 preceding siblings ...)
2016-08-03 13:42 ` [PATCH 04/28] " Chris Wilson
@ 2016-08-03 13:42 ` Chris Wilson
2016-08-03 14:17 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 06/28] drm/i915: Store owning file on the i915_address_space Chris Wilson
` (24 subsequent siblings)
29 siblings, 1 reply; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
Initialising the global GTT is tricky as we wish to use the drm_mm range
manager during the modesetting initialisation (to capture stolen
allocations from the BIOS) before we actually enable GEM. To overcome
this, we currently setup the drm_mm first and then carefully rebind
them.
v2: Fixup after rebasing
v3: GGTT initialisation needs to be split around kicking out conflicts
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
drivers/gpu/drm/i915/i915_drv.c | 19 ---------
drivers/gpu/drm/i915/i915_gem.c | 6 ++-
drivers/gpu/drm/i915/i915_gem_gtt.c | 78 ++++++++++++----------------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 2 +-
drivers/gpu/drm/i915/i915_gem_stolen.c | 17 ++++----
5 files changed, 41 insertions(+), 81 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index ed52f61d8560..50c564089fa8 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -993,8 +993,6 @@ static void intel_sanitize_options(struct drm_i915_private *dev_priv)
static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
{
struct drm_device *dev = &dev_priv->drm;
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
- uint32_t aperture_size;
int ret;
if (i915_inject_load_failure())
@@ -1044,7 +1042,6 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
}
}
-
/* 965GM sometimes incorrectly writes to hardware status page (HWS)
* using 32bit addressing, overwriting memory if HWS is located
* above 4GB.
@@ -1063,19 +1060,6 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
}
}
- aperture_size = ggtt->mappable_end;
-
- ggtt->mappable =
- io_mapping_create_wc(ggtt->mappable_base,
- aperture_size);
- if (!ggtt->mappable) {
- ret = -EIO;
- goto out_ggtt;
- }
-
- ggtt->mtrr = arch_phys_wc_add(ggtt->mappable_base,
- aperture_size);
-
pm_qos_add_request(&dev_priv->pm_qos, PM_QOS_CPU_DMA_LATENCY,
PM_QOS_DEFAULT_VALUE);
@@ -1116,14 +1100,11 @@ out_ggtt:
static void i915_driver_cleanup_hw(struct drm_i915_private *dev_priv)
{
struct drm_device *dev = &dev_priv->drm;
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
if (dev->pdev->msi_enabled)
pci_disable_msi(dev->pdev);
pm_qos_remove_request(&dev_priv->pm_qos);
- arch_phys_wc_del(ggtt->mtrr);
- io_mapping_free(ggtt->mappable);
i915_ggtt_cleanup_hw(dev_priv);
}
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 653197811d43..83e299c429c2 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4543,7 +4543,10 @@ int i915_gem_init(struct drm_device *dev)
intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
i915_gem_init_userptr(dev_priv);
- i915_gem_init_ggtt(dev_priv);
+
+ ret = i915_gem_init_ggtt(dev_priv);
+ if (ret)
+ goto out_unlock;
ret = i915_gem_context_init(dev);
if (ret)
@@ -4634,7 +4637,6 @@ i915_gem_load_init(struct drm_device *dev)
SLAB_HWCACHE_ALIGN,
NULL);
- INIT_LIST_HEAD(&dev_priv->vm_list);
INIT_LIST_HEAD(&dev_priv->context_list);
INIT_LIST_HEAD(&dev_priv->mm.unbound_list);
INIT_LIST_HEAD(&dev_priv->mm.bound_list);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 23a4960deb64..97b017f9e272 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2746,10 +2746,7 @@ static void i915_gtt_color_adjust(struct drm_mm_node *node,
*end -= 4096;
}
-static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
- u64 start,
- u64 mappable_end,
- u64 end)
+int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
{
/* Let GEM Manage all of the aperture.
*
@@ -2761,46 +2758,14 @@ static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
* of the aperture.
*/
struct i915_ggtt *ggtt = &dev_priv->ggtt;
- struct drm_mm_node *entry;
- struct drm_i915_gem_object *obj;
unsigned long hole_start, hole_end;
+ struct drm_mm_node *entry;
int ret;
- BUG_ON(mappable_end > end);
-
- ggtt->base.start = start;
-
- /* Subtract the guard page before address space initialization to
- * shrink the range used by drm_mm */
- ggtt->base.total = end - start - PAGE_SIZE;
- i915_address_space_init(&ggtt->base, dev_priv);
- ggtt->base.total += PAGE_SIZE;
-
ret = intel_vgt_balloon(dev_priv);
if (ret)
return ret;
- if (!HAS_LLC(dev_priv))
- ggtt->base.mm.color_adjust = i915_gtt_color_adjust;
-
- /* Mark any preallocated objects as occupied */
- list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
- struct i915_vma *vma = i915_gem_obj_to_vma(obj, &ggtt->base);
-
- DRM_DEBUG_KMS("reserving preallocated space: %llx + %zx\n",
- i915_gem_obj_ggtt_offset(obj), obj->base.size);
-
- WARN_ON(i915_gem_obj_ggtt_bound(obj));
- ret = drm_mm_reserve_node(&ggtt->base.mm, &vma->node);
- if (ret) {
- DRM_DEBUG_KMS("Reservation failed: %i\n", ret);
- return ret;
- }
- vma->bound |= GLOBAL_BIND;
- __i915_vma_set_map_and_fenceable(vma);
- list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
- }
-
/* Clear any non-preallocated blocks */
drm_mm_for_each_hole(entry, &ggtt->base.mm, hole_start, hole_end) {
DRM_DEBUG_KMS("clearing unused GTT space: [%lx, %lx]\n",
@@ -2810,7 +2775,9 @@ static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
}
/* And finally clear the reserved guard page */
- ggtt->base.clear_range(&ggtt->base, end - PAGE_SIZE, PAGE_SIZE, true);
+ ggtt->base.clear_range(&ggtt->base,
+ ggtt->base.total - PAGE_SIZE, PAGE_SIZE,
+ true);
if (USES_PPGTT(dev_priv) && !USES_FULL_PPGTT(dev_priv)) {
struct i915_hw_ppgtt *ppgtt;
@@ -2849,18 +2816,6 @@ static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
}
/**
- * i915_gem_init_ggtt - Initialize GEM for Global GTT
- * @dev_priv: i915 device
- */
-void i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
-{
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
-
- i915_gem_setup_global_gtt(dev_priv,
- 0, ggtt->mappable_end, ggtt->base.total);
-}
-
-/**
* i915_ggtt_cleanup_hw - Clean up GGTT hardware initialization
* @dev_priv: i915 device
*/
@@ -2884,6 +2839,9 @@ void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv)
}
ggtt->base.cleanup(&ggtt->base);
+
+ arch_phys_wc_del(ggtt->mtrr);
+ io_mapping_free(ggtt->mappable);
}
static unsigned int gen6_get_total_gtt_size(u16 snb_gmch_ctl)
@@ -3274,6 +3232,26 @@ int i915_ggtt_init_hw(struct drm_i915_private *dev_priv)
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int ret;
+ INIT_LIST_HEAD(&dev_priv->vm_list);
+
+ /* Subtract the guard page before address space initialization to
+ * shrink the range used by drm_mm.
+ */
+ ggtt->base.total -= PAGE_SIZE;
+ i915_address_space_init(&ggtt->base, dev_priv);
+ ggtt->base.total += PAGE_SIZE;
+ if (!HAS_LLC(dev_priv))
+ ggtt->base.mm.color_adjust = i915_gtt_color_adjust;
+
+ ggtt->mappable =
+ io_mapping_create_wc(ggtt->mappable_base, ggtt->mappable_end);
+ if (!ggtt->mappable) {
+ ret = -EIO;
+ goto out_gtt_cleanup;
+ }
+
+ ggtt->mtrr = arch_phys_wc_add(ggtt->mappable_base, ggtt->mappable_end);
+
/*
* Initialise stolen early so that we may reserve preallocated
* objects for the BIOS to KMS transition.
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 7db8555aa254..48d50cc3f95a 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -522,7 +522,7 @@ i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
-void i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
+int i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
int i915_ppgtt_init_hw(struct drm_device *dev);
diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c
index 310756c30723..9a8cc8c51077 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/i915_gem_stolen.c
@@ -698,18 +698,17 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev,
*/
vma->node.start = gtt_offset;
vma->node.size = size;
- if (drm_mm_initialized(&ggtt->base.mm)) {
- ret = drm_mm_reserve_node(&ggtt->base.mm, &vma->node);
- if (ret) {
- DRM_DEBUG_KMS("failed to allocate stolen GTT space\n");
- goto err;
- }
- vma->bound |= GLOBAL_BIND;
- __i915_vma_set_map_and_fenceable(vma);
- list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
+ ret = drm_mm_reserve_node(&ggtt->base.mm, &vma->node);
+ if (ret) {
+ DRM_DEBUG_KMS("failed to allocate stolen GTT space\n");
+ goto err;
}
+ vma->bound |= GLOBAL_BIND;
+ __i915_vma_set_map_and_fenceable(vma);
+ list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
+
list_add_tail(&obj->global_list, &dev_priv->mm.bound_list);
i915_gem_object_pin_pages(obj);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 06/28] drm/i915: Store owning file on the i915_address_space
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (4 preceding siblings ...)
2016-08-03 13:42 ` [PATCH 05/28] drm/i915: Split early global GTT initialisation Chris Wilson
@ 2016-08-03 13:42 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 07/28] drm/i915: Count how many VMA are bound for an object Chris Wilson
` (23 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:42 UTC (permalink / raw)
To: intel-gfx
For the global GTT (and aliasing GTT), the address space is owned by the
device (it is a global resource) and so the per-file owner field is
NULL. For per-process GTT (where we create an address space per
context), each is owned by the opening file. We can use this ownership
information to both distinguish GGTT and ppGTT address spaces, as well
as occasionally inspect the owner.
v2: Whitespace, tells us who owns i915_address_space
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 2 +-
drivers/gpu/drm/i915/i915_drv.h | 1 -
drivers/gpu/drm/i915/i915_gem_context.c | 3 ++-
drivers/gpu/drm/i915/i915_gem_gtt.c | 30 +++++++++++++++---------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 17 +++++++++++------
5 files changed, 29 insertions(+), 24 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index e7724be9f4db..9bd5eb3af277 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -354,7 +354,7 @@ static int per_file_stats(int id, void *ptr, void *data)
} else {
struct i915_hw_ppgtt *ppgtt = i915_vm_to_ppgtt(vma->vm);
- if (ppgtt->file_priv != stats->file_priv)
+ if (ppgtt->base.file != stats->file_priv)
continue;
}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 9b18b9c875ec..7452a4015956 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3300,7 +3300,6 @@ i915_vm_to_ppgtt(struct i915_address_space *vm)
return container_of(vm, struct i915_hw_ppgtt, base);
}
-
static inline bool i915_gem_obj_ggtt_bound(struct drm_i915_gem_object *obj)
{
return i915_gem_obj_ggtt_bound_view(obj, &i915_ggtt_view_normal);
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index d7a7cc8b6fa4..24383f0ea3a4 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -327,7 +327,8 @@ i915_gem_create_context(struct drm_device *dev,
return ctx;
if (USES_FULL_PPGTT(dev)) {
- struct i915_hw_ppgtt *ppgtt = i915_ppgtt_create(dev, file_priv);
+ struct i915_hw_ppgtt *ppgtt =
+ i915_ppgtt_create(to_i915(dev), file_priv);
if (IS_ERR(ppgtt)) {
DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n",
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 97b017f9e272..f8e71a4a8a5d 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2104,11 +2104,12 @@ static int gen6_ppgtt_init(struct i915_hw_ppgtt *ppgtt)
return 0;
}
-static int __hw_ppgtt_init(struct drm_device *dev, struct i915_hw_ppgtt *ppgtt)
+static int __hw_ppgtt_init(struct i915_hw_ppgtt *ppgtt,
+ struct drm_i915_private *dev_priv)
{
- ppgtt->base.dev = dev;
+ ppgtt->base.dev = &dev_priv->drm;
- if (INTEL_INFO(dev)->gen < 8)
+ if (INTEL_INFO(dev_priv)->gen < 8)
return gen6_ppgtt_init(ppgtt);
else
return gen8_ppgtt_init(ppgtt);
@@ -2143,15 +2144,17 @@ static void gtt_write_workarounds(struct drm_device *dev)
I915_WRITE(GEN8_L3_LRA_1_GPGPU, GEN9_L3_LRA_1_GPGPU_DEFAULT_VALUE_BXT);
}
-static int i915_ppgtt_init(struct drm_device *dev, struct i915_hw_ppgtt *ppgtt)
+static int i915_ppgtt_init(struct i915_hw_ppgtt *ppgtt,
+ struct drm_i915_private *dev_priv,
+ struct drm_i915_file_private *file_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
- int ret = 0;
+ int ret;
- ret = __hw_ppgtt_init(dev, ppgtt);
+ ret = __hw_ppgtt_init(ppgtt, dev_priv);
if (ret == 0) {
kref_init(&ppgtt->ref);
i915_address_space_init(&ppgtt->base, dev_priv);
+ ppgtt->base.file = file_priv;
}
return ret;
@@ -2183,7 +2186,8 @@ int i915_ppgtt_init_hw(struct drm_device *dev)
}
struct i915_hw_ppgtt *
-i915_ppgtt_create(struct drm_device *dev, struct drm_i915_file_private *fpriv)
+i915_ppgtt_create(struct drm_i915_private *dev_priv,
+ struct drm_i915_file_private *fpriv)
{
struct i915_hw_ppgtt *ppgtt;
int ret;
@@ -2192,14 +2196,12 @@ i915_ppgtt_create(struct drm_device *dev, struct drm_i915_file_private *fpriv)
if (!ppgtt)
return ERR_PTR(-ENOMEM);
- ret = i915_ppgtt_init(dev, ppgtt);
+ ret = i915_ppgtt_init(ppgtt, dev_priv, fpriv);
if (ret) {
kfree(ppgtt);
return ERR_PTR(ret);
}
- ppgtt->file_priv = fpriv;
-
trace_i915_ppgtt_create(&ppgtt->base);
return ppgtt;
@@ -2786,9 +2788,8 @@ int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
if (!ppgtt)
return -ENOMEM;
- ret = __hw_ppgtt_init(&dev_priv->drm, ppgtt);
+ ret = __hw_ppgtt_init(ppgtt, dev_priv);
if (ret) {
- ppgtt->base.cleanup(&ppgtt->base);
kfree(ppgtt);
return ret;
}
@@ -3174,7 +3175,6 @@ int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv)
int ret;
ggtt->base.dev = &dev_priv->drm;
- ggtt->base.is_ggtt = true;
if (INTEL_GEN(dev_priv) <= 5) {
probe = i915_gmch_probe;
@@ -3319,7 +3319,7 @@ void i915_gem_restore_gtt_mappings(struct drm_device *dev)
struct i915_hw_ppgtt *ppgtt;
- if (vm->is_ggtt)
+ if (i915_is_ggtt(vm))
ppgtt = dev_priv->mm.aliasing_ppgtt;
else
ppgtt = i915_vm_to_ppgtt(vm);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 48d50cc3f95a..4498bdd46b6b 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -272,12 +272,19 @@ struct i915_pml4 {
struct i915_address_space {
struct drm_mm mm;
struct drm_device *dev;
+ /* Every address space belongs to a struct file - except for the global
+ * GTT that is owned by the driver (and so @file is set to NULL). In
+ * principle, no information should leak from one context to another
+ * (or between files/processes etc) unless explicitly shared by the
+ * owner. Tracking the owner is important in order to free up per-file
+ * objects along with the file, to aide resource tracking, and to
+ * assign blame.
+ */
+ struct drm_i915_file_private *file;
struct list_head global_link;
u64 start; /* Start offset always 0 for dri2 */
u64 total; /* size addr space maps (ex. 2GB for ggtt) */
- bool is_ggtt;
-
struct i915_page_scratch *scratch_page;
struct i915_page_table *scratch_pt;
struct i915_page_directory *scratch_pd;
@@ -338,7 +345,7 @@ struct i915_address_space {
u32 flags);
};
-#define i915_is_ggtt(V) ((V)->is_ggtt)
+#define i915_is_ggtt(V) (!(V)->file)
/* The Graphics Translation Table is the way in which GEN hardware translates a
* Graphics Virtual Address into a Physical Address. In addition to the normal
@@ -378,8 +385,6 @@ struct i915_hw_ppgtt {
struct i915_page_directory pd; /* GEN6-7 */
};
- struct drm_i915_file_private *file_priv;
-
gen6_pte_t __iomem *pd_addr;
int (*enable)(struct i915_hw_ppgtt *ppgtt);
@@ -527,7 +532,7 @@ void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
int i915_ppgtt_init_hw(struct drm_device *dev);
void i915_ppgtt_release(struct kref *kref);
-struct i915_hw_ppgtt *i915_ppgtt_create(struct drm_device *dev,
+struct i915_hw_ppgtt *i915_ppgtt_create(struct drm_i915_private *dev_priv,
struct drm_i915_file_private *fpriv);
static inline void i915_ppgtt_get(struct i915_hw_ppgtt *ppgtt)
{
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 07/28] drm/i915: Count how many VMA are bound for an object
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (5 preceding siblings ...)
2016-08-03 13:42 ` [PATCH 06/28] drm/i915: Store owning file on the i915_address_space Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 08/28] drm/i915: Be more careful when unbinding vma Chris Wilson
` (22 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
Since we may have VMA allocated for an object, but we interrupted their
binding, there is a disparity between have elements on the obj->vma_list
and being bound. i915_gem_obj_bound_any() does this check, but this is
not rigorously observed - add an explicit count to make it easier.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 12 +++++-------
drivers/gpu/drm/i915/i915_drv.h | 3 ++-
drivers/gpu/drm/i915/i915_gem.c | 30 ++++++++++--------------------
drivers/gpu/drm/i915/i915_gem_shrinker.c | 17 ++++++++---------
drivers/gpu/drm/i915/i915_gem_stolen.c | 1 +
5 files changed, 26 insertions(+), 37 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 9bd5eb3af277..44cf4d0d1cb9 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -174,6 +174,9 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
if (obj->fence_reg != I915_FENCE_REG_NONE)
seq_printf(m, " (fence: %d)", obj->fence_reg);
list_for_each_entry(vma, &obj->vma_list, obj_link) {
+ if (!drm_mm_node_allocated(&vma->node))
+ continue;
+
seq_printf(m, " (%sgtt offset: %08llx, size: %08llx",
vma->is_ggtt ? "g" : "pp",
vma->node.start, vma->node.size);
@@ -335,11 +338,11 @@ static int per_file_stats(int id, void *ptr, void *data)
struct drm_i915_gem_object *obj = ptr;
struct file_stats *stats = data;
struct i915_vma *vma;
- int bound = 0;
stats->count++;
stats->total += obj->base.size;
-
+ if (!obj->bind_count)
+ stats->unbound += obj->base.size;
if (obj->base.name || obj->base.dma_buf)
stats->shared += obj->base.size;
@@ -347,8 +350,6 @@ static int per_file_stats(int id, void *ptr, void *data)
if (!drm_mm_node_allocated(&vma->node))
continue;
- bound++;
-
if (vma->is_ggtt) {
stats->global += vma->node.size;
} else {
@@ -364,9 +365,6 @@ static int per_file_stats(int id, void *ptr, void *data)
stats->inactive += vma->node.size;
}
- if (!bound)
- stats->unbound += obj->base.size;
-
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 7452a4015956..7a9a2d820aca 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2221,6 +2221,8 @@ struct drm_i915_gem_object {
unsigned int frontbuffer_bits:INTEL_FRONTBUFFER_BITS;
unsigned int has_wc_mmap;
+ /** Count of VMA actually bound by this object */
+ unsigned int bind_count;
unsigned int pin_display;
struct sg_table *pages;
@@ -3266,7 +3268,6 @@ i915_gem_obj_ggtt_offset(struct drm_i915_gem_object *o)
return i915_gem_obj_ggtt_offset_view(o, &i915_ggtt_view_normal);
}
-bool i915_gem_obj_bound_any(struct drm_i915_gem_object *o);
bool i915_gem_obj_ggtt_bound_view(struct drm_i915_gem_object *o,
const struct i915_ggtt_view *view);
bool i915_gem_obj_bound(struct drm_i915_gem_object *o,
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 83e299c429c2..a4046fdb12de 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2107,7 +2107,7 @@ i915_gem_object_put_pages(struct drm_i915_gem_object *obj)
if (obj->pages_pin_count)
return -EBUSY;
- BUG_ON(i915_gem_obj_bound_any(obj));
+ GEM_BUG_ON(obj->bind_count);
/* ->put_pages might need to allocate memory for the bit17 swizzle
* array, hence protect them from being reaped by removing them from gtt
@@ -2965,7 +2965,6 @@ static void __i915_vma_iounmap(struct i915_vma *vma)
static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
{
struct drm_i915_gem_object *obj = vma->obj;
- struct drm_i915_private *dev_priv = to_i915(obj->base.dev);
int ret;
if (list_empty(&vma->obj_link))
@@ -2979,7 +2978,8 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
if (vma->pin_count)
return -EBUSY;
- BUG_ON(obj->pages == NULL);
+ GEM_BUG_ON(obj->bind_count == 0);
+ GEM_BUG_ON(!obj->pages);
if (wait) {
ret = i915_gem_object_wait_rendering(obj, false);
@@ -3019,8 +3019,9 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
/* Since the unbound list is global, only move to that list if
* no more VMAs exist. */
- if (list_empty(&obj->vma_list))
- list_move_tail(&obj->global_list, &dev_priv->mm.unbound_list);
+ if (--obj->bind_count == 0)
+ list_move_tail(&obj->global_list,
+ &to_i915(obj->base.dev)->mm.unbound_list);
/* And finally now the object is completely decoupled from this vma,
* we can drop its hold on the backing storage and allow it to be
@@ -3255,6 +3256,7 @@ search_free:
list_move_tail(&obj->global_list, &dev_priv->mm.bound_list);
list_add_tail(&vma->vm_link, &vm->inactive_list);
+ obj->bind_count++;
return vma;
@@ -3450,7 +3452,6 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
{
struct drm_device *dev = obj->base.dev;
struct i915_vma *vma, *next;
- bool bound = false;
int ret = 0;
if (obj->cache_level == cache_level)
@@ -3474,8 +3475,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
ret = i915_vma_unbind(vma);
if (ret)
return ret;
- } else
- bound = true;
+ }
}
/* We can reuse the existing drm_mm nodes but need to change the
@@ -3485,7 +3485,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
* rewrite the PTE in the belief that doing so tramples upon less
* state and so involves less work.
*/
- if (bound) {
+ if (obj->bind_count) {
/* Before we change the PTE, the GPU must not be accessing it.
* If we wait upon the object, we know that all the bound
* VMA are no longer active.
@@ -4223,6 +4223,7 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj)
dev_priv->mm.interruptible = was_interruptible;
}
}
+ GEM_BUG_ON(obj->bind_count);
/* Stolen objects don't hold a ref, but do hold pin count. Fix that up
* before progressing. */
@@ -4840,17 +4841,6 @@ bool i915_gem_obj_ggtt_bound_view(struct drm_i915_gem_object *o,
return false;
}
-bool i915_gem_obj_bound_any(struct drm_i915_gem_object *o)
-{
- struct i915_vma *vma;
-
- list_for_each_entry(vma, &o->vma_list, obj_link)
- if (drm_mm_node_allocated(&vma->node))
- return true;
-
- return false;
-}
-
unsigned long i915_gem_obj_ggtt_size(struct drm_i915_gem_object *o)
{
struct i915_vma *vma;
diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index 5d4772c146b1..b95cd9f404f6 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -48,19 +48,15 @@ static bool mutex_is_locked_by(struct mutex *mutex, struct task_struct *task)
#endif
}
-static int num_vma_bound(struct drm_i915_gem_object *obj)
+static bool any_vma_pinned(struct drm_i915_gem_object *obj)
{
struct i915_vma *vma;
- int count = 0;
- list_for_each_entry(vma, &obj->vma_list, obj_link) {
- if (drm_mm_node_allocated(&vma->node))
- count++;
+ list_for_each_entry(vma, &obj->vma_list, obj_link)
if (vma->pin_count)
- count++;
- }
+ return true;
- return count;
+ return false;
}
static bool swap_available(void)
@@ -82,7 +78,10 @@ static bool can_release_pages(struct drm_i915_gem_object *obj)
* to the GPU, simply unbinding from the GPU is not going to succeed
* in releasing our pin count on the pages themselves.
*/
- if (obj->pages_pin_count != num_vma_bound(obj))
+ if (obj->pages_pin_count > obj->bind_count)
+ return false;
+
+ if (any_vma_pinned(obj))
return false;
/* We can only return physical pages to the system if we can either
diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c
index 9a8cc8c51077..2c321c8df7c3 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/i915_gem_stolen.c
@@ -708,6 +708,7 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev,
vma->bound |= GLOBAL_BIND;
__i915_vma_set_map_and_fenceable(vma);
list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
+ obj->bind_count++;
list_add_tail(&obj->global_list, &dev_priv->mm.bound_list);
i915_gem_object_pin_pages(obj);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 08/28] drm/i915: Be more careful when unbinding vma
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (6 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 07/28] drm/i915: Count how many VMA are bound for an object Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 09/28] drm/i915: Kill drop_pages() Chris Wilson
` (21 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
When we call i915_vma_unbind(), we will wait upon outstanding rendering.
This will also trigger a retirement phase, which may update the object
lists. If, we extend request tracking to the VMA itself (rather than
keep it at the encompassing object), then there is a potential that the
obj->vma_list be modified for other elements upon i915_vma_unbind(). As
a result, if we walk over the object list and call i915_vma_unbind(), we
need to be prepared for that list to change.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.h | 2 ++
drivers/gpu/drm/i915/i915_gem.c | 57 +++++++++++++++++++++++---------
drivers/gpu/drm/i915/i915_gem_shrinker.c | 8 +----
drivers/gpu/drm/i915/i915_gem_userptr.c | 4 +--
4 files changed, 46 insertions(+), 25 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 7a9a2d820aca..a6bab90bf832 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3052,6 +3052,8 @@ int __must_check i915_vma_unbind(struct i915_vma *vma);
* _guarantee_ VMA in question is _not in use_ anywhere.
*/
int __must_check __i915_vma_unbind_no_wait(struct i915_vma *vma);
+
+int i915_gem_object_unbind(struct drm_i915_gem_object *obj);
int i915_gem_object_put_pages(struct drm_i915_gem_object *obj);
void i915_gem_release_all_mmaps(struct drm_i915_private *dev_priv);
void i915_gem_release_mmap(struct drm_i915_gem_object *obj);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index a4046fdb12de..57f43e40a8a5 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -283,18 +283,38 @@ static const struct drm_i915_gem_object_ops i915_gem_phys_ops = {
.release = i915_gem_object_release_phys,
};
+int
+i915_gem_object_unbind(struct drm_i915_gem_object *obj)
+{
+ struct i915_vma *vma;
+ LIST_HEAD(still_in_list);
+ int ret;
+
+ /* The vma will only be freed if it is marked as closed, and if we wait
+ * upon rendering to the vma, we may unbind anything in the list.
+ */
+ while ((vma = list_first_entry_or_null(&obj->vma_list,
+ struct i915_vma,
+ obj_link))) {
+ list_move_tail(&vma->obj_link, &still_in_list);
+ ret = i915_vma_unbind(vma);
+ if (ret)
+ break;
+ }
+ list_splice(&still_in_list, &obj->vma_list);
+
+ return ret;
+}
+
static int
drop_pages(struct drm_i915_gem_object *obj)
{
- struct i915_vma *vma, *next;
int ret;
i915_gem_object_get(obj);
- list_for_each_entry_safe(vma, next, &obj->vma_list, obj_link)
- if (i915_vma_unbind(vma))
- break;
-
- ret = i915_gem_object_put_pages(obj);
+ ret = i915_gem_object_unbind(obj);
+ if (ret == 0)
+ ret = i915_gem_object_put_pages(obj);
i915_gem_object_put(obj);
return ret;
@@ -3450,8 +3470,7 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
enum i915_cache_level cache_level)
{
- struct drm_device *dev = obj->base.dev;
- struct i915_vma *vma, *next;
+ struct i915_vma *vma;
int ret = 0;
if (obj->cache_level == cache_level)
@@ -3462,7 +3481,8 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
* catch the issue of the CS prefetch crossing page boundaries and
* reading an invalid PTE on older architectures.
*/
- list_for_each_entry_safe(vma, next, &obj->vma_list, obj_link) {
+restart:
+ list_for_each_entry(vma, &obj->vma_list, obj_link) {
if (!drm_mm_node_allocated(&vma->node))
continue;
@@ -3471,11 +3491,18 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
return -EBUSY;
}
- if (!i915_gem_valid_gtt_space(vma, cache_level)) {
- ret = i915_vma_unbind(vma);
- if (ret)
- return ret;
- }
+ if (i915_gem_valid_gtt_space(vma, cache_level))
+ continue;
+
+ ret = i915_vma_unbind(vma);
+ if (ret)
+ return ret;
+
+ /* As unbinding may affect other elements in the
+ * obj->vma_list (due to side-effects from retiring
+ * an active vma), play safe and restart the iterator.
+ */
+ goto restart;
}
/* We can reuse the existing drm_mm nodes but need to change the
@@ -3494,7 +3521,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj,
if (ret)
return ret;
- if (!HAS_LLC(dev) && cache_level != I915_CACHE_NONE) {
+ if (!HAS_LLC(obj->base.dev) && cache_level != I915_CACHE_NONE) {
/* Access to snoopable pages through the GTT is
* incoherent and on some machines causes a hard
* lockup. Relinquish the CPU mmaping to force
diff --git a/drivers/gpu/drm/i915/i915_gem_shrinker.c b/drivers/gpu/drm/i915/i915_gem_shrinker.c
index b95cd9f404f6..64d179d4f684 100644
--- a/drivers/gpu/drm/i915/i915_gem_shrinker.c
+++ b/drivers/gpu/drm/i915/i915_gem_shrinker.c
@@ -172,8 +172,6 @@ i915_gem_shrink(struct drm_i915_private *dev_priv,
(obj = list_first_entry_or_null(phase->list,
typeof(*obj),
global_list))) {
- struct i915_vma *vma, *v;
-
list_move_tail(&obj->global_list, &still_in_list);
if (flags & I915_SHRINK_PURGEABLE &&
@@ -193,11 +191,7 @@ i915_gem_shrink(struct drm_i915_private *dev_priv,
i915_gem_object_get(obj);
/* For the unbound phase, this should be a no-op! */
- list_for_each_entry_safe(vma, v,
- &obj->vma_list, obj_link)
- if (i915_vma_unbind(vma))
- break;
-
+ i915_gem_object_unbind(obj);
if (i915_gem_object_put_pages(obj) == 0)
count += obj->base.size >> PAGE_SHIFT;
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index ca8b82ab93d6..e935b327f3f9 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -104,7 +104,6 @@ static void cancel_userptr(struct work_struct *work)
if (obj->pages != NULL) {
struct drm_i915_private *dev_priv = to_i915(dev);
- struct i915_vma *vma, *tmp;
bool was_interruptible;
wait_rendering(obj);
@@ -112,8 +111,7 @@ static void cancel_userptr(struct work_struct *work)
was_interruptible = dev_priv->mm.interruptible;
dev_priv->mm.interruptible = false;
- list_for_each_entry_safe(vma, tmp, &obj->vma_list, obj_link)
- WARN_ON(i915_vma_unbind(vma));
+ WARN_ON(i915_gem_object_unbind(obj));
WARN_ON(i915_gem_object_put_pages(obj));
dev_priv->mm.interruptible = was_interruptible;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 09/28] drm/i915: Kill drop_pages()
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (7 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 08/28] drm/i915: Be more careful when unbinding vma Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 10/28] drm/i915: Introduce i915_gem_active for request tracking Chris Wilson
` (20 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
The drop_pages() function is a dangerous trap in that it can release the
passed in object pointer and so unless the caller is aware, it can
easily trick us into using the stale object afterwards. Move it into its
solitary callsite where we know it is safe.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 20 +++++---------------
1 file changed, 5 insertions(+), 15 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 57f43e40a8a5..90e951202eb8 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -306,20 +306,6 @@ i915_gem_object_unbind(struct drm_i915_gem_object *obj)
return ret;
}
-static int
-drop_pages(struct drm_i915_gem_object *obj)
-{
- int ret;
-
- i915_gem_object_get(obj);
- ret = i915_gem_object_unbind(obj);
- if (ret == 0)
- ret = i915_gem_object_put_pages(obj);
- i915_gem_object_put(obj);
-
- return ret;
-}
-
int
i915_gem_object_attach_phys(struct drm_i915_gem_object *obj,
int align)
@@ -340,7 +326,11 @@ i915_gem_object_attach_phys(struct drm_i915_gem_object *obj,
if (obj->base.filp == NULL)
return -EINVAL;
- ret = drop_pages(obj);
+ ret = i915_gem_object_unbind(obj);
+ if (ret)
+ return ret;
+
+ ret = i915_gem_object_put_pages(obj);
if (ret)
return ret;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 10/28] drm/i915: Introduce i915_gem_active for request tracking
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (8 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 09/28] drm/i915: Kill drop_pages() Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 11/28] drm/i915: Prepare i915_gem_active for annotations Chris Wilson
` (19 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
In the next patch, request tracking is made more generic and for that we
need a new expanded struct and to separate out the logic changes from
the mechanical churn, we split out the structure renaming into this
patch.
v2: Writer's block. Add some spiel about why we track requests.
v3: Now i915_gem_active.
v4: Now with i915_gem_active_set() for attaching to the active request.
v5: Use i915_gem_active_set() from inside the retirement handlers
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 11 +++---
drivers/gpu/drm/i915/i915_drv.h | 9 +++--
drivers/gpu/drm/i915/i915_gem.c | 58 +++++++++++++++---------------
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 4 +--
drivers/gpu/drm/i915/i915_gem_fence.c | 6 ++--
drivers/gpu/drm/i915/i915_gem_request.h | 41 +++++++++++++++++++++
drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +-
drivers/gpu/drm/i915/i915_gem_userptr.c | 2 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 7 ++--
drivers/gpu/drm/i915/intel_display.c | 8 ++---
10 files changed, 93 insertions(+), 55 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 44cf4d0d1cb9..6151460f848f 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -155,10 +155,10 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
obj->base.write_domain);
for_each_engine_id(engine, dev_priv, id)
seq_printf(m, "%x ",
- i915_gem_request_get_seqno(obj->last_read_req[id]));
+ i915_gem_request_get_seqno(obj->last_read[id].request));
seq_printf(m, "] %x %x%s%s%s",
- i915_gem_request_get_seqno(obj->last_write_req),
- i915_gem_request_get_seqno(obj->last_fenced_req),
+ i915_gem_request_get_seqno(obj->last_write.request),
+ i915_gem_request_get_seqno(obj->last_fence.request),
i915_cache_level_str(to_i915(obj->base.dev), obj->cache_level),
obj->dirty ? " dirty" : "",
obj->madv == I915_MADV_DONTNEED ? " purgeable" : "");
@@ -195,9 +195,8 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
*t = '\0';
seq_printf(m, " (%s mappable)", s);
}
- if (obj->last_write_req != NULL)
- seq_printf(m, " (%s)",
- i915_gem_request_get_engine(obj->last_write_req)->name);
+ if (obj->last_write.request)
+ seq_printf(m, " (%s)", obj->last_write.request->engine->name);
if (obj->frontbuffer_bits)
seq_printf(m, " (frontbuffer: 0x%03x)", obj->frontbuffer_bits);
}
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index a6bab90bf832..17a206f19fcb 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -2242,11 +2242,10 @@ struct drm_i915_gem_object {
* requests on one ring where the write request is older than the
* read request. This allows for the CPU to read from an active
* buffer by only waiting for the write to complete.
- * */
- struct drm_i915_gem_request *last_read_req[I915_NUM_ENGINES];
- struct drm_i915_gem_request *last_write_req;
- /** Breadcrumb of last fenced GPU access to the buffer. */
- struct drm_i915_gem_request *last_fenced_req;
+ */
+ struct i915_gem_active last_read[I915_NUM_ENGINES];
+ struct i915_gem_active last_write;
+ struct i915_gem_active last_fence;
/** Current tiling stride for the object, if it's tiled. */
uint32_t stride;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 90e951202eb8..ed868b2f4f71 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1353,23 +1353,23 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
int ret, i;
if (readonly) {
- if (obj->last_write_req != NULL) {
- ret = i915_wait_request(obj->last_write_req);
+ if (obj->last_write.request) {
+ ret = i915_wait_request(obj->last_write.request);
if (ret)
return ret;
- i = obj->last_write_req->engine->id;
- if (obj->last_read_req[i] == obj->last_write_req)
+ i = obj->last_write.request->engine->id;
+ if (obj->last_read[i].request == obj->last_write.request)
i915_gem_object_retire__read(obj, i);
else
i915_gem_object_retire__write(obj);
}
} else {
for (i = 0; i < I915_NUM_ENGINES; i++) {
- if (obj->last_read_req[i] == NULL)
+ if (!obj->last_read[i].request)
continue;
- ret = i915_wait_request(obj->last_read_req[i]);
+ ret = i915_wait_request(obj->last_read[i].request);
if (ret)
return ret;
@@ -1397,9 +1397,9 @@ i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
{
int idx = req->engine->id;
- if (obj->last_read_req[idx] == req)
+ if (obj->last_read[idx].request == req)
i915_gem_object_retire__read(obj, idx);
- else if (obj->last_write_req == req)
+ else if (obj->last_write.request == req)
i915_gem_object_retire__write(obj);
if (!i915_reset_in_progress(&req->i915->gpu_error))
@@ -1428,7 +1428,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
if (readonly) {
struct drm_i915_gem_request *req;
- req = obj->last_write_req;
+ req = obj->last_write.request;
if (req == NULL)
return 0;
@@ -1437,7 +1437,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read_req[i];
+ req = obj->last_read[i].request;
if (req == NULL)
continue;
@@ -2375,7 +2375,7 @@ void i915_vma_move_to_active(struct i915_vma *vma,
obj->active |= intel_engine_flag(engine);
list_move_tail(&obj->engine_list[engine->id], &engine->active_list);
- i915_gem_request_assign(&obj->last_read_req[engine->id], req);
+ i915_gem_active_set(&obj->last_read[engine->id], req);
list_move_tail(&vma->vm_link, &vma->vm->active_list);
}
@@ -2383,10 +2383,10 @@ void i915_vma_move_to_active(struct i915_vma *vma,
static void
i915_gem_object_retire__write(struct drm_i915_gem_object *obj)
{
- GEM_BUG_ON(obj->last_write_req == NULL);
- GEM_BUG_ON(!(obj->active & intel_engine_flag(obj->last_write_req->engine)));
+ GEM_BUG_ON(!obj->last_write.request);
+ GEM_BUG_ON(!(obj->active & intel_engine_flag(obj->last_write.request->engine)));
- i915_gem_request_assign(&obj->last_write_req, NULL);
+ i915_gem_active_set(&obj->last_write, NULL);
intel_fb_obj_flush(obj, true, ORIGIN_CS);
}
@@ -2395,13 +2395,13 @@ i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int idx)
{
struct i915_vma *vma;
- GEM_BUG_ON(obj->last_read_req[idx] == NULL);
+ GEM_BUG_ON(!obj->last_read[idx].request);
GEM_BUG_ON(!(obj->active & (1 << idx)));
list_del_init(&obj->engine_list[idx]);
- i915_gem_request_assign(&obj->last_read_req[idx], NULL);
+ i915_gem_active_set(&obj->last_read[idx], NULL);
- if (obj->last_write_req && obj->last_write_req->engine->id == idx)
+ if (obj->last_write.request && obj->last_write.request->engine->id == idx)
i915_gem_object_retire__write(obj);
obj->active &= ~(1 << idx);
@@ -2420,7 +2420,7 @@ i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int idx)
list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
}
- i915_gem_request_assign(&obj->last_fenced_req, NULL);
+ i915_gem_active_set(&obj->last_fence, NULL);
i915_gem_object_put(obj);
}
@@ -2621,7 +2621,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
struct drm_i915_gem_object,
engine_list[engine->id]);
- if (!list_empty(&obj->last_read_req[engine->id]->list))
+ if (!list_empty(&obj->last_read[engine->id].request->list))
break;
i915_gem_object_retire__read(obj, engine->id);
@@ -2754,7 +2754,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read_req[i];
+ req = obj->last_read[i].request;
if (req == NULL)
continue;
@@ -2830,10 +2830,10 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
i915_gem_object_put(obj);
for (i = 0; i < I915_NUM_ENGINES; i++) {
- if (obj->last_read_req[i] == NULL)
+ if (!obj->last_read[i].request)
continue;
- req[n++] = i915_gem_request_get(obj->last_read_req[i]);
+ req[n++] = i915_gem_request_get(obj->last_read[i].request);
}
mutex_unlock(&dev->struct_mutex);
@@ -2924,12 +2924,12 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,
n = 0;
if (readonly) {
- if (obj->last_write_req)
- req[n++] = obj->last_write_req;
+ if (obj->last_write.request)
+ req[n++] = obj->last_write.request;
} else {
for (i = 0; i < I915_NUM_ENGINES; i++)
- if (obj->last_read_req[i])
- req[n++] = obj->last_read_req[i];
+ if (obj->last_read[i].request)
+ req[n++] = obj->last_read[i].request;
}
for (i = 0; i < n; i++) {
ret = __i915_gem_object_sync(obj, to, req[i]);
@@ -4026,12 +4026,12 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read_req[i];
+ req = obj->last_read[i].request;
if (req)
args->busy |= 1 << (16 + req->engine->exec_id);
}
- if (obj->last_write_req)
- args->busy |= obj->last_write_req->engine->exec_id;
+ if (obj->last_write.request)
+ args->busy |= obj->last_write.request->engine->exec_id;
}
unref:
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index a4b98afc00ca..5e1fb85b708b 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1150,7 +1150,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas,
i915_vma_move_to_active(vma, req);
if (obj->base.write_domain) {
- i915_gem_request_assign(&obj->last_write_req, req);
+ i915_gem_active_set(&obj->last_write, req);
intel_fb_obj_invalidate(obj, ORIGIN_CS);
@@ -1158,7 +1158,7 @@ i915_gem_execbuffer_move_to_active(struct list_head *vmas,
obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS;
}
if (entry->flags & EXEC_OBJECT_NEEDS_FENCE) {
- i915_gem_request_assign(&obj->last_fenced_req, req);
+ i915_gem_active_set(&obj->last_fence, req);
if (entry->flags & __EXEC_OBJECT_HAS_FENCE) {
struct drm_i915_private *dev_priv = engine->i915;
list_move_tail(&dev_priv->fence_regs[obj->fence_reg].lru_list,
diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
index 251d7a95af89..d16b385e79e9 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence.c
@@ -261,12 +261,12 @@ static inline void i915_gem_object_fence_lost(struct drm_i915_gem_object *obj)
static int
i915_gem_object_wait_fence(struct drm_i915_gem_object *obj)
{
- if (obj->last_fenced_req) {
- int ret = i915_wait_request(obj->last_fenced_req);
+ if (obj->last_fence.request) {
+ int ret = i915_wait_request(obj->last_fence.request);
if (ret)
return ret;
- i915_gem_request_assign(&obj->last_fenced_req, NULL);
+ i915_gem_active_set(&obj->last_fence, NULL);
}
return 0;
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 382ca5a163eb..cf2df33f9892 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -249,4 +249,45 @@ static inline bool i915_spin_request(const struct drm_i915_gem_request *request,
__i915_spin_request(request, state, timeout_us));
}
+/* We treat requests as fences. This is not be to confused with our
+ * "fence registers" but pipeline synchronisation objects ala GL_ARB_sync.
+ * We use the fences to synchronize access from the CPU with activity on the
+ * GPU, for example, we should not rewrite an object's PTE whilst the GPU
+ * is reading them. We also track fences at a higher level to provide
+ * implicit synchronisation around GEM objects, e.g. set-domain will wait
+ * for outstanding GPU rendering before marking the object ready for CPU
+ * access, or a pageflip will wait until the GPU is complete before showing
+ * the frame on the scanout.
+ *
+ * In order to use a fence, the object must track the fence it needs to
+ * serialise with. For example, GEM objects want to track both read and
+ * write access so that we can perform concurrent read operations between
+ * the CPU and GPU engines, as well as waiting for all rendering to
+ * complete, or waiting for the last GPU user of a "fence register". The
+ * object then embeds a #i915_gem_active to track the most recent (in
+ * retirement order) request relevant for the desired mode of access.
+ * The #i915_gem_active is updated with i915_gem_active_set() to track the
+ * most recent fence request, typically this is done as part of
+ * i915_vma_move_to_active().
+ *
+ * When the #i915_gem_active completes (is retired), it will
+ * signal its completion to the owner through a callback as well as mark
+ * itself as idle (i915_gem_active.request == NULL). The owner
+ * can then perform any action, such as delayed freeing of an active
+ * resource including itself.
+ */
+struct i915_gem_active {
+ struct drm_i915_gem_request *request;
+};
+
+static inline void
+i915_gem_active_set(struct i915_gem_active *active,
+ struct drm_i915_gem_request *request)
+{
+ i915_gem_request_assign(&active->request, request);
+}
+
+#define for_each_active(mask, idx) \
+ for (; mask ? idx = ffs(mask) - 1, 1 : 0; mask &= ~BIT(idx))
+
#endif /* I915_GEM_REQUEST_H */
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index e83fc2d0433c..00d796da65fb 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -242,7 +242,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,
}
obj->fence_dirty =
- obj->last_fenced_req ||
+ obj->last_fence.request ||
obj->fence_reg != I915_FENCE_REG_NONE;
obj->tiling_mode = args->tiling_mode;
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index e935b327f3f9..32f50a70ea42 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -74,7 +74,7 @@ static void wait_rendering(struct drm_i915_gem_object *obj)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read_req[i];
+ req = obj->last_read[i].request;
if (req == NULL)
continue;
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 226b28efe9ab..d6482e980d5e 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -751,8 +751,8 @@ static void capture_bo(struct drm_i915_error_buffer *err,
err->size = obj->base.size;
err->name = obj->base.name;
for (i = 0; i < I915_NUM_ENGINES; i++)
- err->rseqno[i] = i915_gem_request_get_seqno(obj->last_read_req[i]);
- err->wseqno = i915_gem_request_get_seqno(obj->last_write_req);
+ err->rseqno[i] = i915_gem_request_get_seqno(obj->last_read[i].request);
+ err->wseqno = i915_gem_request_get_seqno(obj->last_write.request);
err->gtt_offset = vma->node.start;
err->read_domains = obj->base.read_domains;
err->write_domain = obj->base.write_domain;
@@ -764,8 +764,7 @@ static void capture_bo(struct drm_i915_error_buffer *err,
err->dirty = obj->dirty;
err->purgeable = obj->madv != I915_MADV_WILLNEED;
err->userptr = obj->userptr.mm != NULL;
- err->engine = obj->last_write_req ?
- i915_gem_request_get_engine(obj->last_write_req)->id : -1;
+ err->engine = obj->last_write.request ? obj->last_write.request->engine->id : -1;
err->cache_level = obj->cache_level;
}
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 993bcfbd7cdc..e858fede37ce 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -11370,7 +11370,7 @@ static bool use_mmio_flip(struct intel_engine_cs *engine,
if (resv && !reservation_object_test_signaled_rcu(resv, false))
return true;
- return engine != i915_gem_request_get_engine(obj->last_write_req);
+ return engine != i915_gem_request_get_engine(obj->last_write.request);
}
static void skl_do_mmio_flip(struct intel_crtc *intel_crtc,
@@ -11673,7 +11673,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
} else if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) {
engine = &dev_priv->engine[BCS];
} else if (INTEL_INFO(dev)->gen >= 7) {
- engine = i915_gem_request_get_engine(obj->last_write_req);
+ engine = i915_gem_request_get_engine(obj->last_write.request);
if (engine == NULL || engine->id != RCS)
engine = &dev_priv->engine[BCS];
} else {
@@ -11695,7 +11695,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
INIT_WORK(&work->mmio_work, intel_mmio_flip_work_func);
i915_gem_request_assign(&work->flip_queued_req,
- obj->last_write_req);
+ obj->last_write.request);
schedule_work(&work->mmio_work);
} else {
@@ -14043,7 +14043,7 @@ intel_prepare_plane_fb(struct drm_plane *plane,
to_intel_plane_state(new_state);
i915_gem_request_assign(&plane_state->wait_req,
- obj->last_write_req);
+ obj->last_write.request);
}
return ret;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 11/28] drm/i915: Prepare i915_gem_active for annotations
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (9 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 10/28] drm/i915: Introduce i915_gem_active for request tracking Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 12/28] drm/i915: Mark up i915_gem_active for locking annotation Chris Wilson
` (18 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
In the future, we will want to add annotations to the i915_gem_active
struct. The API is thus expanded to hide direct access to the contents
of i915_gem_active and mediated instead through a number of helpers.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 13 ++--
drivers/gpu/drm/i915/i915_gem.c | 85 ++++++++++++---------
drivers/gpu/drm/i915/i915_gem_fence.c | 11 ++-
drivers/gpu/drm/i915/i915_gem_request.h | 127 ++++++++++++++++++++++++++++++++
drivers/gpu/drm/i915/i915_gem_tiling.c | 2 +-
drivers/gpu/drm/i915/i915_gem_userptr.c | 8 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 9 ++-
drivers/gpu/drm/i915/intel_display.c | 15 ++--
8 files changed, 204 insertions(+), 66 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 6151460f848f..24ff7f4bc412 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -155,10 +155,10 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
obj->base.write_domain);
for_each_engine_id(engine, dev_priv, id)
seq_printf(m, "%x ",
- i915_gem_request_get_seqno(obj->last_read[id].request));
+ i915_gem_active_get_seqno(&obj->last_read[id]));
seq_printf(m, "] %x %x%s%s%s",
- i915_gem_request_get_seqno(obj->last_write.request),
- i915_gem_request_get_seqno(obj->last_fence.request),
+ i915_gem_active_get_seqno(&obj->last_write),
+ i915_gem_active_get_seqno(&obj->last_fence),
i915_cache_level_str(to_i915(obj->base.dev), obj->cache_level),
obj->dirty ? " dirty" : "",
obj->madv == I915_MADV_DONTNEED ? " purgeable" : "");
@@ -195,8 +195,11 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
*t = '\0';
seq_printf(m, " (%s mappable)", s);
}
- if (obj->last_write.request)
- seq_printf(m, " (%s)", obj->last_write.request->engine->name);
+
+ engine = i915_gem_active_get_engine(&obj->last_write);
+ if (engine)
+ seq_printf(m, " (%s)", engine->name);
+
if (obj->frontbuffer_bits)
seq_printf(m, " (frontbuffer: 0x%03x)", obj->frontbuffer_bits);
}
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index ed868b2f4f71..9c49b0198691 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1349,27 +1349,30 @@ int
i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
bool readonly)
{
+ struct drm_i915_gem_request *request;
struct reservation_object *resv;
int ret, i;
if (readonly) {
- if (obj->last_write.request) {
- ret = i915_wait_request(obj->last_write.request);
+ request = i915_gem_active_peek(&obj->last_write);
+ if (request) {
+ ret = i915_wait_request(request);
if (ret)
return ret;
- i = obj->last_write.request->engine->id;
- if (obj->last_read[i].request == obj->last_write.request)
+ i = request->engine->id;
+ if (i915_gem_active_peek(&obj->last_read[i]) == request)
i915_gem_object_retire__read(obj, i);
else
i915_gem_object_retire__write(obj);
}
} else {
for (i = 0; i < I915_NUM_ENGINES; i++) {
- if (!obj->last_read[i].request)
+ request = i915_gem_active_peek(&obj->last_read[i]);
+ if (!request)
continue;
- ret = i915_wait_request(obj->last_read[i].request);
+ ret = i915_wait_request(request);
if (ret)
return ret;
@@ -1397,9 +1400,9 @@ i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
{
int idx = req->engine->id;
- if (obj->last_read[idx].request == req)
+ if (i915_gem_active_peek(&obj->last_read[idx]) == req)
i915_gem_object_retire__read(obj, idx);
- else if (obj->last_write.request == req)
+ else if (i915_gem_active_peek(&obj->last_write) == req)
i915_gem_object_retire__write(obj);
if (!i915_reset_in_progress(&req->i915->gpu_error))
@@ -1428,20 +1431,20 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
if (readonly) {
struct drm_i915_gem_request *req;
- req = obj->last_write.request;
+ req = i915_gem_active_get(&obj->last_write);
if (req == NULL)
return 0;
- requests[n++] = i915_gem_request_get(req);
+ requests[n++] = req;
} else {
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read[i].request;
+ req = i915_gem_active_get(&obj->last_read[i]);
if (req == NULL)
continue;
- requests[n++] = i915_gem_request_get(req);
+ requests[n++] = req;
}
}
@@ -2383,8 +2386,8 @@ void i915_vma_move_to_active(struct i915_vma *vma,
static void
i915_gem_object_retire__write(struct drm_i915_gem_object *obj)
{
- GEM_BUG_ON(!obj->last_write.request);
- GEM_BUG_ON(!(obj->active & intel_engine_flag(obj->last_write.request->engine)));
+ GEM_BUG_ON(!__i915_gem_active_is_busy(&obj->last_write));
+ GEM_BUG_ON(!(obj->active & intel_engine_flag(i915_gem_active_get_engine(&obj->last_write))));
i915_gem_active_set(&obj->last_write, NULL);
intel_fb_obj_flush(obj, true, ORIGIN_CS);
@@ -2393,15 +2396,17 @@ i915_gem_object_retire__write(struct drm_i915_gem_object *obj)
static void
i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int idx)
{
+ struct intel_engine_cs *engine;
struct i915_vma *vma;
- GEM_BUG_ON(!obj->last_read[idx].request);
+ GEM_BUG_ON(!__i915_gem_active_is_busy(&obj->last_read[idx]));
GEM_BUG_ON(!(obj->active & (1 << idx)));
list_del_init(&obj->engine_list[idx]);
i915_gem_active_set(&obj->last_read[idx], NULL);
- if (obj->last_write.request && obj->last_write.request->engine->id == idx)
+ engine = i915_gem_active_get_engine(&obj->last_write);
+ if (engine && engine->id == idx)
i915_gem_object_retire__write(obj);
obj->active &= ~(1 << idx);
@@ -2621,7 +2626,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
struct drm_i915_gem_object,
engine_list[engine->id]);
- if (!list_empty(&obj->last_read[engine->id].request->list))
+ if (!list_empty(&i915_gem_active_peek(&obj->last_read[engine->id])->list))
break;
i915_gem_object_retire__read(obj, engine->id);
@@ -2754,7 +2759,7 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read[i].request;
+ req = i915_gem_active_peek(&obj->last_read[i]);
if (req == NULL)
continue;
@@ -2794,7 +2799,7 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
{
struct drm_i915_gem_wait *args = data;
struct drm_i915_gem_object *obj;
- struct drm_i915_gem_request *req[I915_NUM_ENGINES];
+ struct drm_i915_gem_request *requests[I915_NUM_ENGINES];
int i, n = 0;
int ret;
@@ -2830,20 +2835,21 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
i915_gem_object_put(obj);
for (i = 0; i < I915_NUM_ENGINES; i++) {
- if (!obj->last_read[i].request)
- continue;
+ struct drm_i915_gem_request *req;
- req[n++] = i915_gem_request_get(obj->last_read[i].request);
+ req = i915_gem_active_get(&obj->last_read[i]);
+ if (req)
+ requests[n++] = req;
}
mutex_unlock(&dev->struct_mutex);
for (i = 0; i < n; i++) {
if (ret == 0)
- ret = __i915_wait_request(req[i], true,
+ ret = __i915_wait_request(requests[i], true,
args->timeout_ns > 0 ? &args->timeout_ns : NULL,
to_rps_client(file));
- i915_gem_request_put(req[i]);
+ i915_gem_request_put(requests[i]);
}
return ret;
@@ -2916,7 +2922,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,
struct drm_i915_gem_request *to)
{
const bool readonly = obj->base.pending_write_domain == 0;
- struct drm_i915_gem_request *req[I915_NUM_ENGINES];
+ struct drm_i915_gem_request *requests[I915_NUM_ENGINES];
int ret, i, n;
if (!obj->active)
@@ -2924,15 +2930,22 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,
n = 0;
if (readonly) {
- if (obj->last_write.request)
- req[n++] = obj->last_write.request;
+ struct drm_i915_gem_request *req;
+
+ req = i915_gem_active_peek(&obj->last_write);
+ if (req)
+ requests[n++] = req;
} else {
- for (i = 0; i < I915_NUM_ENGINES; i++)
- if (obj->last_read[i].request)
- req[n++] = obj->last_read[i].request;
+ for (i = 0; i < I915_NUM_ENGINES; i++) {
+ struct drm_i915_gem_request *req;
+
+ req = i915_gem_active_peek(&obj->last_read[i]);
+ if (req)
+ requests[n++] = req;
+ }
}
for (i = 0; i < n; i++) {
- ret = __i915_gem_object_sync(obj, to, req[i]);
+ ret = __i915_gem_object_sync(obj, to, requests[i]);
if (ret)
return ret;
}
@@ -4021,17 +4034,17 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
args->busy = 0;
if (obj->active) {
+ struct drm_i915_gem_request *req;
int i;
for (i = 0; i < I915_NUM_ENGINES; i++) {
- struct drm_i915_gem_request *req;
-
- req = obj->last_read[i].request;
+ req = i915_gem_active_peek(&obj->last_read[i]);
if (req)
args->busy |= 1 << (16 + req->engine->exec_id);
}
- if (obj->last_write.request)
- args->busy |= obj->last_write.request->engine->exec_id;
+ req = i915_gem_active_peek(&obj->last_write);
+ if (req)
+ args->busy |= req->engine->exec_id;
}
unref:
diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
index d16b385e79e9..9fdbd66128a6 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence.c
@@ -261,14 +261,13 @@ static inline void i915_gem_object_fence_lost(struct drm_i915_gem_object *obj)
static int
i915_gem_object_wait_fence(struct drm_i915_gem_object *obj)
{
- if (obj->last_fence.request) {
- int ret = i915_wait_request(obj->last_fence.request);
- if (ret)
- return ret;
+ int ret;
- i915_gem_active_set(&obj->last_fence, NULL);
- }
+ ret = i915_gem_active_wait(&obj->last_fence);
+ if (ret)
+ return ret;
+ i915_gem_active_set(&obj->last_fence, NULL);
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index cf2df33f9892..6280812e9c7d 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -280,6 +280,15 @@ struct i915_gem_active {
struct drm_i915_gem_request *request;
};
+/**
+ * i915_gem_active_set - updates the tracker to watch the current request
+ * @active - the active tracker
+ * @request - the request to watch
+ *
+ * i915_gem_active_set() watches the given @request for completion. Whilst
+ * that @request is busy, the @active reports busy. When that @request is
+ * retired, the @active tracker is updated to report idle.
+ */
static inline void
i915_gem_active_set(struct i915_gem_active *active,
struct drm_i915_gem_request *request)
@@ -287,6 +296,124 @@ i915_gem_active_set(struct i915_gem_active *active,
i915_gem_request_assign(&active->request, request);
}
+/**
+ * i915_gem_active_peek - report the request being monitored
+ * @active - the active tracker
+ *
+ * i915_gem_active_peek() returns the current request being tracked, or NULL.
+ * It does not obtain a reference on the request for the caller, so the
+ * caller must hold struct_mutex.
+ */
+static inline struct drm_i915_gem_request *
+i915_gem_active_peek(const struct i915_gem_active *active)
+{
+ return active->request;
+}
+
+/**
+ * i915_gem_active_get - return a reference to the active request
+ * @active - the active tracker
+ *
+ * i915_gem_active_get() returns a reference to the active request, or NULL
+ * if the active tracker is idle. The caller must hold struct_mutex.
+ */
+static inline struct drm_i915_gem_request *
+i915_gem_active_get(const struct i915_gem_active *active)
+{
+ struct drm_i915_gem_request *request;
+
+ request = i915_gem_active_peek(active);
+ if (!request || i915_gem_request_completed(request))
+ return NULL;
+
+ return i915_gem_request_get(request);
+}
+
+/**
+ * __i915_gem_active_is_busy - report whether the active tracker is assigned
+ * @active - the active tracker
+ *
+ * __i915_gem_active_is_busy() returns true if the active tracker is currently
+ * assigned to a request. Due to the lazy retiring, that request may be idle
+ * and this may report stale information.
+ */
+static inline bool
+__i915_gem_active_is_busy(const struct i915_gem_active *active)
+{
+ return i915_gem_active_peek(active);
+}
+
+/**
+ * i915_gem_active_is_idle - report whether the active tracker is idle
+ * @active - the active tracker
+ *
+ * i915_gem_active_is_idle() returns true if the active tracker is currently
+ * unassigned or if the request is complete (but not yet retired). Requires
+ * the caller to hold struct_mutex (but that can be relaxed if desired).
+ */
+static inline bool
+i915_gem_active_is_idle(const struct i915_gem_active *active)
+{
+ struct drm_i915_gem_request *request;
+
+ request = i915_gem_active_peek(active);
+ if (!request || i915_gem_request_completed(request))
+ return true;
+
+ return false;
+}
+
+/**
+ * i915_gem_active_wait - waits until the request is completed
+ * @active - the active request on which to wait
+ *
+ * i915_gem_active_wait() waits until the request is completed before
+ * returning. Note that it does not guarantee that the request is
+ * retired first, see i915_gem_active_retire().
+ */
+static inline int __must_check
+i915_gem_active_wait(const struct i915_gem_active *active)
+{
+ struct drm_i915_gem_request *request;
+
+ request = i915_gem_active_peek(active);
+ if (!request)
+ return 0;
+
+ return i915_wait_request(request);
+}
+
+/**
+ * i915_gem_active_retire - waits until the request is retired
+ * @active - the active request on which to wait
+ *
+ * i915_gem_active_retire() waits until the request is completed,
+ * and then ensures that at least the retirement handler for this
+ * @active tracker is called before returning. If the @active
+ * tracker is idle, the function returns immediately.
+ */
+static inline int __must_check
+i915_gem_active_retire(const struct i915_gem_active *active)
+{
+ return i915_gem_active_wait(active);
+}
+
+/* Convenience functions for peeking at state inside active's request whilst
+ * guarded by the struct_mutex.
+ */
+
+static inline uint32_t
+i915_gem_active_get_seqno(const struct i915_gem_active *active)
+{
+ return i915_gem_request_get_seqno(i915_gem_active_peek(active));
+}
+
+static inline struct intel_engine_cs *
+i915_gem_active_get_engine(const struct i915_gem_active *active)
+{
+ return i915_gem_request_get_engine(i915_gem_active_peek(active));
+}
+
#define for_each_active(mask, idx) \
for (; mask ? idx = ffs(mask) - 1, 1 : 0; mask &= ~BIT(idx))
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index 00d796da65fb..8cef2d6b291a 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -242,7 +242,7 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,
}
obj->fence_dirty =
- obj->last_fence.request ||
+ !i915_gem_active_is_idle(&obj->last_fence) ||
obj->fence_reg != I915_FENCE_REG_NONE;
obj->tiling_mode = args->tiling_mode;
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index 32f50a70ea42..00ab5e9d2eb7 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -74,11 +74,9 @@ static void wait_rendering(struct drm_i915_gem_object *obj)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = obj->last_read[i].request;
- if (req == NULL)
- continue;
-
- requests[n++] = i915_gem_request_get(req);
+ req = i915_gem_active_get(&obj->last_read[i]);
+ if (req)
+ requests[n++] = req;
}
mutex_unlock(&dev->struct_mutex);
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index d6482e980d5e..585fe2bdadd5 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -746,13 +746,14 @@ static void capture_bo(struct drm_i915_error_buffer *err,
struct i915_vma *vma)
{
struct drm_i915_gem_object *obj = vma->obj;
+ struct intel_engine_cs *engine;
int i;
err->size = obj->base.size;
err->name = obj->base.name;
for (i = 0; i < I915_NUM_ENGINES; i++)
- err->rseqno[i] = i915_gem_request_get_seqno(obj->last_read[i].request);
- err->wseqno = i915_gem_request_get_seqno(obj->last_write.request);
+ err->rseqno[i] = i915_gem_active_get_seqno(&obj->last_read[i]);
+ err->wseqno = i915_gem_active_get_seqno(&obj->last_write);
err->gtt_offset = vma->node.start;
err->read_domains = obj->base.read_domains;
err->write_domain = obj->base.write_domain;
@@ -764,8 +765,10 @@ static void capture_bo(struct drm_i915_error_buffer *err,
err->dirty = obj->dirty;
err->purgeable = obj->madv != I915_MADV_WILLNEED;
err->userptr = obj->userptr.mm != NULL;
- err->engine = obj->last_write.request ? obj->last_write.request->engine->id : -1;
err->cache_level = obj->cache_level;
+
+ engine = i915_gem_active_get_engine(&obj->last_write);
+ err->engine = engine ? engine->id : -1;
}
static u32 capture_active_bo(struct drm_i915_error_buffer *err,
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index e858fede37ce..8c03d13d494d 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -11370,7 +11370,7 @@ static bool use_mmio_flip(struct intel_engine_cs *engine,
if (resv && !reservation_object_test_signaled_rcu(resv, false))
return true;
- return engine != i915_gem_request_get_engine(obj->last_write.request);
+ return engine != i915_gem_active_get_engine(&obj->last_write);
}
static void skl_do_mmio_flip(struct intel_crtc *intel_crtc,
@@ -11673,7 +11673,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
} else if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) {
engine = &dev_priv->engine[BCS];
} else if (INTEL_INFO(dev)->gen >= 7) {
- engine = i915_gem_request_get_engine(obj->last_write.request);
+ engine = i915_gem_active_get_engine(&obj->last_write);
if (engine == NULL || engine->id != RCS)
engine = &dev_priv->engine[BCS];
} else {
@@ -11694,9 +11694,7 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
if (mmio_flip) {
INIT_WORK(&work->mmio_work, intel_mmio_flip_work_func);
- i915_gem_request_assign(&work->flip_queued_req,
- obj->last_write.request);
-
+ work->flip_queued_req = i915_gem_active_get(&obj->last_write);
schedule_work(&work->mmio_work);
} else {
request = i915_gem_request_alloc(engine, engine->last_context);
@@ -14039,11 +14037,8 @@ intel_prepare_plane_fb(struct drm_plane *plane,
}
if (ret == 0) {
- struct intel_plane_state *plane_state =
- to_intel_plane_state(new_state);
-
- i915_gem_request_assign(&plane_state->wait_req,
- obj->last_write.request);
+ to_intel_plane_state(new_state)->wait_req =
+ i915_gem_active_get(&obj->last_write);
}
return ret;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 12/28] drm/i915: Mark up i915_gem_active for locking annotation
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (10 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 11/28] drm/i915: Prepare i915_gem_active for annotations Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 13/28] drm/i915: Refactor blocking waits Chris Wilson
` (17 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
The future annotations will track the locking used for access to ensure
that it is always sufficient. We make the preparations now to present
the API ahead and to make sure that GCC can eliminate the unused
parameter.
Before: 6298417 3619610 696320 10614347 a1f64b vmlinux
After: 6298417 3619610 696320 10614347 a1f64b vmlinux
(with i915 builtin)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 12 +++++---
drivers/gpu/drm/i915/i915_gem.c | 49 ++++++++++++++++++++++-----------
drivers/gpu/drm/i915/i915_gem_fence.c | 3 +-
drivers/gpu/drm/i915/i915_gem_request.h | 38 +++++++++++++++----------
drivers/gpu/drm/i915/i915_gem_tiling.c | 3 +-
drivers/gpu/drm/i915/i915_gem_userptr.c | 3 +-
drivers/gpu/drm/i915/i915_gpu_error.c | 29 +++++++++++++++----
drivers/gpu/drm/i915/intel_display.c | 12 +++++---
8 files changed, 102 insertions(+), 47 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 24ff7f4bc412..c595cc8169cb 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -155,10 +155,13 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
obj->base.write_domain);
for_each_engine_id(engine, dev_priv, id)
seq_printf(m, "%x ",
- i915_gem_active_get_seqno(&obj->last_read[id]));
+ i915_gem_active_get_seqno(&obj->last_read[id],
+ &obj->base.dev->struct_mutex));
seq_printf(m, "] %x %x%s%s%s",
- i915_gem_active_get_seqno(&obj->last_write),
- i915_gem_active_get_seqno(&obj->last_fence),
+ i915_gem_active_get_seqno(&obj->last_write,
+ &obj->base.dev->struct_mutex),
+ i915_gem_active_get_seqno(&obj->last_fence,
+ &obj->base.dev->struct_mutex),
i915_cache_level_str(to_i915(obj->base.dev), obj->cache_level),
obj->dirty ? " dirty" : "",
obj->madv == I915_MADV_DONTNEED ? " purgeable" : "");
@@ -196,7 +199,8 @@ describe_obj(struct seq_file *m, struct drm_i915_gem_object *obj)
seq_printf(m, " (%s mappable)", s);
}
- engine = i915_gem_active_get_engine(&obj->last_write);
+ engine = i915_gem_active_get_engine(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (engine)
seq_printf(m, " (%s)", engine->name);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 9c49b0198691..ccd2c6c894f4 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1354,21 +1354,24 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
int ret, i;
if (readonly) {
- request = i915_gem_active_peek(&obj->last_write);
+ request = i915_gem_active_peek(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (request) {
ret = i915_wait_request(request);
if (ret)
return ret;
i = request->engine->id;
- if (i915_gem_active_peek(&obj->last_read[i]) == request)
+ if (i915_gem_active_peek(&obj->last_read[i],
+ &obj->base.dev->struct_mutex) == request)
i915_gem_object_retire__read(obj, i);
else
i915_gem_object_retire__write(obj);
}
} else {
for (i = 0; i < I915_NUM_ENGINES; i++) {
- request = i915_gem_active_peek(&obj->last_read[i]);
+ request = i915_gem_active_peek(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (!request)
continue;
@@ -1400,9 +1403,11 @@ i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
{
int idx = req->engine->id;
- if (i915_gem_active_peek(&obj->last_read[idx]) == req)
+ if (i915_gem_active_peek(&obj->last_read[idx],
+ &obj->base.dev->struct_mutex) == req)
i915_gem_object_retire__read(obj, idx);
- else if (i915_gem_active_peek(&obj->last_write) == req)
+ else if (i915_gem_active_peek(&obj->last_write,
+ &obj->base.dev->struct_mutex) == req)
i915_gem_object_retire__write(obj);
if (!i915_reset_in_progress(&req->i915->gpu_error))
@@ -1431,7 +1436,8 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
if (readonly) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_get(&obj->last_write);
+ req = i915_gem_active_get(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (req == NULL)
return 0;
@@ -1440,7 +1446,8 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_get(&obj->last_read[i]);
+ req = i915_gem_active_get(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (req == NULL)
continue;
@@ -2387,7 +2394,9 @@ static void
i915_gem_object_retire__write(struct drm_i915_gem_object *obj)
{
GEM_BUG_ON(!__i915_gem_active_is_busy(&obj->last_write));
- GEM_BUG_ON(!(obj->active & intel_engine_flag(i915_gem_active_get_engine(&obj->last_write))));
+ GEM_BUG_ON(!(obj->active &
+ intel_engine_flag(i915_gem_active_get_engine(&obj->last_write,
+ &obj->base.dev->struct_mutex))));
i915_gem_active_set(&obj->last_write, NULL);
intel_fb_obj_flush(obj, true, ORIGIN_CS);
@@ -2405,7 +2414,8 @@ i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int idx)
list_del_init(&obj->engine_list[idx]);
i915_gem_active_set(&obj->last_read[idx], NULL);
- engine = i915_gem_active_get_engine(&obj->last_write);
+ engine = i915_gem_active_get_engine(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (engine && engine->id == idx)
i915_gem_object_retire__write(obj);
@@ -2626,7 +2636,8 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
struct drm_i915_gem_object,
engine_list[engine->id]);
- if (!list_empty(&i915_gem_active_peek(&obj->last_read[engine->id])->list))
+ if (!list_empty(&i915_gem_active_peek(&obj->last_read[engine->id],
+ &obj->base.dev->struct_mutex)->list))
break;
i915_gem_object_retire__read(obj, engine->id);
@@ -2759,7 +2770,8 @@ i915_gem_object_flush_active(struct drm_i915_gem_object *obj)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_peek(&obj->last_read[i]);
+ req = i915_gem_active_peek(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (req == NULL)
continue;
@@ -2837,7 +2849,8 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_get(&obj->last_read[i]);
+ req = i915_gem_active_get(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (req)
requests[n++] = req;
}
@@ -2932,14 +2945,16 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,
if (readonly) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_peek(&obj->last_write);
+ req = i915_gem_active_peek(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (req)
requests[n++] = req;
} else {
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_peek(&obj->last_read[i]);
+ req = i915_gem_active_peek(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (req)
requests[n++] = req;
}
@@ -4038,11 +4053,13 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
int i;
for (i = 0; i < I915_NUM_ENGINES; i++) {
- req = i915_gem_active_peek(&obj->last_read[i]);
+ req = i915_gem_active_peek(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (req)
args->busy |= 1 << (16 + req->engine->exec_id);
}
- req = i915_gem_active_peek(&obj->last_write);
+ req = i915_gem_active_peek(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (req)
args->busy |= req->engine->exec_id;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
index 9fdbd66128a6..a4ec4fecbea0 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence.c
@@ -263,7 +263,8 @@ i915_gem_object_wait_fence(struct drm_i915_gem_object *obj)
{
int ret;
- ret = i915_gem_active_wait(&obj->last_fence);
+ ret = i915_gem_active_wait(&obj->last_fence,
+ &obj->base.dev->struct_mutex);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 6280812e9c7d..9e56a2c63b0e 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -296,6 +296,12 @@ i915_gem_active_set(struct i915_gem_active *active,
i915_gem_request_assign(&active->request, request);
}
+static inline struct drm_i915_gem_request *
+__i915_gem_active_peek(const struct i915_gem_active *active)
+{
+ return active->request;
+}
+
/**
* i915_gem_active_peek - report the request being monitored
* @active - the active tracker
@@ -305,7 +311,7 @@ i915_gem_active_set(struct i915_gem_active *active,
* caller must hold struct_mutex.
*/
static inline struct drm_i915_gem_request *
-i915_gem_active_peek(const struct i915_gem_active *active)
+i915_gem_active_peek(const struct i915_gem_active *active, struct mutex *mutex)
{
return active->request;
}
@@ -318,11 +324,11 @@ i915_gem_active_peek(const struct i915_gem_active *active)
* if the active tracker is idle. The caller must hold struct_mutex.
*/
static inline struct drm_i915_gem_request *
-i915_gem_active_get(const struct i915_gem_active *active)
+i915_gem_active_get(const struct i915_gem_active *active, struct mutex *mutex)
{
struct drm_i915_gem_request *request;
- request = i915_gem_active_peek(active);
+ request = i915_gem_active_peek(active, mutex);
if (!request || i915_gem_request_completed(request))
return NULL;
@@ -340,7 +346,7 @@ i915_gem_active_get(const struct i915_gem_active *active)
static inline bool
__i915_gem_active_is_busy(const struct i915_gem_active *active)
{
- return i915_gem_active_peek(active);
+ return __i915_gem_active_peek(active);
}
/**
@@ -352,11 +358,12 @@ __i915_gem_active_is_busy(const struct i915_gem_active *active)
* the caller to hold struct_mutex (but that can be relaxed if desired).
*/
static inline bool
-i915_gem_active_is_idle(const struct i915_gem_active *active)
+i915_gem_active_is_idle(const struct i915_gem_active *active,
+ struct mutex *mutex)
{
struct drm_i915_gem_request *request;
- request = i915_gem_active_peek(active);
+ request = i915_gem_active_peek(active, mutex);
if (!request || i915_gem_request_completed(request))
return true;
@@ -372,11 +379,11 @@ i915_gem_active_is_idle(const struct i915_gem_active *active)
* retired first, see i915_gem_active_retire().
*/
static inline int __must_check
-i915_gem_active_wait(const struct i915_gem_active *active)
+i915_gem_active_wait(const struct i915_gem_active *active, struct mutex *mutex)
{
struct drm_i915_gem_request *request;
- request = i915_gem_active_peek(active);
+ request = i915_gem_active_peek(active, mutex);
if (!request)
return 0;
@@ -393,9 +400,10 @@ i915_gem_active_wait(const struct i915_gem_active *active)
* tracker is idle, the function returns immediately.
*/
static inline int __must_check
-i915_gem_active_retire(const struct i915_gem_active *active)
+i915_gem_active_retire(const struct i915_gem_active *active,
+ struct mutex *mutex)
{
- return i915_gem_active_wait(active);
+ return i915_gem_active_wait(active, mutex);
}
/* Convenience functions for peeking at state inside active's request whilst
@@ -403,15 +411,17 @@ i915_gem_active_retire(const struct i915_gem_active *active)
*/
static inline uint32_t
-i915_gem_active_get_seqno(const struct i915_gem_active *active)
+i915_gem_active_get_seqno(const struct i915_gem_active *active,
+ struct mutex *mutex)
{
- return i915_gem_request_get_seqno(i915_gem_active_peek(active));
+ return i915_gem_request_get_seqno(i915_gem_active_peek(active, mutex));
}
static inline struct intel_engine_cs *
-i915_gem_active_get_engine(const struct i915_gem_active *active)
+i915_gem_active_get_engine(const struct i915_gem_active *active,
+ struct mutex *mutex)
{
- return i915_gem_request_get_engine(i915_gem_active_peek(active));
+ return i915_gem_request_get_engine(i915_gem_active_peek(active, mutex));
}
#define for_each_active(mask, idx) \
diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c
index 8cef2d6b291a..fa2eb4a4f212 100644
--- a/drivers/gpu/drm/i915/i915_gem_tiling.c
+++ b/drivers/gpu/drm/i915/i915_gem_tiling.c
@@ -242,7 +242,8 @@ i915_gem_set_tiling(struct drm_device *dev, void *data,
}
obj->fence_dirty =
- !i915_gem_active_is_idle(&obj->last_fence) ||
+ !i915_gem_active_is_idle(&obj->last_fence,
+ &dev->struct_mutex) ||
obj->fence_reg != I915_FENCE_REG_NONE;
obj->tiling_mode = args->tiling_mode;
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index 00ab5e9d2eb7..e57521dbddc6 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -74,7 +74,8 @@ static void wait_rendering(struct drm_i915_gem_object *obj)
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
- req = i915_gem_active_get(&obj->last_read[i]);
+ req = i915_gem_active_get(&obj->last_read[i],
+ &obj->base.dev->struct_mutex);
if (req)
requests[n++] = req;
}
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 585fe2bdadd5..0d2882ae7c89 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -742,18 +742,38 @@ unwind:
#define i915_error_ggtt_object_create(dev_priv, src) \
i915_error_object_create((dev_priv), (src), &(dev_priv)->ggtt.base)
+/* The error capture is special as tries to run underneath the normal
+ * locking rules - so we use the raw version of the i915_gem_active lookup.
+ */
+static inline uint32_t
+__active_get_seqno(struct i915_gem_active *active)
+{
+ return i915_gem_request_get_seqno(__i915_gem_active_peek(active));
+}
+
+static inline int
+__active_get_engine_id(struct i915_gem_active *active)
+{
+ struct intel_engine_cs *engine;
+
+ engine = i915_gem_request_get_engine(__i915_gem_active_peek(active));
+ return engine ? engine->id : -1;
+}
+
static void capture_bo(struct drm_i915_error_buffer *err,
struct i915_vma *vma)
{
struct drm_i915_gem_object *obj = vma->obj;
- struct intel_engine_cs *engine;
int i;
err->size = obj->base.size;
err->name = obj->base.name;
+
for (i = 0; i < I915_NUM_ENGINES; i++)
- err->rseqno[i] = i915_gem_active_get_seqno(&obj->last_read[i]);
- err->wseqno = i915_gem_active_get_seqno(&obj->last_write);
+ err->rseqno[i] = __active_get_seqno(&obj->last_read[i]);
+ err->wseqno = __active_get_seqno(&obj->last_write);
+ err->engine = __active_get_engine_id(&obj->last_write);
+
err->gtt_offset = vma->node.start;
err->read_domains = obj->base.read_domains;
err->write_domain = obj->base.write_domain;
@@ -766,9 +786,6 @@ static void capture_bo(struct drm_i915_error_buffer *err,
err->purgeable = obj->madv != I915_MADV_WILLNEED;
err->userptr = obj->userptr.mm != NULL;
err->cache_level = obj->cache_level;
-
- engine = i915_gem_active_get_engine(&obj->last_write);
- err->engine = engine ? engine->id : -1;
}
static u32 capture_active_bo(struct drm_i915_error_buffer *err,
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index 8c03d13d494d..d54a3ea56536 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -11370,7 +11370,8 @@ static bool use_mmio_flip(struct intel_engine_cs *engine,
if (resv && !reservation_object_test_signaled_rcu(resv, false))
return true;
- return engine != i915_gem_active_get_engine(&obj->last_write);
+ return engine != i915_gem_active_get_engine(&obj->last_write,
+ &obj->base.dev->struct_mutex);
}
static void skl_do_mmio_flip(struct intel_crtc *intel_crtc,
@@ -11673,7 +11674,8 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
} else if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) {
engine = &dev_priv->engine[BCS];
} else if (INTEL_INFO(dev)->gen >= 7) {
- engine = i915_gem_active_get_engine(&obj->last_write);
+ engine = i915_gem_active_get_engine(&obj->last_write,
+ &obj->base.dev->struct_mutex);
if (engine == NULL || engine->id != RCS)
engine = &dev_priv->engine[BCS];
} else {
@@ -11694,7 +11696,8 @@ static int intel_crtc_page_flip(struct drm_crtc *crtc,
if (mmio_flip) {
INIT_WORK(&work->mmio_work, intel_mmio_flip_work_func);
- work->flip_queued_req = i915_gem_active_get(&obj->last_write);
+ work->flip_queued_req = i915_gem_active_get(&obj->last_write,
+ &obj->base.dev->struct_mutex);
schedule_work(&work->mmio_work);
} else {
request = i915_gem_request_alloc(engine, engine->last_context);
@@ -14038,7 +14041,8 @@ intel_prepare_plane_fb(struct drm_plane *plane,
if (ret == 0) {
to_intel_plane_state(new_state)->wait_req =
- i915_gem_active_get(&obj->last_write);
+ i915_gem_active_get(&obj->last_write,
+ &obj->base.dev->struct_mutex);
}
return ret;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 13/28] drm/i915: Refactor blocking waits
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (11 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 12/28] drm/i915: Mark up i915_gem_active for locking annotation Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 14/28] drm/i915: Rename request->list to link for consistency Chris Wilson
` (16 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
Tidy up the for loops that handle waiting for read/write vs read-only
access.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 158 +++++++++++++++++++---------------------
1 file changed, 75 insertions(+), 83 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index ccd2c6c894f4..586b69430e83 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1339,6 +1339,23 @@ put_rpm:
return ret;
}
+static void
+i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
+ struct drm_i915_gem_request *req)
+{
+ int idx = req->engine->id;
+
+ if (i915_gem_active_peek(&obj->last_read[idx],
+ &obj->base.dev->struct_mutex) == req)
+ i915_gem_object_retire__read(obj, idx);
+ else if (i915_gem_active_peek(&obj->last_write,
+ &obj->base.dev->struct_mutex) == req)
+ i915_gem_object_retire__write(obj);
+
+ if (!i915_reset_in_progress(&req->i915->gpu_error))
+ i915_gem_request_retire_upto(req);
+}
+
/**
* Ensures that all rendering to the object has completed and the object is
* safe to unbind from the GTT or access from the CPU.
@@ -1349,39 +1366,34 @@ int
i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
bool readonly)
{
- struct drm_i915_gem_request *request;
struct reservation_object *resv;
- int ret, i;
+ struct i915_gem_active *active;
+ unsigned long active_mask;
+ int idx, ret;
- if (readonly) {
- request = i915_gem_active_peek(&obj->last_write,
- &obj->base.dev->struct_mutex);
- if (request) {
- ret = i915_wait_request(request);
- if (ret)
- return ret;
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
- i = request->engine->id;
- if (i915_gem_active_peek(&obj->last_read[i],
- &obj->base.dev->struct_mutex) == request)
- i915_gem_object_retire__read(obj, i);
- else
- i915_gem_object_retire__write(obj);
- }
+ if (!readonly) {
+ active = obj->last_read;
+ active_mask = obj->active;
} else {
- for (i = 0; i < I915_NUM_ENGINES; i++) {
- request = i915_gem_active_peek(&obj->last_read[i],
- &obj->base.dev->struct_mutex);
- if (!request)
- continue;
+ active_mask = 1;
+ active = &obj->last_write;
+ }
- ret = i915_wait_request(request);
- if (ret)
- return ret;
+ for_each_active(active_mask, idx) {
+ struct drm_i915_gem_request *request;
- i915_gem_object_retire__read(obj, i);
- }
- GEM_BUG_ON(obj->active);
+ request = i915_gem_active_peek(&active[idx],
+ &obj->base.dev->struct_mutex);
+ if (!request)
+ continue;
+
+ ret = i915_wait_request(request);
+ if (ret)
+ return ret;
+
+ i915_gem_object_retire_request(obj, request);
}
resv = i915_gem_object_get_dmabuf_resv(obj);
@@ -1397,23 +1409,6 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
return 0;
}
-static void
-i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
- struct drm_i915_gem_request *req)
-{
- int idx = req->engine->id;
-
- if (i915_gem_active_peek(&obj->last_read[idx],
- &obj->base.dev->struct_mutex) == req)
- i915_gem_object_retire__read(obj, idx);
- else if (i915_gem_active_peek(&obj->last_write,
- &obj->base.dev->struct_mutex) == req)
- i915_gem_object_retire__write(obj);
-
- if (!i915_reset_in_progress(&req->i915->gpu_error))
- i915_gem_request_retire_upto(req);
-}
-
/* A nonblocking variant of the above wait. This is a highly dangerous routine
* as the object state may change during this call.
*/
@@ -1425,34 +1420,31 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
struct drm_device *dev = obj->base.dev;
struct drm_i915_private *dev_priv = to_i915(dev);
struct drm_i915_gem_request *requests[I915_NUM_ENGINES];
+ struct i915_gem_active *active;
+ unsigned long active_mask;
int ret, i, n = 0;
BUG_ON(!mutex_is_locked(&dev->struct_mutex));
BUG_ON(!dev_priv->mm.interruptible);
- if (!obj->active)
+ active_mask = obj->active;
+ if (!active_mask)
return 0;
- if (readonly) {
- struct drm_i915_gem_request *req;
-
- req = i915_gem_active_get(&obj->last_write,
- &obj->base.dev->struct_mutex);
- if (req == NULL)
- return 0;
-
- requests[n++] = req;
+ if (!readonly) {
+ active = obj->last_read;
} else {
- for (i = 0; i < I915_NUM_ENGINES; i++) {
- struct drm_i915_gem_request *req;
+ active_mask = 1;
+ active = &obj->last_write;
+ }
- req = i915_gem_active_get(&obj->last_read[i],
- &obj->base.dev->struct_mutex);
- if (req == NULL)
- continue;
+ for_each_active(active_mask, i) {
+ struct drm_i915_gem_request *req;
+ req = i915_gem_active_get(&active[i],
+ &obj->base.dev->struct_mutex);
+ if (req)
requests[n++] = req;
- }
}
mutex_unlock(&dev->struct_mutex);
@@ -2934,33 +2926,33 @@ int
i915_gem_object_sync(struct drm_i915_gem_object *obj,
struct drm_i915_gem_request *to)
{
- const bool readonly = obj->base.pending_write_domain == 0;
- struct drm_i915_gem_request *requests[I915_NUM_ENGINES];
- int ret, i, n;
+ struct i915_gem_active *active;
+ unsigned long active_mask;
+ int idx;
- if (!obj->active)
- return 0;
+ lockdep_assert_held(&obj->base.dev->struct_mutex);
- n = 0;
- if (readonly) {
- struct drm_i915_gem_request *req;
+ active_mask = obj->active;
+ if (!active_mask)
+ return 0;
- req = i915_gem_active_peek(&obj->last_write,
- &obj->base.dev->struct_mutex);
- if (req)
- requests[n++] = req;
+ if (obj->base.pending_write_domain) {
+ active = obj->last_read;
} else {
- for (i = 0; i < I915_NUM_ENGINES; i++) {
- struct drm_i915_gem_request *req;
-
- req = i915_gem_active_peek(&obj->last_read[i],
- &obj->base.dev->struct_mutex);
- if (req)
- requests[n++] = req;
- }
+ active_mask = 1;
+ active = &obj->last_write;
}
- for (i = 0; i < n; i++) {
- ret = __i915_gem_object_sync(obj, to, requests[i]);
+
+ for_each_active(active_mask, idx) {
+ struct drm_i915_gem_request *request;
+ int ret;
+
+ request = i915_gem_active_peek(&active[idx],
+ &obj->base.dev->struct_mutex);
+ if (!request)
+ continue;
+
+ ret = __i915_gem_object_sync(obj, to, request);
if (ret)
return ret;
}
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 14/28] drm/i915: Rename request->list to link for consistency
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (12 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 13/28] drm/i915: Refactor blocking waits Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 15/28] drm/i915: Remove obsolete i915_gem_object_flush_active() Chris Wilson
` (15 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
We use "list" to denote the list and "link" to denote an element on that
list. Rename request->list to match this idiom.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 4 ++--
drivers/gpu/drm/i915/i915_gem.c | 10 +++++-----
drivers/gpu/drm/i915/i915_gem_request.c | 12 ++++++------
drivers/gpu/drm/i915/i915_gem_request.h | 4 ++--
drivers/gpu/drm/i915/i915_gpu_error.c | 4 ++--
drivers/gpu/drm/i915/intel_ringbuffer.c | 6 +++---
6 files changed, 20 insertions(+), 20 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index c595cc8169cb..fe3c82339c8e 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -746,13 +746,13 @@ static int i915_gem_request_info(struct seq_file *m, void *data)
int count;
count = 0;
- list_for_each_entry(req, &engine->request_list, list)
+ list_for_each_entry(req, &engine->request_list, link)
count++;
if (count == 0)
continue;
seq_printf(m, "%s requests: %d\n", engine->name, count);
- list_for_each_entry(req, &engine->request_list, list) {
+ list_for_each_entry(req, &engine->request_list, link) {
struct task_struct *task;
rcu_read_lock();
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 586b69430e83..04931047cee3 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2475,7 +2475,7 @@ i915_gem_find_active_request(struct intel_engine_cs *engine)
* extra delay for a recent interrupt is pointless. Hence, we do
* not need an engine->irq_seqno_barrier() before the seqno reads.
*/
- list_for_each_entry(request, &engine->request_list, list) {
+ list_for_each_entry(request, &engine->request_list, link) {
if (i915_gem_request_completed(request))
continue;
@@ -2497,7 +2497,7 @@ static void i915_gem_reset_engine_status(struct intel_engine_cs *engine)
ring_hung = engine->hangcheck.score >= HANGCHECK_SCORE_RING_HUNG;
i915_set_reset_status(request->ctx, ring_hung);
- list_for_each_entry_continue(request, &engine->request_list, list)
+ list_for_each_entry_continue(request, &engine->request_list, link)
i915_set_reset_status(request->ctx, false);
}
@@ -2546,7 +2546,7 @@ static void i915_gem_reset_engine_cleanup(struct intel_engine_cs *engine)
request = list_last_entry(&engine->request_list,
struct drm_i915_gem_request,
- list);
+ link);
i915_gem_request_retire_upto(request);
}
@@ -2609,7 +2609,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
request = list_first_entry(&engine->request_list,
struct drm_i915_gem_request,
- list);
+ link);
if (!i915_gem_request_completed(request))
break;
@@ -2629,7 +2629,7 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
engine_list[engine->id]);
if (!list_empty(&i915_gem_active_peek(&obj->last_read[engine->id],
- &obj->base.dev->struct_mutex)->list))
+ &obj->base.dev->struct_mutex)->link))
break;
i915_gem_object_retire__read(obj, engine->id);
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 11c19e7f82fa..7802156d481a 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -160,7 +160,7 @@ i915_gem_request_remove_from_client(struct drm_i915_gem_request *request)
static void i915_gem_request_retire(struct drm_i915_gem_request *request)
{
trace_i915_gem_request_retire(request);
- list_del_init(&request->list);
+ list_del_init(&request->link);
/* We know the GPU must have read the request to have
* sent us the seqno + interrupt, so use the position
@@ -191,12 +191,12 @@ void i915_gem_request_retire_upto(struct drm_i915_gem_request *req)
lockdep_assert_held(&req->i915->drm.struct_mutex);
- if (list_empty(&req->list))
+ if (list_empty(&req->link))
return;
do {
tmp = list_first_entry(&engine->request_list,
- typeof(*tmp), list);
+ typeof(*tmp), link);
i915_gem_request_retire(tmp);
} while (tmp != req);
@@ -317,7 +317,7 @@ i915_gem_request_alloc(struct intel_engine_cs *engine,
/* Move the oldest request to the slab-cache (if not in use!) */
req = list_first_entry_or_null(&engine->request_list,
- typeof(*req), list);
+ typeof(*req), link);
if (req && i915_gem_request_completed(req))
i915_gem_request_retire(req);
@@ -450,7 +450,7 @@ void __i915_add_request(struct drm_i915_gem_request *request,
request->emitted_jiffies = jiffies;
request->previous_seqno = engine->last_submitted_seqno;
smp_store_mb(engine->last_submitted_seqno, request->fence.seqno);
- list_add_tail(&request->list, &engine->request_list);
+ list_add_tail(&request->link, &engine->request_list);
/* Record the position of the start of the request so that
* should we detect the updated seqno part-way through the
@@ -570,7 +570,7 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
might_sleep();
- if (list_empty(&req->list))
+ if (list_empty(&req->link))
return 0;
if (i915_gem_request_completed(req))
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 9e56a2c63b0e..cb3fa58d4eb5 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -105,8 +105,8 @@ struct drm_i915_gem_request {
/** Time at which this request was emitted, in jiffies. */
unsigned long emitted_jiffies;
- /** global list entry for this request */
- struct list_head list;
+ /** engine->request_list entry for this request */
+ struct list_head link;
struct drm_i915_file_private *file_priv;
/** file_priv list entry for this request */
diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c
index 0d2882ae7c89..c19f72e1bcf7 100644
--- a/drivers/gpu/drm/i915/i915_gpu_error.c
+++ b/drivers/gpu/drm/i915/i915_gpu_error.c
@@ -1168,7 +1168,7 @@ static void i915_gem_record_rings(struct drm_i915_private *dev_priv,
i915_gem_record_active_context(engine, error, ee);
count = 0;
- list_for_each_entry(request, &engine->request_list, list)
+ list_for_each_entry(request, &engine->request_list, link)
count++;
ee->num_requests = count;
@@ -1180,7 +1180,7 @@ static void i915_gem_record_rings(struct drm_i915_private *dev_priv,
}
count = 0;
- list_for_each_entry(request, &engine->request_list, list) {
+ list_for_each_entry(request, &engine->request_list, link) {
struct drm_i915_error_request *erq;
if (count >= ee->num_requests) {
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index ecf4278fdabb..76ddc7cc6d25 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2235,7 +2235,7 @@ int intel_engine_idle(struct intel_engine_cs *engine)
req = list_entry(engine->request_list.prev,
struct drm_i915_gem_request,
- list);
+ link);
/* Make sure we do not trigger any retires */
return __i915_wait_request(req,
@@ -2284,7 +2284,7 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
*/
GEM_BUG_ON(!req->reserved_space);
- list_for_each_entry(target, &engine->request_list, list) {
+ list_for_each_entry(target, &engine->request_list, link) {
unsigned space;
/*
@@ -2302,7 +2302,7 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
break;
}
- if (WARN_ON(&target->list == &engine->request_list))
+ if (WARN_ON(&target->link == &engine->request_list))
return -ENOSPC;
return i915_wait_request(target);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 15/28] drm/i915: Remove obsolete i915_gem_object_flush_active()
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (13 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 14/28] drm/i915: Rename request->list to link for consistency Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 16/28] drm/i915: Refactor activity tracking for requests Chris Wilson
` (14 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
Since we track requests, and requests are always added to the GPU fully
formed, we never have to flush the incomplete request and know that the
given request will eventually complete without any further action on our
part.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 59 +++--------------------------------------
1 file changed, 3 insertions(+), 56 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 04931047cee3..c9987e26670e 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2746,35 +2746,6 @@ out_rearm:
}
/**
- * Ensures that an object will eventually get non-busy by flushing any required
- * write domains, emitting any outstanding lazy request and retiring and
- * completed requests.
- * @obj: object to flush
- */
-static int
-i915_gem_object_flush_active(struct drm_i915_gem_object *obj)
-{
- int i;
-
- if (!obj->active)
- return 0;
-
- for (i = 0; i < I915_NUM_ENGINES; i++) {
- struct drm_i915_gem_request *req;
-
- req = i915_gem_active_peek(&obj->last_read[i],
- &obj->base.dev->struct_mutex);
- if (req == NULL)
- continue;
-
- if (i915_gem_request_completed(req))
- i915_gem_object_retire__read(obj, i);
- }
-
- return 0;
-}
-
-/**
* i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
* @dev: drm device pointer
* @data: ioctl data blob
@@ -2820,24 +2791,9 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
return -ENOENT;
}
- /* Need to make sure the object gets inactive eventually. */
- ret = i915_gem_object_flush_active(obj);
- if (ret)
- goto out;
-
if (!obj->active)
goto out;
- /* Do this after OLR check to make sure we make forward progress polling
- * on this IOCTL with a timeout == 0 (like busy ioctl)
- */
- if (args->timeout_ns == 0) {
- ret = -ETIME;
- goto out;
- }
-
- i915_gem_object_put(obj);
-
for (i = 0; i < I915_NUM_ENGINES; i++) {
struct drm_i915_gem_request *req;
@@ -2847,6 +2803,8 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
requests[n++] = req;
}
+out:
+ i915_gem_object_put(obj);
mutex_unlock(&dev->struct_mutex);
for (i = 0; i < n; i++) {
@@ -2857,11 +2815,6 @@ i915_gem_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
i915_gem_request_put(requests[i]);
}
return ret;
-
-out:
- i915_gem_object_put(obj);
- mutex_unlock(&dev->struct_mutex);
- return ret;
}
static int
@@ -4032,13 +3985,8 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
/* Count all active objects as busy, even if they are currently not used
* by the gpu. Users of this interface expect objects to eventually
- * become non-busy without any further actions, therefore emit any
- * necessary flushes here.
+ * become non-busy without any further actions.
*/
- ret = i915_gem_object_flush_active(obj);
- if (ret)
- goto unref;
-
args->busy = 0;
if (obj->active) {
struct drm_i915_gem_request *req;
@@ -4056,7 +4004,6 @@ i915_gem_busy_ioctl(struct drm_device *dev, void *data,
args->busy |= req->engine->exec_id;
}
-unref:
i915_gem_object_put(obj);
unlock:
mutex_unlock(&dev->struct_mutex);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 16/28] drm/i915: Refactor activity tracking for requests
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (14 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 15/28] drm/i915: Remove obsolete i915_gem_object_flush_active() Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 17/28] drm/i915: Track requests inside each intel_ring Chris Wilson
` (13 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
With the introduction of requests, we amplified the number of atomic
refcounted objects we use and update every execbuffer; from none to
several references, and a set of references that need to be changed. We
also introduced interesting side-effects in the order of retiring
requests and objects.
Instead of independently tracking the last request for an object, track
the active objects for each request. The object will reside in the
buffer list of its most recent active request and so we reduce the kref
interchange to a list_move. Now retirements are entirely driven by the
request, dramatically simplifying activity tracking on the object
themselves, and removing the ambiguity between retiring objects and
retiring requests.
Furthermore with the consolidation of managing the activity tracking
centrally, we can look forward to using RCU to enable lockless lookup of
the current active requests for an object. In the future, we will be
able to query the status or wait upon rendering to an object without
even touching the struct_mutex BKL.
All told, less code, simpler and faster, and more extensible.
v2: Add a typedef for the function pointer for convenience later.
v3: Make the noop retirement callback explicit. Allow passing NULL to
the init_request_active() which is expanded to a common noop function.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/Makefile | 1 -
drivers/gpu/drm/i915/i915_drv.h | 10 ---
drivers/gpu/drm/i915/i915_gem.c | 129 ++++++--------------------------
drivers/gpu/drm/i915/i915_gem_debug.c | 70 -----------------
drivers/gpu/drm/i915/i915_gem_fence.c | 11 +--
drivers/gpu/drm/i915/i915_gem_request.c | 50 +++++++++++--
drivers/gpu/drm/i915/i915_gem_request.h | 97 +++++++++++++++++-------
drivers/gpu/drm/i915/intel_engine_cs.c | 1 -
drivers/gpu/drm/i915/intel_ringbuffer.h | 12 ---
9 files changed, 138 insertions(+), 243 deletions(-)
delete mode 100644 drivers/gpu/drm/i915/i915_gem_debug.c
diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
index 6092f0ea24df..dda724f04445 100644
--- a/drivers/gpu/drm/i915/Makefile
+++ b/drivers/gpu/drm/i915/Makefile
@@ -25,7 +25,6 @@ i915-$(CONFIG_DEBUG_FS) += i915_debugfs.o
i915-y += i915_cmd_parser.o \
i915_gem_batch_pool.o \
i915_gem_context.o \
- i915_gem_debug.o \
i915_gem_dmabuf.o \
i915_gem_evict.o \
i915_gem_execbuffer.o \
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 17a206f19fcb..6b57d8ff7218 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -432,8 +432,6 @@ void intel_link_compute_m_n(int bpp, int nlanes,
#define DRIVER_MINOR 6
#define DRIVER_PATCHLEVEL 0
-#define WATCH_LISTS 0
-
struct opregion_header;
struct opregion_acpi;
struct opregion_swsci;
@@ -2153,7 +2151,6 @@ struct drm_i915_gem_object {
struct drm_mm_node *stolen;
struct list_head global_list;
- struct list_head engine_list[I915_NUM_ENGINES];
/** Used in execbuf to temporarily hold a ref */
struct list_head obj_exec_link;
@@ -3463,13 +3460,6 @@ static inline bool i915_gem_object_needs_bit17_swizzle(struct drm_i915_gem_objec
obj->tiling_mode != I915_TILING_NONE;
}
-/* i915_gem_debug.c */
-#if WATCH_LISTS
-int i915_verify_lists(struct drm_device *dev);
-#else
-#define i915_verify_lists(dev) 0
-#endif
-
/* i915_debugfs.c */
#ifdef CONFIG_DEBUG_FS
int i915_debugfs_register(struct drm_i915_private *dev_priv);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index c9987e26670e..7ba02c48bfb7 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -43,10 +43,6 @@
static void i915_gem_object_flush_gtt_write_domain(struct drm_i915_gem_object *obj);
static void i915_gem_object_flush_cpu_write_domain(struct drm_i915_gem_object *obj);
-static void
-i915_gem_object_retire__write(struct drm_i915_gem_object *obj);
-static void
-i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int engine);
static bool cpu_cache_is_coherent(struct drm_device *dev,
enum i915_cache_level level)
@@ -141,7 +137,6 @@ int i915_mutex_lock_interruptible(struct drm_device *dev)
if (ret)
return ret;
- WARN_ON(i915_verify_lists(dev));
return 0;
}
@@ -1339,23 +1334,6 @@ put_rpm:
return ret;
}
-static void
-i915_gem_object_retire_request(struct drm_i915_gem_object *obj,
- struct drm_i915_gem_request *req)
-{
- int idx = req->engine->id;
-
- if (i915_gem_active_peek(&obj->last_read[idx],
- &obj->base.dev->struct_mutex) == req)
- i915_gem_object_retire__read(obj, idx);
- else if (i915_gem_active_peek(&obj->last_write,
- &obj->base.dev->struct_mutex) == req)
- i915_gem_object_retire__write(obj);
-
- if (!i915_reset_in_progress(&req->i915->gpu_error))
- i915_gem_request_retire_upto(req);
-}
-
/**
* Ensures that all rendering to the object has completed and the object is
* safe to unbind from the GTT or access from the CPU.
@@ -1382,18 +1360,10 @@ i915_gem_object_wait_rendering(struct drm_i915_gem_object *obj,
}
for_each_active(active_mask, idx) {
- struct drm_i915_gem_request *request;
-
- request = i915_gem_active_peek(&active[idx],
- &obj->base.dev->struct_mutex);
- if (!request)
- continue;
-
- ret = i915_wait_request(request);
+ ret = i915_gem_active_wait(&active[idx],
+ &obj->base.dev->struct_mutex);
if (ret)
return ret;
-
- i915_gem_object_retire_request(obj, request);
}
resv = i915_gem_object_get_dmabuf_resv(obj);
@@ -1453,11 +1423,8 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
ret = __i915_wait_request(requests[i], true, NULL, rps);
mutex_lock(&dev->struct_mutex);
- for (i = 0; i < n; i++) {
- if (ret == 0)
- i915_gem_object_retire_request(obj, requests[i]);
+ for (i = 0; i < n; i++)
i915_gem_request_put(requests[i]);
- }
return ret;
}
@@ -2376,40 +2343,31 @@ void i915_vma_move_to_active(struct i915_vma *vma,
i915_gem_object_get(obj);
obj->active |= intel_engine_flag(engine);
- list_move_tail(&obj->engine_list[engine->id], &engine->active_list);
i915_gem_active_set(&obj->last_read[engine->id], req);
list_move_tail(&vma->vm_link, &vma->vm->active_list);
}
static void
-i915_gem_object_retire__write(struct drm_i915_gem_object *obj)
+i915_gem_object_retire__write(struct i915_gem_active *active,
+ struct drm_i915_gem_request *request)
{
- GEM_BUG_ON(!__i915_gem_active_is_busy(&obj->last_write));
- GEM_BUG_ON(!(obj->active &
- intel_engine_flag(i915_gem_active_get_engine(&obj->last_write,
- &obj->base.dev->struct_mutex))));
+ struct drm_i915_gem_object *obj =
+ container_of(active, struct drm_i915_gem_object, last_write);
- i915_gem_active_set(&obj->last_write, NULL);
intel_fb_obj_flush(obj, true, ORIGIN_CS);
}
static void
-i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int idx)
+i915_gem_object_retire__read(struct i915_gem_active *active,
+ struct drm_i915_gem_request *request)
{
- struct intel_engine_cs *engine;
+ int idx = request->engine->id;
+ struct drm_i915_gem_object *obj =
+ container_of(active, struct drm_i915_gem_object, last_read[idx]);
struct i915_vma *vma;
- GEM_BUG_ON(!__i915_gem_active_is_busy(&obj->last_read[idx]));
- GEM_BUG_ON(!(obj->active & (1 << idx)));
-
- list_del_init(&obj->engine_list[idx]);
- i915_gem_active_set(&obj->last_read[idx], NULL);
-
- engine = i915_gem_active_get_engine(&obj->last_write,
- &obj->base.dev->struct_mutex);
- if (engine && engine->id == idx)
- i915_gem_object_retire__write(obj);
+ GEM_BUG_ON((obj->active & (1 << idx)) == 0);
obj->active &= ~(1 << idx);
if (obj->active)
@@ -2419,15 +2377,13 @@ i915_gem_object_retire__read(struct drm_i915_gem_object *obj, int idx)
* so that we don't steal from recently used but inactive objects
* (unless we are forced to ofc!)
*/
- list_move_tail(&obj->global_list,
- &to_i915(obj->base.dev)->mm.bound_list);
+ list_move_tail(&obj->global_list, &request->i915->mm.bound_list);
list_for_each_entry(vma, &obj->vma_list, obj_link) {
if (!list_empty(&vma->vm_link))
list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
}
- i915_gem_active_set(&obj->last_fence, NULL);
i915_gem_object_put(obj);
}
@@ -2505,16 +2461,6 @@ static void i915_gem_reset_engine_cleanup(struct intel_engine_cs *engine)
{
struct intel_ring *ring;
- while (!list_empty(&engine->active_list)) {
- struct drm_i915_gem_object *obj;
-
- obj = list_first_entry(&engine->active_list,
- struct drm_i915_gem_object,
- engine_list[engine->id]);
-
- i915_gem_object_retire__read(obj, engine->id);
- }
-
/* Mark all pending requests as complete so that any concurrent
* (lockless) lookup doesn't try and wait upon the request as we
* reset it.
@@ -2586,8 +2532,6 @@ void i915_gem_reset(struct drm_device *dev)
i915_gem_context_reset(dev);
i915_gem_restore_fences(dev);
-
- WARN_ON(i915_verify_lists(dev));
}
/**
@@ -2597,13 +2541,6 @@ void i915_gem_reset(struct drm_device *dev)
void
i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
{
- WARN_ON(i915_verify_lists(engine->dev));
-
- /* Retire requests first as we use it above for the early return.
- * If we retire requests last, we may use a later seqno and so clear
- * the requests lists without clearing the active list, leading to
- * confusion.
- */
while (!list_empty(&engine->request_list)) {
struct drm_i915_gem_request *request;
@@ -2616,26 +2553,6 @@ i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
i915_gem_request_retire_upto(request);
}
-
- /* Move any buffers on the active list that are no longer referenced
- * by the ringbuffer to the flushing/inactive lists as appropriate,
- * before we free the context associated with the requests.
- */
- while (!list_empty(&engine->active_list)) {
- struct drm_i915_gem_object *obj;
-
- obj = list_first_entry(&engine->active_list,
- struct drm_i915_gem_object,
- engine_list[engine->id]);
-
- if (!list_empty(&i915_gem_active_peek(&obj->last_read[engine->id],
- &obj->base.dev->struct_mutex)->link))
- break;
-
- i915_gem_object_retire__read(obj, engine->id);
- }
-
- WARN_ON(i915_verify_lists(engine->dev));
}
void i915_gem_retire_requests(struct drm_i915_private *dev_priv)
@@ -2818,8 +2735,7 @@ out:
}
static int
-__i915_gem_object_sync(struct drm_i915_gem_object *obj,
- struct drm_i915_gem_request *to,
+__i915_gem_object_sync(struct drm_i915_gem_request *to,
struct drm_i915_gem_request *from)
{
int ret;
@@ -2827,9 +2743,6 @@ __i915_gem_object_sync(struct drm_i915_gem_object *obj,
if (to->engine == from->engine)
return 0;
- if (i915_gem_request_completed(from))
- return 0;
-
if (!i915.semaphores) {
ret = __i915_wait_request(from,
from->i915->mm.interruptible,
@@ -2837,8 +2750,6 @@ __i915_gem_object_sync(struct drm_i915_gem_object *obj,
NO_WAITBOOST);
if (ret)
return ret;
-
- i915_gem_object_retire_request(obj, from);
} else {
int idx = intel_engine_sync_index(from->engine, to->engine);
if (from->fence.seqno <= from->engine->semaphore.sync_seqno[idx])
@@ -2905,7 +2816,7 @@ i915_gem_object_sync(struct drm_i915_gem_object *obj,
if (!request)
continue;
- ret = __i915_gem_object_sync(obj, to, request);
+ ret = __i915_gem_object_sync(to, request);
if (ret)
return ret;
}
@@ -3041,7 +2952,6 @@ int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv)
return ret;
}
- WARN_ON(i915_verify_lists(dev));
return 0;
}
@@ -4081,7 +3991,11 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj,
INIT_LIST_HEAD(&obj->global_list);
for (i = 0; i < I915_NUM_ENGINES; i++)
- INIT_LIST_HEAD(&obj->engine_list[i]);
+ init_request_active(&obj->last_read[i],
+ i915_gem_object_retire__read);
+ init_request_active(&obj->last_write,
+ i915_gem_object_retire__write);
+ init_request_active(&obj->last_fence, NULL);
INIT_LIST_HEAD(&obj->obj_exec_link);
INIT_LIST_HEAD(&obj->vma_list);
INIT_LIST_HEAD(&obj->batch_pool_link);
@@ -4574,7 +4488,6 @@ i915_gem_cleanup_engines(struct drm_device *dev)
static void
init_engine_lists(struct intel_engine_cs *engine)
{
- INIT_LIST_HEAD(&engine->active_list);
INIT_LIST_HEAD(&engine->request_list);
}
diff --git a/drivers/gpu/drm/i915/i915_gem_debug.c b/drivers/gpu/drm/i915/i915_gem_debug.c
deleted file mode 100644
index a56516482394..000000000000
--- a/drivers/gpu/drm/i915/i915_gem_debug.c
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
- * Copyright © 2008 Intel Corporation
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice (including the next
- * paragraph) shall be included in all copies or substantial portions of the
- * Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
- *
- * Authors:
- * Keith Packard <keithp@keithp.com>
- *
- */
-
-#include <drm/drmP.h>
-#include <drm/i915_drm.h>
-#include "i915_drv.h"
-
-#if WATCH_LISTS
-int
-i915_verify_lists(struct drm_device *dev)
-{
- static int warned;
- struct drm_i915_private *dev_priv = to_i915(dev);
- struct drm_i915_gem_object *obj;
- struct intel_engine_cs *engine;
- int err = 0;
-
- if (warned)
- return 0;
-
- for_each_engine(engine, dev_priv) {
- list_for_each_entry(obj, &engine->active_list,
- engine_list[engine->id]) {
- if (obj->base.dev != dev ||
- !atomic_read(&obj->base.refcount.refcount)) {
- DRM_ERROR("%s: freed active obj %p\n",
- engine->name, obj);
- err++;
- break;
- } else if (!obj->active ||
- obj->last_read_req[engine->id] == NULL) {
- DRM_ERROR("%s: invalid active obj %p\n",
- engine->name, obj);
- err++;
- } else if (obj->base.write_domain) {
- DRM_ERROR("%s: invalid write obj %p (w %x)\n",
- engine->name,
- obj, obj->base.write_domain);
- err++;
- }
- }
- }
-
- return warned = err;
-}
-#endif /* WATCH_LIST */
diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
index a4ec4fecbea0..dbaab9ce29c9 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence.c
@@ -261,15 +261,8 @@ static inline void i915_gem_object_fence_lost(struct drm_i915_gem_object *obj)
static int
i915_gem_object_wait_fence(struct drm_i915_gem_object *obj)
{
- int ret;
-
- ret = i915_gem_active_wait(&obj->last_fence,
- &obj->base.dev->struct_mutex);
- if (ret)
- return ret;
-
- i915_gem_active_set(&obj->last_fence, NULL);
- return 0;
+ return i915_gem_active_retire(&obj->last_fence,
+ &obj->base.dev->struct_mutex);
}
/**
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 7802156d481a..cdaaeb6dd913 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -22,6 +22,8 @@
*
*/
+#include <linux/prefetch.h>
+
#include "i915_drv.h"
static const char *i915_fence_get_driver_name(struct fence *fence)
@@ -157,8 +159,16 @@ i915_gem_request_remove_from_client(struct drm_i915_gem_request *request)
request->pid = NULL;
}
+void i915_gem_retire_noop(struct i915_gem_active *active,
+ struct drm_i915_gem_request *request)
+{
+ /* Space left intentionally blank */
+}
+
static void i915_gem_request_retire(struct drm_i915_gem_request *request)
{
+ struct i915_gem_active *active, *next;
+
trace_i915_gem_request_retire(request);
list_del_init(&request->link);
@@ -172,6 +182,33 @@ static void i915_gem_request_retire(struct drm_i915_gem_request *request)
*/
request->ring->last_retired_head = request->postfix;
+ /* Walk through the active list, calling retire on each. This allows
+ * objects to track their GPU activity and mark themselves as idle
+ * when their *last* active request is completed (updating state
+ * tracking lists for eviction, active references for GEM, etc).
+ *
+ * As the ->retire() may free the node, we decouple it first and
+ * pass along the auxiliary information (to avoid dereferencing
+ * the node after the callback).
+ */
+ list_for_each_entry_safe(active, next, &request->active_list, link) {
+ /* In microbenchmarks or focusing upon time inside the kernel,
+ * we may spend an inordinate amount of time simply handling
+ * the retirement of requests and processing their callbacks.
+ * Of which, this loop itself is particularly hot due to the
+ * cache misses when jumping around the list of i915_gem_active.
+ * So we try to keep this loop as streamlined as possible and
+ * also prefetch the next i915_gem_active to try and hide
+ * the likely cache miss.
+ */
+ prefetchw(next);
+
+ INIT_LIST_HEAD(&active->link);
+ active->request = NULL;
+
+ active->retire(active, request);
+ }
+
i915_gem_request_remove_from_client(request);
if (request->previous_context) {
@@ -200,8 +237,6 @@ void i915_gem_request_retire_upto(struct drm_i915_gem_request *req)
i915_gem_request_retire(tmp);
} while (tmp != req);
-
- WARN_ON(i915_verify_lists(engine->dev));
}
static int i915_gem_check_wedge(unsigned int reset_counter, bool interruptible)
@@ -336,6 +371,7 @@ i915_gem_request_alloc(struct intel_engine_cs *engine,
engine->fence_context,
seqno);
+ INIT_LIST_HEAD(&req->active_list);
req->i915 = dev_priv;
req->engine = engine;
req->ctx = i915_gem_context_get(ctx);
@@ -570,9 +606,6 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
might_sleep();
- if (list_empty(&req->link))
- return 0;
-
if (i915_gem_request_completed(req))
return 0;
@@ -705,10 +738,13 @@ int i915_wait_request(struct drm_i915_gem_request *req)
{
int ret;
- GEM_BUG_ON(!req);
lockdep_assert_held(&req->i915->drm.struct_mutex);
+ GEM_BUG_ON(list_empty(&req->link));
- ret = __i915_wait_request(req, req->i915->mm.interruptible, NULL, NULL);
+ ret = __i915_wait_request(req,
+ req->i915->mm.interruptible,
+ NULL,
+ NULL);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index cb3fa58d4eb5..70eeae6b0e1a 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -101,6 +101,7 @@ struct drm_i915_gem_request {
* error state dump only).
*/
struct drm_i915_gem_object *batch_obj;
+ struct list_head active_list;
/** Time at which this request was emitted, in jiffies. */
unsigned long emitted_jiffies;
@@ -213,8 +214,12 @@ struct intel_rps_client;
int __i915_wait_request(struct drm_i915_gem_request *req,
bool interruptible,
s64 *timeout,
- struct intel_rps_client *rps);
-int __must_check i915_wait_request(struct drm_i915_gem_request *req);
+ struct intel_rps_client *rps)
+ __attribute__((nonnull(1)));
+
+int __must_check
+i915_wait_request(struct drm_i915_gem_request *req)
+ __attribute__((nonnull));
static inline u32 intel_engine_get_seqno(struct intel_engine_cs *engine);
@@ -276,10 +281,39 @@ static inline bool i915_spin_request(const struct drm_i915_gem_request *request,
* can then perform any action, such as delayed freeing of an active
* resource including itself.
*/
+struct i915_gem_active;
+
+typedef void (*i915_gem_retire_fn)(struct i915_gem_active *,
+ struct drm_i915_gem_request *);
+
struct i915_gem_active {
struct drm_i915_gem_request *request;
+ struct list_head link;
+ i915_gem_retire_fn retire;
};
+void i915_gem_retire_noop(struct i915_gem_active *,
+ struct drm_i915_gem_request *request);
+
+/**
+ * init_request_active - prepares the activity tracker for use
+ * @active - the active tracker
+ * @func - a callback when then the tracker is retired (becomes idle),
+ * can be NULL
+ *
+ * init_request_active() prepares the embedded @active struct for use as
+ * an activity tracker, that is for tracking the last known active request
+ * associated with it. When the last request becomes idle, when it is retired
+ * after completion, the optional callback @func is invoked.
+ */
+static inline void
+init_request_active(struct i915_gem_active *active,
+ i915_gem_retire_fn retire)
+{
+ INIT_LIST_HEAD(&active->link);
+ active->retire = retire ?: i915_gem_retire_noop;
+}
+
/**
* i915_gem_active_set - updates the tracker to watch the current request
* @active - the active tracker
@@ -293,7 +327,8 @@ static inline void
i915_gem_active_set(struct i915_gem_active *active,
struct drm_i915_gem_request *request)
{
- i915_gem_request_assign(&active->request, request);
+ list_move(&active->link, &request->active_list);
+ active->request = request;
}
static inline struct drm_i915_gem_request *
@@ -303,17 +338,23 @@ __i915_gem_active_peek(const struct i915_gem_active *active)
}
/**
- * i915_gem_active_peek - report the request being monitored
+ * i915_gem_active_peek - report the active request being monitored
* @active - the active tracker
*
- * i915_gem_active_peek() returns the current request being tracked, or NULL.
- * It does not obtain a reference on the request for the caller, so the
- * caller must hold struct_mutex.
+ * i915_gem_active_peek() returns the current request being tracked if
+ * still active, or NULL. It does not obtain a reference on the request
+ * for the caller, so the caller must hold struct_mutex.
*/
static inline struct drm_i915_gem_request *
i915_gem_active_peek(const struct i915_gem_active *active, struct mutex *mutex)
{
- return active->request;
+ struct drm_i915_gem_request *request;
+
+ request = active->request;
+ if (!request || i915_gem_request_completed(request))
+ return NULL;
+
+ return request;
}
/**
@@ -326,13 +367,7 @@ i915_gem_active_peek(const struct i915_gem_active *active, struct mutex *mutex)
static inline struct drm_i915_gem_request *
i915_gem_active_get(const struct i915_gem_active *active, struct mutex *mutex)
{
- struct drm_i915_gem_request *request;
-
- request = i915_gem_active_peek(active, mutex);
- if (!request || i915_gem_request_completed(request))
- return NULL;
-
- return i915_gem_request_get(request);
+ return i915_gem_request_get(i915_gem_active_peek(active, mutex));
}
/**
@@ -361,13 +396,7 @@ static inline bool
i915_gem_active_is_idle(const struct i915_gem_active *active,
struct mutex *mutex)
{
- struct drm_i915_gem_request *request;
-
- request = i915_gem_active_peek(active, mutex);
- if (!request || i915_gem_request_completed(request))
- return true;
-
- return false;
+ return !i915_gem_active_peek(active, mutex);
}
/**
@@ -377,6 +406,9 @@ i915_gem_active_is_idle(const struct i915_gem_active *active,
* i915_gem_active_wait() waits until the request is completed before
* returning. Note that it does not guarantee that the request is
* retired first, see i915_gem_active_retire().
+ *
+ * i915_gem_active_wait() returns immediately if the active
+ * request is already complete.
*/
static inline int __must_check
i915_gem_active_wait(const struct i915_gem_active *active, struct mutex *mutex)
@@ -387,7 +419,7 @@ i915_gem_active_wait(const struct i915_gem_active *active, struct mutex *mutex)
if (!request)
return 0;
- return i915_wait_request(request);
+ return __i915_wait_request(request, true, NULL, NULL);
}
/**
@@ -400,10 +432,25 @@ i915_gem_active_wait(const struct i915_gem_active *active, struct mutex *mutex)
* tracker is idle, the function returns immediately.
*/
static inline int __must_check
-i915_gem_active_retire(const struct i915_gem_active *active,
+i915_gem_active_retire(struct i915_gem_active *active,
struct mutex *mutex)
{
- return i915_gem_active_wait(active, mutex);
+ struct drm_i915_gem_request *request;
+ int ret;
+
+ request = active->request;
+ if (!request)
+ return 0;
+
+ ret = __i915_wait_request(request, true, NULL, NULL);
+ if (ret)
+ return ret;
+
+ list_del_init(&active->link);
+ active->request = NULL;
+ active->retire(active, request);
+
+ return 0;
}
/* Convenience functions for peeking at state inside active's request whilst
diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
index 4ec914ed9104..202ad83b3dd1 100644
--- a/drivers/gpu/drm/i915/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/intel_engine_cs.c
@@ -177,7 +177,6 @@ void intel_engine_init_hangcheck(struct intel_engine_cs *engine)
*/
void intel_engine_setup_common(struct intel_engine_cs *engine)
{
- INIT_LIST_HEAD(&engine->active_list);
INIT_LIST_HEAD(&engine->request_list);
INIT_LIST_HEAD(&engine->buffers);
INIT_LIST_HEAD(&engine->execlist_queue);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 170624171887..236e7a2f5689 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -293,18 +293,6 @@ struct intel_engine_cs {
u32 ctx_desc_template;
/**
- * List of objects currently involved in rendering from the
- * ringbuffer.
- *
- * Includes buffers having the contents of their GPU caches
- * flushed, not necessarily primitives. last_read_req
- * represents when the rendering involved will be completed.
- *
- * A reference is held on the buffer while on this list.
- */
- struct list_head active_list;
-
- /**
* List of breadcrumbs associated with GPU requests currently
* outstanding.
*/
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 17/28] drm/i915: Track requests inside each intel_ring
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (15 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 16/28] drm/i915: Refactor activity tracking for requests Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 18/28] drm/i915: Convert intel_overlay to request tracking Chris Wilson
` (12 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
By tracking each request occupying space inside an individual
intel_ring, we can greatly simplify the logic of tracking available
space and not worry about other timelines. (Each ring is an ordered
timeline of committed requests.)
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem_request.c | 2 ++
drivers/gpu/drm/i915/i915_gem_request.h | 3 +++
drivers/gpu/drm/i915/intel_ringbuffer.c | 15 ++++-----------
drivers/gpu/drm/i915/intel_ringbuffer.h | 2 ++
4 files changed, 11 insertions(+), 11 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index cdaaeb6dd913..a91e79fc5553 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -180,6 +180,7 @@ static void i915_gem_request_retire(struct drm_i915_gem_request *request)
* Note this requires that we are always called in request
* completion order.
*/
+ list_del(&request->ring_link);
request->ring->last_retired_head = request->postfix;
/* Walk through the active list, calling retire on each. This allows
@@ -487,6 +488,7 @@ void __i915_add_request(struct drm_i915_gem_request *request,
request->previous_seqno = engine->last_submitted_seqno;
smp_store_mb(engine->last_submitted_seqno, request->fence.seqno);
list_add_tail(&request->link, &engine->request_list);
+ list_add_tail(&request->ring_link, &ring->request_list);
/* Record the position of the start of the request so that
* should we detect the updated seqno part-way through the
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 70eeae6b0e1a..98dedf9152fa 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -109,6 +109,9 @@ struct drm_i915_gem_request {
/** engine->request_list entry for this request */
struct list_head link;
+ /** ring->request_list entry for this request */
+ struct list_head ring_link;
+
struct drm_i915_file_private *file_priv;
/** file_priv list entry for this request */
struct list_head client_list;
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 76ddc7cc6d25..5b0eac2af28c 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2046,6 +2046,8 @@ intel_engine_create_ring(struct intel_engine_cs *engine, int size)
ring->engine = engine;
list_add(&ring->link, &engine->buffers);
+ INIT_LIST_HEAD(&ring->request_list);
+
ring->size = size;
/* Workaround an erratum on the i830 which causes a hang if
* the TAIL pointer points to within the last 2 cachelines
@@ -2266,7 +2268,6 @@ int intel_ring_alloc_request_extras(struct drm_i915_gem_request *request)
static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
{
struct intel_ring *ring = req->ring;
- struct intel_engine_cs *engine = req->engine;
struct drm_i915_gem_request *target;
intel_ring_update_space(ring);
@@ -2284,17 +2285,9 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
*/
GEM_BUG_ON(!req->reserved_space);
- list_for_each_entry(target, &engine->request_list, link) {
+ list_for_each_entry(target, &ring->request_list, ring_link) {
unsigned space;
- /*
- * The request queue is per-engine, so can contain requests
- * from multiple ringbuffers. Here, we must ignore any that
- * aren't from the ringbuffer we're considering.
- */
- if (target->ring != ring)
- continue;
-
/* Would completion of this request free enough space? */
space = __intel_ring_space(target->postfix, ring->tail,
ring->size);
@@ -2302,7 +2295,7 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
break;
}
- if (WARN_ON(&target->link == &engine->request_list))
+ if (WARN_ON(&target->ring_link == &ring->request_list))
return -ENOSPC;
return i915_wait_request(target);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 236e7a2f5689..88952bf10b9d 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -90,6 +90,8 @@ struct intel_ring {
struct intel_engine_cs *engine;
struct list_head link;
+ struct list_head request_list;
+
u32 head;
u32 tail;
int space;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 18/28] drm/i915: Convert intel_overlay to request tracking
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (16 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 17/28] drm/i915: Track requests inside each intel_ring Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 19/28] drm/i915: Move the special case wait-request handling to its one caller Chris Wilson
` (11 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
intel_overlay already tracks its last flip request, along with action to
take after its completion. Refactor intel_overlay to reuse the common
i915_gem_active tracker.
v2: Now using i915_gem_retire_fn typedef
References: https://bugs.freedesktop.org/show_bug.cgi?id=93730
References: https://bugs.freedesktop.org/show_bug.cgi?id=96851
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/intel_overlay.c | 82 +++++++++++++++---------------------
1 file changed, 34 insertions(+), 48 deletions(-)
diff --git a/drivers/gpu/drm/i915/intel_overlay.c b/drivers/gpu/drm/i915/intel_overlay.c
index 651efe4e468e..2c598d63c794 100644
--- a/drivers/gpu/drm/i915/intel_overlay.c
+++ b/drivers/gpu/drm/i915/intel_overlay.c
@@ -183,8 +183,7 @@ struct intel_overlay {
u32 flip_addr;
struct drm_i915_gem_object *reg_bo;
/* flip handling */
- struct drm_i915_gem_request *last_flip_req;
- void (*flip_tail)(struct intel_overlay *);
+ struct i915_gem_active last_flip;
};
static struct overlay_registers __iomem *
@@ -210,23 +209,24 @@ static void intel_overlay_unmap_regs(struct intel_overlay *overlay,
io_mapping_unmap(regs);
}
-static int intel_overlay_do_wait_request(struct intel_overlay *overlay,
+static void intel_overlay_submit_request(struct intel_overlay *overlay,
struct drm_i915_gem_request *req,
- void (*tail)(struct intel_overlay *))
+ i915_gem_retire_fn retire)
{
- int ret;
-
- WARN_ON(overlay->last_flip_req);
- i915_gem_request_assign(&overlay->last_flip_req, req);
+ GEM_BUG_ON(i915_gem_active_peek(&overlay->last_flip,
+ &overlay->i915->drm.struct_mutex));
+ overlay->last_flip.retire = retire;
+ i915_gem_active_set(&overlay->last_flip, req);
i915_add_request(req);
+}
- overlay->flip_tail = tail;
- ret = i915_wait_request(overlay->last_flip_req);
- if (ret)
- return ret;
-
- i915_gem_request_assign(&overlay->last_flip_req, NULL);
- return 0;
+static int intel_overlay_do_wait_request(struct intel_overlay *overlay,
+ struct drm_i915_gem_request *req,
+ i915_gem_retire_fn retire)
+{
+ intel_overlay_submit_request(overlay, req, retire);
+ return i915_gem_active_retire(&overlay->last_flip,
+ &overlay->i915->drm.struct_mutex);
}
static struct drm_i915_gem_request *alloc_request(struct intel_overlay *overlay)
@@ -306,25 +306,32 @@ static int intel_overlay_continue(struct intel_overlay *overlay,
intel_ring_emit(ring, flip_addr);
intel_ring_advance(ring);
- WARN_ON(overlay->last_flip_req);
- i915_gem_request_assign(&overlay->last_flip_req, req);
- i915_add_request(req);
+ intel_overlay_submit_request(overlay, req, NULL);
return 0;
}
-static void intel_overlay_release_old_vid_tail(struct intel_overlay *overlay)
+static void intel_overlay_release_old_vid_tail(struct i915_gem_active *active,
+ struct drm_i915_gem_request *req)
{
+ struct intel_overlay *overlay =
+ container_of(active, typeof(*overlay), last_flip);
struct drm_i915_gem_object *obj = overlay->old_vid_bo;
+ i915_gem_track_fb(obj, NULL,
+ INTEL_FRONTBUFFER_OVERLAY(overlay->crtc->pipe));
+
i915_gem_object_ggtt_unpin(obj);
i915_gem_object_put(obj);
overlay->old_vid_bo = NULL;
}
-static void intel_overlay_off_tail(struct intel_overlay *overlay)
+static void intel_overlay_off_tail(struct i915_gem_active *active,
+ struct drm_i915_gem_request *req)
{
+ struct intel_overlay *overlay =
+ container_of(active, typeof(*overlay), last_flip);
struct drm_i915_gem_object *obj = overlay->vid_bo;
/* never have the overlay hw on without showing a frame */
@@ -387,27 +394,16 @@ static int intel_overlay_off(struct intel_overlay *overlay)
}
intel_ring_advance(ring);
- return intel_overlay_do_wait_request(overlay, req, intel_overlay_off_tail);
+ return intel_overlay_do_wait_request(overlay, req,
+ intel_overlay_off_tail);
}
/* recover from an interruption due to a signal
* We have to be careful not to repeat work forever an make forward progess. */
static int intel_overlay_recover_from_interrupt(struct intel_overlay *overlay)
{
- int ret;
-
- if (overlay->last_flip_req == NULL)
- return 0;
-
- ret = i915_wait_request(overlay->last_flip_req);
- if (ret)
- return ret;
-
- if (overlay->flip_tail)
- overlay->flip_tail(overlay);
-
- i915_gem_request_assign(&overlay->last_flip_req, NULL);
- return 0;
+ return i915_gem_active_retire(&overlay->last_flip,
+ &overlay->i915->drm.struct_mutex);
}
/* Wait for pending overlay flip and release old frame.
@@ -452,13 +448,9 @@ static int intel_overlay_release_old_vid(struct intel_overlay *overlay)
intel_overlay_release_old_vid_tail);
if (ret)
return ret;
- }
+ } else
+ intel_overlay_release_old_vid_tail(&overlay->last_flip, NULL);
- intel_overlay_release_old_vid_tail(overlay);
-
-
- i915_gem_track_fb(overlay->old_vid_bo, NULL,
- INTEL_FRONTBUFFER_OVERLAY(overlay->crtc->pipe));
return 0;
}
@@ -471,7 +463,6 @@ void intel_overlay_reset(struct drm_i915_private *dev_priv)
intel_overlay_release_old_vid(overlay);
- overlay->last_flip_req = NULL;
overlay->old_xscale = 0;
overlay->old_yscale = 0;
overlay->crtc = NULL;
@@ -882,12 +873,7 @@ int intel_overlay_switch_off(struct intel_overlay *overlay)
iowrite32(0, ®s->OCMD);
intel_overlay_unmap_regs(overlay, regs);
- ret = intel_overlay_off(overlay);
- if (ret != 0)
- return ret;
-
- intel_overlay_off_tail(overlay);
- return 0;
+ return intel_overlay_off(overlay);
}
static int check_overlay_possible_on_crtc(struct intel_overlay *overlay,
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 19/28] drm/i915: Move the special case wait-request handling to its one caller
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (17 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 18/28] drm/i915: Convert intel_overlay to request tracking Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 20/28] drm/i915: Disable waitboosting for a saturated engine Chris Wilson
` (10 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem_request.c | 25 -------------------------
drivers/gpu/drm/i915/i915_gem_request.h | 4 ----
drivers/gpu/drm/i915/intel_ringbuffer.c | 18 +++++++++++++-----
3 files changed, 13 insertions(+), 34 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index a91e79fc5553..85ec5ca5c36b 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -731,28 +731,3 @@ complete:
return ret;
}
-
-/**
- * Waits for a request to be signaled, and cleans up the
- * request and object lists appropriately for that event.
- */
-int i915_wait_request(struct drm_i915_gem_request *req)
-{
- int ret;
-
- lockdep_assert_held(&req->i915->drm.struct_mutex);
- GEM_BUG_ON(list_empty(&req->link));
-
- ret = __i915_wait_request(req,
- req->i915->mm.interruptible,
- NULL,
- NULL);
- if (ret)
- return ret;
-
- /* If the GPU hung, we want to keep the requests to find the guilty. */
- if (!i915_reset_in_progress(&req->i915->gpu_error))
- i915_gem_request_retire_upto(req);
-
- return 0;
-}
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 98dedf9152fa..2d8d36976c41 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -220,10 +220,6 @@ int __i915_wait_request(struct drm_i915_gem_request *req,
struct intel_rps_client *rps)
__attribute__((nonnull(1)));
-int __must_check
-i915_wait_request(struct drm_i915_gem_request *req)
- __attribute__((nonnull));
-
static inline u32 intel_engine_get_seqno(struct intel_engine_cs *engine);
/**
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 5b0eac2af28c..542cf585121c 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2269,6 +2269,7 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
{
struct intel_ring *ring = req->ring;
struct drm_i915_gem_request *target;
+ int ret;
intel_ring_update_space(ring);
if (ring->space >= bytes)
@@ -2298,7 +2299,18 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
if (WARN_ON(&target->ring_link == &ring->request_list))
return -ENOSPC;
- return i915_wait_request(target);
+ ret = __i915_wait_request(target, true, NULL, NULL);
+ if (ret)
+ return ret;
+
+ if (i915_reset_in_progress(&target->i915->gpu_error))
+ return -EAGAIN;
+
+ i915_gem_request_retire_upto(target);
+
+ intel_ring_update_space(ring);
+ GEM_BUG_ON(ring->space < bytes);
+ return 0;
}
int intel_ring_begin(struct drm_i915_gem_request *req, int num_dwords)
@@ -2336,10 +2348,6 @@ int intel_ring_begin(struct drm_i915_gem_request *req, int num_dwords)
int ret = wait_for_space(req, wait_bytes);
if (unlikely(ret))
return ret;
-
- intel_ring_update_space(ring);
- if (unlikely(ring->space < wait_bytes))
- return -EAGAIN;
}
if (unlikely(need_wrap)) {
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 20/28] drm/i915: Disable waitboosting for a saturated engine
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (18 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 19/28] drm/i915: Move the special case wait-request handling to its one caller Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 21/28] drm/i915: s/__i915_wait_request/i915_wait_request/ Chris Wilson
` (9 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
If the user floods the GPU with so many requests that the engine stalls
waiting for free space, don't automatically promote the GPU to maximum
frequencies. If the GPU really is saturated with work, it will migrate
to high clocks by itself, otherwise it is merely a user flooding us with
busy-work.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/intel_ringbuffer.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 542cf585121c..4ab6d2365e30 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2299,7 +2299,7 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
if (WARN_ON(&target->ring_link == &ring->request_list))
return -ENOSPC;
- ret = __i915_wait_request(target, true, NULL, NULL);
+ ret = __i915_wait_request(target, true, NULL, NO_WAITBOOST);
if (ret)
return ret;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 21/28] drm/i915: s/__i915_wait_request/i915_wait_request/
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (19 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 20/28] drm/i915: Disable waitboosting for a saturated engine Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 22/28] drm/i915: Double check activity before relocations Chris Wilson
` (8 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
There is only one wait on request function now, so drop the "expert"
indication of leading __.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 18 +++++++++---------
drivers/gpu/drm/i915/i915_gem_request.c | 16 ++++++++--------
drivers/gpu/drm/i915/i915_gem_request.h | 12 ++++++------
drivers/gpu/drm/i915/i915_gem_userptr.c | 2 +-
drivers/gpu/drm/i915/intel_display.c | 14 +++++++-------
drivers/gpu/drm/i915/intel_ringbuffer.c | 8 ++++----
6 files changed, 35 insertions(+), 35 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 7ba02c48bfb7..46e41c42f8b9 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1420,7 +1420,7 @@ i915_gem_object_wait_rendering__nonblocking(struct drm_i915_gem_object *obj,
mutex_unlock(&dev->struct_mutex);
ret = 0;
for (i = 0; ret == 0 && i < n; i++)
- ret = __i915_wait_request(requests[i], true, NULL, rps);
+ ret = i915_wait_request(requests[i], true, NULL, rps);
mutex_lock(&dev->struct_mutex);
for (i = 0; i < n; i++)
@@ -2726,9 +2726,9 @@ out:
for (i = 0; i < n; i++) {
if (ret == 0)
- ret = __i915_wait_request(requests[i], true,
- args->timeout_ns > 0 ? &args->timeout_ns : NULL,
- to_rps_client(file));
+ ret = i915_wait_request(requests[i], true,
+ args->timeout_ns > 0 ? &args->timeout_ns : NULL,
+ to_rps_client(file));
i915_gem_request_put(requests[i]);
}
return ret;
@@ -2744,10 +2744,10 @@ __i915_gem_object_sync(struct drm_i915_gem_request *to,
return 0;
if (!i915.semaphores) {
- ret = __i915_wait_request(from,
- from->i915->mm.interruptible,
- NULL,
- NO_WAITBOOST);
+ ret = i915_wait_request(from,
+ from->i915->mm.interruptible,
+ NULL,
+ NO_WAITBOOST);
if (ret)
return ret;
} else {
@@ -3712,7 +3712,7 @@ i915_gem_ring_throttle(struct drm_device *dev, struct drm_file *file)
if (target == NULL)
return 0;
- ret = __i915_wait_request(target, true, NULL, NULL);
+ ret = i915_wait_request(target, true, NULL, NULL);
i915_gem_request_put(target);
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 85ec5ca5c36b..8549375aa75e 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -70,9 +70,9 @@ static signed long i915_fence_wait(struct fence *fence,
timeout = NULL;
}
- ret = __i915_wait_request(to_request(fence),
- interruptible, timeout,
- NO_WAITBOOST);
+ ret = i915_wait_request(to_request(fence),
+ interruptible, timeout,
+ NO_WAITBOOST);
if (ret == -ETIME)
return 0;
@@ -579,7 +579,7 @@ bool __i915_spin_request(const struct drm_i915_gem_request *req,
}
/**
- * __i915_wait_request - wait until execution of request has finished
+ * i915_wait_request - wait until execution of request has finished
* @req: duh!
* @interruptible: do an interruptible wait (normally yes)
* @timeout: in - how long to wait (NULL forever); out - how much time remaining
@@ -595,10 +595,10 @@ bool __i915_spin_request(const struct drm_i915_gem_request *req,
* Returns 0 if the request was found within the alloted time. Else returns the
* errno with remaining time filled in timeout argument.
*/
-int __i915_wait_request(struct drm_i915_gem_request *req,
- bool interruptible,
- s64 *timeout,
- struct intel_rps_client *rps)
+int i915_wait_request(struct drm_i915_gem_request *req,
+ bool interruptible,
+ s64 *timeout,
+ struct intel_rps_client *rps)
{
int state = interruptible ? TASK_INTERRUPTIBLE : TASK_UNINTERRUPTIBLE;
DEFINE_WAIT(reset);
diff --git a/drivers/gpu/drm/i915/i915_gem_request.h b/drivers/gpu/drm/i915/i915_gem_request.h
index 2d8d36976c41..4241d49a024c 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.h
+++ b/drivers/gpu/drm/i915/i915_gem_request.h
@@ -214,10 +214,10 @@ struct intel_rps_client;
#define IS_RPS_CLIENT(p) (!IS_ERR(p))
#define IS_RPS_USER(p) (!IS_ERR_OR_NULL(p))
-int __i915_wait_request(struct drm_i915_gem_request *req,
- bool interruptible,
- s64 *timeout,
- struct intel_rps_client *rps)
+int i915_wait_request(struct drm_i915_gem_request *req,
+ bool interruptible,
+ s64 *timeout,
+ struct intel_rps_client *rps)
__attribute__((nonnull(1)));
static inline u32 intel_engine_get_seqno(struct intel_engine_cs *engine);
@@ -418,7 +418,7 @@ i915_gem_active_wait(const struct i915_gem_active *active, struct mutex *mutex)
if (!request)
return 0;
- return __i915_wait_request(request, true, NULL, NULL);
+ return i915_wait_request(request, true, NULL, NULL);
}
/**
@@ -441,7 +441,7 @@ i915_gem_active_retire(struct i915_gem_active *active,
if (!request)
return 0;
- ret = __i915_wait_request(request, true, NULL, NULL);
+ ret = i915_wait_request(request, true, NULL, NULL);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c
index e57521dbddc6..651a84ba840c 100644
--- a/drivers/gpu/drm/i915/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/i915_gem_userptr.c
@@ -83,7 +83,7 @@ static void wait_rendering(struct drm_i915_gem_object *obj)
mutex_unlock(&dev->struct_mutex);
for (i = 0; i < n; i++)
- __i915_wait_request(requests[i], false, NULL, NULL);
+ i915_wait_request(requests[i], false, NULL, NULL);
mutex_lock(&dev->struct_mutex);
diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
index d54a3ea56536..da9dcacd49d5 100644
--- a/drivers/gpu/drm/i915/intel_display.c
+++ b/drivers/gpu/drm/i915/intel_display.c
@@ -11463,9 +11463,9 @@ static void intel_mmio_flip_work_func(struct work_struct *w)
struct reservation_object *resv;
if (work->flip_queued_req)
- WARN_ON(__i915_wait_request(work->flip_queued_req,
- false, NULL,
- NO_WAITBOOST));
+ WARN_ON(i915_wait_request(work->flip_queued_req,
+ false, NULL,
+ NO_WAITBOOST));
/* For framebuffer backed by dmabuf, wait for fence */
resv = i915_gem_object_get_dmabuf_resv(obj);
@@ -13508,8 +13508,8 @@ static int intel_atomic_prepare_commit(struct drm_device *dev,
if (!intel_plane_state->wait_req)
continue;
- ret = __i915_wait_request(intel_plane_state->wait_req,
- true, NULL, NULL);
+ ret = i915_wait_request(intel_plane_state->wait_req,
+ true, NULL, NULL);
if (ret) {
/* Any hang should be swallowed by the wait */
WARN_ON(ret == -EIO);
@@ -13621,8 +13621,8 @@ static void intel_atomic_commit_tail(struct drm_atomic_state *state)
if (!intel_plane_state->wait_req)
continue;
- ret = __i915_wait_request(intel_plane_state->wait_req,
- true, NULL, NULL);
+ ret = i915_wait_request(intel_plane_state->wait_req,
+ true, NULL, NULL);
/* EIO should be eaten, and we can't get interrupted in the
* worker, and blocking commits have waited already. */
WARN_ON(ret);
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index 4ab6d2365e30..ac2e610ab37f 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -2240,9 +2240,9 @@ int intel_engine_idle(struct intel_engine_cs *engine)
link);
/* Make sure we do not trigger any retires */
- return __i915_wait_request(req,
- req->i915->mm.interruptible,
- NULL, NULL);
+ return i915_wait_request(req,
+ req->i915->mm.interruptible,
+ NULL, NULL);
}
int intel_ring_alloc_request_extras(struct drm_i915_gem_request *request)
@@ -2299,7 +2299,7 @@ static int wait_for_space(struct drm_i915_gem_request *req, int bytes)
if (WARN_ON(&target->ring_link == &ring->request_list))
return -ENOSPC;
- ret = __i915_wait_request(target, true, NULL, NO_WAITBOOST);
+ ret = i915_wait_request(target, true, NULL, NO_WAITBOOST);
if (ret)
return ret;
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 22/28] drm/i915: Double check activity before relocations
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (20 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 21/28] drm/i915: s/__i915_wait_request/i915_wait_request/ Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 23/28] drm/i915: Move request list retirement to i915_gem_request.c Chris Wilson
` (7 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
If the object is active and we need to perform a relocation upon it, we
need to take the slow relocation path. Before we do, double check the
active requests to see if they have completed.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 16 +++++++++++++++-
1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 5e1fb85b708b..2f7173d7a7b0 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -441,6 +441,20 @@ relocate_entry_clflush(struct drm_i915_gem_object *obj,
return 0;
}
+static bool object_is_idle(struct drm_i915_gem_object *obj)
+{
+ unsigned long active = obj->active;
+ int idx;
+
+ for_each_active(active, idx) {
+ if (!i915_gem_active_is_idle(&obj->last_read[idx],
+ &obj->base.dev->struct_mutex))
+ return false;
+ }
+
+ return true;
+}
+
static int
i915_gem_execbuffer_relocate_entry(struct drm_i915_gem_object *obj,
struct eb_vmas *eb,
@@ -524,7 +538,7 @@ i915_gem_execbuffer_relocate_entry(struct drm_i915_gem_object *obj,
}
/* We can't wait for rendering with pagefaults disabled */
- if (obj->active && pagefault_disabled())
+ if (pagefault_disabled() && !object_is_idle(obj))
return -EFAULT;
if (use_cpu_reloc(obj))
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 23/28] drm/i915: Move request list retirement to i915_gem_request.c
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (21 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 22/28] drm/i915: Double check activity before relocations Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 24/28] drm/i915: i915_vma_move_to_active prep patch Chris Wilson
` (6 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
As the list retirement is now clean of implementation details, we can
move it closer to the request management.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem.c | 44 ---------------------------------
drivers/gpu/drm/i915/i915_gem_request.c | 35 ++++++++++++++++++++++++++
2 files changed, 35 insertions(+), 44 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 46e41c42f8b9..709e527fc0f9 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2534,50 +2534,6 @@ void i915_gem_reset(struct drm_device *dev)
i915_gem_restore_fences(dev);
}
-/**
- * This function clears the request list as sequence numbers are passed.
- * @engine: engine to retire requests on
- */
-void
-i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
-{
- while (!list_empty(&engine->request_list)) {
- struct drm_i915_gem_request *request;
-
- request = list_first_entry(&engine->request_list,
- struct drm_i915_gem_request,
- link);
-
- if (!i915_gem_request_completed(request))
- break;
-
- i915_gem_request_retire_upto(request);
- }
-}
-
-void i915_gem_retire_requests(struct drm_i915_private *dev_priv)
-{
- struct intel_engine_cs *engine;
-
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
- if (dev_priv->gt.active_engines == 0)
- return;
-
- GEM_BUG_ON(!dev_priv->gt.awake);
-
- for_each_engine(engine, dev_priv) {
- i915_gem_retire_requests_ring(engine);
- if (list_empty(&engine->request_list))
- dev_priv->gt.active_engines &= ~intel_engine_flag(engine);
- }
-
- if (dev_priv->gt.active_engines == 0)
- queue_delayed_work(dev_priv->wq,
- &dev_priv->gt.idle_work,
- msecs_to_jiffies(100));
-}
-
static void
i915_gem_retire_work_handler(struct work_struct *work)
{
diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
index 8549375aa75e..6faa84832ade 100644
--- a/drivers/gpu/drm/i915/i915_gem_request.c
+++ b/drivers/gpu/drm/i915/i915_gem_request.c
@@ -731,3 +731,38 @@ complete:
return ret;
}
+
+void i915_gem_retire_requests_ring(struct intel_engine_cs *engine)
+{
+ struct drm_i915_gem_request *request, *next;
+
+ list_for_each_entry_safe(request, next, &engine->request_list, link) {
+ if (!i915_gem_request_completed(request))
+ break;
+
+ i915_gem_request_retire(request);
+ }
+}
+
+void i915_gem_retire_requests(struct drm_i915_private *dev_priv)
+{
+ struct intel_engine_cs *engine;
+
+ lockdep_assert_held(&dev_priv->drm.struct_mutex);
+
+ if (dev_priv->gt.active_engines == 0)
+ return;
+
+ GEM_BUG_ON(!dev_priv->gt.awake);
+
+ for_each_engine(engine, dev_priv) {
+ i915_gem_retire_requests_ring(engine);
+ if (list_empty(&engine->request_list))
+ dev_priv->gt.active_engines &= ~intel_engine_flag(engine);
+ }
+
+ if (dev_priv->gt.active_engines == 0)
+ queue_delayed_work(dev_priv->wq,
+ &dev_priv->gt.idle_work,
+ msecs_to_jiffies(100));
+}
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 24/28] drm/i915: i915_vma_move_to_active prep patch
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (22 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 23/28] drm/i915: Move request list retirement to i915_gem_request.c Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 25/28] drm/i915: Track active vma requests Chris Wilson
` (5 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
This patch is broken out of the next just to remove the code motion from
that patch and make it more readable. What we do here is move the
i915_vma_move_to_active() to i915_gem_execbuffer.c and put the three
stages (read, write, fenced) together so that future modifications to
active handling are all located in the same spot. The importance of this
is so that we can more simply control the order in which the requests
are place in the retirement list (i.e. control the order at which we
retire and so control the lifetimes to avoid having to hold onto
references).
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.h | 3 +-
drivers/gpu/drm/i915/i915_gem.c | 18 --------
drivers/gpu/drm/i915/i915_gem_context.c | 9 ++--
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 65 ++++++++++++++++++----------
drivers/gpu/drm/i915/i915_gem_render_state.c | 2 +-
5 files changed, 51 insertions(+), 46 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 6b57d8ff7218..a1c4c768c0c3 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3153,7 +3153,8 @@ int __must_check i915_mutex_lock_interruptible(struct drm_device *dev);
int i915_gem_object_sync(struct drm_i915_gem_object *obj,
struct drm_i915_gem_request *to);
void i915_vma_move_to_active(struct i915_vma *vma,
- struct drm_i915_gem_request *req);
+ struct drm_i915_gem_request *req,
+ unsigned int flags);
int i915_gem_dumb_create(struct drm_file *file_priv,
struct drm_device *dev,
struct drm_mode_create_dumb *args);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 709e527fc0f9..a907f836a63f 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2330,24 +2330,6 @@ void *i915_gem_object_pin_map(struct drm_i915_gem_object *obj)
return obj->mapping;
}
-void i915_vma_move_to_active(struct i915_vma *vma,
- struct drm_i915_gem_request *req)
-{
- struct drm_i915_gem_object *obj = vma->obj;
- struct intel_engine_cs *engine;
-
- engine = i915_gem_request_get_engine(req);
-
- /* Add a reference if we're newly entering the active list. */
- if (obj->active == 0)
- i915_gem_object_get(obj);
- obj->active |= intel_engine_flag(engine);
-
- i915_gem_active_set(&obj->last_read[engine->id], req);
-
- list_move_tail(&vma->vm_link, &vma->vm->active_list);
-}
-
static void
i915_gem_object_retire__write(struct i915_gem_active *active,
struct drm_i915_gem_request *request)
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 24383f0ea3a4..823e74c86eaa 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -816,8 +816,8 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
* MI_SET_CONTEXT instead of when the next seqno has completed.
*/
if (from != NULL) {
- from->engine[RCS].state->base.read_domains = I915_GEM_DOMAIN_INSTRUCTION;
- i915_vma_move_to_active(i915_gem_obj_to_ggtt(from->engine[RCS].state), req);
+ struct drm_i915_gem_object *obj = from->engine[RCS].state;
+
/* As long as MI_SET_CONTEXT is serializing, ie. it flushes the
* whole damn pipeline, we don't need to explicitly mark the
* object dirty. The only exception is that the context must be
@@ -825,10 +825,11 @@ static int do_rcs_switch(struct drm_i915_gem_request *req)
* able to defer doing this until we know the object would be
* swapped, but there is no way to do that yet.
*/
- from->engine[RCS].state->dirty = 1;
+ obj->base.read_domains = I915_GEM_DOMAIN_INSTRUCTION;
+ i915_vma_move_to_active(i915_gem_obj_to_ggtt(obj), req, 0);
/* obj is kept alive until the next request by its active ref */
- i915_gem_object_ggtt_unpin(from->engine[RCS].state);
+ i915_gem_object_ggtt_unpin(obj);
i915_gem_context_put(from);
}
engine->last_context = i915_gem_context_get(to);
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 2f7173d7a7b0..8bf20f52b79f 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1143,43 +1143,64 @@ i915_gem_validate_context(struct drm_device *dev, struct drm_file *file,
return ctx;
}
+void i915_vma_move_to_active(struct i915_vma *vma,
+ struct drm_i915_gem_request *req,
+ unsigned int flags)
+{
+ struct drm_i915_gem_object *obj = vma->obj;
+ const unsigned int idx = req->engine->id;
+
+ GEM_BUG_ON(!drm_mm_node_allocated(&vma->node));
+
+ obj->dirty = 1; /* be paranoid */
+
+ /* Add a reference if we're newly entering the active list. */
+ if (obj->active == 0)
+ i915_gem_object_get(obj);
+ obj->active |= 1 << idx;
+ i915_gem_active_set(&obj->last_read[idx], req);
+
+ if (flags & EXEC_OBJECT_WRITE) {
+ i915_gem_active_set(&obj->last_write, req);
+
+ intel_fb_obj_invalidate(obj, ORIGIN_CS);
+
+ /* update for the implicit flush after a batch */
+ obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS;
+ }
+
+ if (flags & EXEC_OBJECT_NEEDS_FENCE) {
+ i915_gem_active_set(&obj->last_fence, req);
+ if (flags & __EXEC_OBJECT_HAS_FENCE) {
+ struct drm_i915_private *dev_priv = req->i915;
+
+ list_move_tail(&dev_priv->fence_regs[obj->fence_reg].lru_list,
+ &dev_priv->mm.fence_list);
+ }
+ }
+
+ list_move_tail(&vma->vm_link, &vma->vm->active_list);
+}
+
static void
i915_gem_execbuffer_move_to_active(struct list_head *vmas,
struct drm_i915_gem_request *req)
{
- struct intel_engine_cs *engine = i915_gem_request_get_engine(req);
struct i915_vma *vma;
list_for_each_entry(vma, vmas, exec_list) {
- struct drm_i915_gem_exec_object2 *entry = vma->exec_entry;
struct drm_i915_gem_object *obj = vma->obj;
u32 old_read = obj->base.read_domains;
u32 old_write = obj->base.write_domain;
- obj->dirty = 1; /* be paranoid */
obj->base.write_domain = obj->base.pending_write_domain;
- if (obj->base.write_domain == 0)
+ if (obj->base.write_domain)
+ vma->exec_entry->flags |= EXEC_OBJECT_WRITE;
+ else
obj->base.pending_read_domains |= obj->base.read_domains;
obj->base.read_domains = obj->base.pending_read_domains;
- i915_vma_move_to_active(vma, req);
- if (obj->base.write_domain) {
- i915_gem_active_set(&obj->last_write, req);
-
- intel_fb_obj_invalidate(obj, ORIGIN_CS);
-
- /* update for the implicit flush after a batch */
- obj->base.write_domain &= ~I915_GEM_GPU_DOMAINS;
- }
- if (entry->flags & EXEC_OBJECT_NEEDS_FENCE) {
- i915_gem_active_set(&obj->last_fence, req);
- if (entry->flags & __EXEC_OBJECT_HAS_FENCE) {
- struct drm_i915_private *dev_priv = engine->i915;
- list_move_tail(&dev_priv->fence_regs[obj->fence_reg].lru_list,
- &dev_priv->mm.fence_list);
- }
- }
-
+ i915_vma_move_to_active(vma, req, vma->exec_entry->flags);
trace_i915_gem_object_change_domain(obj, old_read, old_write);
}
}
diff --git a/drivers/gpu/drm/i915/i915_gem_render_state.c b/drivers/gpu/drm/i915/i915_gem_render_state.c
index f85c5505bce2..90236672ac1e 100644
--- a/drivers/gpu/drm/i915/i915_gem_render_state.c
+++ b/drivers/gpu/drm/i915/i915_gem_render_state.c
@@ -217,7 +217,7 @@ int i915_gem_render_state_init(struct drm_i915_gem_request *req)
goto err_unpin;
}
- i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), req);
+ i915_vma_move_to_active(i915_gem_obj_to_ggtt(so.obj), req, 0);
err_unpin:
i915_gem_object_ggtt_unpin(so.obj);
err_obj:
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 25/28] drm/i915: Track active vma requests
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (23 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 24/28] drm/i915: i915_vma_move_to_active prep patch Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 26/28] drm/i915: Release vma when the handle is closed Chris Wilson
` (4 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
Hook the vma itself into the i915_gem_request_retire() so that we can
accurately track when a solitary vma is inactive (as opposed to having
to wait for the entire object to be idle). This improves the interaction
when using multiple contexts (with full-ppgtt) and eliminates some
frequent list walking when retiring objects after a completed request.
A side-effect is that we get an active vma reference for free. The
consequence of this is shown in the next patch...
v2: Update inline names to be consistent with
i915_gem_object_get_active()
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_debugfs.c | 2 +-
drivers/gpu/drm/i915/i915_gem.c | 50 +++++++++++++++++-------------
drivers/gpu/drm/i915/i915_gem_execbuffer.c | 10 +++++-
drivers/gpu/drm/i915/i915_gem_gtt.c | 20 ++++++++++++
drivers/gpu/drm/i915/i915_gem_gtt.h | 33 ++++++++++++++++++++
5 files changed, 91 insertions(+), 24 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index fe3c82339c8e..b35e6174fa7d 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -365,7 +365,7 @@ static int per_file_stats(int id, void *ptr, void *data)
continue;
}
- if (obj->active) /* XXX per-vma statistic */
+ if (i915_vma_is_active(vma))
stats->active += vma->node.size;
else
stats->inactive += vma->node.size;
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index a907f836a63f..f95bfb216022 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2347,7 +2347,6 @@ i915_gem_object_retire__read(struct i915_gem_active *active,
int idx = request->engine->id;
struct drm_i915_gem_object *obj =
container_of(active, struct drm_i915_gem_object, last_read[idx]);
- struct i915_vma *vma;
GEM_BUG_ON((obj->active & (1 << idx)) == 0);
@@ -2359,12 +2358,9 @@ i915_gem_object_retire__read(struct i915_gem_active *active,
* so that we don't steal from recently used but inactive objects
* (unless we are forced to ofc!)
*/
- list_move_tail(&obj->global_list, &request->i915->mm.bound_list);
-
- list_for_each_entry(vma, &obj->vma_list, obj_link) {
- if (!list_empty(&vma->vm_link))
- list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
- }
+ if (obj->bind_count)
+ list_move_tail(&obj->global_list,
+ &request->i915->mm.bound_list);
i915_gem_object_put(obj);
}
@@ -2797,8 +2793,29 @@ static void __i915_vma_iounmap(struct i915_vma *vma)
static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
{
struct drm_i915_gem_object *obj = vma->obj;
+ unsigned long active;
int ret;
+ /* First wait upon any activity as retiring the request may
+ * have side-effects such as unpinning or even unbinding this vma.
+ */
+ active = i915_vma_get_active(vma);
+ if (active && wait) {
+ int idx;
+
+ for_each_active(active, idx) {
+ ret = i915_gem_active_retire(&vma->last_read[idx],
+ &vma->vm->dev->struct_mutex);
+ if (ret)
+ return ret;
+ }
+
+ GEM_BUG_ON(i915_vma_is_active(vma));
+ }
+
+ if (vma->pin_count)
+ return -EBUSY;
+
if (list_empty(&vma->obj_link))
return 0;
@@ -2807,18 +2824,9 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
return 0;
}
- if (vma->pin_count)
- return -EBUSY;
-
GEM_BUG_ON(obj->bind_count == 0);
GEM_BUG_ON(!obj->pages);
- if (wait) {
- ret = i915_gem_object_wait_rendering(obj, false);
- if (ret)
- return ret;
- }
-
if (vma->is_ggtt && vma->ggtt_view.type == I915_GGTT_VIEW_NORMAL) {
i915_gem_object_finish_gtt(obj);
@@ -3201,9 +3209,6 @@ i915_gem_object_flush_cpu_write_domain(struct drm_i915_gem_object *obj)
int
i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
{
- struct drm_device *dev = obj->base.dev;
- struct drm_i915_private *dev_priv = to_i915(dev);
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
uint32_t old_write_domain, old_read_domains;
struct i915_vma *vma;
int ret;
@@ -3256,9 +3261,10 @@ i915_gem_object_set_to_gtt_domain(struct drm_i915_gem_object *obj, bool write)
/* And bump the LRU for this access */
vma = i915_gem_obj_to_ggtt(obj);
- if (vma && drm_mm_node_allocated(&vma->node) && !obj->active)
- list_move_tail(&vma->vm_link,
- &ggtt->base.inactive_list);
+ if (vma &&
+ drm_mm_node_allocated(&vma->node) &&
+ !i915_vma_is_active(vma))
+ list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 8bf20f52b79f..5e3b5054f72d 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1154,7 +1154,13 @@ void i915_vma_move_to_active(struct i915_vma *vma,
obj->dirty = 1; /* be paranoid */
- /* Add a reference if we're newly entering the active list. */
+ /* Add a reference if we're newly entering the active list.
+ * The order in which we add operations to the retirement queue is
+ * vital here: mark_active adds to the start of the callback list,
+ * such that subsequent callbacks are called first. Therefore we
+ * add the active reference first and queue for it to be dropped
+ * *last*.
+ */
if (obj->active == 0)
i915_gem_object_get(obj);
obj->active |= 1 << idx;
@@ -1179,6 +1185,8 @@ void i915_vma_move_to_active(struct i915_vma *vma,
}
}
+ i915_vma_set_active(vma, idx);
+ i915_gem_active_set(&vma->last_read[idx], req);
list_move_tail(&vma->vm_link, &vma->vm->active_list);
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index f8e71a4a8a5d..fc9d6bec7e36 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3332,12 +3332,30 @@ void i915_gem_restore_gtt_mappings(struct drm_device *dev)
i915_ggtt_flush(dev_priv);
}
+static void
+i915_vma_retire(struct i915_gem_active *active,
+ struct drm_i915_gem_request *rq)
+{
+ const unsigned int idx = rq->engine->id;
+ struct i915_vma *vma =
+ container_of(active, struct i915_vma, last_read[idx]);
+
+ GEM_BUG_ON(!i915_vma_has_active_engine(vma, idx));
+
+ i915_vma_clear_active(vma, idx);
+ if (i915_vma_is_active(vma))
+ return;
+
+ list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
+}
+
static struct i915_vma *
__i915_gem_vma_create(struct drm_i915_gem_object *obj,
struct i915_address_space *vm,
const struct i915_ggtt_view *ggtt_view)
{
struct i915_vma *vma;
+ int i;
if (WARN_ON(i915_is_ggtt(vm) != !!ggtt_view))
return ERR_PTR(-EINVAL);
@@ -3349,6 +3367,8 @@ __i915_gem_vma_create(struct drm_i915_gem_object *obj,
INIT_LIST_HEAD(&vma->vm_link);
INIT_LIST_HEAD(&vma->obj_link);
INIT_LIST_HEAD(&vma->exec_list);
+ for (i = 0; i < ARRAY_SIZE(vma->last_read); i++)
+ init_request_active(&vma->last_read[i], i915_vma_retire);
vma->vm = vm;
vma->obj = obj;
vma->is_ggtt = i915_is_ggtt(vm);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 4498bdd46b6b..133d9b160ff1 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -36,6 +36,8 @@
#include <linux/io-mapping.h>
+#include "i915_gem_request.h"
+
struct drm_i915_file_private;
typedef uint32_t gen6_pte_t;
@@ -179,6 +181,9 @@ struct i915_vma {
struct i915_address_space *vm;
void __iomem *iomap;
+ unsigned int active;
+ struct i915_gem_active last_read[I915_NUM_ENGINES];
+
/** Flags and address space this VMA is bound to */
#define GLOBAL_BIND (1<<0)
#define LOCAL_BIND (1<<1)
@@ -222,6 +227,34 @@ struct i915_vma {
#define DRM_I915_GEM_OBJECT_MAX_PIN_COUNT 0xf
};
+static inline unsigned int i915_vma_get_active(const struct i915_vma *vma)
+{
+ return vma->active;
+}
+
+static inline bool i915_vma_is_active(const struct i915_vma *vma)
+{
+ return i915_vma_get_active(vma);
+}
+
+static inline void i915_vma_set_active(struct i915_vma *vma,
+ unsigned int engine)
+{
+ vma->active |= BIT(engine);
+}
+
+static inline void i915_vma_clear_active(struct i915_vma *vma,
+ unsigned int engine)
+{
+ vma->active &= ~BIT(engine);
+}
+
+static inline bool i915_vma_has_active_engine(const struct i915_vma *vma,
+ unsigned int engine)
+{
+ return vma->active & BIT(engine);
+}
+
struct i915_page_dma {
struct page *page;
union {
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 26/28] drm/i915: Release vma when the handle is closed
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (24 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 25/28] drm/i915: Track active vma requests Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 27/28] drm/i915: Mark the context and address space as closed Chris Wilson
` (3 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
In order to prevent a leak of the vma on shared objects, we need to
hook into the object_close callback to destroy the vma on the object for
this file. However, if we destroyed that vma immediately we may cause
unexpected application stalls as we try to unbind a busy vma - hence we
defer the unbind to when we retire the vma.
v2: Keep vma allocated until closed. This is useful for a later
optimisation, but it is required now in order to handle potential
recursion of i915_vma_unbind() by retiring itself.
v3: Comments are important.
Testcase: igt/gem_ppggtt/flink-and-close-vma-leak
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.c | 1 +
drivers/gpu/drm/i915/i915_drv.h | 4 +-
drivers/gpu/drm/i915/i915_gem.c | 88 ++++++++++++++++++-----------------
drivers/gpu/drm/i915/i915_gem_evict.c | 8 +---
drivers/gpu/drm/i915/i915_gem_gtt.c | 25 ++++++++++
drivers/gpu/drm/i915/i915_gem_gtt.h | 1 +
6 files changed, 77 insertions(+), 50 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 50c564089fa8..8cfc264ec9f6 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -2578,6 +2578,7 @@ static struct drm_driver driver = {
.postclose = i915_driver_postclose,
.set_busid = drm_pci_set_busid,
+ .gem_close_object = i915_gem_close_object,
.gem_free_object = i915_gem_free_object,
.gem_vm_ops = &i915_gem_vm_ops,
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index a1c4c768c0c3..f470ea195253 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3014,8 +3014,8 @@ struct drm_i915_gem_object *i915_gem_object_create(struct drm_device *dev,
size_t size);
struct drm_i915_gem_object *i915_gem_object_create_from_data(
struct drm_device *dev, const void *data, size_t size);
+void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file);
void i915_gem_free_object(struct drm_gem_object *obj);
-void i915_gem_vma_destroy(struct i915_vma *vma);
/* Flags used by pin/bind&friends. */
#define PIN_MAPPABLE (1<<0)
@@ -3048,6 +3048,8 @@ int __must_check i915_vma_unbind(struct i915_vma *vma);
* _guarantee_ VMA in question is _not in use_ anywhere.
*/
int __must_check __i915_vma_unbind_no_wait(struct i915_vma *vma);
+void i915_vma_close(struct i915_vma *vma);
+void i915_vma_destroy(struct i915_vma *vma);
int i915_gem_object_unbind(struct drm_i915_gem_object *obj);
int i915_gem_object_put_pages(struct drm_i915_gem_object *obj);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index f95bfb216022..663cddbf9444 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2596,6 +2596,19 @@ out_rearm:
}
}
+void i915_gem_close_object(struct drm_gem_object *gem, struct drm_file *file)
+{
+ struct drm_i915_gem_object *obj = to_intel_bo(gem);
+ struct drm_i915_file_private *fpriv = file->driver_priv;
+ struct i915_vma *vma, *vn;
+
+ mutex_lock(&obj->base.dev->struct_mutex);
+ list_for_each_entry_safe(vma, vn, &obj->vma_list, obj_link)
+ if (vma->vm->file == fpriv)
+ i915_vma_close(vma);
+ mutex_unlock(&obj->base.dev->struct_mutex);
+}
+
/**
* i915_gem_wait_ioctl - implements DRM_IOCTL_I915_GEM_WAIT
* @dev: drm device pointer
@@ -2803,26 +2816,32 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
if (active && wait) {
int idx;
+ /* When a closed VMA is retired, it is unbound - eek.
+ * In order to prevent it from being recursively closed,
+ * take a pin on the vma so that the second unbind is
+ * aborted.
+ */
+ vma->pin_count++;
+
for_each_active(active, idx) {
ret = i915_gem_active_retire(&vma->last_read[idx],
&vma->vm->dev->struct_mutex);
if (ret)
- return ret;
+ break;
}
+ vma->pin_count--;
+ if (ret)
+ return ret;
+
GEM_BUG_ON(i915_vma_is_active(vma));
}
if (vma->pin_count)
return -EBUSY;
- if (list_empty(&vma->obj_link))
- return 0;
-
- if (!drm_mm_node_allocated(&vma->node)) {
- i915_gem_vma_destroy(vma);
- return 0;
- }
+ if (!drm_mm_node_allocated(&vma->node))
+ goto destroy;
GEM_BUG_ON(obj->bind_count == 0);
GEM_BUG_ON(!obj->pages);
@@ -2855,7 +2874,6 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
}
drm_mm_remove_node(&vma->node);
- i915_gem_vma_destroy(vma);
/* Since the unbound list is global, only move to that list if
* no more VMAs exist. */
@@ -2869,6 +2887,10 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
*/
i915_gem_object_unpin_pages(obj);
+destroy:
+ if (unlikely(vma->closed))
+ i915_vma_destroy(vma);
+
return 0;
}
@@ -3043,7 +3065,7 @@ i915_gem_object_bind_to_vm(struct drm_i915_gem_object *obj,
if (offset & (alignment - 1) || offset + size > end) {
ret = -EINVAL;
- goto err_free_vma;
+ goto err_vma;
}
vma->node.start = offset;
vma->node.size = size;
@@ -3055,7 +3077,7 @@ i915_gem_object_bind_to_vm(struct drm_i915_gem_object *obj,
ret = drm_mm_reserve_node(&vm->mm, &vma->node);
}
if (ret)
- goto err_free_vma;
+ goto err_vma;
} else {
if (flags & PIN_HIGH) {
search_flag = DRM_MM_SEARCH_BELOW;
@@ -3080,7 +3102,7 @@ search_free:
if (ret == 0)
goto search_free;
- goto err_free_vma;
+ goto err_vma;
}
}
if (WARN_ON(!i915_gem_valid_gtt_space(vma, obj->cache_level))) {
@@ -3101,8 +3123,7 @@ search_free:
err_remove_node:
drm_mm_remove_node(&vma->node);
-err_free_vma:
- i915_gem_vma_destroy(vma);
+err_vma:
vma = ERR_PTR(ret);
err_unpin:
i915_gem_object_unpin_pages(obj);
@@ -4051,21 +4072,18 @@ void i915_gem_free_object(struct drm_gem_object *gem_obj)
trace_i915_gem_object_destroy(obj);
+ /* All file-owned VMA should have been released by this point through
+ * i915_gem_close_object(), or earlier by i915_gem_context_close().
+ * However, the object may also be bound into the global GTT (e.g.
+ * older GPUs without per-process support, or for direct access through
+ * the GTT either for the user or for scanout). Those VMA still need to
+ * unbound now.
+ */
list_for_each_entry_safe(vma, next, &obj->vma_list, obj_link) {
- int ret;
-
+ GEM_BUG_ON(!vma->is_ggtt);
+ GEM_BUG_ON(i915_vma_is_active(vma));
vma->pin_count = 0;
- ret = __i915_vma_unbind_no_wait(vma);
- if (WARN_ON(ret == -ERESTARTSYS)) {
- bool was_interruptible;
-
- was_interruptible = dev_priv->mm.interruptible;
- dev_priv->mm.interruptible = false;
-
- WARN_ON(i915_vma_unbind(vma));
-
- dev_priv->mm.interruptible = was_interruptible;
- }
+ i915_vma_close(vma);
}
GEM_BUG_ON(obj->bind_count);
@@ -4129,22 +4147,6 @@ struct i915_vma *i915_gem_obj_to_ggtt_view(struct drm_i915_gem_object *obj,
return NULL;
}
-void i915_gem_vma_destroy(struct i915_vma *vma)
-{
- WARN_ON(vma->node.allocated);
-
- /* Keep the vma as a placeholder in the execbuffer reservation lists */
- if (!list_empty(&vma->exec_list))
- return;
-
- if (!vma->is_ggtt)
- i915_ppgtt_put(i915_vm_to_ppgtt(vma->vm));
-
- list_del(&vma->obj_link);
-
- kmem_cache_free(to_i915(vma->obj->base.dev)->vmas, vma);
-}
-
static void
i915_gem_stop_engines(struct drm_device *dev)
{
diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
index 81f7b4383d5e..3437ced76cb6 100644
--- a/drivers/gpu/drm/i915/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/i915_gem_evict.c
@@ -182,8 +182,8 @@ found:
struct i915_vma,
exec_list);
if (drm_mm_scan_remove_block(&vma->node)) {
+ vma->pin_count++;
list_move(&vma->exec_list, &eviction_list);
- i915_gem_object_get(vma->obj);
continue;
}
list_del_init(&vma->exec_list);
@@ -191,18 +191,14 @@ found:
/* Unbinding will emit any required flushes */
while (!list_empty(&eviction_list)) {
- struct drm_i915_gem_object *obj;
-
vma = list_first_entry(&eviction_list,
struct i915_vma,
exec_list);
- obj = vma->obj;
list_del_init(&vma->exec_list);
+ vma->pin_count--;
if (ret == 0)
ret = i915_vma_unbind(vma);
-
- i915_gem_object_put(obj);
}
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index fc9d6bec7e36..c621a243592d 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3347,6 +3347,31 @@ i915_vma_retire(struct i915_gem_active *active,
return;
list_move_tail(&vma->vm_link, &vma->vm->inactive_list);
+ if (unlikely(vma->closed && !vma->pin_count))
+ WARN_ON(i915_vma_unbind(vma));
+}
+
+void i915_vma_destroy(struct i915_vma *vma)
+{
+ GEM_BUG_ON(vma->node.allocated);
+ GEM_BUG_ON(i915_vma_is_active(vma));
+ GEM_BUG_ON(!vma->closed);
+
+ list_del(&vma->vm_link);
+ if (!vma->is_ggtt)
+ i915_ppgtt_put(i915_vm_to_ppgtt(vma->vm));
+
+ kmem_cache_free(to_i915(vma->obj->base.dev)->vmas, vma);
+}
+
+void i915_vma_close(struct i915_vma *vma)
+{
+ GEM_BUG_ON(vma->closed);
+ vma->closed = true;
+
+ list_del_init(&vma->obj_link);
+ if (!i915_vma_is_active(vma) && !vma->pin_count)
+ WARN_ON(__i915_vma_unbind_no_wait(vma));
}
static struct i915_vma *
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 133d9b160ff1..44b05d240e3b 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -189,6 +189,7 @@ struct i915_vma {
#define LOCAL_BIND (1<<1)
unsigned int bound : 4;
bool is_ggtt : 1;
+ bool closed : 1;
/**
* Support different GGTT views into the same object.
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 27/28] drm/i915: Mark the context and address space as closed
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (25 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 26/28] drm/i915: Release vma when the handle is closed Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 13:43 ` [PATCH 28/28] Revert "drm/i915: Clean up associated VMAs on context destruction" Chris Wilson
` (2 subsequent siblings)
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
When the user closes the context mark it and the dependent address space
as closed. As we use an asynchronous destruct method, this has two
purposes. First it allows us to flag the closed context and detect
internal errors if we to create any new objects for it (as it is removed
from the user's namespace, these should be internal bugs only). And
secondly, it allows us to immediately reap stale vma.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.h | 1 +
drivers/gpu/drm/i915/i915_gem.c | 15 ++++++------
drivers/gpu/drm/i915/i915_gem_context.c | 43 ++++++++++++++++++++++++++++-----
drivers/gpu/drm/i915/i915_gem_gtt.c | 9 +++++--
drivers/gpu/drm/i915/i915_gem_gtt.h | 9 +++++++
drivers/gpu/drm/i915/i915_gem_stolen.c | 2 +-
6 files changed, 63 insertions(+), 16 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index f470ea195253..ce472c9bd7f9 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -907,6 +907,7 @@ struct i915_gem_context {
struct list_head link;
u8 remap_slice;
+ bool closed:1;
};
enum fb_op_origin {
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 663cddbf9444..f20af4e9e4bc 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2857,12 +2857,15 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
__i915_vma_iounmap(vma);
}
- trace_i915_vma_unbind(vma);
-
- vma->vm->unbind_vma(vma);
+ if (likely(!vma->vm->closed)) {
+ trace_i915_vma_unbind(vma);
+ vma->vm->unbind_vma(vma);
+ }
vma->bound = 0;
- list_del_init(&vma->vm_link);
+ drm_mm_remove_node(&vma->node);
+ list_move_tail(&vma->vm_link, &vma->vm->unbound_list);
+
if (vma->is_ggtt) {
if (vma->ggtt_view.type == I915_GGTT_VIEW_NORMAL) {
obj->map_and_fenceable = false;
@@ -2873,8 +2876,6 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
vma->ggtt_view.pages = NULL;
}
- drm_mm_remove_node(&vma->node);
-
/* Since the unbound list is global, only move to that list if
* no more VMAs exist. */
if (--obj->bind_count == 0)
@@ -3116,7 +3117,7 @@ search_free:
goto err_remove_node;
list_move_tail(&obj->global_list, &dev_priv->mm.bound_list);
- list_add_tail(&vma->vm_link, &vm->inactive_list);
+ list_move_tail(&vma->vm_link, &vm->inactive_list);
obj->bind_count++;
return vma;
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index 823e74c86eaa..a4ee62335e75 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -156,6 +156,7 @@ void i915_gem_context_free(struct kref *ctx_ref)
lockdep_assert_held(&ctx->i915->drm.struct_mutex);
trace_i915_context_free(ctx);
+ GEM_BUG_ON(!ctx->closed);
/*
* This context is going away and we need to remove all VMAs still
@@ -224,6 +225,37 @@ i915_gem_alloc_context_obj(struct drm_device *dev, size_t size)
return obj;
}
+static void i915_ppgtt_close(struct i915_address_space *vm)
+{
+ struct list_head *phases[] = {
+ &vm->active_list,
+ &vm->inactive_list,
+ &vm->unbound_list,
+ NULL,
+ }, **phase;
+
+ GEM_BUG_ON(vm->closed);
+ vm->closed = true;
+
+ for (phase = phases; *phase; phase++) {
+ struct i915_vma *vma, *vn;
+
+ list_for_each_entry_safe(vma, vn, *phase, vm_link)
+ if (!vma->closed)
+ i915_vma_close(vma);
+ }
+}
+
+static void context_close(struct i915_gem_context *ctx)
+{
+ GEM_BUG_ON(ctx->closed);
+ ctx->closed = true;
+ if (ctx->ppgtt)
+ i915_ppgtt_close(&ctx->ppgtt->base);
+ ctx->file_priv = ERR_PTR(-EBADF);
+ i915_gem_context_put(ctx);
+}
+
static int assign_hw_id(struct drm_i915_private *dev_priv, unsigned *out)
{
int ret;
@@ -305,7 +337,7 @@ __create_hw_context(struct drm_device *dev,
return ctx;
err_out:
- i915_gem_context_put(ctx);
+ context_close(ctx);
return ERR_PTR(ret);
}
@@ -334,7 +366,7 @@ i915_gem_create_context(struct drm_device *dev,
DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n",
PTR_ERR(ppgtt));
idr_remove(&file_priv->context_idr, ctx->user_handle);
- i915_gem_context_put(ctx);
+ context_close(ctx);
return ERR_CAST(ppgtt);
}
@@ -505,7 +537,7 @@ void i915_gem_context_fini(struct drm_device *dev)
lockdep_assert_held(&dev->struct_mutex);
- i915_gem_context_put(dctx);
+ context_close(dctx);
dev_priv->kernel_context = NULL;
ida_destroy(&dev_priv->context_hw_ida);
@@ -515,8 +547,7 @@ static int context_idr_cleanup(int id, void *p, void *data)
{
struct i915_gem_context *ctx = p;
- ctx->file_priv = ERR_PTR(-EBADF);
- i915_gem_context_put(ctx);
+ context_close(ctx);
return 0;
}
@@ -1014,7 +1045,7 @@ int i915_gem_context_destroy_ioctl(struct drm_device *dev, void *data,
}
idr_remove(&file_priv->context_idr, ctx->user_handle);
- i915_gem_context_put(ctx);
+ context_close(ctx);
mutex_unlock(&dev->struct_mutex);
DRM_DEBUG_DRIVER("HW context %d destroyed\n", args->ctx_id);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index c621a243592d..623d49a08894 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2122,6 +2122,7 @@ static void i915_address_space_init(struct i915_address_space *vm,
vm->dev = &dev_priv->drm;
INIT_LIST_HEAD(&vm->active_list);
INIT_LIST_HEAD(&vm->inactive_list);
+ INIT_LIST_HEAD(&vm->unbound_list);
list_add_tail(&vm->global_link, &dev_priv->vm_list);
}
@@ -2214,9 +2215,10 @@ void i915_ppgtt_release(struct kref *kref)
trace_i915_ppgtt_release(&ppgtt->base);
- /* vmas should already be unbound */
+ /* vmas should already be unbound and destroyed */
WARN_ON(!list_empty(&ppgtt->base.active_list));
WARN_ON(!list_empty(&ppgtt->base.inactive_list));
+ WARN_ON(!list_empty(&ppgtt->base.unbound_list));
list_del(&ppgtt->base.global_link);
drm_mm_takedown(&ppgtt->base.mm);
@@ -3382,6 +3384,8 @@ __i915_gem_vma_create(struct drm_i915_gem_object *obj,
struct i915_vma *vma;
int i;
+ GEM_BUG_ON(vm->closed);
+
if (WARN_ON(i915_is_ggtt(vm) != !!ggtt_view))
return ERR_PTR(-EINVAL);
@@ -3389,11 +3393,11 @@ __i915_gem_vma_create(struct drm_i915_gem_object *obj,
if (vma == NULL)
return ERR_PTR(-ENOMEM);
- INIT_LIST_HEAD(&vma->vm_link);
INIT_LIST_HEAD(&vma->obj_link);
INIT_LIST_HEAD(&vma->exec_list);
for (i = 0; i < ARRAY_SIZE(vma->last_read); i++)
init_request_active(&vma->last_read[i], i915_vma_retire);
+ list_add(&vma->vm_link, &vm->unbound_list);
vma->vm = vm;
vma->obj = obj;
vma->is_ggtt = i915_is_ggtt(vm);
@@ -3434,6 +3438,7 @@ i915_gem_obj_lookup_or_create_ggtt_vma(struct drm_i915_gem_object *obj,
if (!vma)
vma = __i915_gem_vma_create(obj, &ggtt->base, view);
+ GEM_BUG_ON(vma->closed);
return vma;
}
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 44b05d240e3b..ae76d490ed5d 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -319,6 +319,8 @@ struct i915_address_space {
u64 start; /* Start offset always 0 for dri2 */
u64 total; /* size addr space maps (ex. 2GB for ggtt) */
+ bool closed;
+
struct i915_page_scratch *scratch_page;
struct i915_page_table *scratch_pt;
struct i915_page_directory *scratch_pd;
@@ -347,6 +349,13 @@ struct i915_address_space {
*/
struct list_head inactive_list;
+ /**
+ * List of vma that have been unbound.
+ *
+ * A reference is not held on the buffer while on this list.
+ */
+ struct list_head unbound_list;
+
/* FIXME: Need a more generic return type */
gen6_pte_t (*pte_encode)(dma_addr_t addr,
enum i915_cache_level level,
diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c
index 2c321c8df7c3..bc91ffe614e2 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/i915_gem_stolen.c
@@ -707,7 +707,7 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev,
vma->bound |= GLOBAL_BIND;
__i915_vma_set_map_and_fenceable(vma);
- list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
+ list_move_tail(&vma->vm_link, &ggtt->base.inactive_list);
obj->bind_count++;
list_add_tail(&obj->global_list, &dev_priv->mm.bound_list);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 28/28] Revert "drm/i915: Clean up associated VMAs on context destruction"
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (26 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 27/28] drm/i915: Mark the context and address space as closed Chris Wilson
@ 2016-08-03 13:43 ` Chris Wilson
2016-08-03 14:15 ` ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Patchwork
2016-08-04 5:19 ` ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers (rev4) Patchwork
29 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 13:43 UTC (permalink / raw)
To: intel-gfx
This reverts commit e9f24d5fb7cf3628b195b18ff3ac4e37937ceeae.
The patch was only a stop-gap measure that fixed half the problem - the
leak of the fbcon when restarting X. A complete solution required
releasing the VMA when the object itself was closed rather than rely on
file/process exit. The previous patches add the VMA tracking necessary
to do close them along with the object, context or file, and so the time
has come to remove the partial fix.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.h | 5 -----
drivers/gpu/drm/i915/i915_gem.c | 14 ++------------
drivers/gpu/drm/i915/i915_gem_context.c | 22 ----------------------
drivers/gpu/drm/i915/i915_gem_gtt.c | 2 +-
4 files changed, 3 insertions(+), 40 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index ce472c9bd7f9..66b98fa4715a 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -3044,11 +3044,6 @@ int i915_vma_bind(struct i915_vma *vma, enum i915_cache_level cache_level,
u32 flags);
void __i915_vma_set_map_and_fenceable(struct i915_vma *vma);
int __must_check i915_vma_unbind(struct i915_vma *vma);
-/*
- * BEWARE: Do not use the function below unless you can _absolutely_
- * _guarantee_ VMA in question is _not in use_ anywhere.
- */
-int __must_check __i915_vma_unbind_no_wait(struct i915_vma *vma);
void i915_vma_close(struct i915_vma *vma);
void i915_vma_destroy(struct i915_vma *vma);
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index f20af4e9e4bc..9bef2c0803a5 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -2803,7 +2803,7 @@ static void __i915_vma_iounmap(struct i915_vma *vma)
vma->iomap = NULL;
}
-static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
+int i915_vma_unbind(struct i915_vma *vma)
{
struct drm_i915_gem_object *obj = vma->obj;
unsigned long active;
@@ -2813,7 +2813,7 @@ static int __i915_vma_unbind(struct i915_vma *vma, bool wait)
* have side-effects such as unpinning or even unbinding this vma.
*/
active = i915_vma_get_active(vma);
- if (active && wait) {
+ if (active) {
int idx;
/* When a closed VMA is retired, it is unbound - eek.
@@ -2895,16 +2895,6 @@ destroy:
return 0;
}
-int i915_vma_unbind(struct i915_vma *vma)
-{
- return __i915_vma_unbind(vma, true);
-}
-
-int __i915_vma_unbind_no_wait(struct i915_vma *vma)
-{
- return __i915_vma_unbind(vma, false);
-}
-
int i915_gem_wait_for_idle(struct drm_i915_private *dev_priv)
{
struct intel_engine_cs *engine;
diff --git a/drivers/gpu/drm/i915/i915_gem_context.c b/drivers/gpu/drm/i915/i915_gem_context.c
index a4ee62335e75..eff6d3953ecd 100644
--- a/drivers/gpu/drm/i915/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/i915_gem_context.c
@@ -134,21 +134,6 @@ static int get_context_size(struct drm_i915_private *dev_priv)
return ret;
}
-static void i915_gem_context_clean(struct i915_gem_context *ctx)
-{
- struct i915_hw_ppgtt *ppgtt = ctx->ppgtt;
- struct i915_vma *vma, *next;
-
- if (!ppgtt)
- return;
-
- list_for_each_entry_safe(vma, next, &ppgtt->base.inactive_list,
- vm_link) {
- if (WARN_ON(__i915_vma_unbind_no_wait(vma)))
- break;
- }
-}
-
void i915_gem_context_free(struct kref *ctx_ref)
{
struct i915_gem_context *ctx = container_of(ctx_ref, typeof(*ctx), ref);
@@ -158,13 +143,6 @@ void i915_gem_context_free(struct kref *ctx_ref)
trace_i915_context_free(ctx);
GEM_BUG_ON(!ctx->closed);
- /*
- * This context is going away and we need to remove all VMAs still
- * around. This is to handle imported shared objects for which
- * destructor did not run when their handles were closed.
- */
- i915_gem_context_clean(ctx);
-
i915_ppgtt_put(ctx->ppgtt);
for (i = 0; i < I915_NUM_ENGINES; i++) {
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 623d49a08894..9d8c44900aea 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -3373,7 +3373,7 @@ void i915_vma_close(struct i915_vma *vma)
list_del_init(&vma->obj_link);
if (!i915_vma_is_active(vma) && !vma->pin_count)
- WARN_ON(__i915_vma_unbind_no_wait(vma));
+ WARN_ON(i915_vma_unbind(vma));
}
static struct i915_vma *
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* Re: [PATCH 02/28] drm/i915: Split GGTT initialisation between probing and setup
2016-08-03 13:42 ` [PATCH 02/28] drm/i915: Split GGTT initialisation between probing and setup Chris Wilson
@ 2016-08-03 13:54 ` Joonas Lahtinen
0 siblings, 0 replies; 41+ messages in thread
From: Joonas Lahtinen @ 2016-08-03 13:54 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On ke, 2016-08-03 at 14:42 +0100, Chris Wilson wrote:
> In order to handle conflicting drivers (i.e. vgacon) having a different
> setup of hardware, we have to remove those other drivers before we try
> to setup our own mappings. This requires us to split GGTT initialisation
> between probing for the hardware location (part of the PCI BAR) and
> later establishing the kernel resources for it.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
The whole probing should be removed and just relayed from the early
quirks code to the driver, but that's going to be after these series.
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Regards, Joonas
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* Re: [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local
2016-08-03 13:42 ` [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local Chris Wilson
@ 2016-08-03 14:00 ` Joonas Lahtinen
2016-08-03 14:07 ` Chris Wilson
2016-08-03 17:14 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Chris Wilson
0 siblings, 2 replies; 41+ messages in thread
From: Joonas Lahtinen @ 2016-08-03 14:00 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On ke, 2016-08-03 at 14:42 +0100, Chris Wilson wrote:
> The ggtt->probe() vfunc is only called once during GGTT initialisation.
> We can keep it in a local for the function, sparing a few bytes in our
> i915_ggtt persistent struct.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
> drivers/gpu/drm/i915/i915_gem_gtt.c | 15 ++++++++-------
> drivers/gpu/drm/i915/i915_gem_gtt.h | 2 --
> 2 files changed, 8 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> index 3fc77776d89b..70178a77d0c8 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> @@ -3215,13 +3215,17 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
> {
> struct drm_i915_private *dev_priv = to_i915(dev);
> struct i915_ggtt *ggtt = &dev_priv->ggtt;
> + int (*probe)(struct i915_ggtt *ggtt);
> int ret;
>
> + ggtt->base.dev = dev;
> + ggtt->base.is_ggtt = true;
> +
> if (INTEL_INFO(dev)->gen <= 5) {
> - ggtt->probe = i915_gmch_probe;
> + probe = i915_gmch_probe;
> ggtt->base.cleanup = i915_gmch_remove;
> } else if (INTEL_INFO(dev)->gen < 8) {
> - ggtt->probe = gen6_gmch_probe;
> + probe = gen6_gmch_probe;
> ggtt->base.cleanup = gen6_gmch_remove;
>
> if (HAS_EDRAM(dev))
> @@ -3235,14 +3239,11 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
> else
> ggtt->base.pte_encode = snb_pte_encode;
> } else {
> - ggtt->probe = gen8_gmch_probe;
> + probe = gen8_gmch_probe;
> ggtt->base.cleanup = gen6_gmch_remove;
> }
>
I'm pretty sure I already saw a version of this with no function
pointer at all? I preferred that version.
Regards, Joonas
> - ggtt->base.dev = dev;
> - ggtt->base.is_ggtt = true;
> -
> - ret = ggtt->probe(ggtt);
> + ret = probe(ggtt);
> if (ret)
> return ret;
>
> diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
> index 5b6744a5c944..defb274f3ba5 100644
> --- a/drivers/gpu/drm/i915/i915_gem_gtt.h
> +++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
> @@ -365,8 +365,6 @@ struct i915_ggtt {
> bool do_idle_maps;
>
> int mtrr;
> -
> - int (*probe)(struct i915_ggtt *ggtt);
> };
>
> struct i915_hw_ppgtt {
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* Re: [PATCH 04/28] drm/i915: Update GGTT initialisation functions to take drm_i915_private
2016-08-03 13:42 ` [PATCH 04/28] " Chris Wilson
@ 2016-08-03 14:05 ` Joonas Lahtinen
0 siblings, 0 replies; 41+ messages in thread
From: Joonas Lahtinen @ 2016-08-03 14:05 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On ke, 2016-08-03 at 14:42 +0100, Chris Wilson wrote:
> Since these are internal functions they operate on drm_i915_private and
> not the drm_device being passed in. So pass in the drm_i915_private
> instead, and remove one layer of dancing. No space wins here, just
> conforming to the norm in function parameters.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Pretty mechanical, so unless you hid an easter egg;
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Regards, Joonas
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* Re: [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local
2016-08-03 14:00 ` Joonas Lahtinen
@ 2016-08-03 14:07 ` Chris Wilson
2016-08-03 17:14 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Chris Wilson
1 sibling, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 14:07 UTC (permalink / raw)
To: Joonas Lahtinen; +Cc: intel-gfx
On Wed, Aug 03, 2016 at 05:00:01PM +0300, Joonas Lahtinen wrote:
> On ke, 2016-08-03 at 14:42 +0100, Chris Wilson wrote:
> > The ggtt->probe() vfunc is only called once during GGTT initialisation.
> > We can keep it in a local for the function, sparing a few bytes in our
> > i915_ggtt persistent struct.
> >
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > ---
> > drivers/gpu/drm/i915/i915_gem_gtt.c | 15 ++++++++-------
> > drivers/gpu/drm/i915/i915_gem_gtt.h | 2 --
> > 2 files changed, 8 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
> > index 3fc77776d89b..70178a77d0c8 100644
> > --- a/drivers/gpu/drm/i915/i915_gem_gtt.c
> > +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
> > @@ -3215,13 +3215,17 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
> > {
> > struct drm_i915_private *dev_priv = to_i915(dev);
> > struct i915_ggtt *ggtt = &dev_priv->ggtt;
> > + int (*probe)(struct i915_ggtt *ggtt);
> > int ret;
> >
> > + ggtt->base.dev = dev;
> > + ggtt->base.is_ggtt = true;
> > +
> > if (INTEL_INFO(dev)->gen <= 5) {
> > - ggtt->probe = i915_gmch_probe;
> > + probe = i915_gmch_probe;
> > ggtt->base.cleanup = i915_gmch_remove;
> > } else if (INTEL_INFO(dev)->gen < 8) {
> > - ggtt->probe = gen6_gmch_probe;
> > + probe = gen6_gmch_probe;
> > ggtt->base.cleanup = gen6_gmch_remove;
> >
> > if (HAS_EDRAM(dev))
> > @@ -3235,14 +3239,11 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
> > else
> > ggtt->base.pte_encode = snb_pte_encode;
> > } else {
> > - ggtt->probe = gen8_gmch_probe;
> > + probe = gen8_gmch_probe;
> > ggtt->base.cleanup = gen6_gmch_remove;
> > }
> >
>
> I'm pretty sure I already saw a version of this with no function
> pointer at all? I preferred that version.
Yes, I do have a patch to change this around to avoid the vfunc, I was
just doing a little drive-by whilst in the vicinity. It's a bigger
patch though - but it's already rebased on top of this, so I'll send it
along.
-Chris
--
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (27 preceding siblings ...)
2016-08-03 13:43 ` [PATCH 28/28] Revert "drm/i915: Clean up associated VMAs on context destruction" Chris Wilson
@ 2016-08-03 14:15 ` Patchwork
2016-08-04 5:19 ` ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers (rev4) Patchwork
29 siblings, 0 replies; 41+ messages in thread
From: Patchwork @ 2016-08-03 14:15 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers
URL : https://patchwork.freedesktop.org/series/10594/
State : failure
== Summary ==
Series 10594v1 Series without cover letter
http://patchwork.freedesktop.org/api/1.0/series/10594/revisions/1/mbox
Test drv_module_reload_basic:
pass -> SKIP (ro-hsw-i3-4010u)
Test gem_exec_gttfill:
Subgroup basic:
pass -> SKIP (fi-snb-i7-2600)
Test kms_cursor_legacy:
Subgroup basic-cursor-vs-flip-varying-size:
pass -> FAIL (ro-ilk1-i5-650)
Subgroup basic-flip-vs-cursor-legacy:
pass -> FAIL (ro-bdw-i5-5250u)
Subgroup basic-flip-vs-cursor-varying-size:
fail -> PASS (ro-snb-i7-2620M)
fail -> PASS (ro-bdw-i5-5250u)
Test kms_flip:
Subgroup basic-flip-vs-wf_vblank:
pass -> FAIL (ro-byt-n2820)
Test kms_pipe_crc_basic:
Subgroup suspend-read-crc-pipe-a:
dmesg-warn -> PASS (ro-bdw-i7-5557U)
fi-hsw-i7-4770k total:240 pass:218 dwarn:0 dfail:0 fail:0 skip:22
fi-skl-i5-6260u total:240 pass:224 dwarn:0 dfail:0 fail:2 skip:14
fi-skl-i7-6700k total:240 pass:208 dwarn:0 dfail:0 fail:4 skip:28
fi-snb-i7-2600 total:240 pass:197 dwarn:0 dfail:0 fail:0 skip:43
ro-bdw-i5-5250u total:240 pass:219 dwarn:4 dfail:0 fail:1 skip:16
ro-bdw-i7-5557U total:240 pass:224 dwarn:0 dfail:0 fail:0 skip:16
ro-bdw-i7-5600u total:240 pass:206 dwarn:0 dfail:0 fail:2 skip:32
ro-byt-n2820 total:240 pass:196 dwarn:0 dfail:0 fail:4 skip:40
ro-hsw-i3-4010u total:240 pass:213 dwarn:0 dfail:0 fail:0 skip:27
ro-hsw-i7-4770r total:240 pass:214 dwarn:0 dfail:0 fail:0 skip:26
ro-ilk-i7-620lm total:240 pass:173 dwarn:1 dfail:0 fail:1 skip:65
ro-ilk1-i5-650 total:235 pass:173 dwarn:0 dfail:0 fail:2 skip:60
ro-ivb-i7-3770 total:240 pass:205 dwarn:0 dfail:0 fail:0 skip:35
ro-ivb2-i7-3770 total:240 pass:209 dwarn:0 dfail:0 fail:0 skip:31
ro-skl3-i5-6260u total:240 pass:222 dwarn:0 dfail:0 fail:4 skip:14
ro-snb-i7-2620M total:240 pass:198 dwarn:0 dfail:0 fail:1 skip:41
fi-kbl-qkkr failed to connect after reboot
ro-bsw-n3050 failed to connect after reboot
Results at /archive/results/CI_IGT_test/RO_Patchwork_1679/
0bee311 drm-intel-nightly: 2016y-08m-03d-13h-02m-57s UTC integration manifest
1fd4595 Revert "drm/i915: Clean up associated VMAs on context destruction"
cd9bda0 drm/i915: Mark the context and address space as closed
4aac19d drm/i915: Release vma when the handle is closed
0557dce drm/i915: Track active vma requests
f3d2ab5 drm/i915: i915_vma_move_to_active prep patch
a357d9b drm/i915: Move request list retirement to i915_gem_request.c
5c957ca drm/i915: Double check activity before relocations
71dc45f drm/i915: s/__i915_wait_request/i915_wait_request/
d720bb7 drm/i915: Disable waitboosting for a saturated engine
e90b1a1 drm/i915: Move the special case wait-request handling to its one caller
c2d7ec3 drm/i915: Convert intel_overlay to request tracking
6119520 drm/i915: Track requests inside each intel_ring
f99267e drm/i915: Refactor activity tracking for requests
98fe9d6 drm/i915: Remove obsolete i915_gem_object_flush_active()
b95b256 drm/i915: Rename request->list to link for consistency
6eec785 drm/i915: Refactor blocking waits
3d3f739 drm/i915: Mark up i915_gem_active for locking annotation
c48ac7c drm/i915: Prepare i915_gem_active for annotations
05740f4 drm/i915: Introduce i915_gem_active for request tracking
59e4dcc drm/i915: Kill drop_pages()
f0061a4 drm/i915: Be more careful when unbinding vma
d618a2d drm/i915: Count how many VMA are bound for an object
3e69021 drm/i915: Store owning file on the i915_address_space
c487558 drm/i915: Split early global GTT initialisation
6521da8 drm/i915: Update GGTT initialisation functions to take drm_i915_private
ac5afb2 drm/i915: Keep ggtt->probe() as a local
447110c drm/i915: Split GGTT initialisation between probing and setup
f2959e3 drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* Re: [PATCH 05/28] drm/i915: Split early global GTT initialisation
2016-08-03 13:42 ` [PATCH 05/28] drm/i915: Split early global GTT initialisation Chris Wilson
@ 2016-08-03 14:17 ` Joonas Lahtinen
0 siblings, 0 replies; 41+ messages in thread
From: Joonas Lahtinen @ 2016-08-03 14:17 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On ke, 2016-08-03 at 14:42 +0100, Chris Wilson wrote:
> - BUG_ON(mappable_end > end);
This gets lost, with that restored somewhere or explained in commit
message.
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Regards, Joonas
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private
2016-08-03 14:00 ` Joonas Lahtinen
2016-08-03 14:07 ` Chris Wilson
@ 2016-08-03 17:14 ` Chris Wilson
2016-08-03 17:14 ` [PATCH 2/3] drm/i915: Split early global GTT initialisation Chris Wilson
` (2 more replies)
1 sibling, 3 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 17:14 UTC (permalink / raw)
To: intel-gfx
Since these are internal functions they operate on drm_i915_private and
not the drm_device being passed in. So pass in the drm_i915_private
instead, and remove one layer of dancing. No space wins here, just
conforming to the norm in function parameters.
v2: Include all the probe functions
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.c | 12 ++--
drivers/gpu/drm/i915/i915_gem.c | 2 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 114 +++++++++++++++++-------------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 10 ++--
4 files changed, 66 insertions(+), 72 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 8996c2a6cd07..ed52f61d8560 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1004,7 +1004,7 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
intel_sanitize_options(dev_priv);
- ret = i915_ggtt_probe_hw(dev);
+ ret = i915_ggtt_probe_hw(dev_priv);
if (ret)
return ret;
@@ -1022,11 +1022,11 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
goto out_ggtt;
}
- ret = i915_ggtt_init_hw(dev);
+ ret = i915_ggtt_init_hw(dev_priv);
if (ret)
return ret;
- ret = i915_ggtt_enable_hw(dev);
+ ret = i915_ggtt_enable_hw(dev_priv);
if (ret) {
DRM_ERROR("failed to enable GGTT\n");
goto out_ggtt;
@@ -1104,7 +1104,7 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
return 0;
out_ggtt:
- i915_ggtt_cleanup_hw(dev);
+ i915_ggtt_cleanup_hw(dev_priv);
return ret;
}
@@ -1124,7 +1124,7 @@ static void i915_driver_cleanup_hw(struct drm_i915_private *dev_priv)
pm_qos_remove_request(&dev_priv->pm_qos);
arch_phys_wc_del(ggtt->mtrr);
io_mapping_free(ggtt->mappable);
- i915_ggtt_cleanup_hw(dev);
+ i915_ggtt_cleanup_hw(dev_priv);
}
/**
@@ -1570,7 +1570,7 @@ static int i915_drm_resume(struct drm_device *dev)
disable_rpm_wakeref_asserts(dev_priv);
- ret = i915_ggtt_enable_hw(dev);
+ ret = i915_ggtt_enable_hw(dev_priv);
if (ret)
DRM_ERROR("failed to re-enable GGTT\n");
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index bc871449751a..653197811d43 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4543,7 +4543,7 @@ int i915_gem_init(struct drm_device *dev)
intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
i915_gem_init_userptr(dev_priv);
- i915_gem_init_ggtt(dev);
+ i915_gem_init_ggtt(dev_priv);
ret = i915_gem_context_init(dev);
if (ret)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 3fc77776d89b..39b3b36e8dea 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2223,17 +2223,16 @@ void i915_ppgtt_release(struct kref *kref)
kfree(ppgtt);
}
-extern int intel_iommu_gfx_mapped;
/* Certain Gen5 chipsets require require idling the GPU before
* unmapping anything from the GTT when VT-d is enabled.
*/
-static bool needs_idle_maps(struct drm_device *dev)
+static bool needs_idle_maps(struct drm_i915_private *dev_priv)
{
#ifdef CONFIG_INTEL_IOMMU
/* Query intel_iommu to see if we need the workaround. Presumably that
* was loaded first.
*/
- if (IS_GEN5(dev) && IS_MOBILE(dev) && intel_iommu_gfx_mapped)
+ if (IS_GEN5(dev_priv) && IS_MOBILE(dev_priv) && intel_iommu_gfx_mapped)
return true;
#endif
return false;
@@ -2746,7 +2745,7 @@ static void i915_gtt_color_adjust(struct drm_mm_node *node,
*end -= 4096;
}
-static int i915_gem_setup_global_gtt(struct drm_device *dev,
+static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
u64 start,
u64 mappable_end,
u64 end)
@@ -2760,7 +2759,6 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
* aperture. One page should be enough to keep any prefetching inside
* of the aperture.
*/
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
struct drm_mm_node *entry;
struct drm_i915_gem_object *obj;
@@ -2781,7 +2779,7 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
if (ret)
return ret;
- if (!HAS_LLC(dev))
+ if (!HAS_LLC(dev_priv))
ggtt->base.mm.color_adjust = i915_gtt_color_adjust;
/* Mark any preallocated objects as occupied */
@@ -2813,14 +2811,14 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
/* And finally clear the reserved guard page */
ggtt->base.clear_range(&ggtt->base, end - PAGE_SIZE, PAGE_SIZE, true);
- if (USES_PPGTT(dev) && !USES_FULL_PPGTT(dev)) {
+ if (USES_PPGTT(dev_priv) && !USES_FULL_PPGTT(dev_priv)) {
struct i915_hw_ppgtt *ppgtt;
ppgtt = kzalloc(sizeof(*ppgtt), GFP_KERNEL);
if (!ppgtt)
return -ENOMEM;
- ret = __hw_ppgtt_init(dev, ppgtt);
+ ret = __hw_ppgtt_init(&dev_priv->drm, ppgtt);
if (ret) {
ppgtt->base.cleanup(&ppgtt->base);
kfree(ppgtt);
@@ -2851,23 +2849,22 @@ static int i915_gem_setup_global_gtt(struct drm_device *dev,
/**
* i915_gem_init_ggtt - Initialize GEM for Global GTT
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-void i915_gem_init_ggtt(struct drm_device *dev)
+void i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
- i915_gem_setup_global_gtt(dev, 0, ggtt->mappable_end, ggtt->base.total);
+ i915_gem_setup_global_gtt(dev_priv,
+ 0, ggtt->mappable_end, ggtt->base.total);
}
/**
* i915_ggtt_cleanup_hw - Clean up GGTT hardware initialization
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-void i915_ggtt_cleanup_hw(struct drm_device *dev)
+void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
if (dev_priv->mm.aliasing_ppgtt) {
@@ -2876,7 +2873,7 @@ void i915_ggtt_cleanup_hw(struct drm_device *dev)
ppgtt->base.cleanup(&ppgtt->base);
}
- i915_gem_cleanup_stolen(dev);
+ i915_gem_cleanup_stolen(&dev_priv->drm);
if (drm_mm_initialized(&ggtt->base.mm)) {
intel_vgt_deballoon(dev_priv);
@@ -2966,17 +2963,16 @@ static size_t gen9_get_stolen_size(u16 gen9_gmch_ctl)
return (gen9_gmch_ctl - 0xf0 + 1) << 22;
}
-static int ggtt_probe_common(struct drm_device *dev,
- size_t gtt_size)
+static int ggtt_probe_common(struct drm_i915_private *dev_priv, size_t gtt_size)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
+ struct pci_dev *pdev = dev_priv->drm.pdev;
struct i915_page_scratch *scratch_page;
phys_addr_t ggtt_phys_addr;
/* For Modern GENs the PTEs and register space are split in the BAR */
- ggtt_phys_addr = pci_resource_start(dev->pdev, 0) +
- (pci_resource_len(dev->pdev, 0) / 2);
+ ggtt_phys_addr = pci_resource_start(pdev, 0) +
+ (pci_resource_len(pdev, 0) / 2);
/*
* On BXT writes larger than 64 bit to the GTT pagetable range will be
@@ -2985,7 +2981,7 @@ static int ggtt_probe_common(struct drm_device *dev,
* resort to an uncached mapping. The WC issue is easily caught by the
* readback check when writing GTT PTE entries.
*/
- if (IS_BROXTON(dev))
+ if (IS_BROXTON(dev_priv))
ggtt->gsm = ioremap_nocache(ggtt_phys_addr, gtt_size);
else
ggtt->gsm = ioremap_wc(ggtt_phys_addr, gtt_size);
@@ -2994,7 +2990,7 @@ static int ggtt_probe_common(struct drm_device *dev,
return -ENOMEM;
}
- scratch_page = alloc_scratch_page(dev);
+ scratch_page = alloc_scratch_page(&dev_priv->drm);
if (IS_ERR(scratch_page)) {
DRM_ERROR("Scratch setup failed\n");
/* iounmap will also get called at remove, but meh */
@@ -3082,24 +3078,24 @@ static void chv_setup_private_ppat(struct drm_i915_private *dev_priv)
static int gen8_gmch_probe(struct i915_ggtt *ggtt)
{
- struct drm_device *dev = ggtt->base.dev;
- struct drm_i915_private *dev_priv = to_i915(dev);
+ struct drm_i915_private *dev_priv = to_i915(ggtt->base.dev);
+ struct pci_dev *pdev = dev_priv->drm.pdev;
u16 snb_gmch_ctl;
int ret;
/* TODO: We're not aware of mappable constraints on gen8 yet */
- ggtt->mappable_base = pci_resource_start(dev->pdev, 2);
- ggtt->mappable_end = pci_resource_len(dev->pdev, 2);
+ ggtt->mappable_base = pci_resource_start(pdev, 2);
+ ggtt->mappable_end = pci_resource_len(pdev, 2);
- if (!pci_set_dma_mask(dev->pdev, DMA_BIT_MASK(39)))
- pci_set_consistent_dma_mask(dev->pdev, DMA_BIT_MASK(39));
+ if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(39)))
+ pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(39));
- pci_read_config_word(dev->pdev, SNB_GMCH_CTRL, &snb_gmch_ctl);
+ pci_read_config_word(pdev, SNB_GMCH_CTRL, &snb_gmch_ctl);
- if (INTEL_INFO(dev)->gen >= 9) {
+ if (INTEL_GEN(dev_priv) >= 9) {
ggtt->stolen_size = gen9_get_stolen_size(snb_gmch_ctl);
ggtt->size = gen8_get_total_gtt_size(snb_gmch_ctl);
- } else if (IS_CHERRYVIEW(dev)) {
+ } else if (IS_CHERRYVIEW(dev_priv)) {
ggtt->stolen_size = chv_get_stolen_size(snb_gmch_ctl);
ggtt->size = chv_get_total_gtt_size(snb_gmch_ctl);
} else {
@@ -3109,12 +3105,12 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt)
ggtt->base.total = (ggtt->size / sizeof(gen8_pte_t)) << PAGE_SHIFT;
- if (IS_CHERRYVIEW(dev) || IS_BROXTON(dev))
+ if (IS_CHERRYVIEW(dev_priv) || IS_BROXTON(dev_priv))
chv_setup_private_ppat(dev_priv);
else
bdw_setup_private_ppat(dev_priv);
- ret = ggtt_probe_common(dev, ggtt->size);
+ ret = ggtt_probe_common(dev_priv, ggtt->size);
ggtt->base.bind_vma = ggtt_bind_vma;
ggtt->base.unbind_vma = ggtt_unbind_vma;
@@ -3132,12 +3128,13 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt)
static int gen6_gmch_probe(struct i915_ggtt *ggtt)
{
- struct drm_device *dev = ggtt->base.dev;
+ struct drm_i915_private *dev_priv = to_i915(ggtt->base.dev);
+ struct pci_dev *pdev = dev_priv->drm.pdev;
u16 snb_gmch_ctl;
int ret;
- ggtt->mappable_base = pci_resource_start(dev->pdev, 2);
- ggtt->mappable_end = pci_resource_len(dev->pdev, 2);
+ ggtt->mappable_base = pci_resource_start(pdev, 2);
+ ggtt->mappable_end = pci_resource_len(pdev, 2);
/* 64/512MB is the current min/max we actually know of, but this is just
* a coarse sanity check.
@@ -3147,15 +3144,15 @@ static int gen6_gmch_probe(struct i915_ggtt *ggtt)
return -ENXIO;
}
- if (!pci_set_dma_mask(dev->pdev, DMA_BIT_MASK(40)))
- pci_set_consistent_dma_mask(dev->pdev, DMA_BIT_MASK(40));
- pci_read_config_word(dev->pdev, SNB_GMCH_CTRL, &snb_gmch_ctl);
+ if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(40)))
+ pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(40));
+ pci_read_config_word(pdev, SNB_GMCH_CTRL, &snb_gmch_ctl);
ggtt->stolen_size = gen6_get_stolen_size(snb_gmch_ctl);
ggtt->size = gen6_get_total_gtt_size(snb_gmch_ctl);
ggtt->base.total = (ggtt->size / sizeof(gen6_pte_t)) << PAGE_SHIFT;
- ret = ggtt_probe_common(dev, ggtt->size);
+ ret = ggtt_probe_common(dev_priv, ggtt->size);
ggtt->base.clear_range = gen6_ggtt_clear_range;
ggtt->base.insert_page = gen6_ggtt_insert_page;
@@ -3176,8 +3173,7 @@ static void gen6_gmch_remove(struct i915_address_space *vm)
static int i915_gmch_probe(struct i915_ggtt *ggtt)
{
- struct drm_device *dev = ggtt->base.dev;
- struct drm_i915_private *dev_priv = to_i915(dev);
+ struct drm_i915_private *dev_priv = to_i915(ggtt->base.dev);
int ret;
ret = intel_gmch_probe(dev_priv->bridge_dev, dev_priv->drm.pdev, NULL);
@@ -3189,7 +3185,7 @@ static int i915_gmch_probe(struct i915_ggtt *ggtt)
intel_gtt_get(&ggtt->base.total, &ggtt->stolen_size,
&ggtt->mappable_base, &ggtt->mappable_end);
- ggtt->do_idle_maps = needs_idle_maps(&dev_priv->drm);
+ ggtt->do_idle_maps = needs_idle_maps(dev_priv);
ggtt->base.insert_page = i915_ggtt_insert_page;
ggtt->base.insert_entries = i915_ggtt_insert_entries;
ggtt->base.clear_range = i915_ggtt_clear_range;
@@ -3209,28 +3205,27 @@ static void i915_gmch_remove(struct i915_address_space *vm)
/**
* i915_ggtt_probe_hw - Probe GGTT hardware location
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-int i915_ggtt_probe_hw(struct drm_device *dev)
+int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int ret;
- if (INTEL_INFO(dev)->gen <= 5) {
+ if (INTEL_GEN(dev_priv) <= 5) {
ggtt->probe = i915_gmch_probe;
ggtt->base.cleanup = i915_gmch_remove;
- } else if (INTEL_INFO(dev)->gen < 8) {
+ } else if (INTEL_GEN(dev_priv) < 8) {
ggtt->probe = gen6_gmch_probe;
ggtt->base.cleanup = gen6_gmch_remove;
- if (HAS_EDRAM(dev))
+ if (HAS_EDRAM(dev_priv))
ggtt->base.pte_encode = iris_pte_encode;
- else if (IS_HASWELL(dev))
+ else if (IS_HASWELL(dev_priv))
ggtt->base.pte_encode = hsw_pte_encode;
- else if (IS_VALLEYVIEW(dev))
+ else if (IS_VALLEYVIEW(dev_priv))
ggtt->base.pte_encode = byt_pte_encode;
- else if (INTEL_INFO(dev)->gen >= 7)
+ else if (INTEL_GEN(dev_priv) >= 7)
ggtt->base.pte_encode = ivb_pte_encode;
else
ggtt->base.pte_encode = snb_pte_encode;
@@ -3239,7 +3234,7 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
ggtt->base.cleanup = gen6_gmch_remove;
}
- ggtt->base.dev = dev;
+ ggtt->base.dev = &dev_priv->drm;
ggtt->base.is_ggtt = true;
ret = ggtt->probe(ggtt);
@@ -3269,11 +3264,10 @@ int i915_ggtt_probe_hw(struct drm_device *dev)
/**
* i915_ggtt_init_hw - Initialize GGTT hardware
- * @dev: DRM device
+ * @dev_priv: i915 device
*/
-int i915_ggtt_init_hw(struct drm_device *dev)
+int i915_ggtt_init_hw(struct drm_i915_private *dev_priv)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int ret;
@@ -3281,7 +3275,7 @@ int i915_ggtt_init_hw(struct drm_device *dev)
* Initialise stolen early so that we may reserve preallocated
* objects for the BIOS to KMS transition.
*/
- ret = i915_gem_init_stolen(dev);
+ ret = i915_gem_init_stolen(&dev_priv->drm);
if (ret)
goto out_gtt_cleanup;
@@ -3292,9 +3286,9 @@ out_gtt_cleanup:
return ret;
}
-int i915_ggtt_enable_hw(struct drm_device *dev)
+int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv)
{
- if (INTEL_INFO(dev)->gen < 6 && !intel_enable_gtt())
+ if (INTEL_GEN(dev_priv) < 6 && !intel_enable_gtt())
return -EIO;
return 0;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index 5b6744a5c944..f85a73fa5477 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -521,11 +521,11 @@ i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
px_dma(ppgtt->base.scratch_pd);
}
-int i915_ggtt_probe_hw(struct drm_device *dev);
-int i915_ggtt_init_hw(struct drm_device *dev);
-int i915_ggtt_enable_hw(struct drm_device *dev);
-void i915_gem_init_ggtt(struct drm_device *dev);
-void i915_ggtt_cleanup_hw(struct drm_device *dev);
+int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
+int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
+int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
+void i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
+void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
int i915_ppgtt_init_hw(struct drm_device *dev);
void i915_ppgtt_release(struct kref *kref);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 2/3] drm/i915: Split early global GTT initialisation
2016-08-03 17:14 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Chris Wilson
@ 2016-08-03 17:14 ` Chris Wilson
2016-08-03 17:14 ` [PATCH 3/3] drm/i915: Rearrange GGTT probing to avoid needing a vfunc Chris Wilson
2016-08-04 6:18 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Joonas Lahtinen
2 siblings, 0 replies; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 17:14 UTC (permalink / raw)
To: intel-gfx
Initialising the global GTT is tricky as we wish to use the drm_mm range
manager during the modesetting initialisation (to capture stolen
allocations from the BIOS) before we actually enable GEM. To overcome
this, we currently setup the drm_mm first and then carefully rebind
them.
v2: Fixup after rebasing
v3: GGTT initialisation needs to be split around kicking out conflicts
v4: Restore an old UMS BUG_ON(mappable > total) as a DRM_ERROR plus
fixup of probe results.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_drv.c | 19 --------
drivers/gpu/drm/i915/i915_gem.c | 6 ++-
drivers/gpu/drm/i915/i915_gem_gtt.c | 87 ++++++++++++++--------------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 2 +-
drivers/gpu/drm/i915/i915_gem_stolen.c | 17 ++++---
5 files changed, 49 insertions(+), 82 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index ed52f61d8560..50c564089fa8 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -993,8 +993,6 @@ static void intel_sanitize_options(struct drm_i915_private *dev_priv)
static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
{
struct drm_device *dev = &dev_priv->drm;
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
- uint32_t aperture_size;
int ret;
if (i915_inject_load_failure())
@@ -1044,7 +1042,6 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
}
}
-
/* 965GM sometimes incorrectly writes to hardware status page (HWS)
* using 32bit addressing, overwriting memory if HWS is located
* above 4GB.
@@ -1063,19 +1060,6 @@ static int i915_driver_init_hw(struct drm_i915_private *dev_priv)
}
}
- aperture_size = ggtt->mappable_end;
-
- ggtt->mappable =
- io_mapping_create_wc(ggtt->mappable_base,
- aperture_size);
- if (!ggtt->mappable) {
- ret = -EIO;
- goto out_ggtt;
- }
-
- ggtt->mtrr = arch_phys_wc_add(ggtt->mappable_base,
- aperture_size);
-
pm_qos_add_request(&dev_priv->pm_qos, PM_QOS_CPU_DMA_LATENCY,
PM_QOS_DEFAULT_VALUE);
@@ -1116,14 +1100,11 @@ out_ggtt:
static void i915_driver_cleanup_hw(struct drm_i915_private *dev_priv)
{
struct drm_device *dev = &dev_priv->drm;
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
if (dev->pdev->msi_enabled)
pci_disable_msi(dev->pdev);
pm_qos_remove_request(&dev_priv->pm_qos);
- arch_phys_wc_del(ggtt->mtrr);
- io_mapping_free(ggtt->mappable);
i915_ggtt_cleanup_hw(dev_priv);
}
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 653197811d43..83e299c429c2 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -4543,7 +4543,10 @@ int i915_gem_init(struct drm_device *dev)
intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
i915_gem_init_userptr(dev_priv);
- i915_gem_init_ggtt(dev_priv);
+
+ ret = i915_gem_init_ggtt(dev_priv);
+ if (ret)
+ goto out_unlock;
ret = i915_gem_context_init(dev);
if (ret)
@@ -4634,7 +4637,6 @@ i915_gem_load_init(struct drm_device *dev)
SLAB_HWCACHE_ALIGN,
NULL);
- INIT_LIST_HEAD(&dev_priv->vm_list);
INIT_LIST_HEAD(&dev_priv->context_list);
INIT_LIST_HEAD(&dev_priv->mm.unbound_list);
INIT_LIST_HEAD(&dev_priv->mm.bound_list);
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 39b3b36e8dea..72231ea4a802 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2745,10 +2745,7 @@ static void i915_gtt_color_adjust(struct drm_mm_node *node,
*end -= 4096;
}
-static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
- u64 start,
- u64 mappable_end,
- u64 end)
+int i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
{
/* Let GEM Manage all of the aperture.
*
@@ -2760,46 +2757,14 @@ static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
* of the aperture.
*/
struct i915_ggtt *ggtt = &dev_priv->ggtt;
- struct drm_mm_node *entry;
- struct drm_i915_gem_object *obj;
unsigned long hole_start, hole_end;
+ struct drm_mm_node *entry;
int ret;
- BUG_ON(mappable_end > end);
-
- ggtt->base.start = start;
-
- /* Subtract the guard page before address space initialization to
- * shrink the range used by drm_mm */
- ggtt->base.total = end - start - PAGE_SIZE;
- i915_address_space_init(&ggtt->base, dev_priv);
- ggtt->base.total += PAGE_SIZE;
-
ret = intel_vgt_balloon(dev_priv);
if (ret)
return ret;
- if (!HAS_LLC(dev_priv))
- ggtt->base.mm.color_adjust = i915_gtt_color_adjust;
-
- /* Mark any preallocated objects as occupied */
- list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) {
- struct i915_vma *vma = i915_gem_obj_to_vma(obj, &ggtt->base);
-
- DRM_DEBUG_KMS("reserving preallocated space: %llx + %zx\n",
- i915_gem_obj_ggtt_offset(obj), obj->base.size);
-
- WARN_ON(i915_gem_obj_ggtt_bound(obj));
- ret = drm_mm_reserve_node(&ggtt->base.mm, &vma->node);
- if (ret) {
- DRM_DEBUG_KMS("Reservation failed: %i\n", ret);
- return ret;
- }
- vma->bound |= GLOBAL_BIND;
- __i915_vma_set_map_and_fenceable(vma);
- list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
- }
-
/* Clear any non-preallocated blocks */
drm_mm_for_each_hole(entry, &ggtt->base.mm, hole_start, hole_end) {
DRM_DEBUG_KMS("clearing unused GTT space: [%lx, %lx]\n",
@@ -2809,7 +2774,9 @@ static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
}
/* And finally clear the reserved guard page */
- ggtt->base.clear_range(&ggtt->base, end - PAGE_SIZE, PAGE_SIZE, true);
+ ggtt->base.clear_range(&ggtt->base,
+ ggtt->base.total - PAGE_SIZE, PAGE_SIZE,
+ true);
if (USES_PPGTT(dev_priv) && !USES_FULL_PPGTT(dev_priv)) {
struct i915_hw_ppgtt *ppgtt;
@@ -2848,18 +2815,6 @@ static int i915_gem_setup_global_gtt(struct drm_i915_private *dev_priv,
}
/**
- * i915_gem_init_ggtt - Initialize GEM for Global GTT
- * @dev_priv: i915 device
- */
-void i915_gem_init_ggtt(struct drm_i915_private *dev_priv)
-{
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
-
- i915_gem_setup_global_gtt(dev_priv,
- 0, ggtt->mappable_end, ggtt->base.total);
-}
-
-/**
* i915_ggtt_cleanup_hw - Clean up GGTT hardware initialization
* @dev_priv: i915 device
*/
@@ -2883,6 +2838,9 @@ void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv)
}
ggtt->base.cleanup(&ggtt->base);
+
+ arch_phys_wc_del(ggtt->mtrr);
+ io_mapping_free(ggtt->mappable);
}
static unsigned int gen6_get_total_gtt_size(u16 snb_gmch_ctl)
@@ -3243,12 +3201,19 @@ int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv)
if ((ggtt->base.total - 1) >> 32) {
DRM_ERROR("We never expected a Global GTT with more than 32bits"
- "of address space! Found %lldM!\n",
+ " of address space! Found %lldM!\n",
ggtt->base.total >> 20);
ggtt->base.total = 1ULL << 32;
ggtt->mappable_end = min(ggtt->mappable_end, ggtt->base.total);
}
+ if (ggtt->mappable_end > ggtt->base.total) {
+ DRM_ERROR("mappable aperture extends past end of GGTT,"
+ " aperture=%llx, total=%llx\n",
+ ggtt->mappable_end, ggtt->base.total);
+ ggtt->mappable_end = ggtt->base.total;
+ }
+
/* GMADR is the PCI mmio aperture into the global GTT. */
DRM_INFO("Memory usable by graphics device = %lluM\n",
ggtt->base.total >> 20);
@@ -3271,6 +3236,26 @@ int i915_ggtt_init_hw(struct drm_i915_private *dev_priv)
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int ret;
+ INIT_LIST_HEAD(&dev_priv->vm_list);
+
+ /* Subtract the guard page before address space initialization to
+ * shrink the range used by drm_mm.
+ */
+ ggtt->base.total -= PAGE_SIZE;
+ i915_address_space_init(&ggtt->base, dev_priv);
+ ggtt->base.total += PAGE_SIZE;
+ if (!HAS_LLC(dev_priv))
+ ggtt->base.mm.color_adjust = i915_gtt_color_adjust;
+
+ ggtt->mappable =
+ io_mapping_create_wc(ggtt->mappable_base, ggtt->mappable_end);
+ if (!ggtt->mappable) {
+ ret = -EIO;
+ goto out_gtt_cleanup;
+ }
+
+ ggtt->mtrr = arch_phys_wc_add(ggtt->mappable_base, ggtt->mappable_end);
+
/*
* Initialise stolen early so that we may reserve preallocated
* objects for the BIOS to KMS transition.
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index f85a73fa5477..bb399930a51a 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -524,7 +524,7 @@ i915_page_dir_dma_addr(const struct i915_hw_ppgtt *ppgtt, const unsigned n)
int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv);
int i915_ggtt_init_hw(struct drm_i915_private *dev_priv);
int i915_ggtt_enable_hw(struct drm_i915_private *dev_priv);
-void i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
+int i915_gem_init_ggtt(struct drm_i915_private *dev_priv);
void i915_ggtt_cleanup_hw(struct drm_i915_private *dev_priv);
int i915_ppgtt_init_hw(struct drm_device *dev);
diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c
index 310756c30723..9a8cc8c51077 100644
--- a/drivers/gpu/drm/i915/i915_gem_stolen.c
+++ b/drivers/gpu/drm/i915/i915_gem_stolen.c
@@ -698,18 +698,17 @@ i915_gem_object_create_stolen_for_preallocated(struct drm_device *dev,
*/
vma->node.start = gtt_offset;
vma->node.size = size;
- if (drm_mm_initialized(&ggtt->base.mm)) {
- ret = drm_mm_reserve_node(&ggtt->base.mm, &vma->node);
- if (ret) {
- DRM_DEBUG_KMS("failed to allocate stolen GTT space\n");
- goto err;
- }
- vma->bound |= GLOBAL_BIND;
- __i915_vma_set_map_and_fenceable(vma);
- list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
+ ret = drm_mm_reserve_node(&ggtt->base.mm, &vma->node);
+ if (ret) {
+ DRM_DEBUG_KMS("failed to allocate stolen GTT space\n");
+ goto err;
}
+ vma->bound |= GLOBAL_BIND;
+ __i915_vma_set_map_and_fenceable(vma);
+ list_add_tail(&vma->vm_link, &ggtt->base.inactive_list);
+
list_add_tail(&obj->global_list, &dev_priv->mm.bound_list);
i915_gem_object_pin_pages(obj);
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* [PATCH 3/3] drm/i915: Rearrange GGTT probing to avoid needing a vfunc
2016-08-03 17:14 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Chris Wilson
2016-08-03 17:14 ` [PATCH 2/3] drm/i915: Split early global GTT initialisation Chris Wilson
@ 2016-08-03 17:14 ` Chris Wilson
2016-08-04 6:34 ` Joonas Lahtinen
2016-08-04 6:18 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Joonas Lahtinen
2 siblings, 1 reply; 41+ messages in thread
From: Chris Wilson @ 2016-08-03 17:14 UTC (permalink / raw)
To: intel-gfx
Since we have a static if-else-chain for device probing of the global
GTT, we do not need to use a function pointer, let alone store it when
we never use it again. So use the if-else-chain to call down into the
device specific probe.
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
---
drivers/gpu/drm/i915/i915_gem_gtt.c | 109 +++++++++++++++++-------------------
drivers/gpu/drm/i915/i915_gem_gtt.h | 3 -
2 files changed, 50 insertions(+), 62 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 72231ea4a802..12522aa4df12 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -2118,7 +2118,6 @@ static void i915_address_space_init(struct i915_address_space *vm,
struct drm_i915_private *dev_priv)
{
drm_mm_init(&vm->mm, vm->start, vm->total);
- vm->dev = &dev_priv->drm;
INIT_LIST_HEAD(&vm->active_list);
INIT_LIST_HEAD(&vm->inactive_list);
list_add_tail(&vm->global_link, &dev_priv->vm_list);
@@ -2921,16 +2920,14 @@ static size_t gen9_get_stolen_size(u16 gen9_gmch_ctl)
return (gen9_gmch_ctl - 0xf0 + 1) << 22;
}
-static int ggtt_probe_common(struct drm_i915_private *dev_priv, size_t gtt_size)
+static int ggtt_probe_common(struct i915_ggtt *ggtt, u64 size)
{
- struct i915_ggtt *ggtt = &dev_priv->ggtt;
- struct pci_dev *pdev = dev_priv->drm.pdev;
+ struct pci_dev *pdev = ggtt->base.dev->pdev;
struct i915_page_scratch *scratch_page;
- phys_addr_t ggtt_phys_addr;
+ phys_addr_t phys_addr;
/* For Modern GENs the PTEs and register space are split in the BAR */
- ggtt_phys_addr = pci_resource_start(pdev, 0) +
- (pci_resource_len(pdev, 0) / 2);
+ phys_addr = pci_resource_start(pdev, 0) + pci_resource_len(pdev, 0) / 2;
/*
* On BXT writes larger than 64 bit to the GTT pagetable range will be
@@ -2939,16 +2936,16 @@ static int ggtt_probe_common(struct drm_i915_private *dev_priv, size_t gtt_size)
* resort to an uncached mapping. The WC issue is easily caught by the
* readback check when writing GTT PTE entries.
*/
- if (IS_BROXTON(dev_priv))
- ggtt->gsm = ioremap_nocache(ggtt_phys_addr, gtt_size);
+ if (IS_BROXTON(ggtt->base.dev))
+ ggtt->gsm = ioremap_nocache(phys_addr, size);
else
- ggtt->gsm = ioremap_wc(ggtt_phys_addr, gtt_size);
+ ggtt->gsm = ioremap_wc(phys_addr, size);
if (!ggtt->gsm) {
- DRM_ERROR("Failed to map the gtt page table\n");
+ DRM_ERROR("Failed to map the ggtt page table\n");
return -ENOMEM;
}
- scratch_page = alloc_scratch_page(&dev_priv->drm);
+ scratch_page = alloc_scratch_page(ggtt->base.dev);
if (IS_ERR(scratch_page)) {
DRM_ERROR("Scratch setup failed\n");
/* iounmap will also get called at remove, but meh */
@@ -3034,12 +3031,20 @@ static void chv_setup_private_ppat(struct drm_i915_private *dev_priv)
I915_WRITE(GEN8_PRIVATE_PAT_HI, pat >> 32);
}
+static void gen6_gmch_remove(struct i915_address_space *vm)
+{
+ struct i915_ggtt *ggtt = container_of(vm, struct i915_ggtt, base);
+
+ iounmap(ggtt->gsm);
+ free_scratch_page(vm->dev, vm->scratch_page);
+}
+
static int gen8_gmch_probe(struct i915_ggtt *ggtt)
{
struct drm_i915_private *dev_priv = to_i915(ggtt->base.dev);
struct pci_dev *pdev = dev_priv->drm.pdev;
+ unsigned int size;
u16 snb_gmch_ctl;
- int ret;
/* TODO: We're not aware of mappable constraints on gen8 yet */
ggtt->mappable_base = pci_resource_start(pdev, 2);
@@ -3052,24 +3057,23 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt)
if (INTEL_GEN(dev_priv) >= 9) {
ggtt->stolen_size = gen9_get_stolen_size(snb_gmch_ctl);
- ggtt->size = gen8_get_total_gtt_size(snb_gmch_ctl);
+ size = gen8_get_total_gtt_size(snb_gmch_ctl);
} else if (IS_CHERRYVIEW(dev_priv)) {
ggtt->stolen_size = chv_get_stolen_size(snb_gmch_ctl);
- ggtt->size = chv_get_total_gtt_size(snb_gmch_ctl);
+ size = chv_get_total_gtt_size(snb_gmch_ctl);
} else {
ggtt->stolen_size = gen8_get_stolen_size(snb_gmch_ctl);
- ggtt->size = gen8_get_total_gtt_size(snb_gmch_ctl);
+ size = gen8_get_total_gtt_size(snb_gmch_ctl);
}
- ggtt->base.total = (ggtt->size / sizeof(gen8_pte_t)) << PAGE_SHIFT;
+ ggtt->base.total = (size / sizeof(gen8_pte_t)) << PAGE_SHIFT;
if (IS_CHERRYVIEW(dev_priv) || IS_BROXTON(dev_priv))
chv_setup_private_ppat(dev_priv);
else
bdw_setup_private_ppat(dev_priv);
- ret = ggtt_probe_common(dev_priv, ggtt->size);
-
+ ggtt->base.cleanup = gen6_gmch_remove;
ggtt->base.bind_vma = ggtt_bind_vma;
ggtt->base.unbind_vma = ggtt_unbind_vma;
ggtt->base.insert_page = gen8_ggtt_insert_page;
@@ -3081,15 +3085,15 @@ static int gen8_gmch_probe(struct i915_ggtt *ggtt)
if (IS_CHERRYVIEW(dev_priv))
ggtt->base.insert_entries = gen8_ggtt_insert_entries__BKL;
- return ret;
+ return ggtt_probe_common(ggtt, size);
}
static int gen6_gmch_probe(struct i915_ggtt *ggtt)
{
struct drm_i915_private *dev_priv = to_i915(ggtt->base.dev);
struct pci_dev *pdev = dev_priv->drm.pdev;
+ unsigned int size;
u16 snb_gmch_ctl;
- int ret;
ggtt->mappable_base = pci_resource_start(pdev, 2);
ggtt->mappable_end = pci_resource_len(pdev, 2);
@@ -3097,7 +3101,7 @@ static int gen6_gmch_probe(struct i915_ggtt *ggtt)
/* 64/512MB is the current min/max we actually know of, but this is just
* a coarse sanity check.
*/
- if ((ggtt->mappable_end < (64<<20) || (ggtt->mappable_end > (512<<20)))) {
+ if (ggtt->mappable_end < (64<<20) || ggtt->mappable_end > (512<<20)) {
DRM_ERROR("Unknown GMADR size (%llx)\n", ggtt->mappable_end);
return -ENXIO;
}
@@ -3107,26 +3111,34 @@ static int gen6_gmch_probe(struct i915_ggtt *ggtt)
pci_read_config_word(pdev, SNB_GMCH_CTRL, &snb_gmch_ctl);
ggtt->stolen_size = gen6_get_stolen_size(snb_gmch_ctl);
- ggtt->size = gen6_get_total_gtt_size(snb_gmch_ctl);
- ggtt->base.total = (ggtt->size / sizeof(gen6_pte_t)) << PAGE_SHIFT;
- ret = ggtt_probe_common(dev_priv, ggtt->size);
+ size = gen6_get_total_gtt_size(snb_gmch_ctl);
+ ggtt->base.total = (size / sizeof(gen6_pte_t)) << PAGE_SHIFT;
ggtt->base.clear_range = gen6_ggtt_clear_range;
ggtt->base.insert_page = gen6_ggtt_insert_page;
ggtt->base.insert_entries = gen6_ggtt_insert_entries;
ggtt->base.bind_vma = ggtt_bind_vma;
ggtt->base.unbind_vma = ggtt_unbind_vma;
+ ggtt->base.cleanup = gen6_gmch_remove;
+
+ if (HAS_EDRAM(dev_priv))
+ ggtt->base.pte_encode = iris_pte_encode;
+ else if (IS_HASWELL(dev_priv))
+ ggtt->base.pte_encode = hsw_pte_encode;
+ else if (IS_VALLEYVIEW(dev_priv))
+ ggtt->base.pte_encode = byt_pte_encode;
+ else if (INTEL_GEN(dev_priv) >= 7)
+ ggtt->base.pte_encode = ivb_pte_encode;
+ else
+ ggtt->base.pte_encode = snb_pte_encode;
- return ret;
+ return ggtt_probe_common(ggtt, size);
}
-static void gen6_gmch_remove(struct i915_address_space *vm)
+static void i915_gmch_remove(struct i915_address_space *vm)
{
- struct i915_ggtt *ggtt = container_of(vm, struct i915_ggtt, base);
-
- iounmap(ggtt->gsm);
- free_scratch_page(vm->dev, vm->scratch_page);
+ intel_gmch_remove();
}
static int i915_gmch_probe(struct i915_ggtt *ggtt)
@@ -3149,6 +3161,7 @@ static int i915_gmch_probe(struct i915_ggtt *ggtt)
ggtt->base.clear_range = i915_ggtt_clear_range;
ggtt->base.bind_vma = ggtt_bind_vma;
ggtt->base.unbind_vma = ggtt_unbind_vma;
+ ggtt->base.cleanup = i915_gmch_remove;
if (unlikely(ggtt->do_idle_maps))
DRM_INFO("applying Ironlake quirks for intel_iommu\n");
@@ -3156,11 +3169,6 @@ static int i915_gmch_probe(struct i915_ggtt *ggtt)
return 0;
}
-static void i915_gmch_remove(struct i915_address_space *vm)
-{
- intel_gmch_remove();
-}
-
/**
* i915_ggtt_probe_hw - Probe GGTT hardware location
* @dev_priv: i915 device
@@ -3170,32 +3178,15 @@ int i915_ggtt_probe_hw(struct drm_i915_private *dev_priv)
struct i915_ggtt *ggtt = &dev_priv->ggtt;
int ret;
- if (INTEL_GEN(dev_priv) <= 5) {
- ggtt->probe = i915_gmch_probe;
- ggtt->base.cleanup = i915_gmch_remove;
- } else if (INTEL_GEN(dev_priv) < 8) {
- ggtt->probe = gen6_gmch_probe;
- ggtt->base.cleanup = gen6_gmch_remove;
-
- if (HAS_EDRAM(dev_priv))
- ggtt->base.pte_encode = iris_pte_encode;
- else if (IS_HASWELL(dev_priv))
- ggtt->base.pte_encode = hsw_pte_encode;
- else if (IS_VALLEYVIEW(dev_priv))
- ggtt->base.pte_encode = byt_pte_encode;
- else if (INTEL_GEN(dev_priv) >= 7)
- ggtt->base.pte_encode = ivb_pte_encode;
- else
- ggtt->base.pte_encode = snb_pte_encode;
- } else {
- ggtt->probe = gen8_gmch_probe;
- ggtt->base.cleanup = gen6_gmch_remove;
- }
-
ggtt->base.dev = &dev_priv->drm;
ggtt->base.is_ggtt = true;
- ret = ggtt->probe(ggtt);
+ if (INTEL_GEN(dev_priv) <= 5)
+ ret = i915_gmch_probe(ggtt);
+ else if (INTEL_GEN(dev_priv) < 8)
+ ret = gen6_gmch_probe(ggtt);
+ else
+ ret = gen8_gmch_probe(ggtt);
if (ret)
return ret;
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
index bb399930a51a..48ce7222cc32 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
@@ -354,7 +354,6 @@ struct i915_ggtt {
size_t stolen_usable_size; /* Total size minus BIOS reserved */
size_t stolen_reserved_base;
size_t stolen_reserved_size;
- size_t size; /* Total size of Global GTT */
u64 mappable_end; /* End offset that we can CPU map */
struct io_mapping *mappable; /* Mapping to our CPU mappable region */
phys_addr_t mappable_base; /* PA of our GMADR */
@@ -365,8 +364,6 @@ struct i915_ggtt {
bool do_idle_maps;
int mtrr;
-
- int (*probe)(struct i915_ggtt *ggtt);
};
struct i915_hw_ppgtt {
--
2.8.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply related [flat|nested] 41+ messages in thread
* ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers (rev4)
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
` (28 preceding siblings ...)
2016-08-03 14:15 ` ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Patchwork
@ 2016-08-04 5:19 ` Patchwork
29 siblings, 0 replies; 41+ messages in thread
From: Patchwork @ 2016-08-04 5:19 UTC (permalink / raw)
To: Chris Wilson; +Cc: intel-gfx
== Series Details ==
Series: series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers (rev4)
URL : https://patchwork.freedesktop.org/series/10594/
State : failure
== Summary ==
Applying: drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers
Applying: drm/i915: Split GGTT initialisation between probing and setup
Applying: drm/i915: Rearrange GGTT probing to avoid needing a vfunc
fatal: sha1 information is lacking or useless (drivers/gpu/drm/i915/i915_gem_gtt.c).
error: could not build fake ancestor
Patch failed at 0003 drm/i915: Rearrange GGTT probing to avoid needing a vfunc
The copy of the patch that failed is found in: .git/rebase-apply/patch
When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* Re: [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private
2016-08-03 17:14 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Chris Wilson
2016-08-03 17:14 ` [PATCH 2/3] drm/i915: Split early global GTT initialisation Chris Wilson
2016-08-03 17:14 ` [PATCH 3/3] drm/i915: Rearrange GGTT probing to avoid needing a vfunc Chris Wilson
@ 2016-08-04 6:18 ` Joonas Lahtinen
2 siblings, 0 replies; 41+ messages in thread
From: Joonas Lahtinen @ 2016-08-04 6:18 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On ke, 2016-08-03 at 18:14 +0100, Chris Wilson wrote:
> Since these are internal functions they operate on drm_i915_private and
> not the drm_device being passed in. So pass in the drm_i915_private
> instead, and remove one layer of dancing. No space wins here, just
> conforming to the norm in function parameters.
>
> v2: Include all the probe functions
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Mechanical.
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Regards, Joonas
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
* Re: [PATCH 3/3] drm/i915: Rearrange GGTT probing to avoid needing a vfunc
2016-08-03 17:14 ` [PATCH 3/3] drm/i915: Rearrange GGTT probing to avoid needing a vfunc Chris Wilson
@ 2016-08-04 6:34 ` Joonas Lahtinen
0 siblings, 0 replies; 41+ messages in thread
From: Joonas Lahtinen @ 2016-08-04 6:34 UTC (permalink / raw)
To: Chris Wilson, intel-gfx
On ke, 2016-08-03 at 18:14 +0100, Chris Wilson wrote:
> Since we have a static if-else-chain for device probing of the global
> GTT, we do not need to use a function pointer, let alone store it when
> we never use it again. So use the if-else-chain to call down into the
> device specific probe.
>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
ggtt->size can be added back when there's need for it, so;
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Regards, Joonas
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
^ permalink raw reply [flat|nested] 41+ messages in thread
end of thread, other threads:[~2016-08-04 6:34 UTC | newest]
Thread overview: 41+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-08-03 13:42 Next batch of almost reviewed patches, up to the VMA leak fix Chris Wilson
2016-08-03 13:42 ` [PATCH 01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Chris Wilson
2016-08-03 13:42 ` [PATCH 02/28] drm/i915: Split GGTT initialisation between probing and setup Chris Wilson
2016-08-03 13:54 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 03/28] drm/i915: Keep ggtt->probe() as a local Chris Wilson
2016-08-03 14:00 ` Joonas Lahtinen
2016-08-03 14:07 ` Chris Wilson
2016-08-03 17:14 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Chris Wilson
2016-08-03 17:14 ` [PATCH 2/3] drm/i915: Split early global GTT initialisation Chris Wilson
2016-08-03 17:14 ` [PATCH 3/3] drm/i915: Rearrange GGTT probing to avoid needing a vfunc Chris Wilson
2016-08-04 6:34 ` Joonas Lahtinen
2016-08-04 6:18 ` [PATCH 1/3] drm/i915: Update GGTT initialisation functions to take drm_i915_private Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 04/28] " Chris Wilson
2016-08-03 14:05 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 05/28] drm/i915: Split early global GTT initialisation Chris Wilson
2016-08-03 14:17 ` Joonas Lahtinen
2016-08-03 13:42 ` [PATCH 06/28] drm/i915: Store owning file on the i915_address_space Chris Wilson
2016-08-03 13:43 ` [PATCH 07/28] drm/i915: Count how many VMA are bound for an object Chris Wilson
2016-08-03 13:43 ` [PATCH 08/28] drm/i915: Be more careful when unbinding vma Chris Wilson
2016-08-03 13:43 ` [PATCH 09/28] drm/i915: Kill drop_pages() Chris Wilson
2016-08-03 13:43 ` [PATCH 10/28] drm/i915: Introduce i915_gem_active for request tracking Chris Wilson
2016-08-03 13:43 ` [PATCH 11/28] drm/i915: Prepare i915_gem_active for annotations Chris Wilson
2016-08-03 13:43 ` [PATCH 12/28] drm/i915: Mark up i915_gem_active for locking annotation Chris Wilson
2016-08-03 13:43 ` [PATCH 13/28] drm/i915: Refactor blocking waits Chris Wilson
2016-08-03 13:43 ` [PATCH 14/28] drm/i915: Rename request->list to link for consistency Chris Wilson
2016-08-03 13:43 ` [PATCH 15/28] drm/i915: Remove obsolete i915_gem_object_flush_active() Chris Wilson
2016-08-03 13:43 ` [PATCH 16/28] drm/i915: Refactor activity tracking for requests Chris Wilson
2016-08-03 13:43 ` [PATCH 17/28] drm/i915: Track requests inside each intel_ring Chris Wilson
2016-08-03 13:43 ` [PATCH 18/28] drm/i915: Convert intel_overlay to request tracking Chris Wilson
2016-08-03 13:43 ` [PATCH 19/28] drm/i915: Move the special case wait-request handling to its one caller Chris Wilson
2016-08-03 13:43 ` [PATCH 20/28] drm/i915: Disable waitboosting for a saturated engine Chris Wilson
2016-08-03 13:43 ` [PATCH 21/28] drm/i915: s/__i915_wait_request/i915_wait_request/ Chris Wilson
2016-08-03 13:43 ` [PATCH 22/28] drm/i915: Double check activity before relocations Chris Wilson
2016-08-03 13:43 ` [PATCH 23/28] drm/i915: Move request list retirement to i915_gem_request.c Chris Wilson
2016-08-03 13:43 ` [PATCH 24/28] drm/i915: i915_vma_move_to_active prep patch Chris Wilson
2016-08-03 13:43 ` [PATCH 25/28] drm/i915: Track active vma requests Chris Wilson
2016-08-03 13:43 ` [PATCH 26/28] drm/i915: Release vma when the handle is closed Chris Wilson
2016-08-03 13:43 ` [PATCH 27/28] drm/i915: Mark the context and address space as closed Chris Wilson
2016-08-03 13:43 ` [PATCH 28/28] Revert "drm/i915: Clean up associated VMAs on context destruction" Chris Wilson
2016-08-03 14:15 ` ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers Patchwork
2016-08-04 5:19 ` ✗ Ro.CI.BAT: failure for series starting with [01/28] drm/i915: Amalgamate GGTT/ppGTT vma debug list walkers (rev4) Patchwork
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox