From cd66407810869709e0d6764409137d48c6812cfc Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Wed, 2 Oct 2013 10:34:19 -0700 Subject: drm/i915: disable LVDS clock gating on CPT v2 Needed to prevent display corruption in high res panels. v2: use correct unit names (Rodrigo) Signed-off-by: Jesse Barnes Tested-by: Ulrich Drepper Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_reg.h | 2 ++ drivers/gpu/drm/i915/intel_pm.c | 4 +++- 2 files changed, 5 insertions(+), 1 deletion(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 38f96f65d87a..fd721ea8728e 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -4279,7 +4279,9 @@ #define FDI_RX_CHICKEN(pipe) _PIPE(pipe, _FDI_RXA_CHICKEN, _FDI_RXB_CHICKEN) #define SOUTH_DSPCLK_GATE_D 0xc2020 +#define PCH_DPLUNIT_CLOCK_GATE_DISABLE (1<<30) #define PCH_DPLSUNIT_CLOCK_GATE_DISABLE (1<<29) +#define PCH_CPUNIT_CLOCK_GATE_DISABLE (1<<14) #define PCH_LP_PARTITION_LEVEL_DISABLE (1<<12) /* CPU: FDI_TX */ diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index f4c5e95b2d6f..26c2ea3e985c 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4759,7 +4759,9 @@ static void cpt_init_clock_gating(struct drm_device *dev) * gating for the panel power sequencer or it will fail to * start up when no ports are active. */ - I915_WRITE(SOUTH_DSPCLK_GATE_D, PCH_DPLSUNIT_CLOCK_GATE_DISABLE); + I915_WRITE(SOUTH_DSPCLK_GATE_D, PCH_DPLSUNIT_CLOCK_GATE_DISABLE | + PCH_DPLUNIT_CLOCK_GATE_DISABLE | + PCH_CPUNIT_CLOCK_GATE_DISABLE); I915_WRITE(SOUTH_CHICKEN2, I915_READ(SOUTH_CHICKEN2) | DPLS_EDP_PPS_FIX_DIS); /* The below fixes the weird display corruption, a few pixels shifted -- cgit From b35b380ed46bb01726bec1795e6443e625306757 Mon Sep 17 00:00:00 2001 From: Ben Widawsky Date: Wed, 16 Oct 2013 09:18:21 -0700 Subject: drm/i915: Make PTE valid encoding optional We need this to work around a corruption when the boot kernel image loads the hibernated kernel image from swap on Haswell systems - somehow not everything is properly shut off. This is just the prep work, the next patch will implement the actual workaround. Signed-off-by: Ben Widawsky [danvet: Add a commit message suitable for -fixes and add cc: stable] Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_drv.h | 3 ++- drivers/gpu/drm/i915/i915_gem_gtt.c | 35 ++++++++++++++++++++--------------- 2 files changed, 22 insertions(+), 16 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 35874b3a86dc..3979a81dd6ee 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -497,7 +497,8 @@ struct i915_address_space { /* FIXME: Need a more generic return type */ gen6_gtt_pte_t (*pte_encode)(dma_addr_t addr, - enum i915_cache_level level); + enum i915_cache_level level, + bool valid); /* Create a valid PTE */ void (*clear_range)(struct i915_address_space *vm, unsigned int first_entry, unsigned int num_entries); diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 212f6d8c35ec..32aa69d7ef20 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -58,9 +58,10 @@ #define HSW_WT_ELLC_LLC_AGE0 HSW_CACHEABILITY_CONTROL(0x6) static gen6_gtt_pte_t snb_pte_encode(dma_addr_t addr, - enum i915_cache_level level) + enum i915_cache_level level, + bool valid) { - gen6_gtt_pte_t pte = GEN6_PTE_VALID; + gen6_gtt_pte_t pte = valid ? GEN6_PTE_VALID : 0; pte |= GEN6_PTE_ADDR_ENCODE(addr); switch (level) { @@ -79,9 +80,10 @@ static gen6_gtt_pte_t snb_pte_encode(dma_addr_t addr, } static gen6_gtt_pte_t ivb_pte_encode(dma_addr_t addr, - enum i915_cache_level level) + enum i915_cache_level level, + bool valid) { - gen6_gtt_pte_t pte = GEN6_PTE_VALID; + gen6_gtt_pte_t pte = valid ? GEN6_PTE_VALID : 0; pte |= GEN6_PTE_ADDR_ENCODE(addr); switch (level) { @@ -105,9 +107,10 @@ static gen6_gtt_pte_t ivb_pte_encode(dma_addr_t addr, #define BYT_PTE_SNOOPED_BY_CPU_CACHES (1 << 2) static gen6_gtt_pte_t byt_pte_encode(dma_addr_t addr, - enum i915_cache_level level) + enum i915_cache_level level, + bool valid) { - gen6_gtt_pte_t pte = GEN6_PTE_VALID; + gen6_gtt_pte_t pte = valid ? GEN6_PTE_VALID : 0; pte |= GEN6_PTE_ADDR_ENCODE(addr); /* Mark the page as writeable. Other platforms don't have a @@ -122,9 +125,10 @@ static gen6_gtt_pte_t byt_pte_encode(dma_addr_t addr, } static gen6_gtt_pte_t hsw_pte_encode(dma_addr_t addr, - enum i915_cache_level level) + enum i915_cache_level level, + bool valid) { - gen6_gtt_pte_t pte = GEN6_PTE_VALID; + gen6_gtt_pte_t pte = valid ? GEN6_PTE_VALID : 0; pte |= HSW_PTE_ADDR_ENCODE(addr); if (level != I915_CACHE_NONE) @@ -134,9 +138,10 @@ static gen6_gtt_pte_t hsw_pte_encode(dma_addr_t addr, } static gen6_gtt_pte_t iris_pte_encode(dma_addr_t addr, - enum i915_cache_level level) + enum i915_cache_level level, + bool valid) { - gen6_gtt_pte_t pte = GEN6_PTE_VALID; + gen6_gtt_pte_t pte = valid ? GEN6_PTE_VALID : 0; pte |= HSW_PTE_ADDR_ENCODE(addr); switch (level) { @@ -245,7 +250,7 @@ static void gen6_ppgtt_clear_range(struct i915_address_space *vm, unsigned first_pte = first_entry % I915_PPGTT_PT_ENTRIES; unsigned last_pte, i; - scratch_pte = vm->pte_encode(vm->scratch.addr, I915_CACHE_LLC); + scratch_pte = vm->pte_encode(vm->scratch.addr, I915_CACHE_LLC, true); while (num_entries) { last_pte = first_pte + num_entries; @@ -282,7 +287,7 @@ static void gen6_ppgtt_insert_entries(struct i915_address_space *vm, dma_addr_t page_addr; page_addr = sg_page_iter_dma_address(&sg_iter); - pt_vaddr[act_pte] = vm->pte_encode(page_addr, cache_level); + pt_vaddr[act_pte] = vm->pte_encode(page_addr, cache_level, true); if (++act_pte == I915_PPGTT_PT_ENTRIES) { kunmap_atomic(pt_vaddr); act_pt++; @@ -536,7 +541,7 @@ static void gen6_ggtt_insert_entries(struct i915_address_space *vm, for_each_sg_page(st->sgl, &sg_iter, st->nents, 0) { addr = sg_page_iter_dma_address(&sg_iter); - iowrite32(vm->pte_encode(addr, level), >t_entries[i]); + iowrite32(vm->pte_encode(addr, level, true), >t_entries[i]); i++; } @@ -548,7 +553,7 @@ static void gen6_ggtt_insert_entries(struct i915_address_space *vm, */ if (i != 0) WARN_ON(readl(>t_entries[i-1]) != - vm->pte_encode(addr, level)); + vm->pte_encode(addr, level, true)); /* This next bit makes the above posting read even more important. We * want to flush the TLBs only after we're certain all the PTE updates @@ -573,7 +578,7 @@ static void gen6_ggtt_clear_range(struct i915_address_space *vm, first_entry, num_entries, max_entries)) num_entries = max_entries; - scratch_pte = vm->pte_encode(vm->scratch.addr, I915_CACHE_LLC); + scratch_pte = vm->pte_encode(vm->scratch.addr, I915_CACHE_LLC, true); for (i = 0; i < num_entries; i++) iowrite32(scratch_pte, >t_base[i]); readl(gtt_base); -- cgit From 828c79087cec61eaf4c76bb32c222fbe35ac3930 Mon Sep 17 00:00:00 2001 From: Ben Widawsky Date: Wed, 16 Oct 2013 09:21:30 -0700 Subject: drm/i915: Disable GGTT PTEs on GEN6+ suspend Once the machine gets to a certain point in the suspend process, we expect the GPU to be idle. If it is not, we might corrupt memory. Empirically (with an early version of this patch) we have seen this is not the case. We cannot currently explain why the latent GPU writes occur. In the technical sense, this patch is a workaround in that we have an issue we can't explain, and the patch indirectly solves the issue. However, it's really better than a workaround because we understand why it works, and it really should be a safe thing to do in all cases. The noticeable effect other than the debug messages would be an increase in the suspend time. I have not measure how expensive it actually is. I think it would be good to spend further time to root cause why we're seeing these latent writes, but it shouldn't preclude preventing the fallout. NOTE: It should be safe (and makes some sense IMO) to also keep the VALID bit unset on resume when we clear_range(). I've opted not to do this as properly clearing those bits at some later point would be extra work. v2: Fix bugzilla link Bugzilla: http://bugs.freedesktop.org/show_bug.cgi?id=65496 Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=59321 Tested-by: Takashi Iwai Tested-by: Paulo Zanoni Signed-off-by: Ben Widawsky Tested-By: Todd Previte Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_drv.c | 5 ++- drivers/gpu/drm/i915/i915_drv.h | 5 ++- drivers/gpu/drm/i915/i915_gem_gtt.c | 76 ++++++++++++++++++++++++++++++++----- drivers/gpu/drm/i915/i915_reg.h | 4 ++ 4 files changed, 78 insertions(+), 12 deletions(-) (limited to 'drivers/gpu/drm') diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 69d8ed5416c3..2ad27880cd04 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -505,6 +505,8 @@ static int i915_drm_freeze(struct drm_device *dev) intel_modeset_suspend_hw(dev); } + i915_gem_suspend_gtt_mappings(dev); + i915_save_state(dev); intel_opregion_fini(dev); @@ -648,7 +650,8 @@ static int i915_drm_thaw(struct drm_device *dev) mutex_lock(&dev->struct_mutex); i915_gem_restore_gtt_mappings(dev); mutex_unlock(&dev->struct_mutex); - } + } else if (drm_core_check_feature(dev, DRIVER_MODESET)) + i915_check_and_clear_faults(dev); __i915_drm_thaw(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 3979a81dd6ee..ab0f2c0a440c 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -501,7 +501,8 @@ struct i915_address_space { bool valid); /* Create a valid PTE */ void (*clear_range)(struct i915_address_space *vm, unsigned int first_entry, - unsigned int num_entries); + unsigned int num_entries, + bool use_scratch); void (*insert_entries)(struct i915_address_space *vm, struct sg_table *st, unsigned int first_entry, @@ -2066,6 +2067,8 @@ void i915_ppgtt_bind_object(struct i915_hw_ppgtt *ppgtt, void i915_ppgtt_unbind_object(struct i915_hw_ppgtt *ppgtt, struct drm_i915_gem_object *obj); +void i915_check_and_clear_faults(struct drm_device *dev); +void i915_gem_suspend_gtt_mappings(struct drm_device *dev); void i915_gem_restore_gtt_mappings(struct drm_device *dev); int __must_check i915_gem_gtt_prepare_object(struct drm_i915_gem_object *obj); void i915_gem_gtt_bind_object(struct drm_i915_gem_object *obj, diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c index 32aa69d7ef20..1f7b4caefb6e 100644 --- a/drivers/gpu/drm/i915/i915_gem_gtt.c +++ b/drivers/gpu/drm/i915/i915_gem_gtt.c @@ -241,7 +241,8 @@ static int gen6_ppgtt_enable(struct drm_device *dev) /* PPGTT support for Sandybdrige/Gen6 and later */ static void gen6_ppgtt_clear_range(struct i915_address_space *vm, unsigned first_entry, - unsigned num_entries) + unsigned num_entries, + bool use_scratch) { struct i915_hw_ppgtt *ppgtt = container_of(vm, struct i915_hw_ppgtt, base); @@ -372,7 +373,7 @@ static int gen6_ppgtt_init(struct i915_hw_ppgtt *ppgtt) } ppgtt->base.clear_range(&ppgtt->base, 0, - ppgtt->num_pd_entries * I915_PPGTT_PT_ENTRIES); + ppgtt->num_pd_entries * I915_PPGTT_PT_ENTRIES, true); ppgtt->pd_offset = first_pd_entry_in_global_pt * sizeof(gen6_gtt_pte_t); @@ -449,7 +450,8 @@ void i915_ppgtt_unbind_object(struct i915_hw_ppgtt *ppgtt, { ppgtt->base.clear_range(&ppgtt->base, i915_gem_obj_ggtt_offset(obj) >> PAGE_SHIFT, - obj->base.size >> PAGE_SHIFT); + obj->base.size >> PAGE_SHIFT, + true); } extern int intel_iommu_gfx_mapped; @@ -490,15 +492,65 @@ static void undo_idling(struct drm_i915_private *dev_priv, bool interruptible) dev_priv->mm.interruptible = interruptible; } +void i915_check_and_clear_faults(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + struct intel_ring_buffer *ring; + int i; + + if (INTEL_INFO(dev)->gen < 6) + return; + + for_each_ring(ring, dev_priv, i) { + u32 fault_reg; + fault_reg = I915_READ(RING_FAULT_REG(ring)); + if (fault_reg & RING_FAULT_VALID) { + DRM_DEBUG_DRIVER("Unexpected fault\n" + "\tAddr: 0x%08lx\\n" + "\tAddress space: %s\n" + "\tSource ID: %d\n" + "\tType: %d\n", + fault_reg & PAGE_MASK, + fault_reg & RING_FAULT_GTTSEL_MASK ? "GGTT" : "PPGTT", + RING_FAULT_SRCID(fault_reg), + RING_FAULT_FAULT_TYPE(fault_reg)); + I915_WRITE(RING_FAULT_REG(ring), + fault_reg & ~RING_FAULT_VALID); + } + } + POSTING_READ(RING_FAULT_REG(&dev_priv->ring[RCS])); +} + +void i915_gem_suspend_gtt_mappings(struct drm_device *dev) +{ + struct drm_i915_private *dev_priv = dev->dev_private; + + /* Don't bother messing with faults pre GEN6 as we have little + * documentation supporting that it's a good idea. + */ + if (INTEL_INFO(dev)->gen < 6) + return; + + i915_check_and_clear_faults(dev); + + dev_priv->gtt.base.clear_range(&dev_priv->gtt.base, + dev_priv->gtt.base.start / PAGE_SIZE, + dev_priv->gtt.base.total / PAGE_SIZE, + false); +} + void i915_gem_restore_gtt_mappings(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; struct drm_i915_gem_object *obj; + i915_check_and_clear_faults(dev); + /* First fill our portion of the GTT with scratch pages */ dev_priv->gtt.base.clear_range(&dev_priv->gtt.base, dev_priv->gtt.base.start / PAGE_SIZE, - dev_priv->gtt.base.total / PAGE_SIZE); + dev_priv->gtt.base.total / PAGE_SIZE, + true); list_for_each_entry(obj, &dev_priv->mm.bound_list, global_list) { i915_gem_clflush_object(obj, obj->pin_display); @@ -565,7 +617,8 @@ static void gen6_ggtt_insert_entries(struct i915_address_space *vm, static void gen6_ggtt_clear_range(struct i915_address_space *vm, unsigned int first_entry, - unsigned int num_entries) + unsigned int num_entries, + bool use_scratch) { struct drm_i915_private *dev_priv = vm->dev->dev_private; gen6_gtt_pte_t scratch_pte, __iomem *gtt_base = @@ -578,7 +631,8 @@ static void gen6_ggtt_clear_range(struct i915_address_space *vm, first_entry, num_entries, max_entries)) num_entries = max_entries; - scratch_pte = vm->pte_encode(vm->scratch.addr, I915_CACHE_LLC, true); + scratch_pte = vm->pte_encode(vm->scratch.addr, I915_CACHE_LLC, use_scratch); + for (i = 0; i < num_entries; i++) iowrite32(scratch_pte, >t_base[i]); readl(gtt_base); @@ -599,7 +653,8 @@ static void i915_ggtt_insert_entries(struct i915_address_space *vm, static void i915_ggtt_clear_range(struct i915_address_space *vm, unsigned int first_entry, - unsigned int num_entries) + unsigned int num_entries, + bool unused) { intel_gtt_clear_range(first_entry, num_entries); } @@ -627,7 +682,8 @@ void i915_gem_gtt_unbind_object(struct drm_i915_gem_object *obj) dev_priv->gtt.base.clear_range(&dev_priv->gtt.base, entry, - obj->base.size >> PAGE_SHIFT); + obj->base.size >> PAGE_SHIFT, + true); obj->has_global_gtt_mapping = 0; } @@ -714,11 +770,11 @@ void i915_gem_setup_global_gtt(struct drm_device *dev, const unsigned long count = (hole_end - hole_start) / PAGE_SIZE; DRM_DEBUG_KMS("clearing unused GTT space: [%lx, %lx]\n", hole_start, hole_end); - ggtt_vm->clear_range(ggtt_vm, hole_start / PAGE_SIZE, count); + ggtt_vm->clear_range(ggtt_vm, hole_start / PAGE_SIZE, count, true); } /* And finally clear the reserved guard page */ - ggtt_vm->clear_range(ggtt_vm, end / PAGE_SIZE - 1, 1); + ggtt_vm->clear_range(ggtt_vm, end / PAGE_SIZE - 1, 1, true); } static bool diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index fd721ea8728e..ef9b35479f01 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -604,6 +604,10 @@ #define ARB_MODE_SWIZZLE_IVB (1<<5) #define RENDER_HWS_PGA_GEN7 (0x04080) #define RING_FAULT_REG(ring) (0x4094 + 0x100*(ring)->id) +#define RING_FAULT_GTTSEL_MASK (1<<11) +#define RING_FAULT_SRCID(x) ((x >> 3) & 0xff) +#define RING_FAULT_FAULT_TYPE(x) ((x >> 1) & 0x3) +#define RING_FAULT_VALID (1<<0) #define DONE_REG 0x40b0 #define BSD_HWS_PGA_GEN7 (0x04180) #define BLT_HWS_PGA_GEN7 (0x04280) -- cgit