From a9c1f90c8e1792127f8348c1cc2565b79b2ef20a Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Thu, 22 Aug 2013 21:09:00 +0300 Subject: [PATCH 01/29] drm/i915: Don't mask EI UP interrupt on IVB|SNB Submitting a batchbuffer which simulates a gpu hang by doing MI_BATCH_BUFFER_START into itself, to test hangcheck, started to hard hang the whole box (IVB). Bisecting lead to this commit: commit 664b422c2966cd39b8f67e8d53a566ea8c877cd6 Author: Vinit Azad Date: Wed Aug 14 13:34:33 2013 -0700 drm/i915: Only unmask required PM interrupts Experimenting with the mask register showed that unmasking EI UP will prevent the hard hang in IVB and SNB. HSW doesn't hang with EI UP masked. Considering we are just disabling interrupts that aren't even delivered to driver, this change is more likely to paper over some weirdness in gpu's internal state machine. But until better explanation can be found, let's trade little bit of power for stability on these architectures. v2: - Unmask EI_EXPIRED directly in I915_WRITE (Vinit) v3: - Only unmask on SNB and IVB Cc: Vinit Azad Signed-off-by: Mika Kuoppala Acked-by: Vinit Azad Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_pm.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 46056820d1d2..6b1d00389952 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -3447,14 +3447,24 @@ int intel_enable_rc6(const struct drm_device *dev) static void gen6_enable_rps_interrupts(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; + u32 enabled_intrs; spin_lock_irq(&dev_priv->irq_lock); WARN_ON(dev_priv->rps.pm_iir); snb_enable_pm_irq(dev_priv, GEN6_PM_RPS_EVENTS); I915_WRITE(GEN6_PMIIR, GEN6_PM_RPS_EVENTS); spin_unlock_irq(&dev_priv->irq_lock); + /* only unmask PM interrupts we need. Mask all others. */ - I915_WRITE(GEN6_PMINTRMSK, ~GEN6_PM_RPS_EVENTS); + enabled_intrs = GEN6_PM_RPS_EVENTS; + + /* IVB and SNB hard hangs on looping batchbuffer + * if GEN6_PM_UP_EI_EXPIRED is masked. + */ + if (INTEL_INFO(dev)->gen <= 7 && !IS_HASWELL(dev)) + enabled_intrs |= GEN6_PM_RP_UP_EI_EXPIRED; + + I915_WRITE(GEN6_PMINTRMSK, ~enabled_intrs); } static void gen6_enable_rps(struct drm_device *dev) From 521198a2e7095c8c7daa8d7d3a76a110c346be6f Mon Sep 17 00:00:00 2001 From: Mika Kuoppala Date: Fri, 23 Aug 2013 16:52:30 +0300 Subject: [PATCH 02/29] drm/i915: sanitize forcewake registers on reset In reset we try to restore the forcewake state to pre reset state, using forcewake_count. The reset doesn't seem to clear the forcewake bits so we get warn on forcewake ack register not clearing. Use same mechanism as intel_uncore_sanitize() does when loading driver to reset the forcewake bits, right after the chip has been reset. Reviewed-by: Chris Wilson Signed-off-by: Mika Kuoppala Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_uncore.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c index 8f5bc869c023..8649f1c36b00 100644 --- a/drivers/gpu/drm/i915/intel_uncore.c +++ b/drivers/gpu/drm/i915/intel_uncore.c @@ -261,7 +261,7 @@ void intel_uncore_init(struct drm_device *dev) } } -void intel_uncore_sanitize(struct drm_device *dev) +static void intel_uncore_forcewake_reset(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; @@ -272,6 +272,11 @@ void intel_uncore_sanitize(struct drm_device *dev) if (IS_IVYBRIDGE(dev) || IS_HASWELL(dev)) __gen6_gt_force_wake_mt_reset(dev_priv); } +} + +void intel_uncore_sanitize(struct drm_device *dev) +{ + intel_uncore_forcewake_reset(dev); /* BIOS often leaves RC6 enabled, but disable it for hw init */ intel_disable_gt_powersave(dev); @@ -549,6 +554,8 @@ static int gen6_do_reset(struct drm_device *dev) /* Spin waiting for the device to ack the reset request */ ret = wait_for((__raw_i915_read32(dev_priv, GEN6_GDRST) & GEN6_GRDOM_FULL) == 0, 500); + intel_uncore_forcewake_reset(dev); + /* If reset with a user forcewake, try to restore, otherwise turn it off */ if (dev_priv->uncore.forcewake_count) dev_priv->uncore.funcs.force_wake_get(dev_priv); From 97e4eed7dc532e20208b0bdf7ad1136569da2f35 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 26 Aug 2013 16:18:54 +0100 Subject: [PATCH 03/29] drm/i915: Adjust available RPS information through sysfs for vlv Valleyview has its own render power state implementation with different capability knobs - it has no RP0,RP1,RPn but rather RPe. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67734 Signed-off-by: Chris Wilson Tested-by: kobe.qin@intel.com Reviewed-by: Jani Nikula Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_sysfs.c | 36 ++++++++++++++++++++++++++----- 1 file changed, 31 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c index a777e7f3b0df..c8c4112de110 100644 --- a/drivers/gpu/drm/i915/i915_sysfs.c +++ b/drivers/gpu/drm/i915/i915_sysfs.c @@ -224,6 +224,18 @@ static ssize_t gt_cur_freq_mhz_show(struct device *kdev, return snprintf(buf, PAGE_SIZE, "%d\n", ret); } +static ssize_t vlv_rpe_freq_mhz_show(struct device *kdev, + struct device_attribute *attr, char *buf) +{ + struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); + struct drm_device *dev = minor->dev; + struct drm_i915_private *dev_priv = dev->dev_private; + + return snprintf(buf, PAGE_SIZE, "%d\n", + vlv_gpu_freq(dev_priv->mem_freq, + dev_priv->rps.rpe_delay)); +} + static ssize_t gt_max_freq_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf) { struct drm_minor *minor = container_of(kdev, struct drm_minor, kdev); @@ -366,6 +378,7 @@ static DEVICE_ATTR(gt_cur_freq_mhz, S_IRUGO, gt_cur_freq_mhz_show, NULL); static DEVICE_ATTR(gt_max_freq_mhz, S_IRUGO | S_IWUSR, gt_max_freq_mhz_show, gt_max_freq_mhz_store); static DEVICE_ATTR(gt_min_freq_mhz, S_IRUGO | S_IWUSR, gt_min_freq_mhz_show, gt_min_freq_mhz_store); +static DEVICE_ATTR(vlv_rpe_freq_mhz, S_IRUGO, vlv_rpe_freq_mhz_show, NULL); static ssize_t gt_rp_mhz_show(struct device *kdev, struct device_attribute *attr, char *buf); static DEVICE_ATTR(gt_RP0_freq_mhz, S_IRUGO, gt_rp_mhz_show, NULL); @@ -409,6 +422,14 @@ static const struct attribute *gen6_attrs[] = { NULL, }; +static const struct attribute *vlv_attrs[] = { + &dev_attr_gt_cur_freq_mhz.attr, + &dev_attr_gt_max_freq_mhz.attr, + &dev_attr_gt_min_freq_mhz.attr, + &dev_attr_vlv_rpe_freq_mhz.attr, + NULL, +}; + static ssize_t error_state_read(struct file *filp, struct kobject *kobj, struct bin_attribute *attr, char *buf, loff_t off, size_t count) @@ -492,11 +513,13 @@ void i915_setup_sysfs(struct drm_device *dev) DRM_ERROR("l3 parity sysfs setup failed\n"); } - if (INTEL_INFO(dev)->gen >= 6) { + ret = 0; + if (IS_VALLEYVIEW(dev)) + ret = sysfs_create_files(&dev->primary->kdev.kobj, vlv_attrs); + else if (INTEL_INFO(dev)->gen >= 6) ret = sysfs_create_files(&dev->primary->kdev.kobj, gen6_attrs); - if (ret) - DRM_ERROR("gen6 sysfs setup failed\n"); - } + if (ret) + DRM_ERROR("RPS sysfs setup failed\n"); ret = sysfs_create_bin_file(&dev->primary->kdev.kobj, &error_state_attr); @@ -507,7 +530,10 @@ void i915_setup_sysfs(struct drm_device *dev) void i915_teardown_sysfs(struct drm_device *dev) { sysfs_remove_bin_file(&dev->primary->kdev.kobj, &error_state_attr); - sysfs_remove_files(&dev->primary->kdev.kobj, gen6_attrs); + if (IS_VALLEYVIEW(dev)) + sysfs_remove_files(&dev->primary->kdev.kobj, vlv_attrs); + else + sysfs_remove_files(&dev->primary->kdev.kobj, gen6_attrs); device_remove_bin_file(&dev->primary->kdev, &dpf_attrs); #ifdef CONFIG_PM sysfs_unmerge_group(&dev->primary->kdev.kobj, &rc6_attr_group); From 10603caacf599297c7da0c4f4db440d015b8131a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 26 Aug 2013 19:51:06 -0300 Subject: [PATCH 04/29] drm/i915: Apply the force-detect VGA w/a to Valleyview It appears that Valleyview shares its VGA encoder with more recent siblings and requires the same forced detection cycle after a hardware reset before we can rely on hotplugging. Reported-and-tested-by: kobeqin Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=67733 Tested-by: kobeqin Signed-off-by: Chris Wilson [danvet: Check for gen >= 5 insted, acked by Chris.] Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_crt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_crt.c b/drivers/gpu/drm/i915/intel_crt.c index b5a3875f22c7..ea9022ef15d5 100644 --- a/drivers/gpu/drm/i915/intel_crt.c +++ b/drivers/gpu/drm/i915/intel_crt.c @@ -688,7 +688,7 @@ static void intel_crt_reset(struct drm_connector *connector) struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crt *crt = intel_attached_crt(connector); - if (HAS_PCH_SPLIT(dev)) { + if (INTEL_INFO(dev)->gen >= 5) { u32 adpa; adpa = I915_READ(crt->adpa_reg); From 8e8c06cd34dbd68d36b0c561b8850478ddc5fb84 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 26 Aug 2013 19:51:01 -0300 Subject: [PATCH 05/29] drm/i915: Report requested frequency alongside current frequency in debugfs It can be useful to compare at times the current vs requested frequency of the GPU, so provide the contents of RPNSWREQ alonside CAGF. Signed-off-by: Chris Wilson Reviewed-by: Rodrigo Vivi Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_debugfs.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 55ab9246e1b9..a6f4cb5af185 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -857,7 +857,7 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused) u32 gt_perf_status = I915_READ(GEN6_GT_PERF_STATUS); u32 rp_state_limits = I915_READ(GEN6_RP_STATE_LIMITS); u32 rp_state_cap = I915_READ(GEN6_RP_STATE_CAP); - u32 rpstat, cagf; + u32 rpstat, cagf, reqf; u32 rpupei, rpcurup, rpprevup; u32 rpdownei, rpcurdown, rpprevdown; int max_freq; @@ -869,6 +869,14 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused) gen6_gt_force_wake_get(dev_priv); + reqf = I915_READ(GEN6_RPNSWREQ); + reqf &= ~GEN6_TURBO_DISABLE; + if (IS_HASWELL(dev)) + reqf >>= 24; + else + reqf >>= 25; + reqf *= GT_FREQUENCY_MULTIPLIER; + rpstat = I915_READ(GEN6_RPSTAT1); rpupei = I915_READ(GEN6_RP_CUR_UP_EI); rpcurup = I915_READ(GEN6_RP_CUR_UP); @@ -893,6 +901,7 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused) gt_perf_status & 0xff); seq_printf(m, "Render p-state limit: %d\n", rp_state_limits & 0xff); + seq_printf(m, "RPNSWREQ: %dMHz\n", reqf); seq_printf(m, "CAGF: %dMHz\n", cagf); seq_printf(m, "RP CUR UP EI: %dus\n", rpupei & GEN6_CURICONT_MASK); From b8d88d1d40c18d42e3fedc289832e28d14bc1f00 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 28 Aug 2013 10:57:59 +0200 Subject: [PATCH 06/29] drm/i915: tune down hangcheck noise We already have a big splashing *ERROR* for all the relevant cases of hangs, so this one here is redudant. And it results in an unclean dmesg when running with simulated hangs. Regression has been introduced in commit 05407ff889ceebe383aa5907219f86582ef96b72 Author: Mika Kuoppala Date: Thu May 30 09:04:29 2013 +0300 drm/i915: detect hang using per ring hangcheck_score Cc: Mika Kuoppala Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68641 Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_irq.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index a03b445ceb5f..4c6853f57d62 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -2027,9 +2027,9 @@ static void i915_hangcheck_elapsed(unsigned long data) for_each_ring(ring, dev_priv, i) { if (ring->hangcheck.score > FIRE) { - DRM_ERROR("%s on %s\n", - stuck[i] ? "stuck" : "no progress", - ring->name); + DRM_INFO("%s on %s\n", + stuck[i] ? "stuck" : "no progress", + ring->name); rings_hung++; } } From a52690e445637dda7f71878965d64d9b6a15a2b7 Mon Sep 17 00:00:00 2001 From: Imre Deak Date: Tue, 27 Aug 2013 12:24:09 +0300 Subject: [PATCH 07/29] drm/i915: fix lvds/dp panel fitter setting If need to enable the panel fitter, the crtc timings have to be programmed according to the panel's native (fixed) mode. This isn't the case atm, since after the encoder changes adjusted_mode to fixed mode the crtc_* timing fields of adjusted_mode will stay at their original non-native values that the user passed in. This results in a corrupted output. One exception is when we have a second pass of computing encoder configs due to bandwidth limitation, since then we'll set adjusted_mode.crtc_* fields to the fixed mode values set in the first pass; so in this case things will work out. Fix this by updating the adjusted_mode.crtc_* fields when we set the fixed panel mode. This regression has been introduced in commit 135c81b8c3c9a70d7b55758c9c2a247a4abb7b64 Author: Daniel Vetter Date: Sun Jul 21 21:37:09 2013 +0200 drm/i915: clean up crtc timings computation Signed-off-by: Imre Deak Reviewed-by: Mika Kuoppala Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_panel.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c index a43c33bc4a35..913cb9d7fd32 100644 --- a/drivers/gpu/drm/i915/intel_panel.c +++ b/drivers/gpu/drm/i915/intel_panel.c @@ -50,6 +50,8 @@ intel_fixed_panel_mode(struct drm_display_mode *fixed_mode, adjusted_mode->vtotal = fixed_mode->vtotal; adjusted_mode->clock = fixed_mode->clock; + + drm_mode_set_crtcinfo(adjusted_mode, 0); } /* adjusted_mode has been preset to be the panel's fixed mode */ From 0d1aacac36530fce058d7a0db3da7befd5765417 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 26 Aug 2013 20:58:11 +0100 Subject: [PATCH 08/29] drm/i915: Embed the ring->private within the struct intel_ring_buffer We now have more devices using ring->private than not, and they all want the same structure. Worse, I would like to use a scratch page from outside of intel_ringbuffer.c and so for convenience would like to reuse ring->private. Embed the object into the struct intel_ringbuffer so that we can keep the code clean. Signed-off-by: Chris Wilson Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gpu_error.c | 2 +- drivers/gpu/drm/i915/intel_ringbuffer.c | 99 ++++++++----------------- drivers/gpu/drm/i915/intel_ringbuffer.h | 6 +- 3 files changed, 35 insertions(+), 72 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gpu_error.c b/drivers/gpu/drm/i915/i915_gpu_error.c index 558e568d5b45..aba9d7498996 100644 --- a/drivers/gpu/drm/i915/i915_gpu_error.c +++ b/drivers/gpu/drm/i915/i915_gpu_error.c @@ -641,7 +641,7 @@ i915_error_first_batchbuffer(struct drm_i915_private *dev_priv, if (WARN_ON(ring->id != RCS)) return NULL; - obj = ring->private; + obj = ring->scratch.obj; if (acthd >= i915_gem_obj_ggtt_offset(obj) && acthd < i915_gem_obj_ggtt_offset(obj) + obj->base.size) return i915_error_object_create(dev_priv, obj); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index f05cceac5a52..460ee1026fca 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -33,16 +33,6 @@ #include "i915_trace.h" #include "intel_drv.h" -/* - * 965+ support PIPE_CONTROL commands, which provide finer grained control - * over cache flushing. - */ -struct pipe_control { - struct drm_i915_gem_object *obj; - volatile u32 *cpu_page; - u32 gtt_offset; -}; - static inline int ring_space(struct intel_ring_buffer *ring) { int space = (ring->head & HEAD_ADDR) - (ring->tail + I915_RING_FREE_SPACE); @@ -175,8 +165,7 @@ gen4_render_ring_flush(struct intel_ring_buffer *ring, static int intel_emit_post_sync_nonzero_flush(struct intel_ring_buffer *ring) { - struct pipe_control *pc = ring->private; - u32 scratch_addr = pc->gtt_offset + 128; + u32 scratch_addr = ring->scratch.gtt_offset + 128; int ret; @@ -213,8 +202,7 @@ gen6_render_ring_flush(struct intel_ring_buffer *ring, u32 invalidate_domains, u32 flush_domains) { u32 flags = 0; - struct pipe_control *pc = ring->private; - u32 scratch_addr = pc->gtt_offset + 128; + u32 scratch_addr = ring->scratch.gtt_offset + 128; int ret; /* Force SNB workarounds for PIPE_CONTROL flushes */ @@ -306,8 +294,7 @@ gen7_render_ring_flush(struct intel_ring_buffer *ring, u32 invalidate_domains, u32 flush_domains) { u32 flags = 0; - struct pipe_control *pc = ring->private; - u32 scratch_addr = pc->gtt_offset + 128; + u32 scratch_addr = ring->scratch.gtt_offset + 128; int ret; /* @@ -481,68 +468,43 @@ out: static int init_pipe_control(struct intel_ring_buffer *ring) { - struct pipe_control *pc; - struct drm_i915_gem_object *obj; int ret; - if (ring->private) + if (ring->scratch.obj) return 0; - pc = kmalloc(sizeof(*pc), GFP_KERNEL); - if (!pc) - return -ENOMEM; - - obj = i915_gem_alloc_object(ring->dev, 4096); - if (obj == NULL) { + ring->scratch.obj = i915_gem_alloc_object(ring->dev, 4096); + if (ring->scratch.obj == NULL) { DRM_ERROR("Failed to allocate seqno page\n"); ret = -ENOMEM; goto err; } - i915_gem_object_set_cache_level(obj, I915_CACHE_LLC); + i915_gem_object_set_cache_level(ring->scratch.obj, I915_CACHE_LLC); - ret = i915_gem_obj_ggtt_pin(obj, 4096, true, false); + ret = i915_gem_obj_ggtt_pin(ring->scratch.obj, 4096, true, false); if (ret) goto err_unref; - pc->gtt_offset = i915_gem_obj_ggtt_offset(obj); - pc->cpu_page = kmap(sg_page(obj->pages->sgl)); - if (pc->cpu_page == NULL) { + ring->scratch.gtt_offset = i915_gem_obj_ggtt_offset(ring->scratch.obj); + ring->scratch.cpu_page = kmap(sg_page(ring->scratch.obj->pages->sgl)); + if (ring->scratch.cpu_page == NULL) { ret = -ENOMEM; goto err_unpin; } DRM_DEBUG_DRIVER("%s pipe control offset: 0x%08x\n", - ring->name, pc->gtt_offset); - - pc->obj = obj; - ring->private = pc; + ring->name, ring->scratch.gtt_offset); return 0; err_unpin: - i915_gem_object_unpin(obj); + i915_gem_object_unpin(ring->scratch.obj); err_unref: - drm_gem_object_unreference(&obj->base); + drm_gem_object_unreference(&ring->scratch.obj->base); err: - kfree(pc); return ret; } -static void -cleanup_pipe_control(struct intel_ring_buffer *ring) -{ - struct pipe_control *pc = ring->private; - struct drm_i915_gem_object *obj; - - obj = pc->obj; - - kunmap(sg_page(obj->pages->sgl)); - i915_gem_object_unpin(obj); - drm_gem_object_unreference(&obj->base); - - kfree(pc); -} - static int init_render_ring(struct intel_ring_buffer *ring) { struct drm_device *dev = ring->dev; @@ -607,16 +569,16 @@ static void render_ring_cleanup(struct intel_ring_buffer *ring) { struct drm_device *dev = ring->dev; - if (!ring->private) + if (ring->scratch.obj == NULL) return; - if (HAS_BROKEN_CS_TLB(dev)) - drm_gem_object_unreference(to_gem_object(ring->private)); + if (INTEL_INFO(dev)->gen >= 5) { + kunmap(sg_page(ring->scratch.obj->pages->sgl)); + i915_gem_object_unpin(ring->scratch.obj); + } - if (INTEL_INFO(dev)->gen >= 5) - cleanup_pipe_control(ring); - - ring->private = NULL; + drm_gem_object_unreference(&ring->scratch.obj->base); + ring->scratch.obj = NULL; } static void @@ -742,8 +704,7 @@ do { \ static int pc_render_add_request(struct intel_ring_buffer *ring) { - struct pipe_control *pc = ring->private; - u32 scratch_addr = pc->gtt_offset + 128; + u32 scratch_addr = ring->scratch.gtt_offset + 128; int ret; /* For Ironlake, MI_USER_INTERRUPT was deprecated and apparently @@ -761,7 +722,7 @@ pc_render_add_request(struct intel_ring_buffer *ring) intel_ring_emit(ring, GFX_OP_PIPE_CONTROL(4) | PIPE_CONTROL_QW_WRITE | PIPE_CONTROL_WRITE_FLUSH | PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE); - intel_ring_emit(ring, pc->gtt_offset | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, ring->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT); intel_ring_emit(ring, ring->outstanding_lazy_request); intel_ring_emit(ring, 0); PIPE_CONTROL_FLUSH(ring, scratch_addr); @@ -780,7 +741,7 @@ pc_render_add_request(struct intel_ring_buffer *ring) PIPE_CONTROL_WRITE_FLUSH | PIPE_CONTROL_TEXTURE_CACHE_INVALIDATE | PIPE_CONTROL_NOTIFY); - intel_ring_emit(ring, pc->gtt_offset | PIPE_CONTROL_GLOBAL_GTT); + intel_ring_emit(ring, ring->scratch.gtt_offset | PIPE_CONTROL_GLOBAL_GTT); intel_ring_emit(ring, ring->outstanding_lazy_request); intel_ring_emit(ring, 0); intel_ring_advance(ring); @@ -814,15 +775,13 @@ ring_set_seqno(struct intel_ring_buffer *ring, u32 seqno) static u32 pc_render_get_seqno(struct intel_ring_buffer *ring, bool lazy_coherency) { - struct pipe_control *pc = ring->private; - return pc->cpu_page[0]; + return ring->scratch.cpu_page[0]; } static void pc_render_set_seqno(struct intel_ring_buffer *ring, u32 seqno) { - struct pipe_control *pc = ring->private; - pc->cpu_page[0] = seqno; + ring->scratch.cpu_page[0] = seqno; } static bool @@ -1141,8 +1100,7 @@ i830_dispatch_execbuffer(struct intel_ring_buffer *ring, intel_ring_emit(ring, MI_NOOP); intel_ring_advance(ring); } else { - struct drm_i915_gem_object *obj = ring->private; - u32 cs_offset = i915_gem_obj_ggtt_offset(obj); + u32 cs_offset = ring->scratch.gtt_offset; if (len > I830_BATCH_LIMIT) return -ENOSPC; @@ -1835,7 +1793,8 @@ int intel_init_render_ring_buffer(struct drm_device *dev) return ret; } - ring->private = obj; + ring->scratch.obj = obj; + ring->scratch.gtt_offset = i915_gem_obj_ggtt_offset(obj); } return intel_init_ring_buffer(dev, ring); diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 432ad5311ba6..68b1ca974d59 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -155,7 +155,11 @@ struct intel_ring_buffer { struct intel_ring_hangcheck hangcheck; - void *private; + struct { + struct drm_i915_gem_object *obj; + u32 gtt_offset; + volatile u32 *cpu_page; + } scratch; }; static inline bool From ffe74d75502e3a9b0791240b5562bcbecc6ab8dc Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 26 Aug 2013 20:58:12 +0100 Subject: [PATCH 09/29] drm/i915: Use RCS flips on Ivybridge+ MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit RCS flips do work on Iybridge+ so long as we can unmask the messages through DERRMR. However, there are quite a few workarounds mentioned regarding unmasking more than one event or triggering more than one message through DERRMR. Those workarounds in principle prevent us from performing pipelined flips (and asynchronous flips across multiple planes) and equally apply to the "known good" BCS ring. Given that it already appears to work, and also appears to work with unmasking all 3 planes at once (and queuing flips across multiple planes), be brave. Bugzlla: https://bugs.freedesktop.org/show_bug.cgi?id=67600 Signed-off-by: Chris Wilson Lightly-tested-by: Stephane Marchesin Cc: Stephane Marchesin Cc: Ben Widawsky Tested-by: Stéphane Marchesin Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_reg.h | 18 +++++++++++++ drivers/gpu/drm/i915/intel_display.c | 40 +++++++++++++++++++++------- 2 files changed, 49 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index b6a58f720f9a..b2fa2a4c4454 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -245,6 +245,7 @@ * address/value pairs. Don't overdue it, though, x <= 2^4 must hold! */ #define MI_LOAD_REGISTER_IMM(x) MI_INSTR(0x22, 2*x-1) +#define MI_STORE_REGISTER_MEM(x) MI_INSTR(0x24, 2*x-1) #define MI_FLUSH_DW MI_INSTR(0x26, 1) /* for GEN6 */ #define MI_FLUSH_DW_STORE_INDEX (1<<21) #define MI_INVALIDATE_TLB (1<<18) @@ -693,6 +694,23 @@ #define FPGA_DBG_RM_NOCLAIM (1<<31) #define DERRMR 0x44050 +#define DERRMR_PIPEA_SCANLINE (1<<0) +#define DERRMR_PIPEA_PRI_FLIP_DONE (1<<1) +#define DERRMR_PIPEA_SPR_FLIP_DONE (1<<2) +#define DERRMR_PIPEA_VBLANK (1<<3) +#define DERRMR_PIPEA_HBLANK (1<<5) +#define DERRMR_PIPEB_SCANLINE (1<<8) +#define DERRMR_PIPEB_PRI_FLIP_DONE (1<<9) +#define DERRMR_PIPEB_SPR_FLIP_DONE (1<<10) +#define DERRMR_PIPEB_VBLANK (1<<11) +#define DERRMR_PIPEB_HBLANK (1<<13) +/* Note that PIPEC is not a simple translation of PIPEA/PIPEB */ +#define DERRMR_PIPEC_SCANLINE (1<<14) +#define DERRMR_PIPEC_PRI_FLIP_DONE (1<<15) +#define DERRMR_PIPEC_SPR_FLIP_DONE (1<<20) +#define DERRMR_PIPEC_VBLANK (1<<21) +#define DERRMR_PIPEC_HBLANK (1<<22) + /* GM45+ chicken bits -- debug workaround bits that may be required * for various sorts of correct behavior. The top 16 bits of each are diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 38452d82ac7d..a98392f9894e 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7828,12 +7828,6 @@ err: return ret; } -/* - * On gen7 we currently use the blit ring because (in early silicon at least) - * the render ring doesn't give us interrpts for page flip completion, which - * means clients will hang after the first flip is queued. Fortunately the - * blit ring generates interrupts properly, so use it instead. - */ static int intel_gen7_queue_flip(struct drm_device *dev, struct drm_crtc *crtc, struct drm_framebuffer *fb, @@ -7842,9 +7836,13 @@ static int intel_gen7_queue_flip(struct drm_device *dev, { struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *intel_crtc = to_intel_crtc(crtc); - struct intel_ring_buffer *ring = &dev_priv->ring[BCS]; + struct intel_ring_buffer *ring; uint32_t plane_bit = 0; - int ret; + int len, ret; + + ring = obj->ring; + if (ring == NULL || ring->id != RCS) + ring = &dev_priv->ring[BCS]; ret = intel_pin_and_fence_fb_obj(dev, obj, ring); if (ret) @@ -7866,10 +7864,34 @@ static int intel_gen7_queue_flip(struct drm_device *dev, goto err_unpin; } - ret = intel_ring_begin(ring, 4); + len = 4; + if (ring->id == RCS) + len += 6; + + ret = intel_ring_begin(ring, len); if (ret) goto err_unpin; + /* Unmask the flip-done completion message. Note that the bspec says that + * we should do this for both the BCS and RCS, and that we must not unmask + * more than one flip event at any time (or ensure that one flip message + * can be sent by waiting for flip-done prior to queueing new flips). + * Experimentation says that BCS works despite DERRMR masking all + * flip-done completion events and that unmasking all planes at once + * for the RCS also doesn't appear to drop events. Setting the DERRMR + * to zero does lead to lockups within MI_DISPLAY_FLIP. + */ + if (ring->id == RCS) { + intel_ring_emit(ring, MI_LOAD_REGISTER_IMM(1)); + intel_ring_emit(ring, DERRMR); + intel_ring_emit(ring, ~(DERRMR_PIPEA_PRI_FLIP_DONE | + DERRMR_PIPEB_PRI_FLIP_DONE | + DERRMR_PIPEC_PRI_FLIP_DONE)); + intel_ring_emit(ring, MI_STORE_REGISTER_MEM(1)); + intel_ring_emit(ring, DERRMR); + intel_ring_emit(ring, ring->scratch.gtt_offset + 256); + } + intel_ring_emit(ring, MI_DISPLAY_FLIP_I915 | plane_bit); intel_ring_emit(ring, (fb->pitches[0] | obj->tiling_mode)); intel_ring_emit(ring, i915_gem_obj_ggtt_offset(obj) + intel_crtc->dspaddr_offset); From fac15c108248e8d592fb8f4cbcf26d98b3485526 Mon Sep 17 00:00:00 2001 From: Joe Perches Date: Thu, 29 Aug 2013 13:11:07 -0700 Subject: [PATCH 10/29] i915_gem: Convert kmem_cache_alloc(...GFP_ZERO) to kmem_cache_zalloc The helper exists, might as well use it instead of __GFP_ZERO. Signed-off-by: Joe Perches Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 2d1cb10d846f..61313054fce6 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -212,7 +212,7 @@ i915_gem_get_aperture_ioctl(struct drm_device *dev, void *data, void *i915_gem_object_alloc(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; - return kmem_cache_alloc(dev_priv->slab, GFP_KERNEL | __GFP_ZERO); + return kmem_cache_zalloc(dev_priv->slab, GFP_KERNEL); } void i915_gem_object_free(struct drm_i915_gem_object *obj) From a0a1807544fe59b42d3760ee912ea4c6741298f5 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Fri, 26 Jul 2013 13:32:51 -0700 Subject: [PATCH 11/29] drm/i915: split PCI IDs out into i915_drm.h v4 For use by userspace (at some point in the future) and other kernel code. v2: move PCI IDs to uabi (Chris) move PCI IDs to drm/ (Dave) v3: fixup Quanta detection - needs to come first (Daniel) v4: fix up PCI match structure init for easier use by userspace (Chris) Signed-off-by: Jesse Barnes Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_drv.c | 164 +++++-------------------- include/drm/i915_drm.h | 2 + include/drm/i915_pciids.h | 211 ++++++++++++++++++++++++++++++++ 3 files changed, 247 insertions(+), 130 deletions(-) create mode 100644 include/drm/i915_pciids.h diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index ccb28ead3501..69d8ed5416c3 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -157,25 +157,6 @@ MODULE_PARM_DESC(prefault_disable, static struct drm_driver driver; extern int intel_agp_enabled; -#define INTEL_VGA_DEVICE(id, info) { \ - .class = PCI_BASE_CLASS_DISPLAY << 16, \ - .class_mask = 0xff0000, \ - .vendor = 0x8086, \ - .device = id, \ - .subvendor = PCI_ANY_ID, \ - .subdevice = PCI_ANY_ID, \ - .driver_data = (unsigned long) info } - -#define INTEL_QUANTA_VGA_DEVICE(info) { \ - .class = PCI_BASE_CLASS_DISPLAY << 16, \ - .class_mask = 0xff0000, \ - .vendor = 0x8086, \ - .device = 0x16a, \ - .subvendor = 0x152d, \ - .subdevice = 0x8990, \ - .driver_data = (unsigned long) info } - - static const struct intel_device_info intel_i830_info = { .gen = 2, .is_mobile = 1, .cursor_needs_physical = 1, .num_pipes = 2, .has_overlay = 1, .overlay_needs_physical = 1, @@ -350,118 +331,41 @@ static const struct intel_device_info intel_haswell_m_info = { .has_vebox_ring = 1, }; +/* + * Make sure any device matches here are from most specific to most + * general. For example, since the Quanta match is based on the subsystem + * and subvendor IDs, we need it to come before the more general IVB + * PCI ID matches, otherwise we'll use the wrong info struct above. + */ +#define INTEL_PCI_IDS \ + INTEL_I830_IDS(&intel_i830_info), \ + INTEL_I845G_IDS(&intel_845g_info), \ + INTEL_I85X_IDS(&intel_i85x_info), \ + INTEL_I865G_IDS(&intel_i865g_info), \ + INTEL_I915G_IDS(&intel_i915g_info), \ + INTEL_I915GM_IDS(&intel_i915gm_info), \ + INTEL_I945G_IDS(&intel_i945g_info), \ + INTEL_I945GM_IDS(&intel_i945gm_info), \ + INTEL_I965G_IDS(&intel_i965g_info), \ + INTEL_G33_IDS(&intel_g33_info), \ + INTEL_I965GM_IDS(&intel_i965gm_info), \ + INTEL_GM45_IDS(&intel_gm45_info), \ + INTEL_G45_IDS(&intel_g45_info), \ + INTEL_PINEVIEW_IDS(&intel_pineview_info), \ + INTEL_IRONLAKE_D_IDS(&intel_ironlake_d_info), \ + INTEL_IRONLAKE_M_IDS(&intel_ironlake_m_info), \ + INTEL_SNB_D_IDS(&intel_sandybridge_d_info), \ + INTEL_SNB_M_IDS(&intel_sandybridge_m_info), \ + INTEL_IVB_Q_IDS(&intel_ivybridge_q_info), /* must be first IVB */ \ + INTEL_IVB_M_IDS(&intel_ivybridge_m_info), \ + INTEL_IVB_D_IDS(&intel_ivybridge_d_info), \ + INTEL_HSW_D_IDS(&intel_haswell_d_info), \ + INTEL_HSW_M_IDS(&intel_haswell_m_info), \ + INTEL_VLV_M_IDS(&intel_valleyview_m_info), \ + INTEL_VLV_D_IDS(&intel_valleyview_d_info) + static const struct pci_device_id pciidlist[] = { /* aka */ - INTEL_VGA_DEVICE(0x3577, &intel_i830_info), /* I830_M */ - INTEL_VGA_DEVICE(0x2562, &intel_845g_info), /* 845_G */ - INTEL_VGA_DEVICE(0x3582, &intel_i85x_info), /* I855_GM */ - INTEL_VGA_DEVICE(0x358e, &intel_i85x_info), - INTEL_VGA_DEVICE(0x2572, &intel_i865g_info), /* I865_G */ - INTEL_VGA_DEVICE(0x2582, &intel_i915g_info), /* I915_G */ - INTEL_VGA_DEVICE(0x258a, &intel_i915g_info), /* E7221_G */ - INTEL_VGA_DEVICE(0x2592, &intel_i915gm_info), /* I915_GM */ - INTEL_VGA_DEVICE(0x2772, &intel_i945g_info), /* I945_G */ - INTEL_VGA_DEVICE(0x27a2, &intel_i945gm_info), /* I945_GM */ - INTEL_VGA_DEVICE(0x27ae, &intel_i945gm_info), /* I945_GME */ - INTEL_VGA_DEVICE(0x2972, &intel_i965g_info), /* I946_GZ */ - INTEL_VGA_DEVICE(0x2982, &intel_i965g_info), /* G35_G */ - INTEL_VGA_DEVICE(0x2992, &intel_i965g_info), /* I965_Q */ - INTEL_VGA_DEVICE(0x29a2, &intel_i965g_info), /* I965_G */ - INTEL_VGA_DEVICE(0x29b2, &intel_g33_info), /* Q35_G */ - INTEL_VGA_DEVICE(0x29c2, &intel_g33_info), /* G33_G */ - INTEL_VGA_DEVICE(0x29d2, &intel_g33_info), /* Q33_G */ - INTEL_VGA_DEVICE(0x2a02, &intel_i965gm_info), /* I965_GM */ - INTEL_VGA_DEVICE(0x2a12, &intel_i965gm_info), /* I965_GME */ - INTEL_VGA_DEVICE(0x2a42, &intel_gm45_info), /* GM45_G */ - INTEL_VGA_DEVICE(0x2e02, &intel_g45_info), /* IGD_E_G */ - INTEL_VGA_DEVICE(0x2e12, &intel_g45_info), /* Q45_G */ - INTEL_VGA_DEVICE(0x2e22, &intel_g45_info), /* G45_G */ - INTEL_VGA_DEVICE(0x2e32, &intel_g45_info), /* G41_G */ - INTEL_VGA_DEVICE(0x2e42, &intel_g45_info), /* B43_G */ - INTEL_VGA_DEVICE(0x2e92, &intel_g45_info), /* B43_G.1 */ - INTEL_VGA_DEVICE(0xa001, &intel_pineview_info), - INTEL_VGA_DEVICE(0xa011, &intel_pineview_info), - INTEL_VGA_DEVICE(0x0042, &intel_ironlake_d_info), - INTEL_VGA_DEVICE(0x0046, &intel_ironlake_m_info), - INTEL_VGA_DEVICE(0x0102, &intel_sandybridge_d_info), - INTEL_VGA_DEVICE(0x0112, &intel_sandybridge_d_info), - INTEL_VGA_DEVICE(0x0122, &intel_sandybridge_d_info), - INTEL_VGA_DEVICE(0x0106, &intel_sandybridge_m_info), - INTEL_VGA_DEVICE(0x0116, &intel_sandybridge_m_info), - INTEL_VGA_DEVICE(0x0126, &intel_sandybridge_m_info), - INTEL_VGA_DEVICE(0x010A, &intel_sandybridge_d_info), - INTEL_VGA_DEVICE(0x0156, &intel_ivybridge_m_info), /* GT1 mobile */ - INTEL_VGA_DEVICE(0x0166, &intel_ivybridge_m_info), /* GT2 mobile */ - INTEL_VGA_DEVICE(0x0152, &intel_ivybridge_d_info), /* GT1 desktop */ - INTEL_VGA_DEVICE(0x0162, &intel_ivybridge_d_info), /* GT2 desktop */ - INTEL_VGA_DEVICE(0x015a, &intel_ivybridge_d_info), /* GT1 server */ - INTEL_QUANTA_VGA_DEVICE(&intel_ivybridge_q_info), /* Quanta transcode */ - INTEL_VGA_DEVICE(0x016a, &intel_ivybridge_d_info), /* GT2 server */ - INTEL_VGA_DEVICE(0x0402, &intel_haswell_d_info), /* GT1 desktop */ - INTEL_VGA_DEVICE(0x0412, &intel_haswell_d_info), /* GT2 desktop */ - INTEL_VGA_DEVICE(0x0422, &intel_haswell_d_info), /* GT3 desktop */ - INTEL_VGA_DEVICE(0x040a, &intel_haswell_d_info), /* GT1 server */ - INTEL_VGA_DEVICE(0x041a, &intel_haswell_d_info), /* GT2 server */ - INTEL_VGA_DEVICE(0x042a, &intel_haswell_d_info), /* GT3 server */ - INTEL_VGA_DEVICE(0x0406, &intel_haswell_m_info), /* GT1 mobile */ - INTEL_VGA_DEVICE(0x0416, &intel_haswell_m_info), /* GT2 mobile */ - INTEL_VGA_DEVICE(0x0426, &intel_haswell_m_info), /* GT2 mobile */ - INTEL_VGA_DEVICE(0x040B, &intel_haswell_d_info), /* GT1 reserved */ - INTEL_VGA_DEVICE(0x041B, &intel_haswell_d_info), /* GT2 reserved */ - INTEL_VGA_DEVICE(0x042B, &intel_haswell_d_info), /* GT3 reserved */ - INTEL_VGA_DEVICE(0x040E, &intel_haswell_d_info), /* GT1 reserved */ - INTEL_VGA_DEVICE(0x041E, &intel_haswell_d_info), /* GT2 reserved */ - INTEL_VGA_DEVICE(0x042E, &intel_haswell_d_info), /* GT3 reserved */ - INTEL_VGA_DEVICE(0x0C02, &intel_haswell_d_info), /* SDV GT1 desktop */ - INTEL_VGA_DEVICE(0x0C12, &intel_haswell_d_info), /* SDV GT2 desktop */ - INTEL_VGA_DEVICE(0x0C22, &intel_haswell_d_info), /* SDV GT3 desktop */ - INTEL_VGA_DEVICE(0x0C0A, &intel_haswell_d_info), /* SDV GT1 server */ - INTEL_VGA_DEVICE(0x0C1A, &intel_haswell_d_info), /* SDV GT2 server */ - INTEL_VGA_DEVICE(0x0C2A, &intel_haswell_d_info), /* SDV GT3 server */ - INTEL_VGA_DEVICE(0x0C06, &intel_haswell_m_info), /* SDV GT1 mobile */ - INTEL_VGA_DEVICE(0x0C16, &intel_haswell_m_info), /* SDV GT2 mobile */ - INTEL_VGA_DEVICE(0x0C26, &intel_haswell_m_info), /* SDV GT3 mobile */ - INTEL_VGA_DEVICE(0x0C0B, &intel_haswell_d_info), /* SDV GT1 reserved */ - INTEL_VGA_DEVICE(0x0C1B, &intel_haswell_d_info), /* SDV GT2 reserved */ - INTEL_VGA_DEVICE(0x0C2B, &intel_haswell_d_info), /* SDV GT3 reserved */ - INTEL_VGA_DEVICE(0x0C0E, &intel_haswell_d_info), /* SDV GT1 reserved */ - INTEL_VGA_DEVICE(0x0C1E, &intel_haswell_d_info), /* SDV GT2 reserved */ - INTEL_VGA_DEVICE(0x0C2E, &intel_haswell_d_info), /* SDV GT3 reserved */ - INTEL_VGA_DEVICE(0x0A02, &intel_haswell_d_info), /* ULT GT1 desktop */ - INTEL_VGA_DEVICE(0x0A12, &intel_haswell_d_info), /* ULT GT2 desktop */ - INTEL_VGA_DEVICE(0x0A22, &intel_haswell_d_info), /* ULT GT3 desktop */ - INTEL_VGA_DEVICE(0x0A0A, &intel_haswell_d_info), /* ULT GT1 server */ - INTEL_VGA_DEVICE(0x0A1A, &intel_haswell_d_info), /* ULT GT2 server */ - INTEL_VGA_DEVICE(0x0A2A, &intel_haswell_d_info), /* ULT GT3 server */ - INTEL_VGA_DEVICE(0x0A06, &intel_haswell_m_info), /* ULT GT1 mobile */ - INTEL_VGA_DEVICE(0x0A16, &intel_haswell_m_info), /* ULT GT2 mobile */ - INTEL_VGA_DEVICE(0x0A26, &intel_haswell_m_info), /* ULT GT3 mobile */ - INTEL_VGA_DEVICE(0x0A0B, &intel_haswell_d_info), /* ULT GT1 reserved */ - INTEL_VGA_DEVICE(0x0A1B, &intel_haswell_d_info), /* ULT GT2 reserved */ - INTEL_VGA_DEVICE(0x0A2B, &intel_haswell_d_info), /* ULT GT3 reserved */ - INTEL_VGA_DEVICE(0x0A0E, &intel_haswell_m_info), /* ULT GT1 reserved */ - INTEL_VGA_DEVICE(0x0A1E, &intel_haswell_m_info), /* ULT GT2 reserved */ - INTEL_VGA_DEVICE(0x0A2E, &intel_haswell_m_info), /* ULT GT3 reserved */ - INTEL_VGA_DEVICE(0x0D02, &intel_haswell_d_info), /* CRW GT1 desktop */ - INTEL_VGA_DEVICE(0x0D12, &intel_haswell_d_info), /* CRW GT2 desktop */ - INTEL_VGA_DEVICE(0x0D22, &intel_haswell_d_info), /* CRW GT3 desktop */ - INTEL_VGA_DEVICE(0x0D0A, &intel_haswell_d_info), /* CRW GT1 server */ - INTEL_VGA_DEVICE(0x0D1A, &intel_haswell_d_info), /* CRW GT2 server */ - INTEL_VGA_DEVICE(0x0D2A, &intel_haswell_d_info), /* CRW GT3 server */ - INTEL_VGA_DEVICE(0x0D06, &intel_haswell_m_info), /* CRW GT1 mobile */ - INTEL_VGA_DEVICE(0x0D16, &intel_haswell_m_info), /* CRW GT2 mobile */ - INTEL_VGA_DEVICE(0x0D26, &intel_haswell_m_info), /* CRW GT3 mobile */ - INTEL_VGA_DEVICE(0x0D0B, &intel_haswell_d_info), /* CRW GT1 reserved */ - INTEL_VGA_DEVICE(0x0D1B, &intel_haswell_d_info), /* CRW GT2 reserved */ - INTEL_VGA_DEVICE(0x0D2B, &intel_haswell_d_info), /* CRW GT3 reserved */ - INTEL_VGA_DEVICE(0x0D0E, &intel_haswell_d_info), /* CRW GT1 reserved */ - INTEL_VGA_DEVICE(0x0D1E, &intel_haswell_d_info), /* CRW GT2 reserved */ - INTEL_VGA_DEVICE(0x0D2E, &intel_haswell_d_info), /* CRW GT3 reserved */ - INTEL_VGA_DEVICE(0x0f30, &intel_valleyview_m_info), - INTEL_VGA_DEVICE(0x0f31, &intel_valleyview_m_info), - INTEL_VGA_DEVICE(0x0f32, &intel_valleyview_m_info), - INTEL_VGA_DEVICE(0x0f33, &intel_valleyview_m_info), - INTEL_VGA_DEVICE(0x0157, &intel_valleyview_m_info), - INTEL_VGA_DEVICE(0x0155, &intel_valleyview_d_info), + INTEL_PCI_IDS, {0, 0, 0} }; diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h index 63d609d8a3f6..7276a72710e2 100644 --- a/include/drm/i915_drm.h +++ b/include/drm/i915_drm.h @@ -26,6 +26,7 @@ #ifndef _I915_DRM_H_ #define _I915_DRM_H_ +#include #include /* For use by IPS driver */ @@ -34,4 +35,5 @@ extern bool i915_gpu_raise(void); extern bool i915_gpu_lower(void); extern bool i915_gpu_busy(void); extern bool i915_gpu_turbo_disable(void); + #endif /* _I915_DRM_H_ */ diff --git a/include/drm/i915_pciids.h b/include/drm/i915_pciids.h new file mode 100644 index 000000000000..8a10f5c354e6 --- /dev/null +++ b/include/drm/i915_pciids.h @@ -0,0 +1,211 @@ +/* + * Copyright 2013 Intel Corporation + * All Rights Reserved. + * + * 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, sub license, 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. + */ +#ifndef _I915_PCIIDS_H +#define _I915_PCIIDS_H + +/* + * A pci_device_id struct { + * __u32 vendor, device; + * __u32 subvendor, subdevice; + * __u32 class, class_mask; + * kernel_ulong_t driver_data; + * }; + * Don't use C99 here because "class" is reserved and we want to + * give userspace flexibility. + */ +#define INTEL_VGA_DEVICE(id, info) { \ + 0x8086, id, \ + ~0, ~0, \ + 0x030000, 0xff0000, \ + (unsigned long) info } + +#define INTEL_QUANTA_VGA_DEVICE(info) { \ + 0x8086, 0x16a, \ + 0x152d, 0x8990, \ + 0x030000, 0xff0000, \ + (unsigned long) info } + +#define INTEL_I830_IDS(info) \ + INTEL_VGA_DEVICE(0x3577, info) + +#define INTEL_I845G_IDS(info) \ + INTEL_VGA_DEVICE(0x2562, info) + +#define INTEL_I85X_IDS(info) \ + INTEL_VGA_DEVICE(0x3582, info), /* I855_GM */ \ + INTEL_VGA_DEVICE(0x358e, info) + +#define INTEL_I865G_IDS(info) \ + INTEL_VGA_DEVICE(0x2572, info) /* I865_G */ + +#define INTEL_I915G_IDS(info) \ + INTEL_VGA_DEVICE(0x2582, info), /* I915_G */ \ + INTEL_VGA_DEVICE(0x258a, info) /* E7221_G */ + +#define INTEL_I915GM_IDS(info) \ + INTEL_VGA_DEVICE(0x2592, info) /* I915_GM */ + +#define INTEL_I945G_IDS(info) \ + INTEL_VGA_DEVICE(0x2772, info) /* I945_G */ + +#define INTEL_I945GM_IDS(info) \ + INTEL_VGA_DEVICE(0x27a2, info), /* I945_GM */ \ + INTEL_VGA_DEVICE(0x27ae, info) /* I945_GME */ + +#define INTEL_I965G_IDS(info) \ + INTEL_VGA_DEVICE(0x2972, info), /* I946_GZ */ \ + INTEL_VGA_DEVICE(0x2982, info), /* G35_G */ \ + INTEL_VGA_DEVICE(0x2992, info), /* I965_Q */ \ + INTEL_VGA_DEVICE(0x29a2, info) /* I965_G */ + +#define INTEL_G33_IDS(info) \ + INTEL_VGA_DEVICE(0x29b2, info), /* Q35_G */ \ + INTEL_VGA_DEVICE(0x29c2, info), /* G33_G */ \ + INTEL_VGA_DEVICE(0x29d2, info) /* Q33_G */ + +#define INTEL_I965GM_IDS(info) \ + INTEL_VGA_DEVICE(0x2a02, info), /* I965_GM */ \ + INTEL_VGA_DEVICE(0x2a12, info) /* I965_GME */ + +#define INTEL_GM45_IDS(info) \ + INTEL_VGA_DEVICE(0x2a42, info) /* GM45_G */ + +#define INTEL_G45_IDS(info) \ + INTEL_VGA_DEVICE(0x2e02, info), /* IGD_E_G */ \ + INTEL_VGA_DEVICE(0x2e12, info), /* Q45_G */ \ + INTEL_VGA_DEVICE(0x2e22, info), /* G45_G */ \ + INTEL_VGA_DEVICE(0x2e32, info), /* G41_G */ \ + INTEL_VGA_DEVICE(0x2e42, info), /* B43_G */ \ + INTEL_VGA_DEVICE(0x2e92, info) /* B43_G.1 */ + +#define INTEL_PINEVIEW_IDS(info) \ + INTEL_VGA_DEVICE(0xa001, info), \ + INTEL_VGA_DEVICE(0xa011, info) + +#define INTEL_IRONLAKE_D_IDS(info) \ + INTEL_VGA_DEVICE(0x0042, info) + +#define INTEL_IRONLAKE_M_IDS(info) \ + INTEL_VGA_DEVICE(0x0046, info) + +#define INTEL_SNB_D_IDS(info) \ + INTEL_VGA_DEVICE(0x0102, info), \ + INTEL_VGA_DEVICE(0x0112, info), \ + INTEL_VGA_DEVICE(0x0122, info), \ + INTEL_VGA_DEVICE(0x010A, info) + +#define INTEL_SNB_M_IDS(info) \ + INTEL_VGA_DEVICE(0x0106, info), \ + INTEL_VGA_DEVICE(0x0116, info), \ + INTEL_VGA_DEVICE(0x0126, info) + +#define INTEL_IVB_M_IDS(info) \ + INTEL_VGA_DEVICE(0x0156, info), /* GT1 mobile */ \ + INTEL_VGA_DEVICE(0x0166, info) /* GT2 mobile */ + +#define INTEL_IVB_D_IDS(info) \ + INTEL_VGA_DEVICE(0x0152, info), /* GT1 desktop */ \ + INTEL_VGA_DEVICE(0x0162, info), /* GT2 desktop */ \ + INTEL_VGA_DEVICE(0x015a, info), /* GT1 server */ \ + INTEL_VGA_DEVICE(0x016a, info) /* GT2 server */ + +#define INTEL_IVB_Q_IDS(info) \ + INTEL_QUANTA_VGA_DEVICE(info) /* Quanta transcode */ + +#define INTEL_HSW_D_IDS(info) \ + INTEL_VGA_DEVICE(0x0402, info), /* GT1 desktop */ \ + INTEL_VGA_DEVICE(0x0412, info), /* GT2 desktop */ \ + INTEL_VGA_DEVICE(0x0422, info), /* GT3 desktop */ \ + INTEL_VGA_DEVICE(0x040a, info), /* GT1 server */ \ + INTEL_VGA_DEVICE(0x041a, info), /* GT2 server */ \ + INTEL_VGA_DEVICE(0x042a, info), /* GT3 server */ \ + INTEL_VGA_DEVICE(0x040B, info), /* GT1 reserved */ \ + INTEL_VGA_DEVICE(0x041B, info), /* GT2 reserved */ \ + INTEL_VGA_DEVICE(0x042B, info), /* GT3 reserved */ \ + INTEL_VGA_DEVICE(0x040E, info), /* GT1 reserved */ \ + INTEL_VGA_DEVICE(0x041E, info), /* GT2 reserved */ \ + INTEL_VGA_DEVICE(0x042E, info), /* GT3 reserved */ \ + INTEL_VGA_DEVICE(0x0C02, info), /* SDV GT1 desktop */ \ + INTEL_VGA_DEVICE(0x0C12, info), /* SDV GT2 desktop */ \ + INTEL_VGA_DEVICE(0x0C22, info), /* SDV GT3 desktop */ \ + INTEL_VGA_DEVICE(0x0C0A, info), /* SDV GT1 server */ \ + INTEL_VGA_DEVICE(0x0C1A, info), /* SDV GT2 server */ \ + INTEL_VGA_DEVICE(0x0C2A, info), /* SDV GT3 server */ \ + INTEL_VGA_DEVICE(0x0C0B, info), /* SDV GT1 reserved */ \ + INTEL_VGA_DEVICE(0x0C1B, info), /* SDV GT2 reserved */ \ + INTEL_VGA_DEVICE(0x0C2B, info), /* SDV GT3 reserved */ \ + INTEL_VGA_DEVICE(0x0C0E, info), /* SDV GT1 reserved */ \ + INTEL_VGA_DEVICE(0x0C1E, info), /* SDV GT2 reserved */ \ + INTEL_VGA_DEVICE(0x0C2E, info), /* SDV GT3 reserved */ \ + INTEL_VGA_DEVICE(0x0A02, info), /* ULT GT1 desktop */ \ + INTEL_VGA_DEVICE(0x0A12, info), /* ULT GT2 desktop */ \ + INTEL_VGA_DEVICE(0x0A22, info), /* ULT GT3 desktop */ \ + INTEL_VGA_DEVICE(0x0A0A, info), /* ULT GT1 server */ \ + INTEL_VGA_DEVICE(0x0A1A, info), /* ULT GT2 server */ \ + INTEL_VGA_DEVICE(0x0A2A, info), /* ULT GT3 server */ \ + INTEL_VGA_DEVICE(0x0A0B, info), /* ULT GT1 reserved */ \ + INTEL_VGA_DEVICE(0x0A1B, info), /* ULT GT2 reserved */ \ + INTEL_VGA_DEVICE(0x0A2B, info), /* ULT GT3 reserved */ \ + INTEL_VGA_DEVICE(0x0D02, info), /* CRW GT1 desktop */ \ + INTEL_VGA_DEVICE(0x0D12, info), /* CRW GT2 desktop */ \ + INTEL_VGA_DEVICE(0x0D22, info), /* CRW GT3 desktop */ \ + INTEL_VGA_DEVICE(0x0D0A, info), /* CRW GT1 server */ \ + INTEL_VGA_DEVICE(0x0D1A, info), /* CRW GT2 server */ \ + INTEL_VGA_DEVICE(0x0D2A, info), /* CRW GT3 server */ \ + INTEL_VGA_DEVICE(0x0D0B, info), /* CRW GT1 reserved */ \ + INTEL_VGA_DEVICE(0x0D1B, info), /* CRW GT2 reserved */ \ + INTEL_VGA_DEVICE(0x0D2B, info), /* CRW GT3 reserved */ \ + INTEL_VGA_DEVICE(0x0D0E, info), /* CRW GT1 reserved */ \ + INTEL_VGA_DEVICE(0x0D1E, info), /* CRW GT2 reserved */ \ + INTEL_VGA_DEVICE(0x0D2E, info) /* CRW GT3 reserved */ \ + +#define INTEL_HSW_M_IDS(info) \ + INTEL_VGA_DEVICE(0x0406, info), /* GT1 mobile */ \ + INTEL_VGA_DEVICE(0x0416, info), /* GT2 mobile */ \ + INTEL_VGA_DEVICE(0x0426, info), /* GT2 mobile */ \ + INTEL_VGA_DEVICE(0x0C06, info), /* SDV GT1 mobile */ \ + INTEL_VGA_DEVICE(0x0C16, info), /* SDV GT2 mobile */ \ + INTEL_VGA_DEVICE(0x0C26, info), /* SDV GT3 mobile */ \ + INTEL_VGA_DEVICE(0x0A06, info), /* ULT GT1 mobile */ \ + INTEL_VGA_DEVICE(0x0A16, info), /* ULT GT2 mobile */ \ + INTEL_VGA_DEVICE(0x0A26, info), /* ULT GT3 mobile */ \ + INTEL_VGA_DEVICE(0x0A0E, info), /* ULT GT1 reserved */ \ + INTEL_VGA_DEVICE(0x0A1E, info), /* ULT GT2 reserved */ \ + INTEL_VGA_DEVICE(0x0A2E, info), /* ULT GT3 reserved */ \ + INTEL_VGA_DEVICE(0x0D06, info), /* CRW GT1 mobile */ \ + INTEL_VGA_DEVICE(0x0D16, info), /* CRW GT2 mobile */ \ + INTEL_VGA_DEVICE(0x0D26, info) /* CRW GT3 mobile */ + +#define INTEL_VLV_M_IDS(info) \ + INTEL_VGA_DEVICE(0x0f30, info), \ + INTEL_VGA_DEVICE(0x0f31, info), \ + INTEL_VGA_DEVICE(0x0f32, info), \ + INTEL_VGA_DEVICE(0x0f33, info), \ + INTEL_VGA_DEVICE(0x0157, info) + +#define INTEL_VLV_D_IDS(info) \ + INTEL_VGA_DEVICE(0x0155, info) + +#endif /* _I915_PCIIDS_H */ From 814c5f1f52a4beb3710317022acd6ad34fc0b6b9 Mon Sep 17 00:00:00 2001 From: Jesse Barnes Date: Fri, 26 Jul 2013 13:32:52 -0700 Subject: [PATCH 12/29] x86: add early quirk for reserving Intel graphics stolen memory v5 Systems with Intel graphics controllers set aside memory exclusively for gfx driver use. This memory is not always marked in the E820 as reserved or as RAM, and so is subject to overlap from E820 manipulation later in the boot process. On some systems, MMIO space is allocated on top, despite the efforts of the "RAM buffer" approach, which simply rounds memory boundaries up to 64M to try to catch space that may decode as RAM and so is not suitable for MMIO. v2: use read_pci_config for 32 bit reads instead of adding a new one (Chris) add gen6 stolen size function (Chris) v3: use a function pointer (Chris) drop gen2 bits (Daniel) v4: call e820_sanitize_map after adding the region v5: fixup comments (Peter) simplify loop (Chris) Acked-by: Ingo Molnar Signed-off-by: Jesse Barnes Acked-by: H. Peter Anvin Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66726 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=66844 Signed-off-by: Daniel Vetter --- arch/x86/kernel/early-quirks.c | 154 ++++++++++++++++++++++++++++++++ drivers/gpu/drm/i915/i915_reg.h | 15 ---- include/drm/i915_drm.h | 32 +++++++ 3 files changed, 186 insertions(+), 15 deletions(-) diff --git a/arch/x86/kernel/early-quirks.c b/arch/x86/kernel/early-quirks.c index 63bdb29b2549..b3cd3ebae077 100644 --- a/arch/x86/kernel/early-quirks.c +++ b/arch/x86/kernel/early-quirks.c @@ -12,6 +12,7 @@ #include #include #include +#include #include #include #include @@ -216,6 +217,157 @@ static void __init intel_remapping_check(int num, int slot, int func) } +/* + * Systems with Intel graphics controllers set aside memory exclusively + * for gfx driver use. This memory is not marked in the E820 as reserved + * or as RAM, and so is subject to overlap from E820 manipulation later + * in the boot process. On some systems, MMIO space is allocated on top, + * despite the efforts of the "RAM buffer" approach, which simply rounds + * memory boundaries up to 64M to try to catch space that may decode + * as RAM and so is not suitable for MMIO. + * + * And yes, so far on current devices the base addr is always under 4G. + */ +static u32 __init intel_stolen_base(int num, int slot, int func) +{ + u32 base; + + /* + * For the PCI IDs in this quirk, the stolen base is always + * in 0x5c, aka the BDSM register (yes that's really what + * it's called). + */ + base = read_pci_config(num, slot, func, 0x5c); + base &= ~((1<<20) - 1); + + return base; +} + +#define KB(x) ((x) * 1024) +#define MB(x) (KB (KB (x))) +#define GB(x) (MB (KB (x))) + +static size_t __init gen3_stolen_size(int num, int slot, int func) +{ + size_t stolen_size; + u16 gmch_ctrl; + + gmch_ctrl = read_pci_config_16(0, 0, 0, I830_GMCH_CTRL); + + switch (gmch_ctrl & I855_GMCH_GMS_MASK) { + case I855_GMCH_GMS_STOLEN_1M: + stolen_size = MB(1); + break; + case I855_GMCH_GMS_STOLEN_4M: + stolen_size = MB(4); + break; + case I855_GMCH_GMS_STOLEN_8M: + stolen_size = MB(8); + break; + case I855_GMCH_GMS_STOLEN_16M: + stolen_size = MB(16); + break; + case I855_GMCH_GMS_STOLEN_32M: + stolen_size = MB(32); + break; + case I915_GMCH_GMS_STOLEN_48M: + stolen_size = MB(48); + break; + case I915_GMCH_GMS_STOLEN_64M: + stolen_size = MB(64); + break; + case G33_GMCH_GMS_STOLEN_128M: + stolen_size = MB(128); + break; + case G33_GMCH_GMS_STOLEN_256M: + stolen_size = MB(256); + break; + case INTEL_GMCH_GMS_STOLEN_96M: + stolen_size = MB(96); + break; + case INTEL_GMCH_GMS_STOLEN_160M: + stolen_size = MB(160); + break; + case INTEL_GMCH_GMS_STOLEN_224M: + stolen_size = MB(224); + break; + case INTEL_GMCH_GMS_STOLEN_352M: + stolen_size = MB(352); + break; + default: + stolen_size = 0; + break; + } + + return stolen_size; +} + +static size_t __init gen6_stolen_size(int num, int slot, int func) +{ + u16 gmch_ctrl; + + gmch_ctrl = read_pci_config_16(num, slot, func, SNB_GMCH_CTRL); + gmch_ctrl >>= SNB_GMCH_GMS_SHIFT; + gmch_ctrl &= SNB_GMCH_GMS_MASK; + + return gmch_ctrl << 25; /* 32 MB units */ +} + +typedef size_t (*stolen_size_fn)(int num, int slot, int func); + +static struct pci_device_id intel_stolen_ids[] __initdata = { + INTEL_I915G_IDS(gen3_stolen_size), + INTEL_I915GM_IDS(gen3_stolen_size), + INTEL_I945G_IDS(gen3_stolen_size), + INTEL_I945GM_IDS(gen3_stolen_size), + INTEL_VLV_M_IDS(gen3_stolen_size), + INTEL_VLV_D_IDS(gen3_stolen_size), + INTEL_PINEVIEW_IDS(gen3_stolen_size), + INTEL_I965G_IDS(gen3_stolen_size), + INTEL_G33_IDS(gen3_stolen_size), + INTEL_I965GM_IDS(gen3_stolen_size), + INTEL_GM45_IDS(gen3_stolen_size), + INTEL_G45_IDS(gen3_stolen_size), + INTEL_IRONLAKE_D_IDS(gen3_stolen_size), + INTEL_IRONLAKE_M_IDS(gen3_stolen_size), + INTEL_SNB_D_IDS(gen6_stolen_size), + INTEL_SNB_M_IDS(gen6_stolen_size), + INTEL_IVB_M_IDS(gen6_stolen_size), + INTEL_IVB_D_IDS(gen6_stolen_size), + INTEL_HSW_D_IDS(gen6_stolen_size), + INTEL_HSW_M_IDS(gen6_stolen_size), +}; + +static void __init intel_graphics_stolen(int num, int slot, int func) +{ + size_t size; + int i; + u32 start; + u16 device, subvendor, subdevice; + + device = read_pci_config_16(num, slot, func, PCI_DEVICE_ID); + subvendor = read_pci_config_16(num, slot, func, + PCI_SUBSYSTEM_VENDOR_ID); + subdevice = read_pci_config_16(num, slot, func, PCI_SUBSYSTEM_ID); + + for (i = 0; i < ARRAY_SIZE(intel_stolen_ids); i++) { + if (intel_stolen_ids[i].device == device) { + stolen_size_fn stolen_size = + (stolen_size_fn)intel_stolen_ids[i].driver_data; + size = stolen_size(num, slot, func); + start = intel_stolen_base(num, slot, func); + if (size && start) { + /* Mark this space as reserved */ + e820_add_region(start, size, E820_RESERVED); + sanitize_e820_map(e820.map, + ARRAY_SIZE(e820.map), + &e820.nr_map); + } + return; + } + } +} + #define QFLAG_APPLY_ONCE 0x1 #define QFLAG_APPLIED 0x2 #define QFLAG_DONE (QFLAG_APPLY_ONCE|QFLAG_APPLIED) @@ -251,6 +403,8 @@ static struct chipset early_qrk[] __initdata = { PCI_BASE_CLASS_BRIDGE, 0, intel_remapping_check }, { PCI_VENDOR_ID_INTEL, 0x3406, PCI_CLASS_BRIDGE_HOST, PCI_BASE_CLASS_BRIDGE, 0, intel_remapping_check }, + { PCI_VENDOR_ID_INTEL, PCI_ANY_ID, PCI_CLASS_DISPLAY_VGA, PCI_ANY_ID, + QFLAG_APPLY_ONCE, intel_graphics_stolen }, {} }; diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index b2fa2a4c4454..dcc58ea157fe 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -33,21 +33,6 @@ #define _MASKED_BIT_ENABLE(a) (((a) << 16) | (a)) #define _MASKED_BIT_DISABLE(a) ((a) << 16) -/* - * The Bridge device's PCI config space has information about the - * fb aperture size and the amount of pre-reserved memory. - * This is all handled in the intel-gtt.ko module. i915.ko only - * cares about the vga bit for the vga rbiter. - */ -#define INTEL_GMCH_CTRL 0x52 -#define INTEL_GMCH_VGA_DISABLE (1 << 1) -#define SNB_GMCH_CTRL 0x50 -#define SNB_GMCH_GGMS_SHIFT 8 /* GTT Graphics Memory Size */ -#define SNB_GMCH_GGMS_MASK 0x3 -#define SNB_GMCH_GMS_SHIFT 3 /* Graphics Mode Select */ -#define SNB_GMCH_GMS_MASK 0x1f - - /* PCI config space */ #define HPLLCC 0xc0 /* 855 only */ diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h index 7276a72710e2..3abfa6ea226e 100644 --- a/include/drm/i915_drm.h +++ b/include/drm/i915_drm.h @@ -36,4 +36,36 @@ extern bool i915_gpu_lower(void); extern bool i915_gpu_busy(void); extern bool i915_gpu_turbo_disable(void); +/* + * The Bridge device's PCI config space has information about the + * fb aperture size and the amount of pre-reserved memory. + * This is all handled in the intel-gtt.ko module. i915.ko only + * cares about the vga bit for the vga rbiter. + */ +#define INTEL_GMCH_CTRL 0x52 +#define INTEL_GMCH_VGA_DISABLE (1 << 1) +#define SNB_GMCH_CTRL 0x50 +#define SNB_GMCH_GGMS_SHIFT 8 /* GTT Graphics Memory Size */ +#define SNB_GMCH_GGMS_MASK 0x3 +#define SNB_GMCH_GMS_SHIFT 3 /* Graphics Mode Select */ +#define SNB_GMCH_GMS_MASK 0x1f + +#define I830_GMCH_CTRL 0x52 + +#define I855_GMCH_GMS_MASK 0xF0 +#define I855_GMCH_GMS_STOLEN_0M 0x0 +#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) +#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) +#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) +#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) +#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) +#define I915_GMCH_GMS_STOLEN_48M (0x6 << 4) +#define I915_GMCH_GMS_STOLEN_64M (0x7 << 4) +#define G33_GMCH_GMS_STOLEN_128M (0x8 << 4) +#define G33_GMCH_GMS_STOLEN_256M (0x9 << 4) +#define INTEL_GMCH_GMS_STOLEN_96M (0xa << 4) +#define INTEL_GMCH_GMS_STOLEN_160M (0xb << 4) +#define INTEL_GMCH_GMS_STOLEN_224M (0xc << 4) +#define INTEL_GMCH_GMS_STOLEN_352M (0xd << 4) + #endif /* _I915_DRM_H_ */ From 1f5d76dbb636c73912c9ff1c90ff46dd2273f098 Mon Sep 17 00:00:00 2001 From: Paulo Zanoni Date: Fri, 23 Aug 2013 19:51:28 -0300 Subject: [PATCH 13/29] drm/i915: enable trickle feed on Haswell MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit We shouldn't disable the trickle feed bits on Haswell. Our documentation explicitly says the trickle feed bits of PRI_CTL and CUR_CTL should not be programmed to 1, and the hardware engineer also asked us to not program the SPR_CTL field to 1. Leaving the bits as 1 could cause underflows. Reported-by: Arthur Runyan Signed-off-by: Paulo Zanoni Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_reg.h | 1 + drivers/gpu/drm/i915/intel_display.c | 10 +++++++--- drivers/gpu/drm/i915/intel_pm.c | 2 -- drivers/gpu/drm/i915/intel_sprite.c | 7 +++++-- 4 files changed, 13 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index dcc58ea157fe..c159e1a6810f 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3313,6 +3313,7 @@ #define MCURSOR_PIPE_A 0x00 #define MCURSOR_PIPE_B (1 << 28) #define MCURSOR_GAMMA_ENABLE (1 << 26) +#define CURSOR_TRICKLE_FEED_DISABLE (1 << 14) #define _CURABASE (dev_priv->info->display_mmio_offset + 0x70084) #define _CURAPOS (dev_priv->info->display_mmio_offset + 0x70088) #define CURSOR_POS_MASK 0x007FF diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index a98392f9894e..a1e502b07266 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -2077,8 +2077,10 @@ static int ironlake_update_plane(struct drm_crtc *crtc, else dspcntr &= ~DISPPLANE_TILED; - /* must disable */ - dspcntr |= DISPPLANE_TRICKLE_FEED_DISABLE; + if (IS_HASWELL(dev)) + dspcntr &= ~DISPPLANE_TRICKLE_FEED_DISABLE; + else + dspcntr |= DISPPLANE_TRICKLE_FEED_DISABLE; I915_WRITE(reg, dspcntr); @@ -6762,8 +6764,10 @@ static void ivb_update_cursor(struct drm_crtc *crtc, u32 base) cntl &= ~(CURSOR_MODE | MCURSOR_GAMMA_ENABLE); cntl |= CURSOR_MODE_DISABLE; } - if (IS_HASWELL(dev)) + if (IS_HASWELL(dev)) { cntl |= CURSOR_PIPE_CSC_ENABLE; + cntl &= ~CURSOR_TRICKLE_FEED_DISABLE; + } I915_WRITE(CURCNTR_IVB(pipe), cntl); intel_crtc->cursor_visible = visible; diff --git a/drivers/gpu/drm/i915/intel_pm.c b/drivers/gpu/drm/i915/intel_pm.c index 6b1d00389952..0c115cc4899f 100644 --- a/drivers/gpu/drm/i915/intel_pm.c +++ b/drivers/gpu/drm/i915/intel_pm.c @@ -4960,8 +4960,6 @@ static void haswell_init_clock_gating(struct drm_device *dev) I915_READ(GEN7_SQ_CHICKEN_MBCUNIT_CONFIG) | GEN7_SQ_CHICKEN_MBCUNIT_SQINTMOB); - g4x_disable_trickle_feed(dev); - /* WaVSRefCountFullforceMissDisable:hsw */ gen7_setup_fixed_func_scheduler(dev_priv); diff --git a/drivers/gpu/drm/i915/intel_sprite.c b/drivers/gpu/drm/i915/intel_sprite.c index 78b621cdd108..ad6ec4b39005 100644 --- a/drivers/gpu/drm/i915/intel_sprite.c +++ b/drivers/gpu/drm/i915/intel_sprite.c @@ -260,8 +260,11 @@ ivb_update_plane(struct drm_plane *plane, struct drm_crtc *crtc, if (obj->tiling_mode != I915_TILING_NONE) sprctl |= SPRITE_TILED; - /* must disable */ - sprctl |= SPRITE_TRICKLE_FEED_DISABLE; + if (IS_HASWELL(dev)) + sprctl &= ~SPRITE_TRICKLE_FEED_DISABLE; + else + sprctl |= SPRITE_TRICKLE_FEED_DISABLE; + sprctl |= SPRITE_ENABLE; if (IS_HASWELL(dev)) From 5cfacdedb1a94efd29faeaab53f939554a3f5943 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 26 Aug 2013 19:50:55 -0300 Subject: [PATCH 14/29] drm/i915: Pin pages whilst mapping the dma-buf As we attempt to kmalloc after calling get_pages, there is a possibility that the shrinker may reap the pages we just acquired. To prevent this we need to increment the pages_pin_count early, so rearrange the code and error paths to make it so. Signed-off-by: Chris Wilson Reviewed-by: Damien Lespiau Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem_dmabuf.c | 41 ++++++++++++++------------ 1 file changed, 22 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_dmabuf.c b/drivers/gpu/drm/i915/i915_gem_dmabuf.c index e918b05fcbdd..7d5752fda5f1 100644 --- a/drivers/gpu/drm/i915/i915_gem_dmabuf.c +++ b/drivers/gpu/drm/i915/i915_gem_dmabuf.c @@ -42,27 +42,24 @@ static struct sg_table *i915_gem_map_dma_buf(struct dma_buf_attachment *attachme ret = i915_mutex_lock_interruptible(obj->base.dev); if (ret) - return ERR_PTR(ret); + goto err; ret = i915_gem_object_get_pages(obj); - if (ret) { - st = ERR_PTR(ret); - goto out; - } + if (ret) + goto err_unlock; + + i915_gem_object_pin_pages(obj); /* Copy sg so that we make an independent mapping */ st = kmalloc(sizeof(struct sg_table), GFP_KERNEL); if (st == NULL) { - st = ERR_PTR(-ENOMEM); - goto out; + ret = -ENOMEM; + goto err_unpin; } ret = sg_alloc_table(st, obj->pages->nents, GFP_KERNEL); - if (ret) { - kfree(st); - st = ERR_PTR(ret); - goto out; - } + if (ret) + goto err_free; src = obj->pages->sgl; dst = st->sgl; @@ -73,17 +70,23 @@ static struct sg_table *i915_gem_map_dma_buf(struct dma_buf_attachment *attachme } if (!dma_map_sg(attachment->dev, st->sgl, st->nents, dir)) { - sg_free_table(st); - kfree(st); - st = ERR_PTR(-ENOMEM); - goto out; + ret =-ENOMEM; + goto err_free_sg; } - i915_gem_object_pin_pages(obj); - -out: mutex_unlock(&obj->base.dev->struct_mutex); return st; + +err_free_sg: + sg_free_table(st); +err_free: + kfree(st); +err_unpin: + i915_gem_object_unpin_pages(obj); +err_unlock: + mutex_unlock(&obj->base.dev->struct_mutex); +err: + return ERR_PTR(ret); } static void i915_gem_unmap_dma_buf(struct dma_buf_attachment *attachment, From f22d776f3e280e605819fc09fc35db0d802d36ce Mon Sep 17 00:00:00 2001 From: Alex Williamson Date: Thu, 15 Aug 2013 16:37:53 -0600 Subject: [PATCH 15/29] vgaarb: Don't disable resources that are not owned MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit If a device does not own a resource then we don't need to disable it. This resolves the case where an Intel IGD device can be configured to disable decode of VGA memory but we still need the arbiter to handle VGA I/O port routing. When the IGD device is in conflict, only PCI_COMMAND_IO should be disabled since VGA memory does not require arbitration on this device. Signed-off-by: Alex Williamson Cc: Dave Airlie Acked-by: Dave Airlie Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter --- drivers/gpu/vga/vgaarb.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/vga/vgaarb.c b/drivers/gpu/vga/vgaarb.c index e893f6e1937d..ea564711bb80 100644 --- a/drivers/gpu/vga/vgaarb.c +++ b/drivers/gpu/vga/vgaarb.c @@ -257,9 +257,9 @@ static struct vga_device *__vga_tryget(struct vga_device *vgadev, if (!conflict->bridge_has_one_vga) { vga_irq_set_state(conflict, false); flags |= PCI_VGA_STATE_CHANGE_DECODES; - if (lwants & (VGA_RSRC_LEGACY_MEM|VGA_RSRC_NORMAL_MEM)) + if (match & (VGA_RSRC_LEGACY_MEM|VGA_RSRC_NORMAL_MEM)) pci_bits |= PCI_COMMAND_MEMORY; - if (lwants & (VGA_RSRC_LEGACY_IO|VGA_RSRC_NORMAL_IO)) + if (match & (VGA_RSRC_LEGACY_IO|VGA_RSRC_NORMAL_IO)) pci_bits |= PCI_COMMAND_IO; } @@ -267,11 +267,11 @@ static struct vga_device *__vga_tryget(struct vga_device *vgadev, flags |= PCI_VGA_STATE_CHANGE_BRIDGE; pci_set_vga_state(conflict->pdev, false, pci_bits, flags); - conflict->owns &= ~lwants; + conflict->owns &= ~match; /* If he also owned non-legacy, that is no longer the case */ - if (lwants & VGA_RSRC_LEGACY_MEM) + if (match & VGA_RSRC_LEGACY_MEM) conflict->owns &= ~VGA_RSRC_NORMAL_MEM; - if (lwants & VGA_RSRC_LEGACY_IO) + if (match & VGA_RSRC_LEGACY_IO) conflict->owns &= ~VGA_RSRC_NORMAL_IO; } From 5c0f6ee766d7c689e71a7df5e5fa87941affd75a Mon Sep 17 00:00:00 2001 From: Alex Williamson Date: Thu, 15 Aug 2013 16:37:59 -0600 Subject: [PATCH 16/29] vgaarb: Fix VGA decodes changes MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When VGA decodes change we need to do a bit more evaluation of exactly what has changed. We don't necessarily give up all the old owns resources and we need to account for resources with locks. The new algorithm is: If something is added, update decodes. If legacy resources were added and none were there before, we have a new participant. If something is removed, update decodes. If we previously owned it, we no longer own it. If it was previously locked, invalidate all locks and release it. If legacy resources were removed and none are left, remove the participant from VGA arbitration. Previously we updated decodes, released ownership of everything that was previously decoded, ignored all locks, and went off looking for another device to transfer VGA to. In a test case where Intel IGD removes only legacy VGA memory decoding, this left the arbiter switching to discrete graphics without actually disabling legacy VGA IO from the IGD. As a bonus, we bumped up the count of VGA arbitration participants for no good reason. Signed-off-by: Alex Williamson Cc: Dave Airlie Acked-by: Dave Airlie Reviewed-by: Ville Syrjälä [danvet: Kill now unused variables, reported by the 0-day kernel builtbot.] Signed-off-by: Daniel Vetter --- drivers/gpu/vga/vgaarb.c | 41 +++++++++++++++++----------------------- 1 file changed, 17 insertions(+), 24 deletions(-) diff --git a/drivers/gpu/vga/vgaarb.c b/drivers/gpu/vga/vgaarb.c index ea564711bb80..af0259708358 100644 --- a/drivers/gpu/vga/vgaarb.c +++ b/drivers/gpu/vga/vgaarb.c @@ -644,10 +644,12 @@ bail: static inline void vga_update_device_decodes(struct vga_device *vgadev, int new_decodes) { - int old_decodes; - struct vga_device *new_vgadev, *conflict; + int old_decodes, decodes_removed, decodes_unlocked; old_decodes = vgadev->decodes; + decodes_removed = ~new_decodes & old_decodes; + decodes_unlocked = vgadev->locks & decodes_removed; + vgadev->owns &= ~decodes_removed; vgadev->decodes = new_decodes; pr_info("vgaarb: device changed decodes: PCI:%s,olddecodes=%s,decodes=%s:owns=%s\n", @@ -656,31 +658,22 @@ static inline void vga_update_device_decodes(struct vga_device *vgadev, vga_iostate_to_str(vgadev->decodes), vga_iostate_to_str(vgadev->owns)); - - /* if we own the decodes we should move them along to - another card */ - if ((vgadev->owns & old_decodes) && (vga_count > 1)) { - /* set us to own nothing */ - vgadev->owns &= ~old_decodes; - list_for_each_entry(new_vgadev, &vga_list, list) { - if ((new_vgadev != vgadev) && - (new_vgadev->decodes & VGA_RSRC_LEGACY_MASK)) { - pr_info("vgaarb: transferring owner from PCI:%s to PCI:%s\n", pci_name(vgadev->pdev), pci_name(new_vgadev->pdev)); - conflict = __vga_tryget(new_vgadev, VGA_RSRC_LEGACY_MASK); - if (!conflict) - __vga_put(new_vgadev, VGA_RSRC_LEGACY_MASK); - break; - } - } + /* if we removed locked decodes, lock count goes to zero, and release */ + if (decodes_unlocked) { + if (decodes_unlocked & VGA_RSRC_LEGACY_IO) + vgadev->io_lock_cnt = 0; + if (decodes_unlocked & VGA_RSRC_LEGACY_MEM) + vgadev->mem_lock_cnt = 0; + __vga_put(vgadev, decodes_unlocked); } /* change decodes counter */ - if (old_decodes != new_decodes) { - if (new_decodes & (VGA_RSRC_LEGACY_IO | VGA_RSRC_LEGACY_MEM)) - vga_decode_count++; - else - vga_decode_count--; - } + if (old_decodes & VGA_RSRC_LEGACY_MASK && + !(new_decodes & VGA_RSRC_LEGACY_MASK)) + vga_decode_count--; + if (!(old_decodes & VGA_RSRC_LEGACY_MASK) && + new_decodes & VGA_RSRC_LEGACY_MASK) + vga_decode_count++; pr_debug("vgaarb: decoding count now is: %d\n", vga_decode_count); } From 81b5c7bc8de3e6f63419139c2fc91bf81dea8a7d Mon Sep 17 00:00:00 2001 From: Alex Williamson Date: Wed, 28 Aug 2013 09:39:08 -0600 Subject: [PATCH 17/29] i915: Update VGA arbiter support for newer devices MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is intended to add VGA arbiter support for Intel HD graphics on Core processors. The old GMCH registers no longer exist, so even though it appears that i915 participates in VGA arbitration, it doesn't work. On Intel HD graphics we already attempt to disable VGA regions of the device. This makes registering as a VGA client unnecessary since we don't intend to operate differently depending on how many VGA devices are present. We can disable VGA memory regions by clearing the memory enable bit in the VGA MSR. That only leaves VGA IO, which we update the VGA arbiter to know that we don't participate in VGA memory arbitration. We also add a hook on unload to re-enable memory and reinstate VGA memory arbitration. v3: Use explicit LEGACY_IO | LEGACY_MEM when restoring rather than LEGACY_MASK, per Ville's comments. v2: I915_READ/WRITE accessors don't work in i915_disable_vga, use inb/outb directly. Also, on the driver unbind VGA enable path, acquire legacy IO to re-enable VGA memory. Correct comment. Signed-off-by: Alex Williamson Reviewed-by: Ville Syrjälä [danvet: Add patch changelog. Also squash in a fixup to have a dummy static inline for vga_set_legacy_decoding for CONFIG_VGA_ARB=n as reported by the 0-day kernel build bot.] Signed-off-by: Daniel Vetter fixup 2 --- drivers/gpu/drm/i915/i915_dma.c | 9 ++++++--- drivers/gpu/drm/i915/intel_display.c | 25 +++++++++++++++++++++++++ include/linux/vgaarb.h | 7 +++++++ 3 files changed, 38 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index fdaa0915ce56..3de60503378e 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1290,9 +1290,12 @@ static int i915_load_modeset_init(struct drm_device *dev) * then we do not take part in VGA arbitration and the * vga_client_register() fails with -ENODEV. */ - ret = vga_client_register(dev->pdev, dev, NULL, i915_vga_set_decode); - if (ret && ret != -ENODEV) - goto out; + if (!HAS_PCH_SPLIT(dev)) { + ret = vga_client_register(dev->pdev, dev, NULL, + i915_vga_set_decode); + if (ret && ret != -ENODEV) + goto out; + } intel_register_dsm_handler(); diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index a1e502b07266..73dec62f871f 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10041,6 +10041,15 @@ static void i915_disable_vga(struct drm_device *dev) outb(SR01, VGA_SR_INDEX); sr1 = inb(VGA_SR_DATA); outb(sr1 | 1<<5, VGA_SR_DATA); + + /* Disable VGA memory on Intel HD */ + if (HAS_PCH_SPLIT(dev)) { + outb(inb(VGA_MSR_READ) & ~VGA_MSR_MEM_EN, VGA_MSR_WRITE); + vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO | + VGA_RSRC_NORMAL_IO | + VGA_RSRC_NORMAL_MEM); + } + vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); udelay(300); @@ -10048,6 +10057,20 @@ static void i915_disable_vga(struct drm_device *dev) POSTING_READ(vga_reg); } +static void i915_enable_vga(struct drm_device *dev) +{ + /* Enable VGA memory on Intel HD */ + if (HAS_PCH_SPLIT(dev)) { + vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO); + outb(inb(VGA_MSR_READ) | VGA_MSR_MEM_EN, VGA_MSR_WRITE); + vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO | + VGA_RSRC_LEGACY_MEM | + VGA_RSRC_NORMAL_IO | + VGA_RSRC_NORMAL_MEM); + vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); + } +} + void intel_modeset_init_hw(struct drm_device *dev) { intel_init_power_well(dev); @@ -10539,6 +10562,8 @@ void intel_modeset_cleanup(struct drm_device *dev) intel_disable_fbc(dev); + i915_enable_vga(dev); + intel_disable_gt_powersave(dev); ironlake_teardown_rc6(dev); diff --git a/include/linux/vgaarb.h b/include/linux/vgaarb.h index 2c02f3a8d2ba..80cf8173a65b 100644 --- a/include/linux/vgaarb.h +++ b/include/linux/vgaarb.h @@ -65,8 +65,15 @@ struct pci_dev; * out of the arbitration process (and can be safe to take * interrupts at any time. */ +#if defined(CONFIG_VGA_ARB) extern void vga_set_legacy_decoding(struct pci_dev *pdev, unsigned int decodes); +#else +static inline void vga_set_legacy_decoding(struct pci_dev *pdev, + unsigned int decodes) +{ +} +#endif /** * vga_get - acquire & locks VGA resources From d2933a5b8f8f11cbdf9d2e44f0c7c7abeeb64e6b Mon Sep 17 00:00:00 2001 From: Damien Lespiau Date: Fri, 30 Aug 2013 15:39:26 +0100 Subject: [PATCH 18/29] drm/i915: Don't call sg_free_table() if sg_alloc_table() fails One needs to call __sg_free_table() if __sg_alloc_table() fails, but sg_alloc_table() does that for us already. Signed-off-by: Damien Lespiau Reviewd-by: Chris Wilson Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem.c | 1 - 1 file changed, 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 61313054fce6..f21a0c36a40b 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1774,7 +1774,6 @@ i915_gem_object_get_pages_gtt(struct drm_i915_gem_object *obj) page_count = obj->base.size / PAGE_SIZE; if (sg_alloc_table(st, page_count, GFP_KERNEL)) { - sg_free_table(st); kfree(st); return -ENOMEM; } From 4c6df4b4ca1b26f4532d403b544f649a1c801fd1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 2 Sep 2013 21:13:39 +0300 Subject: [PATCH 19/29] drm/i915: Fix pipe config warnings when dealing with LVDS fixed mode MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit intel_fixed_panel_mode() overwrote the adjusted_mode with the fixed mode only partially. Notably it forgot to copy over the sync flags. The LVDS code however programmed the hardware with the sync flags from fixed mode, and then later the pipe config comparison obviously failed as we filled out the adjusted_mode in get_config from the real registers. Just call drm_mode_copy() in intel_fixed_panel_mode() to copy over the whole thing, and then just use adjusted_mode in the LVDS code to figure out which sync settings the hardware needs. Also constify the fixed_mode argument to intel_fixed_panel_mode(). Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_drv.h | 2 +- drivers/gpu/drm/i915/intel_lvds.c | 8 ++++---- drivers/gpu/drm/i915/intel_panel.c | 14 ++------------ 3 files changed, 7 insertions(+), 17 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index 176080822a74..dbfe5f7bb3de 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -551,7 +551,7 @@ extern int intel_panel_init(struct intel_panel *panel, struct drm_display_mode *fixed_mode); extern void intel_panel_fini(struct intel_panel *panel); -extern void intel_fixed_panel_mode(struct drm_display_mode *fixed_mode, +extern void intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, struct drm_display_mode *adjusted_mode); extern void intel_pch_panel_fitting(struct intel_crtc *crtc, struct intel_crtc_config *pipe_config, diff --git a/drivers/gpu/drm/i915/intel_lvds.c b/drivers/gpu/drm/i915/intel_lvds.c index 4d33278e31fb..831a5c021c4b 100644 --- a/drivers/gpu/drm/i915/intel_lvds.c +++ b/drivers/gpu/drm/i915/intel_lvds.c @@ -128,8 +128,8 @@ static void intel_pre_enable_lvds(struct intel_encoder *encoder) struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; struct intel_crtc *crtc = to_intel_crtc(encoder->base.crtc); - struct drm_display_mode *fixed_mode = - lvds_encoder->attached_connector->base.panel.fixed_mode; + const struct drm_display_mode *adjusted_mode = + &crtc->config.adjusted_mode; int pipe = crtc->pipe; u32 temp; @@ -183,9 +183,9 @@ static void intel_pre_enable_lvds(struct intel_encoder *encoder) temp &= ~LVDS_ENABLE_DITHER; } temp &= ~(LVDS_HSYNC_POLARITY | LVDS_VSYNC_POLARITY); - if (fixed_mode->flags & DRM_MODE_FLAG_NHSYNC) + if (adjusted_mode->flags & DRM_MODE_FLAG_NHSYNC) temp |= LVDS_HSYNC_POLARITY; - if (fixed_mode->flags & DRM_MODE_FLAG_NVSYNC) + if (adjusted_mode->flags & DRM_MODE_FLAG_NVSYNC) temp |= LVDS_VSYNC_POLARITY; I915_WRITE(lvds_encoder->reg, temp); diff --git a/drivers/gpu/drm/i915/intel_panel.c b/drivers/gpu/drm/i915/intel_panel.c index 913cb9d7fd32..42114ecbae0e 100644 --- a/drivers/gpu/drm/i915/intel_panel.c +++ b/drivers/gpu/drm/i915/intel_panel.c @@ -36,20 +36,10 @@ #define PCI_LBPC 0xf4 /* legacy/combination backlight modes */ void -intel_fixed_panel_mode(struct drm_display_mode *fixed_mode, +intel_fixed_panel_mode(const struct drm_display_mode *fixed_mode, struct drm_display_mode *adjusted_mode) { - adjusted_mode->hdisplay = fixed_mode->hdisplay; - adjusted_mode->hsync_start = fixed_mode->hsync_start; - adjusted_mode->hsync_end = fixed_mode->hsync_end; - adjusted_mode->htotal = fixed_mode->htotal; - - adjusted_mode->vdisplay = fixed_mode->vdisplay; - adjusted_mode->vsync_start = fixed_mode->vsync_start; - adjusted_mode->vsync_end = fixed_mode->vsync_end; - adjusted_mode->vtotal = fixed_mode->vtotal; - - adjusted_mode->clock = fixed_mode->clock; + drm_mode_copy(adjusted_mode, fixed_mode); drm_mode_set_crtcinfo(adjusted_mode, 0); } From d4d36014ca37e6fa8271b0690e678b76456faa01 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 2 Sep 2013 20:56:23 +0200 Subject: [PATCH 20/29] drm/i915: fix up the relocate_entry refactoring Somehow we've lost the error handling in the patch split-up between the internal and external patch. This regression has been introduced in commit 5032d871f7d300aee10c309ea004eb4f851553fe Author: Rafael Barbalho Date: Wed Aug 21 17:10:51 2013 +0100 drm/i915: Cleaning up the relocate entry function This bug is exercised by igt/gem_reloc_vs_gpu/interruptible. Cc: Rafael Barbalho Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem_execbuffer.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index 792c52a235ee..bf345777ae9f 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -310,6 +310,9 @@ i915_gem_execbuffer_relocate_entry(struct drm_i915_gem_object *obj, else ret = relocate_entry_gtt(obj, reloc); + if (ret) + return ret; + /* and update the user's relocation entry */ reloc->presumed_offset = target_offset; From 645416f5adc87c8fae44289cdba7562f3ade8f5c Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Mon, 2 Sep 2013 16:22:25 +0200 Subject: [PATCH 21/29] drm/i915: fix hpd work vs. flush_work in the pageflip code deadlock Historically we've run our own driver hotplug handling in our own work-queue, which then launched the drm core hotplug handling in the system workqueue. This is important since we flush our own driver workqueue in the pageflip code while hodling modeset locks, and only the drm hotplug code grabbed these locks. But with commit 69787f7da6b2adc4054357a661aaa1701a9ca76f Author: Daniel Vetter Date: Tue Oct 23 18:23:34 2012 +0000 drm: run the hpd irq event code directly this was changed and now we could deadlock in our flip handler if there's a hotplug work blocking the progress of the crucial unpin works. So this broke the careful deadlock avoidance implemented in commit b4a98e57fc27854b5938fc8b08b68e5e68b91e1f Author: Chris Wilson Date: Thu Nov 1 09:26:26 2012 +0000 drm/i915: Flush outstanding unpin tasks before pageflipping Since the rule thus far has been that work items on our own workqueue may never grab modeset locks simply restore that rule again. v2: Add a comment to the declaration of dev_priv->wq to warn readers about the tricky implications of using it. Suggested by Chris Wilson. Cc: Chris Wilson Cc: Stuart Abercrombie Reported-by: Stuart Abercrombie References: http://permalink.gmane.org/gmane.comp.freedesktop.xorg.drivers.intel/26239 Cc: stable@vger.kernel.org Reviewed-by: Chris Wilson [danvet: Squash in a comment at the place where we schedule the work. Requested after-the-fact by Chris on irc since the hpd work isn't the only place we botch this.] Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_drv.h | 7 +++++++ drivers/gpu/drm/i915/i915_irq.c | 9 +++++++-- 2 files changed, 14 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 52a3785a3fdf..35874b3a86dc 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1236,6 +1236,13 @@ typedef struct drm_i915_private { unsigned int fsb_freq, mem_freq, is_ddr3; + /** + * wq - Driver workqueue for GEM. + * + * NOTE: Work items scheduled here are not allowed to grab any modeset + * locks, for otherwise the flushing done in the pageflip code will + * result in deadlocks. + */ struct workqueue_struct *wq; /* Display functions */ diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 4c6853f57d62..9e48cf27db5e 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1027,8 +1027,13 @@ static inline void intel_hpd_irq_handler(struct drm_device *dev, dev_priv->display.hpd_irq_setup(dev); spin_unlock(&dev_priv->irq_lock); - queue_work(dev_priv->wq, - &dev_priv->hotplug_work); + /* + * Our hotplug handler can grab modeset locks (by calling down into the + * fb helpers). Hence it must not be run on our own dev-priv->wq work + * queue for otherwise the flush_work in the pageflip code will + * deadlock. + */ + schedule_work(&dev_priv->hotplug_work); } static void gmbus_irq_handler(struct drm_device *dev) From eeb4793779060e0ae27fc8a85b8dac2ab3620934 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 3 Sep 2013 20:40:36 +0200 Subject: [PATCH 22/29] drm/i915: handle sdvo input pixel multiplier correctly again MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The sdvo input timing needs to be the actual mode, the sdvo encoder automatically adjusts for the need of pixel doubling or quadrupling. This was lost in pipe config conversion of the pixel multiplier in commit 6cc5f341b5830541a1b6945435ca90c69b1b8b21 Author: Daniel Vetter Date: Wed Mar 27 00:44:53 2013 +0100 drm/i915: add pipe_config->pixel_multiplier While at it ditch the intel_ prefix from the crtc in intel_sdvo_mode_set. Cc: Jesse Barnes Cc: Ville Syrjälä Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_sdvo.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c index 317e058fb3cf..85037b9d4934 100644 --- a/drivers/gpu/drm/i915/intel_sdvo.c +++ b/drivers/gpu/drm/i915/intel_sdvo.c @@ -1151,11 +1151,10 @@ static void intel_sdvo_mode_set(struct intel_encoder *intel_encoder) { struct drm_device *dev = intel_encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; - struct drm_crtc *crtc = intel_encoder->base.crtc; - struct intel_crtc *intel_crtc = to_intel_crtc(crtc); + struct intel_crtc *crtc = to_intel_crtc(intel_encoder->base.crtc); struct drm_display_mode *adjusted_mode = - &intel_crtc->config.adjusted_mode; - struct drm_display_mode *mode = &intel_crtc->config.requested_mode; + &crtc->config.adjusted_mode; + struct drm_display_mode *mode = &crtc->config.requested_mode; struct intel_sdvo *intel_sdvo = to_sdvo(intel_encoder); u32 sdvox; struct intel_sdvo_in_out_map in_out; @@ -1213,13 +1212,15 @@ static void intel_sdvo_mode_set(struct intel_encoder *intel_encoder) * adjusted_mode. */ intel_sdvo_get_dtd_from_mode(&input_dtd, adjusted_mode); + input_dtd.part1.clock /= crtc->config.pixel_multiplier; + if (intel_sdvo->is_tv || intel_sdvo->is_lvds) input_dtd.part2.sdvo_flags = intel_sdvo->dtd_sdvo_flags; if (!intel_sdvo_set_input_timing(intel_sdvo, &input_dtd)) DRM_INFO("Setting input timings on %s failed\n", SDVO_NAME(intel_sdvo)); - switch (intel_crtc->config.pixel_multiplier) { + switch (crtc->config.pixel_multiplier) { default: WARN(1, "unknown pixel mutlipler specified\n"); case 1: rate = SDVO_CLOCK_RATE_MULT_1X; break; @@ -1252,9 +1253,9 @@ static void intel_sdvo_mode_set(struct intel_encoder *intel_encoder) } if (INTEL_PCH_TYPE(dev) >= PCH_CPT) - sdvox |= SDVO_PIPE_SEL_CPT(intel_crtc->pipe); + sdvox |= SDVO_PIPE_SEL_CPT(crtc->pipe); else - sdvox |= SDVO_PIPE_SEL(intel_crtc->pipe); + sdvox |= SDVO_PIPE_SEL(crtc->pipe); if (intel_sdvo->has_hdmi_audio) sdvox |= SDVO_AUDIO_ENABLE; @@ -1264,7 +1265,7 @@ static void intel_sdvo_mode_set(struct intel_encoder *intel_encoder) } else if (IS_I945G(dev) || IS_I945GM(dev) || IS_G33(dev)) { /* done in crtc_mode_set as it lives inside the dpll register */ } else { - sdvox |= (intel_crtc->config.pixel_multiplier - 1) + sdvox |= (crtc->config.pixel_multiplier - 1) << SDVO_PORT_MULTIPLY_SHIFT; } From a2dc53e7dc4d22f20aecdfa10f85004de442e4d0 Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Tue, 3 Sep 2013 20:40:37 +0200 Subject: [PATCH 23/29] drm/i915: fix i9xx_crtc_clock_get for multiplied pixels MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The dpll actually runs at the port clock so we don't need to multiply it again with the pixel multiplier to get the adjusted_mode.clock. This is in contrast to the ironlake pixel clock readout code which uses the fdi dotclock: That one does _not_ run with multiplied pixels. This issue goes back to the original clock readout code added in commit f1f644dc66cbaf5a4c7dcde683361536b41885b9 Author: Jesse Barnes Date: Thu Jun 27 00:39:25 2013 +0300 drm/i915: get mode clock when reading the pipe config v9 Cc: Jesse Barnes Cc: Ville Syrjälä Reviewed-by: Ville Syrjälä Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_display.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 73dec62f871f..0534d5b9b3e3 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7313,8 +7313,7 @@ static void i9xx_crtc_clock_get(struct intel_crtc *crtc, } } - pipe_config->adjusted_mode.clock = clock.dot * - pipe_config->pixel_multiplier; + pipe_config->adjusted_mode.clock = clock.dot; } static void ironlake_crtc_clock_get(struct intel_crtc *crtc, From 57094f82465002fbde1447e2fd850e1179bf6d86 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 4 Sep 2013 10:45:50 +0100 Subject: [PATCH 24/29] drm/i915: Hold an object reference whilst we shrink it Whilst running the shrinker, we need to hold a reference as we unbind the objects, or else we may end up waiting for and retiring requests, which in turn may result in this object being freed. This is very similar to the eviction code which also has to be very careful to keep a reference to its objects as it retires and unbinds them. Another similarity, that Ben pointed out, is that as we may call retire-requests, the unbound_list is outside of our control. We must only process a single element of that list at a time, that is we can not rely on the "safe" next pointer being valid after a call to i915_vma_unbind(). BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 IP: [] i915_gem_gtt_finish_object+0x68/0xbd [i915] PGD 758d3067 PUD ac0d6067 PMD 0 Oops: 0000 [#1] SMP Modules linked in: dm_mod snd_hda_codec_realtek iTCO_wdt iTCO_vendor_support pcspkr snd_hda_intel i2c_i801 snd_hda_codec snd_hwdep snd_pcm snd_page_alloc snd_timer snd lpc_ich mfd_core soundcore battery ac option usb_wwan usbserial uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_core videodev i915 video button drm_kms_helper drm acpi_cpufreq mperf freq_table CPU: 1 PID: 16835 Comm: fbo-maxsize Not tainted 3.11.0-rc7_nightlytop_8fdad4_20130902_+ #7977 task: ffff8800712106d0 ti: ffff880028e4a000 task.ti: ffff880028e4a000 RIP: 0010:[] [] i915_gem_gtt_finish_object+0x68/0xbd [i915] RSP: 0018:ffff880028e4b9e8 EFLAGS: 00010246 RAX: 0000000000000000 RBX: ffff880145734000 RCX: ffff880145735328 RDX: ffff8801457353fc RSI: 0000000000000000 RDI: ffff88007597cc00 RBP: ffff88007597cc00 R08: 0000000000000001 R09: ffff88014f257f00 R10: ffffea0001d65f00 R11: 0000000000bba60b R12: ffff880149e5b000 R13: ffff880145734001 R14: ffff88007597ccc8 R15: ffff88007597cc00 FS: 00007ff5bc919740(0000) GS:ffff88014f240000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000008 CR3: 0000000028f4c000 CR4: 00000000001407e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Stack: 0000000000000000 ffff88007597cc00 ffff8801440d6840 0000000000000000 ffff880145734000 ffffffffa007c854 0000000000000010 ffff88007597c900 0000000000018000 00000000004a1201 ffff88007597cc60 ffffffffa007d183 Call Trace: [] ? i915_vma_unbind+0xe2/0x1d1 [i915] [] ? __i915_gem_shrink+0xf1/0x162 [i915] [] ? i915_gem_object_get_pages_gtt+0xfa/0x303 [i915] [] ? i915_gem_object_get_pages+0x54/0x89 [i915] [] ? i915_gem_object_pin+0x238/0x5ce [i915] [] ? __sg_page_iter_next+0x2b/0x58 [] ? gen6_ppgtt_insert_entries+0xf2/0x114 [i915] [] ? i915_gem_execbuffer_reserve_vma.isra.13+0x79/0x18d [i915] [] ? i915_gem_execbuffer_reserve+0x21d/0x347 [i915] [] ? i915_gem_do_execbuffer.isra.17+0x4f3/0xe61 [i915] [] ? i915_gem_object_get_pages+0x54/0x89 [i915] [] ? i915_gem_pwrite_ioctl+0x743/0x7a5 [i915] [] ? i915_gem_execbuffer2+0x15e/0x1e4 [i915] [] ? drm_ioctl+0x2a5/0x3c4 [drm] [] ? i915_gem_execbuffer+0x37f/0x37f [i915] [] ? __do_page_fault+0x3ab/0x449 [] ? do_mmap_pgoff+0x2b2/0x341 [] ? vfs_ioctl+0x1e/0x31 [] ? do_vfs_ioctl+0x3ad/0x3ef [] ? SyS_ioctl+0x4e/0x7e [] ? system_call_fastpath+0x16/0x1b Code: 52 0c a0 48 c7 c6 22 30 0d a0 31 c0 e8 ef 00 f9 ff bf c6 a7 00 00 e8 90 5d 24 e1 f6 85 13 01 00 00 10 75 44 48 8b 85 18 01 00 00 <8b> 50 08 48 8b 30 49 8b 84 24 88 02 00 00 48 89 c7 48 81 c7 98 RIP [] i915_gem_gtt_finish_object+0x68/0xbd [i915] RSP CR2: 0000000000000008 Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68171 Signed-off-by: Chris Wilson Cc: stable@vger.kernel.org [danvet: Bikeshed the comments a bit as discussed with Chris.] Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem.c | 45 ++++++++++++++++++++++++++++----- 1 file changed, 39 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index f21a0c36a40b..d9e337feef14 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1695,6 +1695,7 @@ static long __i915_gem_shrink(struct drm_i915_private *dev_priv, long target, bool purgeable_only) { + struct list_head still_bound_list; struct drm_i915_gem_object *obj, *next; long count = 0; @@ -1709,23 +1710,55 @@ __i915_gem_shrink(struct drm_i915_private *dev_priv, long target, } } - list_for_each_entry_safe(obj, next, &dev_priv->mm.bound_list, - global_list) { + /* + * As we may completely rewrite the bound list whilst unbinding + * (due to retiring requests) we have to strictly process only + * one element of the list at the time, and recheck the list + * on every iteration. + */ + INIT_LIST_HEAD(&still_bound_list); + while (count < target && !list_empty(&dev_priv->mm.bound_list)) { struct i915_vma *vma, *v; + obj = list_first_entry(&dev_priv->mm.bound_list, + typeof(*obj), global_list); + list_move_tail(&obj->global_list, &still_bound_list); + if (!i915_gem_object_is_purgeable(obj) && purgeable_only) continue; + /* + * Hold a reference whilst we unbind this object, as we may + * end up waiting for and retiring requests. This might + * release the final reference (held by the active list) + * and result in the object being freed from under us. + * in this object being freed. + * + * Note 1: Shrinking the bound list is special since only active + * (and hence bound objects) can contain such limbo objects, so + * we don't need special tricks for shrinking the unbound list. + * The only other place where we have to be careful with active + * objects suddenly disappearing due to retiring requests is the + * eviction code. + * + * Note 2: Even though the bound list doesn't hold a reference + * to the object we can safely grab one here: The final object + * unreferencing and the bound_list are both protected by the + * dev->struct_mutex and so we won't ever be able to observe an + * object on the bound_list with a reference count equals 0. + */ + drm_gem_object_reference(&obj->base); + list_for_each_entry_safe(vma, v, &obj->vma_list, vma_link) if (i915_vma_unbind(vma)) break; - if (!i915_gem_object_put_pages(obj)) { + if (i915_gem_object_put_pages(obj) == 0) count += obj->base.size >> PAGE_SHIFT; - if (count >= target) - return count; - } + + drm_gem_object_unreference(&obj->base); } + list_splice(&still_bound_list, &dev_priv->mm.bound_list); return count; } From 122f46badaafbe651f05c2c0f24cadee692f761b Mon Sep 17 00:00:00 2001 From: Daniel Vetter Date: Wed, 4 Sep 2013 17:36:14 +0200 Subject: [PATCH 25/29] drm/i915: fix gpu hang vs. flip stall deadlocks MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Since we've started to clean up pending flips when the gpu hangs in commit 96a02917a0131e52efefde49c2784c0421d6c439 Author: Ville Syrjälä Date: Mon Feb 18 19:08:49 2013 +0200 drm/i915: Finish page flips and update primary planes after a GPU reset the gpu reset work now also grabs modeset locks. But since work items on our private work queue are not allowed to do that due to the flush_workqueue from the pageflip code this results in a neat deadlock: INFO: task kms_flip:14676 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kms_flip D ffff88019283a5c0 0 14676 13344 0x00000004 ffff88018e62dbf8 0000000000000046 ffff88013bdb12e0 ffff88018e62dfd8 ffff88018e62dfd8 00000000001d3b00 ffff88019283a5c0 ffff88018ec21000 ffff88018f693f00 ffff88018eece000 ffff88018e62dd60 ffff88018eece898 Call Trace: [] schedule+0x60/0x62 [] intel_crtc_wait_for_pending_flips+0xb2/0x114 [i915] [] ? finish_wait+0x60/0x60 [] intel_crtc_set_config+0x7f3/0x81e [i915] [] drm_mode_set_config_internal+0x4f/0xc6 [drm] [] drm_mode_setcrtc+0x44d/0x4f9 [drm] [] ? might_fault+0x38/0x86 [] drm_ioctl+0x2f9/0x447 [drm] [] ? trace_hardirqs_off+0xd/0xf [] ? drm_mode_setplane+0x343/0x343 [drm] [] ? mntput_no_expire+0x3e/0x13d [] vfs_ioctl+0x18/0x34 [] do_vfs_ioctl+0x396/0x454 [] ? sysret_check+0x1b/0x56 [] SyS_ioctl+0x52/0x7d [] system_call_fastpath+0x16/0x1b 2 locks held by kms_flip/14676: #0: (&dev->mode_config.mutex){+.+.+.}, at: [] drm_modeset_lock_all+0x22/0x59 [drm] #1: (&crtc->mutex){+.+.+.}, at: [] drm_modeset_lock_all+0x48/0x59 [drm] INFO: task kworker/u8:4:175 blocked for more than 120 seconds. "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. kworker/u8:4 D ffff88018de9a5c0 0 175 2 0x00000000 Workqueue: i915 i915_error_work_func [i915] ffff88018e37dc30 0000000000000046 ffff8801938ab8a0 ffff88018e37dfd8 ffff88018e37dfd8 00000000001d3b00 ffff88018de9a5c0 ffff88018ec21018 0000000000000246 ffff88018e37dca0 000000005a865a86 ffff88018de9a5c0 Call Trace: [] schedule+0x60/0x62 [] schedule_preempt_disabled+0x9/0xb [] mutex_lock_nested+0x205/0x3b1 [] ? intel_display_handle_reset+0x7e/0xbd [i915] [] ? intel_display_handle_reset+0x7e/0xbd [i915] [] intel_display_handle_reset+0x7e/0xbd [i915] [] i915_error_work_func+0x128/0x147 [i915] [] process_one_work+0x1d4/0x35a [] ? process_one_work+0x15b/0x35a [] worker_thread+0x144/0x1f0 [] ? rescuer_thread+0x275/0x275 [] kthread+0xac/0xb4 [] ? finish_task_switch+0x3b/0xc0 [] ? __kthread_parkme+0x60/0x60 [] ret_from_fork+0x7c/0xb0 [] ? __kthread_parkme+0x60/0x60 3 locks held by kworker/u8:4/175: #0: (i915){.+.+.+}, at: [] process_one_work+0x15b/0x35a #1: ((&dev_priv->gpu_error.work)){+.+.+.}, at: [] process_one_work+0x15b/0x35a #2: (&crtc->mutex){+.+.+.}, at: [] intel_display_handle_reset+0x7e/0xbd [i915] This blew up while running kms_flip/flip-vs-panning-vs-hang-interruptible on one of my older machines. Unfortunately (despite the proper lockdep annotations for flush_workqueue) lockdep still doesn't detect this correctly, so we need to rely on chance to discover these bugs. Apply the usual bugfix and schedule the reset work on the system workqueue to keep our own driver workqueue free of any modeset lock grabbing. Note that this is not a terribly serious regression since before the offending commit we'd simply have stalled userspace forever due to failing to abort all outstanding pageflips. v2: Add a comment as requested by Chris. Cc: Thomas Gleixner Cc: stable@vger.kernel.org Cc: Ville Syrjälä Cc: Chris Wilson Reviewed-by: Chris Wilson Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_irq.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c index 9e48cf27db5e..83cce0cdb769 100644 --- a/drivers/gpu/drm/i915/i915_irq.c +++ b/drivers/gpu/drm/i915/i915_irq.c @@ -1660,7 +1660,13 @@ void i915_handle_error(struct drm_device *dev, bool wedged) wake_up_all(&ring->irq_queue); } - queue_work(dev_priv->wq, &dev_priv->gpu_error.work); + /* + * Our reset work can grab modeset locks (since it needs to reset the + * state of outstanding pagelips). Hence it must not be run on our own + * dev-priv->wq work queue for otherwise the flush_work in the pageflip + * code will deadlock. + */ + schedule_work(&dev_priv->gpu_error.work); } static void __always_unused i915_pageflip_stall_check(struct drm_device *dev, int pipe) From 6644a4e9422fe19293599d1fef0bd292f4ca72a6 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Thu, 5 Sep 2013 13:40:25 +0100 Subject: [PATCH 26/29] drm/i915: Skip stolen region initialisation if none is reserved Paulo reported that if he set the amount of reserved memory to 0, then we emitted a warning about a conflict before disabling our use of stolen memory. This was introduced with commit eaba1b8f3379b5d100bd146b9a41d28348bdfd09 Author: Chris Wilson Date: Thu Jul 4 12:28:35 2013 +0100 drm/i915: Verify that our stolen memory doesn't conflict and is simply fixed by checking for a no reservation first. Reported-by: Paulo Zanoni Signed-off-by: Chris Wilson Cc: Paulo Zanoni Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem_stolen.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/gpu/drm/i915/i915_gem_stolen.c b/drivers/gpu/drm/i915/i915_gem_stolen.c index 9969d10b80f5..e15a1d90037d 100644 --- a/drivers/gpu/drm/i915/i915_gem_stolen.c +++ b/drivers/gpu/drm/i915/i915_gem_stolen.c @@ -201,6 +201,9 @@ int i915_gem_init_stolen(struct drm_device *dev) struct drm_i915_private *dev_priv = dev->dev_private; int bios_reserved = 0; + if (dev_priv->gtt.stolen_size == 0) + return 0; + dev_priv->mm.stolen_base = i915_stolen_to_physical(dev); if (dev_priv->mm.stolen_base == 0) return 0; From 1c5fd08520de35c7ac0d70e2ed3c3bb200830447 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Wed, 4 Sep 2013 10:54:30 +0100 Subject: [PATCH 27/29] drm/i915: Confine page flips to BCS on Valleyview Once again we find that Valleyview is ever so subtlety different from the rest of its gen7 brethen. In this case, Valleyview has no support for pageflipping from the RCS ring. Fixes a regression from commit ffe74d75502e3a9b0791240b5562bcbecc6ab8dc Author: Chris Wilson Date: Mon Aug 26 20:58:12 2013 +0100 drm/i915: Use RCS flips on Ivybridge+ Reported-by: "Lee, Chon Ming" Signed-off-by: Chris Wilson Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=68968 Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_display.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0534d5b9b3e3..0ee89011c19a 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -7844,7 +7844,7 @@ static int intel_gen7_queue_flip(struct drm_device *dev, int len, ret; ring = obj->ring; - if (ring == NULL || ring->id != RCS) + if (IS_VALLEYVIEW(dev) || ring == NULL || ring->id != RCS) ring = &dev_priv->ring[BCS]; ret = intel_pin_and_fence_fb_obj(dev, obj, ring); From cac6a5ae0118832936eb162ec4cedb30f2422bcc Mon Sep 17 00:00:00 2001 From: Jani Nikula Date: Fri, 23 Aug 2013 10:50:39 +0300 Subject: [PATCH 28/29] drm/i915: try not to lose backlight CBLV precision ACPI has _BCM and _BQC methods to set and query the backlight brightness, respectively. The ACPI opregion has variables BCLP and CBLV to hold the requested and current backlight brightness, respectively. The BCLP variable has range 0..255 while the others have range 0..100. This means the _BCM method has to scale the brightness for BCLP, and the gfx driver has to scale the requested value back for CBLV. If the _BQC method uses the CBLV variable (apparently some implementations do, some don't) for current backlight level reporting, there's room for rounding errors. Use DIV_ROUND_UP for scaling back to CBLV to get back to the same values that were passed to _BCM, presuming the _BCM simply uses bclp = (in * 255) / 100 for scaling to BCLP. Reference: https://gist.github.com/aaronlu/6314920 Reported-by: Aaron Lu Signed-off-by: Jani Nikula Reviewed-by: Aaron Lu Cc: stable@vger.kernel.org Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/intel_opregion.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_opregion.c b/drivers/gpu/drm/i915/intel_opregion.c index cfb8fb68f09c..119771ff46ab 100644 --- a/drivers/gpu/drm/i915/intel_opregion.c +++ b/drivers/gpu/drm/i915/intel_opregion.c @@ -173,7 +173,7 @@ static u32 asle_set_backlight(struct drm_device *dev, u32 bclp) return ASLE_BACKLIGHT_FAILED; intel_panel_set_backlight(dev, bclp, 255); - iowrite32((bclp*0x64)/0xff | ASLE_CBLV_VALID, &asle->cblv); + iowrite32(DIV_ROUND_UP(bclp * 100, 255) | ASLE_CBLV_VALID, &asle->cblv); return 0; } From 6e1b4fdad5157bb9e88777d525704aba24389bee Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Thu, 5 Sep 2013 20:40:52 +0300 Subject: [PATCH 29/29] drm/i915: Delay disabling of VGA memory until vgacon->fbcon handoff is done MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit When transitioning away from vgacon the system tries to save the current contents of the VGA memory, so that it can be cleanly handed off to fbcon (or whatever comes afterwards). The recent change commit 81b5c7bc8de3e6f63419139c2fc91bf81dea8a7d Author: Alex Williamson Date: Wed Aug 28 09:39:08 2013 -0600 i915: Update VGA arbiter support for newer devices caused i915 to disable VGA memory decode for the IGD when i915 is initializing. Unfortunately that happens before the vgacon->fbcon handoff so vgacon_save_screen() will read out all ones from the VGA memory. After the handoff fbcon will inherit the bogus state from vgacon, and pre-fills the fb with matching contents. The end result is a white rectangle in the top left corner of the screen, the size of which matches the now inactive VGA console. To remedy the situation delay the disabling of VGA memory until the vgacon->fbcon handoff has happened. Also rename i915_enable_vga to i915_enable_vga_mem to make the relationship between these functions clearer. Cc: Alex Williamson Signed-off-by: Ville Syrjälä Signed-off-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_dma.c | 6 ++++++ drivers/gpu/drm/i915/intel_display.c | 27 ++++++++++++++++----------- drivers/gpu/drm/i915/intel_drv.h | 1 + 3 files changed, 23 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 3de60503378e..9b265a4c6a3d 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1351,6 +1351,12 @@ static int i915_load_modeset_init(struct drm_device *dev) */ intel_fbdev_initial_config(dev); + /* + * Must do this after fbcon init so that + * vgacon_save_screen() works during the handover. + */ + i915_disable_vga_mem(dev); + /* Only enable hotplug handling once the fbdev is fully set up. */ dev_priv->enable_hotplug_processing = true; diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 0ee89011c19a..2489d0b4c7d2 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -10040,15 +10040,6 @@ static void i915_disable_vga(struct drm_device *dev) outb(SR01, VGA_SR_INDEX); sr1 = inb(VGA_SR_DATA); outb(sr1 | 1<<5, VGA_SR_DATA); - - /* Disable VGA memory on Intel HD */ - if (HAS_PCH_SPLIT(dev)) { - outb(inb(VGA_MSR_READ) & ~VGA_MSR_MEM_EN, VGA_MSR_WRITE); - vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO | - VGA_RSRC_NORMAL_IO | - VGA_RSRC_NORMAL_MEM); - } - vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); udelay(300); @@ -10056,7 +10047,7 @@ static void i915_disable_vga(struct drm_device *dev) POSTING_READ(vga_reg); } -static void i915_enable_vga(struct drm_device *dev) +static void i915_enable_vga_mem(struct drm_device *dev) { /* Enable VGA memory on Intel HD */ if (HAS_PCH_SPLIT(dev)) { @@ -10070,6 +10061,19 @@ static void i915_enable_vga(struct drm_device *dev) } } +void i915_disable_vga_mem(struct drm_device *dev) +{ + /* Disable VGA memory on Intel HD */ + if (HAS_PCH_SPLIT(dev)) { + vga_get_uninterruptible(dev->pdev, VGA_RSRC_LEGACY_IO); + outb(inb(VGA_MSR_READ) & ~VGA_MSR_MEM_EN, VGA_MSR_WRITE); + vga_set_legacy_decoding(dev->pdev, VGA_RSRC_LEGACY_IO | + VGA_RSRC_NORMAL_IO | + VGA_RSRC_NORMAL_MEM); + vga_put(dev->pdev, VGA_RSRC_LEGACY_IO); + } +} + void intel_modeset_init_hw(struct drm_device *dev) { intel_init_power_well(dev); @@ -10348,6 +10352,7 @@ void i915_redisable_vga(struct drm_device *dev) if (I915_READ(vga_reg) != VGA_DISP_DISABLE) { DRM_DEBUG_KMS("Something enabled VGA plane, disabling it\n"); i915_disable_vga(dev); + i915_disable_vga_mem(dev); } } @@ -10561,7 +10566,7 @@ void intel_modeset_cleanup(struct drm_device *dev) intel_disable_fbc(dev); - i915_enable_vga(dev); + i915_enable_vga_mem(dev); intel_disable_gt_powersave(dev); diff --git a/drivers/gpu/drm/i915/intel_drv.h b/drivers/gpu/drm/i915/intel_drv.h index dbfe5f7bb3de..a47799e832c6 100644 --- a/drivers/gpu/drm/i915/intel_drv.h +++ b/drivers/gpu/drm/i915/intel_drv.h @@ -792,5 +792,6 @@ extern void hsw_pc8_disable_interrupts(struct drm_device *dev); extern void hsw_pc8_restore_interrupts(struct drm_device *dev); extern void intel_aux_display_runtime_get(struct drm_i915_private *dev_priv); extern void intel_aux_display_runtime_put(struct drm_i915_private *dev_priv); +extern void i915_disable_vga_mem(struct drm_device *dev); #endif /* __INTEL_DRV_H__ */