From 0ee537abbd10a9abf11e1c22ee32a68e8c12ed4a Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 6 Mar 2011 09:03:16 +0000 Subject: [PATCH 1/4] Revert "drm/i915: fix corruptions on i8xx due to relaxed fencing" This reverts commit c2e0eb167070a6e9dcb49c84c13c79a30d672431. As it turns out, userspace already depends upon being able to enable tiling on existing bo which it promises to be large enough for its purposes i.e. it will not access beyond the end of the last full-tile row. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=35016 Reported-and-tested-by: Kamal Mostafa Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_gem_tiling.c | 16 +--------------- 1 file changed, 1 insertion(+), 15 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 79a04fde69b5..22a32b9932c5 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -184,7 +184,7 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev) static bool i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode) { - int tile_width, tile_height; + int tile_width; /* Linear is always fine */ if (tiling_mode == I915_TILING_NONE) @@ -215,20 +215,6 @@ i915_tiling_ok(struct drm_device *dev, int stride, int size, int tiling_mode) } } - if (IS_GEN2(dev) || - (tiling_mode == I915_TILING_Y && HAS_128_BYTE_Y_TILING(dev))) - tile_height = 32; - else - tile_height = 8; - /* i8xx is strange: It has 2 interleaved rows of tiles, so needs an even - * number of tile rows. */ - if (IS_GEN2(dev)) - tile_height *= 2; - - /* Size needs to be aligned to a full tile row */ - if (size & (tile_height * stride - 1)) - return false; - /* 965+ just needs multiples of tile width */ if (INTEL_INFO(dev)->gen >= 4) { if (stride & (tile_width - 1)) From 91355834646328e7edc6bd25176ae44bcd7386c7 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 4 Mar 2011 19:22:40 +0000 Subject: [PATCH 2/4] drm/i915: Do not overflow the MMADDR write FIFO Whilst the GT is powered down (rc6), writes to MMADDR are placed in a FIFO by the System Agent. This is a limited resource, only 64 entries, of which 20 are reserved for Display and PCH writes, and so we must take care not to queue up too many writes. To avoid this, there is counter which we can poll to ensure there are sufficient free entries in the fifo. "Issuing a write to a full FIFO is not supported; at worst it could result in corruption or a system hang." Reported-and-Tested-by: Matt Turner Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=34056 Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_debugfs.c | 4 ++-- drivers/gpu/drm/i915/i915_drv.c | 14 ++++++++++++-- drivers/gpu/drm/i915/i915_drv.h | 20 +++++++++++++++----- drivers/gpu/drm/i915/i915_reg.h | 2 ++ drivers/gpu/drm/i915/intel_display.c | 8 ++++---- drivers/gpu/drm/i915/intel_ringbuffer.h | 13 +++++++------ 6 files changed, 42 insertions(+), 19 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c index 3601466c5502..4ff9b6cc973f 100644 --- a/drivers/gpu/drm/i915/i915_debugfs.c +++ b/drivers/gpu/drm/i915/i915_debugfs.c @@ -865,7 +865,7 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused) int max_freq; /* RPSTAT1 is in the GT power well */ - __gen6_force_wake_get(dev_priv); + __gen6_gt_force_wake_get(dev_priv); seq_printf(m, "GT_PERF_STATUS: 0x%08x\n", gt_perf_status); seq_printf(m, "RPSTAT1: 0x%08x\n", I915_READ(GEN6_RPSTAT1)); @@ -888,7 +888,7 @@ static int i915_cur_delayinfo(struct seq_file *m, void *unused) seq_printf(m, "Max non-overclocked (RP0) frequency: %dMHz\n", max_freq * 100); - __gen6_force_wake_put(dev_priv); + __gen6_gt_force_wake_put(dev_priv); } else { seq_printf(m, "no P-state info available\n"); } diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 0ad533f06af9..37d672a116db 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -254,7 +254,7 @@ void intel_detect_pch (struct drm_device *dev) } } -void __gen6_force_wake_get(struct drm_i915_private *dev_priv) +void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv) { int count; @@ -270,12 +270,22 @@ void __gen6_force_wake_get(struct drm_i915_private *dev_priv) udelay(10); } -void __gen6_force_wake_put(struct drm_i915_private *dev_priv) +void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv) { I915_WRITE_NOTRACE(FORCEWAKE, 0); POSTING_READ(FORCEWAKE); } +void __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv) +{ + int loop = 500; + u32 fifo = I915_READ_NOTRACE(GT_FIFO_FREE_ENTRIES); + while (fifo < 20 && loop--) { + udelay(10); + fifo = I915_READ_NOTRACE(GT_FIFO_FREE_ENTRIES); + } +} + static int i915_drm_freeze(struct drm_device *dev) { struct drm_i915_private *dev_priv = dev->dev_private; diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 65dfe81d0035..549c046b4ecc 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1353,22 +1353,32 @@ __i915_write(64, q) * must be set to prevent GT core from power down and stale values being * returned. */ -void __gen6_force_wake_get(struct drm_i915_private *dev_priv); -void __gen6_force_wake_put (struct drm_i915_private *dev_priv); -static inline u32 i915_safe_read(struct drm_i915_private *dev_priv, u32 reg) +void __gen6_gt_force_wake_get(struct drm_i915_private *dev_priv); +void __gen6_gt_force_wake_put(struct drm_i915_private *dev_priv); +void __gen6_gt_wait_for_fifo(struct drm_i915_private *dev_priv); + +static inline u32 i915_gt_read(struct drm_i915_private *dev_priv, u32 reg) { u32 val; if (dev_priv->info->gen >= 6) { - __gen6_force_wake_get(dev_priv); + __gen6_gt_force_wake_get(dev_priv); val = I915_READ(reg); - __gen6_force_wake_put(dev_priv); + __gen6_gt_force_wake_put(dev_priv); } else val = I915_READ(reg); return val; } +static inline void i915_gt_write(struct drm_i915_private *dev_priv, + u32 reg, u32 val) +{ + if (dev_priv->info->gen >= 6) + __gen6_gt_wait_for_fifo(dev_priv); + I915_WRITE(reg, val); +} + static inline void i915_write(struct drm_i915_private *dev_priv, u32 reg, u64 val, int len) { diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index 729d4233b763..3e6f486f4605 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -3261,6 +3261,8 @@ #define FORCEWAKE 0xA18C #define FORCEWAKE_ACK 0x130090 +#define GT_FIFO_FREE_ENTRIES 0x120008 + #define GEN6_RPNSWREQ 0xA008 #define GEN6_TURBO_DISABLE (1<<31) #define GEN6_FREQUENCY(x) ((x)<<25) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index e79b25bbee6c..49fb54fd9a18 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -1219,7 +1219,7 @@ static void sandybridge_blit_fbc_update(struct drm_device *dev) u32 blt_ecoskpd; /* Make sure blitter notifies FBC of writes */ - __gen6_force_wake_get(dev_priv); + __gen6_gt_force_wake_get(dev_priv); blt_ecoskpd = I915_READ(GEN6_BLITTER_ECOSKPD); blt_ecoskpd |= GEN6_BLITTER_FBC_NOTIFY << GEN6_BLITTER_LOCK_SHIFT; @@ -1230,7 +1230,7 @@ static void sandybridge_blit_fbc_update(struct drm_device *dev) GEN6_BLITTER_LOCK_SHIFT); I915_WRITE(GEN6_BLITTER_ECOSKPD, blt_ecoskpd); POSTING_READ(GEN6_BLITTER_ECOSKPD); - __gen6_force_wake_put(dev_priv); + __gen6_gt_force_wake_put(dev_priv); } static void ironlake_enable_fbc(struct drm_crtc *crtc, unsigned long interval) @@ -6282,7 +6282,7 @@ void gen6_enable_rps(struct drm_i915_private *dev_priv) * userspace... */ I915_WRITE(GEN6_RC_STATE, 0); - __gen6_force_wake_get(dev_priv); + __gen6_gt_force_wake_get(dev_priv); /* disable the counters and set deterministic thresholds */ I915_WRITE(GEN6_RC_CONTROL, 0); @@ -6380,7 +6380,7 @@ void gen6_enable_rps(struct drm_i915_private *dev_priv) /* enable all PM interrupts */ I915_WRITE(GEN6_PMINTRMSK, 0); - __gen6_force_wake_put(dev_priv); + __gen6_gt_force_wake_put(dev_priv); } void intel_enable_clock_gating(struct drm_device *dev) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h index 6d6fde85a636..34306865a5df 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.h +++ b/drivers/gpu/drm/i915/intel_ringbuffer.h @@ -14,22 +14,23 @@ struct intel_hw_status_page { struct drm_i915_gem_object *obj; }; -#define I915_RING_READ(reg) i915_safe_read(dev_priv, reg) +#define I915_RING_READ(reg) i915_gt_read(dev_priv, reg) +#define I915_RING_WRITE(reg, val) i915_gt_write(dev_priv, reg, val) #define I915_READ_TAIL(ring) I915_RING_READ(RING_TAIL((ring)->mmio_base)) -#define I915_WRITE_TAIL(ring, val) I915_WRITE(RING_TAIL((ring)->mmio_base), val) +#define I915_WRITE_TAIL(ring, val) I915_RING_WRITE(RING_TAIL((ring)->mmio_base), val) #define I915_READ_START(ring) I915_RING_READ(RING_START((ring)->mmio_base)) -#define I915_WRITE_START(ring, val) I915_WRITE(RING_START((ring)->mmio_base), val) +#define I915_WRITE_START(ring, val) I915_RING_WRITE(RING_START((ring)->mmio_base), val) #define I915_READ_HEAD(ring) I915_RING_READ(RING_HEAD((ring)->mmio_base)) -#define I915_WRITE_HEAD(ring, val) I915_WRITE(RING_HEAD((ring)->mmio_base), val) +#define I915_WRITE_HEAD(ring, val) I915_RING_WRITE(RING_HEAD((ring)->mmio_base), val) #define I915_READ_CTL(ring) I915_RING_READ(RING_CTL((ring)->mmio_base)) -#define I915_WRITE_CTL(ring, val) I915_WRITE(RING_CTL((ring)->mmio_base), val) +#define I915_WRITE_CTL(ring, val) I915_RING_WRITE(RING_CTL((ring)->mmio_base), val) -#define I915_WRITE_IMR(ring, val) I915_WRITE(RING_IMR((ring)->mmio_base), val) #define I915_READ_IMR(ring) I915_RING_READ(RING_IMR((ring)->mmio_base)) +#define I915_WRITE_IMR(ring, val) I915_RING_WRITE(RING_IMR((ring)->mmio_base), val) #define I915_READ_NOPID(ring) I915_RING_READ(RING_NOPID((ring)->mmio_base)) #define I915_READ_SYNC_0(ring) I915_RING_READ(RING_SYNC_0((ring)->mmio_base)) From a1656b9090f7008d2941c314f5a64724bea2ae37 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Fri, 4 Mar 2011 18:48:03 +0000 Subject: [PATCH 3/4] drm/i915: Disable GPU semaphores by default Andi Kleen narrowed his GPU hangs on his Sugar Bay (SNB desktop) rev 09 down to the use of GPU semaphores, and we already know that they appear broken up to Huron River (mobile) rev 08. (I'm optimistic that disabling GPU semaphores is simply hiding another bug by the latency and side-effects of the additional device interaction it introduces...) However, use of semaphores is a massive performance improvement... Only as long as the system remains stable. Enable at your peril. Reported-by: Andi Kleen Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=33921 Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_drv.c | 3 +++ drivers/gpu/drm/i915/i915_drv.h | 1 + drivers/gpu/drm/i915/i915_gem_execbuffer.c | 4 ++-- 3 files changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 37d672a116db..22ec066adae6 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -46,6 +46,9 @@ module_param_named(fbpercrtc, i915_fbpercrtc, int, 0400); unsigned int i915_powersave = 1; module_param_named(powersave, i915_powersave, int, 0600); +unsigned int i915_semaphores = 0; +module_param_named(semaphores, i915_semaphores, int, 0600); + unsigned int i915_enable_rc6 = 0; module_param_named(i915_enable_rc6, i915_enable_rc6, int, 0600); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 549c046b4ecc..d023b9b33d60 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -956,6 +956,7 @@ extern struct drm_ioctl_desc i915_ioctls[]; extern int i915_max_ioctl; extern unsigned int i915_fbpercrtc; extern unsigned int i915_powersave; +extern unsigned int i915_semaphores; extern unsigned int i915_lvds_downclock; extern unsigned int i915_panel_use_ssc; extern unsigned int i915_enable_rc6; diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c index d2f445e825f2..50ab1614571c 100644 --- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c +++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c @@ -772,8 +772,8 @@ i915_gem_execbuffer_sync_rings(struct drm_i915_gem_object *obj, if (from == NULL || to == from) return 0; - /* XXX gpu semaphores are currently causing hard hangs on SNB mobile */ - if (INTEL_INFO(obj->base.dev)->gen < 6 || IS_MOBILE(obj->base.dev)) + /* XXX gpu semaphores are implicated in various hard hangs on SNB */ + if (INTEL_INFO(obj->base.dev)->gen < 6 || !i915_semaphores) return i915_gem_object_wait_rendering(obj, true); idx = intel_ring_sync_index(from, to); From 467cffba85791cdfce38c124d75bd578f4bb8625 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 7 Mar 2011 10:42:03 +0000 Subject: [PATCH 4/4] drm/i915: Rebind the buffer if its alignment constraints changes with tiling Early gen3 and gen2 chipset do not have the relaxed per-surface tiling constraints of the later chipsets, so we need to check that the GTT alignment is correct for the new tiling. If it is not, we need to rebind. Reported-by: Daniel Vetter Reviewed-by: Daniel Vetter Signed-off-by: Chris Wilson --- drivers/gpu/drm/i915/i915_drv.h | 3 +++ drivers/gpu/drm/i915/i915_gem.c | 2 +- drivers/gpu/drm/i915/i915_gem_tiling.c | 21 +++++++++++++++++---- 3 files changed, 21 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index d023b9b33d60..456f40484838 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1178,6 +1178,9 @@ void i915_gem_detach_phys_object(struct drm_device *dev, void i915_gem_free_all_phys_object(struct drm_device *dev); void i915_gem_release(struct drm_device *dev, struct drm_file *file); +uint32_t +i915_gem_get_unfenced_gtt_alignment(struct drm_i915_gem_object *obj); + /* i915_gem_gtt.c */ void i915_gem_restore_gtt_mappings(struct drm_device *dev); int __must_check i915_gem_gtt_bind_object(struct drm_i915_gem_object *obj); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index cf4f74c7c6fb..36e66cc5225e 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1398,7 +1398,7 @@ i915_gem_get_gtt_alignment(struct drm_i915_gem_object *obj) * Return the required GTT alignment for an object, only taking into account * unfenced tiled surface requirements. */ -static uint32_t +uint32_t i915_gem_get_unfenced_gtt_alignment(struct drm_i915_gem_object *obj) { struct drm_device *dev = obj->base.dev; diff --git a/drivers/gpu/drm/i915/i915_gem_tiling.c b/drivers/gpu/drm/i915/i915_gem_tiling.c index 22a32b9932c5..d64843e18df2 100644 --- a/drivers/gpu/drm/i915/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/i915_gem_tiling.c @@ -349,14 +349,27 @@ i915_gem_set_tiling(struct drm_device *dev, void *data, (obj->gtt_offset + obj->base.size <= dev_priv->mm.gtt_mappable_end && i915_gem_object_fence_ok(obj, args->tiling_mode)); - obj->tiling_changed = true; - obj->tiling_mode = args->tiling_mode; - obj->stride = args->stride; + /* Rebind if we need a change of alignment */ + if (!obj->map_and_fenceable) { + u32 unfenced_alignment = + i915_gem_get_unfenced_gtt_alignment(obj); + if (obj->gtt_offset & (unfenced_alignment - 1)) + ret = i915_gem_object_unbind(obj); + } + + if (ret == 0) { + obj->tiling_changed = true; + obj->tiling_mode = args->tiling_mode; + obj->stride = args->stride; + } } + /* we have to maintain this existing ABI... */ + args->stride = obj->stride; + args->tiling_mode = obj->tiling_mode; drm_gem_object_unreference(&obj->base); mutex_unlock(&dev->struct_mutex); - return 0; + return ret; } /**