All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/3] drm/i915: Prepare for larger CSB status FIFO size
@ 2018-12-18 13:24 Mika Kuoppala
  2018-12-18 13:24 ` [PATCH 2/3] drm/i915/icl: Switch to using 12 deep CSB status FIFO Mika Kuoppala
                   ` (11 more replies)
  0 siblings, 12 replies; 19+ messages in thread
From: Mika Kuoppala @ 2018-12-18 13:24 UTC (permalink / raw)
  To: intel-gfx

Make csb entry count variable in preparation for larger
CSB status FIFO size found on gen11+ hardware.

v2: adapt to hwsp access only (Chris)
    non continuous mmio (Daniele)

Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
---
 drivers/gpu/drm/i915/intel_engine_cs.c  | 23 +++++++++++++----------
 drivers/gpu/drm/i915/intel_lrc.c        | 10 +++++++---
 drivers/gpu/drm/i915/intel_lrc.h        | 19 ++++++++++++-------
 drivers/gpu/drm/i915/intel_ringbuffer.h |  5 +++++
 4 files changed, 37 insertions(+), 20 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
index 6f165f9ad2bf..f5960b3dc5d7 100644
--- a/drivers/gpu/drm/i915/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/intel_engine_cs.c
@@ -1305,37 +1305,40 @@ static void intel_engine_print_registers(const struct intel_engine_cs *engine,
 	}
 
 	if (HAS_EXECLISTS(dev_priv)) {
-		const u32 *hws = &engine->status_page.page_addr[I915_HWS_CSB_BUF0_INDEX];
+		const u32 *hws = execlists->csb_status;
+		const u8 entries = execlists->csb_size;
 		unsigned int idx;
 		u8 read, write;
 
-		drm_printf(m, "\tExeclist status: 0x%08x %08x\n",
+		drm_printf(m, "\tExeclist status: 0x%08x %08x, entries %u\n",
 			   I915_READ(RING_EXECLIST_STATUS_LO(engine)),
-			   I915_READ(RING_EXECLIST_STATUS_HI(engine)));
+			   I915_READ(RING_EXECLIST_STATUS_HI(engine)),
+			   entries);
 
 		read = execlists->csb_head;
 		write = READ_ONCE(*execlists->csb_write);
 
 		drm_printf(m, "\tExeclist CSB read %d, write %d [mmio:%d], tasklet queued? %s (%s)\n",
 			   read, write,
-			   GEN8_CSB_WRITE_PTR(I915_READ(RING_CONTEXT_STATUS_PTR(engine))),
+			   I915_READ(RING_CONTEXT_STATUS_PTR(engine)) &
+			   GEN11_CSB_WRITE_PTR_MASK,
 			   yesno(test_bit(TASKLET_STATE_SCHED,
 					  &engine->execlists.tasklet.state)),
 			   enableddisabled(!atomic_read(&engine->execlists.tasklet.count)));
-		if (read >= GEN8_CSB_ENTRIES)
+		if (read >= entries)
 			read = 0;
-		if (write >= GEN8_CSB_ENTRIES)
+		if (write >= entries)
 			write = 0;
 		if (read > write)
-			write += GEN8_CSB_ENTRIES;
+			write += entries;
 		while (read < write) {
-			idx = ++read % GEN8_CSB_ENTRIES;
+			idx = ++read % entries;
 			drm_printf(m, "\tExeclist CSB[%d]: 0x%08x [mmio:0x%08x], context: %d [mmio:%d]\n",
 				   idx,
 				   hws[idx * 2],
-				   I915_READ(RING_CONTEXT_STATUS_BUF_LO(engine, idx)),
+				   I915_READ(RING_CSB_LO(engine, idx)),
 				   hws[idx * 2 + 1],
-				   I915_READ(RING_CONTEXT_STATUS_BUF_HI(engine, idx)));
+				   I915_READ(RING_CSB_HI(engine, idx)));
 		}
 
 		rcu_read_lock();
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 1c9748f391fe..bd35ac3b7b77 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -760,7 +760,7 @@ invalidate_csb_entries(const u32 *first, const u32 *last)
 
 static void reset_csb_pointers(struct intel_engine_execlists *execlists)
 {
-	const unsigned int reset_value = GEN8_CSB_ENTRIES - 1;
+	const unsigned int reset_value = execlists->csb_size - 1;
 
 	/*
 	 * After a reset, the HW starts writing into CSB entry [0]. We
@@ -867,7 +867,7 @@ static void process_csb(struct intel_engine_cs *engine)
 	struct intel_engine_execlists * const execlists = &engine->execlists;
 	struct execlist_port *port = execlists->port;
 	const u32 * const buf = execlists->csb_status;
-	u8 head, tail;
+	u8 head, tail, entries;
 
 	/*
 	 * Note that csb_write, csb_status may be either in HWSP or mmio.
@@ -885,6 +885,8 @@ static void process_csb(struct intel_engine_cs *engine)
 	if (unlikely(head == tail))
 		return;
 
+	entries = execlists->csb_size;
+
 	/*
 	 * Hopefully paired with a wmb() in HW!
 	 *
@@ -900,7 +902,7 @@ static void process_csb(struct intel_engine_cs *engine)
 		unsigned int status;
 		unsigned int count;
 
-		if (++head == GEN8_CSB_ENTRIES)
+		if (++head == entries)
 			head = 0;
 
 		/*
@@ -2252,6 +2254,8 @@ static int logical_ring_init(struct intel_engine_cs *engine)
 	execlists->csb_write =
 		&engine->status_page.page_addr[intel_hws_csb_write_index(i915)];
 
+	execlists->csb_size = GEN8_CSB_ENTRIES;
+
 	reset_csb_pointers(execlists);
 
 	return 0;
diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h
index f5a5502ecf70..206b0da6fbcc 100644
--- a/drivers/gpu/drm/i915/intel_lrc.h
+++ b/drivers/gpu/drm/i915/intel_lrc.h
@@ -36,9 +36,13 @@
 #define	  CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT	(1 << 0)
 #define   CTX_CTRL_RS_CTX_ENABLE                (1 << 1)
 #define	  CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT	(1 << 2)
-#define RING_CONTEXT_STATUS_BUF_BASE(engine)	_MMIO((engine)->mmio_base + 0x370)
-#define RING_CONTEXT_STATUS_BUF_LO(engine, i)	_MMIO((engine)->mmio_base + 0x370 + (i) * 8)
-#define RING_CONTEXT_STATUS_BUF_HI(engine, i)	_MMIO((engine)->mmio_base + 0x370 + (i) * 8 + 4)
+
+#define _RING_CSB_OFFSET(i)			((i) < 6 ? \
+						 (0x370 + (i) * 8) : \
+						 (0x3c0 + ((i) - 6) * 8))
+#define RING_CSB_OFFSET(engine, i)		((engine)->mmio_base + _RING_CSB_OFFSET(i))
+#define RING_CSB_LO(engine, i)			_MMIO(RING_CSB_OFFSET(engine, i))
+#define RING_CSB_HI(engine, i)			_MMIO(RING_CSB_OFFSET(engine, i) + 4)
 #define RING_CONTEXT_STATUS_PTR(engine)		_MMIO((engine)->mmio_base + 0x3a0)
 #define RING_EXECLIST_SQ_CONTENTS(engine)	_MMIO((engine)->mmio_base + 0x510)
 #define RING_EXECLIST_CONTROL(engine)		_MMIO((engine)->mmio_base + 0x550)
@@ -55,10 +59,11 @@
 #define GEN8_CSB_PTR_MASK 0x7
 #define GEN8_CSB_READ_PTR_MASK (GEN8_CSB_PTR_MASK << 8)
 #define GEN8_CSB_WRITE_PTR_MASK (GEN8_CSB_PTR_MASK << 0)
-#define GEN8_CSB_WRITE_PTR(csb_status) \
-	(((csb_status) & GEN8_CSB_WRITE_PTR_MASK) >> 0)
-#define GEN8_CSB_READ_PTR(csb_status) \
-	(((csb_status) & GEN8_CSB_READ_PTR_MASK) >> 8)
+
+#define GEN11_CSB_ENTRIES 12
+#define GEN11_CSB_PTR_MASK 0xf
+#define GEN11_CSB_READ_PTR_MASK (GEN11_CSB_PTR_MASK << 8)
+#define GEN11_CSB_WRITE_PTR_MASK (GEN11_CSB_PTR_MASK << 0)
 
 enum {
 	INTEL_CONTEXT_SCHEDULE_IN = 0,
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 1ae74e579386..64a54d8f9b2b 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -332,6 +332,11 @@ struct intel_engine_execlists {
 	 */
 	u32 preempt_complete_status;
 
+	/**
+	 * @csb_size: context status buffer FIFO size
+	 */
+	u8 csb_size;
+
 	/**
 	 * @csb_head: context status buffer head
 	 */
-- 
2.17.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 19+ messages in thread
* [PATCH 1/3] drm/i915: Prepare for larger CSB status FIFO size
@ 2018-11-27 17:38 Mika Kuoppala
  2018-11-27 17:38 ` [PATCH 2/3] drm/i915/icl: Switch to using 12 deep CSB status FIFO Mika Kuoppala
  0 siblings, 1 reply; 19+ messages in thread
From: Mika Kuoppala @ 2018-11-27 17:38 UTC (permalink / raw)
  To: intel-gfx

Make csb entry count variable in preparation for larger
CSB status FIFO size found on gen11+ hardware.

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Signed-off-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
---
 drivers/gpu/drm/i915/intel_engine_cs.c  | 16 +++++++++-------
 drivers/gpu/drm/i915/intel_lrc.c        | 24 +++++++++++++++++-------
 drivers/gpu/drm/i915/intel_lrc.h        |  9 +++++----
 drivers/gpu/drm/i915/intel_ringbuffer.h | 15 +++++++++------
 4 files changed, 40 insertions(+), 24 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c
index 759c0fd58f8c..218ef817da1d 100644
--- a/drivers/gpu/drm/i915/intel_engine_cs.c
+++ b/drivers/gpu/drm/i915/intel_engine_cs.c
@@ -1347,27 +1347,29 @@ static void intel_engine_print_registers(const struct intel_engine_cs *engine,
 		unsigned int idx;
 		u8 read, write;
 
-		drm_printf(m, "\tExeclist status: 0x%08x %08x\n",
+		drm_printf(m, "\tExeclist status: 0x%08x %08x, entries %u\n",
 			   I915_READ(RING_EXECLIST_STATUS_LO(engine)),
-			   I915_READ(RING_EXECLIST_STATUS_HI(engine)));
+			   I915_READ(RING_EXECLIST_STATUS_HI(engine)),
+			   execlists->csb_entries);
 
 		read = execlists->csb_head;
 		write = READ_ONCE(*execlists->csb_write);
 
 		drm_printf(m, "\tExeclist CSB read %d, write %d [mmio:%d], tasklet queued? %s (%s)\n",
 			   read, write,
-			   GEN8_CSB_WRITE_PTR(I915_READ(RING_CONTEXT_STATUS_PTR(engine))),
+			   I915_READ(RING_CONTEXT_STATUS_PTR(engine)) &
+			   GEN11_CSB_WRITE_PTR_MASK,
 			   yesno(test_bit(TASKLET_STATE_SCHED,
 					  &engine->execlists.tasklet.state)),
 			   enableddisabled(!atomic_read(&engine->execlists.tasklet.count)));
-		if (read >= GEN8_CSB_ENTRIES)
+		if (read >= execlists->csb_entries)
 			read = 0;
-		if (write >= GEN8_CSB_ENTRIES)
+		if (write >= execlists->csb_entries)
 			write = 0;
 		if (read > write)
-			write += GEN8_CSB_ENTRIES;
+			write += execlists->csb_entries;
 		while (read < write) {
-			idx = ++read % GEN8_CSB_ENTRIES;
+			idx = ++read % execlists->csb_entries;
 			drm_printf(m, "\tExeclist CSB[%d]: 0x%08x [mmio:0x%08x], context: %d [mmio:%d]\n",
 				   idx,
 				   hws[idx * 2],
diff --git a/drivers/gpu/drm/i915/intel_lrc.c b/drivers/gpu/drm/i915/intel_lrc.c
index 08fd9b12e4d7..5487fe496bb6 100644
--- a/drivers/gpu/drm/i915/intel_lrc.c
+++ b/drivers/gpu/drm/i915/intel_lrc.c
@@ -767,6 +767,7 @@ execlists_cancel_port_requests(struct intel_engine_execlists * const execlists)
 
 static void reset_csb_pointers(struct intel_engine_execlists *execlists)
 {
+	u32 reset_val;
 	/*
 	 * After a reset, the HW starts writing into CSB entry [0]. We
 	 * therefore have to set our HEAD pointer back one entry so that
@@ -776,8 +777,19 @@ static void reset_csb_pointers(struct intel_engine_execlists *execlists)
 	 * inline comparison of our cached head position against the last HW
 	 * write works even before the first interrupt.
 	 */
-	execlists->csb_head = execlists->csb_write_reset;
-	WRITE_ONCE(*execlists->csb_write, execlists->csb_write_reset);
+	execlists->csb_head = execlists->csb_entries - 1;
+
+	if (execlists_mmio_mode(execlists)) {
+		const u32 mask = execlists->csb_entries == GEN8_CSB_ENTRIES ?
+			GEN8_CSB_WRITE_PTR_MASK :
+			GEN11_CSB_WRITE_PTR_MASK;
+
+		reset_val = _MASKED_FIELD(mask, execlists->csb_head);
+	} else {
+		reset_val = execlists->csb_head;
+	}
+
+	WRITE_ONCE(*execlists->csb_write, reset_val);
 }
 
 static void nop_submission_tasklet(unsigned long data)
@@ -895,7 +907,7 @@ static void process_csb(struct intel_engine_cs *engine)
 		unsigned int status;
 		unsigned int count;
 
-		if (++head == GEN8_CSB_ENTRIES)
+		if (++head == execlists->csb_entries)
 			head = 0;
 
 		/*
@@ -2264,6 +2276,8 @@ static int logical_ring_init(struct intel_engine_cs *engine)
 			upper_32_bits(ce->lrc_desc);
 	}
 
+	execlists->csb_entries = GEN8_CSB_ENTRIES;
+
 	execlists->csb_read =
 		i915->regs + i915_mmio_reg_offset(RING_CONTEXT_STATUS_PTR(engine));
 	if (csb_force_mmio(i915)) {
@@ -2271,16 +2285,12 @@ static int logical_ring_init(struct intel_engine_cs *engine)
 			(i915->regs + i915_mmio_reg_offset(RING_CONTEXT_STATUS_BUF_LO(engine, 0)));
 
 		execlists->csb_write = (u32 __force *)execlists->csb_read;
-		execlists->csb_write_reset =
-			_MASKED_FIELD(GEN8_CSB_WRITE_PTR_MASK,
-				      GEN8_CSB_ENTRIES - 1);
 	} else {
 		execlists->csb_status =
 			&engine->status_page.page_addr[I915_HWS_CSB_BUF0_INDEX];
 
 		execlists->csb_write =
 			&engine->status_page.page_addr[intel_hws_csb_write_index(i915)];
-		execlists->csb_write_reset = GEN8_CSB_ENTRIES - 1;
 	}
 	reset_csb_pointers(execlists);
 
diff --git a/drivers/gpu/drm/i915/intel_lrc.h b/drivers/gpu/drm/i915/intel_lrc.h
index f5a5502ecf70..e865cba43a26 100644
--- a/drivers/gpu/drm/i915/intel_lrc.h
+++ b/drivers/gpu/drm/i915/intel_lrc.h
@@ -55,10 +55,11 @@
 #define GEN8_CSB_PTR_MASK 0x7
 #define GEN8_CSB_READ_PTR_MASK (GEN8_CSB_PTR_MASK << 8)
 #define GEN8_CSB_WRITE_PTR_MASK (GEN8_CSB_PTR_MASK << 0)
-#define GEN8_CSB_WRITE_PTR(csb_status) \
-	(((csb_status) & GEN8_CSB_WRITE_PTR_MASK) >> 0)
-#define GEN8_CSB_READ_PTR(csb_status) \
-	(((csb_status) & GEN8_CSB_READ_PTR_MASK) >> 8)
+
+#define GEN11_CSB_ENTRIES 12
+#define GEN11_CSB_PTR_MASK 0xf
+#define GEN11_CSB_READ_PTR_MASK (GEN11_CSB_PTR_MASK << 8)
+#define GEN11_CSB_WRITE_PTR_MASK (GEN11_CSB_PTR_MASK << 0)
 
 enum {
 	INTEL_CONTEXT_SCHEDULE_IN = 0,
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.h b/drivers/gpu/drm/i915/intel_ringbuffer.h
index 8a2270b209b0..d9520a52fe29 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.h
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.h
@@ -339,13 +339,9 @@ struct intel_engine_execlists {
 	u32 preempt_complete_status;
 
 	/**
-	 * @csb_write_reset: reset value for CSB write pointer
-	 *
-	 * As the CSB write pointer maybe either in HWSP or as a field
-	 * inside an mmio register, we want to reprogram it slightly
-	 * differently to avoid later confusion.
+	 * @csb_entries: number of CSB entries
 	 */
-	u32 csb_write_reset;
+	u8 csb_entries;
 
 	/**
 	 * @csb_head: context status buffer head
@@ -714,6 +710,13 @@ execlists_is_active(const struct intel_engine_execlists *execlists,
 	return test_bit(bit, (unsigned long *)&execlists->active);
 }
 
+static inline bool
+execlists_mmio_mode(const struct intel_engine_execlists * const execlists)
+{
+	return (unsigned long)execlists->csb_read ==
+		(unsigned long)execlists->csb_write;
+}
+
 void execlists_user_begin(struct intel_engine_execlists *execlists,
 			  const struct execlist_port *port);
 void execlists_user_end(struct intel_engine_execlists *execlists);
-- 
2.17.1

_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

^ permalink raw reply related	[flat|nested] 19+ messages in thread
end of thread, other threads:[~2018-12-19 16:48 UTC | newest]

Thread overview: 19+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2018-12-18 13:24 [PATCH 1/3] drm/i915: Prepare for larger CSB status FIFO size Mika Kuoppala
2018-12-18 13:24 ` [PATCH 2/3] drm/i915/icl: Switch to using 12 deep CSB status FIFO Mika Kuoppala
2018-12-18 13:24 ` [PATCH 3/3] drm/i915/icl: Introduce gen11 flush/invalidate Mika Kuoppala
2018-12-18 13:47 ` [PATCH 1/3] drm/i915: Prepare for larger CSB status FIFO size Chris Wilson
2018-12-19 12:27   ` Mika Kuoppala
2018-12-19 12:40     ` Chris Wilson
2018-12-19 13:17       ` Mika Kuoppala
2018-12-19 13:22         ` Mika Kuoppala
2018-12-19 13:26       ` Mika Kuoppala
2018-12-18 14:05 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/3] " Patchwork
2018-12-18 14:22 ` ✗ Fi.CI.BAT: failure " Patchwork
2018-12-19 12:36 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/3] drm/i915: Prepare for larger CSB status FIFO size (rev2) Patchwork
2018-12-19 13:05 ` ✗ Fi.CI.BAT: failure " Patchwork
2018-12-19 13:29 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/3] drm/i915: Prepare for larger CSB status FIFO size (rev3) Patchwork
2018-12-19 13:58 ` ✓ Fi.CI.BAT: success " Patchwork
2018-12-19 14:46 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [1/3] drm/i915: Prepare for larger CSB status FIFO size (rev4) Patchwork
2018-12-19 15:08 ` ✓ Fi.CI.BAT: success " Patchwork
2018-12-19 16:48 ` ✗ Fi.CI.IGT: failure " Patchwork
  -- strict thread matches above, loose matches on Subject: below --
2018-11-27 17:38 [PATCH 1/3] drm/i915: Prepare for larger CSB status FIFO size Mika Kuoppala
2018-11-27 17:38 ` [PATCH 2/3] drm/i915/icl: Switch to using 12 deep CSB status FIFO Mika Kuoppala

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.