[Intel-gfx] [PATCH v7] drm/i915: handle uncore spinlock when not available

[Sebastian Andrzej Siewior]

On 2023-12-01 12:00:32 [+0200], Luca Coelho wrote:
> To handle this, split the spin_lock/unlock_irqsave/restore() into
> spin_lock/unlock() followed by a call to local_irq_save/restore() and

On PREEMPT_RT spinlock_t becomes a sleeping lock so this split is not
working. See Documentation/locking/locktypes.rst

What I don't understand: do you have to keep the interrupts disabled
for some reasons or is it just to avoid using _irqsave() twice?

I do have more i915 related patches in the PREEMPT_RT queue and I can
post them if you folks have time for it. For now, let me show what I did
here:

----------->8-------------

From: Mike Galbraith <umgwanakikbuti at gmail.com>
Date: Sat, 27 Feb 2016 08:09:11 +0100
Subject: [PATCH 03/10] drm/i915: Use preempt_disable/enable_rt() where
 recommended

Mario Kleiner suggest in commit
  ad3543ede630f ("drm/intel: Push get_scanout_position() timestamping into kms driver.")

a spots where preemption should be disabled on PREEMPT_RT. The
difference is that on PREEMPT_RT the intel_uncore::lock disables neither
preemption nor interrupts and so region remains preemptible.

The area covers only register reads and writes. The part that worries me
is:
- __intel_get_crtc_scanline() the worst case is 100us if no match is
  found.

- intel_crtc_scanlines_since_frame_timestamp() not sure how long this
  may take in the worst case.

It was in the RT queue for a while and nobody complained.
Disable preemption on PREEPMPT_RT during timestamping.

[bigeasy: patch description.]

Cc: Mario Kleiner <mario.kleiner.de at gmail.com>
Signed-off-by: Mike Galbraith <umgwanakikbuti at gmail.com>
Signed-off-by: Thomas Gleixner <tglx at linutronix.de>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy at linutronix.de>
---
 drivers/gpu/drm/i915/display/intel_vblank.c |   38 ++++++++++++++++++++--------
 1 file changed, 28 insertions(+), 10 deletions(-)

--- a/drivers/gpu/drm/i915/display/intel_vblank.c
+++ b/drivers/gpu/drm/i915/display/intel_vblank.c
@@ -275,6 +275,26 @@ int intel_crtc_scanline_to_hw(struct int
  * all register accesses to the same cacheline to be serialized,
  * otherwise they may hang.
  */
+static void intel_vblank_section_enter_irqsave(struct drm_i915_private *i915, unsigned long *flags)
+	__acquires(i915->uncore.lock)
+{
+#ifdef I915
+	spin_lock_irqsave(&i915->uncore.lock, *flags);
+#else
+	*flags = NULL;
+#endif
+}
+
+static void intel_vblank_section_exit_irqrestore(struct drm_i915_private *i915, unsigned long flags)
+	__releases(i915->uncore.lock)
+{
+#ifdef I915
+	spin_unlock_irqrestore(&i915->uncore.lock, flags);
+#else
+	if (flags)
+		;
+#endif
+}
 static void intel_vblank_section_enter(struct drm_i915_private *i915)
 	__acquires(i915->uncore.lock)
 {
@@ -332,10 +352,10 @@ static bool i915_get_crtc_scanoutpos(str
 	 * timing critical raw register reads, potentially with
 	 * preemption disabled, so the following code must not block.
 	 */
-	local_irq_save(irqflags);
-	intel_vblank_section_enter(dev_priv);
+	intel_vblank_section_enter_irqsave(dev_priv, &irqflags);
 
-	/* preempt_disable_rt() should go right here in PREEMPT_RT patchset. */
+	if (IS_ENABLED(CONFIG_PREEMPT_RT))
+		preempt_disable();
 
 	/* Get optional system timestamp before query. */
 	if (stime)
@@ -399,10 +419,10 @@ static bool i915_get_crtc_scanoutpos(str
 	if (etime)
 		*etime = ktime_get();
 
-	/* preempt_enable_rt() should go right here in PREEMPT_RT patchset. */
+	if (IS_ENABLED(CONFIG_PREEMPT_RT))
+		preempt_enable();
 
-	intel_vblank_section_exit(dev_priv);
-	local_irq_restore(irqflags);
+	intel_vblank_section_exit_irqrestore(dev_priv, irqflags);
 
 	/*
 	 * While in vblank, position will be negative
@@ -440,13 +460,11 @@ int intel_get_crtc_scanline(struct intel
 	unsigned long irqflags;
 	int position;
 
-	local_irq_save(irqflags);
-	intel_vblank_section_enter(dev_priv);
+	intel_vblank_section_enter_irqsave(dev_priv, &irqflags);
 
 	position = __intel_get_crtc_scanline(crtc);
 
-	intel_vblank_section_exit(dev_priv);
-	local_irq_restore(irqflags);
+	intel_vblank_section_exit_irqrestore(dev_priv, irqflags);
 
 	return position;
 }