Re: [PATCH] drm/i915/fbdev: Stop repeating tile configuration on stagnation

On Mon, Feb 20, 2017 at 04:37:47PM +0100, Tomasz Lis wrote:

I tested this variant (reverted the change to "pass" variable before
testing), and it fixes the issue with count=0 as well as possible
infinite loop issue.

But something needs to be done with the "pass" var, one way or the
other - commented below.


W dniu 2017-02-18 o 16:37, Chris Wilson pisze:

quoted

If we cease making progress in finding matching outputs for a tiled
configuration, stop looping over the remaining unconfigured outputs.

Fixes: b0ee9e7fa5b4 ("drm/fb: add support for tiled monitor configurations. (v2)")
Cc: Tomasz Lis <redacted>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Daniel Vetter <redacted>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Sean Paul <redacted>
Cc: <redacted> # v3.19+
---
 drivers/gpu/drm/i915/intel_fbdev.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c
index 041322fef607..32d0af6b07a7 100644
--- a/drivers/gpu/drm/i915/intel_fbdev.c
+++ b/drivers/gpu/drm/i915/intel_fbdev.c

@@ -348,7 +348,7 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
 				    bool *enabled, int width, int height)
 {
 	struct drm_i915_private *dev_priv = to_i915(fb_helper->dev);
-	unsigned long conn_configured, mask;
+	unsigned long conn_configured, conn_seq, mask;
 	unsigned int count = min(fb_helper->connector_count, BITS_PER_LONG);
 	int i, j;
 	bool *save_enabled;

@@ -365,6 +365,7 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
 	mask = GENMASK(count - 1, 0);
 	conn_configured = 0;
 retry:
+	conn_seq = conn_configured;
 	for (i = 0; i < count; i++) {
 		struct drm_fb_helper_connector *fb_conn;
 		struct drm_connector *connector;

@@ -489,10 +490,8 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
 		conn_configured |= BIT(i);
 	}
-	if ((conn_configured & mask) != mask) {
-		pass++;

This doesn't seem right; increasing the amount of passes should
stay, or the use of "pass" variable should be completely replaced by
conn_seq.

Good catch.

diff --git a/drivers/gpu/drm/i915/intel_fbdev.c b/drivers/gpu/drm/i915/intel_fbdev.c
index 32d0af6b07a7..f7e9a4e69595 100644
--- a/drivers/gpu/drm/i915/intel_fbdev.c
+++ b/drivers/gpu/drm/i915/intel_fbdev.c

@@ -355,7 +355,6 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
        bool fallback = true;
        int num_connectors_enabled = 0;
        int num_connectors_detected = 0;
-       int pass = 0;
 
        save_enabled = kcalloc(count, sizeof(bool), GFP_KERNEL);
        if (!save_enabled)

@@ -379,7 +378,7 @@ static bool intel_fb_initial_config(struct drm_fb_helper *fb_helper,
                if (conn_configured & BIT(i))
                        continue;
 
-               if (pass == 0 && !connector->has_tile)
+               if (conn_seq == 0 && !connector->has_tile)
                        continue;
 
                if (connector->status == connector_status_connected)

Suggestions for a better name than conn_seq much appreciated.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre