i965: improve debug logging

Looking for memory leaks that were causing crashes in my environment
in a situation where valgrind would not work, I ended up improving
the i965 debug traces so I could better see where the memory was
being allocated and where it was going, in the regions and miptrees
code, and in the state caches.  These traces were specific enough
that external scripts could determine what elements were not being
released, and where the memory leaks were.

I also ended up creating my own backtrace code in intel_regions.c,
to determine exactly where regions were being allocated and for what,
since valgrind wasn't working.  Because it was useful, I left it in,
but disabled and compiled out.  It can be activated by changing a flag
at the top of the file.

3 files changed, 86 insertions, 14 deletions

diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index 320d886c992..e40d7a04164 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -521,6 +521,9 @@ brw_clear_cache(struct brw_context *brw, struct brw_cache *cache)
 void
 brw_state_cache_check_size(struct brw_context *brw)
 {
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s (n_items=%d)\n", __FUNCTION__, brw->cache.n_items);
+
    /* un-tuned guess.  We've got around 20 state objects for a total of around
     * 32k, so 1000 of them is around 1.5MB.
     */
@@ -537,6 +540,9 @@ brw_destroy_cache(struct brw_context *brw, struct brw_cache *cache)
 {
    GLuint i;
 
+   if (INTEL_DEBUG & DEBUG_STATE)
+      _mesa_printf("%s\n", __FUNCTION__);
+
    brw_clear_cache(brw, cache);
    for (i = 0; i < BRW_MAX_CACHE; i++) {
       dri_bo_unreference(cache->last_bo[i]);
diff --git a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
index 6e1e034e53d..f3652720ece 100644
--- a/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/intel/intel_mipmap_tree.c
@@ -62,9 +62,10 @@ intel_miptree_create_internal(struct intel_context *intel,
    GLboolean ok;
    struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1);
 
-   DBG("%s target %s format %s level %d..%d\n", __FUNCTION__,
+   DBG("%s target %s format %s level %d..%d <-- %p\n", __FUNCTION__,
        _mesa_lookup_enum_by_nr(target),
-       _mesa_lookup_enum_by_nr(internal_format), first_level, last_level);
+       _mesa_lookup_enum_by_nr(internal_format), 
+       first_level, last_level, mt);
 
    mt->target = target_to_target(target);
    mt->internal_format = internal_format;
@@ -89,6 +90,7 @@ intel_miptree_create_internal(struct intel_context *intel,
 
    if (!ok) {
       free(mt);
+      DBG("%s not okay - returning NULL\n", __FUNCTION__);
       return NULL;
    }
 
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index 0aa5b8c02c9..534e75efe1f 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -52,12 +52,66 @@
 
 #define FILE_DEBUG_FLAG DEBUG_REGION
 
+/* This should be set to the maximum backtrace size desired.
+ * Set it to 0 to disable backtrace debugging.
+ */
+#define DEBUG_BACKTRACE_SIZE 0
+
+#if DEBUG_BACKTRACE_SIZE == 0
+/* Use the standard debug output */
+#define _DBG(...) DBG(__VA_ARGS__)
+#else
+/* Use backtracing debug output */
+#define _DBG(...) {debug_backtrace(); DBG(__VA_ARGS__);}
+
+/* Backtracing debug support */
+#include <execinfo.h>
+
+static void
+debug_backtrace(void)
+{
+   void *trace[DEBUG_BACKTRACE_SIZE];
+   char **strings = NULL;
+   int traceSize;
+   register int i;
+
+   traceSize = backtrace(trace, DEBUG_BACKTRACE_SIZE);
+   strings = backtrace_symbols(trace, traceSize);
+   if (strings == NULL) {
+      DBG("no backtrace:");
+      return;
+   }
+
+   /* Spit out all the strings with a colon separator.  Ignore
+    * the first, since we don't really care about the call
+    * to debug_backtrace() itself.  Skip until the final "/" in
+    * the trace to avoid really long lines.
+    */
+   for (i = 1; i < traceSize; i++) {
+      char *p = strings[i], *slash = strings[i];
+      while (*p) {
+         if (*p++ == '/') {
+            slash = p;
+         }
+      }
+
+      DBG("%s:", slash);
+   }
+
+   /* Free up the memory, and we're done */
+   free(strings);
+}
+
+#endif
+
+
+
 /* XXX: Thread safety?
  */
 GLubyte *
 intel_region_map(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!region->map_refcount++) {
       if (region->pbo)
          intel_region_cow(intel, region);
@@ -72,7 +126,7 @@ intel_region_map(struct intel_context *intel, struct intel_region *region)
 void
 intel_region_unmap(struct intel_context *intel, struct intel_region *region)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s %p\n", __FUNCTION__, region);
    if (!--region->map_refcount) {
       dri_bo_unmap(region->buffer);
       region->map = NULL;
@@ -87,10 +141,10 @@ intel_region_alloc_internal(struct intel_context *intel,
 {
    struct intel_region *region;
 
-   DBG("%s\n", __FUNCTION__);
-
-   if (buffer == NULL)
+   if (buffer == NULL) {
+      _DBG("%s <-- NULL\n", __FUNCTION__);
       return NULL;
+   }
 
    region = calloc(sizeof(*region), 1);
    region->cpp = cpp;
@@ -104,6 +158,7 @@ intel_region_alloc_internal(struct intel_context *intel,
    region->tiling = I915_TILING_NONE;
    region->bit_6_swizzle = I915_BIT_6_SWIZZLE_NONE;
 
+   _DBG("%s <-- %p\n", __FUNCTION__, region);
    return region;
 }
 
@@ -158,7 +213,7 @@ void
 intel_region_reference(struct intel_region **dst, struct intel_region *src)
 {
    if (src)
-      DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
+      _DBG("%s %p %d\n", __FUNCTION__, src, src->refcount);
 
    assert(*dst == NULL);
    if (src) {
@@ -172,10 +227,12 @@ intel_region_release(struct intel_region **region_handle)
 {
    struct intel_region *region = *region_handle;
 
-   if (region == NULL)
+   if (region == NULL) {
+      _DBG("%s NULL\n", __FUNCTION__);
       return;
+   }
 
-   DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
+   _DBG("%s %p %d\n", __FUNCTION__, region, region->refcount - 1);
 
    ASSERT(region->refcount > 0);
    region->refcount--;
@@ -251,7 +308,7 @@ intel_region_data(struct intel_context *intel,
 {
    GLboolean locked = GL_FALSE;
 
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -293,7 +350,7 @@ intel_region_copy(struct intel_context *intel,
                   GLuint src_offset,
                   GLuint srcx, GLuint srcy, GLuint width, GLuint height)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;
@@ -326,7 +383,7 @@ intel_region_fill(struct intel_context *intel,
                   GLuint dstx, GLuint dsty,
                   GLuint width, GLuint height, GLuint color)
 {
-   DBG("%s\n", __FUNCTION__);
+   _DBG("%s\n", __FUNCTION__);
 
    if (intel == NULL)
       return;   
@@ -356,6 +413,8 @@ intel_region_attach_pbo(struct intel_context *intel,
    if (region->pbo == pbo)
       return;
 
+   _DBG("%s %p %p\n", __FUNCTION__, region, pbo);
+
    /* If there is already a pbo attached, break the cow tie now.
     * Don't call intel_region_release_pbo() as that would
     * unnecessarily allocate a new buffer we would have to immediately
@@ -385,6 +444,7 @@ void
 intel_region_release_pbo(struct intel_context *intel,
                          struct intel_region *region)
 {
+   _DBG("%s %p\n", __FUNCTION__, region);
    assert(region->buffer == region->pbo->buffer);
    region->pbo->region = NULL;
    region->pbo = NULL;
@@ -412,7 +472,7 @@ intel_region_cow(struct intel_context *intel, struct intel_region *region)
 
    assert(region->cpp * region->pitch * region->height == pbo->Base.Size);
 
-   DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size);
+   _DBG("%s %p (%d bytes)\n", __FUNCTION__, region, pbo->Base.Size);
 
    /* Now blit from the texture buffer to the new buffer: 
     */
@@ -459,6 +519,10 @@ intel_recreate_static(struct intel_context *intel,
    if (region == NULL) {
       region = calloc(sizeof(*region), 1);
       region->refcount = 1;
+      _DBG("%s creating new region %p\n", __FUNCTION__, region);
+   }
+   else {
+      _DBG("%s %p\n", __FUNCTION__, region);
    }
 
    if (intel->ctx.Visual.rgbBits == 24)