Merge commit 'origin/gallium-0.1' into gallium-0.2

Conflicts:

	Makefile
	progs/demos/Makefile
	progs/glsl/Makefile
	progs/redbook/Makefile
	progs/samples/Makefile
	progs/tests/Makefile
	progs/trivial/Makefile
	progs/xdemos/Makefile
	src/gallium/Makefile
	src/mesa/main/attrib.c
	src/mesa/main/bufferobj.c
	src/mesa/vbo/vbo_exec_draw.c

18 files changed, 206 insertions, 537 deletions

diff --git a/src/gallium/winsys/drm/intel/common/glthread.h b/src/gallium/winsys/drm/intel/common/glthread.h
deleted file mode 100644
index b8e9d5f59b8..00000000000
--- a/src/gallium/winsys/drm/intel/common/glthread.h
+++ /dev/null
@@ -1,359 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.2
- *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the "Software"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included
- * in all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
- * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
- * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
- * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
- */
-
-
-/*
- * Thread support for gl dispatch.
- *
- * Initial version by John Stone ([email protected]) ([email protected])
- *                and Christoph Poliwoda ([email protected])
- * Revised by Keith Whitwell
- * Adapted for new gl dispatcher by Brian Paul
- *
- *
- *
- * DOCUMENTATION
- *
- * This thread module exports the following types:
- *   _glthread_TSD     Thread-specific data area
- *   _glthread_Thread  Thread datatype
- *   _glthread_Mutex   Mutual exclusion lock
- *
- * Macros:
- *   _glthread_DECLARE_STATIC_MUTEX(name)   Declare a non-local mutex
- *   _glthread_INIT_MUTEX(name)             Initialize a mutex
- *   _glthread_LOCK_MUTEX(name)             Lock a mutex
- *   _glthread_UNLOCK_MUTEX(name)           Unlock a mutex
- *
- * Functions:
- *   _glthread_GetID(v)      Get integer thread ID
- *   _glthread_InitTSD()     Initialize thread-specific data
- *   _glthread_GetTSD()      Get thread-specific data
- *   _glthread_SetTSD()      Set thread-specific data
- *
- */
-
-/*
- * If this file is accidentally included by a non-threaded build,
- * it should not cause the build to fail, or otherwise cause problems.
- * In general, it should only be included when needed however.
- */
-
-#ifndef GLTHREAD_H
-#define GLTHREAD_H
-
-
-#if defined(USE_MGL_NAMESPACE)
-#define _glapi_Dispatch _mglapi_Dispatch
-#endif
-
-
-
-#if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\
-     defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \
-    && !defined(THREADS)
-# define THREADS
-#endif
-
-#ifdef VMS
-#include <GL/vms_x_fix.h>
-#endif
-
-/*
- * POSIX threads. This should be your choice in the Unix world
- * whenever possible.  When building with POSIX threads, be sure
- * to enable any compiler flags which will cause the MT-safe
- * libc (if one exists) to be used when linking, as well as any
- * header macros for MT-safe errno, etc.  For Solaris, this is the -mt
- * compiler flag.  On Solaris with gcc, use -D_REENTRANT to enable
- * proper compiling for MT-safe libc etc.
- */
-#if defined(PTHREADS)
-#include <pthread.h> /* POSIX threads headers */
-
-typedef struct {
-   pthread_key_t  key;
-   int initMagic;
-} _glthread_TSD;
-
-typedef pthread_t _glthread_Thread;
-
-typedef pthread_mutex_t _glthread_Mutex;
-
-#define _glthread_DECLARE_STATIC_MUTEX(name) \
-   static _glthread_Mutex name = PTHREAD_MUTEX_INITIALIZER
-
-#define _glthread_INIT_MUTEX(name) \
-   pthread_mutex_init(&(name), NULL)
-
-#define _glthread_DESTROY_MUTEX(name) \
-   pthread_mutex_destroy(&(name))
-
-#define _glthread_LOCK_MUTEX(name) \
-   (void) pthread_mutex_lock(&(name))
-
-#define _glthread_UNLOCK_MUTEX(name) \
-   (void) pthread_mutex_unlock(&(name))
-
-typedef pthread_cond_t _glthread_Cond;
-
-#define _glthread_DECLARE_STATIC_COND(name) \
-   static _glthread_Cond name = PTHREAD_COND_INITIALIZER
-
-#define _glthread_INIT_COND(cond)			\
-   pthread_cond_init(&(cond), NULL)
-
-#define _glthread_DESTROY_COND(name) \
-   pthread_cond_destroy(&(name))
-
-#define _glthread_COND_WAIT(cond, mutex) \
-  pthread_cond_wait(&(cond), &(mutex))
-
-#define _glthread_COND_SIGNAL(cond) \
-  pthread_cond_signal(&(cond))
-
-#define _glthread_COND_BROADCAST(cond) \
-  pthread_cond_broadcast(&(cond))
-
-
-#else /* PTHREADS */
-
-typedef unsigned int _glthread_Cond;
-#define _glthread_DECLARE_STATIC_COND(name) \
-//  #warning Condition variables not implemented.
-
-#define _glthread_INIT_COND(cond)	    \
-  abort();
-
-#define _glthread_DESTROY_COND(name) \
-  abort();
-
-#define _glthread_COND_WAIT(cond, mutex) \
-  abort();
-
-#define _glthread_COND_SIGNAL(cond) \
-  abort();
-
-#define _glthread_COND_BROADCAST(cond) \
-  abort();
-
-#endif
-
-
-/*
- * Solaris threads. Use only up to Solaris 2.4.
- * Solaris 2.5 and higher provide POSIX threads.
- * Be sure to compile with -mt on the Solaris compilers, or
- * use -D_REENTRANT if using gcc.
- */
-#ifdef SOLARIS_THREADS
-#include <thread.h>
-
-typedef struct {
-   thread_key_t key;
-   mutex_t      keylock;
-   int          initMagic;
-} _glthread_TSD;
-
-typedef thread_t _glthread_Thread;
-
-typedef mutex_t _glthread_Mutex;
-
-/* XXX need to really implement mutex-related macros */
-#define _glthread_DECLARE_STATIC_MUTEX(name)  static _glthread_Mutex name = 0
-#define _glthread_INIT_MUTEX(name)  (void) name
-#define _glthread_DESTROY_MUTEX(name) (void) name
-#define _glthread_LOCK_MUTEX(name)  (void) name
-#define _glthread_UNLOCK_MUTEX(name)  (void) name
-
-#endif /* SOLARIS_THREADS */
-
-
-
-
-/*
- * Windows threads. Should work with Windows NT and 95.
- * IMPORTANT: Link with multithreaded runtime library when THREADS are
- * used!
- */
-#ifdef WIN32_THREADS
-#include <windows.h>
-
-typedef struct {
-   DWORD key;
-   int   initMagic;
-} _glthread_TSD;
-
-typedef HANDLE _glthread_Thread;
-
-typedef CRITICAL_SECTION _glthread_Mutex;
-
-#define _glthread_DECLARE_STATIC_MUTEX(name)  /*static*/ _glthread_Mutex name = {0,0,0,0,0,0}
-#define _glthread_INIT_MUTEX(name)  InitializeCriticalSection(&name)
-#define _glthread_DESTROY_MUTEX(name)  DeleteCriticalSection(&name)
-#define _glthread_LOCK_MUTEX(name)  EnterCriticalSection(&name)
-#define _glthread_UNLOCK_MUTEX(name)  LeaveCriticalSection(&name)
-
-#endif /* WIN32_THREADS */
-
-
-
-
-/*
- * XFree86 has its own thread wrapper, Xthreads.h
- * We wrap it again for GL.
- */
-#ifdef USE_XTHREADS
-#include <X11/Xthreads.h>
-
-typedef struct {
-   xthread_key_t key;
-   int initMagic;
-} _glthread_TSD;
-
-typedef xthread_t _glthread_Thread;
-
-typedef xmutex_rec _glthread_Mutex;
-
-#ifdef XMUTEX_INITIALIZER
-#define _glthread_DECLARE_STATIC_MUTEX(name) \
-   static _glthread_Mutex name = XMUTEX_INITIALIZER
-#else
-#define _glthread_DECLARE_STATIC_MUTEX(name) \
-   static _glthread_Mutex name
-#endif
-
-#define _glthread_INIT_MUTEX(name) \
-   xmutex_init(&(name))
-
-#define _glthread_DESTROY_MUTEX(name) \
-   xmutex_clear(&(name))
-
-#define _glthread_LOCK_MUTEX(name) \
-   (void) xmutex_lock(&(name))
-
-#define _glthread_UNLOCK_MUTEX(name) \
-   (void) xmutex_unlock(&(name))
-
-#endif /* USE_XTHREADS */
-
-
-
-/*
- * BeOS threads. R5.x required.
- */
-#ifdef BEOS_THREADS
-
-#include <kernel/OS.h>
-#include <support/TLS.h>
-
-typedef struct {
-   int32        key;
-   int          initMagic;
-} _glthread_TSD;
-
-typedef thread_id _glthread_Thread;
-
-/* Use Benaphore, aka speeder semaphore */
-typedef struct {
-    int32   lock;
-    sem_id  sem;
-} benaphore;
-typedef benaphore _glthread_Mutex;
-
-#define _glthread_DECLARE_STATIC_MUTEX(name)  static _glthread_Mutex name = { 0, 0 }
-#define _glthread_INIT_MUTEX(name)    	name.sem = create_sem(0, #name"_benaphore"), name.lock = 0
-#define _glthread_DESTROY_MUTEX(name) 	delete_sem(name.sem), name.lock = 0
-#define _glthread_LOCK_MUTEX(name)    	if (name.sem == 0) _glthread_INIT_MUTEX(name); \
-									  	if (atomic_add(&(name.lock), 1) >= 1) acquire_sem(name.sem)
-#define _glthread_UNLOCK_MUTEX(name)  	if (atomic_add(&(name.lock), -1) > 1) release_sem(name.sem)
-
-#endif /* BEOS_THREADS */
-
-
-
-#ifndef THREADS
-
-/*
- * THREADS not defined
- */
-
-typedef GLuint _glthread_TSD;
-
-typedef GLuint _glthread_Thread;
-
-typedef GLuint _glthread_Mutex;
-
-#define _glthread_DECLARE_STATIC_MUTEX(name)  static _glthread_Mutex name = 0
-
-#define _glthread_INIT_MUTEX(name)  (void) name
-
-#define _glthread_DESTROY_MUTEX(name)  (void) name
-
-#define _glthread_LOCK_MUTEX(name)  (void) name
-
-#define _glthread_UNLOCK_MUTEX(name)  (void) name
-
-#endif /* THREADS */
-
-
-
-/*
- * Platform independent thread specific data API.
- */
-
-extern unsigned long
-_glthread_GetID(void);
-
-
-extern void
-_glthread_InitTSD(_glthread_TSD *);
-
-
-extern void *
-_glthread_GetTSD(_glthread_TSD *);
-
-
-extern void
-_glthread_SetTSD(_glthread_TSD *, void *);
-
-#if defined(GLX_USE_TLS)
-
-extern __thread struct _glapi_table * _glapi_tls_Dispatch
-    __attribute__((tls_model("initial-exec")));
-
-#define GET_DISPATCH() _glapi_tls_Dispatch
-
-#elif !defined(GL_CALL)
-# if defined(THREADS)
-#  define GET_DISPATCH() \
-   ((__builtin_expect( _glapi_Dispatch != NULL, 1 )) \
-       ? _glapi_Dispatch : _glapi_get_dispatch())
-# else
-#  define GET_DISPATCH() _glapi_Dispatch
-# endif /* defined(THREADS) */
-#endif  /* ndef GL_CALL */
-
-
-#endif /* THREADS_H */
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c
index b6d901f85e4..517a97b3ee5 100644
--- a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c
@@ -33,7 +33,7 @@
 #include <xf86drm.h>
 #include <stdlib.h>
 #include <stdio.h>
-#include "glthread.h"
+#include "pipe/p_thread.h"
 #include "errno.h"
 #include "ws_dri_bufmgr.h"
 #include "string.h"
@@ -51,8 +51,8 @@
  * driBufferObject mutex - > this rw lock.
  */
 
-_glthread_DECLARE_STATIC_MUTEX(bmMutex);
-_glthread_DECLARE_STATIC_COND(bmCond);
+pipe_static_mutex(bmMutex);
+pipe_static_condvar(bmCond);
 
 static int kernelReaders = 0;
 static int num_buffers = 0;
@@ -241,29 +241,29 @@ static int drmBOResetList(drmBOList *list)
 
 void driWriteLockKernelBO(void)
 {
-    _glthread_LOCK_MUTEX(bmMutex);
+    pipe_mutex_lock(bmMutex);
     while(kernelReaders != 0)
-	_glthread_COND_WAIT(bmCond, bmMutex);
+	pipe_condvar_wait(bmCond, bmMutex);
 }
 
 void driWriteUnlockKernelBO(void)
 {
-    _glthread_UNLOCK_MUTEX(bmMutex);
+    pipe_mutex_unlock(bmMutex);
 }
 
 void driReadLockKernelBO(void)
 {
-    _glthread_LOCK_MUTEX(bmMutex);
+    pipe_mutex_lock(bmMutex);
     kernelReaders++;
-    _glthread_UNLOCK_MUTEX(bmMutex);
+    pipe_mutex_unlock(bmMutex);
 }
 
 void driReadUnlockKernelBO(void)
 {
-    _glthread_LOCK_MUTEX(bmMutex);
+    pipe_mutex_lock(bmMutex);
     if (--kernelReaders == 0)
-        _glthread_COND_BROADCAST(bmCond);
-    _glthread_UNLOCK_MUTEX(bmMutex);
+       pipe_condvar_broadcast(bmCond);
+    pipe_mutex_unlock(bmMutex);
 }
 
 
@@ -277,7 +277,7 @@ void driReadUnlockKernelBO(void)
 typedef struct _DriBufferObject
 {
    DriBufferPool *pool;
-   _glthread_Mutex mutex;
+   pipe_mutex mutex;
    int refCount;
    const char *name;
    uint64_t flags;
@@ -318,12 +318,12 @@ driBOKernel(struct _DriBufferObject *buf)
    drmBO *ret;
 
    driReadLockKernelBO();
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    assert(buf->private != NULL);
    ret = buf->pool->kernel(buf->pool, buf->private);
    if (!ret)
       BM_CKFATAL(-EINVAL);
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
    driReadUnlockKernelBO();
 
    return ret;
@@ -338,9 +338,9 @@ driBOWaitIdle(struct _DriBufferObject *buf, int lazy)
    * that time??
    */
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy));
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 }
 
 void *
@@ -353,11 +353,11 @@ driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint)
       return buf->userData;
    }
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    assert(buf->private != NULL);
    retval = buf->pool->map(buf->pool, buf->private, flags, hint,
 			   &buf->mutex, &virtual);
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 
    return retval == 0 ? virtual : NULL;
 }
@@ -369,9 +369,9 @@ driBOUnmap(struct _DriBufferObject *buf)
       return;
 
    assert(buf->private != NULL);
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private));
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 }
 
 unsigned long
@@ -381,9 +381,9 @@ driBOOffset(struct _DriBufferObject *buf)
 
    assert(buf->private != NULL);
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    ret = buf->pool->offset(buf->pool, buf->private);
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
    return ret;
 }
 
@@ -394,9 +394,9 @@ driBOPoolOffset(struct _DriBufferObject *buf)
 
    assert(buf->private != NULL);
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    ret = buf->pool->poolOffset(buf->pool, buf->private);
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
    return ret;
 }
 
@@ -408,9 +408,9 @@ driBOFlags(struct _DriBufferObject *buf)
    assert(buf->private != NULL);
 
    driReadLockKernelBO();
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    ret = buf->pool->flags(buf->pool, buf->private);
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
    driReadUnlockKernelBO();
    return ret;
 }
@@ -418,12 +418,12 @@ driBOFlags(struct _DriBufferObject *buf)
 struct _DriBufferObject *
 driBOReference(struct _DriBufferObject *buf)
 {
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    if (++buf->refCount == 1) {
-      _glthread_UNLOCK_MUTEX(buf->mutex);
+      pipe_mutex_unlock(buf->mutex);
       BM_CKFATAL(-EINVAL);
    }
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
    return buf;
 }
 
@@ -435,10 +435,10 @@ driBOUnReference(struct _DriBufferObject *buf)
    if (!buf)
       return;
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    tmp = --buf->refCount;
    if (!tmp) {
-      _glthread_UNLOCK_MUTEX(buf->mutex);
+      pipe_mutex_unlock(buf->mutex);
       if (buf->private) {
 	 if (buf->createdByReference)
 	    buf->pool->unreference(buf->pool, buf->private);
@@ -451,7 +451,7 @@ driBOUnReference(struct _DriBufferObject *buf)
 	 num_buffers--;
       free(buf);
    } else
-     _glthread_UNLOCK_MUTEX(buf->mutex);
+     pipe_mutex_unlock(buf->mutex);
 
 }
 
@@ -469,7 +469,7 @@ driBOData(struct _DriBufferObject *buf,
 
    assert(!buf->userBuffer); /* XXX just do a memcpy? */
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    pool = buf->pool;
 
    if (pool == NULL && newPool != NULL) {
@@ -556,7 +556,7 @@ driBOData(struct _DriBufferObject *buf,
    }
 
  out:
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 
    return retval;
 }
@@ -569,7 +569,7 @@ driBOSubData(struct _DriBufferObject *buf,
 
    assert(!buf->userBuffer); /* XXX just do a memcpy? */
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    if (size && data) {
       BM_CKFATAL(buf->pool->map(buf->pool, buf->private,
                                 DRM_BO_FLAG_WRITE, 0, &buf->mutex,
@@ -577,7 +577,7 @@ driBOSubData(struct _DriBufferObject *buf,
       memcpy((unsigned char *) virtual + offset, data, size);
       BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private));
    }
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 }
 
 void
@@ -588,21 +588,21 @@ driBOGetSubData(struct _DriBufferObject *buf,
 
    assert(!buf->userBuffer); /* XXX just do a memcpy? */
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    if (size && data) {
       BM_CKFATAL(buf->pool->map(buf->pool, buf->private,
                                 DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual));
       memcpy(data, (unsigned char *) virtual + offset, size);
       BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private));
    }
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 }
 
 void
 driBOSetReferenced(struct _DriBufferObject *buf,
 		   unsigned long handle)
 {
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    if (buf->private != NULL) {
       assert((size_t)"Invalid buffer for setReferenced\n" & 0);
       BM_CKFATAL(-EINVAL);
@@ -619,7 +619,7 @@ driBOSetReferenced(struct _DriBufferObject *buf,
    }
    buf->createdByReference = TRUE;
    buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags;
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 }
 
 int
@@ -644,8 +644,8 @@ driGenBuffers(struct _DriBufferPool *pool,
       if (!buf)
 	 return -ENOMEM;
 
-      _glthread_INIT_MUTEX(buf->mutex);
-      _glthread_LOCK_MUTEX(buf->mutex);
+      pipe_mutex_init(buf->mutex);
+      pipe_mutex_lock(buf->mutex);
       buf->refCount = 1;
       buf->flags = flags;
       buf->hint = hint;
@@ -653,7 +653,7 @@ driGenBuffers(struct _DriBufferPool *pool,
       buf->alignment = alignment;
       buf->pool = pool;
       buf->createdByReference = 0;
-      _glthread_UNLOCK_MUTEX(buf->mutex);
+      pipe_mutex_unlock(buf->mutex);
       buffers[i] = buf;
    }
    return 0;
@@ -818,7 +818,7 @@ driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf,
 {
    int newItem;
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    BM_CKFATAL(driAddValidateItem(&list->drmBuffers,
 				 buf->pool->kernel(buf->pool, buf->private),
                                  flags, mask, itemLoc, node));
@@ -827,7 +827,7 @@ driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf,
    if (newItem)
      buf->refCount++;
 
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 }
 
 drmBOList *driGetdrmBOList(struct _DriBufferList *list)
@@ -845,10 +845,10 @@ void driPutdrmBOList(struct _DriBufferList *list)
 void
 driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence)
 {
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    if (buf->pool->fence)
        BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence));
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 
 }
 
@@ -908,10 +908,10 @@ driBOValidateUserList(struct _DriBufferList * list)
 
     while (curBuf) {
 	buf = (struct _DriBufferObject *) drmBOListBuf(curBuf);
-	_glthread_LOCK_MUTEX(buf->mutex);
+	pipe_mutex_lock(buf->mutex);
 	if (buf->pool->validate)
 	    BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex));
-	_glthread_UNLOCK_MUTEX(buf->mutex);
+	pipe_mutex_unlock(buf->mutex);
 	curBuf = drmBOListNext(&list->driBuffers, curBuf);
     }
 }
@@ -929,9 +929,9 @@ driBOSize(struct _DriBufferObject *buf)
 {
   unsigned long size;
 
-   _glthread_LOCK_MUTEX(buf->mutex);
+   pipe_mutex_lock(buf->mutex);
    size = buf->pool->size(buf->pool, buf->private);
-   _glthread_UNLOCK_MUTEX(buf->mutex);
+   pipe_mutex_unlock(buf->mutex);
 
   return size;
 
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h
index bf607989241..ad3b6f3931c 100644
--- a/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h
@@ -33,14 +33,14 @@
 #define _PSB_BUFPOOL_H_
 
 #include <xf86drm.h>
-#include <glthread.h>
+#include "pipe/p_thread.h"
 struct _DriFenceObject;
 
 typedef struct _DriBufferPool
 {
    int fd;
    int (*map) (struct _DriBufferPool * pool, void *private,
-               unsigned flags, int hint, _glthread_Mutex *mutex,
+               unsigned flags, int hint, pipe_mutex *mutex,
 	       void **virtual);
    int (*unmap) (struct _DriBufferPool * pool, void *private);
    int (*destroy) (struct _DriBufferPool * pool, void *private);
@@ -55,8 +55,8 @@ typedef struct _DriBufferPool
    int (*fence) (struct _DriBufferPool * pool, void *private,
                  struct _DriFenceObject * fence);
    drmBO *(*kernel) (struct _DriBufferPool * pool, void *private);
-   int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex);
-   int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex,
+   int (*validate) (struct _DriBufferPool * pool, void *private, pipe_mutex *mutex);
+   int (*waitIdle) (struct _DriBufferPool *pool, void *private, pipe_mutex *mutex,
 		    int lazy);
    int (*setStatus)  (struct _DriBufferPool *pool, void *private,
 		      uint64_t flag_diff, uint64_t old_flags);
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c
index 40929efa2f9..54618b1c82a 100644
--- a/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c
@@ -113,7 +113,7 @@ pool_unreference(struct _DriBufferPool *pool, void *private)
 
 static int
 pool_map(struct _DriBufferPool *pool, void *private, unsigned flags,
-         int hint, _glthread_Mutex *mutex, void **virtual)
+         int hint, pipe_mutex *mutex, void **virtual)
 {
    drmBO *buf = (drmBO *) private;
    int ret;
@@ -202,7 +202,7 @@ pool_kernel(struct _DriBufferPool *pool, void *private)
 }
 
 static int
-pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex,
+pool_waitIdle(struct _DriBufferPool *pool, void *private, pipe_mutex *mutex,
 	      int lazy)
 {
    drmBO *buf = (drmBO *) private;
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c
index b56bc269da5..831c75d30cc 100644
--- a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c
@@ -1,5 +1,5 @@
 #include "ws_dri_fencemgr.h"
-#include "glthread.h"
+#include "pipe/p_thread.h"
 #include <xf86mm.h>
 #include <string.h>
 #include <unistd.h>
@@ -20,7 +20,7 @@ struct _DriFenceMgr {
     /*
      * These members are protected by this->mutex
      */
-    _glthread_Mutex mutex;
+    pipe_mutex mutex;
     int refCount;
     drmMMListHead *heads;
     int num_fences;
@@ -44,7 +44,7 @@ struct _DriFenceObject {
     /*
      * These members are protected by this->mutex.
      */
-    _glthread_Mutex mutex;
+    pipe_mutex mutex;
     uint32_t signaled_type;
     void *private;
 };
@@ -65,8 +65,8 @@ driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info)
   if (!tmp)
       return NULL;
 
-  _glthread_INIT_MUTEX(tmp->mutex);
-  _glthread_LOCK_MUTEX(tmp->mutex);
+  pipe_mutex_init(tmp->mutex);
+  pipe_mutex_lock(tmp->mutex);
   tmp->refCount = 1;
   tmp->info = *info;
   tmp->num_fences = 0;
@@ -77,7 +77,7 @@ driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info)
   for (i=0; i<tmp->info.num_classes; ++i) {
       DRMINITLISTHEAD(&tmp->heads[i]);
   }
-  _glthread_UNLOCK_MUTEX(tmp->mutex);
+  pipe_mutex_unlock(tmp->mutex);
   return tmp;
 
   out_err:
@@ -95,13 +95,13 @@ driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr)
     if (--mgr->refCount == 0)
 	free(mgr);
     else
-	_glthread_UNLOCK_MUTEX(mgr->mutex);
+	pipe_mutex_unlock(mgr->mutex);
 }
 
 void
 driFenceMgrUnReference(struct _DriFenceMgr **pMgr)
 {
-    _glthread_LOCK_MUTEX((*pMgr)->mutex);
+    pipe_mutex_lock((*pMgr)->mutex);
     driFenceMgrUnrefUnlock(pMgr);
 }
 
@@ -143,9 +143,9 @@ driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr,
 	 */
 
 	++entry->refCount;
-	_glthread_UNLOCK_MUTEX(mgr->mutex);
-	_glthread_LOCK_MUTEX(entry->mutex);
-	_glthread_LOCK_MUTEX(mgr->mutex);
+	pipe_mutex_unlock(mgr->mutex);
+	pipe_mutex_lock(entry->mutex);
+	pipe_mutex_lock(mgr->mutex);
 
 	prev = list->prev;
 
@@ -157,7 +157,7 @@ driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr,
 		 * Somebody else removed the entry from the list.
 		 */
 
-		_glthread_UNLOCK_MUTEX(entry->mutex);
+		pipe_mutex_unlock(entry->mutex);
 		driFenceUnReferenceLocked(&entry);
 		return;
 	}
@@ -167,7 +167,7 @@ driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr,
 	    DRMLISTDELINIT(list);
 	    mgr->info.unreference(mgr, &entry->private);
 	}
-	_glthread_UNLOCK_MUTEX(entry->mutex);
+	pipe_mutex_unlock(entry->mutex);
 	driFenceUnReferenceLocked(&entry);
 	list = prev;
     }
@@ -181,7 +181,7 @@ driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type,
     struct _DriFenceMgr *mgr = fence->mgr;
     int ret = 0;
 
-    _glthread_LOCK_MUTEX(fence->mutex);
+    pipe_mutex_lock(fence->mutex);
 
     if ((fence->signaled_type & fence_type) == fence_type)
 	goto out0;
@@ -190,16 +190,16 @@ driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type,
     if (ret)
 	goto out0;
 
-    _glthread_LOCK_MUTEX(mgr->mutex);
-    _glthread_UNLOCK_MUTEX(fence->mutex);
+    pipe_mutex_lock(mgr->mutex);
+    pipe_mutex_unlock(fence->mutex);
 
     driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class,
 				  fence_type);
-    _glthread_UNLOCK_MUTEX(mgr->mutex);
+    pipe_mutex_unlock(mgr->mutex);
     return 0;
 
   out0:
-    _glthread_UNLOCK_MUTEX(fence->mutex);
+    pipe_mutex_unlock(fence->mutex);
     return ret;
 }
 
@@ -207,9 +207,9 @@ uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence)
 {
     uint32_t ret;
 
-    _glthread_LOCK_MUTEX(fence->mutex);
+    pipe_mutex_lock(fence->mutex);
     ret = fence->signaled_type;
-    _glthread_UNLOCK_MUTEX(fence->mutex);
+    pipe_mutex_unlock(fence->mutex);
 
     return ret;
 }
@@ -221,7 +221,7 @@ driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type,
     int ret = 0;
     struct _DriFenceMgr *mgr;
 
-    _glthread_LOCK_MUTEX(fence->mutex);
+    pipe_mutex_lock(fence->mutex);
     mgr = fence->mgr;
     *signaled = fence->signaled_type;
     if ((fence->signaled_type & flush_type) == flush_type)
@@ -236,25 +236,25 @@ driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type,
     if ((fence->signaled_type | *signaled) == fence->signaled_type)
 	goto out0;
 
-    _glthread_LOCK_MUTEX(mgr->mutex);
-    _glthread_UNLOCK_MUTEX(fence->mutex);
+    pipe_mutex_lock(mgr->mutex);
+    pipe_mutex_unlock(fence->mutex);
 
     driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class,
 				  *signaled);
 
-    _glthread_UNLOCK_MUTEX(mgr->mutex);
+    pipe_mutex_unlock(mgr->mutex);
     return 0;
   out0:
-    _glthread_UNLOCK_MUTEX(fence->mutex);
+    pipe_mutex_unlock(fence->mutex);
     return ret;
 }
 
 struct _DriFenceObject *
 driFenceReference(struct _DriFenceObject *fence)
 {
-    _glthread_LOCK_MUTEX(fence->mgr->mutex);
+    pipe_mutex_lock(fence->mgr->mutex);
     ++fence->refCount;
-    _glthread_UNLOCK_MUTEX(fence->mgr->mutex);
+    pipe_mutex_unlock(fence->mgr->mutex);
     return fence;
 }
 
@@ -267,7 +267,7 @@ driFenceUnReference(struct _DriFenceObject **pFence)
 	return;
 
     mgr = (*pFence)->mgr;
-    _glthread_LOCK_MUTEX(mgr->mutex);
+    pipe_mutex_lock(mgr->mutex);
     ++mgr->refCount;
     driFenceUnReferenceLocked(pFence);
     driFenceMgrUnrefUnlock(&mgr);
@@ -294,15 +294,15 @@ struct _DriFenceObject
 	return NULL;
     }
 
-    _glthread_INIT_MUTEX(fence->mutex);
-    _glthread_LOCK_MUTEX(fence->mutex);
-    _glthread_LOCK_MUTEX(mgr->mutex);
+    pipe_mutex_init(fence->mutex);
+    pipe_mutex_lock(fence->mutex);
+    pipe_mutex_lock(mgr->mutex);
     fence->refCount = 1;
     DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]);
     fence->mgr = mgr;
     ++mgr->refCount;
     ++mgr->num_fences;
-    _glthread_UNLOCK_MUTEX(mgr->mutex);
+    pipe_mutex_unlock(mgr->mutex);
     fence->fence_class = fence_class;
     fence->fence_type = fence_type;
     fence->signaled_type = 0;
@@ -312,7 +312,7 @@ struct _DriFenceObject
 	memcpy(fence->private, private, private_size);
     }
 
-    _glthread_UNLOCK_MUTEX(fence->mutex);
+    pipe_mutex_unlock(fence->mutex);
     return fence;
 }
 
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c
index a80555c9c71..60924eac9ee 100644
--- a/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c
@@ -33,7 +33,7 @@
 #include <stdlib.h>
 #include <errno.h>
 #include "pipe/p_debug.h"
-#include "glthread.h"
+#include "pipe/p_thread.h"
 #include "ws_dri_bufpool.h"
 #include "ws_dri_bufmgr.h"
 
@@ -60,14 +60,14 @@ pool_destroy(struct _DriBufferPool *pool, void *private)
 
 static int
 pool_waitIdle(struct _DriBufferPool *pool, void *private,
-	      _glthread_Mutex *mutex, int lazy)
+	      pipe_mutex *mutex, int lazy)
 {
     return 0;
 }
 
 static int
 pool_map(struct _DriBufferPool *pool, void *private, unsigned flags,
-         int hint, _glthread_Mutex *mutex, void **virtual)
+         int hint, pipe_mutex *mutex, void **virtual)
 {
     *virtual = (void *)((unsigned long *)private + 2);
     return 0;
diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c
index dfcf6d6b19a..391cea50a7f 100644
--- a/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c
+++ b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c
@@ -37,7 +37,7 @@
 #include "ws_dri_bufpool.h"
 #include "ws_dri_fencemgr.h"
 #include "ws_dri_bufmgr.h"
-#include "glthread.h"
+#include "pipe/p_thread.h"
 
 #define DRI_SLABPOOL_ALLOC_RETRIES 100
 
@@ -53,7 +53,7 @@ struct _DriSlabBuffer {
     uint32_t start;
     uint32_t fenceType;
     int unFenced;
-    _glthread_Cond event;
+    pipe_condvar event;
 };
 
 struct _DriKernelBO {
@@ -84,7 +84,7 @@ struct _DriSlabSizeHeader {
     uint32_t numDelayed;
     struct _DriSlabPool *slabPool;
     uint32_t bufSize;
-    _glthread_Mutex mutex;
+    pipe_mutex mutex;
 };
 
 struct _DriFreeSlabManager {
@@ -94,7 +94,7 @@ struct _DriFreeSlabManager {
     drmMMListHead timeoutList;
     drmMMListHead unCached;
     drmMMListHead cached;
-    _glthread_Mutex mutex;
+    pipe_mutex mutex;
 };
 
 
@@ -196,7 +196,7 @@ driSetKernelBOFree(struct _DriFreeSlabManager *fMan,
 {
     struct timeval time;
 
-    _glthread_LOCK_MUTEX(fMan->mutex);
+    pipe_mutex_lock(fMan->mutex);
     gettimeofday(&time, NULL);
     driTimeAdd(&time, &fMan->slabTimeout);
 
@@ -210,7 +210,7 @@ driSetKernelBOFree(struct _DriFreeSlabManager *fMan,
     DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList);
     driFreeTimeoutKBOsLocked(fMan, &time);
 
-    _glthread_UNLOCK_MUTEX(fMan->mutex);
+    pipe_mutex_unlock(fMan->mutex);
 }
 
 /*
@@ -237,7 +237,7 @@ driAllocKernelBO(struct _DriSlabSizeHeader *header)
 
     size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize;
     size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1);
-    _glthread_LOCK_MUTEX(fMan->mutex);
+    pipe_mutex_lock(fMan->mutex);
 
     kbo = NULL;
 
@@ -269,7 +269,7 @@ driAllocKernelBO(struct _DriSlabSizeHeader *header)
 	DRMLISTDELINIT(&kbo->timeoutHead);
     }
 
-    _glthread_UNLOCK_MUTEX(fMan->mutex);
+    pipe_mutex_unlock(fMan->mutex);
 
     if (kbo) {
         uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags;
@@ -360,7 +360,7 @@ driAllocSlab(struct _DriSlabSizeHeader *header)
 	buf->start = i* header->bufSize;
 	buf->mapCount = 0;
 	buf->isSlabBuffer = 1;
-	_glthread_INIT_COND(buf->event);
+	pipe_condvar_init(buf->event);
 	DRMLISTADDTAIL(&buf->head, &slab->freeBuffers);
 	slab->numFree++;
 	buf++;
@@ -494,23 +494,23 @@ driSlabAllocBuffer(struct _DriSlabSizeHeader *header)
     drmMMListHead *list;
     int count = DRI_SLABPOOL_ALLOC_RETRIES;
 
-    _glthread_LOCK_MUTEX(header->mutex);
+    pipe_mutex_lock(header->mutex);
     while(header->slabs.next == &header->slabs && count > 0) {
         driSlabCheckFreeLocked(header, 0);
 	if (header->slabs.next != &header->slabs)
 	  break;
 
-	_glthread_UNLOCK_MUTEX(header->mutex);
+	pipe_mutex_unlock(header->mutex);
 	if (count != DRI_SLABPOOL_ALLOC_RETRIES)
 	    usleep(1);
-	_glthread_LOCK_MUTEX(header->mutex);
+	pipe_mutex_lock(header->mutex);
 	(void) driAllocSlab(header);
 	count--;
     }
 
     list = header->slabs.next;
     if (list == &header->slabs) {
-	_glthread_UNLOCK_MUTEX(header->mutex);
+	pipe_mutex_unlock(header->mutex);
 	return NULL;
     }
     slab = DRMLISTENTRY(struct _DriSlab, list, head);
@@ -520,7 +520,7 @@ driSlabAllocBuffer(struct _DriSlabSizeHeader *header)
     list = slab->freeBuffers.next;
     DRMLISTDELINIT(list);
 
-    _glthread_UNLOCK_MUTEX(header->mutex);
+    pipe_mutex_unlock(header->mutex);
     buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head);
     return buf;
 }
@@ -618,7 +618,7 @@ pool_destroy(struct _DriBufferPool *driPool, void *private)
     slab = buf->parent;
     header = slab->header;
 
-    _glthread_LOCK_MUTEX(header->mutex);
+    pipe_mutex_lock(header->mutex);
     buf->unFenced = 0;
     buf->mapCount = 0;
 
@@ -631,18 +631,18 @@ pool_destroy(struct _DriBufferPool *driPool, void *private)
 	driSlabFreeBufferLocked(buf);
     }
 
-    _glthread_UNLOCK_MUTEX(header->mutex);
+    pipe_mutex_unlock(header->mutex);
     return 0;
 }
 
 static int
 pool_waitIdle(struct _DriBufferPool *driPool, void *private,
-	      _glthread_Mutex *mutex, int lazy)
+	      pipe_mutex *mutex, int lazy)
 {
    struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
 
    while(buf->unFenced)
-       _glthread_COND_WAIT(buf->event, *mutex);
+       pipe_condvar_wait(buf->event, *mutex);
 
    if (!buf->fence)
      return 0;
@@ -655,7 +655,7 @@ pool_waitIdle(struct _DriBufferPool *driPool, void *private,
 
 static int
 pool_map(struct _DriBufferPool *pool, void *private, unsigned flags,
-         int hint, _glthread_Mutex *mutex, void **virtual)
+         int hint, pipe_mutex *mutex, void **virtual)
 {
    struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
    int busy;
@@ -689,7 +689,7 @@ pool_unmap(struct _DriBufferPool *pool, void *private)
 
    --buf->mapCount;
    if (buf->mapCount == 0 && buf->isSlabBuffer)
-       _glthread_COND_BROADCAST(buf->event);
+      pipe_condvar_broadcast(buf->event);
 
    return 0;
 }
@@ -760,7 +760,7 @@ pool_fence(struct _DriBufferPool *pool, void *private,
    buf->fenceType = bo->fenceFlags;
 
    buf->unFenced = 0;
-   _glthread_COND_BROADCAST(buf->event);
+   pipe_condvar_broadcast(buf->event);
 
    return 0;
 }
@@ -775,7 +775,7 @@ pool_kernel(struct _DriBufferPool *pool, void *private)
 
 static int
 pool_validate(struct _DriBufferPool *pool, void *private,
-	      _glthread_Mutex *mutex)
+	      pipe_mutex *mutex)
 {
    struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private;
 
@@ -783,7 +783,7 @@ pool_validate(struct _DriBufferPool *pool, void *private,
        return 0;
 
    while(buf->mapCount != 0)
-       _glthread_COND_WAIT(buf->event, *mutex);
+      pipe_condvar_wait(buf->event, *mutex);
 
    buf->unFenced = 1;
    return 0;
@@ -799,8 +799,8 @@ driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec)
     if (!tmp)
 	return NULL;
 
-    _glthread_INIT_MUTEX(tmp->mutex);
-    _glthread_LOCK_MUTEX(tmp->mutex);
+    pipe_mutex_init(tmp->mutex);
+    pipe_mutex_lock(tmp->mutex);
     tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000;
     tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000;
     tmp->slabTimeout.tv_usec -=  tmp->slabTimeout.tv_sec*1000000;
@@ -814,7 +814,7 @@ driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec)
     DRMINITLISTHEAD(&tmp->timeoutList);
     DRMINITLISTHEAD(&tmp->unCached);
     DRMINITLISTHEAD(&tmp->cached);
-    _glthread_UNLOCK_MUTEX(tmp->mutex);
+    pipe_mutex_unlock(tmp->mutex);
 
     return tmp;
 }
@@ -827,9 +827,9 @@ driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan)
     time = fMan->nextCheck;
     driTimeAdd(&time, &fMan->checkInterval);
 
-    _glthread_LOCK_MUTEX(fMan->mutex);
+    pipe_mutex_lock(fMan->mutex);
     driFreeTimeoutKBOsLocked(fMan, &time);
-    _glthread_UNLOCK_MUTEX(fMan->mutex);
+    pipe_mutex_unlock(fMan->mutex);
 
     assert(fMan->timeoutList.next == &fMan->timeoutList);
     assert(fMan->unCached.next == &fMan->unCached);
@@ -842,8 +842,8 @@ static void
 driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size,
 		  struct _DriSlabSizeHeader *header)
 {
-    _glthread_INIT_MUTEX(header->mutex);
-    _glthread_LOCK_MUTEX(header->mutex);
+    pipe_mutex_init(header->mutex);
+    pipe_mutex_lock(header->mutex);
 
     DRMINITLISTHEAD(&header->slabs);
     DRMINITLISTHEAD(&header->freeSlabs);
@@ -853,7 +853,7 @@ driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size,
     header->slabPool = pool;
     header->bufSize = size;
 
-    _glthread_UNLOCK_MUTEX(header->mutex);
+    pipe_mutex_unlock(header->mutex);
 }
 
 static void
@@ -862,7 +862,7 @@ driFinishSizeHeader(struct _DriSlabSizeHeader *header)
     drmMMListHead *list, *next;
     struct _DriSlabBuffer *buf;
 
-    _glthread_LOCK_MUTEX(header->mutex);
+    pipe_mutex_lock(header->mutex);
     for (list = header->delayedBuffers.next, next = list->next;
 	 list != &header->delayedBuffers;
 	 list = next, next = list->next) {
@@ -875,7 +875,7 @@ driFinishSizeHeader(struct _DriSlabSizeHeader *header)
 	header->numDelayed--;
 	driSlabFreeBufferLocked(buf);
     }
-    _glthread_UNLOCK_MUTEX(header->mutex);
+    pipe_mutex_unlock(header->mutex);
 }
 
 static void
diff --git a/src/gallium/winsys/drm/intel/dri/intel_lock.c b/src/gallium/winsys/drm/intel/dri/intel_lock.c
index 406284c98fb..ad1c202429e 100644
--- a/src/gallium/winsys/drm/intel/dri/intel_lock.c
+++ b/src/gallium/winsys/drm/intel/dri/intel_lock.c
@@ -27,7 +27,7 @@
 
 
 #include "main/glheader.h"
-#include "glapi/glthread.h"
+#include "pipe/p_thread.h"
 #include <GL/internal/glcore.h>
 #include "state_tracker/st_public.h"
 #include "intel_context.h"
@@ -35,7 +35,7 @@
 
 
 
-_glthread_DECLARE_STATIC_MUTEX( lockMutex );
+pipe_static_mutex( lockMutex );
 
 
 static void
@@ -72,7 +72,7 @@ void LOCK_HARDWARE( struct intel_context *intel )
 {
     char __ret = 0;
 
-    _glthread_LOCK_MUTEX(lockMutex);
+    pipe_mutex_lock(lockMutex);
     assert(!intel->locked);
 
     DRM_CAS(intel->driHwLock, intel->hHWContext,
@@ -96,7 +96,7 @@ void UNLOCK_HARDWARE( struct intel_context *intel )
 
    DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext);
 
-   _glthread_UNLOCK_MUTEX(lockMutex);
+   pipe_mutex_unlock(lockMutex);
 
    DBG(LOCK, "%s - unlocked\n", __progname);
 }
diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.c b/src/gallium/winsys/drm/intel/dri/intel_screen.c
index 46d4861e77c..3a486481f56 100644
--- a/src/gallium/winsys/drm/intel/dri/intel_screen.c
+++ b/src/gallium/winsys/drm/intel/dri/intel_screen.c
@@ -83,7 +83,7 @@ intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys,
                                      buffer);
 
    /* Unref the buffer we don't need it anyways */
-   pipe_buffer_reference(screen->winsys, &buffer, NULL);
+   pipe_buffer_reference(screen, &buffer, NULL);
 
    surface = screen->get_tex_surface(screen,
                                      texture,
diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c
index ae81d7f801b..2fd190da52e 100644
--- a/src/gallium/winsys/egl_xlib/sw_winsys.c
+++ b/src/gallium/winsys/egl_xlib/sw_winsys.c
@@ -216,7 +216,7 @@ surface_release(struct pipe_winsys *winsys, struct pipe_surface **s)
    surf->refcount--;
    if (surf->refcount == 0) {
       if (surf->buffer)
-         pipe_buffer_reference(winsys, &surf->buffer, NULL);
+         winsys_buffer_reference(winsys, &surf->buffer, NULL);
       free(surf);
    }
    *s = NULL;
diff --git a/src/gallium/winsys/gdi/wmesa.c b/src/gallium/winsys/gdi/wmesa.c
index 730fb1b5417..ed3dd2b9277 100644
--- a/src/gallium/winsys/gdi/wmesa.c
+++ b/src/gallium/winsys/gdi/wmesa.c
@@ -463,7 +463,7 @@ wm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s)
    surf->refcount--;
    if (surf->refcount == 0) {
       if (surf->buffer)
-	pipe_buffer_reference(winsys, &surf->buffer, NULL);
+	winsys_buffer_reference(winsys, &surf->buffer, NULL);
       free(surf);
    }
    *s = NULL;
diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile
index ec92c790685..11c76324113 100644
--- a/src/gallium/winsys/xlib/Makefile
+++ b/src/gallium/winsys/xlib/Makefile
@@ -33,10 +33,7 @@ XLIB_WINSYS_SOURCES = \
 XLIB_WINSYS_OBJECTS = $(XLIB_WINSYS_SOURCES:.c=.o)
 
 
-ifeq ($(CONFIG_NAME), linux-cell)
-# The SPU code is in a separate .a file, unfortunately
-CELL_SPU_LIB = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a
-endif
+# Note: CELL_SPU_LIB is only defined for cell configs
 
 LIBS = \
 	$(GALLIUM_DRIVERS) \
diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript
index 8650f595a72..324fbef306a 100644
--- a/src/gallium/winsys/xlib/SConscript
+++ b/src/gallium/winsys/xlib/SConscript
@@ -36,8 +36,10 @@ if env['platform'] == 'linux' \
         drivers += [trace]
 
     # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions
-    env.SharedLibrary(
+    libgl = env.SharedLibrary(
         target ='GL',
         source = sources,
         LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'],
     )
+
+    env.InstallSharedLibrary(libgl, version=(1, 5))
diff --git a/src/gallium/winsys/xlib/glxapi.c b/src/gallium/winsys/xlib/glxapi.c
index c2ccce6f520..c059fc3edb5 100644
--- a/src/gallium/winsys/xlib/glxapi.c
+++ b/src/gallium/winsys/xlib/glxapi.c
@@ -37,6 +37,7 @@
 #include "main/glheader.h"
 #include "glapi/glapi.h"
 #include "glxapi.h"
+#include "pipe/p_thread.h"
 
 
 extern struct _glxapi_table *_real_GetGLXDispatchTable(void);
@@ -127,26 +128,13 @@ get_dispatch(Display *dpy)
 /**
  * GLX API current context.
  */
-#if defined(GLX_USE_TLS)
-PUBLIC __thread void * CurrentContext
-    __attribute__((tls_model("initial-exec")));
-#elif defined(THREADS)
-static _glthread_TSD ContextTSD;         /**< Per-thread context pointer */
-#else
-static GLXContext CurrentContext = 0;
-#endif
+pipe_tsd ContextTSD;
 
 
 static void
 SetCurrentContext(GLXContext c)
 {
-#if defined(GLX_USE_TLS)
-   CurrentContext = c;
-#elif defined(THREADS)
-   _glthread_SetTSD(&ContextTSD, c);
-#else
-   CurrentContext = c;
-#endif
+   pipe_tsd_set(&ContextTSD, c);
 }
 
 
@@ -238,13 +226,7 @@ glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value)
 GLXContext PUBLIC
 glXGetCurrentContext(void)
 {
-#if defined(GLX_USE_TLS)
-   return CurrentContext;
-#elif defined(THREADS)
-   return (GLXContext) _glthread_GetTSD(&ContextTSD);
-#else
-   return CurrentContext;
-#endif
+   return (GLXContext) pipe_tsd_get(&ContextTSD);
 }
 
 
diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c
index 72563404202..28bd6ceab42 100644
--- a/src/gallium/winsys/xlib/xm_api.c
+++ b/src/gallium/winsys/xlib/xm_api.c
@@ -62,7 +62,6 @@
 #include "xmesaP.h"
 #include "main/context.h"
 #include "main/framebuffer.h"
-#include "glapi/glthread.h"
 
 #include "state_tracker/st_public.h"
 #include "state_tracker/st_context.h"
@@ -75,7 +74,7 @@
 /**
  * Global X driver lock
  */
-_glthread_Mutex _xmesa_lock;
+pipe_mutex _xmesa_lock;
 
 
 int xmesa_mode;
@@ -245,10 +244,10 @@ xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b,
 #else
    Status stat;
 
-   _glthread_LOCK_MUTEX(_xmesa_lock);
+   pipe_mutex_lock(_xmesa_lock);
    XSync(b->xm_visual->display, 0); /* added for Chromium */
    stat = get_drawable_size(dpy, b->drawable, width, height);
-   _glthread_UNLOCK_MUTEX(_xmesa_lock);
+   pipe_mutex_unlock(_xmesa_lock);
 
    if (!stat) {
       /* probably querying a window that's recently been destroyed */
@@ -350,12 +349,17 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type,
 
    if (vis->mesa_visual.depthBits == 0)
       depthFormat = PIPE_FORMAT_NONE;
+#ifdef GALLIUM_CELL /* XXX temporary for Cell! */
+   else
+      depthFormat = PIPE_FORMAT_S8Z24_UNORM;
+#else
    else if (vis->mesa_visual.depthBits <= 16)
-      depthFormat = PIPE_FORMAT_Z16_UNORM;
+      depthFormat = PIPE_FORMAT_Z16UNORM;
    else if (vis->mesa_visual.depthBits <= 24)
       depthFormat = PIPE_FORMAT_S8Z24_UNORM;
    else
       depthFormat = PIPE_FORMAT_Z32_UNORM;
+#endif
 
    if (vis->mesa_visual.stencilBits == 8) {
       if (depthFormat == PIPE_FORMAT_S8Z24_UNORM)
@@ -779,7 +783,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
    uint pf;
 
    if (firstTime) {
-      _glthread_INIT_MUTEX(_xmesa_lock);
+      pipe_mutex_init(_xmesa_lock);
       firstTime = GL_FALSE;
    }
 
@@ -1299,6 +1303,7 @@ void XMesaFlush( XMesaContext c )
 #ifdef XFree86Server
       /* NOT_NEEDED */
 #else
+      st_finish(c->st);
       XSync( c->xm_visual->display, False );
 #endif
    }
diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c
index 68ead7f528e..c4a30d37023 100644
--- a/src/gallium/winsys/xlib/xm_winsys.c
+++ b/src/gallium/winsys/xlib/xm_winsys.c
@@ -276,6 +276,39 @@ xm_buffer_destroy(struct pipe_winsys *pws,
 
 
 /**
+ * For Cell.  Basically, rearrange the pixels/quads from this layout:
+ *  +--+--+--+--+
+ *  |p0|p1|p2|p3|....
+ *  +--+--+--+--+
+ *
+ * to this layout:
+ *  +--+--+
+ *  |p0|p1|....
+ *  +--+--+
+ *  |p2|p3|
+ *  +--+--+
+ */
+static void
+twiddle_tile(uint *tile)
+{
+   uint tile2[TILE_SIZE * TILE_SIZE];
+   int y, x;
+
+   for (y = 0; y < TILE_SIZE; y+=2) {
+      for (x = 0; x < TILE_SIZE; x+=2) {
+         int k = 4 * (y/2 * TILE_SIZE/2 + x/2);
+         tile2[y * TILE_SIZE + (x + 0)] = tile[k];
+         tile2[y * TILE_SIZE + (x + 1)] = tile[k+1];
+         tile2[(y + 1) * TILE_SIZE + (x + 0)] = tile[k+2];
+         tile2[(y + 1) * TILE_SIZE + (x + 1)] = tile[k+3];
+      }
+   }
+   memcpy(tile, tile2, sizeof(tile2));
+}
+
+
+
+/**
  * Display a surface that's in a tiled configuration.  That is, all the
  * pixels for a TILE_SIZExTILE_SIZE block are contiguous in memory.
  */
@@ -287,16 +320,16 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf)
    const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE;
    uint x, y;
 
-   /* check that the XImage has been previously initialized */
-   assert(ximage->format);
-   assert(ximage->bitmap_unit);
-
    if (XSHM_ENABLED(xm_buf) && (xm_buf->tempImage == NULL)) {
       alloc_shm_ximage(xm_buf, b, TILE_SIZE, TILE_SIZE);
    }
 
    ximage = (XSHM_ENABLED(xm_buf)) ? xm_buf->tempImage : b->tempImage;
 
+   /* check that the XImage has been previously initialized */
+   assert(ximage->format);
+   assert(ximage->bitmap_unit);
+
    if (!XSHM_ENABLED(xm_buf)) {
       /* update XImage's fields */
       ximage->width = TILE_SIZE;
@@ -306,24 +339,32 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf)
 
    for (y = 0; y < surf->height; y += TILE_SIZE) {
       for (x = 0; x < surf->width; x += TILE_SIZE) {
-         int dx = x;
-         int dy = y;
          int tx = x / TILE_SIZE;
          int ty = y / TILE_SIZE;
          int offset = ty * tilesPerRow + tx;
+         int w = TILE_SIZE;
+         int h = TILE_SIZE;
+
+         if (y + h > surf->height)
+            h = surf->height - y;
+         if (x + w > surf->width)
+            w = surf->width - x;
 
          offset *= 4 * TILE_SIZE * TILE_SIZE;
 
          ximage->data = (char *) xm_buf->data + offset;
 
+         twiddle_tile((uint *) ximage->data);
+
          if (XSHM_ENABLED(xm_buf)) {
 #if defined(USE_XSHM) && !defined(XFree86Server)
             XShmPutImage(b->xm_visual->display, b->drawable, b->gc,
-                         ximage, 0, 0, x, y, TILE_SIZE, TILE_SIZE, False);
+                         ximage, 0, 0, x, y, w, h, False);
 #endif
-         } else {
+         }
+         else {
             XPutImage(b->xm_visual->display, b->drawable, b->gc,
-                      ximage, 0, 0, dx, dy, TILE_SIZE, TILE_SIZE);
+                      ximage, 0, 0, x, y, w, h);
          }
       }
    }
@@ -543,7 +584,7 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s)
    surf->refcount--;
    if (surf->refcount == 0) {
       if (surf->buffer)
-	pipe_buffer_reference(winsys, &surf->buffer, NULL);
+	winsys_buffer_reference(winsys, &surf->buffer, NULL);
       free(surf);
    }
    *s = NULL;
diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c
index 35c4ebc4ba3..b7c10b6bcae 100644
--- a/src/gallium/winsys/xlib/xm_winsys_aub.c
+++ b/src/gallium/winsys/xlib/xm_winsys_aub.c
@@ -308,7 +308,7 @@ aub_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s)
    surf->refcount--;
    if (surf->refcount == 0) {
       if (surf->buffer)
-         pipe_buffer_reference(winsys, &surf->buffer, NULL);
+         winsys_buffer_reference(winsys, &surf->buffer, NULL);
       free(surf);
    }
    *s = NULL;
diff --git a/src/gallium/winsys/xlib/xmesaP.h b/src/gallium/winsys/xlib/xmesaP.h
index 9b15b2ddf99..fcaeee52bcf 100644
--- a/src/gallium/winsys/xlib/xmesaP.h
+++ b/src/gallium/winsys/xlib/xmesaP.h
@@ -35,9 +35,10 @@
 
 #include "state_tracker/st_context.h"
 #include "state_tracker/st_public.h"
+#include "pipe/p_thread.h"
 
 
-extern _glthread_Mutex _xmesa_lock;
+extern pipe_mutex _xmesa_lock;
 
 extern XMesaBuffer XMesaBufferList;