Add extension clCreateBufferFromFdINTEL to create cl buffer by external buffer object's fd.

Before this patch, Beignet can only create cl buffer from external bo by
its handle using clCreateBufferFromLibvaIntel. Render node is the first
choice of accessing gpu in currect Beignet implementation. DRM_IOCTL_GEM_OPEN
is used by clCreateBufferFromLibvaIntel but forbidden in Render node mode.
So it's necessary to add this extension to support buffer sharing between
different libraries.

v2:
Seperate clCreateMemObjectFromFdIntel into two extensions:
clCreateBufferFromFdINTEL and clCreateImageFromFdINTEL.

v3:
Fix rebase conflict: add a parameter when invoke cl_mem_allocate.

Signed-off-by: Chuanbo Weng <chuanbo.weng@intel.com>
Reviewed-by: "Yang, Rong R" <rong.r.yang@intel.com>

7 files changed, 109 insertions, 3 deletions

diff --git a/include/CL/cl_intel.h b/include/CL/cl_intel.h
index 28bcb62b..01da5533 100644
--- a/include/CL/cl_intel.h
+++ b/include/CL/cl_intel.h
@@ -133,6 +133,22 @@ typedef CL_API_ENTRY cl_int (CL_API_CALL *clGetMemObjectFdIntel_fn)(
                              cl_mem       /* Memory Obejct */,
                              int*         /* returned fd */);
 
+typedef struct _cl_import_buffer_info_intel {
+    int                     fd;
+    int                     size;
+} cl_import_buffer_info_intel;
+
+/* Create memory object from external buffer object by fd */
+extern CL_API_ENTRY cl_mem CL_API_CALL
+clCreateBufferFromFdINTEL(cl_context                            /* context */,
+                          const cl_import_buffer_info_intel *   /* info */,
+                          cl_int *                              /* errcode_ret */);
+
+typedef CL_API_ENTRY cl_mem (CL_API_CALL *clCreateBufferFromFdINTEL_fn)(
+                             cl_context                            /* context */,
+                             const cl_import_buffer_info_intel *   /* info */,
+                             cl_int *                              /* errcode_ret */);
+
 #ifdef __cplusplus
 }
 #endif
diff --git a/src/cl_api.c b/src/cl_api.c
index a18bc998..7a75ca3a 100644
--- a/src/cl_api.c
+++ b/src/cl_api.c
@@ -3190,6 +3190,7 @@ internal_clGetExtensionFunctionAddress(const char *func_name)
   EXTFUNC(clCreateBufferFromLibvaIntel)
   EXTFUNC(clCreateImageFromLibvaIntel)
   EXTFUNC(clGetMemObjectFdIntel)
+  EXTFUNC(clCreateBufferFromFdINTEL)
   return NULL;
 }
 
@@ -3358,3 +3359,25 @@ clGetMemObjectFdIntel(cl_context context,
 error:
   return err;
 }
+
+cl_mem
+clCreateBufferFromFdINTEL(cl_context context,
+                          const cl_import_buffer_info_intel* info,
+                          cl_int *errorcode_ret)
+{
+  cl_mem mem = NULL;
+  cl_int err = CL_SUCCESS;
+  CHECK_CONTEXT (context);
+
+  if (!info) {
+    err = CL_INVALID_VALUE;
+    goto error;
+  }
+
+  mem = cl_mem_new_buffer_from_fd(context, info->fd, info->size, &err);
+
+error:
+  if (errorcode_ret)
+    *errorcode_ret = err;
+  return mem;
+}
diff --git a/src/cl_driver.h b/src/cl_driver.h
index 4ffca09a..13cd6ba2 100644
--- a/src/cl_driver.h
+++ b/src/cl_driver.h
@@ -381,6 +381,9 @@ extern cl_buffer_get_fd_cb *cl_buffer_get_fd;
 typedef int (cl_buffer_get_tiling_align_cb)(cl_context ctx, uint32_t tiling_mode, uint32_t dim);
 extern cl_buffer_get_tiling_align_cb *cl_buffer_get_tiling_align;
 
+typedef cl_buffer (cl_buffer_get_buffer_from_fd_cb)(cl_context ctx, int fd, int size);
+extern cl_buffer_get_buffer_from_fd_cb *cl_buffer_get_buffer_from_fd;
+
 /* Get the device id */
 typedef int (cl_driver_get_device_id_cb)(void);
 extern cl_driver_get_device_id_cb *cl_driver_get_device_id;
diff --git a/src/cl_driver_defs.c b/src/cl_driver_defs.c
index b77acdc4..b3e8403d 100644
--- a/src/cl_driver_defs.c
+++ b/src/cl_driver_defs.c
@@ -53,6 +53,7 @@ LOCAL cl_buffer_get_buffer_from_libva_cb *cl_buffer_get_buffer_from_libva = NULL
 LOCAL cl_buffer_get_image_from_libva_cb *cl_buffer_get_image_from_libva = NULL;
 LOCAL cl_buffer_get_fd_cb *cl_buffer_get_fd = NULL;
 LOCAL cl_buffer_get_tiling_align_cb *cl_buffer_get_tiling_align = NULL;
+LOCAL cl_buffer_get_buffer_from_fd_cb *cl_buffer_get_buffer_from_fd = NULL;
 
 /* cl_khr_gl_sharing */
 LOCAL cl_gl_acquire_texture_cb *cl_gl_acquire_texture = NULL;
diff --git a/src/cl_mem.c b/src/cl_mem.c
index be64abe4..6bbd27ab 100644
--- a/src/cl_mem.c
+++ b/src/cl_mem.c
@@ -2206,3 +2206,33 @@ cl_mem_get_fd(cl_mem mem,
 	err = CL_INVALID_OPERATION;
   return err;
 }
+
+LOCAL cl_mem cl_mem_new_buffer_from_fd(cl_context ctx,
+                                       int fd,
+                                       int buffer_sz,
+                                       cl_int* errcode)
+{
+  cl_int err = CL_SUCCESS;
+  cl_mem mem = NULL;
+
+  mem = cl_mem_allocate(CL_MEM_BUFFER_TYPE, ctx, 0, 0, CL_FALSE, NULL, NULL, &err);
+  if (mem == NULL || err != CL_SUCCESS)
+    goto error;
+
+  mem->bo = cl_buffer_get_buffer_from_fd(ctx, fd, buffer_sz);
+  if (mem->bo == NULL) {
+    err = CL_MEM_OBJECT_ALLOCATION_FAILURE;
+    goto error;
+  }
+  mem->size = buffer_sz;
+
+exit:
+  if (errcode)
+    *errcode = err;
+  return mem;
+
+error:
+  cl_mem_delete(mem);
+  mem = NULL;
+  goto exit;
+}
diff --git a/src/cl_mem.h b/src/cl_mem.h
index 4970a755..91c109b7 100644
--- a/src/cl_mem.h
+++ b/src/cl_mem.h
@@ -295,8 +295,13 @@ extern cl_mem cl_mem_new_libva_image(cl_context ctx,
                                      cl_image_format fmt,
                                      size_t row_pitch,
                                      cl_int *errcode);
+
 extern cl_int cl_mem_get_fd(cl_mem mem, int* fd);
 
+extern cl_mem cl_mem_new_buffer_from_fd(cl_context ctx,
+                                        int fd,
+                                        int buffer_sz,
+                                        cl_int* errcode);
 
 #endif /* __CL_MEM_H__ */
 
diff --git a/src/intel/intel_driver.c b/src/intel/intel_driver.c
index 782a2deb..a9c147a9 100644
--- a/src/intel/intel_driver.c
+++ b/src/intel/intel_driver.c
@@ -371,7 +371,7 @@ intel_driver_unlock_hardware(intel_driver_t *driver)
 }
 
 LOCAL dri_bo*
-intel_driver_share_buffer(intel_driver_t *driver, const char *sname, uint32_t name)
+intel_driver_share_buffer_from_name(intel_driver_t *driver, const char *sname, uint32_t name)
 {
   dri_bo *bo = intel_bo_gem_create_from_name(driver->bufmgr,
                                              sname,
@@ -383,6 +383,19 @@ intel_driver_share_buffer(intel_driver_t *driver, const char *sname, uint32_t na
   return bo;
 }
 
+LOCAL dri_bo*
+intel_driver_share_buffer_from_fd(intel_driver_t *driver, int fd, int size)
+{
+  dri_bo *bo = drm_intel_bo_gem_create_from_prime(driver->bufmgr,
+                                                  fd,
+                                                  size);
+  if (bo == NULL) {
+    fprintf(stderr, "drm_intel_bo_gem_create_from_prime create bo(size %d) from fd %d failed: %s\n", fd, size, strerror(errno));
+    return NULL;
+  }
+  return bo;
+}
+
 LOCAL uint32_t
 intel_driver_shared_name(intel_driver_t *driver, dri_bo *bo)
 {
@@ -697,7 +710,7 @@ cl_buffer intel_share_buffer_from_libva(cl_context ctx,
 {
   drm_intel_bo *intel_bo;
 
-  intel_bo = intel_driver_share_buffer((intel_driver_t *)ctx->drv, "shared from libva", bo_name);
+  intel_bo = intel_driver_share_buffer_from_name((intel_driver_t *)ctx->drv, "shared from libva", bo_name);
 
   if (intel_bo == NULL)
     return NULL;
@@ -715,7 +728,7 @@ cl_buffer intel_share_image_from_libva(cl_context ctx,
   drm_intel_bo *intel_bo;
   uint32_t intel_tiling, intel_swizzle_mode;
 
-  intel_bo = intel_driver_share_buffer((intel_driver_t *)ctx->drv, "shared from libva", bo_name);
+  intel_bo = intel_driver_share_buffer_from_name((intel_driver_t *)ctx->drv, "shared from libva", bo_name);
 
   drm_intel_bo_get_tiling(intel_bo, &intel_tiling, &intel_swizzle_mode);
   image->tiling = get_cl_tiling(intel_tiling);
@@ -723,6 +736,20 @@ cl_buffer intel_share_image_from_libva(cl_context ctx,
   return (cl_buffer)intel_bo;
 }
 
+cl_buffer intel_share_buffer_from_fd(cl_context ctx,
+                                     int fd,
+                                     int buffer_size)
+{
+  drm_intel_bo *intel_bo;
+
+  intel_bo = intel_driver_share_buffer_from_fd((intel_driver_t *)ctx->drv, fd, buffer_size);
+
+  if (intel_bo == NULL)
+    return NULL;
+
+  return (cl_buffer)intel_bo;
+}
+
 static cl_buffer intel_buffer_alloc_userptr(cl_buffer_mgr bufmgr, const char* name, void *data,size_t size, unsigned long flags)
 {
 #ifdef HAS_USERPTR
@@ -877,5 +904,6 @@ intel_setup_callbacks(void)
   cl_buffer_wait_rendering = (cl_buffer_wait_rendering_cb *) drm_intel_bo_wait_rendering;
   cl_buffer_get_fd = (cl_buffer_get_fd_cb *) drm_intel_bo_gem_export_to_prime;
   cl_buffer_get_tiling_align = (cl_buffer_get_tiling_align_cb *)intel_buffer_get_tiling_align;
+  cl_buffer_get_buffer_from_fd = (cl_buffer_get_buffer_from_fd_cb *) intel_share_buffer_from_fd;
   intel_set_gpgpu_callbacks(intel_get_device_id());
 }