PVRSRVCreateDCSwapChain Error On Android

Hi all,

I’m working with GLES on Android 4.0.4 (AOSP), using a Galaxy Nexus maguro phone. However when I run the command line OpenGL tests, built from sources in frameworks/base/opengl/tests, they fail with the following output from logcat:

D/libEGL (12589): loaded /system/lib/egl/libGLES_android.so

D/libEGL (12589): loaded /vendor/lib/egl/libEGL_POWERVR_SGX540_120.so

D/libEGL (12589): loaded /vendor/lib/egl/libGLESv1_CM_POWERVR_SGX540_120.so

D/libEGL (12589): loaded /vendor/lib/egl/libGLESv2_POWERVR_SGX540_120.so

E/IMGSRV (12589): :0: PVRSRVCreateDCSwapChain: Error - 10 returned

E/IMGSRV (12589): :0: framebuffer_device_open: Failed to create flip chain

E/FramebufferNativeWindow(12589): couldn’t open framebuffer HAL (Bad address)

The errors come from a call to android_createDisplaySurface. Here are some relevant code snippets:


342EGLNativeWindowType android_createDisplaySurface(void)


344 FramebufferNativeWindow* w;

345 w = new FramebufferNativeWindow();

346 if (w->getDevice() == NULL) {

347 // get a ref so it can be destroyed when we exit this block

348 sp ref(w);

349 return NULL;

350 }

351 return (EGLNativeWindowType)w;



83 if (hw_get_module(GRALLOC_HARDWARE_MODULE_ID, &module) == 0) {

85 int err;

87 err = framebuffer_open(module, &fbDev);

88 LOGE_IF(err, “couldn’t open framebuffer HAL (%s)”, strerror(-err));


151static inline int framebuffer_open(const struct hw_module_t* module,

152 struct framebuffer_device_t** device) {

153 return module->methods->open(module,

154 GRALLOC_HARDWARE_FB0, (struct hw_device_t**)device);


Any ideas about what’s happening here? Any other information I could provide to help diagnose the problem?

Ultimately, my aim is to do some offscreen rendering using GLES from native code.

Many thanks,


Hi Joel,

Are you building with the vanilla Android 4.0.4 (AOSP) code, or have you modified it at all? Would a factory image be suitable for the work you are doing?



Thanks for your response Joe.

I have got some modifications from TI that enable the writeback pipeline, but I don’t think I can share them. I haven’t tried the same tests on vanilla 4.0.4 yet. It is certainly possible that those modifications have had an effect.

At the moment I seem to have worked around the issue by using SurfaceTextureClient, like in



Hi Joel,

Glad to hear you’ve found a way to work around the problem :slight_smile:

Our Android expert is out of the office this week. I’ll ask him to look at this once he’s back.