ENGR00213170-1 [MX6SL] Add resource needed by GPU

Prepare resourec such as memory, interrupt, clock, regester address
needed by GPU.

Signed-off-by: Larry Li <b20787@freescale.com>

3 files changed, 196 insertions, 2 deletions

diff --git a/arch/arm/mach-mx6/board-mx6sl_arm2.c b/arch/arm/mach-mx6/board-mx6sl_arm2.c
index 3d1bba15c626..2d42966dbfb1 100755
--- a/arch/arm/mach-mx6/board-mx6sl_arm2.c
+++ b/arch/arm/mach-mx6/board-mx6sl_arm2.c
@@ -563,6 +563,10 @@ static struct mxc_dvfs_platform_data mx6sl_arm2_dvfscore_data = {
 	.delay_time		= 80,
 };
 
+static struct viv_gpu_platform_data imx6q_gpu_pdata __initdata = {
+	.reserved_mem_size = SZ_128M,
+};
+
 void __init early_console_setup(unsigned long base, struct clk *clk);
 
 static inline void mx6_arm2_init_uart(void)
@@ -1112,6 +1116,8 @@ static void __init mx6_arm2_init(void)
 
 	imx6q_add_viim();
 	imx6q_add_imx2_wdt(0, NULL);
+
+	imx_add_viv_gpu(&imx6_gpu_data, &imx6q_gpu_pdata);
 }
 
 extern void __iomem *twd_base;
@@ -1134,7 +1140,17 @@ static struct sys_timer mxc_timer = {
 
 static void __init mx6_arm2_reserve(void)
 {
-
+#if defined(CONFIG_MXC_GPU_VIV) || defined(CONFIG_MXC_GPU_VIV_MODULE)
+	phys_addr_t phys;
+
+	if (imx6q_gpu_pdata.reserved_mem_size) {
+		phys = memblock_alloc_base(imx6q_gpu_pdata.reserved_mem_size,
+					   SZ_4K, MEMBLOCK_ALLOC_ACCESSIBLE);
+		memblock_free(phys, imx6q_gpu_pdata.reserved_mem_size);
+		memblock_remove(phys, imx6q_gpu_pdata.reserved_mem_size);
+		imx6q_gpu_pdata.reserved_mem_base = phys;
+	}
+#endif
 }
 
 MACHINE_START(MX6SL_ARM2, "Freescale i.MX 6SoloLite Armadillo2 Board")
diff --git a/arch/arm/mach-mx6/clock_mx6sl.c b/arch/arm/mach-mx6/clock_mx6sl.c
index 877b37d20a2d..0c17ed4bfd8f 100755
--- a/arch/arm/mach-mx6/clock_mx6sl.c
+++ b/arch/arm/mach-mx6/clock_mx6sl.c
@@ -3331,6 +3331,170 @@ static struct clk aips_tz1_clk = {
 	.disable = _clk_disable_inwait,
 };
 
+/* Raster 2D and OpenVG GPU core share the same clock gate bits,
+ * so we abstract gpu2d_core_clk to handle clock gate enabling
+ * and disabling for both 2D core. These two 2D core have
+ * different multiplexer, so gpu2d_axi_clk and openvg_axi_clk
+ * are provided for them to control the multiplexer individually.
+*/
+static struct clk gpu2d_core_clk = {
+	__INIT_CLK_DEBUG(gpu2d_core_clk)
+	.enable = _clk_enable,
+	.enable_reg = MXC_CCM_CCGR1,
+	.enable_shift = MXC_CCM_CCGRx_CG13_OFFSET,
+	.disable = _clk_disable,
+	.flags = AHB_HIGH_SET_POINT | CPU_FREQ_TRIG_UPDATE,
+};
+
+static unsigned long _clk_openvg_axi_round_rate(struct clk *clk,
+						unsigned long rate)
+{
+	u32 div;
+	u32 parent_rate = clk_get_rate(clk->parent);
+
+	div = parent_rate / rate;
+
+	/* Make sure rate is not greater than the maximum value for the clock.
+	 * Also prevent a div of 0.
+	 */
+	if (div == 0)
+		div++;
+
+	if (div > 8)
+		div = 8;
+
+	return parent_rate / div;
+}
+
+static int _clk_openvg_axi_set_parent(struct clk *clk, struct clk *parent)
+{
+	int mux;
+	u32 reg = __raw_readl(MXC_CCM_CBCMR)
+		& ~MXC_CCM_CBCMR_GPU3D_CORE_CLK_SEL_MASK;
+
+	mux = _get_mux6(parent, &pll3_pfd1_540M,
+		&pll3_usb_otg_main_clk,
+		&pll2_528_bus_main_clk, &pll2_pfd2_400M, NULL, NULL);
+	reg |= (mux << MXC_CCM_CBCMR_GPU3D_CORE_CLK_SEL_OFFSET);
+	__raw_writel(reg, MXC_CCM_CBCMR);
+
+	return 0;
+}
+
+static unsigned long _clk_openvg_axi_get_rate(struct clk *clk)
+{
+	u32 reg, div;
+
+	reg = __raw_readl(MXC_CCM_CBCMR);
+	div = ((reg & MXC_CCM_CBCMR_GPU3D_CORE_PODF_MASK) >>
+			MXC_CCM_CBCMR_GPU3D_CORE_PODF_OFFSET) + 1;
+
+	return clk_get_rate(clk->parent) / div;
+}
+
+static int _clk_openvg_axi_set_rate(struct clk *clk, unsigned long rate)
+{
+	u32 reg, div;
+	u32 parent_rate = clk_get_rate(clk->parent);
+
+	div = parent_rate / rate;
+	if (div == 0)
+		div++;
+	if (div > 8)
+		div = 8;
+
+	reg = __raw_readl(MXC_CCM_CBCMR);
+	reg &= ~MXC_CCM_CBCMR_GPU3D_CORE_PODF_MASK;
+	reg |= (div - 1) << MXC_CCM_CBCMR_GPU3D_CORE_PODF_OFFSET;
+	__raw_writel(reg, MXC_CCM_CBCMR);
+
+	return 0;
+}
+
+static struct clk openvg_axi_clk = {
+	__INIT_CLK_DEBUG(openvg_axi_clk)
+	.parent = &pll2_528_bus_main_clk,
+	.set_parent = _clk_openvg_axi_set_parent,
+	.set_rate = _clk_openvg_axi_set_rate,
+	.get_rate = _clk_openvg_axi_get_rate,
+	.round_rate = _clk_openvg_axi_round_rate,
+	.flags = AHB_HIGH_SET_POINT | CPU_FREQ_TRIG_UPDATE,
+};
+
+static unsigned long _clk_gpu2d_axi_round_rate(struct clk *clk,
+						unsigned long rate)
+{
+	u32 div;
+	u32 parent_rate = clk_get_rate(clk->parent);
+
+	div = parent_rate / rate;
+
+	/* Make sure rate is not greater than the maximum value for the clock.
+	 * Also prevent a div of 0.
+	 */
+	if (div == 0)
+		div++;
+
+	if (div > 8)
+		div = 8;
+
+	return parent_rate / div;
+}
+
+static int _clk_gpu2d_axi_set_parent(struct clk *clk, struct clk *parent)
+{
+	int mux;
+	u32 reg = __raw_readl(MXC_CCM_CBCMR) &
+				~MXC_CCM_CBCMR_GPU3D_SHADER_CLK_SEL_MASK;
+
+	mux = _get_mux6(parent, &pll2_pfd2_400M, &pll3_usb_otg_main_clk,
+		&pll3_pfd1_540M, &pll2_528_bus_main_clk, NULL, NULL);
+	reg |= (mux << MXC_CCM_CBCMR_GPU3D_SHADER_CLK_SEL_OFFSET);
+	__raw_writel(reg, MXC_CCM_CBCMR);
+
+	return 0;
+}
+
+static unsigned long _clk_gpu2d_axi_get_rate(struct clk *clk)
+{
+	u32 reg, div = 1;
+
+	reg = __raw_readl(MXC_CCM_CBCMR);
+	div = ((reg & MXC_CCM_CBCMR_GPU3D_SHADER_PODF_MASK) >>
+			MXC_CCM_CBCMR_GPU3D_SHADER_PODF_OFFSET) + 1;
+
+	return clk_get_rate(clk->parent) / div;
+}
+
+static int _clk_gpu2d_axi_set_rate(struct clk *clk, unsigned long rate)
+{
+	u32 reg, div;
+	u32 parent_rate = clk_get_rate(clk->parent);
+
+	div = parent_rate / rate;
+	if (div == 0)
+		div++;
+	if (((parent_rate / div) != rate) || (div > 8))
+		return -EINVAL;
+
+	reg = __raw_readl(MXC_CCM_CBCMR);
+	reg &= ~MXC_CCM_CBCMR_GPU3D_SHADER_PODF_MASK;
+	reg |= (div - 1) << MXC_CCM_CBCMR_GPU3D_SHADER_PODF_OFFSET;
+	__raw_writel(reg, MXC_CCM_CBCMR);
+
+	return 0;
+}
+
+static struct clk gpu2d_axi_clk = {
+	__INIT_CLK_DEBUG(gpu2d_axi_clk)
+	.parent = &pll2_528_bus_main_clk,
+	.set_parent = _clk_gpu2d_axi_set_parent,
+	.set_rate = _clk_gpu2d_axi_set_rate,
+	.get_rate = _clk_gpu2d_axi_get_rate,
+	.round_rate = _clk_gpu2d_axi_round_rate,
+	.flags = AHB_HIGH_SET_POINT | CPU_FREQ_TRIG_UPDATE,
+};
+
 /* set the parent by the ipcg table */
 
 static struct clk pwm_clk[] = {
@@ -3711,6 +3875,9 @@ static struct clk_lookup lookups[] = {
 	_REGISTER_CLOCK("mxs-perfmon.2", "perfmon", perfmon2_clk),
 	_REGISTER_CLOCK(NULL, "anaclk_1", anaclk_1),
 	_REGISTER_CLOCK(NULL, "anaclk_2", anaclk_2),
+	_REGISTER_CLOCK(NULL, "gpu2d_clk", gpu2d_core_clk),
+	_REGISTER_CLOCK(NULL, "gpu2d_axi_clk", gpu2d_axi_clk),
+	_REGISTER_CLOCK(NULL, "openvg_axi_clk", openvg_axi_clk),
 };
 
 static void clk_tree_init(void)
diff --git a/arch/arm/plat-mxc/devices/platform-viv_gpu.c b/arch/arm/plat-mxc/devices/platform-viv_gpu.c
index dad9d5a2df3b..40cd3fca56d7 100644
--- a/arch/arm/plat-mxc/devices/platform-viv_gpu.c
+++ b/arch/arm/plat-mxc/devices/platform-viv_gpu.c
@@ -22,6 +22,16 @@
 #include <mach/devices-common.h>
 
 #ifdef CONFIG_ARCH_MX6
+#ifdef CONFIG_SOC_IMX6SL
+const struct imx_viv_gpu_data imx6_gpu_data __initconst = {
+	.iobase_3d = 0,
+	.irq_3d = -1,
+	.iobase_2d = MX6SL_GPU_2D_ARB_BASE_ADDR,
+	.irq_2d = MXC_INT_GPU2D_IRQ,
+	.iobase_vg = OPENVG_ARB_BASE_ADDR,
+	.irq_vg = MXC_INT_OPENVG_XAQ2,
+};
+#else
 const struct imx_viv_gpu_data imx6_gpu_data __initconst = {
 	.iobase_3d = GPU_3D_ARB_BASE_ADDR,
 	.irq_3d = MXC_INT_GPU3D_IRQ,
@@ -31,6 +41,7 @@ const struct imx_viv_gpu_data imx6_gpu_data __initconst = {
 	.irq_vg = MXC_INT_OPENVG_XAQ2,
 };
 #endif
+#endif
 
 struct platform_device *__init imx_add_viv_gpu(
 		const struct imx_viv_gpu_data *data,
@@ -71,7 +82,7 @@ struct platform_device *__init imx_add_viv_gpu(
 		},
 	};
 
-	if (cpu_is_mx6q())
+	if (cpu_is_mx6q() || cpu_is_mx6sl())
 		res_count = ARRAY_SIZE(res);
 	else if (cpu_is_mx6dl())
 		/* No openVG on i.mx6 Solo/DL */