xtensa: add support for the xtensa processor architecture [1/2]

The Xtensa processor architecture is a configurable, extensible,
and synthesizable 32-bit RISC processor core provided by Cadence.

This is the first part of the basic architecture port with changes to
common files. The 'arch/xtensa' directory, and boards and additional
drivers will be in separate commits.

Signed-off-by: Chris Zankel <chris@zankel.net>
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Tom Rini <trini@konsulko.com>

MAINTAINERS

@@ -454,6 +454,11 @@ S:	Maintained
 T:	git git://git.denx.de/u-boot-x86.git
 F:	arch/x86/
 
+XTENSA
+M:	Max Filippov <jcmvbkbc@gmail.com>
+S:	Maintained
+F:	arch/xtensa/
+
 THE REST
 M:	Tom Rini <trini@konsulko.com>
 L:	u-boot@lists.denx.de

Makefile

@@ -557,6 +557,14 @@ else
 include/config/auto.conf: ;
 endif # $(dot-config)
 
+#
+# Xtensa linker script cannot be preprocessed with -ansi because of
+# preprocessor operations on strings that don't make C identifiers.
+#
+ifeq ($(CONFIG_XTENSA),)
+LDPPFLAGS	+= -ansi
+endif
+
 ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE
 KBUILD_CFLAGS	+= -Os
 else
@@ -1312,7 +1320,7 @@ $(timestamp_h): $(srctree)/Makefile FORCE
 
 # ---------------------------------------------------------------------------
 quiet_cmd_cpp_lds = LDS     $@
-cmd_cpp_lds = $(CPP) -Wp,-MD,$(depfile) $(cpp_flags) $(LDPPFLAGS) -ansi \
+cmd_cpp_lds = $(CPP) -Wp,-MD,$(depfile) $(cpp_flags) $(LDPPFLAGS) \
 		-D__ASSEMBLY__ -x assembler-with-cpp -P -o $@ $<
 
 u-boot.lds: $(LDSCRIPT) prepare FORCE

cmd/bdinfo.c

@@ -524,6 +524,14 @@ int do_bdinfo(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
 	return 0;
 }
 
+#elif defined(CONFIG_XTENSA)
+
+int do_bdinfo(cmd_tbl_t *cmdtp, int flag, int argc, char * const argv[])
+{
+	print_std_bdinfo(gd->bd);
+	return 0;
+}
+
 #else
  #error "a case for this architecture does not exist!"
 #endif

common/board_f.c

@@ -54,6 +54,8 @@
 #include <asm/sections.h>
 #if defined(CONFIG_X86) || defined(CONFIG_ARC)
 #include <asm/init_helpers.h>
+#endif
+#if defined(CONFIG_X86) || defined(CONFIG_ARC) || defined(CONFIG_XTENSA)
 #include <asm/relocate.h>
 #endif
 #ifdef CONFIG_SANDBOX
@@ -271,7 +273,8 @@ static int setup_mon_len(void)
 	gd->mon_len = (ulong)&__bss_end - (ulong)_start;
 #elif defined(CONFIG_SANDBOX) || defined(CONFIG_EFI_APP)
 	gd->mon_len = (ulong)&_end - (ulong)_init;
-#elif defined(CONFIG_BLACKFIN) || defined(CONFIG_NIOS2)
+#elif defined(CONFIG_BLACKFIN) || defined(CONFIG_NIOS2) || \
+	defined(CONFIG_XTENSA)
 	gd->mon_len = CONFIG_SYS_MONITOR_LEN;
 #elif defined(CONFIG_NDS32)
 	gd->mon_len = (ulong)(&__bss_end) - (ulong)(&_start);
@@ -971,7 +974,7 @@ static init_fnc_t init_sequence_f[] = {
 	 *  - board info struct
 	 */
 	setup_dest_addr,
-#if defined(CONFIG_BLACKFIN)
+#if defined(CONFIG_BLACKFIN) || defined(CONFIG_XTENSA)
 	/* Blackfin u-boot monitor should be on top of the ram */
 	reserve_uboot,
 #endif
@@ -1003,7 +1006,7 @@ static init_fnc_t init_sequence_f[] = {
 # endif
 #endif /* CONFIG_DM_VIDEO */
 	reserve_trace,
-#if !defined(CONFIG_BLACKFIN)
+#if !defined(CONFIG_BLACKFIN) && !defined(CONFIG_XTENSA)
 	reserve_uboot,
 #endif
 #ifndef CONFIG_SPL_BUILD
@@ -1036,6 +1039,9 @@ static init_fnc_t init_sequence_f[] = {
 	clear_bss,
 	do_elf_reloc_fixups,
 #endif
+#if defined(CONFIG_XTENSA)
+	clear_bss,
+#endif
 #if !defined(CONFIG_ARM) && !defined(CONFIG_SANDBOX)
 	jump_to_copy,
 #endif

common/image.c

@@ -93,6 +93,7 @@ static const table_entry_t uimage_arch[] = {
 	{	IH_ARCH_ARM64,		"arm64",	"AArch64",	},
 	{	IH_ARCH_ARC,		"arc",		"ARC",		},
 	{	IH_ARCH_X86_64,		"x86_64",	"AMD x86_64",	},
+	{	IH_ARCH_XTENSA,		"xtensa",	"Xtensa",	},
 	{	-1,			"",		"",		},
 };
 

doc/README.xtensa

@@ -0,0 +1,97 @@
+U-Boot for the Xtensa Architecture
+==================================
+
+Xtensa Architecture and Diamond Cores
+-------------------------------------
+
+Xtensa is a configurable processor architecture from Tensilica, Inc.
+Diamond Cores are pre-configured instances available for license and
+SoC cores in the same manner as ARM, MIPS, etc.
+
+Xtensa licensees create their own Xtensa cores with selected features
+and custom instructions, registers and co-processors. The custom core
+is configured with Tensilica tools and built with Tensilica's Xtensa
+Processor Generator.
+
+There are an effectively infinite number of CPUs in the Xtensa
+architecture family. It is, however, not feasible to support individual
+Xtensa CPUs in U-Boot. Therefore, there is only a single 'xtensa' CPU
+in the cpu tree of U-Boot.
+
+In the same manner as the Linux port to Xtensa, U-Boot adapts to an
+individual Xtensa core configuration using a set of macros provided with
+the particular core. This is part of what is known as the hardware
+abstraction layer (HAL). For the purpose of U-Boot, the HAL consists only
+of a few header files. These provide CPP macros that customize sources,
+Makefiles, and the linker script.
+
+
+Adding support for an additional processor configuration
+--------------------------------------------------------
+
+The header files for one particular processor configuration are inside
+a variant-specific directory located in the arch/xtensa/include/asm
+directory. The name of that directory starts with 'arch-' followed by
+the name for the processor configuration, for example, arch-dc233c for
+the Diamond DC233 processor.
+
+    core.h	Definitions for the core itself.
+
+The following files are part of the overlay but not used by U-Boot.
+
+    tie.h	Co-processors and custom extensions defined
+		in the Tensilica Instruction Extension (TIE)
+		language.
+    tie-asm.h	Assembly macros to access custom-defined registers
+		and states.
+
+
+Global Data Pointer, Exported Function Stubs, and the ABI
+---------------------------------------------------------
+
+To support standalone applications launched with the "go" command,
+U-Boot provides a jump table of entrypoints to exported functions
+(grep for EXPORT_FUNC). The implementation for Xtensa depends on
+which ABI (or function calling convention) is used.
+
+Windowed ABI presents unique difficulties with the approach based on
+keeping global data pointer in dedicated register. Because the register
+window rotates during a call, there is no register that is constantly
+available for the gd pointer. Therefore, on xtensa gd is a simple
+global variable. Another difficulty arises from the requirement to have
+an 'entry' at the beginning of a function, which rotates the register
+file and reserves a stack frame. This is an integral part of the
+windowed ABI implemented in hardware. It makes using a jump table to an
+arbitrary (separately compiled) function a bit tricky. Use of a simple
+wrapper is also very tedious due to the need to move all possible
+register arguments and adjust the stack to handle arguments that cannot
+be passed in registers. The most efficient approach is to have the jump
+table perform the 'entry' so as to pretend it's the start of the real
+function. This requires decoding the target function's 'entry'
+instruction to determine the stack frame size, and adjusting the stack
+pointer accordingly, then jumping into the target function just after
+the 'entry'. Decoding depends on the processor's endianness so uses the
+HAL. The implementation (12 instructions) is in examples/stubs.c.
+
+
+Access to Invalid Memory Addresses
+----------------------------------
+
+U-Boot does not check if memory addresses given as arguments to commands
+such as "md" are valid. There are two possible types of invalid
+addresses: an area of physical address space may not be mapped to RAM
+or peripherals, or in the presence of MMU an area of virtual address
+space may not be mapped to physical addresses.
+
+Accessing first type of invalid addresses may result in hardware lockup,
+reading of meaningless data, written data being ignored or an exception,
+depending on the CPU wiring to the system. Accessing second type of
+invalid addresses always ends with an exception.
+
+U-Boot for Xtensa provides a special memory exception handler that
+reports such access attempts and resets the board.
+
+
+------------------------------------------------------------------------------
+Chris Zankel
+Ross Morley

examples/standalone/stubs.c

@@ -240,6 +240,53 @@ gd_t *global_data;
 "	ld	r10, [r10, %1]\n" \
 "	j	[r10]\n" \
 	: : "i"(offsetof(gd_t, jt)), "i"(FO(x)) : "r10");
+#elif defined(CONFIG_XTENSA)
+/*
+ * Global data ptr is in global_data, jump table ptr is in jt.
+ * Windowed ABI: Jump just past 'entry' in target and adjust stack frame
+ * (extract stack frame size from target 'entry' instruction).
+ */
+
+static void **jt;
+
+#if defined(__XTENSA_CALL0_ABI__)
+#define EXPORT_FUNC(f, a, x, ...)	\
+	asm volatile (			\
+"	.extern jt\n"			\
+"	.globl " #x "\n"		\
+"	.align 4\n"			\
+#x ":\n"				\
+"	l32i	a8, %0, 0\n"		\
+"	l32i	a8, a8, %1\n"		\
+"	jx	a8\n"			\
+	: : "r"(jt), "i" (FO(x)) : "a8");
+#elif defined(__XTENSA_WINDOWED_ABI__)
+#if XCHAL_HAVE_BE
+# define SFT "8"
+#else
+# define SFT "12"
+#endif
+#define EXPORT_FUNC(f, a, x, ...)	\
+	asm volatile (			\
+"	.extern jt\n"			\
+"	.globl " #x "\n"		\
+"	.align 4\n"			\
+#x ":\n"				\
+"	entry	sp, 16\n"		\
+"	l32i	a8, %0, 0\n"		\
+"	l32i	a8, a8, %1\n"		\
+"	l32i	a9, a8, 0\n"		\
+"	extui	a9, a9, " SFT ", 12\n"	\
+"	subx8	a9, a9, sp\n"		\
+"	movi	a10, 16\n"		\
+"	sub	a9, a10, a9\n"		\
+"	movsp	sp, a9\n"		\
+"	addi	a8, a8, 3\n"		\
+"	jx	a8\n"			\
+	: : "r"(jt), "i" (FO(x)) : "a8", "a9", "a10");
+#else
+#error Unsupported Xtensa ABI
+#endif
 #else
 /*"	addi	$sp, $sp, -24\n"	\
 "	br	$r16\n"			\*/

include/image.h

@@ -200,6 +200,7 @@ enum {
 	IH_ARCH_ARM64,			/* ARM64	*/
 	IH_ARCH_ARC,			/* Synopsys DesignWare ARC */
 	IH_ARCH_X86_64,			/* AMD x86_64, Intel and Via */
+	IH_ARCH_XTENSA,			/* Xtensa	*/
 
 	IH_ARCH_COUNT,
 };

include/linux/stat.h

@@ -126,7 +126,7 @@ struct stat {
 
 #endif	/* __MIPS__ */
 
-#if defined(__AVR32__) || defined(__SH__)
+#if defined(__AVR32__) || defined(__SH__) || defined(__XTENSA__)
 
 struct stat {
 	unsigned long st_dev;
@@ -149,7 +149,7 @@ struct stat {
 	unsigned long  __unused5;
 };
 
-#endif /* __AVR32__ || __SH__ */
+#endif /* __AVR32__ || __SH__  || __XTENSA__ */
 
 #ifdef __cplusplus
 }