plat-vexpress: linking supports pagable tee.bin

* Adds scripts to create a pagable tee.bin
* Updates link script and link.mk to support pagable tee.bin
* Binary format of tee.bin is changed from a raw format to a header
  followed by the binary data as described by the header. This requires
  updates in ARM-TF OP-TEE Dispatcher and QEMU virt bios.

fvp:
* Changes OP-TEE load address to start of TZDRAM_BASE

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Pascal Brand <pascal.brand@linaro.org>

3 files changed, 231 insertions, 51 deletions

diff --git a/core/arch/arm32/plat-vexpress/kern.ld.S b/core/arch/arm32/plat-vexpress/kern.ld.S
index ee1449e..e1714ae 100644
--- a/core/arch/arm32/plat-vexpress/kern.ld.S
+++ b/core/arch/arm32/plat-vexpress/kern.ld.S
@@ -64,7 +64,15 @@ SECTIONS
 		KEEP(*(.text.boot.vectab1))
 		KEEP(*(.text.boot.vectab2))
 		KEEP(*(.text.boot))
-		*(.text* .sram.text.glue_7* .gnu.linkonce.t.*)
+
+#ifdef CFG_WITH_PAGER
+		*(.text)
+/* Include list of sections needed for paging */
+#include <text_unpaged.ld.S>
+#else
+		*(.text .text.*)
+#endif
+		*(.sram.text.glue_7* .gnu.linkonce.t.*)
 		__text_end = .;
 	}
 
@@ -96,19 +104,26 @@ SECTIONS
 	.plt : { *(.plt) }
 
 	/* .ARM.exidx is sorted, so has to go in its own output section.  */
-	__exidx_start = .;
-	.ARM.exidx : { *(.ARM.exidx* .gnu.linkonce.armexidx.*) }
-	__exidx_end = .;
+	.ARM.exidx : {
+		__exidx_start = .;
+		*(.ARM.exidx* .gnu.linkonce.armexidx.*)
+		__exidx_end = .;
+	}
 
 	.rodata : ALIGN(4) {
 		__rodata_start = .;
-		*(.rodata .rodata.* .gnu.linkonce.r.*)
-		. = ALIGN(4);
+		*(.gnu.linkonce.r.*)
+#ifdef CFG_WITH_PAGER
+		*(.rodata)
+#include <rodata_unpaged.ld.S>
+#else
+		*(.rodata .rodata.*)
 
+		. = ALIGN(4);
 		__start_ta_head_section = . ;
-		*(ta_head_section)
+		KEEP(*(ta_head_section))
 		__stop_ta_head_section = . ;
-
+#endif
 		. = ALIGN(4);
 		__rodata_end = .;
 	}
@@ -120,6 +135,7 @@ SECTIONS
 		/* in one segment binaries, the rom data address is on top of the ram data address */
 		__data_start = .;
 		*(.data .data.* .gnu.linkonce.d.*)
+		. = ALIGN(4);
 	}
 
 	.ctors : ALIGN(4) {
@@ -145,13 +161,20 @@ SECTIONS
 		. = ALIGN(4);
 		__bss_end = .;
 	}
-	__heap2_start = .;
 
-	/*
-	 * We're keeping track of the padding added before the .nozi
-	 * section so we can do something useful with this otherwise wasted
-	 * memory.
-	 */
+	.heap1 (NOLOAD) : {
+		/*
+		 * We're keeping track of the padding added before the
+		 * .nozi section so we can do something useful with
+		 * this otherwise wasted memory.
+		 */
+		__heap1_start = .;
+#ifndef CFG_WITH_PAGER
+		. += HEAP_SIZE;
+#endif
+		. = ALIGN(16 * 1024);
+		__heap1_end = .;
+	}
 
  	/*
 	 * Uninitialized data that shouldn't be zero initialized at
@@ -159,43 +182,89 @@ SECTIONS
 	 *
 	 * L1 mmu table requires 16 KiB alignment
 	 */
-	.nozi : ALIGN(16 * 1024) {
-		__heap2_end = .;
+	.nozi (NOLOAD) : ALIGN(16 * 1024) {
 		KEEP(*(.nozi .nozi.*))
 	}
 
-
-#if PLATFORM_FLAVOR_IS(fvp)
-	/*
-	 * The base address of the coherent memory section must be
-	 * page-aligned (4K) to guarantee that the coherent data are stored
-	 * on their own pages and are not mixed with normal data.  This is
-	 * required to set up the correct memory attributes for the
-	 * coherent data page tables.
-	 */
-	coherent_ram (NOLOAD) : ALIGN(4096) {
-		__COHERENT_RAM_START__ = .;
-		*(tzfw_coherent_mem)
-		__COHERENT_RAM_END_UNALIGNED__ = .;
+#ifdef CFG_WITH_PAGER
+	.heap2 (NOLOAD) : {
+		__heap2_start = .;
 		/*
-		 * Memory page(s) mapped to this section will be marked as
-		 * device memory.  No other unexpected data must creep in.
-		 * Ensure the rest of the current memory page is unused.
+		 * Reserve additional memory for heap, the total should
+		 * be at least HEAP_SIZE, but count what has already been
+		 * reserved in .heap1
 		 */
-		. = NEXT(4096);
-		__COHERENT_RAM_END__ = .;
+		. += HEAP_SIZE - (__heap1_end - __heap1_start);
+		. = ALIGN(4 * 1024);
+		__heap2_end = .;
 	}
-#endif
 
-	__heap1_start = .;
-	. += 0x40000 /*256KiB*/;
-	__heap1_end = .;
+	.text_init : ALIGN(4 * 1024) {
+		__text_init_start = .;
+/*
+ * Include list of sections needed for boot initialization, this list
+ * overlaps with unpaged.ld.S but since unpaged.ld.S is first all those
+ * sections will go into the unpaged area.
+ */
+#include <text_init.ld.S>
+		. = ALIGN(4);
+		__text_init_end = .;
+	}
+
+	.rodata_init : ALIGN(4) {
+		__rodata_init_start = .;
+#include <rodata_init.ld.S>
+		. = ALIGN(4);
+		__rodata_init_end = .;
+	}
+	__init_start = __text_init_start;
+	__init_end = .;
+	__init_size = __init_end - __text_init_start;
+	__init_mem_usage = __init_end - CFG_TEE_LOAD_ADDR;
+
+	.text_pagable : ALIGN(4) {
+		__text_pagable_start = .;
+		*(.text*)
+		. = ALIGN(4);
+		__text_pagable_end = .;
+	}
+
+	.rodata_pagable : ALIGN(4) {
+		__rodata_pagable_start = .;
+		*(.rodata*)
+		. = ALIGN(4);
+		__start_ta_head_section = . ;
+		KEEP(*(ta_head_section))
+		__stop_ta_head_section = . ;
+		. = ALIGN(4 * 1024);
+		__rodata_pagable_end = .;
+	}
+
+	__pagable_part_start = __rodata_init_end;
+	__pagable_part_end = __rodata_pagable_end;
+	__pagable_start = __text_init_start;
+	__pagable_end = __pagable_part_end;
+
+	ASSERT(CFG_TEE_LOAD_ADDR >= CFG_TEE_RAM_START,
+		"Load address before start of physical memory")
+	ASSERT(CFG_TEE_LOAD_ADDR < (CFG_TEE_RAM_START + CFG_TEE_RAM_PH_SIZE),
+		"Load address after end of physical memory")
+	ASSERT(__init_end < (CFG_TEE_RAM_START + CFG_TEE_RAM_PH_SIZE),
+		"OP-TEE can't fit init part into available physical memory")
+
+
+#endif /*CFG_WITH_PAGER*/
 
 	_end = .;
 
+#ifndef CFG_WITH_PAGER
+	__init_size = __data_end - CFG_TEE_LOAD_ADDR;
+	__init_mem_usage = _end - CFG_TEE_LOAD_ADDR;
+#endif
 	. = CFG_TEE_RAM_START + CFG_TEE_RAM_VA_SIZE;
 	_end_of_ram = .;
 
 	/* Strip unnecessary stuff */
 	/DISCARD/ : { *(.comment .note .eh_frame) }
+
 }
diff --git a/core/arch/arm32/plat-vexpress/link.mk b/core/arch/arm32/plat-vexpress/link.mk
index 9f90e44..e17948c 100644
--- a/core/arch/arm32/plat-vexpress/link.mk
+++ b/core/arch/arm32/plat-vexpress/link.mk
@@ -6,16 +6,12 @@ link-script-dep = $(link-out-dir)/.kern.ld.d
 
 AWK	 = awk
 
-all: $(link-out-dir)/tee.elf $(link-out-dir)/tee.dmp $(link-out-dir)/tee.bin
-all: $(link-out-dir)/tee.symb_sizes
-cleanfiles += $(link-out-dir)/tee.elf $(link-out-dir)/tee.dmp $(link-out-dir)/tee.map
-cleanfiles += $(link-out-dir)/tee.bin
-cleanfiles += $(link-out-dir)/tee.symb_sizes
-cleanfiles += $(link-script-pp) $(link-script-dep)
 
 link-ldflags  = $(LDFLAGS)
 link-ldflags += -T $(link-script-pp) -Map=$(link-out-dir)/tee.map
 link-ldflags += --sort-section=alignment
+link-ldflags += --fatal-warnings
+link-ldflags += --print-gc-sections
 
 link-ldadd  = $(LDADD)
 link-ldadd += $(addprefix -L,$(libdirs))
@@ -25,30 +21,147 @@ ldargs-tee.elf := $(link-ldflags) $(objs) $(link-ldadd) $(libgcc)
 link-script-cppflags := -DASM=1 \
 	$(filter-out $(CPPFLAGS_REMOVE) $(cppflags-remove), \
 		$(nostdinc) $(CPPFLAGS) \
-		$(addprefix -I,$(incdirs$(sm))) $(cppflags$(sm)))
+		$(addprefix -I,$(incdirs$(sm)) $(link-out-dir)) \
+		$(cppflags$(sm)))
+
+entries-unpaged += tee_pager_abort_handler
+entries-unpaged += pager_load_rodata
+entries-unpaged += thread_init_vbar
+entries-unpaged += sm_init
+entries-unpaged += core_init_mmu_regs
+entries-unpaged += main_cpu_on_handler
+entries-unpaged += main_init_secondary
+entries-unpaged += stack_tmp_top
+objs-unpaged := \
+	$(filter-out $(addprefix $(out-dir)/, $(objs-unpaged-rem)), $(objs))
+ldargs-unpaged := -i --gc-sections \
+	$(addprefix -u, $(entries-unpaged)) \
+	$(objs-unpaged) $(link-ldadd) $(libgcc)
+cleanfiles += $(link-out-dir)/unpaged.o
+$(link-out-dir)/unpaged.o: $(objs-unpaged) $(libdeps) $(MAKEFILE_LIST)
+	@echo '  LD      $@'
+	$(q)$(LD) $(ldargs-unpaged) -o $@
 
+cleanfiles += $(link-out-dir)/text_unpaged.ld.S:
+$(link-out-dir)/text_unpaged.ld.S: $(link-out-dir)/unpaged.o
+	@echo '  GEN     $@'
+	$(q)$(READELF) -a -W $< | ${AWK} -f ./scripts/gen_ld_text_sects.awk > $@
+
+cleanfiles += $(link-out-dir)/rodata_unpaged.ld.S:
+$(link-out-dir)/rodata_unpaged.ld.S: $(link-out-dir)/unpaged.o
+	@echo '  GEN     $@'
+	$(q)$(READELF) -a -W $< | \
+		${AWK} -f ./scripts/gen_ld_rodata_sects.awk > $@
+
+objs-init-rem += core/arch/arm32/tee/arch_svc.o
+objs-init-rem += core/arch/arm32/tee/arch_svc_asm.o
+objs-init-rem += core/arch/arm32/plat-vexpress/plat_tee_func.o
+objs-init-rem += core/arch/arm32/tee/init.o
+entries-init += _start
+objs-init := \
+	$(filter-out $(addprefix $(out-dir)/, $(objs-init-rem)), $(objs))
+ldargs-init := -i --gc-sections \
+	$(addprefix -u, $(entries-init)) \
+	$(objs-init) $(link-ldadd) $(libgcc)
+cleanfiles += $(link-out-dir)/init.o
+$(link-out-dir)/init.o: $(objs-init) $(libdeps) $(MAKEFILE_LIST)
+	@echo '  LD      $@'
+	$(q)$(LD) $(ldargs-init) -o $@
+
+cleanfiles += $(link-out-dir)/text_init.ld.S:
+$(link-out-dir)/text_init.ld.S: $(link-out-dir)/init.o
+	@echo '  GEN     $@'
+	$(q)$(READELF) -a -W $< | ${AWK} -f ./scripts/gen_ld_text_sects.awk > $@
+
+cleanfiles += $(link-out-dir)/rodata_init.ld.S:
+$(link-out-dir)/rodata_init.ld.S: $(link-out-dir)/init.o
+	@echo '  GEN     $@'
+	$(q)$(READELF) -a -W $< | \
+		${AWK} -f ./scripts/gen_ld_rodata_sects.awk > $@
 
 -include $(link-script-dep)
 
-$(link-script-pp): $(link-script)
+link-script-extra-deps += $(link-out-dir)/text_unpaged.ld.S
+link-script-extra-deps += $(link-out-dir)/rodata_unpaged.ld.S
+link-script-extra-deps += $(link-out-dir)/text_init.ld.S
+link-script-extra-deps += $(link-out-dir)/rodata_init.ld.S
+cleanfiles += $(link-script-pp) $(link-script-dep)
+$(link-script-pp): $(link-script) $(link-script-extra-deps)
 	@echo '  CPP     $@'
 	@mkdir -p $(dir $@)
 	$(q)$(CPP) -Wp,-P,-MT,$@,-MD,$(link-script-dep) \
 		$(link-script-cppflags) $< > $@
 
-
+all: $(link-out-dir)/tee.elf
+cleanfiles += $(link-out-dir)/tee.elf $(link-out-dir)/tee.map
 $(link-out-dir)/tee.elf: $(objs) $(libdeps) $(link-script-pp)
 	@echo '  LD      $@'
 	$(q)$(LD) $(ldargs-tee.elf) -o $@
 
+all: $(link-out-dir)/tee.dmp
+cleanfiles += $(link-out-dir)/tee.dmp
 $(link-out-dir)/tee.dmp: $(link-out-dir)/tee.elf
 	@echo '  OBJDUMP $@'
 	$(q)$(OBJDUMP) -l -x -d $< > $@
 
-$(link-out-dir)/tee.bin: $(link-out-dir)/tee.elf
+pagable_sections := .*_pagable
+init_sections := .*_init
+cleanfiles += $(link-out-dir)/tee-pager.bin
+$(link-out-dir)/tee-pager.bin: $(link-out-dir)/tee.elf
 	@echo '  OBJCOPY $@'
-	$(q)$(OBJCOPY) -O binary $< $@
+	$(q)$(OBJCOPY) -O binary \
+		--remove-section="$(pagable_sections)" \
+		--remove-section="$(init_sections)" \
+		$< $@
 
+cleanfiles += $(link-out-dir)/tee-pagable.bin
+$(link-out-dir)/tee-pagable.bin: $(link-out-dir)/tee.elf
+	@echo '  OBJCOPY $@'
+	$(q)$(OBJCOPY) -O binary \
+		--only-section="$(init_sections)" \
+		--only-section="$(pagable_sections)" \
+		$< $@
+
+cleanfiles += $(link-out-dir)/tee-init_size.txt
+$(link-out-dir)/tee-init_size.txt: $(link-out-dir)/tee.elf
+	@echo '  GEN     $@'
+	@echo -n 0x > $@
+	$(q)$(NM) $< | grep __init_size | sed 's/ .*$$//' >> $@
+
+cleanfiles += $(link-out-dir)/tee-init_load_addr.txt
+$(link-out-dir)/tee-init_load_addr.txt: $(link-out-dir)/tee.elf
+	@echo '  GEN     $@'
+	@echo -n 0x > $@
+	$(q)$(NM) $< | grep ' _start' | sed 's/ .*$$//' >> $@
+
+cleanfiles += $(link-out-dir)/tee-init_mem_usage.txt
+$(link-out-dir)/tee-init_mem_usage.txt: $(link-out-dir)/tee.elf
+	@echo '  GEN     $@'
+	@echo -n 0x > $@
+	$(q)$(NM) $< | grep ' __init_mem_usage' | sed 's/ .*$$//' >> $@
+
+all: $(link-out-dir)/tee.bin
+cleanfiles += $(link-out-dir)/tee.bin
+$(link-out-dir)/tee.bin: $(link-out-dir)/tee-pager.bin \
+			 $(link-out-dir)/tee-pagable.bin \
+			 $(link-out-dir)/tee-init_size.txt \
+			 $(link-out-dir)/tee-init_load_addr.txt \
+			 $(link-out-dir)/tee-init_mem_usage.txt \
+			./scripts/gen_hashed_bin.py
+	@echo '  GEN     $@'
+	$(q)./scripts/gen_hashed_bin.py \
+		--arch 0 \
+		--init_size `cat $(link-out-dir)/tee-init_size.txt` \
+		--init_load_addr_lo \
+			`cat $(link-out-dir)/tee-init_load_addr.txt` \
+		--init_mem_usage `cat $(link-out-dir)/tee-init_mem_usage.txt` \
+		--tee_pager_bin $(link-out-dir)/tee-pager.bin \
+		--tee_pagable_bin $(link-out-dir)/tee-pagable.bin \
+		--out $@
+
+
+all: $(link-out-dir)/tee.symb_sizes
+cleanfiles += $(link-out-dir)/tee.symb_sizes
 $(link-out-dir)/tee.symb_sizes: $(link-out-dir)/tee.elf
 	@echo '  GEN     $@'
 	$(q)$(NM) --print-size --reverse-sort --size-sort $< > $@
diff --git a/core/arch/arm32/plat-vexpress/platform_config.h b/core/arch/arm32/plat-vexpress/platform_config.h
index d34e9af..bdab66d 100644
--- a/core/arch/arm32/plat-vexpress/platform_config.h
+++ b/core/arch/arm32/plat-vexpress/platform_config.h
@@ -119,8 +119,6 @@
 
 #define CFG_TEE_CORE_NB_CORE	8
 
-#define CFG_TEE_LOAD_ADDR	(TZDRAM_BASE + 0x1000)
-
 #define CFG_SHMEM_START		(DRAM0_BASE + 0x1000000)
 #define CFG_SHMEM_SIZE		0x100000