towards multi-arch

arch/bochs.mk

@@ -0,0 +1,30 @@
+FAIL_DOWNLOAD_URL = ${FAIL_DOWNLOAD_BASE}?job=build-bochs-generic-tools
+
+CFLAGS += -m32
+LDFLAGS += -Wl,-T linker.ld $^ -m32 -static -nostdlib  -Wl,--build-id=none
+
+################################################################
+# Build Targets
+${BUILD_DIR}/%/system.o: %.c
+	mkdir -p $(shell dirname $@)
+	${CC} $< -o $@ ${CFLAGS} -c -ffunction-sections
+
+${BUILD_DIR}/startup.o: arch/bochs/startup.s
+	${CC} $< ${CFLAGS} -c -o $@ -ffunction-sections
+
+${BUILD_DIR}/%/system.elf: ${BUILD_DIR}/%/system.o ${BUILD_DIR}/startup.o
+	${CC} ${LDFLAGS} -o $@
+
+${BUILD_DIR}/%/system.iso: ${BUILD_DIR}/%/system.elf
+	rm -rf $(shell dirname $<)/grub
+	mkdir -p $(shell dirname $<)/grub/boot/grub
+	cp grub.cfg $(shell dirname $<)/grub/boot/grub
+	cp $< $(shell dirname $<)/grub/boot/system.elf
+	grub-mkrescue -o $@ $(shell dirname $<)/grub
+
+define arch-make-targets
+build-$1: ${BUILD_DIR}/$1/system.iso
+endef
+
+
+

arch/bochs/lib.c

@@ -0,0 +1,4 @@
+#pragma once
+
+#define ARCH_ASM_CLOBBER_ALL "eax", "ebx", "ecx", "edx", "esi", "edi", "ebp"
+

arch/bochs/startup.s

@@ -0,0 +1,59 @@
+## Bare bone boot.s from wiki.osdev.org
+
+# multiboot header
+.section .rodata.multiboot
+.align 4
+
+# magic number
+.long 0x1BADB002
+
+# flags: align, meminfo
+.long 0x3
+
+# checksum: -(magic+flags)
+.long -(0x1BADB002 + 0x3)
+
+
+
+# the initial kernel stack
+.section .kernel_stack
+.global os_stack
+.size os_stack, 4096
+#.Lstack_bottom:
+os_stack:
+.byte 0
+#.skip 16384 # 16 KiB
+.skip 4094 # 4 KiB
+.byte 0
+.Lstack_top:
+
+
+
+# The linker script specifies _start as the entry point to the kernel and the
+# bootloader will jump to this position once the kernel has been loaded. It
+# doesn't make sense to return from this function as the bootloader is gone.
+.section .text.startup
+.global _start
+.type _start, @function
+_start:
+	# Welcome to kernel mode! 
+	# To set up a stack, we simply set the esp register to point to the top of
+	# our stack (as it grows downwards).
+	movl $.Lstack_top, %esp
+	# We are now ready to actually execute C code. (see ./startup.cc)
+	call os_main
+
+	# In case the function returns, we'll want to put the computer into an
+	# infinite loop. To do that, we use the clear interrupt ('cli') instruction
+	# to disable interrupts, the halt instruction ('hlt') to stop the CPU until
+	# the next interrupt arrives, and jumping to the halt instruction if it ever
+	# continues execution, just to be safe. We will create a local label rather
+	# than real symbol and jump to there endlessly.
+	cli
+	hlt
+.Lhang:
+	jmp .Lhang
+
+# Set the size of the _start symbol to the current location '.' minus its start.
+# This is useful when debugging or when you implement call tracing.
+.size _start, . - _start

arch/riscv-common.mk

@@ -0,0 +1,33 @@
+CC      := clang-11
+CFLAGS  += -nostdlib -fno-inline-functions -fno-unroll-loops -mcmodel=medium -gdwarf-5
+LDFLAGS += -nostdlib -nostartfiles -static -fuse-ld=lld  -Wl,-nostdlib -Wl,--script=arch/riscv/linker.ld
+
+
+${BUILD_DIR}/startup.o: arch/riscv/bootcode/${ARCH}.c
+	@mkdir -p $(dir $@)
+	${CC} ${CFLAGS} -o $@ -c $^
+
+${BUILD_DIR}/%/system.o: %.c
+	@mkdir -p $(dir $@)
+	${CC} ${CFLAGS} -o $@ -c $^
+
+${BUILD_DIR}/%/system.elf:  ${BUILD_DIR}/startup.o ${BUILD_DIR}/%/system.o
+	${CC} ${CFLAGS} ${LDFLAGS} -o $@ $^
+
+
+define arch-make-targets
+build-$1: ${BUILD_DIR}/$1/system.elf
+
+trace-$1: ${BUILD_DIR}/$1/system.elf
+	${FAIL_TRACE} \
+	-Wf,--elf-file      -Wf,$$<             \
+	-Wf,--start-symbol  -Wf,start_trace 	\
+	-Wf,--end-symbol    -Wf,stop_trace  	\
+	-Wf,--check-bounds                      \
+	-Wf,--state-file=$$(dir $$<)/state      \
+	-Wf,--trace-file=$$(dir $$<)/trace.pb   \
+	$$<
+
+endef
+
+

arch/riscv/bootcode/boot.h

@@ -0,0 +1,30 @@
+#pragma once
+
+#if defined(__riscv) && !defined(__CHERI__)
+#define RISCV 1
+#define ARCH_ASM_CLOBBER_ALL NO_CAP_REGS
+
+#elif defined(__riscv) && defined(__CHERI__)
+#define RISCVCHERI 1
+#define ARCH_ASM_CLOBBER_ALL NO_CAPS_REGS, CAP_REGS
+
+#else
+#warning compiling for unknown architecture, using stdlib malloc.
+#endif
+
+#define NO_CAP_REGS "ra","sp","gp","tp","t0","t1","t2","t3","t4","t5","t6","s0","s1","s2","s3","s4","s5","s6","s7","s8","s9","s10","s11","a0","a1","a2","a3","a4","a5","a6","a7"
+
+#define CAP_REGS "cra","csp","cgp","ctp","ct0","ct1","ct2","ct3","ct4","ct5","ct6","cs0","cs1","cs2","cs3","cs4","cs5","cs6","cs7","cs8","cs9","cs10","cs11","ca0","ca1","ca2","ca3","ca4","ca5","ca6","ca7"
+
+#ifndef __ASSEMBLER__
+
+#if __riscv_xlen == 32
+typedef unsigned int size_t;
+#elif __riscv_xlen == 64
+typedef unsigned long size_t;
+#elif DEBUG
+#else
+#warning Unknown __riscv_xlen value
+typedef unsigned long size_t;
+#endif
+#endif

arch/riscv/bootcode/cheri32.c

@@ -0,0 +1,30 @@
+#include <cheri_init_globals.h>
+#include "boot.h"
+
+asm (
+".text" "\n"
+".option push" "\n"
+".option nocapmode" "\n"
+".global _start" "\n"
+"_start:" "\n"
+"       lla     sp, __sp" "\n"
+"       lla      t0, __stack_size" "\n"
+"       cfromptr        csp, ddc, sp" "\n"
+"       csetbounds      csp, csp, t0" "\n"
+"       cincoffset      csp, csp, t0" "\n"
+"       lla     t0, _start_purecap" "\n"
+"       cfromptr        ct0, ddc, t0" "\n"
+"       li      t1, 1" "\n"
+"       csetflags       ct0, ct0, t1" "\n"
+"       cjr     ct0" "\n"
+".option pop" "\n"
+);
+
+extern void os_main();
+
+void _start_purecap(void) {
+        cheri_init_globals_3(__builtin_cheri_global_data_get(),
+                __builtin_cheri_program_counter_get(),
+                __builtin_cheri_global_data_get());
+        os_main();
+}

arch/riscv/bootcode/cheri64.c

@@ -0,0 +1 @@
+cheri32.c

arch/riscv/bootcode/riscv32.c

@@ -0,0 +1,12 @@
+#include "boot.h"
+
+asm (
+".text" "\n"
+".global _start" "\n"
+"_start:" "\n"
+"       la     sp, __sp" "\n"
+"       la     t0, __stack_size" "\n"
+"       add    sp, sp, t0" "\n"
+"       la     t0,  os_main" "\n"
+"       jr     t0" "\n"
+       );

arch/riscv/bootcode/riscv64.c

@@ -0,0 +1 @@
+riscv32.c

arch/riscv/lib.c

@@ -0,0 +1,3 @@
+#include "arch/riscv/bootcode/boot.h"
+
+

arch/riscv/linker.ld

@@ -0,0 +1,254 @@
+
+/*======================================================================*/
+/* Default maven linker script                                          */
+/*======================================================================*/
+/* This is the default linker script for maven. It is based off of the
+   mips idt32.ld linker script. I have added many more comments and
+   tried to clean things up a bit. For more information about standard
+   MIPS sections see Section 9.5 of "See MIPS Run Linux" by Dominic
+   Sweetman. For more generic information about the init, fini, ctors,
+   and dtors sections see the paper titled "ELF From the Programmers
+   Perspective" by Hongiu Lu. */
+
+/*----------------------------------------------------------------------*/
+/* Setup                                                                */
+/*----------------------------------------------------------------------*/
+
+/* The OUTPUT_ARCH command specifies the machine architecture where the
+   argument is one of the names used in the BFD library. More
+   specifically one of the entires in bfd/cpu-mips.c */
+
+OUTPUT_ARCH( "riscv" )
+
+/* The ENTRY command specifies the entry point (ie. first instruction to
+   execute). The symbol _start is defined in crt0.S */
+
+ENTRY( _start )
+
+
+/*----------------------------------------------------------------------*/
+/* Sections                                                             */
+/*----------------------------------------------------------------------*/
+/* This is where we specify how the input sections map to output
+   sections. The .= commands set the location counter, and the
+   sections are inserted in increasing address order according to the
+   location counter. The following statement will take all of the .bar
+   input sections and reloate them into the .foo output section which
+   starts at address 0x1000.
+
+    . = 0.x1000;
+    .foo : { *(.bar) }
+
+   If we wrap an input specification with a KEEP command then it
+   prevents it from being eliminted during "link-time garbage
+   collection". I'm not sure what this is, so I just followed what was
+   done in idt32.ld.
+
+   We can also set a global external symbol to a specific address in the
+   output binary with this syntax:
+
+    _etext = .;
+    PROVIDE( etext = . );
+
+   This will set the global symbol _ftext to the current location. If we
+   wrap this in a PROVIDE commad, the symbol will only be set if it is
+   not defined. We do this with symbols which don't begin with an
+   underscore since technically in ansi C someone might have a function
+   with the same name (eg. etext).
+
+   If we need to label the beginning of a section we need to make sure
+   that the linker doesn't insert an orphan section inbetween where we
+   set the symbol and the actual begining of the section. We can do that
+   by assigning the location dot to itself.
+
+    . = .
+    _ftext = .;
+    .text :
+    { }
+
+   */
+
+SECTIONS
+{
+  /* Defines for stack and heap size. These are used by the startup code. */
+  __stack_size = 0x800;
+  __heap_size = 0x800;
+
+  /* Define data/bss padding size, defaults to zero but can be override by passing -Wl,--defsym=__padding_size=0x123 */
+  __padding_size = DEFINED(__padding_size) ? __padding_size: 0x0;
+
+  /*--------------------------------------------------------------------*/
+  /* Code and read-only segment                                         */
+  /*--------------------------------------------------------------------*/
+
+  /* Begining of code and text segment */
+  . = 0x80000000;
+  _ftext = .;
+
+  /* text: Program code section */
+  .text :
+  {
+    *(.text)
+    *(.text.*)
+    *(.gnu.linkonce.t.*)
+  }
+
+  /* init: Code to execute before main (called by crt0.S) */
+  .init :
+  {
+    KEEP( *(.init) )
+  }
+
+  /* fini: Code to execute after main (called by crt0.S) */
+  .fini :
+  {
+    KEEP( *(.fini) )
+  }
+
+  /* rodata: Read-only data */
+  .rodata :
+  {
+    *(.rdata)
+    *(.rodata)
+    *(.rodata.*)
+    *(.gnu.linkonce.r.*)
+  }
+
+  /* End of code and read-only segment */
+  PROVIDE( etext = . );
+  _etext = .;
+
+  /*--------------------------------------------------------------------*/
+  /* Global constructor/destructor segement                             */
+  /*--------------------------------------------------------------------*/
+
+  .preinit_array     :
+  {
+    PROVIDE_HIDDEN (__preinit_array_start = .);
+    KEEP (*(.preinit_array))
+    PROVIDE_HIDDEN (__preinit_array_end = .);
+  }
+
+  .init_array     :
+  {
+    PROVIDE_HIDDEN (__init_array_start = .);
+    KEEP (*(SORT(.init_array.*)))
+    KEEP (*(.init_array ))
+    PROVIDE_HIDDEN (__init_array_end = .);
+  }
+
+  .fini_array     :
+  {
+    PROVIDE_HIDDEN (__fini_array_start = .);
+    KEEP (*(SORT(.fini_array.*)))
+    KEEP (*(.fini_array ))
+    PROVIDE_HIDDEN (__fini_array_end = .);
+  }
+
+
+  /*--------------------------------------------------------------------*/
+  /* Other misc gcc segments (this was in idt32.ld)                     */
+  /*--------------------------------------------------------------------*/
+  /* I am not quite sure about these sections but it seems they are for
+     C++ exception handling. I think .jcr is for "Java Class
+     Registration" but it seems to end up in C++ binaries as well. */
+
+  .eh_frame_hdr     : { *(.eh_frame_hdr)     }
+  .eh_frame         : { KEEP( *(.eh_frame) ) }
+  .gcc_except_table : { *(.gcc_except_table) }
+  .jcr              : { KEEP (*(.jcr))       }
+
+  /*--------------------------------------------------------------------*/
+  /* Initialized data segment                                           */
+  /*--------------------------------------------------------------------*/
+
+  .padding_before : {
+    . += __padding_size;
+  }
+
+  /* Start of initialized data segment */
+  . = ALIGN(16);
+   _fdata = .;
+
+  /* data: Writable data */
+  .data : 
+  {
+    *(.data)
+    *(.data.*)
+    *(.gnu.linkonce.d.*)
+  }
+
+  /* End of initialized data segment */
+  PROVIDE( edata = . );
+  _edata = .;
+
+  /* Have _gp point to middle of sdata/sbss to maximize displacement range */
+  . = ALIGN(16);
+  _gp = . + 0x800;
+
+  /* Writable small data segment */
+  .sdata : 
+  {
+    *(.sdata)
+    *(.sdata.*)
+    *(.srodata.*)
+    *(.gnu.linkonce.s.*)
+  }
+
+  /*--------------------------------------------------------------------*/
+  /* Uninitialized data segment                                         */
+  /*--------------------------------------------------------------------*/
+
+  /* Start of uninitialized data segment */
+  . = ALIGN(16);
+  _fbss = .;
+
+  /* Writable uninitialized small data segment */
+  .sbss :
+  {
+    *(.sbss)
+    *(.sbss.*)
+    *(.gnu.linkonce.sb.*)
+  }
+
+  /* bss: Uninitialized writeable data section */
+  . = .;
+  _bss_start = .;
+  .bss : 
+  {
+    *(.bss)
+    *(.bss.*)
+    *(.gnu.linkonce.b.*)
+    *(COMMON)
+  }
+
+  .padding_after : {
+    . += __padding_size;
+  }
+
+  /*--------------------------------------------------------------------*/
+  /* Stack and Heap storage                                             */
+  /*--------------------------------------------------------------------*/
+  .heap : ALIGN(16) {
+      *(.heap)
+      *(.heap.*)
+  }
+
+  .stack : ALIGN(16) {
+      PROVIDE(__sp = .);
+      . += __stack_size;
+   }
+
+  /*--------------------------------------------------------------------*/
+  /* HTIF, isolated on a seperate page.                                */
+  /*--------------------------------------------------------------------*/
+  .htif (NOLOAD) : ALIGN(0x1000) {
+      tohost = .;
+      PROVIDE(tohost = tohost);
+      . += 8;
+      fromhost = .;
+      PROVIDE(fromhost = fromhost);
+      . += 8;
+  }
+  . = ALIGN(0x1000);
+}

arch/riscv32

@@ -0,0 +1 @@
+riscv

arch/riscv32.mk

@@ -0,0 +1,5 @@
+FAIL_DOWNLOAD_URL = ${FAIL_DOWNLOAD_BASE}?job=build-riscv-generic-tools%3A+%5Briscv64%5D
+
+include arch/riscv-common.mk
+
+CFLAGS += -march=rv32im -target riscv32-unknown-freebsd

arch/riscv64

@@ -0,0 +1 @@
+riscv