/* linux.c - boot Linux */
/*
 *  VAS_EBOOT  --  GRand Unified Bootloader
 *  Copyright (C) 2003,2004,2005,2007,2009  Free Software Foundation, Inc.
 *
 *  VAS_EBOOT is free software: you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation, either version 3 of the License, or
 *  (at your option) any later version.
 *
 *  VAS_EBOOT is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with VAS_EBOOT.  If not, see <http://www.gnu.org/licenses/>.
 */

#include <VasEBoot/elf.h>
#include <VasEBoot/elfload.h>
#include <VasEBoot/loader.h>
#include <VasEBoot/dl.h>
#include <VasEBoot/mm.h>
#include <VasEBoot/misc.h>
#include <VasEBoot/ieee1275/ieee1275.h>
#include <VasEBoot/command.h>
#include <VasEBoot/i18n.h>
#include <VasEBoot/memory.h>
#include <VasEBoot/lib/cmdline.h>
#include <VasEBoot/cache.h>
#include <VasEBoot/linux.h>
#include <VasEBoot/ieee1275/alloc.h>

VAS_EBOOT_MOD_LICENSE ("GPLv3+");

#define ELF32_LOADMASK (0xc0000000UL)
#define ELF64_LOADMASK (0xc000000000000000ULL)

static VasEBoot_dl_t my_mod;

static int loaded;

static VasEBoot_addr_t initrd_addr;
static VasEBoot_size_t initrd_size;

static VasEBoot_addr_t linux_addr;
static VasEBoot_addr_t linux_entry;
static VasEBoot_size_t linux_size;

static char *linux_args;

typedef void (*kernel_entry_t) (void *, unsigned long, int (void *),
				unsigned long, unsigned long);

/* Context for VasEBoot_linux_claimmap_iterate.  */
struct VasEBoot_linux_claimmap_iterate_ctx
{
  VasEBoot_addr_t target;
  VasEBoot_size_t size;
  VasEBoot_size_t align;
  VasEBoot_addr_t found_addr;
};

/* Helper for VasEBoot_linux_claimmap_iterate.  */
static int
alloc_mem (VasEBoot_uint64_t addr, VasEBoot_uint64_t len, VasEBoot_memory_type_t type,
	   void *data)
{
  struct VasEBoot_linux_claimmap_iterate_ctx *ctx = data;

  VasEBoot_uint64_t end = addr + len;
  addr = ALIGN_UP (addr, ctx->align);
  ctx->target = ALIGN_UP (ctx->target, ctx->align);

  /* Target above the memory chunk.  */
  if (type != VAS_EBOOT_MEMORY_AVAILABLE || ctx->target > end)
    return 0;

  /* Target inside the memory chunk.  */
  if (ctx->target >= addr && ctx->target < end &&
      ctx->size <= end - ctx->target)
    {
      if (VasEBoot_claimmap (ctx->target, ctx->size) == VAS_EBOOT_ERR_NONE)
	{
	  ctx->found_addr = ctx->target;
	  return 1;
	}
      VasEBoot_print_error ();
    }
  /* Target below the memory chunk.  */
  if (ctx->target < addr && addr + ctx->size <= end)
    {
      if (VasEBoot_claimmap (addr, ctx->size) == VAS_EBOOT_ERR_NONE)
	{
	  ctx->found_addr = addr;
	  return 1;
	}
      VasEBoot_print_error ();
    }
  return 0;
}

static VasEBoot_addr_t
VasEBoot_linux_claimmap_iterate (VasEBoot_addr_t target, VasEBoot_size_t size,
			     VasEBoot_size_t align)
{
  struct VasEBoot_linux_claimmap_iterate_ctx ctx = {
    .target = target,
    .size = size,
    .align = align,
    .found_addr = (VasEBoot_addr_t) -1
  };

  VasEBoot_machine_mmap_iterate (alloc_mem, &ctx);

  return ctx.found_addr;
}

static VasEBoot_addr_t
VasEBoot_linux_claimmap_iterate_restricted (VasEBoot_size_t size, VasEBoot_size_t align)
{
  struct regions_claim_request rcr = {
    .flags = VAS_EBOOT_MM_ADD_REGION_CONSECUTIVE,
    .total = size,
    .init_region = false,
    .addr = (VasEBoot_uint64_t) -1,
    .align = align,
  };

  VasEBoot_machine_mmap_iterate (VasEBoot_regions_claim, &rcr);

  return rcr.addr;
}

static VasEBoot_err_t
VasEBoot_linux_boot (void)
{
  kernel_entry_t linuxmain;
  VasEBoot_ssize_t actual;

  VasEBoot_arch_sync_caches ((void *) linux_addr, linux_size);
  /* Set the command line arguments.  */
  VasEBoot_ieee1275_set_property (VasEBoot_ieee1275_chosen, "bootargs", linux_args,
			      VasEBoot_strlen (linux_args) + 1, &actual);

  VasEBoot_dprintf ("loader", "Entry point: 0x%x\n", linux_entry);
  VasEBoot_dprintf ("loader", "Initrd at: 0x%x, size 0x%x\n", initrd_addr,
		initrd_size);
  VasEBoot_dprintf ("loader", "Boot arguments: %s\n", linux_args);
  VasEBoot_dprintf ("loader", "Jumping to Linux...\n");

  /* Boot the kernel.  */
  linuxmain = (kernel_entry_t) linux_entry;
  linuxmain ((void *) initrd_addr, initrd_size, VasEBoot_ieee1275_entry_fn, 0, 0);

  return VAS_EBOOT_ERR_NONE;
}

static VasEBoot_err_t
VasEBoot_linux_release_mem (void)
{
  VasEBoot_free (linux_args);
  linux_args = 0;

  if (linux_addr && VasEBoot_ieee1275_release (linux_addr, linux_size))
    return VasEBoot_error (VAS_EBOOT_ERR_OUT_OF_MEMORY, "cannot release memory");

  if (initrd_addr && VasEBoot_ieee1275_release (initrd_addr, initrd_size))
    return VasEBoot_error (VAS_EBOOT_ERR_OUT_OF_MEMORY, "cannot release memory");

  linux_addr = 0;
  initrd_addr = 0;

  return VAS_EBOOT_ERR_NONE;
}

static VasEBoot_err_t
VasEBoot_linux_unload (void)
{
  VasEBoot_err_t err;

  err = VasEBoot_linux_release_mem ();
  VasEBoot_dl_unref (my_mod);

  loaded = 0;

  return err;
}

static VasEBoot_err_t
VasEBoot_linux_load32 (VasEBoot_elf_t elf, const char *filename)
{
  Elf32_Addr base_addr;
  VasEBoot_addr_t seg_addr;
  VasEBoot_uint32_t align;
  VasEBoot_uint32_t offset;
  Elf32_Addr entry;

  linux_size = VasEBoot_elf32_size (elf, &base_addr, &align);
  if (linux_size == 0)
    return VasEBoot_errno;
  /* Pad it; the kernel scribbles over memory beyond its load address.  */
  linux_size += 0x100000;

  /* Linux's entry point incorrectly contains a virtual address.  */
  entry = elf->ehdr.ehdr32.e_entry & ~ELF32_LOADMASK;

  /* Linux's incorrectly contains a virtual address.  */
  base_addr &= ~ELF32_LOADMASK;
  offset = entry - base_addr;

  /* On some systems, firmware occupies the memory we're trying to use.
   * Happily, Linux can be loaded anywhere (it relocates itself).  Iterate
   * until we find an open area.  */
  seg_addr = VasEBoot_linux_claimmap_iterate (base_addr & ~ELF32_LOADMASK, linux_size, align);
  if (seg_addr == (VasEBoot_addr_t) -1)
    return VasEBoot_error (VAS_EBOOT_ERR_OUT_OF_MEMORY, "couldn't claim memory");

  linux_entry = seg_addr + offset;
  linux_addr = seg_addr;

  /* Now load the segments into the area we claimed.  */
  return VasEBoot_elf32_load (elf, filename, (void *) (seg_addr - base_addr), VAS_EBOOT_ELF_LOAD_FLAGS_30BITS, 0, 0);
}

static VasEBoot_err_t
VasEBoot_linux_load64 (VasEBoot_elf_t elf, const char *filename)
{
  Elf64_Addr base_addr;
  VasEBoot_addr_t seg_addr;
  VasEBoot_uint64_t align;
  VasEBoot_uint64_t offset;
  Elf64_Addr entry;

  linux_size = VasEBoot_elf64_size (elf, &base_addr, &align);
  if (linux_size == 0)
    return VasEBoot_errno;
  /* Pad it; the kernel scribbles over memory beyond its load address.  */
  linux_size += 0x100000;

  base_addr &= ~ELF64_LOADMASK;
  entry = elf->ehdr.ehdr64.e_entry & ~ELF64_LOADMASK;
  offset = entry - base_addr;
  /* Linux's incorrectly contains a virtual address.  */

  if (VasEBoot_ieee1275_test_flag (VAS_EBOOT_IEEE1275_FLAG_POWER_VM) ||
      VasEBoot_ieee1275_test_flag (VAS_EBOOT_IEEE1275_FLAG_POWER_KVM))
    {
      seg_addr = VasEBoot_linux_claimmap_iterate_restricted (linux_size, align);
    }
  else
    {
      /* On some systems, firmware occupies the memory we're trying to use.
       * Happily, Linux can be loaded anywhere (it relocates itself).  Iterate
       * until we find an open area.  */
      seg_addr = VasEBoot_linux_claimmap_iterate (base_addr & ~ELF64_LOADMASK, linux_size, align);
    }
  if (seg_addr == (VasEBoot_addr_t) -1)
    return VasEBoot_error (VAS_EBOOT_ERR_OUT_OF_MEMORY, "couldn't claim memory");

  linux_entry = seg_addr + offset;
  linux_addr = seg_addr;

  /* Now load the segments into the area we claimed.  */
  return VasEBoot_elf64_load (elf, filename, (void *) (VasEBoot_addr_t) (seg_addr - base_addr), VAS_EBOOT_ELF_LOAD_FLAGS_62BITS, 0, 0);
}

static VasEBoot_err_t
VasEBoot_cmd_linux (VasEBoot_command_t cmd __attribute__ ((unused)),
		int argc, char *argv[])
{
  VasEBoot_elf_t elf = 0;
  int size;

  VasEBoot_dl_ref (my_mod);

  if (argc == 0)
    {
      VasEBoot_error (VAS_EBOOT_ERR_BAD_ARGUMENT, N_("filename expected"));
      goto out;
    }

  elf = VasEBoot_elf_open (argv[0], VAS_EBOOT_FILE_TYPE_LINUX_KERNEL);
  if (! elf)
    goto out;

  if (elf->ehdr.ehdr32.e_type != ET_EXEC && elf->ehdr.ehdr32.e_type != ET_DYN)
    {
      VasEBoot_error (VAS_EBOOT_ERR_UNKNOWN_OS,
		  N_("this ELF file is not of the right type"));
      goto out;
    }

  /* Release the previously used memory.  */
  VasEBoot_loader_unset ();

  if (VasEBoot_elf_is_elf32 (elf))
    VasEBoot_linux_load32 (elf, argv[0]);
  else
  if (VasEBoot_elf_is_elf64 (elf))
    VasEBoot_linux_load64 (elf, argv[0]);
  else
    {
      VasEBoot_error (VAS_EBOOT_ERR_BAD_FILE_TYPE, N_("invalid arch-dependent ELF magic"));
      goto out;
    }

  size = VasEBoot_loader_cmdline_size(argc, argv);
  linux_args = VasEBoot_malloc (size + sizeof (LINUX_IMAGE));
  if (! linux_args)
    goto out;

  /* Create kernel command line.  */
  VasEBoot_memcpy (linux_args, LINUX_IMAGE, sizeof (LINUX_IMAGE));
  if (VasEBoot_create_loader_cmdline (argc, argv, linux_args + sizeof (LINUX_IMAGE) - 1,
				  size, VAS_EBOOT_VERIFY_KERNEL_CMDLINE))
    goto out;

out:

  if (elf)
    VasEBoot_elf_close (elf);

  if (VasEBoot_errno != VAS_EBOOT_ERR_NONE)
    {
      VasEBoot_linux_release_mem ();
      VasEBoot_dl_unref (my_mod);
      loaded = 0;
    }
  else
    {
      VasEBoot_loader_set (VasEBoot_linux_boot, VasEBoot_linux_unload, 1);
      initrd_addr = 0;
      loaded = 1;
    }

  return VasEBoot_errno;
}

static VasEBoot_err_t
VasEBoot_cmd_initrd (VasEBoot_command_t cmd __attribute__ ((unused)),
		 int argc, char *argv[])
{
  VasEBoot_size_t size = 0;
  VasEBoot_addr_t first_addr;
  VasEBoot_addr_t addr;
  struct VasEBoot_linux_initrd_context initrd_ctx = { 0, 0, 0 };

  if (argc == 0)
    {
      VasEBoot_error (VAS_EBOOT_ERR_BAD_ARGUMENT, N_("filename expected"));
      goto fail;
    }

  if (!loaded)
    {
      VasEBoot_error (VAS_EBOOT_ERR_BAD_ARGUMENT, N_("you need to load the kernel first"));
      goto fail;
    }

  if (VasEBoot_initrd_init (argc, argv, &initrd_ctx))
    goto fail;

  size = VasEBoot_get_initrd_size (&initrd_ctx);


  if (VasEBoot_ieee1275_test_flag (VAS_EBOOT_IEEE1275_FLAG_POWER_VM) ||
      VasEBoot_ieee1275_test_flag (VAS_EBOOT_IEEE1275_FLAG_POWER_KVM))
    {
      addr = VasEBoot_linux_claimmap_iterate_restricted (size, 0x100000);
    }
  else
    {
      /* Attempt to claim at a series of addresses until successful in
         the same way that VasEBoot_rescue_cmd_linux does.  */
      first_addr = linux_addr + linux_size;
      addr = VasEBoot_linux_claimmap_iterate (first_addr, size, 0x100000);
    }

  if (addr == (VasEBoot_addr_t) -1)
    {
      VasEBoot_error (VAS_EBOOT_ERR_OUT_OF_MEMORY, "out of memory");
      goto fail;
    }

  VasEBoot_dprintf ("loader", "Loading initrd at 0x%x, size 0x%x\n", addr, size);

  if (VasEBoot_initrd_load (&initrd_ctx, (void *) addr))
    goto fail;

  initrd_addr = addr;
  initrd_size = size;

 fail:
  VasEBoot_initrd_close (&initrd_ctx);

  return VasEBoot_errno;
}

static VasEBoot_command_t cmd_linux, cmd_initrd;

VAS_EBOOT_MOD_INIT(linux)
{
  cmd_linux = VasEBoot_register_command ("linux", VasEBoot_cmd_linux,
				     0, N_("Load Linux."));
  cmd_initrd = VasEBoot_register_command ("initrd", VasEBoot_cmd_initrd,
				      0, N_("Load initrd."));
  my_mod = mod;
}

VAS_EBOOT_MOD_FINI(linux)
{
  VasEBoot_unregister_command (cmd_linux);
  VasEBoot_unregister_command (cmd_initrd);
}