* stage2/terminfo.c (ti_set_term): Use a pointer to struct
terminfo instead to avoid GCC's bug, which inserts a reference
to memcpy implicitly.
(ti_get_term): Likewise.
All callers are fixed.
* stage2/terminfo.h (ti_set_term): Updated.
(ti_get_term): Likewise.
* stage2/shared.h (struct linux_kernel_header): New member,
initrd_max_address. Defined in the boot protocol 2.03 or higher.
* stage2/boot.c (load_initrd): If the boot protocol is greater
than or equal to 2.03, use the field ``initrd_max_address''
instead of LINUX_INITRD_MAX_ADDRESS.
* stage2/boot.c (load_image): Check if DATA_LEN plus SECTOR_SIZE
is less than or equal to MULTIBOOT_SEARCH, instead of if
DATA_LEN is less than or equal to MULTIBOOT_SEARCH.
Reported by Neelkanth Natu <neelnatu@yahoo.com>.
From Alexander Langer <alex@big.endian.de>:
* stage2/freebsd.h (RB_GDB): New macro.
(RB_MUTE): Likewise.
(RB_MULTIPLE): Likewise.
* stage2/boot.c (bsd_boot): Set the bits of RB_MULTIPLE, RB_GDB
and RB_MUTE when "-D", "-g" and "-m" are specified,
respectively.
* stage2/boot.c (load_image): Rewrite the Linux booting support
radically. Now it should work even on a machine having, say,
only 128KB, theoretically. Of course, GRUB itself doesn't work
on such a system, though.
(load_initrd): Initialize LH based on CUR_ADDR, because the
location becomes dynamic.
* stage2/shared.h (LINUX_MAX_SETUP_SECTS): Set to 64.
(LINUX_HEAP_END_OFFSET): Set to (0x9000 - 0x200).
(LINUX_STAGING_AREA): Removed.
(LINUX_SETUP): Likewise.
(LINUX_KERNEL): Likewise.
(LINUX_KERNEL_MAXLEN): Likewise.
(LINUX_SETUP_SEG): Likewise.
(LINUX_INIT_SEG): Likewise.
(LINUX_SETUP_STACK): Set to 0x9000.
(LINUX_BZIMAGE_ADDR): New macro.
(LINUX_ZIMAGE_ADDR): Likewise.
(LINUX_OLD_REAL_MODE_ADDR): Likewise.
(CL_MY_LOCATION): Removed.
(CL_MY_END_ADDR): Likewise.
(CL_BASE_ADDR): Likewise.
(CL_MAGIC): Renamed to ...
(LINUX_CL_MAGIC): ... this.
(LINUX_CL_OFFSET): New macro.
(LINUX_CL_END_OFFSET): Likewise.
(LINUX_SETUP_MOVE_SIZE): Likewise.
(struct linux_kernel_header): Change the type of the member
"cmd_line_ptr" to char *.
(linux_data_tmp_addr): Declared.
(linux_data_real_addr): Likewise.
* stage2/asm.S [!STAGE1_5] (linux_data_tmp_addr): New variable.
[!STAGE1_5] (linux_data_real_addr): Likewise.
[!STAGE1_5] (big_linux_boot): Copy the real mode part from
LINUX_DATA_TMP_ADDR to LINUX_DATA_REAL_ADDR.
* grub/asmstub.c (linux_data_tmp_addr): New variable.
(linux_data_real_addr): Likewise.
From Mark Kettenis <kettenis@chello.nl>:
* stage2/boot.c (load_image): Recognize newer FreeBSD kernels.
* stage2/i386-elf.h (EI_OSABI): New macro.
(EI_ABIVERSION): Likewise.
(ELFOSABI_FREEBSD): Likewise.
(EI_PAD): Set to 9.
Define the behavior of the boot loader when the load end address
and the bss end address are zero in the Multiboot Specification,
and add the support into GRUB. I've modified a patch from Yuri
Zaporogets <yuriz@ukr.net>.
* stage2/boot.c (load_image): In the case of Multiboot a.out
kludge, set the load end address to the load address plus the
size of the OS image file, if it is zero. Similarly, set the bss
end address to the load end address, if it is zero.
* util/mbchk.c (check_multiboot): Don't check if the load
address is greater than or equal to the load end address, if the
load end address is zero. Don't check if the load end address is
greater than the bss end address, if the bss end address is
zero. And, don't check if the load end address is less than or
equal to the entry address, if the load end address is zero.
* docs/multiboot.texi (The address fields of Multiboot header):
Added descriptions about the behavior of the boot loader when
LOAD_END_ADDR is zero and BSS_END_ADDR is zero.
From Jean-Jacques Michel <jjmichel@linbox.com>:
* stage2/boot.c (load_image): For Linux, check if DATA_LEN is
greater than MULTIBOOT_SEARCH. If that's true, read the rest
after copying data already read in BUFFER.
* stage2/boot.c (load_image): For Linux, don't check if the
length of protected mode code is greater than or equal to the
expected length minus 16. Instead, just check if no error
occurred. That was problematic, because memdisk has no protected
mode code. Also, I don't see what the magic number 16 was for.
* stage2/boot.c (linux_mem_size): New variable.
(load_image): Check a mem= option and set LINUX_MEM_SIZE to the
specified memory size, if any. Otherwise, to zero. When an
overflow is detected, use LINUX_INITRD_MAX_ADDRESS instead.
(load_initrd): If LINUX_MEM_SIZE is non-zero, use it instead of
the actual memory size.
* stage2/char_io.c (safe_parse_maxint): Use ERR_NUMBER_OVERFLOW
instead of ERR_NUMBER_PARSING, when an overflow occurs.
* stage2/common.c [!STAGE1_5] (err_list): Added
ERR_NUMBER_OVERFLOW.
* stage2/shared.h (ERR_NUMBER_OVERFLOW): New constant.
assignments received by the FSF.
Jason Thomas was added.
Erich Stefan Boleyn was moved into those who assigned their copyrights.
Gordon Matzigkeit was moved into the others, because he still wasn't listed
in fencepost.gnu.org:/gd/gnuorg/copyright.list.
The remaining problem is that we haven't had copyright assignments of
VaX#8, Heiko Schroeder and Gordon Matzigkeit. As for the first two, Erich
is now trying to contact them by proxy. So I'm just waiting.
As for Gordon.... oh, what are you doing, Gordon?! Hurry up, please!
NOTE: Even the word "mess" is not still enough to represent how wrong and
bad the Linux video support is, but I don't want to hear complaints from
Linux fanatics any more.
proski