Now to start testing.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@810 f3766cd6-281f-0410-b1cd-43a5c92072e9
It also fixes lzma compression in lar to fix the silent memory
corruption that was possible when files didn't compress well.
It adds some comments to both files and the file that calls them.
Signed-off-by: Myles Watson <mylesgw@gmail.com>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@658 f3766cd6-281f-0410-b1cd-43a5c92072e9
LAR archive, the LAR utility will segfault. This is reproduced easily by
zerofilling the LAR, then adding anything to it.
Looking at the code, the reason is obvious:
lar_empty_offset() can return an error code (-1). None of the callers
check for an error code, they simply assume the return value is valid.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Peter Stuge <peter@stuge.se>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@647 f3766cd6-281f-0410-b1cd-43a5c92072e9
makes use of functions that were already defined. It also adds greedy name
matching for listing and extracting archives, which allows recursive descent
into the lar directory structure.
changes file-by-file:
util/lar/lar.c:
add more options to the usage message
use get_larsize() instead of using larsize
rearrange errors from parsing args to be more correct
util/lar/stream.c:
change elfname size to MAX_PATHLEN instead of 64
make file_in_list greedy with filename matches
change total_size calculation to include file names
change lar_add_entry to use header_len function instead of reinventing
Signed-off-by: Myles Watson <mylesgw@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@632 f3766cd6-281f-0410-b1cd-43a5c92072e9
file struct for pathname and compression, so that directories can be correctly
recursed.
file-by-file:
util/lar/lar.c:
make error messages more verbose
pass a pointer to the file structure instead of the name
parse the name here with lar_process_name
util/lar/lib.c:
change handle_directory to use a path name and respect nocompress
change add_files to use pre-processed names
use sensible defaults for new file members when listing or extracting
free pathname if allocated
util/lar/lib.h:
add new members to struct file
change prototypes of add_files and lar_add_file
util/lar/stream.c:
change lar_add_file to use pathname and compression from struct file
Signed-off-by: Myles Watson <mylesgw@gmail.com>
Acked-by: Peter Stuge <peter@stuge.se>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@623 f3766cd6-281f-0410-b1cd-43a5c92072e9
file is not found instead of seg faulting.
test with:
lar -a coreboot.rom nonexistant_file.bin
Signed-off-by: Myles Watson <mylesgw@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@615 f3766cd6-281f-0410-b1cd-43a5c92072e9
number during boot.
Convert process_file() to use enum compalgo instead of hardcoded
"1","2","3" and change the control structure from a series of if()
statements to a switch() statement.
Uppercasing enum compalgo also found a name clash between NONE as
compression algo and NONE as operation mode of util/lar.
Compile and boot tested on Qemu.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@606 f3766cd6-281f-0410-b1cd-43a5c92072e9
> Alix1c won't boot with the zero decompress code.
> I think the code is using the wrong address on decompress.
Indeed, r601 broke all targets, you were just lucky that qemu didn't
explode as well.
It's the seemingly easy patches which break booting. With your hint, I
found the bug. Myles made a small, but important mistake with the memset
for the "zeroes" decompression.
The memset zeroed the archive instead of the destination. No wonder it
did explode.
This patch fixes it and also reverts the emergency commit r604 because
that one is no longer necessary.
Ron tested on the Alix1c, boots fine, ethernet and IDE working.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@605 f3766cd6-281f-0410-b1cd-43a5c92072e9
it stands, is causing coreboot on the alix1c to hang. I don't know why.
But this patch is the difference between works and hangs. I think
the memset in the decompress is not using the right start value? Not sure.
This simple change allows us to fix the zero decompress later, when
we figure it out. I just lost a few hours to this and I want to make
sure the repo works before I go to bed.
This fix tested on alix1c. Alix1c now works fine.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@604 f3766cd6-281f-0410-b1cd-43a5c92072e9
for this is that currently, if you select no compression, the bss
segment of filo takes up 153K with just zeroes. With this patch, it
always takes up a lar header + 1 byte. I left the one byte so that
the checksum wouldn't be broken.
This patch could have taken out the calloc in the compression area,
but since it only uses compile-time memory, I decided to keep this
simple.
Myles
Signed-off-by: Myles Watson <myles@pel.cs.byu.edu>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@601 f3766cd6-281f-0410-b1cd-43a5c92072e9
buffer overflow checks. It exits with an error instead of
overflowing.
Signed-off-by: Myles Watson <myles@pel.cs.byu.edu>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@594 f3766cd6-281f-0410-b1cd-43a5c92072e9
architecture. The function prototypes had a size mismatch, which overwrote
other things on the stack. Now the prototypes use int for lengths.
Signed-off-by: Myles Watson <myles@pel.cs.byu.edu>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@592 f3766cd6-281f-0410-b1cd-43a5c92072e9
I did it because I wanted to know the differences between preparsing
elf files and compressing the pieces.
Signed-off-by: Myles Watson <myles@pel.cs.byu.edu>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@588 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Note: the following discussion applies to this patch.
This patch is basically limited. It is mostly useful for newly-created
LARs.
On Feb 9, 2008 1:06 PM, Carl-Daniel Hailfinger
<c-d.hailfinger.devel.2006@gmx.net> wrote:
> If I read the code correctly, the command will try to fill the first
> occurrence of free space in the lar, not the biggest one. That means
> we
> might have to call it multiple times.
hmm. I obviously did not think of this.
> And a general problem with the code: If it finds the offset of the
> first
> empty area, it checks whether the area is big enough. If not, it does
> not search for another empty area, but returns instead with an error.
Actually, I think as written it is broken for the general case of lots
of empty spots. It should just look for the room left from start of
empty space to next LAR header. It's just that I have never had a LAR
that has more than one block of empty space, and that empty space is
always at the gap before the bootblock.
How about this. I'd like to commit now, and then fix this later, as I
really need this code in there for speed and it works fine with all
"fresh" images built with v3.
> Should we really call this "zerofill"?
"emptyfill"? I'll let it go for now but you make good point.
> Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
OK, this is clearly a work in progress but even in this form it is so
useful I'm going to commit it with a note, "This needs more work", and
with luck someone else will improve it once we need it improved -- the
basic structure is there.
Index: util/lar/stream.c
Add hlen (header len) and maxsize (max size left in lar) functions.
Index: util/lar/lib.h
Add function prototypes and new ZEROFILL enum.
Index: util/lar/lar.c
Add -z option and command support.
git-svn-id: svn://coreboot.org/repository/coreboot-v3@583 f3766cd6-281f-0410-b1cd-43a5c92072e9
This is not terribly clean but it works.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Tested with a qemu boot to work correctly.
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@580 f3766cd6-281f-0410-b1cd-43a5c92072e9
- I left LB_TAG_ intact because they are used by the payloads
- file renames are still missing. see next commit
- some lb_ renames might be missing. feel free to provide patches.
Signed-off-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/coreboot-v3@564 f3766cd6-281f-0410-b1cd-43a5c92072e9
The message appears if a file has not been found in the LAR archive and
is triggered even by lookup routines. Normal capitalization helps
reducing the frightening effect of the message.
Correct a few typos in other areas of the LAR code as well.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@538 f3766cd6-281f-0410-b1cd-43a5c92072e9
number from ELF as segment number in the file. This works nicely when
there are no skips (e.g. not PT_LOAD segments, which are discarded).
If one segment is skipped, we get a bump:
normal/payload0/segment0 (27288 bytes, lzma compressed to 14506 bytes @0x64c0)
normal/payload0/segment2 (211136 bytes, lzma compressed to 70905 bytes @0x9dc0)
The LAR loader wont load segment2, and in this particular case, grub2-lb
will only boot into rescue mode (segment0 contains it).
Attached patch adds a counter for segment number in the LAR utility to
solve this bug:
normal/payload0/segment0 (27288 bytes, lzma compressed to 14506 bytes @0x64c0)
normal/payload0/segment1 (211136 bytes, lzma compressed to 70905 bytes @0x9dc0)
Also the eagle eyed can see that I merged in Uwe's multiple-payload
patch into current stage1, which includes the segment support. And this
means that grub2-lb without any hacks works when loaded from LAR segments.
Signed-off-by: Alex Beregszaszi <alex@rtfs.hu>
Acked-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@498 f3766cd6-281f-0410-b1cd-43a5c92072e9
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@494 f3766cd6-281f-0410-b1cd-43a5c92072e9
The system will default to old behaviour. See Kconfig in the root.
I still wish to kill ELF mode very soon, however.
LAR is a very capable format. With two simple extensions, we can use
LAR to replace all that we are using ELF for now. This change can
really make life better:
1. we can use streaming decompress instead of the current "uncompress
elf to memory and then copy segments" approach. So we can get rid of
THIS hardcode:
#define UNCOMPRESS_AREA (0x400000)
2. A simple lar l can show ALL segments, including payload segments
3. It's really easy to see where things will go in memory, and catch problems
4. We can figure out an ELF input file is bogus BEFORE we flash, not
AFTER we flash and try to boot it
5. did I mention streaming decompress?
6. We no longer have to worry about where we decompress the elf in
memory (this problem was causing trouble when the payload was a linux
kernel -- it was so big)
7. Since we have a load address, we can create this lar entry:
normal/cmdline
and specify that it be loaded at a place where linux will find it as
the cmdline.
8. The decision on whether to XIP can be made in the LAR entry, not in
hardcode. For example, if initram needs to be XIP, set the load
address to 0xffffffff. Done.
The change is simple. Add a load address and entry point to the lar
header. Extend the lar tool to parse the elf file and create multiple
lar segments. It looks like this:
normal/payload0 (33192 bytes, lzma compressed to 18088 bytes @0x38
load @0x100000, entry 0x105258)
normal/payload1 (72 bytes, lzma compressed to 47 bytes @0x4718 load
@0x1225a0, entry 0x105258)
normal/option_table (932 bytes @0x4798 load @0, entry 0)
normal/stage2 (33308 bytes, lzma compressed to 15474 bytes @0x4b78
load @0, entry 0)
normal/initram (4208 bytes @0x8828 load @0, entry 0)
linuxbios.bootblock (16384 bytes @0xfc000 load @0, entry 0)
note that the payload is now payload/segment0, payload/segment1, etc. I've extended
linuxbios to look for these. Note that you can now see all the things
that get loaded ;they're no longer hidden in an ELF header somewhere.
Elf failures are gone!
Note that I've left legacy elf support in, for now, but recommend we
get rid of it as soon as possible.
patch attached. This is a first pass. lar.c needs some refactoring but
I want to get the cmdline going. You can now have a linux payload and
it will uncompress with no problems.
This has been tested with filo and BOCHS.
This patch includes ONLY the lar changes, the other changes are next.
Signed-off-by: Ronald G. Minnich <rminnich@gmail.com>
Acked-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@481 f3766cd6-281f-0410-b1cd-43a5c92072e9
and ends at the end of a temporary buffer, which doesn't look
right to me (and segfaults if between those two there's an unmapped
region, as happened here).
Signed-off-by: Patrick Georgi <patrick@georgi-clan.de>
Acked-by: Stefan Reinauer <stepan@coresystems.de>
Acked-by: Uwe Hermann <uwe@hermann-uwe.de>
git-svn-id: svn://coreboot.org/repository/LinuxBIOSv3@480 f3766cd6-281f-0410-b1cd-43a5c92072e9
