This fixes a double free crash that occurs when a call to
cbfs_image_from_file() fails in cbfs_extract() and falls though to
cbfs_image_delete() with a NULL-pointer.
To reproduce the crash pass the following arguments where the files
passed, in fact, do not exist. As follows:
./cbfstool build/coreboot.rom extract -n config -f /tmp/config.txt
Change-Id: I2213ff175d0703705a0ec10271b30bb26b6f8d0a
Signed-off-by: Edward O'Callaghan <eocallaghan@alterapraxis.com>
Reviewed-on: http://review.coreboot.org/5353
Tested-by: build bot (Jenkins)
Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
elfparsing.h serves as the header to working with the elf
parser. Additionally, only include what is needed by the other
files. Many had no reason to be including elf.h aside from fixing
compilation problems when including cbfs.h.
Change-Id: I9eb5f09f3122aa18beeca52d2e4dc2102d70fb9d
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: http://review.coreboot.org/5370
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Marc Jones <marc.jones@se-eng.com>
Rather than using [hn]to[nh] whenever accessing a member of the CBFS
header, deserialize the header when opening the CBFS image. The header
is no longer a pointer inside the CBFS buffer, but a separate struct,
a copy of the original header in a host-friendly format. This kills
more of the ntohl usage.
Change-Id: I5f8a5818b9d5a2d1152b1906249c4a5847d02bac
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Reviewed-on: http://review.coreboot.org/5121
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
This change adds a header serialization function. Programmers can thus just
set up a header as needed, without worrying about forgetting if and how to
use the [hn]to[hn]* functions.
In the long term, we will work to remove swab.h, i.e. we need to get to the
point where programmers don't have to try to remember [hn]to[nh]* and where
it goes. To date, even the best programmers we have have made an error with
those functions, and those errors have persisted for 6 or 7 years now. It's
very easy to make that mistake.
BUG=None
TEST=Build a peppy image and verify that it's bit for bit the same. All
chromebooks use this code and build and boot correctly.
BRANCH=None
Change-Id: I0f9b8e7cac5f52d0ea330ba948650fa0803aa0d5
Signed-off-by: Ronald G. Minnich <rminnich@google.com>
Reviewed-on: https://chromium-review.googlesource.com/181552
Reviewed-by: Ronald Minnich <rminnich@chromium.org>
Commit-Queue: Ronald Minnich <rminnich@chromium.org>
Tested-by: Ronald Minnich <rminnich@chromium.org>
Reviewed-on: http://review.coreboot.org/5100
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
This completes the improvements to the ELF file parsing code. We can
now parse section headers too, across all 4 combinations of word size
and endianness. I had hoped to completely remove the use of htonl
until I found it in cbfs_image.c. That's a battle for another day.
There's now a handy macro to create magic numbers in host byte order.
I'm using it for all the PAYLOAD_SEGMENT_* constants and maybe
we can use it for the others too, but this is sensitive code and
I'd rather change one thing at a time.
To maximize the ease of use for users, elf parsing is accomplished with
just one function:
int
elf_headers(const struct buffer *pinput,
Elf64_Ehdr *ehdr,
Elf64_Phdr **pphdr,
Elf64_Shdr **pshdr)
which requires the ehdr and pphdr pointers to be non-NULL, but allows
the pshdr to be NULL. If pshdr is NULL, the code will not try to read
in section headers.
To satisfy our powerful scripts, I had to remove the ^M from an unrelated
microcode file.
BUG=None
TEST=Build a peppy image (known to boot) with old and new versions and verify they are bit-for-bit the same. This was also fully tested across all chromebooks for building and booting and running chromeos.
BRANCH=None
Change-Id: I54dad887d922428b6175fdb6a9cdfadd8a6bb889
Signed-off-by: Ronald G. Minnich <rminnich@google.com>
Reviewed-on: https://chromium-review.googlesource.com/181272
Reviewed-by: Ronald Minnich <rminnich@chromium.org>
Commit-Queue: Ronald Minnich <rminnich@chromium.org>
Tested-by: Ronald Minnich <rminnich@chromium.org>
Signed-off-by: Ronald G. Minnich <rminnich@google.com>
Reviewed-on: http://review.coreboot.org/5098
Tested-by: build bot (Jenkins)
Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Add XDR functions and use them to convert the ELF headers
to native headers, using the Elf64 structs to ensure we accomodate
all word sizes. Also, use these XDR functions for output.
This may seem overly complex but it turned out to be much the easiest
way to do this. Note that the basic elf parsing function
in cbfs-mkstage.c now works over all ELF files, for all architectures,
endian, and word size combinations. At the same time, the basic elf
parsing in cbfs-mkstage.c is a loop that has no architecture-specific
conditionals.
Add -g to the LDFLAGS while we're here. It's on the CFLAGS so there is
no harm done.
This code has been tested on all chromebooks that use coreboot to date.
I added most of the extra checks from ChromeOS and they triggered a
lot of warnings, hence the other changes. I had to take -Wshadow back
out due to the many errors it triggers in LZMA.
BUG=None
TEST=Build and boot for Peppy; works fine. Build and boot for nyan,
works fine. Build for qemu targets and armv8 targets.
BRANCH=None
Change-Id: I5a4cee9854799189115ac701e22efc406a8d902f
Signed-off-by: Ronald G. Minnich <rminnich@google.com>
Reviewed-on: https://chromium-review.googlesource.com/178606
Reviewed-by: Ronald Minnich <rminnich@chromium.org>
Commit-Queue: Ronald Minnich <rminnich@chromium.org>
Tested-by: Ronald Minnich <rminnich@chromium.org>
Reviewed-on: http://review.coreboot.org/4817
Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Tested-by: build bot (Jenkins)
Reviewed-by: Ronald G. Minnich <rminnich@gmail.com>
After removing a file sandwiched between two other files, that file
could no longer be re-added at the same location. cbfstool tried to
add the file, and a new "empty" entry, which, together, would no
longer fit, so it continued checking for the next available space.
Change the behavior to add the file if there is enough space for the
file alone, then only add the "empty" entry if there is enough space
for it.
Change-Id: Iad3897dd28cf12f12ae877cfd83e1990fa7d2f0f
Signed-off-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: http://review.coreboot.org/4772
Tested-by: build bot (Jenkins)
Reviewed-by: Aaron Durbin <adurbin@google.com>
- The read-only structures are const now
- cosmetic fixes
- put { on a new line for functions
- move code after structures
Change-Id: Ib9131b80242b91bd5105feaebdf8306a844da1cc
Signed-off-by: Stefan Reinauer <reinauer@google.com>
Reviewed-on: http://review.coreboot.org/2922
Reviewed-by: Aaron Durbin <adurbin@google.com>
Tested-by: build bot (Jenkins)
When calculating initial CBFS empty entry space, the size of header itself must
be not included (with the reserved space for entry name). This is a regression
of the old cbfstool size bug.
Before this fix, in build process we see:
OBJCOPY cbfs/fallback/romstage_null.bin
W: CBFS image was created with old cbfstool with size bug.
Fixing size in last entry...
And checking the output binary:
cbfstool build/coreboot.pre1 print -v -v
DEBUG: read_cbfs_image: build/coreboot.pre1 (262144 bytes)
DEBUG: x86sig: 0xfffffd30, offset: 0x3fd30
W: CBFS image was created with old cbfstool with size bug.
Fixing size in last entry...
DEBUG: Last entry has been changed from 0x3fd40 to 0x3fd00.
coreboot.pre1: 256 kB, bootblksz 688, romsize 262144, offset 0x0 align: 64
Name Offset Type Size
(empty) 0x0 null 261296
DEBUG: cbfs_file=0x0, offset=0x28, content_address=0x28+0x3fcb0
After this fix, no more alerts in build process.
Verified to build successfully on x86/qemu and arm/snow configurations.
Change-Id: I35c96f4c10a41bae671148a0e08988fa3bf6b7d3
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2731
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
cbfstool usage change:
"-a" for "cbfstool locate" can specify base address alignment.
To support putting a blob in aligned location (ex, microcode needs to be aligned
in 0x10), alignment (-a) is implemented into "locate" command.
Verified by manually testing a file (324 bytes) with alignment=0x10:
cbfstool coreboot.rom locate -f test -n test -a 0x10
# output: 0x71fdd0
cbfstool coreboot.rom add -f test -n test -t raw -b 0x71fdd0
cbfstool coreboot.rom print -v -v
# output: test 0x71fd80 raw 324
# output: cbfs_file=0x71fd80, offset=0x50, content_address=0x71fdd0+0x144
Also verified to be compatible with old behavior by building i386/axus/tc320
(with page limitation 0x40000):
cbfstool coreboot.rom locate -f romstage_null.bin -n romstage -P 0x40000
# output: 0x44
cbfstool coreboot.rom locate -f x.bin -n romstage -P 0x40000 -a 0x30
# output: 0x60
Change-Id: I78b549fe6097ce5cb6162b09f064853827069637
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2824
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins)
On platforms with CBFS data filling end of ROM image without bootblock in the
end (ex, ARM), calculation of "next valid entry" may exceed ROM image buffer in
memory and raise segmentation fault when we try to compare its magic value.
To fix this, always check if the entry address is inside ROM image buffer.
Verified to build and boot successfully on qemu/x86 and armv7/snow.
Change-Id: I117d6767a5403be636eea2b23be1dcf2e1c88839
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2330
Tested-by: build bot (Jenkins)
Reviewed-by: David Hendricks <dhendrix@chromium.org>
For arm/snow, current bootblock is larger than previously assigned CBFS offset
and will fail to boot. To prevent this happening again in future, cbfstool now
checks if CBFS will overlap bootblock.
A sample error message:
E: Bootblock (0x0+0x71d4) overlap CBFS data (0x5000)
E: Failed to create build/coreboot.pre1.tmp.
arm/snow offset is also enlarged and moved to Kconfig variable.
Change-Id: I4556aef27ff716556040312ae8ccb78078abc82d
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2295
Tested-by: build bot (Jenkins)
Reviewed-by: David Hendricks <dhendrix@chromium.org>
The "add" command is compatible with all legacy usage. Also, to support
platforms without top-aligned address, all address-type params (-b, -H, -l) can
now be ROM offset (address < 0x8000000) or x86 top-aligned address (address >
0x80000000).
Example:
cbfstool coreboot.rom add -f config -n config -t raw -b 0x2000
cbfstool coreboot.rom add -f stage -n newstage -b 0xffffd1c0
Verified boot-able on both ARM(snow) and x86(QEMU) system.
Change-Id: I485e4e88b5e269494a4b138e0a83f793ffc5a084
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2216
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Usage Changes: To support platforms with different memory layout, "create" takes
two extra optional parameters:
"-b": base address (or offset) for bootblock. When omitted, put bootblock in
end of ROM (x86 style).
"-H": header offset. When omitted, put header right before bootblock,
and update a top-aligned virtual address reference in end of ROM.
Example: (can be found in ARM MAkefile):
cbfstool coreboot.rom create -m armv7 -s 4096K -B bootblock.bin \
-a 64 -b 0x0000 -H 0x2040 -o 0x5000
Verified to boot on ARM (Snow) and X86 (QEMU).
Change-Id: Ida2a9e32f9a459787b577db5e6581550d9d7017b
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2214
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
To support platforms without top-aligned address mapping like ARM, "locate"
command now outputs platform independent ROM offset by default. To retrieve x86
style top-aligned virtual address, add "-T".
To test:
cbfstool coreboot.rom locate -f stage -n stage -a 0x100000 -T
# Example output: 0xffffdc10
Change-Id: I474703c4197b36524b75407a91faab1194edc64d
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2213
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Old cbfstool may produce CBFS image with calculation error in size of last empty
entry, and then corrupts master header data when you really use every bit in
last entry. This fix will correct free space size when you load ROM images with
cbfs_image_from_file.
Change-Id: I2ada319728ef69ab9296ae446c77d37e05d05fce
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2211
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
To delete a component (file) from existing CBFS ROM image.
To test:
cbfstool coreboot.rom remove -n fallback/romstage
# and compare with old cbfstool output result.
Change-Id: If39ef9be0b34d8e3df77afb6c9f944e02f08bc4e
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2208
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Change the "extract" command to use cbfs_export_entry API. Nothing changed in
its usage.
To verify, run "cbfstool coreboot.rom extract -f blah -n blah" and check if the
raw type file is correctly extracted.
Change-Id: I1ed280d47a2224a9d1213709f6b459b403ce5055
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2207
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Process CBFS ROM image by new cbfs_image API.
To verify, run "cbfstool coreboot.rom print -v" and compare with old cbfstool.
Change-Id: I3a5a9ef176596d825e6cdba28a8ad732f69f5600
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2206
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Current cbfstool implementation is relying on global variables to pass processed
data, and the calculation of address is based on x86 architecture (ex, always
assuming 0x0000 as invalid address), not easy to be used on platforms without
top-aligned memory mapping. This CL is a first step to start a new cbfstool
without global variables, and to prevent assuming memory layout in x86 mode.
The first published APIs are for reading and writing existing CBFS ROM image
files (and to find file entries in a ROM file).
Read cbfs_image.h for detail usage of each API function.
Change-Id: I28c737c8f290e51332119188248ac9e28042024c
Signed-off-by: Hung-Te Lin <hungte@chromium.org>
Reviewed-on: http://review.coreboot.org/2194
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Edward O'Callaghan