| Age | Commit message (Collapse) | Author |
|---|
|
The CBFS stage header is part of the file data (not the header) from
CBFS's point of view, which is problematic for verification: in pre-RAM
environments, there's usually not enough scratch space in CBFS_CACHE to
load the full stage into memory, so it must be directly loaded into its
final destination. However, that destination is decided from reading the
stage header. There's no way we can verify the stage header without
loading the whole file and we can't load the file without trusting the
information in the stage header.
To solve this problem, this patch changes the CBFS stage format to move
the stage header out of the file contents and into a separate CBFS
attribute. Attributes are part of the metadata, so they have already
been verified before the file is loaded.
Since CBFS stages are generally only meant to be used by coreboot itself
and the coreboot build system builds cbfstool and all stages together in
one go, maintaining backwards-compatibility should not be necessary. An
older version of coreboot will build the old version of cbfstool and a
newer version of coreboot will build the new version of cbfstool before
using it to add stages to the final image, thus cbfstool and coreboot's
stage loader should stay in sync. This only causes problems when someone
stashes away a copy of cbfstool somewhere and later uses it to try to
extract stages from a coreboot image built from a different revision...
a debugging use-case that is hopefully rare enough that affected users
can manually deal with finding a matching version of cbfstool.
The SELF (payload) format, on the other hand, is designed to be used for
binaries outside of coreboot that may use independent build systems and
are more likely to be added with a potentially stale copy of cbfstool,
so it would be more problematic to make a similar change for SELFs. It
is not necessary for verification either, since they're usually only
used in post-RAM environments and selfload() already maps SELFs to
CBFS_CACHE before loading them to their final destination anyway (so
they can be hashed at that time).
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I8471ad7494b07599e24e82b81e507fcafbad808a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/46484
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
This patchs adds a new CBFS primitive that allows callers to pass in an
allocator function that will be called once the size of the file to load
is known, to decide on its final location. This can be useful for
loading a CBFS file straight into CBMEM, for example. The new primitive
is combined with cbfs_map() and cbfs_load() into a single underlying
function that can handle all operations, to reduce the amount of code
that needs to be duplicated (especially later when file verification is
added). Also add a new variation that allows restraining or querying the
CBFS type of a file as it is being loaded, and reorganize the
documentation/definition of all these accessors and variations in the
header file a little.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I5fe0645387c0e9053ad5c15744437940fc904392
Reviewed-on: https://review.coreboot.org/c/coreboot/+/49334
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
<endian.h> should never be included directly in commonlib files and
should instead be chain-included via <commonlib/bsd/sysincludes.h>.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: Ibc67ea97da36ec58738236ef22f961d9bbaf8574
Reviewed-on: https://review.coreboot.org/c/coreboot/+/50630
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
Attribute tags are defined as hexadecimal constants, not decimal, so it
makes more sense to print them like that in error messages as well.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I3a5a6a8c9b8d24e57633595fc47221a483d8593a
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48836
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
cbfstool has always had a CBFS_FILENAME_ALIGN that forces the filename
field to be aligned upwards to the next 16-byte boundary. This was
presumably done to align the file contents (which used to come
immediately after the filename field).
However, this hasn't really worked right ever since we introduced CBFS
attributes. Attributes come between the filename and the contents, so
what this code currently does is fill up the filename field with extra
NUL-bytes to the boundary, and then just put the attributes behind it
with whatever size they may be. The file contents don't end up with any
alignment guarantee and the filename field is just wasting space.
This patch removes the old FILENAME_ALIGN, and instead adds a new
alignment of 4 for the attributes. 4 seems like a reasonable alignment
to enforce since all existing attributes (with the exception of weird
edge cases with the padding attribute) already use sizes divisible by 4
anyway, and the common attribute header fields have a natural alignment
of 4. This means file contents will also have a minimum alignment
guarantee of 4 -- files requiring a larger guarantee can still be added
with the --alignment flag as usual.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I43f3906977094df87fdc283221d8971a6df01b53
Reviewed-on: https://review.coreboot.org/c/coreboot/+/47827
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
After the mcache is copied into CBMEM, it has *just* the right size to
fit the final tag with no room to spare. That means the test to check if
we walked over the end must be `current + sizeof(tag) <= end`, not
`current + sizeof(tag) < end`.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I25a0d774fb3294bb4d15f31f432940bfccc84af0
Reviewed-on: https://review.coreboot.org/c/coreboot/+/48277
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
|
|
This patch adds the first stage of the new CONFIG_CBFS_VERIFICATION
feature. It's not useful to end-users in this stage so it cannot be
selected in menuconfig (and should not be used other than for
development) yet. With this patch coreboot can verify the metadata hash
of the RO CBFS when it starts booting, but it does not verify individual
files yet. Likewise, verifying RW CBFSes with vboot is not yet
supported.
Verification is bootstrapped from a "metadata hash anchor" structure
that is embedded in the bootblock code and marked by a unique magic
number. This anchor contains both the CBFS metadata hash and a separate
hash for the FMAP which is required to find the primary CBFS. Both are
verified on first use in the bootblock (and halt the system on failure).
The CONFIG_TOCTOU_SAFETY option is also added for illustrative purposes
to show some paths that need to be different when full protection
against TOCTOU (time-of-check vs. time-of-use) attacks is desired. For
normal verification it is sufficient to check the FMAP and the CBFS
metadata hash only once in the bootblock -- for TOCTOU verification we
do the same, but we need to be extra careful that we do not re-read the
FMAP or any CBFS metadata in later stages. This is mostly achieved by
depending on the CBFS metadata cache and FMAP cache features, but we
allow for one edge case in case the RW CBFS metadata cache overflows
(which may happen during an RW update and could otherwise no longer be
fixed because mcache size is defined by RO code). This code is added to
demonstrate design intent but won't really matter until RW CBFS
verification can be supported.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I8930434de55eb938b042fdada9aa90218c0b5a34
Reviewed-on: https://review.coreboot.org/c/coreboot/+/41120
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
|
|
This patch reduces some code duplication in cbfstool by switching it to
use the CBFS data structure definitions in commonlib rather than its own
private copy. In addition, replace a few custom helpers related to hash
algorithms with the official vboot APIs of the same purpose.
Signed-off-by: Julius Werner <jwerner@chromium.org>
Change-Id: I22eae1bcd76d85fff17749617cfe4f1de55603f4
Reviewed-on: https://review.coreboot.org/c/coreboot/+/41117
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-by: Wim Vervoorn <wvervoorn@eltan.com>
|
|
This patch adapts cbfs_load() and cbfs_map() to use the new CBFS API
directly, rather than through cbfs_boot_locate(). For cbfs_load() this
means that attribute metadata does not need to be read twice.
Change-Id: I754cc34b1c1471129e15475aa0f1891e02439a02
Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/39305
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
This patch adds a new CBFS "mcache" (metadata cache) -- a memory buffer
that stores the headers of all CBFS files. Similar to the existing FMAP
cache, this cache should reduce the amount of SPI accesses we need to do
every boot: rather than having to re-read all CBFS headers from SPI
flash every time we're looking for a file, we can just walk the same
list in this in-memory copy and finally use it to directly access the
flash at the right position for the file data.
This patch adds the code to support the cache but doesn't enable it on
any platform. The next one will turn it on by default.
Change-Id: I5b1084bfdad1c6ab0ee1b143ed8dd796827f4c65
Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/38423
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
This patch adds a new CBFS implementation that is intended to replace
the existing commonlib/cbfs.c. The new implementation is designed to
meet a bunch of current and future goals that in aggregate make it
easier to start from scratch than to adapt the exisiting implementation:
1. Be BSD-licensed so it can evetually be shared with libpayload.
2. Allow generating/verifying a metadata hash for future CBFS per-file
verification (see [1][2]).
3. Be very careful about reading (not mmaping) all data only once, to be
suitable for eventual TOCTOU-safe verification.
4. Make it possible to efficiently implement all current and future
firmware use cases (both with and without verification).
The main primitive is the cbfs_walk() function which will traverse a
CBFS and call a callback for every file. cbfs_lookup() uses this to
implement the most common use case of finding a file so that it can be
read. A host application using this code (e.g. coreboot, libpayload,
cbfstool) will need to provide a <cbfs_glue.h> header to provide the
glue to access the respective CBFS storage backend implementation.
This patch merely adds the code, the next patch will integrate it into
coreboot.
[1]: https://www.youtube.com/watch?v=Hs_EhewBgtM
[2]: https://osfc.io/uploads/talk/paper/47/The_future_of_firmware_verification_in_coreboot.pdf
(Note: In early discussions the metadata hash was called "master hash".)
Change-Id: Ica64c1751fa37686814c0247460c399261d5814c
Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/38421
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Change-Id: Ib3db132aa649b6895f41290df049f87a9f36dc52
Signed-off-by: Elyes HAOUAS <ehaouas@noos.fr>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/44606
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Michael Niewöhner <foss@mniewoehner.de>
|
|
Certain non-Linux OSes require an include file in different
places.
Build tested on Linux, FreeBSD.
Change-Id: Icd81c2a96c608589ce2ec8f4b883fd4e584776b1
Signed-off-by: Idwer Vollering <vidwer@gmail.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/38648
Reviewed-by: Angel Pons <th3fanbus@gmail.com>
Reviewed-by: Julius Werner <jwerner@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
Add CBFS_TYPE_BOOTBLOCK to allow accessing the bootblock as a CBFS file.
BUG=N/A
TEST=build
Change-Id: Ibb03ac3a6cd8711e0402e47335bb8e110c6ef61a
Signed-off-by: Wim Vervoorn <wvervoorn@eltan.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/41824
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Reviewed-by: Frans Hendriks <fhendriks@eltan.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
|
|
This change adds a helper macro POWER_OF_2 that is useful for
calculating the requested power of 2.
Change-Id: Ie70f93b6ac175699c11cae7d8f023a52cce01e88
Signed-off-by: Furquan Shaikh <furquan@google.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/39881
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Aaron Durbin <adurbin@chromium.org>
|
|
Fix two out-of-bounds reads in lz4 decompression:
1) LZ4_decompress_generic could read one byte past the input buffer when
decoding variable length literals due to a missing bounds check. This
issue was resolved in libpayload, commonlib and cbfstool
2) ulz4fn could read up to 4 bytes past the input buffer when reading a
lz4_block_header due to a missing bounds check. This issue was resolved
in libpayload and commonlib.
Change-Id: I5afdf7e1d43ecdb06c7b288be46813c1017569fc
Signed-off-by: Alex Rebert <alexandre.rebert@gmail.com>
Found-by: Mayhem
Reviewed-on: https://review.coreboot.org/c/coreboot/+/39174
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
|
|
This patch creates a new commonlib/bsd subdirectory with a similar
purpose to the existing commonlib, with the difference that all files
under this subdirectory shall be licensed under the BSD-3-Clause license
(or compatible permissive license). The goal is to allow more code to be
shared with libpayload in the future.
Initially, I'm going to move a few files there that have already been
BSD-licensed in the existing commonlib. I am also exracting most
contents of the often-needed <commonlib/helpers.h> as long as they have
either been written by me (and are hereby relicensed) or have an
existing equivalent in BSD-licensed libpayload code. I am also
relicensing <commonlib/compression.h> (written by me) and
<commonlib/compiler.h> (same stuff exists in libpayload).
Finally, I am extracting the cb_err error code definitions from
<types.h> into a new BSD-licensed header so that future commonlib/bsd
code can build upon a common set of error values. I am making the
assumption here that the enum constants and the half-sentence fragments
of documentation next to them by themselves do not meet the threshold of
copyrightability.
Change-Id: I316cea70930f131e8e93d4218542ddb5ae4b63a2
Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/38420
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Rudolph <siro@das-labor.org>
|
