This cleans up the vboot functions which handle display so they don't
need to pass it around. Eventually, it'll be absorbed by vb2_context.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; build_packages --board=reef chromeos-firmware; boot reef
Change-Id: I58169dfd37abe657f9b9aa339cc72ffa398329e0
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/414288
Reviewed-by: Shelley Chen <shchen@chromium.org>
Passing the vb2 context around allows using more of the vb2 functions in
future changes, and prepares for a future where we directly use the
context as it was set up in firmware verification.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge
Change-Id: I8efa606dbdec5d195b66eb899e76fdc84337ad36
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/404997
Reviewed-by: Shelley Chen <shchen@chromium.org>
Now that LoadKernel() uses a stream API for its partition data, it
doesn't care about those fields. They're blindly passed to
cgptlib_internal, which does similar checks in CheckParameters() and
CheckHeader(). So, don't duplicate the checks.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge
Change-Id: I72375496e5df7b7c17df25d358f2555fe41fe520
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/407053
Reviewed-by: Daisuke Nojiri <dnojiri@chromium.org>
LoadKernel() was a big function which did everything from looping over
partitions on a drive to loading the data within them to calling the
low-level verification functions on that data. Split it apart into more
manageable chunks. This also reduces indentation of the inner parts of
the code, whic increases readability.
No outwardly-visible functionality changes.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge
Change-Id: Iea79e70163f5d9f1a9d0d897e4a9bacc925a742d
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/404919
Reviewed-by: Daisuke Nojiri <dnojiri@chromium.org>
Previously, vb2_unpack_key() actually unpacked a key buffer. Callers
that had a vb2_packed_key had to typecast it back to a uint8_t buffer to
unpack it. Rename vb2_unpack_key() to vb2_unpack_key_buffer(), and make
vb2_unpack_key() unpack a vb2_packed_key.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge;
emerge-samus and boot it
Change-Id: I9ee38a819c59cc58a72ead78cf5ddf3d0f301ae7
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/400906
Reviewed-by: Daisuke Nojiri <dnojiri@chromium.org>
Originally, vboot1 code used VbExMalloc() and VbExFree() since it needed
to talk to EFI firmware that didn't have standard malloc() and free().
Now, coreboot and depthcharge implement them as wrappers around those
standard calls. vboot2 code already calls them directly, so let vboot1
code do that too.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge
Change-Id: I49ad0e32e38d278dc3589bfaf494bcf0e4b0a4bd
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/400905
At this point, all that's left are a few constants in the cryptolib
header files, and they're only used by host-side code. So move them to
a host-side header file and get rid of cryptolib.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge
Change-Id: I2235f0e84e13fef313afe54e749b73744b157884
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/400903
Reviewed-by: Daisuke Nojiri <dnojiri@chromium.org>
Originally, we didn't trust the firmware to provide these functions from
a standard library. Now, with coreboot, we do.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; emerge-kevin coreboot depthcharge
Change-Id: I4e624c40085f2b665275a38624340b2f6aabcf11
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/399120
Reviewed-by: Daisuke Nojiri <dnojiri@chromium.org>
This change replaces all calls to the old vboot1 SHA library with their
vboot2 equivalents.
This is the first in a long series of changes to move the core vboot kernel
verification into vb2, and the control/display loop out to depthcharge.
BUG=chromium:611535
BRANCH=none
TEST=make runtests; build samus firmware and boot it
Change-Id: I31986eb766176c0e39a192c5ce15730471c3cf94
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/344342
Tested-by: Daisuke Nojiri <dnojiri@chromium.org>
Reviewed-by: Daisuke Nojiri <dnojiri@chromium.org>
This adds RW firmware support for the optional firmware management
parameters TPM space.
System-level tests require CL:339262 to add cryptohome support.
BUG=chromium:601492
BRANCH=baytrail and newer platforms
TEST=make -j runtests
Or better, COV=1 make, and then make sure all new code is covered.
Change-Id: Ifaf644c80809552d5961615be6017c2a332a034b
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/339234
When a read fails in getting the GPT, just zero the contents of the
buffer and carry on.
Some testing changes are required for this. When a read of the GPT
fails, it is no longer fatal, so tests of that have been adjusted.
Tests have been improved to show that the GPT is automatically
repaired when a read error occurs.
There was one test which checked that a zero-sized disk would fail
to load a kernel, but it was surrounded by a number of mocked
functions which normally do that error checking, and it amounted
to the same test as read failure; that test was deleted.
BUG=chrome-os-partner:35440
TEST=vboot tests pass
BRANCH=none
Change-Id: I0c05813e7492920433733947d3fb74a7e4aa66f2
Signed-off-by: Dan Ehrenberg <dehrenberg@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/266882
Reviewed-by: Randall Spangler <rspangler@chromium.org>
This patch makes some small modifications to cgpt and vboot to
root out the last vestigates of a fixed 128-entry GPT:
- Get rid of the TOTAL_ENTRIES_SIZE constant and all users.
- Reduce MAX_NUMBER_OF_ENTRIES to 128 (which is what the GPT
spec specifies) so that this can be used for things like memory
allocations without additional overhead.
- Base the amount of GPT read/written on the number of entries
specified in the GPT header on disk/flash.
BUG=chromium:433433
TEST=make runalltests
TEST=Modified fmap to make an 8k RW_GPT, wrote a GPT with cgpt, then
rebooted and found that the GPT was correctly read after restarting
and the appropriate mtd partitions were present.
BRANCH=none
Change-Id: I45317377da20259caf04a7a4fa077a892b03c45f
Signed-off-by: Dan Ehrenberg <dehrenberg@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/238245
Reviewed-by: Randall Spangler <rspangler@chromium.org>
This patch reinstates the external GPT support which was previously
committed and reverted. Improvements since last time include:
- Cleaned-up internal interface based on code review
- Function correctly on legacy bootloaders (e.g., depthcharge before
NAND-related patches are added)
- Better comments
- Treat new field values = 0 -> not use new feature
- Tests are added to ensure external GPT flag is passed down properly
The original commit had change-id
I5a77e417aea8ee9442d18c200d1b073aa5375ecf
Its commit message is reproduced below, and then an additional test.
----
To support an external GPT, disks have two new attributes:
- A binary flag indicating whether the GPT is in the same address
space as the payloads or a separate one.
- The number of sectors of the streaming portion of storage, as
opposed to the portion containing the GPT.
These have been added elsewhere to GptData (in cgptlib) and BlockDev
(in depthcharge). This patch adds the plumbing between those, including
in the DiskInfo interface between the firmware and vboot.
BUG=chromium:425677
BRANCH=none
TEST=Interactively wrote the GPT with cgpt and observed the following
boot with depthcharge to read the GPT from SPI and then read from
the proper locations in NAND flash.
TEST=make runalltests passes.
TEST=boots from USB with depthcharge from HEAD.
Change-Id: Ia7956517a7b9da0301f01fac5a10204f6d78cf4f
Signed-off-by: Dan Ehrenberg <dehrenberg@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/234640
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
The standard says that entries table must be at least 16384 bytes. On
some of our devices, the NOR section is only 8 KiB and used to store
both primary and secondary tables. On this device, we can only store 24
entries.
Therefore, this CL adds support for non-standard entry table. It adjusts
the MIN_NUMBER_OF_ENTRIES to 16, and replaces GPT_ENTRIES_SECTORS with
CalculateEntriesSectors.
BUG=chromium:441812
BRANCH=none
TEST=unittest
Change-Id: I6b85b35ce5612c7abb22142f8252bd0d45b676c5
Reviewed-on: https://chromium-review.googlesource.com/234996
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Commit-Queue: Nam Nguyen <namnguyen@chromium.org>
Tested-by: Nam Nguyen <namnguyen@chromium.org>
- Rename drive_sectors to streaming_drive_sectors, to contrast with
gpt_drive_sectors
- Replace stored_on_device field with flags field for future
extensibility
BUG=chromium:433433
TEST=make runtests
BRANCH=none
Change-Id: I785a3b735b8eb96f647a334659329db3ee43eb80
Signed-off-by: Dan Ehrenberg <dehrenberg@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/234283
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
This reverts commit 5040a945df.
This patch breaks booting any image (both fixed and
removable) on Veyron_Pinky (and presumably every other
non-NAND board?). By the power vested in me through the
office of ChromeOS tree sheriff (well, five hours early
but whatever) it is hereby reverted!
BUG=chromium:425677
BRANCH=none
TEST=Can successfully boot on Veyron_Pinky again.
Change-Id: I9323a3d5e34491337fc7eb09dd00d845ac42997d
Reviewed-on: https://chromium-review.googlesource.com/229963
Reviewed-by: Julius Werner <jwerner@chromium.org>
Commit-Queue: Julius Werner <jwerner@chromium.org>
Tested-by: Julius Werner <jwerner@chromium.org>
To support an external GPT, disks have two new attributes:
- A binary flag indicating whether the GPT is in the same address
space as the payloads or a separate one.
- The number of sectors of the streaming portion of storage, as
opposed to the portion containing the GPT.
These have been added elsewhere to GptData (in cgptlib) and BlockDev
(in depthcharge). This patch adds the plumbing between those, including
in the DiskInfo interface between the firmware and vboot.
BUG=chromium:425677
BRANCH=none
TEST=Interactively wrote the GPT with cgpt and observed the following
boot with depthcharge to read the GPT from SPI and then read from
the proper locations in NAND flash.
make runalltests passes.
Signed-off-by: Dan Ehrenberg <dehrenberg@chromium.org>
Change-Id: I5a77e417aea8ee9442d18c200d1b073aa5375ecf
Reviewed-on: https://chromium-review.googlesource.com/228943
Reviewed-by: Randall Spangler <rspangler@chromium.org>
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Previously, my_lba of the secondary GPT was recorded as if that
GPT was written at the end of the device. This patch tweaks
my_lba to report where it is in the random-access GPT address space,
namely at the end of that space.
TEST=Compiled it into the firmware and observed the firmware to update
the my_lba field of the secondary GPT.
BRANCH=none
BUG=chromium:425677
Signed-off-by: Dan Ehrenberg <dehrenberg@chromium.org>
Change-Id: I88791fb2cff1086351ca8a3adeef675c4a88cc9a
Reviewed-on: https://chromium-review.googlesource.com/228942
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Commit-Queue: Daniel Ehrenberg <dehrenberg@chromium.org>
Tested-by: Daniel Ehrenberg <dehrenberg@chromium.org>
This CL allows the GPT headers and partition entry arrays to be stored
in a NOR flash device. Instead of treating both the NOR and NAND devices
as one (in a sandwich way), this CL writes and reads the GPT structs
independently of the actual device that houses the partitions.
Therefore, the first usable LBA of the partitions will be at 0, and the
last usable LBA is at the end of the NAND.
+------------------------+
| NOR houses GPT structs |
+------------------------+
|
0 | Index into
v v
+------------------------+
| NAND houses partitions |
+------------------------+
Note that the "my_lba", "alternate_lba", "entries_lba" in the GPT headers
are no longer meaningful.
Consumers of cgptlib will have to set "stored_on_device" to either
GPT_STORED_ON_DEVICE or GPT_STORED_OFF_DEVICE, and "gpt_drive_sectors"
to the number of 512-byte sectors available to store GPT structs.
The NOR read and write operations are done by "flashrom".
BUG=chromium:425677
BRANCH=none
TEST=unittest
TEST=build with DEBUG, cgpt create/add/show on a stumpy-moblab
Change-Id: I083b3c94da3b0bb3da1a7b10c6969774080a2afd
Reviewed-on: https://chromium-review.googlesource.com/226800
Reviewed-by: Nam Nguyen <namnguyen@chromium.org>
Commit-Queue: Nam Nguyen <namnguyen@chromium.org>
Tested-by: Nam Nguyen <namnguyen@chromium.org>
This is necessary to support reading the kernel from raw NAND flash,
where the driver may need to skip over bad sectors, and absolute
sector addressing is thus not practical.
The impact is relatively minor. Vboot only did two reads per kernel
anyway, one for the first 64KB of the partition and a second for the
rest of the kernel data.
Firmware which uses vboot will need to implement the streaming APIs.
Or, as a really easy workaround, just copy the implementation from
firmware/stub/vboot_api_stub_stream.c, which translates from the new
streaming API to the old sector-based disk API.
BUG=chromium:403432
BRANCH=none
TEST=make runtests; passes.
CQ-DEPEND=CL:221992, CL:222885, CL:222945
Change-Id: I7437b489650c95c09ac68b67d4d86f9e15c2fa73
Signed-off-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/222410
Reviewed-by: Daniel Ehrenberg <dehrenberg@chromium.org>
In AllocAndReadGptData() the code was changed to use the GPT header
to determine the LBA of the GPT entries. This change did not account
for devices that have an invalid header and it can attempt to read
from invalid block addresses on a device.
This commit happened here:
a2d72f7 vboot: cgpt: Refer to partition entries by entries_lba.
https://chromium-review.googlesource.com/213861
The subsequent steps in vboot, LoadKernel->GptInit->GptRepair will
fix a missing header and entries, so it is only necessary for one of
the headers to be valid.
This is commonly the case with a new USB stick that has an image
written to it as only the primary header will be valid in this case.
However it is also true if the primary header has been corrupted and
the secondary header is still valid.
The code has been changed to call CheckHeader() on the primary and
secondary headers before attempting to use the 'entries_lba' field
to read the entries from the device. AllocAndReadGptData() now only
fails if both headers are invalid.
A number of new unit tests are created to check for these failure
conditions. In order to support this I had to extend the vboot_kernel
test infrastructure to have a buffer for the mocked disk data instead
of just ignoring reads and writes. This is because many of the existing
tests assumed they could have an invalid GPT header and still pass.
Now that the header is checked it is necessary for a valid header to
be created before the tests can pass.
BUG=chrome-os-partner:32386
BRANCH=samus,auron
TEST=All unit tests pass when running 'make runtests'
In addition real-world testing was done by corrupting the primary
and/or secondary headers of USB stick to ensure that it will
successfully boot if one of the headers is valid.
Change-Id: I7f840a44742fa3ba9a124df29ab5749e4c5a40c1
Signed-off-by: Duncan Laurie <dlaurie@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/220757
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Reviewed-by: Nam Nguyen <namnguyen@chromium.org>
This CL accesses the partition entry array through its header's
entries_lba value.
Previously, we assume the primary entry array lies on third sector, and
the secondary array lies (1 + 32) sectors from disk end. This assumption
was fine, even Wikipedia assumed the same.
But in order for us to support writing boot code to the third sector (as
required by some Freescale board), the primary entry array must be moved
to another location. Therefore, we must use "entries_lba" to locate the
arrays from now on.
BRANCH=none
BUG=chromium:406432
TEST=unittest
TEST=`cgpt create -p` and then `cgpt show`. Make sure the table
header and entries are properly moved.
Change-Id: Ia9008b0bb204f290b1f6240df562ce7d3a9bbff2
Reviewed-on: https://chromium-review.googlesource.com/213861
Reviewed-by: Bill Richardson <wfrichar@chromium.org>
Tested-by: Bill Richardson <wfrichar@chromium.org>
Commit-Queue: Nam Nguyen <namnguyen@chromium.org>
Tested-by: Nam Nguyen <namnguyen@chromium.org>
Add checks that the vboot library does not leak memory. This works by
tracking VbExMalloc() calls and making sure that they have an associated
VbExFree().
Adjust host_signature to use VbExFree() instead of free(), so that this
scheme works correctly for existing code.
BUG=chrome-os-partner:21115
BRANCH=pit
TEST=FEATURES=test emerge-peach_pit vboot_reference
Change-Id: I6ccccfbcc162fc43fb75862cd0eddad78ce8b18a
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/66175
At present reading data from storage in Vboot is a little fragmented. For
the firmware image, we expect the boot loader to handle this. For the disk
we have a block-level API. For the GBB (which also sits in the firmware
image) we expect the entire thing to be read before Vboot is called.
Add the concept of a region, and an API to read from a region. At present,
and most pressing, is reading from a GBB region. In the future this could
be extended to other parts of the firmware or even the disk.
Move all access to the GBB into this API so that the boot loader can provide
either a GBB region in one large contiguous chunk, or a function to deal with
read requests from vboot.
The call to VbExRegionRead() is behind a flag since not all boot loaders
support it yet.
The main change for boot loaders which don't support this new API is that
vboot will do more behind the scenes. For example, it will allocate memory
for chunks of data that it reads from the GBB, rather than just accessing it
directly. This approach is considerably simpler than trying to pass char **
everywhere and have vboot decide whether something needs to be allocated or
not.
The tests are updated, mainly to include setting up a GBB structure
accessible from VbCommonParams, which is now required by the firmware and
kernel functions. In normal operation this is set up at the start of
VbLoadFIrmware() and VbSelectAndLoadKernel() but for tests which call
children of these functions directly, the GBB structure must be set up
manually by the test.
BUG=chrome-os-partner:21115
BRANCH=none
TEST=manual
FEATURES=test sudo -E emerge vboot_reference
Change-Id: If2b8bbe467fdbd643239d8d9b5d7aa98df4d286f
Signed-off-by: Simon Glass <sjg@chromium.org>
Signed-off-by: David Hendricks <dhendrix@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/63336
Reviewed-by: Randall Spangler <rspangler@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/167361
At present reading data from storage in Vboot is a little fragmented. For
the firmware image, we expect the boot loader to handle this. For the disk
we have a block-level API. For the GBB (which also sits in the firmware
image) we expect the entire thing to be read before Vboot is called.
Add the concept of a region, and an API to read from a region. At present,
and most pressing, is reading from a GBB region. In the future this could
be extended to other parts of the firmware or even the disk.
Move all access to the GBB into this API so that the boot loader can provide
either a GBB region in one large contiguous chunk, or a function to deal with
read requests from vboot.
The call to VbExRegionRead() is behind a flag since not all boot loaders
support it yet.
The main change for boot loaders which don't support this new API is that
vboot will do more behind the scenes. For example, it will allocate memory
for chunks of data that it reads from the GBB, rather than just accessing it
directly. This approach is considerably simpler than trying to pass char **
everywhere and have vboot decide whether something needs to be allocated or
not.
The tests are updated, mainly to include setting up a GBB structure
accessible from VbCommonParams, which is now required by the firmware and
kernel functions. In normal operation this is set up at the start of
VbLoadFIrmware() and VbSelectAndLoadKernel() but for tests which call
children of these functions directly, the GBB structure must be set up
manually by the test.
BUG=chrome-os-partner:21115
BRANCH=none
TEST=manual
FEATURES=test sudo -E emerge vboot_reference
Change-Id: I2c19e9dc2ed602d0642bbf4f7d27f79fe9fad873
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/63336
Reviewed-by: Randall Spangler <rspangler@chromium.org>
