Thanks both for the reviews, and sorry for the CI breakage.
v3 keeps "avb init" lazy, as Igor suggested.
Patch 1 is the rename, unchanged, with the tags from v2.
Patch 2 is new:
validating an interface name without resolving a device needs a
name-to-uclass lookup and the block layer only had the reverse. Say if you
would rather not grow a block layer header for one caller, the alternative
is a static helper in cmd/avb.c.
Patch 3 is the reworked AVB change.
Based on u-boot-dfu-next, not on your test suite series. The fixup you
posted is no longer needed, since "avb init" resolves no device and the
test no longer refers to mmc_dev; it applies on top of your series with
git am -3 and passes there too.
On the total_compute motivation: that board carries only CONFIG_LIBAVB
since 867e16ae05e2 dropped AVB_VERIFY and CMD_AVB, so its
"avb init ${blk_dev} 0" fails as an unknown command, not a usage error.
test_avb.py on sandbox gives 6 passed, 2 skipped, where v2 gives 5 failed.
Also tested on Raspberry Pi 4 and 5, booting Android from SD on both and
from USB mass storage on the Pi 4.
Jan Kehren (3):
avb: Drop the mmc naming from the block IO helpers
blk: Add a name-to-interface-type lookup
avb: Take a storage interface in the AVB operations
boot/bootmeth_android.c | 2 +-
cmd/avb.c | 25 +++-
common/Kconfig | 7 +-
common/avb_verify.c | 167 ++++++++++++++-----------
doc/android/avb2.rst | 14 ++-
drivers/block/blk-uclass.c | 4 +-
drivers/block/blk_legacy.c | 7 ++
include/avb_verify.h | 38 ++++--
include/blk.h | 11 ++
test/py/tests/test_android/test_avb.py | 28 +++++
10 files changed, 209 insertions(+), 94 deletions(-)
base-commit: 25a459fd4d2eeef909dd4ea8a577ba82b1fc5e43
--
2.43.0
