core: fix RPMB rollback vulnerability

Normal world is used to pass the RPMB request to the eMMC. If normal
world saves a write request and returns an error instead it can be used
at a later stage where OP-TEE doesn't expect a certain block to be
updated. For more details on possible attacks and mitigations see [1]
and [2].

The mitigation consists of two parts, while initializing and later how
each write request is handled.

While initializing the RPMB file system we don't have a spare dummy
block so the alternative method of reading a block and writing it again
is used instead.

For normal write request all errors after the request message has been
created will be retried 10 times. If a write request fails after 10
retries RPMB is disabled entirely until next boot. An eventual
requesting TA is with an unexpected error code since we can't tell if
the request has been committed to storage or not.

Link: [1] https://www.westerndigital.com/support/productsecurity/wdc-20008-replay-attack-vulnerabilities-rpmb-protocol-applications
Link: [2] https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/collateral/white-paper/white-paper-replay-protected-memory-block-protocol-vulernabilities.pdf

Acked-by: Joakim Bech <joakim.bech@linaro.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>

0 files changed, 0 insertions, 0 deletions