From 27e1b65aceef7b051bcfa74e1020f82a2210ed95 Mon Sep 17 00:00:00 2001 From: Gilad Ben-Yossef Date: Sun, 23 Apr 2017 12:26:15 +0300 Subject: staging: ccree: add TODO list Add TODO list for moving out of staging tree for ccree crypto driver Signed-off-by: Gilad Ben-Yossef Signed-off-by: Greg Kroah-Hartman --- drivers/staging/ccree/TODO | 30 ++++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) create mode 100644 drivers/staging/ccree/TODO (limited to 'drivers/staging/ccree/TODO') diff --git a/drivers/staging/ccree/TODO b/drivers/staging/ccree/TODO new file mode 100644 index 000000000000..c9f5754d062d --- /dev/null +++ b/drivers/staging/ccree/TODO @@ -0,0 +1,30 @@ + + +************************************************************************* +* * +* Arm Trust Zone CryptoCell REE Linux driver upstreaming TODO items * +* * +************************************************************************* + +ccree specific items +a.k.a stuff fixing for this driver to move out of staging +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ + +1. Move to using Crypto Engine to handle backlog queueing. +2. Remove synchronous algorithm support leftovers. +3. Separate platform specific code for FIPS and power management into separate platform modules. +4. Drop legacy kernel support code. +5. Move most (all?) #ifdef CONFIG into inline functions. +6. Remove all unused definitions. +7. Re-factor to accomediate newer/older HW revisions besides the 712. +8. Handle the many checkpatch errors. +9. Implement ahash import/export correctly. +10. Go through a proper review of DT bindings and sysfs ABI +11. Sort out FIPS mode: bake tests into testmgr, sort out behaviour on error, + figure if 3DES weak key check is needed + +Kernel infrastructure items +a.k.a stuff we either neither need to fix in the kernel or understand what we're doing wrong +~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ +1. ahash import/export context has a PAGE_SIZE/8 size limit. We need more. +2. Crypto Engine seems to be built for HW with hardware queue depth of 1, we have 600++. -- cgit