[testutils] make entropy_testutils work for darjeeling
#26269
Conversation
moidx
requested review from
jadephilipoom and
vogelpi
and removed request for
a team
|
|
||
| #define MODULE_ID MAKE_MODULE_ID('e', 'n', 'y') | ||
|
|
||
| status_t entropy_testutils_entropy_src_init(void) { | ||
| #ifdef HAS_ENTROPY_SRC |
There was a problem hiding this comment.
This is more of an open question: should this function be moved to the entropy_src testutils? I would expect that users of the entropy complex that don't specifically care of the entropy source will use entropy_testutils_auto_mode_init or entropy_testutils_boot_mode_init (which call this function anyway).
There was a problem hiding this comment.
There're at least 2 tests that I know that needs this for earlgrey while also exists in integrated_dev, which is why I decided to keep it entropy_testutils.
There was a problem hiding this comment.
Maybe let me reformulate the question: what is this function supposed to achieve that cannot be done by using one of the other initialization functions? It seems its purpose is to initialize specifically the entropy_src so maybe those tests should be switch to of the more general init functions? Can you point to those tests?
There was a problem hiding this comment.
There's ast_clk_rst_inputs.c which uses this function, but taking another look it seems that master and integrated_dev has diverged a bit, so probably having conditional compilation inside itself might make more sense.
Another example is entropy_src_csrng_test which mostly tests CSRNG.
There was a problem hiding this comment.
I agree on that one:
- entropy_src_csrng_test really just tests CSRNG and for this the entropy_src needs to be configured, too. On Darjeeling, we'll simply not configure ENTROPY_SRC.
- ast_clk_rst_inputs tests some features of AST including whether it consumes entropy from EDN (not affected by this code) and whether it outputs raw entropy. The latter is checked by looking at the fill status of the observe FIFO inside ENTROPY_SRC. For Darjeeling the test needs to adapted in a way to check that CSRNG can be reseeded with "raw" entropy from AST.
|
|
||
| #define MODULE_ID MAKE_MODULE_ID('e', 'n', 'y') | ||
|
|
||
| status_t entropy_testutils_entropy_src_init(void) { | ||
| #ifdef HAS_ENTROPY_SRC |
There was a problem hiding this comment.
I agree on that one:
- entropy_src_csrng_test really just tests CSRNG and for this the entropy_src needs to be configured, too. On Darjeeling, we'll simply not configure ENTROPY_SRC.
- ast_clk_rst_inputs tests some features of AST including whether it consumes entropy from EDN (not affected by this code) and whether it outputs raw entropy. The latter is checked by looking at the fill status of the observe FIFO inside ENTROPY_SRC. For Darjeeling the test needs to adapted in a way to check that CSRNG can be reseeded with "raw" entropy from AST.
This initializes entropy src with default configuration. `entropy_testutils_config_default` is removed. The idea is that this function will become a no-op for darjeeling when the user only needs the entropy src to be initialized but doesn't need special config. The old `entropy_testutils_config_default`, for most users, are immediately passed to initialize entropy_src, with only a single user tweaking it. This user is updated to just explicitly list all config, and other users are updated to call the new initialization function instead. Signed-off-by: Gary Guo <[email protected]>
…_check This function is still needed for darjeeling, but it doesn't have an entropy_src block and DIF cannot be used. This function is updated to implicitly check for the entropy_src internally, so it can work for both earlgrey and darjeeling. Signed-off-by: Gary Guo <[email protected]>
There's a part of entropy_testutils that is dedicated to entropy_src, rather than the entropy generation as a whole. This is now moved to entropy_src_testutils, allowing tops that does not have entropy_src to continue to use entropy_testutils. Signed-off-by: Gary Guo <[email protected]>
Signed-off-by: Gary Guo <[email protected]>
This involves the entropy_src specific part of to a new
entropy_src_testutils.Fix #26193