Refine selftest further #125
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mkannwischer
force-pushed
the
refine-selftest
branch
from
d6b9d95 to
e3974bd
Compare
These are not necessarily restricted to GPRs as code may hold pointers in FPU registers temporarily.
Previously the selftest would determine which registers are supposed to contain addresses (and allocate memory accordingly) by going through all memory operations and assuming that address registers used in those are the inputs holding the addresses initially. This assumption does not hold as (1) addresses may be moved around (2) addresses may actually be a result of a previous computation. One example where this did not work was vmov rX, sY ldr tZ, [rX] here the selftest would assume that rX needs to be an address. However, sY is the register holding the address on entry of the code. This commit refines the code determining the address registers by recursively going through the DFG and determining the _input_ registers that affect values being used as addresses. Note that in the above examples rX would not be marked as an address, at it is not an input. Hopefully this approach is more robust.
This allows to access memory at negative offsets.
mkannwischer
force-pushed
the
refine-selftest
branch
from
d5a636f to
ba11ddb
Compare
When the selftest fails with an emulation failure in unicorn, we would previously not print any information hinting at the underlying problem. This commit improves debuggability by printing the UcError exception thrown by Unicorn. Signed-off-by: Hanno Becker <[email protected]>
hanno-becker
force-pushed
the
refine-selftest
branch
from
9bc198e to
4313b3d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We encountered further issues with the selftest in the
armv7mbranch. This PR refines it furtherWhen unicorn fails to run code, currently this fails silently and the selftest continues to run failing in the assertions checking the output resulting in misleading error messages. This is changed now to fail immediately with an appropriate eror message.
The configuration option
selftest_address_gprswas misnamed as addresses can also be contained in non-gpr-registers. I renamed it toselftest_address_registersPreviously the selftest would determine which registers are supposed to contain addresses (and allocate memory
accordingly) by going through all memory operations and assuming that address registers used in those are the
inputs holding the addresses initially.
This assumption does not hold as
(1) addresses may be moved around
(2) addresses may actually be a result of a previous computation.
One example where this did not work was
Here the selftest would assume that rX needs to be an address. However, sY is the register holding the address on entry of the code.
This commit refines the code determining the address registers by recursively going through the DFG and determining the input registers that affect values being used as addresses.
Note that in the above examples rX would not be marked as an address, at it is not an input.
Hopefully this approach is more robust.
Allow accessing memory at negative offsets. Sometimes code has pointers that do not point at the start of the buffer, but somewhere in the middle. If those are then used with negative offsets, unicorn fails as the memory is potentially undefined. This is fixed by allocating a double-sized buffer and placing the pointer in the middle.
The stack pointer is currently initialized to the top of the stack. However, if code is using the stack pointer (and the sp is not decremented as a part of the code snipped), then we get memory access errors from unicorn. This is fixed by allocating some stack space (the default buffer size) before calling the function.
Building on top of Refine selftest for Armv7-M #122, we also need to skip the selftest if symbolics are present in the preamble or postamble in the case of sw piplining. This PR adds the check in the same way as Refine selftest for Armv7-M #122 did for the main body of the code.