Smcsrind exception fixex #1896
Smcsrind exception fixex #1896
Conversation
|
Haven't had the chance to review yet, but please clean up the commit history to get rid of the merge commit and the commits with debug prints. In this case, squashing everything into a single commit is probably appropriate. |
JJ-Gaisler
force-pushed
the
smcsrind_fix
branch
2 times, most recently
from
44a4f87 to
c9b298b
Compare
|
I was just extending my testing and I noticed that checks for the s/hcontext bits is also missing, I'll be adding these soon. Should I also add hedelegh as a read-only zero CSR and incorporate the check for p1p13 while at it? |
|
May as well, thank you! |
|
You can add them to this PR if you prefer. Up to you. |
missing proxy csr and check for csr privilege in sscsring_reg_csr_t since mireg uses the same class. Add stateen checks for scontext/hcontext Fix for issue: 1893
JJ-Gaisler
force-pushed
the
smcsrind_fix
branch
from
c9b298b to
a8e9876
Compare
|
I couldn't find mstateen0_priv113 nor csr_hedelegh in the encoding.h file. I have added these constants in the code inline for now, I am waiting for riscv/riscv-opcodes#333 I have a test case which loops over CSRs that are affected by stateen, it checks that the correct trap is raised in all privilege modes. The test can be run with: |
JJ-Gaisler
force-pushed
the
smcsrind_fix
branch
from
65b1d3f to
2db2820
Compare
|
riscv-opcodes PR has been merged. |
JJ-Gaisler
force-pushed
the
smcsrind_fix
branch
from
2db2820 to
89386f3
Compare
Ensuring No negation on -2^63(int64_t) is performed during mulh/mulhsu Signed-off-by: Nicolas Brunie <82109999+nibrunieAtSi5@users.noreply.github.com> Update riscv/arith.h Signed-off-by: Nicolas Brunie <82109999+nibrunieAtSi5@users.noreply.github.com>
Pointer masking needs to be reapplied after computing the address of the tail of a misaligned access in case there's a carry-out into the MSBs. Resolves riscv-software-src#1895
…d besti for shamt condition
…sing vector CSRs" This reverts commit a17842c.
Fixes for: #1893
I have implemented smcdeleg in another branch and tested this implementation with smcdeleg and from the way I interpret the spec this should fix the issues I was seeing before. Let me know if something is poorly explained or incorrect.