search

riscv-software-src / riscv-pk

RISC-V Proxy Kernel
Other
595 stars 308 forks source link

Add ENOSYS stub for syscall 278, getrandom #300

Closed egouriou-rivos closed 1 year ago

egouriou-rivos commented 1 year ago

Binaries built with a recent LLVM and current riscv-gnu-toolchain and targeting riscv64-unknown-linux-gnu fail to execute under Spike+pk with this output:

$ /scratch/projects/riscv-isa-sim//build/spike \
    --varch=vlen:128,elen:64 --isa=rv64gcv_zvkned \
    /rivos/riscv-pk/riscv64-unknown-linux-gnu/bin/pk aes-cbc-test
bbl loader
bad syscall #278!

Syscall 278 is getrandom which is invoked prior to reaching main(). Providing a syscall stub to return ENOSYS is enough to please the caller. This has the additional benefit of not requiring to implement getrandom() in a portable way.

$ /scratch/projects/riscv-isa-sim//build/spike \
    --varch=vlen:128,elen:64 --isa=rv64gcv_zvkned \
    /rivos/riscv-pk/riscv64-unknown-linux-gnu/bin/pk aes-cbc-test
bbl loader
vlen: 128 bits
--- Running 'CBCKeySbox256' test suite...
...
egouriou-rivos commented 1 year ago

@aswaterman for your consideration. Please let me know if you want me to file an issue to document the pre-existing failure.

aswaterman commented 1 year ago

I’m certainly OK with the stub, but I’m confused as to why it is needed. Reading the code, it appears that sys_stub_nosys is already called whenever there is no such call defined. Where is the “bad syscall” printout coming from?

egouriou-rivos commented 1 year ago

Oh, I just didn't pay enough attention when moving from our internal version to /master . Rivos' internal version of riscv-pk is behind and doesn't contain ca69d69a.

commit ca69d69a304d5668a8c95d1de5c6c6fa71ed9936
Author: Andrew Waterman <andrew@sifive.com>
Date:   Mon Apr 11 01:22:16 2022 -0700

    Handle unimplemented syscalls gracefully

    We previously kernel-panicked because that made it more obvious when a
    syscall implementation was missing.  These days, it's more common that
    the C library will do something sensible in response to returning -ENOSYS.
    Favor that approach to avoid frustrating users.

Never mind then. I'll move our clone of riscv-pk forward. Thanks.

aswaterman commented 1 year ago

Ah, right, I forgot about that. Glad you figured it out.