riscv: Correct the handling of unexpected ebreak in do_trap_break()
authorVincent Chen <vincent.chen@sifive.com>
Mon, 23 Sep 2019 00:45:16 +0000 (08:45 +0800)
committerPaul Walmsley <paul.walmsley@sifive.com>
Mon, 7 Oct 2019 19:59:41 +0000 (12:59 -0700)
commit8bb0daef64e5a92db63ad1d3bbf9e280a7b3612a
treeb56848d150a8be8b712e8ef975a5d9830238381a
parente0c0fc18f10d5080cddde0e81505fd3e952c20c4
riscv: Correct the handling of unexpected ebreak in do_trap_break()

For the kernel space, all ebreak instructions are determined at compile
time because the kernel space debugging module is currently unsupported.
Hence, it should be treated as a bug if an ebreak instruction which does
not belong to BUG_TRAP_TYPE_WARN or BUG_TRAP_TYPE_BUG is executed in
kernel space. For the userspace, debugging module or user problem may
intentionally insert an ebreak instruction to trigger a SIGTRAP signal.
To approach the above two situations, the do_trap_break() will direct
the BUG_TRAP_TYPE_NONE ebreak exception issued in kernel space to die()
and will send a SIGTRAP to the trapped process only when the ebreak is
in userspace.

Signed-off-by: Vincent Chen <vincent.chen@sifive.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
[paul.walmsley@sifive.com: fixed checkpatch issue]
Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com>
arch/riscv/kernel/traps.c