[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87sfgqr7d4.fsf@all.your.base.are.belong.to.us>
Date: Wed, 04 Jan 2023 10:12:23 +0100
From: Björn Töpel <bjorn@...nel.org>
To: "liaochang (A)" <liaochang1@...wei.com>,
Chen Guokai <chenguokai17@...ls.ucas.ac.cn>,
paul.walmsley@...ive.com, palmer@...belt.com,
aou@...s.berkeley.edu, rostedt@...dmis.org, mingo@...hat.com,
sfr@...b.auug.org.au
Cc: linux-riscv@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH v5 7/9] riscv/kprobe: Prepare detour buffer for
optimized kprobe
"liaochang (A)" <liaochang1@...wei.com> writes:
>>> +SYM_ENTRY(optprobe_template_insn, SYM_L_GLOBAL, SYM_A_NONE)
>>> + /*
>>> + * Step3:
>>> + * NOPS will be replaced by the probed instruction, at worst case 3 RVC
>>> + * and 1 RVI instructions is about to execute out of line.
>>> + */
>>> + nop
>>
>> A nop here will be either a compressed nop or a non-compressed,
>> depending on the build (C-enabled or not), right? Maybe be explicit to
>> the assembler what you want?
>>
>
> You are right, if CONFIG_RISCV_ISA_C is disabled, two NOP is enough for 2 RVI execute out of line,
> if CONFIG_RISCV_ISA_C is enabled, it needs eight C.NOP here for the worst case (3 RVC + 1 RVI).
>
> I will use {C}.NOP explicitly for different configure in next revision, thanks.
What I meant was that "nop" can expand to compressed instructions, and
you should be explicit. So you know how it's expanded by the
compiler/assembler.
An example:
$ cat bar.S
.text
bar:
nop
nop
$ riscv64-linux-gnu-gcc -O2 -o bar.o -c bar.S && riscv64-linux-gnu-objdump -M no-aliases -d bar.o
bar.o: file format elf64-littleriscv
Disassembly of section .text:
0000000000000000 <bar>:
0: 0001 c.addi zero,0
2: 0001 c.addi zero,0
vs
$ cat foo.S
.text
foo:
.option norvc
nop
nop
$ riscv64-linux-gnu-gcc -O2 -o foo.o -c foo.S && riscv64-linux-gnu-objdump -M no-aliases -d foo.o
foo.o: file format elf64-littleriscv
Disassembly of section .text:
0000000000000000 <foo>:
0: 00000013 addi zero,zero,0
4: 00000013 addi zero,zero,0
Björn
Powered by blists - more mailing lists