Re: [PATCH v12 0/5] Add kprobe and kretprobe support for LoongArch

[Huacai Chen]

Hi, Jeff,

The code has been updated here,
https://github.com/loongson/linux/commits/loongarch-next, you can test
again.

Huacai

On Tue, Feb 7, 2023 at 11:14 AM Tiezhu Yang <yangtiezhu@xxxxxxxxxxx> wrote:
>
>
>
> On 02/06/2023 08:48 PM, Jeff Xie wrote:
> > On Mon, Feb 6, 2023 at 8:13 PM Huacai Chen <chenhuacai@xxxxxxxxxx> wrote:
> >>
> >> Hi, Jeff,
> >>
> >> Now I add kprobes on ftrace support in
> >> https://github.com/loongson/linux/commits/loongarch-next, please test
> >> again. Thank you.
> >>
> >
> > When using the kprobe example module kprobe_example.ko, I haven't seen
> > any errors.
> >
> > But when using the ftrace to probe the symbol + offset, the kernel will panic:
> > e.g. probe the scheduler_tick+4 is fine, but when probe the
> > scheduler_tick+5, the kernel will panic.
> >
>
> Thanks for your test.
>
> We can see that the instruction address is 4-byte alignment,
> this is because the instruction length is 32-bit on LoongArch.
>
> $ objdump -d vmlinux > dump.txt
> $ grep -A 20 scheduler_tick dump.txt | head -21
> 9000000000279fc8 <scheduler_tick>:
> 9000000000279fc8:       03400000        andi            $zero, $zero, 0x0
> 9000000000279fcc:       03400000        andi            $zero, $zero, 0x0
> 9000000000279fd0:       02ff4063        addi.d          $sp, $sp, -48(0xfd0)
> 9000000000279fd4:       29c08077        st.d            $s0, $sp, 32(0x20)
> 9000000000279fd8:       29c06078        st.d            $s1, $sp, 24(0x18)
> 9000000000279fdc:       29c04079        st.d            $s2, $sp, 16(0x10)
> 9000000000279fe0:       29c0207a        st.d            $s3, $sp, 8(0x8)
> 9000000000279fe4:       29c0a061        st.d            $ra, $sp, 40(0x28)
> 9000000000279fe8:       2700007b        stptr.d         $s4, $sp, 0
> 9000000000279fec:       24001844        ldptr.w         $a0, $tp, 24(0x18)
> 9000000000279ff0:       1a02edd9        pcalau12i       $s2, 5998(0x176e)
> 9000000000279ff4:       1a034bac        pcalau12i       $t0, 6749(0x1a5d)
> 9000000000279ff8:       02f56339        addi.d          $s2, $s2, -680(0xd58)
> 9000000000279ffc:       00410c9a        slli.d          $s3, $a0, 0x3
> 900000000027a000:       28aae18d        ld.w            $t1, $t0, -1352(0xab8)
> 900000000027a004:       380c6b2e        ldx.d           $t2, $s2, $s3
> 900000000027a008:       1a022fcc        pcalau12i       $t0, 4478(0x117e)
> 900000000027a00c:       02f20198        addi.d          $s1, $t0, -896(0xc80)
> 900000000027a010:       00150317        move            $s0, $s1
> 900000000027a014:       004081ac        slli.w          $t0, $t1, 0x0
>
> So we should check the probe address at the beginning of
> arch_prepare_kprobe(), some other archs do the same thing.
>
> $ git diff
> diff --git a/arch/loongarch/kernel/kprobes.c
> b/arch/loongarch/kernel/kprobes.c
> index bdab707b6edf..56c8c4b09a42 100644
> --- a/arch/loongarch/kernel/kprobes.c
> +++ b/arch/loongarch/kernel/kprobes.c
> @@ -79,6 +79,9 @@ NOKPROBE_SYMBOL(arch_prepare_simulate);
>
>   int arch_prepare_kprobe(struct kprobe *p)
>   {
> +       if ((unsigned long)p->addr & 0x3)
> +               return -EILSEQ;
> +
>          /* copy instruction */
>          p->opcode = *p->addr;
>
>
> Thanks,
> Tiezhu
>
>