linux-kernel - Re: [PATCH] riscv/kprobe: Fix instruction simulation of JALR

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAJF2gTQifm_uJ4pEFTZdfj2NWbNA1Xr3s7iT=drDhk_BQfTbBg@mail.gmail.com>
Date:   Sat, 21 Jan 2023 10:59:51 +0800
From:   Guo Ren <guoren@...nel.org>
To:     Liao Chang <liaochang1@...wei.com>
Cc:     paul.walmsley@...ive.com, palmer@...belt.com,
        aou@...s.berkeley.edu, mhiramat@...nel.org, me@...ki.ch,
        penberg@...nel.org, linux-riscv@...ts.infradead.org,
        linux-kernel@...r.kernel.org, bjorn@...nel.org, chenlifu@...wei.com
Subject: Re: [PATCH] riscv/kprobe: Fix instruction simulation of JALR

On Mon, Jan 16, 2023 at 2:46 PM Liao Chang <liaochang1@...wei.com> wrote:
>
> Set kprobe at 'jalr 1140(ra)' of vfs_write results in the following
> crash:
>
> [   32.092235] Unable to handle kernel access to user memory without uaccess routines at virtual address 00aaaaaad77b1170
> [   32.093115] Oops [#1]
> [   32.093251] Modules linked in:
> [   32.093626] CPU: 0 PID: 135 Comm: ftracetest Not tainted 6.2.0-rc2-00013-gb0aa5e5df0cb-dirty #16
> [   32.093985] Hardware name: riscv-virtio,qemu (DT)
> [   32.094280] epc : ksys_read+0x88/0xd6
> [   32.094855]  ra : ksys_read+0xc0/0xd6
> [   32.095016] epc : ffffffff801cda80 ra : ffffffff801cdab8 sp : ff20000000d7bdc0
> [   32.095227]  gp : ffffffff80f14000 tp : ff60000080f9cb40 t0 : ffffffff80f13e80
> [   32.095500]  t1 : ffffffff8000c29c t2 : ffffffff800dbc54 s0 : ff20000000d7be60
> [   32.095716]  s1 : 0000000000000000 a0 : ffffffff805a64ae a1 : ffffffff80a83708
> [   32.095921]  a2 : ffffffff80f160a0 a3 : 0000000000000000 a4 : f229b0afdb165300
> [   32.096171]  a5 : f229b0afdb165300 a6 : ffffffff80eeebd0 a7 : 00000000000003ff
> [   32.096411]  s2 : ff6000007ff76800 s3 : fffffffffffffff7 s4 : 00aaaaaad77b1170
> [   32.096638]  s5 : ffffffff80f160a0 s6 : ff6000007ff76800 s7 : 0000000000000030
> [   32.096865]  s8 : 00ffffffc3d97be0 s9 : 0000000000000007 s10: 00aaaaaad77c9410
> [   32.097092]  s11: 0000000000000000 t3 : ffffffff80f13e48 t4 : ffffffff8000c29c
> [   32.097317]  t5 : ffffffff8000c29c t6 : ffffffff800dbc54
> [   32.097505] status: 0000000200000120 badaddr: 00aaaaaad77b1170 cause: 000000000000000d
> [   32.098011] [<ffffffff801cdb72>] ksys_write+0x6c/0xd6
> [   32.098222] [<ffffffff801cdc06>] sys_write+0x2a/0x38
> [   32.098405] [<ffffffff80003c76>] ret_from_syscall+0x0/0x2
>
> Since the rs1 and rd might be the same one, such as 'jalr 1140(ra)',
> hence it requires obtaining the target address from rs1 followed by
> updating rd.
>
> Fixes: c22b0bcb1dd0 ("riscv: Add kprobes supported")
> Signed-off-by: Liao Chang <liaochang1@...wei.com>
> ---
>  arch/riscv/kernel/probes/simulate-insn.c | 6 +-----
>  1 file changed, 1 insertion(+), 5 deletions(-)
>
> diff --git a/arch/riscv/kernel/probes/simulate-insn.c b/arch/riscv/kernel/probes/simulate-insn.c
> index d73e96f6ed7c..02aa262fb2ac 100644
> --- a/arch/riscv/kernel/probes/simulate-insn.c
> +++ b/arch/riscv/kernel/probes/simulate-insn.c
> @@ -71,17 +71,13 @@ bool __kprobes simulate_jalr(u32 opcode, unsigned long addr, struct pt_regs *reg
>         u32 rd_index = (opcode >> 7) & 0x1f;
>         u32 rs1_index = (opcode >> 15) & 0x1f;
>
> -       ret = rv_insn_reg_set_val(regs, rd_index, addr + 4);
> -       if (!ret)
> -               return ret;
> -
>         ret = rv_insn_reg_get_val(regs, rs1_index, &base_addr);
>         if (!ret)
>                 return ret;
Ah, thank you, If rd == rs1,  the regs-><rs1> got a overwritten value,
my mistake. We should read rs1 firstly.

But for the patch, just exchanging the position is enough; keep the
original sequence for coding convention.
eg:
@@ -71,11 +71,11 @@ bool __kprobes simulate_jalr(u32 opcode, unsigned
long addr, struct pt_regs *reg
        u32 rd_index = (opcode >> 7) & 0x1f;
        u32 rs1_index = (opcode >> 15) & 0x1f;

-       ret = rv_insn_reg_set_val(regs, rd_index, addr + 4);
+       ret = rv_insn_reg_get_val(regs, rs1_index, &base_addr);
        if (!ret)
                return ret;

-       ret = rv_insn_reg_get_val(regs, rs1_index, &base_addr);
+       ret = rv_insn_reg_set_val(regs, rd_index, addr + 4);
        if (!ret)
                return ret;

Then:
Reviewed-by: Guo Ren <guoren@...nel.org>



>
>         instruction_pointer_set(regs, (base_addr + sign_extend32((imm), 11))&~1);
>
> -       return ret;
> +       return rv_insn_reg_set_val(regs, rd_index, addr + 4);
>  }
>
>  #define auipc_rd_idx(opcode) \
> --
> 2.25.1
>


-- 
Best Regards
 Guo Ren