[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <22dbe32a-0757-97ad-8885-0f2ba55bb743@huawei.com>
Date: Fri, 27 Nov 2020 19:43:19 +0800
From: Kefeng Wang <wangkefeng.wang@...wei.com>
To: Chen Huang <chenhuang5@...wei.com>,
Paul Walmsley <paul.walmsley@...ive.com>,
Palmer Dabbelt <palmer@...belt.com>,
Albert Ou <aou@...s.berkeley.edu>
CC: Andrew Morton <akpm@...ux-foundation.org>,
<linux-riscv@...ts.infradead.org>, <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] riscv/stacktrace: fix stackframe without ra on the top
On 2020/11/27 16:45, Chen Huang wrote:
> When a function doesn't have a callee, then it will not push ra
> into the stack, such as lkdtm_BUG() function:
>
> addi sp,sp,-16
> sd s0,8(sp)
> addi s0,sp,16
> ebreak
>
> Then we use pt_regs as a parameter to walk_stackframe(), for the
> struct stackframe use {fp,ra} to get information from stack, it
> will get the wrong value. And the call trace will be:
>
> [<ffffffe00066c56c>] lkdtm_BUG+0x6/0x8
> ---[ end trace 18da3fbdf08e25d5 ]---
>
> It should be that:
> [<ffffffe00066c56c>] lkdtm_BUG+0x6/0x8
> [<ffffffe0008b24a4>] lkdtm_do_action+0x14/0x1c
> [<ffffffe00066c372>] direct_entry+0xc0/0x10a
> [<ffffffe000439f86>] full_proxy_write+0x42/0x6a
> [<ffffffe000309626>] vfs_write+0x7e/0x214
> [<ffffffe00030992a>] ksys_write+0x98/0xc0
> [<ffffffe000309960>] sys_write+0xe/0x16
> [<ffffffe0002014bc>] ret_from_syscall+0x0/0x2
> ---[ end trace 61917f3d9a9fadcd ]---
>
> Signed-off-by: Chen Huang <chenhuang5@...wei.com>
> ---
> arch/riscv/kernel/stacktrace.c | 9 ++++++++-
> 1 file changed, 8 insertions(+), 1 deletion(-)
>
> diff --git a/arch/riscv/kernel/stacktrace.c b/arch/riscv/kernel/stacktrace.c
> index 595342910c3f..d1d4c18335c4 100644
> --- a/arch/riscv/kernel/stacktrace.c
> +++ b/arch/riscv/kernel/stacktrace.c
> @@ -57,7 +57,14 @@ void notrace walk_stackframe(struct task_struct *task, struct pt_regs *regs,
> /* Unwind stack frame */
> frame = (struct stackframe *)fp - 1;
> sp = fp;
> - fp = frame->fp;
> + if (regs && (frame->fp & 0x7) && (pc == regs->epc)) {
> + fp = frame->ra;
> + pc = regs->ra;
> + continue;
> + } else {
> + fp = frame->fp;
> + }
> +
> pc = ftrace_graph_ret_addr(current, NULL, frame->ra,
> (unsigned long *)(fp - 8));
Better to move above code into else branch
> }
Powered by blists - more mailing lists