On Wed, Jun 12, 2024 at 12:29 PM HAGIO KAZUHITO(萩尾 一仁) <k-hagio-ab@nec.com> wrote:

On 2024/06/12 12:42, Lianbo Jiang wrote:
> The commit 48764a14bc58 may cause a regression issue when the CONFIG_X86_FRED
> is not enabled, this is because the SIZE(fred_frame) will call the
> SIZE_verify() to determine if the fred_frame is valid, otherwise it will
> emit an error:
>
>    crash> bt 1
>
>    bt: invalid structure size: fred_frame
>          FILE: x86_64.c  LINE: 4089  FUNCTION: x86_64_low_budget_back_trace_cmd()
>
>    [/home/k-hagio/bin/crash] error trace: 588df3 => 5cbc72 => 5eb3e1 => 5eb366
>    PID: 1        TASK: ffff9f94c024b980  CPU: 2    COMMAND: "systemd"
>       #0 [ffffade44001bca8] __schedule at ffffffffb948ebbb
>       #1 [ffffade44001bd10] schedule at ffffffffb948f04d
>       #2 [ffffade44001bd20] schedule_hrtimeout_range_clock at ffffffffb9494fef
>       #3 [ffffade44001bda8] ep_poll at ffffffffb8c91be8
>       #4 [ffffade44001be48] do_epoll_wait at ffffffffb8c91d11
>       #5 [ffffade44001be80] __x64_sys_epoll_wait at ffffffffb8c92590
>       #6 [ffffade44001bed0] do_syscall_64 at ffffffffb947f459
>       #7 [ffffade44001bf50] entry_SYSCALL_64_after_hwframe at ffffffffb96000ea
>
>        5eb366: SIZE_verify.part.42+70
>        5eb3e1: SIZE_verify+49
>        5cbc72: x86_64_low_budget_back_trace_cmd+3010
>        588df3: back_trace+1523
>
>    bt: invalid structure size: fred_frame
>          FILE: x86_64.c  LINE: 4089  FUNCTION: x86_64_low_budget_back_trace_cmd()
>
> Let's replace the SIZE(fred_frame) with the VALID_SIZE(fred_frame) to
> fix it.
>
> Fixes: 48764a14bc58 ("x86_64: fix for adding top_of_kernel_stack_padding for kernel stack")
> Reported-by: Kazuhito Hagio <k-hagio-ab@nec.com>
> Signed-off-by: Lianbo Jiang <lijiang@redhat.com>
> ---
>   x86_64.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/x86_64.c b/x86_64.c
> index 6777c93e6b47..469d26b05e24 100644
> --- a/x86_64.c
> +++ b/x86_64.c
> @@ -4086,7 +4086,7 @@ in_exception_stack:
>   
>           if (!irq_eframe && !is_kernel_thread(bt->tc->task) &&
>               (GET_STACKBASE(bt->tc->task) == bt->stackbase)) {
> -             long stack_padding_size = SIZE(fred_frame) > 0 ? (2*8) : 0;
> +             long stack_padding_size = VALID_SIZE(fred_frame) ? (2*8) : 0;
>               user_mode_eframe = bt->stacktop - SIZE(pt_regs);
>               if (last_process_stack_eframe < user_mode_eframe)
>                       x86_64_exception_frame(EFRAME_PRINT, 0, bt->stackbuf +
> @@ -4408,7 +4408,7 @@ in_exception_stack:
>   
>           if (!irq_eframe && !is_kernel_thread(bt->tc->task) &&
>               (GET_STACKBASE(bt->tc->task) == bt->stackbase)) {
> -             long stack_padding_size = SIZE(fred_frame) > 0 ? (2*8) : 0;
> +             long stack_padding_size = VALID_SIZE(fred_frame) ? (2*8) : 0;
>               user_mode_eframe = bt->stacktop - SIZE(pt_regs);
>               if (last_process_stack_eframe < user_mode_eframe)
>                       x86_64_exception_frame(EFRAME_PRINT, 0, bt->stackbuf +

Thanks, tested ok with kernels configured without CONFIG_X86_FRED=y.


Thank you for helping with the test, Kazu.

Lianbo
 
Thanks,
Kazu