On Thu, 18 Apr 2019 10:41:40 +0200
Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> The per cpu stack trace buffer usage pattern is odd at best. The buffer has
> place for 512 stack trace entries on 64-bit and 1024 on 32-bit. When
> interrupts or exceptions nest after the per cpu buffer was acquired the
> stacktrace length is hardcoded to 8 entries. 512/1024 stack trace entries
> in kernel stacks are unrealistic so the buffer is a complete waste.
>
> Split the buffer into chunks of 64 stack entries which is plenty. This
> allows nesting contexts (interrupts, exceptions) to utilize the cpu buffer
> for stack retrieval and avoids the fixed length allocation along with the
> conditional execution pathes.
>
> Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> ---
> kernel/trace/trace.c | 77 +++++++++++++++++++++++++--------------------------
> 1 file changed, 39 insertions(+), 38 deletions(-)
>
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -2749,12 +2749,21 @@ trace_function(struct trace_array *tr,
>
> #ifdef CONFIG_STACKTRACE
>
> -#define FTRACE_STACK_MAX_ENTRIES (PAGE_SIZE / sizeof(unsigned long))
> +/* 64 entries for kernel stacks are plenty */
> +#define FTRACE_KSTACK_ENTRIES 64
> +
> struct ftrace_stack {
> - unsigned long calls[FTRACE_STACK_MAX_ENTRIES];
> + unsigned long calls[FTRACE_KSTACK_ENTRIES];
> };
>
> -static DEFINE_PER_CPU(struct ftrace_stack, ftrace_stack);
> +/* This allows 8 level nesting which is plenty */
Can we make this 4 level nesting and increase the size? (I can see us
going more than 64 deep, kernel developers never cease to amaze me ;-)
That's all we need:
Context: Normal, softirq, irq, NMI
Is there any other way to nest?
-- Steve
> +#define FTRACE_KSTACK_NESTING (PAGE_SIZE / sizeof(struct ftrace_stack))
> +
> +struct ftrace_stacks {
> + struct ftrace_stack stacks[FTRACE_KSTACK_NESTING];
> +};
> +
> +static DEFINE_PER_CPU(struct ftrace_stacks, ftrace_stacks);
> static DEFINE_PER_CPU(int, ftrace_stack_reserve);
>
>
