[Skiboot] [PATCH 08/15] core/trace: Export trace buffers to sysfs
Oliver
oohall at gmail.com
Mon Mar 25 12:36:22 AEDT 2019
On Mon, Mar 25, 2019 at 11:18 AM Jordan Niethe <jniethe5 at gmail.com> wrote:
>
> Every property in the device tree under /ibm,opal/firmware/exports has a
> sysfs node created in /firmware/opal/exports. Add properties with the
> physical address and size for each trace buffer so they are exported.
> ---
> core/test/run-trace.c | 1 +
> core/trace.c | 10 ++++++++++
> 2 files changed, 11 insertions(+)
>
> diff --git a/core/test/run-trace.c b/core/test/run-trace.c
> index 971b967a5719..6cb9a582078d 100644
> --- a/core/test/run-trace.c
> +++ b/core/test/run-trace.c
> @@ -269,6 +269,7 @@ int main(void)
> unsigned int i, j;
>
> opal_node = dt_new_root("opal");
> + dt_new(dt_new(opal_node, "firmware"), "exports");
> for (i = 0; i < CPUS; i++) {
> fake_cpus[i].server_no = i;
> fake_cpus[i].is_secondary = (i & 0x1);
> diff --git a/core/trace.c b/core/trace.c
> index b33a5bcf2318..dfb5cfad045c 100644
> --- a/core/trace.c
> +++ b/core/trace.c
> @@ -19,6 +19,7 @@
> #include <lock.h>
> #include <string.h>
> #include <stdlib.h>
> +#include <inttypes.h>
> #include <cpu.h>
> #include <device.h>
> #include <libfdt.h>
> @@ -163,12 +164,21 @@ static void trace_add_dt_props(void)
> {
> unsigned int i;
> u64 *prop, tmask;
> + struct dt_node *exports;
> + char tname[256];
>
> prop = malloc(sizeof(u64) * 2 * debug_descriptor.num_traces);
>
> + exports = dt_find_by_path(opal_node, "firmware/exports");
> for (i = 0; i < debug_descriptor.num_traces; i++) {
> prop[i * 2] = cpu_to_fdt64(debug_descriptor.trace_phys[i]);
> prop[i * 2 + 1] = cpu_to_fdt64(debug_descriptor.trace_size[i]);
> +
> + snprintf(tname, sizeof(tname), "trace-%x-%"PRIx64,
> + debug_descriptor.trace_cpu[i],
> + debug_descriptor.trace_phys[i]);
> + dt_add_property_u64s(exports, tname, debug_descriptor.trace_phys[i],
> + debug_descriptor.trace_size[i]);
This is going to be a bit spammy especially on the bigger systems. A
dual socket P9 can have up to 44 cores so we're going to have a fair
amount of junk in there. It would be nice if we could have some kind
of categorisation, (e.g. subnodes that turn into subdirectories of
exports/), but that would require a kernel change. Maybe a nvram
option to flatten the exports directory would do the trick? Stewart,
what do you think?
Code looks fine though.
> }
>
> dt_add_property(opal_node, "ibm,opal-traces",
> --
> 2.20.1
>
> _______________________________________________
> Skiboot mailing list
> Skiboot at lists.ozlabs.org
> https://lists.ozlabs.org/listinfo/skiboot
More information about the Skiboot
mailing list