Re: [PATCH] Add gpu memory tracepoints

[Yiwei Zhang]

On Mon, Feb 10, 2020 at 5:16 PM <zzyiwei@xxxxxxxxxx> wrote:
>
> From: Yiwei Zhang <zzyiwei@xxxxxxxxxx>
>
> This change adds the below gpu memory tracepoint:
> gpu_mem/gpu_mem_total: track global or process gpu memory total counters
>
> Signed-off-by: Yiwei Zhang <zzyiwei@xxxxxxxxxx>
> ---
>  include/trace/events/gpu_mem.h | 64 ++++++++++++++++++++++++++++++++++
>  kernel/trace/Kconfig           |  3 ++
>  kernel/trace/Makefile          |  1 +
>  kernel/trace/trace_gpu_mem.c   | 13 +++++++
>  4 files changed, 81 insertions(+)
>  create mode 100644 include/trace/events/gpu_mem.h
>  create mode 100644 kernel/trace/trace_gpu_mem.c
>
> diff --git a/include/trace/events/gpu_mem.h b/include/trace/events/gpu_mem.h
> new file mode 100644
> index 000000000000..3b632a2b5100
> --- /dev/null
> +++ b/include/trace/events/gpu_mem.h
> @@ -0,0 +1,64 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * GPU memory trace points
> + *
> + * Copyright (C) 2020 Google, Inc.
> + */
> +
> +#undef TRACE_SYSTEM
> +#define TRACE_SYSTEM gpu_mem
> +
> +#if !defined(_TRACE_GPU_MEM_H) || defined(TRACE_HEADER_MULTI_READ)
> +#define _TRACE_GPU_MEM_H
> +
> +#include <linux/tracepoint.h>
> +
> +/*
> + * The gpu_memory_total event indicates that there's an update to either the
> + * global or process total gpu memory counters.
> + *
> + * This event should be emitted whenever the kernel device driver allocates,
> + * frees, imports, unimports memory in the GPU addressable space.
> + *
> + * @gpu_id: This is the gpu id.
> + *
> + * @pid: Put 0 for global total, while positive pid for process total.
> + *
> + * @size: Virtual size of the allocation in bytes.
> + *
> + */
> +TRACE_EVENT(gpu_mem_total,
> +       TP_PROTO(
> +               uint32_t gpu_id,
> +               uint32_t pid,
> +               uint64_t size
> +       ),
> +       TP_ARGS(
> +               gpu_id,
> +               pid,
> +               size
> +       ),
> +       TP_STRUCT__entry(
> +               __field(uint32_t, gpu_id)
> +               __field(uint32_t, pid)
> +               __field(uint64_t, size)
> +       ),
> +       TP_fast_assign(
> +               __entry->gpu_id = gpu_id;
> +               __entry->pid = pid;
> +               __entry->size = size;
> +       ),
> +       TP_printk(
> +               "gpu_id=%u "
> +               "pid=%u "
> +               "size=%llu",
> +               __entry->gpu_id,
> +               __entry->pid,
> +               __entry->size
> +       )
> +);
> +
> +#endif /* _TRACE_GPU_MEM_H */
> +
> +/* This part must be outside protection */
> +#include <trace/define_trace.h>
> diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
> index 91e885194dbc..cb404755b0a6 100644
> --- a/kernel/trace/Kconfig
> +++ b/kernel/trace/Kconfig
> @@ -85,6 +85,9 @@ config EVENT_TRACING
>  config CONTEXT_SWITCH_TRACER
>         bool
>
> +config TRACE_GPU_MEM
> +       bool
> +
>  config RING_BUFFER_ALLOW_SWAP
>         bool
>         help
> diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile
> index f9dcd19165fa..267985313dca 100644
> --- a/kernel/trace/Makefile
> +++ b/kernel/trace/Makefile
> @@ -47,6 +47,7 @@ obj-$(CONFIG_PREEMPTIRQ_DELAY_TEST) += preemptirq_delay_test.o
>  obj-$(CONFIG_SYNTH_EVENT_GEN_TEST) += synth_event_gen_test.o
>  obj-$(CONFIG_KPROBE_EVENT_GEN_TEST) += kprobe_event_gen_test.o
>  obj-$(CONFIG_CONTEXT_SWITCH_TRACER) += trace_sched_switch.o
> +obj-$(CONFIG_TRACE_GPU_MEM) += trace_gpu_mem.o
>  obj-$(CONFIG_FUNCTION_TRACER) += trace_functions.o
>  obj-$(CONFIG_PREEMPTIRQ_TRACEPOINTS) += trace_preemptirq.o
>  obj-$(CONFIG_IRQSOFF_TRACER) += trace_irqsoff.o
> diff --git a/kernel/trace/trace_gpu_mem.c b/kernel/trace/trace_gpu_mem.c
> new file mode 100644
> index 000000000000..01e855897b6d
> --- /dev/null
> +++ b/kernel/trace/trace_gpu_mem.c
> @@ -0,0 +1,13 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * GPU memory trace points
> + *
> + * Copyright (C) 2020 Google, Inc.
> + */
> +
> +#include <linux/module.h>
> +
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/gpu_mem.h>
> +
> +EXPORT_TRACEPOINT_SYMBOL(gpu_mem_total);
> --
> 2.25.0.341.g760bfbb309-goog
>