Re: [PATCH RFC 3/4] drivers: irqchip: pdc: log PDC info in FTRACE

From: Steven Rostedt
Date: Tue Jan 23 2018 - 13:13:33 EST


On Tue, 23 Jan 2018 10:56:55 -0700
Lina Iyer <ilina@xxxxxxxxxxxxxx> wrote:

> From: Archana Sathyakumar <asathyak@xxxxxxxxxxxxxx>

Hi Lina and Archana,

>
> Log key PDC pin configuration in FTRACE.
>
> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> Signed-off-by: Archana Sathyakumar <asathyak@xxxxxxxxxxxxxx>
> Signed-off-by: Lina Iyer <ilina@xxxxxxxxxxxxxx>
> ---
> drivers/irqchip/qcom-pdc.c | 6 ++++++
> include/trace/events/pdc.h | 50 ++++++++++++++++++++++++++++++++++++++++++++++
> 2 files changed, 56 insertions(+)
> create mode 100644 include/trace/events/pdc.h
>
> diff --git a/drivers/irqchip/qcom-pdc.c b/drivers/irqchip/qcom-pdc.c
> index 57f1348bd81c..9b626e9f3a29 100644
> --- a/drivers/irqchip/qcom-pdc.c
> +++ b/drivers/irqchip/qcom-pdc.c
> @@ -26,6 +26,8 @@
> #include <linux/platform_device.h>
> #include <linux/slab.h>
> #include <linux/types.h>
> +#define CREATE_TRACE_POINTS
> +#include "trace/events/pdc.h"
>
> #include "qcom-pdc.h"
>
> @@ -78,6 +80,7 @@ static inline void pdc_enable_intr(struct irq_data *d, bool on)
> break;
> udelay(5);
> } while (1);
> + trace_irq_pin_config("enable", (u32)pin_out, (u32)d->hwirq, 0, on);

I'm concerned with that string you are passing in.

> spin_unlock_irqrestore(&pdc_lock, flags);
> WARN_ON(r_enable != enable);
> }
> @@ -161,6 +164,9 @@ static int qcom_pdc_gic_set_type(struct irq_data *d, unsigned int type)
> udelay(5);
> } while (1);
>
> + trace_irq_pin_config("type_config", (u32)pin_out, (u32)d->hwirq,

This one too.

> + pdc_type, 0);
> +
> /*
> * If type is edge triggered, forward that as Rising edge as PDC
> * takes care of converting falling edge to rising edge signal
> diff --git a/include/trace/events/pdc.h b/include/trace/events/pdc.h
> new file mode 100644
> index 000000000000..bec54c414880
> --- /dev/null
> +++ b/include/trace/events/pdc.h
> @@ -0,0 +1,50 @@
> +/* Copyright (c) 2017-2018, The Linux Foundation. All rights reserved.
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 and
> + * only version 2 as published by the Free Software Foundation.
> + *
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> + * GNU General Public License for more details.
> + */
> +
> +#undef TRACE_SYSTEM
> +#define TRACE_SYSTEM pdc
> +
> +#if !defined(_TRACE_PDC_) || defined(TRACE_HEADER_MULTI_READ)
> +#define _TRACE_PDC_H_
> +
> +#include <linux/tracepoint.h>
> +
> +TRACE_EVENT(irq_pin_config,
> +
> + TP_PROTO(char *func, u32 pin, u32 hwirq, u32 type, u32 enable),
> +
> + TP_ARGS(func, pin, hwirq, type, enable),
> +
> + TP_STRUCT__entry(
> + __field(char *, func)
> + __field(u32, pin)
> + __field(u32, hwirq)
> + __field(u32, type)
> + __field(u32, enable)
> + ),
> +
> + TP_fast_assign(
> + __entry->pin = pin;
> + __entry->func = func;
> + __entry->hwirq = hwirq;
> + __entry->type = type;
> + __entry->enable = enable;
> + ),
> +
> + TP_printk("%s hwirq:%u pin:%u type:%u enable:%u",

%s is dereferencing a pointer, which works fine when reading the ASCII
translated trace files (/sys/kernel/debug/tracing/trace and friends).
But it breaks when we use trace-cmd or perf. Because they record the
binary data and they do not have access to memory inside the kernel.

What I would recommend is in pdc.h have:

#define PDC_ENTRY 1
#define PDC_TYPE_CONFIG 2

Then have the tracepoints do instead of:

trace_irq_pin_config("enable", ...);
trace_irq_pin_config("type_config", ...);

Have them do:

trace_irq_pin_config(PDC_ENTRY, ...);
trace_irq_pin_config(PDC_TYPE_CONFIG, ...);

And then have:

TP_PROTO(u32 func, ...)

TP_STRUCT__entry(
__field(u32, func)
[...]

TP_fast_assign(
__entry->func = func;
[...]

TP_printk("%s hwirq...",
print_symbolic(__entry->func,
{ PDC_ENTRY, "entry" },
{ PDC_TYPE_CONFIG, "type_config" }),
[...]

And you will get the same effect, with trace-cmd and perf still working
as expected.

-- Steve


> + __entry->func, __entry->pin, __entry->hwirq, __entry->type,
> + __entry->enable)
> +);
> +
> +#endif
> +#define TRACE_INCLUDE_FILE pdc
> +#include <trace/define_trace.h>