[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c0cbdd85-9702-40ab-bc97-b51b02b9fc89@linux.alibaba.com>
Date: Mon, 23 Jun 2025 11:04:40 +0800
From: Shuai Xue <xueshuai@...ux.alibaba.com>
To: Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>,
Lukas Wunner <lukas@...ner.de>, rostedt@...dmis.org
Cc: rostedt@...dmis.org, linux-pci@...r.kernel.org,
LKML <linux-kernel@...r.kernel.org>, linux-edac@...r.kernel.org,
linux-trace-kernel@...r.kernel.org, helgaas@...nel.org, bhelgaas@...gle.com,
tony.luck@...el.com, bp@...en8.de, mhiramat@...nel.org,
mathieu.desnoyers@...icios.com, oleg@...hat.com, naveen@...nel.org,
davem@...emloft.net, anil.s.keshavamurthy@...el.com, mark.rutland@....com,
peterz@...radead.org, tianruidong@...ux.alibaba.com
Subject: Re: [PATCH v8] PCI: hotplug: Add a generic RAS tracepoint for hotplug
event
在 2025/6/2 14:30, Ilpo Järvinen 写道:
> On Mon, 12 May 2025, Shuai Xue wrote:
>
>> Hotplug events are critical indicators for analyzing hardware health,
>> particularly in AI supercomputers where surprise link downs can
>> significantly impact system performance and reliability.
>>
>> To this end, define a new TRACING_SYSTEM named pci, add a generic RAS
>> tracepoint for hotplug event to help healthy check, and generate
>> tracepoints for pcie hotplug event. Add enum pci_hotplug_event in
>> include/uapi/linux/pci.h so applications like rasdaemon can register
>> tracepoint event handlers for it.
>>
>> The output like below:
>>
>> $ echo 1 > /sys/kernel/debug/tracing/events/pci/pci_hp_event/enable
>> $ cat /sys/kernel/debug/tracing/trace_pipe
>> <...>-206 [001] ..... 40.373870: pci_hp_event: 0000:00:02.0 slot:10, event:Link Down
>>
>> <...>-206 [001] ..... 40.374871: pci_hp_event: 0000:00:02.0 slot:10, event:Card not present
>>
>> Suggested-by: Lukas Wunner <lukas@...ner.de>
>> Suggested-by: Steven Rostedt <rostedt@...dmis.org>
>> Signed-off-by: Shuai Xue <xueshuai@...ux.alibaba.com>
>> Reviewed-by: Lukas Wunner <lukas@...ner.de>
>> Reviewed-by: Jonathan Cameron <Jonathan.Cameron@...wei.com>
>> ---
>> changes since v7:
>> - replace the TRACE_INCLUDE_PATH to avoid macro conflict per Steven
>> - pick up Reviewed-by from Lukas Wunner
>> ---
>> drivers/pci/hotplug/Makefile | 3 ++
>> drivers/pci/hotplug/pciehp_ctrl.c | 33 ++++++++++++---
>> drivers/pci/hotplug/trace.h | 68 +++++++++++++++++++++++++++++++
>> include/uapi/linux/pci.h | 7 ++++
>> 4 files changed, 105 insertions(+), 6 deletions(-)
>> create mode 100644 drivers/pci/hotplug/trace.h
>>
>> diff --git a/drivers/pci/hotplug/Makefile b/drivers/pci/hotplug/Makefile
>> index 40aaf31fe338..a1a9d1e98962 100644
>> --- a/drivers/pci/hotplug/Makefile
>> +++ b/drivers/pci/hotplug/Makefile
>> @@ -3,6 +3,9 @@
>> # Makefile for the Linux kernel pci hotplug controller drivers.
>> #
>>
>> +# define_trace.h needs to know how to find our header
>> +CFLAGS_pciehp_ctrl.o := -I$(src)
>> +
>> obj-$(CONFIG_HOTPLUG_PCI) += pci_hotplug.o
>> obj-$(CONFIG_HOTPLUG_PCI_COMPAQ) += cpqphp.o
>> obj-$(CONFIG_HOTPLUG_PCI_IBM) += ibmphp.o
>> diff --git a/drivers/pci/hotplug/pciehp_ctrl.c b/drivers/pci/hotplug/pciehp_ctrl.c
>> index d603a7aa7483..f9beb4d3a9b8 100644
>> --- a/drivers/pci/hotplug/pciehp_ctrl.c
>> +++ b/drivers/pci/hotplug/pciehp_ctrl.c
>> @@ -23,6 +23,9 @@
>> #include "../pci.h"
>> #include "pciehp.h"
>>
>> +#define CREATE_TRACE_POINTS
>> +#include "trace.h"
>
> Hi,
>
> Instead of spreading tracepoint creating into subdriver code like this,
> should we place it into one place, e.g., drivers/pci/pci-trace.c (which is
> what I seem to have used in my yet to be submitted patch that adds
> tracepoints into bwctrl link speed events)?
>
>> diff --git a/drivers/pci/hotplug/trace.h b/drivers/pci/hotplug/trace.h
>> new file mode 100644
>> index 000000000000..21329c198019
>> --- /dev/null
>> +++ b/drivers/pci/hotplug/trace.h
>
> Perhaps in include/trace/events/pci.h (or
> include/trace/events/pci-hotplug.h)?
>
> I don't know what is the general rule having them inside drivers/ vs
> include/trace/events, Documentation/trace/tracepoints.rst only mentions
> the latter, but there seems to be plenty under drivers/ too.
>
Hi, Ilpo,
I don't know either.
I think it is tracepoint stuff and it is up to @Steve.
@Steve, which way do you prefer?
Thanks for help.
Best Regards,
Shuai
Powered by blists - more mailing lists