All of lore.kernel.org
 help / color / mirror / Atom feed
From: Mark Rutland <mark.rutland@arm.com>
To: Douglas Anderson <dianders@chromium.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>, Sumit Garg <sumit.garg@linaro.org>,
	Daniel Thompson <daniel.thompson@linaro.org>,
	Marc Zyngier <maz@kernel.org>,
	linux-perf-users@vger.kernel.org, ito-yuichi@fujitsu.com,
	Chen-Yu Tsai <wens@csie.org>, Ard Biesheuvel <ardb@kernel.org>,
	Stephen Boyd <swboyd@chromium.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	linux-arm-kernel@lists.infradead.org,
	kgdb-bugreport@lists.sourceforge.net,
	Masayoshi Mizuma <msys.mizuma@gmail.com>,
	"Rafael J . Wysocki" <rafael.j.wysocki@intel.com>,
	Lecopzer Chen <lecopzer.chen@mediatek.com>,
	Ben Dooks <ben-linux@fluff.org>, Ingo Molnar <mingo@kernel.org>,
	Josh Poimboeuf <jpoimboe@kernel.org>,
	Valentin Schneider <vschneid@redhat.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v9 4/7] arm64: smp: Assign and setup the debug IPI
Date: Mon, 7 Aug 2023 11:17:45 +0100	[thread overview]
Message-ID: <ZNDEyT3mHCl0UQIV@FVFF77S0Q05N.cambridge.arm.com> (raw)
In-Reply-To: <20230601143109.v9.4.I6d7f7d5fa0aa293c8c3374194947254b93114d37@changeid>

On Thu, Jun 01, 2023 at 02:31:48PM -0700, Douglas Anderson wrote:
> From: Sumit Garg <sumit.garg@linaro.org>
> 
> All current arm64 interrupt controllers have at least 8
> IPIs. Currently we are only using 7 of them on arm64. Let's use the
> 8th one as a debug IPI. This uses the new "debug IPI" infrastructure
> which will try to allocate this IPI as an NMI/pseudo NMI if possible.
> 
> Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> Signed-off-by: Douglas Anderson <dianders@chromium.org>

I think with my suggestion on the prior patch, we don't need the additional
logic here.

> ---
> I could imagine that people object to using up the last free IPI on
> interrupt controllers with only 8 IPIs. However, it shouldn't be a big
> deal. If we later need an extra IPI, it shouldn't be too hard to
> combine some of the existing ones. Presumably we could just get rid of
> the "crash stop" IPI and have the normal "stop" IPI do the crash if
> "waiting_for_crash_ipi" is non-zero

TBH, I'd love to unify the logic for IPI_CPU_STOP and IPI_CPU_CRASH_STOP as
they have a bunch of pointless divergence.

We could also remove IPI_WAKEUP and replace that with something else (e.g.
IPI_CALL_FUNC with an empty function) in order to free up an IPI.

I'd *also* prefer to have separate IPI_CPU_BACKTRACE and IPI_CPU_DEBUG as the
backtrace logic is used for a bunch of things other than KGDB (e.g. RCU stalls,
panic), and that would clearly separate the logic for the two cases.

Thanks,
Mark.

> 
> Changes in v9:
> - Add a warning if we don't have enough IPIs for the NMI IPI
> - Renamed "NMI IPI" to "debug IPI" since it might not be backed by NMI.
> - Update commit description
> 
> Changes in v8:
> - debug_ipi_setup() and debug_ipi_teardown() no longer take cpu param
> 
>  arch/arm64/kernel/smp.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c
> index edd63894d61e..db019b49d3bd 100644
> --- a/arch/arm64/kernel/smp.c
> +++ b/arch/arm64/kernel/smp.c
> @@ -53,6 +53,8 @@
>  
>  #include <trace/events/ipi.h>
>  
> +#include "ipi_debug.h"
> +
>  DEFINE_PER_CPU_READ_MOSTLY(int, cpu_number);
>  EXPORT_PER_CPU_SYMBOL(cpu_number);
>  
> @@ -935,6 +937,8 @@ static void ipi_setup(int cpu)
>  
>  	for (i = 0; i < nr_ipi; i++)
>  		enable_percpu_irq(ipi_irq_base + i, 0);
> +
> +	debug_ipi_setup();
>  }
>  
>  #ifdef CONFIG_HOTPLUG_CPU
> @@ -947,6 +951,8 @@ static void ipi_teardown(int cpu)
>  
>  	for (i = 0; i < nr_ipi; i++)
>  		disable_percpu_irq(ipi_irq_base + i);
> +
> +	debug_ipi_teardown();
>  }
>  #endif
>  
> @@ -968,6 +974,11 @@ void __init set_smp_ipi_range(int ipi_base, int n)
>  		irq_set_status_flags(ipi_base + i, IRQ_HIDDEN);
>  	}
>  
> +	if (n > nr_ipi)
> +		set_smp_debug_ipi(ipi_base + nr_ipi);
> +	else
> +		WARN(1, "Not enough IPIs for NMI IPI\n");
> +
>  	ipi_irq_base = ipi_base;
>  
>  	/* Setup the boot CPU immediately */
> -- 
> 2.41.0.rc2.161.g9c6817b8e7-goog
> 

WARNING: multiple messages have this Message-ID (diff)
From: Mark Rutland <mark.rutland@arm.com>
To: Douglas Anderson <dianders@chromium.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>, Sumit Garg <sumit.garg@linaro.org>,
	Daniel Thompson <daniel.thompson@linaro.org>,
	Marc Zyngier <maz@kernel.org>,
	linux-perf-users@vger.kernel.org, ito-yuichi@fujitsu.com,
	Chen-Yu Tsai <wens@csie.org>, Ard Biesheuvel <ardb@kernel.org>,
	Stephen Boyd <swboyd@chromium.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	linux-arm-kernel@lists.infradead.org,
	kgdb-bugreport@lists.sourceforge.net,
	Masayoshi Mizuma <msys.mizuma@gmail.com>,
	"Rafael J . Wysocki" <rafael.j.wysocki@intel.com>,
	Lecopzer Chen <lecopzer.chen@mediatek.com>,
	Ben Dooks <ben-linux@fluff.org>, Ingo Molnar <mingo@kernel.org>,
	Josh Poimboeuf <jpoimboe@kernel.org>,
	Valentin Schneider <vschneid@redhat.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v9 4/7] arm64: smp: Assign and setup the debug IPI
Date: Mon, 7 Aug 2023 11:17:45 +0100	[thread overview]
Message-ID: <ZNDEyT3mHCl0UQIV@FVFF77S0Q05N.cambridge.arm.com> (raw)
In-Reply-To: <20230601143109.v9.4.I6d7f7d5fa0aa293c8c3374194947254b93114d37@changeid>

On Thu, Jun 01, 2023 at 02:31:48PM -0700, Douglas Anderson wrote:
> From: Sumit Garg <sumit.garg@linaro.org>
> 
> All current arm64 interrupt controllers have at least 8
> IPIs. Currently we are only using 7 of them on arm64. Let's use the
> 8th one as a debug IPI. This uses the new "debug IPI" infrastructure
> which will try to allocate this IPI as an NMI/pseudo NMI if possible.
> 
> Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
> Signed-off-by: Douglas Anderson <dianders@chromium.org>

I think with my suggestion on the prior patch, we don't need the additional
logic here.

> ---
> I could imagine that people object to using up the last free IPI on
> interrupt controllers with only 8 IPIs. However, it shouldn't be a big
> deal. If we later need an extra IPI, it shouldn't be too hard to
> combine some of the existing ones. Presumably we could just get rid of
> the "crash stop" IPI and have the normal "stop" IPI do the crash if
> "waiting_for_crash_ipi" is non-zero

TBH, I'd love to unify the logic for IPI_CPU_STOP and IPI_CPU_CRASH_STOP as
they have a bunch of pointless divergence.

We could also remove IPI_WAKEUP and replace that with something else (e.g.
IPI_CALL_FUNC with an empty function) in order to free up an IPI.

I'd *also* prefer to have separate IPI_CPU_BACKTRACE and IPI_CPU_DEBUG as the
backtrace logic is used for a bunch of things other than KGDB (e.g. RCU stalls,
panic), and that would clearly separate the logic for the two cases.

Thanks,
Mark.

> 
> Changes in v9:
> - Add a warning if we don't have enough IPIs for the NMI IPI
> - Renamed "NMI IPI" to "debug IPI" since it might not be backed by NMI.
> - Update commit description
> 
> Changes in v8:
> - debug_ipi_setup() and debug_ipi_teardown() no longer take cpu param
> 
>  arch/arm64/kernel/smp.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/arch/arm64/kernel/smp.c b/arch/arm64/kernel/smp.c
> index edd63894d61e..db019b49d3bd 100644
> --- a/arch/arm64/kernel/smp.c
> +++ b/arch/arm64/kernel/smp.c
> @@ -53,6 +53,8 @@
>  
>  #include <trace/events/ipi.h>
>  
> +#include "ipi_debug.h"
> +
>  DEFINE_PER_CPU_READ_MOSTLY(int, cpu_number);
>  EXPORT_PER_CPU_SYMBOL(cpu_number);
>  
> @@ -935,6 +937,8 @@ static void ipi_setup(int cpu)
>  
>  	for (i = 0; i < nr_ipi; i++)
>  		enable_percpu_irq(ipi_irq_base + i, 0);
> +
> +	debug_ipi_setup();
>  }
>  
>  #ifdef CONFIG_HOTPLUG_CPU
> @@ -947,6 +951,8 @@ static void ipi_teardown(int cpu)
>  
>  	for (i = 0; i < nr_ipi; i++)
>  		disable_percpu_irq(ipi_irq_base + i);
> +
> +	debug_ipi_teardown();
>  }
>  #endif
>  
> @@ -968,6 +974,11 @@ void __init set_smp_ipi_range(int ipi_base, int n)
>  		irq_set_status_flags(ipi_base + i, IRQ_HIDDEN);
>  	}
>  
> +	if (n > nr_ipi)
> +		set_smp_debug_ipi(ipi_base + nr_ipi);
> +	else
> +		WARN(1, "Not enough IPIs for NMI IPI\n");
> +
>  	ipi_irq_base = ipi_base;
>  
>  	/* Setup the boot CPU immediately */
> -- 
> 2.41.0.rc2.161.g9c6817b8e7-goog
> 

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

  reply	other threads:[~2023-08-07 10:17 UTC|newest]

Thread overview: 71+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-06-01 21:31 [PATCH v9 0/7] arm64: Add debug IPI for backtraces / kgdb; try to use NMI for it Douglas Anderson
2023-06-01 21:31 ` Douglas Anderson
2023-06-01 21:31 ` [PATCH v9 1/7] irqchip/gic-v3: Enable support for SGIs to act as NMIs Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-08-07  9:50   ` Mark Rutland
2023-08-07  9:50     ` Mark Rutland
2023-08-07 11:22     ` Sumit Garg
2023-08-07 11:22       ` Sumit Garg
2023-08-07 13:25       ` Mark Rutland
2023-08-07 13:25         ` Mark Rutland
2023-06-01 21:31 ` [PATCH v9 2/7] arm64: idle: Tag the arm64 idle functions as __cpuidle Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-08-07  9:52   ` Mark Rutland
2023-08-07  9:52     ` Mark Rutland
2023-06-01 21:31 ` [PATCH v9 3/7] arm64: Add framework for a debug IPI Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-08-07 10:12   ` Mark Rutland
2023-08-07 10:12     ` Mark Rutland
2023-08-21 22:16     ` Doug Anderson
2023-08-21 22:16       ` Doug Anderson
2023-08-22  6:42       ` Mark Rutland
2023-08-22  6:42         ` Mark Rutland
2023-06-01 21:31 ` [PATCH v9 4/7] arm64: smp: Assign and setup the " Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-08-07 10:17   ` Mark Rutland [this message]
2023-08-07 10:17     ` Mark Rutland
2023-06-01 21:31 ` [PATCH v9 5/7] arm64: ipi_debug: Add support for backtrace using " Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-08-07 10:23   ` Mark Rutland
2023-08-07 10:23     ` Mark Rutland
2023-08-22  0:06     ` Doug Anderson
2023-08-22  0:06       ` Doug Anderson
2023-08-22  6:35       ` Mark Rutland
2023-08-22  6:35         ` Mark Rutland
2023-06-01 21:31 ` [PATCH v9 6/7] kgdb: Provide a stub kgdb_nmicallback() if !CONFIG_KGDB Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-06-15 18:14   ` Doug Anderson
2023-06-15 18:14     ` Doug Anderson
2023-06-26 14:30     ` Daniel Thompson
2023-06-26 14:30       ` Daniel Thompson
2023-08-07 10:27   ` Mark Rutland
2023-08-07 10:27     ` Mark Rutland
2023-08-07 10:29     ` Mark Rutland
2023-08-07 10:29       ` Mark Rutland
2023-06-01 21:31 ` [PATCH v9 7/7] arm64: kgdb: Roundup cpus using the debug IPI Douglas Anderson
2023-06-01 21:31   ` Douglas Anderson
2023-08-07 10:28   ` Mark Rutland
2023-08-07 10:28     ` Mark Rutland
2023-08-07 10:47     ` Marc Zyngier
2023-08-07 10:47       ` Marc Zyngier
2023-08-07 10:54       ` Mark Rutland
2023-08-07 10:54         ` Mark Rutland
2023-08-07 11:08         ` Marc Zyngier
2023-08-07 11:08           ` Marc Zyngier
2023-08-07 11:13           ` Mark Rutland
2023-08-07 11:13             ` Mark Rutland
2023-08-07 15:24     ` Daniel Thompson
2023-08-07 15:24       ` Daniel Thompson
2023-08-07 16:04       ` Mark Rutland
2023-08-07 16:04         ` Mark Rutland
2023-08-08 11:17         ` Daniel Thompson
2023-08-08 11:17           ` Daniel Thompson
2023-06-02  5:19 ` [PATCH v9 0/7] arm64: Add debug IPI for backtraces / kgdb; try to use NMI for it Sumit Garg
2023-06-02  5:19   ` Sumit Garg
2023-07-24 15:55 ` Doug Anderson
2023-08-07 10:41   ` Mark Rutland
2023-08-07 10:41     ` Mark Rutland
2023-08-07 12:46     ` Sumit Garg
2023-08-07 12:46       ` Sumit Garg
2023-08-07 14:43       ` Mark Rutland
2023-08-07 14:43         ` Mark Rutland

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ZNDEyT3mHCl0UQIV@FVFF77S0Q05N.cambridge.arm.com \
    --to=mark.rutland@arm.com \
    --cc=ardb@kernel.org \
    --cc=ben-linux@fluff.org \
    --cc=catalin.marinas@arm.com \
    --cc=daniel.thompson@linaro.org \
    --cc=dianders@chromium.org \
    --cc=ito-yuichi@fujitsu.com \
    --cc=jpoimboe@kernel.org \
    --cc=kgdb-bugreport@lists.sourceforge.net \
    --cc=lecopzer.chen@mediatek.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-perf-users@vger.kernel.org \
    --cc=maz@kernel.org \
    --cc=mingo@kernel.org \
    --cc=msys.mizuma@gmail.com \
    --cc=peterz@infradead.org \
    --cc=rafael.j.wysocki@intel.com \
    --cc=sumit.garg@linaro.org \
    --cc=swboyd@chromium.org \
    --cc=tglx@linutronix.de \
    --cc=vschneid@redhat.com \
    --cc=wens@csie.org \
    --cc=will@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.