[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <53962765-53b9-dfdc-a5b2-a3133a924c12@arm.com>
Date: Thu, 21 Oct 2021 15:33:01 +0100
From: James Clark <james.clark@....com>
To: Leo Yan <leo.yan@...aro.org>,
Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will@...nel.org>,
Mark Rutland <mark.rutland@....com>,
Kees Cook <keescook@...omium.org>,
Ard Biesheuvel <ardb@...nel.org>,
Sami Tolvanen <samitolvanen@...gle.com>,
Nicholas Piggin <npiggin@...il.com>,
James Morse <james.morse@....com>,
Marc Zyngier <maz@...nel.org>, Joey Gouly <joey.gouly@....com>,
Peter Collingbourne <pcc@...gle.com>,
Vincenzo Frascino <vincenzo.frascino@....com>,
"Peter Zijlstra (Intel)" <peterz@...radead.org>,
Stephane Eranian <eranian@...gle.com>,
linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: [RFCv1 1/4] arm64: Use static key for tracing PID in CONTEXTIDR
On 21/10/2021 14:45, Leo Yan wrote:
> The kernel provides CONFIG_PID_IN_CONTEXTIDR for tracing PID into system
> register CONTEXTIDR; we need to statically enable this configuration
> when build kernel image to use this feature.
>
> On the other hand, hardware tracing modules (e.g. Arm CoreSight, SPE,
> etc) rely on this feature to provide context info in their tracing data.
> If kernel has not enabled configuration CONFIG_PID_IN_CONTEXTIDR, then
> tracing modules have no chance to capture PID related info.
>
> This patch introduces static key for tracing PID in CONTEXTIDR, it
> provides a possibility for device driver to dynamically enable and
> disable tracing PID into CONTEXTIDR as needed.
>
> As the first step, the kernel increases the static key if
> CONFIG_PID_IN_CONTEXTIDR is enabled when booting kernel, in this case
> kernel will always trace PID into CONTEXTIDR at the runtime. This means
> before and after applying this patch, the semantics for
> CONFIG_PID_IN_CONTEXTIDR are consistent.
>
> Signed-off-by: Leo Yan <leo.yan@...aro.org>
> ---
> arch/arm64/include/asm/mmu_context.h | 4 +++-
> arch/arm64/kernel/process.c | 11 +++++++++++
> 2 files changed, 14 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm64/include/asm/mmu_context.h b/arch/arm64/include/asm/mmu_context.h
> index f4ba93d4ffeb..e1f33616f83a 100644
> --- a/arch/arm64/include/asm/mmu_context.h
> +++ b/arch/arm64/include/asm/mmu_context.h
> @@ -26,9 +26,11 @@
>
> extern bool rodata_full;
>
> +DECLARE_STATIC_KEY_FALSE(contextidr_in_use);
> +
> static inline void contextidr_thread_switch(struct task_struct *next)
> {
> - if (!IS_ENABLED(CONFIG_PID_IN_CONTEXTIDR))
> + if (!static_branch_unlikely(&contextidr_in_use))
> return;
>
> write_sysreg(task_pid_nr(next), contextidr_el1);
> diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c
> index 40adb8cdbf5a..d744c0c7e4c4 100644
> --- a/arch/arm64/kernel/process.c
> +++ b/arch/arm64/kernel/process.c
> @@ -61,6 +61,9 @@ unsigned long __stack_chk_guard __ro_after_init;
> EXPORT_SYMBOL(__stack_chk_guard);
> #endif
>
> +DEFINE_STATIC_KEY_FALSE(contextidr_in_use);
> +EXPORT_SYMBOL_GPL(contextidr_in_use);
> +
> /*
> * Function pointers to optional machine specific functions
> */
> @@ -721,3 +724,11 @@ int arch_elf_adjust_prot(int prot, const struct arch_elf_state *state,
> return prot;
> }
> #endif
> +
> +static int __init contextidr_init(void)
> +{
> + if (IS_ENABLED(CONFIG_PID_IN_CONTEXTIDR))
> + static_branch_inc(&contextidr_in_use);
> + return 0;
> +}
> +early_initcall(contextidr_init);
Hi Leo,
Can you skip this early_initcall() part if you do something like this:
DECLARE_STATIC_KEY_MAYBE(CONFIG_PID_IN_CONTEXTIDR, contextidr_in_use)
It seems like there is a way to conditionally initialise it to true.
James
Powered by blists - more mailing lists