Re: [PATCH] semaphore: Add might_sleep() to down_*() family

From: Will Deacon <will@kernel.org>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: Guenter Roeck <linux@roeck-us.net>,
	Xiaoming Ni <nixiaoming@huawei.com>,
	linux-kernel@vger.kernel.org, peterz@infradead.org,
	mingo@redhat.com, longman@redhat.com, boqun.feng@gmail.com,
	wangle6@huawei.com, xiaoqian9@huawei.com, shaolexi@huawei.com,
	linux-acpi@vger.kernel.org,
	Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>,
	Hanjun Guo <guohanjun@huawei.com>,
	Sudeep Holla <sudeep.holla@arm.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Subject: Re: [PATCH] semaphore: Add might_sleep() to down_*() family
Date: Tue, 31 Aug 2021 13:34:01 +0100	[thread overview]
Message-ID: <20210831123401.GA31712@willie-the-truck> (raw)
In-Reply-To: <871r69ersb.ffs@tglx>

On Tue, Aug 31, 2021 at 02:13:08PM +0200, Thomas Gleixner wrote:
> On Tue, Aug 31 2021 at 04:13, Guenter Roeck wrote:
> > On Mon, Aug 09, 2021 at 10:12:15AM +0800, Xiaoming Ni wrote:
> >> Semaphore is sleeping lock. Add might_sleep() to down*() family
> >> (with exception of down_trylock()) to detect atomic context sleep.
> >> 
> >> Previously discussed with Peter Zijlstra, see link:
> >>  https://lore.kernel.org/lkml/20210806082320.GD22037@worktop.programming.kicks-ass.net
> >> 
> >> Signed-off-by: Xiaoming Ni <nixiaoming@huawei.com>
> >> Acked-by: Will Deacon <will@kernel.org>
> >
> > This patch results in the following traceback on all arm64 boots with
> > EFI BIOS.
> 
> That's what this change was supposed to catch :)
> 
> > The problem is only seen with CONFIG_ACPI_PPTT=y, and thus only on arm64.
> 
> The below should fix this.
> 
> Thanks,
> 
>         tglx
> ---
> Subject: drivers: base: cacheinfo: Get rid of DEFINE_SMP_CALL_CACHE_FUNCTION()
> From: Thomas Gleixner <tglx@linutronix.de>
> Date: Tue, 31 Aug 2021 13:48:34 +0200
> 
> DEFINE_SMP_CALL_CACHE_FUNCTION() was usefel before the CPU hotplug rework

typo: "usefel"

> to ensure that the cache related functions are called on the upcoming CPU
> because the notifier itself could run on any online CPU.
> 
> The hotplug state machine guarantees that the callbacks are invoked on the
> upcoming CPU. So there is no need to have this SMP function call
> obfuscation. That indirection was missed when the hotplug notifiers were
> converted.
> 
> This also solves the problem of ARM64 init_cache_level() invoking ACPI
> functions which take a semaphore in that context. That's invalid as SMP
> function calls run with interrupts disabled. Running it just from the
> callback in context of the CPU hotplug thread solves this.
>  
> Reported-by: Guenter Roeck <linux@roeck-us.net>
> Fixes: 8571890e1513 ("arm64: Add support for ACPI based firmware tables")
> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> ---
>  arch/arm64/kernel/cacheinfo.c   |    7 ++-----
>  arch/mips/kernel/cacheinfo.c    |    7 ++-----
>  arch/riscv/kernel/cacheinfo.c   |    7 ++-----
>  arch/x86/kernel/cpu/cacheinfo.c |    7 ++-----
>  include/linux/cacheinfo.h       |   18 ------------------
>  5 files changed, 8 insertions(+), 38 deletions(-)
> 
> --- a/arch/arm64/kernel/cacheinfo.c
> +++ b/arch/arm64/kernel/cacheinfo.c
> @@ -43,7 +43,7 @@ static void ci_leaf_init(struct cacheinf
>  	this_leaf->type = type;
>  }
>  
> -static int __init_cache_level(unsigned int cpu)
> +int init_cache_level(unsigned int cpu)
>  {
>  	unsigned int ctype, level, leaves, fw_level;
>  	struct cpu_cacheinfo *this_cpu_ci = get_cpu_cacheinfo(cpu);
> @@ -78,7 +78,7 @@ static int __init_cache_level(unsigned i
>  	return 0;
>  }
>  
> -static int __populate_cache_leaves(unsigned int cpu)
> +int populate_cache_leaves(unsigned int cpu)
>  {
>  	unsigned int level, idx;
>  	enum cache_type type;
> @@ -97,6 +97,3 @@ static int __populate_cache_leaves(unsig
>  	}
>  	return 0;
>  }
> -
> -DEFINE_SMP_CALL_CACHE_FUNCTION(init_cache_level)
> -DEFINE_SMP_CALL_CACHE_FUNCTION(populate_cache_leaves)

Glad to see the back of this:

Acked-by: Will Deacon <will@kernel.org>

Will