Re: [PATCH v2 2/9] x86: numa: check the node id consistently for x86

From: Yunsheng Lin <linyunsheng@huawei.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: dalias@libc.org, linux-sh@vger.kernel.org,
	catalin.marinas@arm.com, dave.hansen@linux.intel.com,
	heiko.carstens@de.ibm.com, linuxarm@huawei.com,
	jiaxun.yang@flygoat.com, linux-mips@vger.kernel.org,
	mwb@linux.vnet.ibm.com, paulus@samba.org, hpa@zytor.com,
	sparclinux@vger.kernel.org, chenhc@lemote.com, will@kernel.org,
	bp@alien8.de, linux-s390@vger.kernel.org,
	ysato@users.sourceforge.jp, mpe@ellerman.id.au, x86@kernel.org,
	rppt@linux.ibm.com, borntraeger@de.ibm.com, dledford@redhat.com,
	mingo@redhat.com, jeffrey.t.kirsher@intel.com,
	benh@kernel.crashing.org, jhogan@kernel.org,
	nfont@linux.vnet.ibm.com, mattst88@gmail.com,
	len.brown@intel.com, gor@linux.ibm.com,
	anshuman.khandual@arm.com, ink@jurassic.park.msu.ru,
	luto@kernel.org, tglx@linutronix.de,
	naveen.n.rao@linux.vnet.ibm.com,
	linux-arm-kernel@lists.infradead.org, rth@twiddle.net,
	axboe@kernel.dk, linuxppc-dev@lists.ozlabs.org,
	linux-kernel@vger.kernel.org, ralf@linux-mips.org,
	tbogendoerfer@suse.de, paul.burton@mips.com,
	linux-alpha@vger.kernel.org, cai@lca.pw,
	akpm@linux-foundation.org, robin.murphy@arm.com,
	davem@davemloft.net
Subject: Re: [PATCH v2 2/9] x86: numa: check the node id consistently for x86
Date: Mon, 02 Sep 2019 12:25:24 +0000	[thread overview]
Message-ID: <5fa2aa99-89fa-cd41-b090-36a23cfdeb73@huawei.com> (raw)
In-Reply-To: <20190902072542.GN2369@hirez.programming.kicks-ass.net>

On 2019/9/2 15:25, Peter Zijlstra wrote:
> On Mon, Sep 02, 2019 at 01:46:51PM +0800, Yunsheng Lin wrote:
>> On 2019/9/1 0:12, Peter Zijlstra wrote:
> 
>>> 1) because even it is not set, the device really does belong to a node.
>>> It is impossible a device will have magic uniform access to memory when
>>> CPUs cannot.
>>
>> So it means dev_to_node() will return either NUMA_NO_NODE or a
>> valid node id?
> 
> NUMA_NO_NODE := -1, which is not a valid node number. It is also, like I
> said, not a valid device location on a NUMA system.
> 
> Just because ACPI/BIOS is shit, doesn't mean the device doesn't have a
> node association. It just means we don't know and might have to guess.

How do we guess the device's location when ACPI/BIOS does not set it?

It seems dev_to_node() does not do anything about that and leave the
job to the caller or whatever function that get called with its return
value, such as cpumask_of_node().

> 
>>> 2) is already true today, cpumask_of_node() requires a valid node_id.
>>
>> Ok, most of the user does check node_id before calling
>> cpumask_of_node(), but does a little different type of checking:
>>
>> 1) some does " < 0" check;
>> 2) some does "= NUMA_NO_NODE" check;
>> 3) some does ">= MAX_NUMNODES" check;
>> 4) some does "< 0 || >= MAX_NUMNODES || !node_online(node)" check.
> 
> The one true way is:
> 
> 	'(unsigned)node_id >= nr_node_ids'

I missed the magic of the "unsigned" in your previous reply.

> 
>>> 3) is just wrong and increases overhead for everyone.
>>
>> Ok, cpumask_of_node() is also used in some critical path such
>> as scheduling, which may not need those checking, the overhead
>> is unnecessary.
>>
>> But for non-critical path such as setup or configuration path,
>> it better to have consistent checking, and also simplify the
>> user code that calls cpumask_of_node().
>>
>> Do you think it is worth the trouble to add a new function
>> such as cpumask_of_node_check(maybe some other name) to do
>> consistent checking?
>>
>> Or caller just simply check if dev_to_node()'s return value is
>> NUMA_NO_NODE before calling cpumask_of_node()?
> 
> It is not a matter of convenience. The function is called
> cpumask_of_node(), when node < 0 || node >= nr_node_ids, it is not a
> valid node, therefore the function shouldn't return anything except an
> error.
what do you mean by error? What I can think is three type of errors:
1) return NULL, this way it seems cpumask_of_node() also leave the
   job to the function that calls it.
2) cpu_none_mask, I am not sure what this means, maybe it means there
   is no cpu on the same node with the device?
3) give a warning, stack dump, or even a BUG_ON?

I would prefer the second one, and implement the third one when the
CONFIG_DEBUG_PER_CPU_MAPS is selected.

Any suggestion?

> 
> Also note that the CONFIG_DEBUG_PER_CPU_MAPS version of
> cpumask_of_node() already does this (although it wants the below fix).

Thanks for the note and example.