linux-kernel - Re: [PATCH 1/2] memory tier: Introduce sysfs for tier interleave weights.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20231002112622.0000220a@Huawei.com>
Date:   Mon, 2 Oct 2023 11:26:22 +0100
From:   Jonathan Cameron <Jonathan.Cameron@...wei.com>
To:     Ravi Jonnalagadda <ravis.opensrc@...ron.com>
CC:     <linux-mm@...r.kernel.org>, <linux-cxl@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>, <linux-arch@...r.kernel.org>,
        <linux-api@...r.kernel.org>, <luto@...nel.org>,
        <tglx@...utronix.de>, <mingo@...hat.com>, <bp@...en8.de>,
        <dietmar.eggemann@....com>, <vincent.guittot@...aro.org>,
        <dave.hansen@...ux.intel.com>, <hpa@...or.com>, <arnd@...db.de>,
        <akpm@...ux-foundation.org>, <x86@...nel.org>,
        <aneesh.kumar@...ux.ibm.com>, <gregory.price@...verge.com>,
        <ying.huang@...el.com>, <jgroves@...ron.com>,
        <sthanneeru@...ron.com>, <emirakhur@...ron.com>,
        <vtanna@...ron.com>
Subject: Re: [PATCH 1/2] memory tier: Introduce sysfs for tier interleave
 weights.

On Wed, 27 Sep 2023 15:20:01 +0530
Ravi Jonnalagadda <ravis.opensrc@...ron.com> wrote:

> From: Srinivasulu Thanneeru <sthanneeru@...ron.com>
> 
> Allocating pages across tiers is accomplished by provisioning
> interleave weights for each tier, with the distribution based on
> these weight values.
> By default, all tiers will have a weight of 1, which means
> default standard page allocation. By default all nodes within
> tier will have weight of 1.
> 
> Signed-off-by: Srinivasulu Thanneeru <sthanneeru@...ron.com>
> Co-authored-by: Ravi Jonnalagadda <ravis.opensrc@...ron.com>

ABI docs?  

Documentation/ABI/testing/sysfs-kernel-mm-memory-tiers

A few trivial comments inline.
> ---
>  include/linux/memory-tiers.h |  2 ++
>  mm/memory-tiers.c            | 46 +++++++++++++++++++++++++++++++++++-
>  2 files changed, 47 insertions(+), 1 deletion(-)
> 
> diff --git a/include/linux/memory-tiers.h b/include/linux/memory-tiers.h
> index 437441cdf78f..c62d286749d0 100644
> --- a/include/linux/memory-tiers.h
> +++ b/include/linux/memory-tiers.h
> @@ -19,6 +19,8 @@
>   */
>  #define MEMTIER_ADISTANCE_DRAM	((4 * MEMTIER_CHUNK_SIZE) + (MEMTIER_CHUNK_SIZE >> 1))
>  
> +#define MAX_TIER_INTERLEAVE_WEIGHT 100
> +
>  struct memory_tier;
>  struct memory_dev_type {
>  	/* list of memory types that are part of same tier as this type */
> diff --git a/mm/memory-tiers.c b/mm/memory-tiers.c
> index 37a4f59d9585..7e06c9e0fa41 100644
> --- a/mm/memory-tiers.c
> +++ b/mm/memory-tiers.c
> @@ -13,6 +13,11 @@ struct memory_tier {
>  	struct list_head list;
>  	/* list of all memory types part of this tier */
>  	struct list_head memory_types;
> +	/*
> +	 * By default all tiers will have weight as 1, which means they
> +	 * follow default standard allocation.
> +	 */
> +	unsigned short interleave_weight;

If you are going to use fixed size, keep it going.
u16 (u8 as per below comment probably makes more sense)


>  	/*
>  	 * start value of abstract distance. memory tier maps
>  	 * an abstract distance  range,
> @@ -145,8 +150,45 @@ static ssize_t nodelist_show(struct device *dev,
>  }
>  static DEVICE_ATTR_RO(nodelist);
>  
> +static ssize_t interleave_weight_show(struct device *dev,
> +				      struct device_attribute *attr, char *buf)
> +{
> +	int ret;
> +	struct memory_tier *tier = to_memory_tier(dev);
> +
> +	mutex_lock(&memory_tier_lock);
> +	ret = sysfs_emit(buf, "%u\n", tier->interleave_weight);
> +	mutex_unlock(&memory_tier_lock);

For this one

	guard(mutex)(&memory_tier_lock);
	return sysfs_emit()...

would perhaps be slightly nicer
 (see below)

> +
> +	return ret;
> +}
> +
> +static ssize_t interleave_weight_store(struct device *dev,
> +				       struct device_attribute *attr,
> +				       const char *buf, size_t size)
> +{
> +	unsigned short value;
> +	int ret;
> +	struct memory_tier *tier = to_memory_tier(dev);
> +
> +	ret = kstrtou16(buf, 0, &value);

Why u16?  Max is 100.  I'd not mind if you just put it in an
unsigned int, but seems odd to chose a specific size and
pick one that is twice as big as needed!

> +
> +	if (ret)
> +		return ret;
> +	if (value > MAX_TIER_INTERLEAVE_WEIGHT)
> +		return -EINVAL;
> +
> +	mutex_lock(&memory_tier_lock);

You could play with the new cleanup.h	 toys though it doesn't save a lot here.

	scoped_guard(mutex)(&memory_tier_lock)
		tier->interleave_weight = value;

> +	tier->interleave_weight = value;
> +	mutex_unlock(&memory_tier_lock);
> +
> +	return size;
> +}
> +static DEVICE_ATTR_RW(interleave_weight);
> +
>  static struct attribute *memtier_dev_attrs[] = {
>  	&dev_attr_nodelist.attr,
> +	&dev_attr_interleave_weight.attr,
>  	NULL
>  };
>  
> @@ -489,8 +531,10 @@ static struct memory_tier *set_node_memory_tier(int node)
>  	memtype = node_memory_types[node].memtype;
>  	node_set(node, memtype->nodes);
>  	memtier = find_create_memory_tier(memtype);
> -	if (!IS_ERR(memtier))
> +	if (!IS_ERR(memtier)) {
>  		rcu_assign_pointer(pgdat->memtier, memtier);
> +		memtier->interleave_weight = 1;
> +	}
>  	return memtier;
>  }
>