linux-arm-kernel.lists.infradead.org archive mirror
 help / color / mirror / Atom feed
From: Greg KH <gregkh@linuxfoundation.org>
To: Kohei Tarumizu <tarumizu.kohei@fujitsu.com>
Cc: rafael@kernel.org, peterz@infradead.org, catalin.marinas@arm.com,
	linus.walleij@linaro.org, dave.hansen@linux.intel.com,
	conor.dooley@microchip.com, hpa@zytor.com, will@kernel.org,
	mchehab+huawei@kernel.org, x86@kernel.org, mingo@redhat.com,
	eugenis@google.com, arnd@arndb.de, bp@alien8.de,
	tglx@linutronix.de, pcc@google.com,
	linux-arm-kernel@lists.infradead.org, marcos@orca.pet,
	tony.luck@intel.com, marcan@marcan.st,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 1/8] drivers: base: Add hardware prefetch control core driver
Date: Wed, 18 May 2022 09:09:12 +0200	[thread overview]
Message-ID: <YoSbmA+5m1Le0XGe@kroah.com> (raw)
In-Reply-To: <20220518063032.2377351-2-tarumizu.kohei@fujitsu.com>

On Wed, May 18, 2022 at 03:30:25PM +0900, Kohei Tarumizu wrote:
> Adds a register/unregister function to provide sysfs interface to
> control CPU's hardware prefetch behavior. It creates the
> "prefetch_control" sysfs directory and some attributes.
> 
> Attributes are hardware dependent, so it must be implemented for each
> hardware. If CPU has a hardware prefetch behavior, call this function
> to create sysfs.
> 
> Following patches add support for A64FX and x86.
> 
> Signed-off-by: Kohei Tarumizu <tarumizu.kohei@fujitsu.com>
> ---
>  drivers/base/pfctl.c  | 180 ++++++++++++++++++++++++++++++++++++++++++
>  include/linux/pfctl.h |  14 ++++
>  2 files changed, 194 insertions(+)
>  create mode 100644 drivers/base/pfctl.c
>  create mode 100644 include/linux/pfctl.h
> 
> diff --git a/drivers/base/pfctl.c b/drivers/base/pfctl.c
> new file mode 100644
> index 000000000000..08ee8faaf277
> --- /dev/null
> +++ b/drivers/base/pfctl.c
> @@ -0,0 +1,180 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Hardware prefetch control support via sysfs.
> + *
> + * Copyright 2022 FUJITSU LIMITED
> + *
> + * See Documentation/ABI/testing/sysfs-devices-system-cpu for more information.
> + */
> +
> +#include <linux/cacheinfo.h>
> +#include <linux/cpu.h>
> +#include <linux/device.h>
> +#include <linux/pfctl.h>
> +#include <linux/slab.h>
> +
> +#ifdef pr_fmt
> +#undef pr_fmt
> +#endif
> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> +
> +const struct pfctl_group *pgroups;
> +enum cpuhp_state hp_online;
> +
> +static struct device_attribute **
> +get_pfctl_attribute(unsigned int level, enum cache_type type)
> +{
> +	int i;
> +
> +	for (i = 0; pgroups[i].attrs; i++)
> +		if ((level == pgroups[i].level) && (type == pgroups[i].type))
> +			return pgroups[i].attrs;
> +
> +	return NULL;
> +}
> +
> +static int remove_pfctl_attr(struct device *index_dev, void *data)
> +{
> +	struct cacheinfo *leaf = dev_get_drvdata(index_dev);
> +	struct device_attribute **attrs;
> +	struct device *pfctl_dev;
> +	int i;
> +
> +	attrs = get_pfctl_attribute(leaf->level, leaf->type);
> +	if (!attrs)
> +		return 0;
> +
> +	pfctl_dev = device_find_child_by_name(index_dev, "prefetch_control");
> +	if (!pfctl_dev)
> +		return 0;
> +
> +	for (i = 0; attrs[i]; i++)
> +		device_remove_file(pfctl_dev, attrs[i]);

This feels wrong, attributes should be groups and be automatically added
and removed by the driver core that way.  Not as lists of attributes
like this, as that will race and be wrong.

Use a list of attribute groups please.

> +
> +	device_unregister(pfctl_dev);
> +	put_device(pfctl_dev);
> +
> +	pfctl_dev = NULL;
> +
> +	return 0;
> +}
> +
> +static int create_pfctl_attr(struct device *index_dev, void *data)
> +{
> +	struct cacheinfo *leaf = dev_get_drvdata(index_dev);
> +	struct device_attribute **attrs;
> +	struct device *pfctl_dev;
> +	int i, err;
> +
> +	attrs = get_pfctl_attribute(leaf->level, leaf->type);
> +	if (!attrs)
> +		return 0;
> +
> +	pfctl_dev = cpu_device_create(index_dev, NULL, NULL,
> +				      "prefetch_control");
> +	if (IS_ERR(pfctl_dev))
> +		return PTR_ERR(pfctl_dev);
> +
> +	for (i = 0; attrs[i]; i++) {
> +		err = device_create_file(pfctl_dev, attrs[i]);

You just raced with userspace and lost :(

Please use attribute groups instead of manually adding files after the
device is created and userspace is notified that it was present.

That also makes your clean up logic much simpler (i.e. none as the drive
core did it for you already.)

> +		if (err) {
> +			while (--i >= 0)
> +				device_remove_file(pfctl_dev, attrs[i]);
> +
> +			device_unregister(pfctl_dev);
> +			pfctl_dev = NULL;
> +
> +			return err;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +static int pfctl_online(unsigned int cpu)
> +{
> +	struct device *cpu_dev = get_cpu_device(cpu);
> +	struct device *cache_dev;
> +	int ret;
> +
> +	cache_dev = device_find_child_by_name(cpu_dev, "cache");
> +	if (!cache_dev)
> +		return -ENODEV;
> +
> +	ret = device_for_each_child(cache_dev, NULL, create_pfctl_attr);
> +	if (ret < 0)
> +		device_for_each_child(cache_dev, NULL, remove_pfctl_attr);
> +
> +	put_device(cache_dev);
> +
> +	return ret;
> +}
> +
> +static int pfctl_prepare_down(unsigned int cpu)
> +{
> +	struct device *cpu_dev = get_cpu_device(cpu);
> +	struct device *cache_dev;
> +
> +	cache_dev = device_find_child_by_name(cpu_dev, "cache");
> +	if (!cache_dev)
> +		return 0;
> +
> +	device_for_each_child(cache_dev, NULL, remove_pfctl_attr);
> +
> +	put_device(cache_dev);
> +
> +	return 0;
> +}
> +
> +/**
> + * pfctl_register_attrs - register a Hardware Prefetch Control attributes
> + * @pfctl_groups: pfctl_groups contains device attribute group to control the
> + *                hardware prefetch register.
> + *
> + * Note: Call this function after the cache device is initialized because it
> + * requires access to the cache device. (e.g. Call at the late_initcall)
> + *
> + * Context: Any context.
> + * Return: 0 on success, negative error code on failure.
> + */
> +int pfctl_register_attrs(const struct pfctl_group *pfctl_groups)
> +{
> +	int ret;
> +
> +	if (pgroups)
> +		return -EEXIST;
> +
> +	pgroups = pfctl_groups;
> +
> +	ret = cpuhp_setup_state(CPUHP_AP_ONLINE_DYN, "base/pfctl:online",
> +				pfctl_online, pfctl_prepare_down);
> +	if (ret < 0) {
> +		pr_err("failed to register hotplug callbacks\n");
> +		pgroups = NULL;
> +		return ret;
> +	}
> +
> +	hp_online = ret;
> +
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(pfctl_register_attrs);
> +
> +/**
> + * pfctl_unregister_attrs - unregister the Hardware Prefetch Control driver
> + * @pfctl_groups: Used to verify that this function is called by the same driver
> + *                that called pfctl_register_attrs.
> + *
> + * Context: Any context.
> + * Return: nothing.
> + */
> +void pfctl_unregister_attrs(const struct pfctl_group *pfctl_groups)
> +{
> +	if (!pgroups || (pfctl_groups != pgroups))
> +		return;
> +
> +	cpuhp_remove_state(hp_online);
> +
> +	pgroups = NULL;
> +}
> +EXPORT_SYMBOL_GPL(pfctl_unregister_attrs);
> diff --git a/include/linux/pfctl.h b/include/linux/pfctl.h
> new file mode 100644
> index 000000000000..ecdab78be09f
> --- /dev/null
> +++ b/include/linux/pfctl.h
> @@ -0,0 +1,14 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +#ifndef _LINUX_PFCTL_H
> +#define _LINUX_PFCTL_H
> +
> +struct pfctl_group {
> +	unsigned int		level;
> +	enum cache_type		type;
> +	struct device_attribute	**attrs;

Attribute groups please.

> +};
> +
> +int pfctl_register_attrs(const struct pfctl_group *pfctl_groups);
> +void pfctl_unregister_attrs(const struct pfctl_group *pfctl_groups);

Why do you think this needs to be in the driver core?  Why isn't this
just a normal cpu driver?

thanks,

greg k-h

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

  reply	other threads:[~2022-05-18  7:10 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-05-18  6:30 [PATCH v4 0/8] Add hardware prefetch control driver for A64FX and x86 Kohei Tarumizu
2022-05-18  6:30 ` [PATCH v4 1/8] drivers: base: Add hardware prefetch control core driver Kohei Tarumizu
2022-05-18  7:09   ` Greg KH [this message]
2022-05-18 12:38     ` tarumizu.kohei
2022-05-18  6:30 ` [PATCH v4 2/8] drivers: base: Add Kconfig/Makefile to build " Kohei Tarumizu
2022-05-18  7:04   ` Greg KH
2022-05-20  6:42     ` tarumizu.kohei
2022-05-18  6:30 ` [PATCH v4 3/8] soc: fujitsu: Add hardware prefetch control support for A64FX Kohei Tarumizu
2022-05-18  7:10   ` Greg KH
2022-05-20  7:06     ` tarumizu.kohei
2022-05-18  6:30 ` [PATCH v4 4/8] soc: fujitsu: Add Kconfig/Makefile to build hardware prefetch control driver Kohei Tarumizu
2022-05-18  6:30 ` [PATCH v4 5/8] arm64: Create cache sysfs directory without ACPI PPTT for hardware prefetch control Kohei Tarumizu
2022-05-18  7:09   ` Greg KH
2022-05-20  7:00     ` tarumizu.kohei
2022-05-18  6:30 ` [PATCH v4 6/8] x86: Add hardware prefetch control support for x86 Kohei Tarumizu
2022-05-18  6:43   ` Greg KH
2022-05-20  6:30     ` tarumizu.kohei
2022-05-18  6:44   ` Greg KH
2022-05-20  6:40     ` tarumizu.kohei
2022-05-18  6:30 ` [PATCH v4 7/8] x86: Add Kconfig/Makefile to build hardware prefetch control driver Kohei Tarumizu
2022-05-18  6:43   ` Greg KH
2022-05-20  6:35     ` tarumizu.kohei
2022-05-18  6:30 ` [PATCH v4 8/8] docs: ABI: Add sysfs documentation interface of " Kohei Tarumizu
2022-05-19  8:29 ` [PATCH v4 0/8] Add hardware prefetch control driver for A64FX and x86 Hector Martin
2022-05-20  8:31   ` tarumizu.kohei

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=YoSbmA+5m1Le0XGe@kroah.com \
    --to=gregkh@linuxfoundation.org \
    --cc=arnd@arndb.de \
    --cc=bp@alien8.de \
    --cc=catalin.marinas@arm.com \
    --cc=conor.dooley@microchip.com \
    --cc=dave.hansen@linux.intel.com \
    --cc=eugenis@google.com \
    --cc=hpa@zytor.com \
    --cc=linus.walleij@linaro.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=marcan@marcan.st \
    --cc=marcos@orca.pet \
    --cc=mchehab+huawei@kernel.org \
    --cc=mingo@redhat.com \
    --cc=pcc@google.com \
    --cc=peterz@infradead.org \
    --cc=rafael@kernel.org \
    --cc=tarumizu.kohei@fujitsu.com \
    --cc=tglx@linutronix.de \
    --cc=tony.luck@intel.com \
    --cc=will@kernel.org \
    --cc=x86@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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).