From: Guenter Roeck <linux@roeck-us.net>
To: Jean Delvare <jdelvare@suse.de>
Cc: Jonathan Cameron <jic23@kernel.org>,
Zhang Rui <rui.zhang@intel.com>,
Eduardo Valentin <edubezval@gmail.com>,
Punit Agrawal <punit.agrawal@arm.com>,
linux-pm@vger.kernel.org, linux-iio@vger.kernel.org,
linux-hwmon@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3 2/9] hwmon: (core) New hwmon registration API
Date: Sun, 16 Oct 2016 11:20:21 -0700 [thread overview]
Message-ID: <a3b080a0-c929-7e72-3540-e49991512be5@roeck-us.net> (raw)
In-Reply-To: <20161007143213.0ba633fe@endymion>
Hi Jean,
On 10/07/2016 05:32 AM, Jean Delvare wrote:
[ ... ]
>
>>>> + name = (char *)template;
>>>> + } else {
>>>> + name = devm_kzalloc(dev, strlen(template) + 16, GFP_KERNEL);
>>>> + if (!name)
>>>> + return ERR_PTR(-ENOMEM);
>>>> + scnprintf(name, strlen(template) + 16, template,
>>>> + index + hwmon_attr_base(type));
>>>> + }
>>>> +
>>>> + hattr = devm_kzalloc(dev, sizeof(*hattr), GFP_KERNEL);
>>>> + if (!hattr)
>>>> + return ERR_PTR(-ENOMEM);
>>>
>>> So basically you are doing 1 or 2 memory allocations for every
>>> attribute? Looks quite bad from a memory fragmentation
>>> perspective :-( Not to mention performance. Given that you have all the
>>> data at hand before you start, can't you preallocate an array with the
>>> right number of hattr and pick from it? That would at least solve half
>>> of the problem.
>
> FTR I took a quick look at the iio code and there seems to be something
> like the idea above implemented in iio_device_register_sysfs(). But
> attributes themselves as instantiated by iio_device_register_sysfs()
> are still allocated individually. But hey I'm not familiar with the iio
> code anyway, I'm sure you know it better than I do.
>
>>> Something similar for string allocation may work too, although it's
>>> somewhat more complex due to the variable length. But I think the
>>> abituguru3 driver is doing it right, so maybe you can too.
>>
>> I'll look into it.
>
Merging name allocation and hattr allocation turned out to be easy by adding
the name to struct hwmon_device_attribute. However, allocating an array of
struct hwmon_device_attribute for all attributes in one go is more difficult.
Main issue is that we _don't_ really know how many attributes are going to be
created; we only know the ceiling, which is calculated in __hwmon_create_attrs().
That number is currently used to allocate the array of attributes. The extra
pointers did not seem that important there (to me). However, allocating an
array of struct hwmon_device_attribute would result in memory being allocated
for non-existing attributes as well (ie those for which is_visible returns 0),
and that would at least potentially be much more. We can either accept that,
or I would have to implement a second pass over the attributes to determine
how many are actually implemented, or we could leave the per-attribute
allocation unchanged.
Thoughts ?
Thanks,
Guenter
next prev parent reply other threads:[~2016-10-16 18:20 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-07-25 3:32 [PATCH v3 0/9] hwmon: " Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 1/9] hwmon: (core) Order include files alphabetically Guenter Roeck
2016-09-27 21:13 ` Jean Delvare
2016-07-25 3:32 ` [PATCH v3 2/9] hwmon: (core) New hwmon registration API Guenter Roeck
2016-08-12 8:16 ` Keerthy
2016-08-12 12:40 ` Guenter Roeck
2016-10-04 8:45 ` Jean Delvare
2016-10-04 19:37 ` Guenter Roeck
2016-10-07 12:32 ` Jean Delvare
2016-10-10 23:48 ` Guenter Roeck
2016-10-11 12:29 ` Jean Delvare
2016-10-16 18:20 ` Guenter Roeck [this message]
2016-07-25 3:32 ` [PATCH v3 3/9] hwmon: (core) Add voltage attribute support to new API Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 4/9] hwmon: (core) Add current " Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 5/9] hwmon: (core) Add power " Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 6/9] hwmon: (core) Add energy and humidity " Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 7/9] hwmon: (core) Add fan " Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 8/9] hwmon: (core) Document new kernel API Guenter Roeck
2016-07-25 3:32 ` [PATCH v3 9/9] hwmon: (core) Add basic pwm attribute support to new API Guenter Roeck
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=a3b080a0-c929-7e72-3540-e49991512be5@roeck-us.net \
--to=linux@roeck-us.net \
--cc=edubezval@gmail.com \
--cc=jdelvare@suse.de \
--cc=jic23@kernel.org \
--cc=linux-hwmon@vger.kernel.org \
--cc=linux-iio@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-pm@vger.kernel.org \
--cc=punit.agrawal@arm.com \
--cc=rui.zhang@intel.com \
/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
all inboxes | Powered by JetHome®