linux-kernel - Re: [PATCH] drm/panfrost: Dynamically allocate pm

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: <5119045c-3c30-3625-d7ba-3100eda3ccbd@collabora.com>
Date:   Tue, 15 Feb 2022 10:23:50 +0100
From:   AngeloGioacchino Del Regno 
        <angelogioacchino.delregno@...labora.com>
To:     Alyssa Rosenzweig <alyssa@...labora.com>,
        Alyssa Rosenzweig <alyssa.rosenzweig@...labora.com>
Cc:     dri-devel@...ts.freedesktop.org, Rob Herring <robh@...nel.org>,
        Tomeu Vizoso <tomeu.vizoso@...labora.com>,
        Steven Price <steven.price@....com>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] drm/panfrost: Dynamically allocate pm_domains

Il 14/02/22 21:55, Alyssa Rosenzweig ha scritto:
> mali_kbase hardcodes MAX_PM_DOMAINS (=5 for the mt8192 kernel). I have
> no real objection to it but Angelo did. Maybe should've marked this RFC.

Clarifying, the suggested patch was not a big objection, but I think that it
would be a nice preventive cleanup that is useful for the power tree that has
to be managed on MT8192 and possibly on other SoCs.

I would expect to see a variable amount of PM domains to take care of as time
goes by (with new SoCs, not necessarily only MediaTek) due to granular power
optimizations but, at the same time, dynamically allocating the pm_domain_devs
and links structures makes this driver to also be nice with older platforms,
where memory is a little more constrained, allowing Linux to have a lighter
footprint, even if not by much.

Logic for this footprint saving is "a little here, a little there, becomes a
bit more considerable" (of course, being aware of both upsides and downsides
in dynamically allocating things, and avoiding to write gigabytes of text to
explain common knowledge).

P.S.: Thank you all!

Regards,
Angelo

> 
> On Mon, Feb 14, 2022 at 03:31:32PM -0500, Alyssa Rosenzweig wrote:
>> MT8192 requires 5 power domains. Rather than bump MAX_PM_DOMAINS and
>> waste memory on every supported Panfrost chip, instead dynamically
>> allocate pm_domain_devs and pm_domain_links. This adds some flexibility;
>> it seems inevitable a new MediaTek device will require more than 5
>> domains.
>>
>> On non-MediaTek devices, this saves a small amount of memory.
>>
>> Suggested-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@...labora.com>
>> Signed-off-by: Alyssa Rosenzweig <alyssa.rosenzweig@...labora.com>
>> ---
>>   drivers/gpu/drm/panfrost/panfrost_device.c | 14 ++++++++++----
>>   drivers/gpu/drm/panfrost/panfrost_device.h |  5 ++---
>>   2 files changed, 12 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/panfrost/panfrost_device.c b/drivers/gpu/drm/panfrost/panfrost_device.c
>> index ee612303f076..661cdec320af 100644
>> --- a/drivers/gpu/drm/panfrost/panfrost_device.c
>> +++ b/drivers/gpu/drm/panfrost/panfrost_device.c
>> @@ -127,7 +127,10 @@ static void panfrost_pm_domain_fini(struct panfrost_device *pfdev)
>>   {
>>   	int i;
>>   
>> -	for (i = 0; i < ARRAY_SIZE(pfdev->pm_domain_devs); i++) {
>> +	if (!pfdev->pm_domain_devs || !pfdev->pm_domain_links)
>> +		return;
>> +
>> +	for (i = 0; i < pfdev->comp->num_pm_domains; i++) {
>>   		if (!pfdev->pm_domain_devs[i])
>>   			break;
>>   
>> @@ -161,9 +164,12 @@ static int panfrost_pm_domain_init(struct panfrost_device *pfdev)
>>   		return -EINVAL;
>>   	}
>>   
>> -	if (WARN(num_domains > ARRAY_SIZE(pfdev->pm_domain_devs),
>> -			"Too many supplies in compatible structure.\n"))
>> -		return -EINVAL;
>> +	pfdev->pm_domain_devs = devm_kcalloc(pfdev->dev, num_domains,
>> +					     sizeof(*pfdev->pm_domain_devs),
>> +					     GFP_KERNEL);
>> +	pfdev->pm_domain_links = devm_kcalloc(pfdev->dev, num_domains,
>> +					      sizeof(*pfdev->pm_domain_links),
>> +					      GFP_KERNEL);
>>   
>>   	for (i = 0; i < num_domains; i++) {
>>   		pfdev->pm_domain_devs[i] =
>> diff --git a/drivers/gpu/drm/panfrost/panfrost_device.h b/drivers/gpu/drm/panfrost/panfrost_device.h
>> index 8b25278f34c8..98e3039696f9 100644
>> --- a/drivers/gpu/drm/panfrost/panfrost_device.h
>> +++ b/drivers/gpu/drm/panfrost/panfrost_device.h
>> @@ -22,7 +22,6 @@ struct panfrost_job;
>>   struct panfrost_perfcnt;
>>   
>>   #define NUM_JOB_SLOTS 3
>> -#define MAX_PM_DOMAINS 3
>>   
>>   struct panfrost_features {
>>   	u16 id;
>> @@ -87,8 +86,8 @@ struct panfrost_device {
>>   	struct regulator_bulk_data *regulators;
>>   	struct reset_control *rstc;
>>   	/* pm_domains for devices with more than one. */
>> -	struct device *pm_domain_devs[MAX_PM_DOMAINS];
>> -	struct device_link *pm_domain_links[MAX_PM_DOMAINS];
>> +	struct device **pm_domain_devs;
>> +	struct device_link **pm_domain_links;
>>   	bool coherent;
>>   
>>   	struct panfrost_features features;
>> -- 
>> 2.34.1
>>