Received-SPF: pass (google.com: domain of linux-kernel+bounces-193100-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Feedback-ID: i5ec1447f:Fastmail
User-Agent: Cyrus-JMAP/3.11.0-alpha0-491-g033e30d24-fm-20240520.001-g033e30d2
Precedence: bulk
MIME-Version: 1.0
Message-Id: <ac597520-fb5b-40e5-ae1f-de825450d2db@app.fastmail.com>
In-Reply-To: <5f4799b1-0606-46a9-a347-5a03738db341@amd.com>
References: <20240528013626.14066-1-luke@ljones.dev>
 <20240528013626.14066-9-luke@ljones.dev>
 <6f4bc109-00d0-47b0-a581-b96a6152545c@amd.com>
 <4d6b9171-7248-4937-87de-7e921ed8e507@app.fastmail.com>
 <5f4799b1-0606-46a9-a347-5a03738db341@amd.com>
Date: Wed, 29 May 2024 09:04:11 +1200
From: "Luke Jones" <luke@ljones.dev>
To: "Mario Limonciello" <mario.limonciello@amd.com>,
 "Hans de Goede" <hdegoede@redhat.com>
Cc: =?UTF-8?Q?Ilpo_J=C3=A4rvinen?= <ilpo.jarvinen@linux.intel.com>,
 corentin.chary@gmail.com, platform-driver-x86@vger.kernel.org,
 linux-kernel@vger.kernel.org
Subject: Re: [PATCH 8/9] platform/x86: asus-wmi: add apu_mem setting
Content-Type: text/plain


On Wed, 29 May 2024, at 1:27 AM, Mario Limonciello wrote:
> >>> diff --git a/drivers/platform/x86/asus-wmi.c b/drivers/platform/x86/asus-wmi.c
> >>> index f62a36dfcd4b..4b5fbae8c563 100644
> >>> --- a/drivers/platform/x86/asus-wmi.c
> >>> +++ b/drivers/platform/x86/asus-wmi.c
> >>> @@ -855,6 +855,112 @@ static DEVICE_ATTR_RW(cores_enabled);
> >>>    WMI_SIMPLE_SHOW(cores_max, "0x%x\n", ASUS_WMI_DEVID_CORES_MAX);
> >>>    static DEVICE_ATTR_RO(cores_max);
> >>>    
> >>> +/* Device memory available to APU */
> >>> +
> >>> +static ssize_t apu_mem_show(struct device *dev,
> >>> + struct device_attribute *attr, char *buf)
> >>> +{
> >>> + struct asus_wmi *asus = dev_get_drvdata(dev);
> >>> + int err;
> >>> + u32 mem;
> >>> +
> >>> + err = asus_wmi_get_devstate(asus, ASUS_WMI_DEVID_APU_MEM, &mem);
> >>> + if (err < 0)
> >>> + return err;
> >>> +
> >>> + switch (mem) {
> >>> + case 256:
> >>> + mem = 0;
> >>> + break;
> >>> + case 258:
> >>> + mem = 1;
> >>> + break;
> >>> + case 259:
> >>> + mem = 2;
> >>> + break;
> >>> + case 260:
> >>> + mem = 3;
> >>> + break;
> >>> + case 261:
> >>> + mem = 4;
> >>> + break;
> >>> + case 262:
> >>> + mem = 8;
> >>> + break;
> >>> + case 263:
> >>> + mem = 5;
> >>> + break;
> >>> + case 264:
> >>> + mem = 6;
> >>> + break;
> >>> + case 265:
> >>> + mem = 7;
> >>> + break;
> >>> + default:
> >>> + mem = 4;
> >>> + break;
> >>> + }
> >>> +
> >>> + return sysfs_emit(buf, "%d\n", mem);
> >>> +}
> >>> +
> >>> +static ssize_t apu_mem_store(struct device *dev,
> >>> +     struct device_attribute *attr,
> >>> +     const char *buf, size_t count)
> >>> +{
> >>> + struct asus_wmi *asus = dev_get_drvdata(dev);
> >>> + int result, err;
> >>> + u32 mem;
> >>> +
> >>> + result = kstrtou32(buf, 10, &mem);
> >>> + if (result)
> >>> + return result;
> >>> +
> >>> + switch (mem) {
> >>> + case 0:
> >>> + mem = 0;
> >>> + break;
> >>> + case 1:
> >>> + mem = 258;
> >>> + break;
> >>> + case 2:
> >>> + mem = 259;
> >>> + break;
> >>> + case 3:
> >>> + mem = 260;
> >>> + break;
> >>> + case 4:
> >>> + mem = 261;
> >>> + break;
> >>> + case 5:
> >>> + mem = 263;
> >>> + break;
> >>> + case 6:
> >>> + mem = 264;
> >>> + break;
> >>> + case 7:
> >>> + mem = 265;
> >>> + break;
> >>> + case 8:
> >>> + mem = 262;
> >>
> >> Is case 8 a mistake, or intentionally out of order?
> > 
> > Do you mean the `mem = <val>`? Those aren't in order, and I thought it was easier to read if the switch was ordered.
> > 
> 
> I'm wondering if case 5 should be 262, case 6 263, case 7 264 and case 8 
> 265.  It just stood out to me.

Yeah it's weird but that is what it is. Also verified in ghelper which calls the same WMI interfaces in Windows.

> 
> If that's all intended then no concerns and I agree sorting the case is 
> better.
> 
> >>
> >>> + break;
> >>> + default:
> >>> + return -EIO;
> >>> + }
> >>> +
> >>> + err = asus_wmi_set_devstate(ASUS_WMI_DEVID_APU_MEM, mem, &result);
> >>> + if (err) {
> >>> + pr_warn("Failed to set apu_mem: %d\n", err);
> >>> + return err;
> >>> + }
> >>> +
> >>> + pr_info("APU memory changed, reboot required\n");
> >>
> >> If you're logging something into the logs for this, I'd say make it more
> >> useful.
> >>
> >> "APU memory changed to %d MB"
> > 
> > Agreed. There's probably a few other spots I can do this also.
> > 
> >>
> >>> + sysfs_notify(&asus->platform_device->dev.kobj, NULL, "apu_mem");
> >>
> >> So this is a case that the BIOS attributes API I mentioned before would
> >> be REALLY useful.  There is a pending_reboot sysfs file that userspace
> >> can query to know if a given setting requires a reboot or not.
> >>
> >> Fwupd also uses this attribute to know to delay BIOS updates until the
> >> system has been rebooted.
> > 
> > Oh! Yes I'll queue that as an additional patch. There's at least 2 or 3 other spots where that would be good to have.
> > 
> 
> For any "new" attributes it's better to put them in that API than code 
> duplication of the BIOS attributes API as well as a random sysfs file 
> API that you can never discard.

Do you mean the firmware_attributes API? If so, I'm not opposed to adding all the existing ROG attributes to it also.

If I'm understanding the docs correctly, for example this apu_mem attr would then become:
- /sys/class/firmware-attributes/asus-bios/attributes/apu_mem/type
- /sys/class/firmware-attributes/*/attributes/apu_mem/current_value
- /sys/class/firmware-attributes/*/attributes/apu_mem/default_value
- /sys/class/firmware-attributes/*/attributes/apu_mem/display_name
- /sys/class/firmware-attributes/*/attributes/apu_mem/possible_values
- ..etc

That's absolutely much better than what I've been doing and I wish I'd known about it sooner.

So if I go ahead and convert all the new attr to this are there any issues with also converting much of the previous attr? And I'm aware of "don't break userspace" so really I'm a bit unsure how best to manage that (would a new module be better here also? "asus-bios.c" perhaps).

What I don't want is a split between platform and firmware_attributes.

> 
> >>> +
> >>> + return count;
> >>> +}
> >>> +static DEVICE_ATTR_RW(apu_mem);
> >>> +
> >>>    /* Tablet mode ****************************************************************/
> >>>    
> >>>    static void asus_wmi_tablet_mode_get_state(struct asus_wmi *asus)
> >>> @@ -4100,6 +4206,7 @@ static struct attribute *platform_attributes[] = {
> >>>    &dev_attr_panel_fhd.attr,
> >>>    &dev_attr_cores_enabled.attr,
> >>>    &dev_attr_cores_max.attr,
> >>> + &dev_attr_apu_mem.attr,
> >>>    &dev_attr_mini_led_mode.attr,
> >>>    &dev_attr_available_mini_led_mode.attr,
> >>>    NULL
> >>> @@ -4176,6 +4283,8 @@ static umode_t asus_sysfs_is_visible(struct kobject *kobj,
> >>>    else if (attr == &dev_attr_cores_enabled.attr
> >>>    || attr == &dev_attr_cores_max.attr)
> >>>    ok = asus_wmi_dev_is_present(asus, ASUS_WMI_DEVID_CORES_SET);
> >>> + else if (attr == &dev_attr_apu_mem.attr)
> >>> + ok = asus_wmi_dev_is_present(asus, ASUS_WMI_DEVID_APU_MEM);
> >>>    else if (attr == &dev_attr_mini_led_mode.attr)
> >>>    ok = asus->mini_led_dev_id != 0;
> >>>    else if (attr == &dev_attr_available_mini_led_mode.attr)
> >>> diff --git a/include/linux/platform_data/x86/asus-wmi.h b/include/linux/platform_data/x86/asus-wmi.h
> >>> index 5a56e7e97785..efe608861e55 100644
> >>> --- a/include/linux/platform_data/x86/asus-wmi.h
> >>> +++ b/include/linux/platform_data/x86/asus-wmi.h
> >>> @@ -121,6 +121,9 @@
> >>>     /* Maximum Intel E-core and P-core availability */
> >>>    #define ASUS_WMI_DEVID_CORES_MAX 0x001200D3
> >>>    
> >>> +/* Set the memory available to the APU */
> >>> +#define ASUS_WMI_DEVID_APU_MEM 0x000600C1
> >>> +
> >>>    /* MCU powersave mode */
> >>>    #define ASUS_WMI_DEVID_MCU_POWERSAVE   0x001200E2
> >>>    
> >>
> 
>