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: <3691edd4-7d04-42c5-9f5f-8f57a4da2740@gmail.com>
Date: Sat, 18 Oct 2025 03:23:53 +0200
From: Denis Benato <benato.denis96@...il.com>
To: Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>
Cc: LKML <linux-kernel@...r.kernel.org>, platform-driver-x86@...r.kernel.org,
 Hans de Goede <hdegoede@...hat.com>,
 "Limonciello, Mario" <mario.limonciello@....com>,
 "Luke D . Jones" <luke@...nes.dev>, Alok Tiwari <alok.a.tiwari@...cle.com>,
 Derek John Clark <derekjohn.clark@...il.com>,
 Mateusz Schyboll <dragonn@...pl>, porfet828@...il.com
Subject: Re: [PATCH v14 4/9] platform/x86: asus-armoury: add apu-mem control
 support


On 10/17/25 14:16, Ilpo Järvinen wrote:
> On Wed, 15 Oct 2025, Denis Benato wrote:
>
>> From: "Luke D. Jones" <luke@...nes.dev>
>>
>> Implement the APU memory size control under the asus-armoury module using
>> the fw_attributes class.
>>
>> This allows the APU allocated memory size to be adjusted depending on
>> the users priority. A reboot is required after change.
>>
>> Signed-off-by: Luke D. Jones <luke@...nes.dev>
>> Signed-off-by: Denis Benato <benato.denis96@...il.com>
>> ---
>>  drivers/platform/x86/asus-armoury.c        | 81 ++++++++++++++++++++++
>>  include/linux/platform_data/x86/asus-wmi.h |  2 +
>>  2 files changed, 83 insertions(+)
>>
>> diff --git a/drivers/platform/x86/asus-armoury.c b/drivers/platform/x86/asus-armoury.c
>> index 68ce2c159ae1..3b49a27e397d 100644
>> --- a/drivers/platform/x86/asus-armoury.c
>> +++ b/drivers/platform/x86/asus-armoury.c
>> @@ -394,6 +394,86 @@ static ssize_t egpu_enable_current_value_store(struct kobject *kobj, struct kobj
>>  WMI_SHOW_INT(egpu_enable_current_value, "%d\n", ASUS_WMI_DEVID_EGPU);
>>  ATTR_GROUP_BOOL_CUSTOM(egpu_enable, "egpu_enable", "Enable the eGPU (also disables dGPU)");
>>  
>> +/* Device memory available to APU */
>> +
>> +/* Values map for APU memory: some looks out of order but are actually correct */
>> +static u32 apu_mem_map[] = {
>> +	[0] = 0x000, /* called "AUTO" on the BIOS, is the minimum available */
>> +	[1] = 0x102,
>> +	[2] = 0x103,
>> +	[3] = 0x104,
>> +	[4] = 0x105,
>> +	[5] = 0x107,
>> +	[6] = 0x108,
>> +	[7] = 0x109,
>> +	[8] = 0x106,
> Is BIT(8) actually telling non-AUTO here? I mean, it's not set for the 
> AUTO case. If it has this special meaning, maybe that should be left out 
> of this array and handled by the show/store functions as it seems 
> different for the auto depending on which way the information is being 
> passed.
When 0x000 is set the interface will answer with value 0x100,
so it seems a bit confusing associating BIT(8) to "non-auto".

Honestly I don't see any particular pattern around these values.


I have tried to see if I could set 0x100 or 0x101 and my hardware
stopped booting until a hard reset was performed.

I am not seeing any particular reason to further modify code here.

>> +};
>> +
>> +static ssize_t apu_mem_current_value_show(struct kobject *kobj, struct kobj_attribute *attr,
>> +					  char *buf)
>> +{
>> +	int err;
>> +	u32 mem;
>> +
>> +	err = asus_wmi_get_devstate_dsts(ASUS_WMI_DEVID_APU_MEM, &mem);
>> +	if (err)
>> +		return err;
>> +
>> +	if ((mem & ASUS_WMI_DSTS_PRESENCE_BIT) == 0)
>> +		return -ENODEV;
>> +
>> +	mem &= ~ASUS_WMI_DSTS_PRESENCE_BIT;
>> +
>> +	/* After 0x000 is set, a read will return 0x100 */
>> +	if (mem == 0x100)
>> +		return sysfs_emit(buf, "0\n");
>> +
>> +	for (unsigned int i = 0; i < ARRAY_SIZE(apu_mem_map); i++) {
>> +		if (apu_mem_map[i] == mem)
>> +			return sysfs_emit(buf, "%u\n", i);
>> +	}
>> +
>> +	pr_warn("Unrecognised value for APU mem 0x%08x\n", mem);
>> +	return sysfs_emit(buf, "%u\n", mem);
> ??
>
> Should this return -EIO or something like that instead. It definitely 
> looks wrong to first log it as %x and then output to sysfs as %u.
>
>> +}
>> +
>> +static ssize_t apu_mem_current_value_store(struct kobject *kobj, struct kobj_attribute *attr,
>> +					   const char *buf, size_t count)
>> +{
>> +	int result, err;
>> +	u32 requested, mem;
>> +
>> +	result = kstrtou32(buf, 10, &requested);
>> +	if (result)
>> +		return result;
>> +
>> +	if (requested >= ARRAY_SIZE(apu_mem_map))
>> +		return -EINVAL;
>> +
>> +	mem = apu_mem_map[requested];
>> +
>> +	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 to %uGB, reboot required\n", requested+1);
> requested + 1
>
> Please mention GB up where the array is.
>
>> +	sysfs_notify(kobj, NULL, attr->attr.name);
>> +
>> +	asus_set_reboot_and_signal_event();
>> +
>> +	return count;
>> +}
>> +
>> +static ssize_t apu_mem_possible_values_show(struct kobject *kobj, struct kobj_attribute *attr,
>> +					    char *buf)
>> +{
>> +	BUILD_BUG_ON(ARRAY_SIZE(apu_mem_map) != 9);
>> +	return sysfs_emit(buf, "0;1;2;3;4;5;6;7;8\n");
> You could make this future proof with a for loop and sysfs_emit_at() and 
> drop the BUILD_BUG_ON(). Maybe all these multi-valued show functions could 
> share a helper which does that.
Thanks, due to changes to egpu interface I will introduce such helper in 0/2.
>> +}
>> +ATTR_GROUP_ENUM_CUSTOM(apu_mem, "apu_mem", "Set available system RAM (in GB) for the APU to use");
>> +
>>  /* Simple attribute creation */
>>  ATTR_GROUP_ENUM_INT_RO(charge_mode, "charge_mode", ASUS_WMI_DEVID_CHARGE_MODE, "0;1;2",
>>  		       "Show the current mode of charging");
>> @@ -414,6 +494,7 @@ static const struct asus_attr_group armoury_attr_groups[] = {
>>  	{ &egpu_connected_attr_group, ASUS_WMI_DEVID_EGPU_CONNECTED },
>>  	{ &egpu_enable_attr_group, ASUS_WMI_DEVID_EGPU },
>>  	{ &dgpu_disable_attr_group, ASUS_WMI_DEVID_DGPU },
>> +	{ &apu_mem_attr_group, ASUS_WMI_DEVID_APU_MEM },
>>  
>>  	{ &charge_mode_attr_group, ASUS_WMI_DEVID_CHARGE_MODE },
>>  	{ &boot_sound_attr_group, ASUS_WMI_DEVID_BOOT_SOUND },
>> diff --git a/include/linux/platform_data/x86/asus-wmi.h b/include/linux/platform_data/x86/asus-wmi.h
>> index 10acd5d52e38..a4f6bab93a6f 100644
>> --- a/include/linux/platform_data/x86/asus-wmi.h
>> +++ b/include/linux/platform_data/x86/asus-wmi.h
>> @@ -137,6 +137,8 @@
>>  /* dgpu on/off */
>>  #define ASUS_WMI_DEVID_DGPU		0x00090020
>>  
>> +#define ASUS_WMI_DEVID_APU_MEM		0x000600C1
>> +
>>  /* gpu mux switch, 0 = dGPU, 1 = Optimus */
>>  #define ASUS_WMI_DEVID_GPU_MUX		0x00090016
>>  #define ASUS_WMI_DEVID_GPU_MUX_VIVO	0x00090026
>>

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ