linux-kernel - Re: [PATCH 5/9] platform/x86: intel_pmc_core: Get LPM requirements for Tiger Lake

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]
Date:   Wed, 7 Apr 2021 16:27:46 +0200
From:   Hans de Goede <hdegoede@...hat.com>
To:     "David E. Box" <david.e.box@...ux.intel.com>,
        irenic.rajneesh@...il.com, mgross@...ux.intel.com,
        gayatri.kammela@...el.com
Cc:     platform-driver-x86@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 5/9] platform/x86: intel_pmc_core: Get LPM requirements
 for Tiger Lake

Hi,

On 4/1/21 5:05 AM, David E. Box wrote:
> From: Gayatri Kammela <gayatri.kammela@...el.com>
> 
> Platforms that support low power modes (LPM) such as Tiger Lake maintain
> requirements for each sub-state that a readable in the PMC. However, unlike
> LPM status registers, requirement registers are not memory mapped but are
> available from an ACPI _DSM. Collect the requirements for Tiger Lake using
> the _DSM method and store in a buffer.
> 
> Signed-off-by: Gayatri Kammela <gayatri.kammela@...el.com>
> Co-developed-by: David E. Box <david.e.box@...ux.intel.com>
> Signed-off-by: David E. Box <david.e.box@...ux.intel.com>
> ---
>  drivers/platform/x86/intel_pmc_core.c | 49 +++++++++++++++++++++++++++
>  drivers/platform/x86/intel_pmc_core.h |  2 ++
>  2 files changed, 51 insertions(+)
> 
> diff --git a/drivers/platform/x86/intel_pmc_core.c b/drivers/platform/x86/intel_pmc_core.c
> index ba0db301f07b..0ec26a4c715e 100644
> --- a/drivers/platform/x86/intel_pmc_core.c
> +++ b/drivers/platform/x86/intel_pmc_core.c
> @@ -23,7 +23,9 @@
>  #include <linux/slab.h>
>  #include <linux/suspend.h>
>  #include <linux/uaccess.h>
> +#include <linux/uuid.h>
>  
> +#include <acpi/acpi_bus.h>
>  #include <asm/cpu_device_id.h>
>  #include <asm/intel-family.h>
>  #include <asm/msr.h>
> @@ -31,6 +33,9 @@
>  
>  #include "intel_pmc_core.h"
>  
> +#define ACPI_S0IX_DSM_UUID		"57a6512e-3979-4e9d-9708-ff13b2508972"
> +#define ACPI_GET_LOW_MODE_REGISTERS	1
> +
>  /* PKGC MSRs are common across Intel Core SoCs */
>  static const struct pmc_bit_map msr_map[] = {
>  	{"Package C2",                  MSR_PKG_C2_RESIDENCY},
> @@ -587,6 +592,46 @@ static const struct pmc_reg_map tgl_reg_map = {
>  	.lpm_live_status_offset = TGL_LPM_LIVE_STATUS_OFFSET,
>  };
>  
> +static void pmc_core_get_tgl_lpm_reqs(struct platform_device *pdev)
> +{
> +	struct pmc_dev *pmcdev = platform_get_drvdata(pdev);
> +	const int num_maps = pmcdev->map->lpm_num_maps;
> +	size_t lpm_size = LPM_MAX_NUM_MODES * num_maps * 4;
> +	union acpi_object *out_obj;
> +	struct acpi_device *adev;
> +	guid_t s0ix_dsm_guid;
> +	u32 *lpm_req_regs;
> +
> +	adev = ACPI_COMPANION(&pdev->dev);
> +	if (!adev)
> +		return;
> +
> +	lpm_req_regs = devm_kzalloc(&pdev->dev, lpm_size * sizeof(u32),
> +				     GFP_KERNEL);
> +	if (!lpm_req_regs)
> +		return;
> +
> +	guid_parse(ACPI_S0IX_DSM_UUID, &s0ix_dsm_guid);
> +
> +	out_obj = acpi_evaluate_dsm(adev->handle, &s0ix_dsm_guid, 0,
> +				    ACPI_GET_LOW_MODE_REGISTERS, NULL);

Since you are using ACPI functions here, maybe change:

	depends on PCI

In the config INTEL_PMC_CORE Kconfig entry to:

	depends on PCI && ACPI

Note all functions you use are stubbed when !ACPI, so this
should build fine without this, but it will turn this function
into a no-op. If you prefer not to add the depends on ACPI that
is fine too.



> +	if (out_obj && out_obj->type == ACPI_TYPE_BUFFER) {
> +		u32 *addr = (u32 *)out_obj->buffer.pointer;
> +		int size = out_obj->buffer.length;
> +
> +		if (size != lpm_size)

	You're leaking lpm_req_regs here (sort of) maybe devm_free it here ?

> +			return;
> +
> +		memcpy_fromio(lpm_req_regs, addr, lpm_size);

This is wrong, the memory in an ACPI buffer is not IO-mem it is normal memory.

> +	} else
> +		acpi_handle_debug(adev->handle,
> +				  "_DSM function 0 evaluation failed\n");
> +
> +	ACPI_FREE(out_obj);
> +
> +	pmcdev->lpm_req_regs = lpm_req_regs;

You do this even if the "if (out_obj && out_obj->type == ACPI_TYPE_BUFFER)"
check above failed, making pmcdev->lpm_req_regs point to a block of
memory filled with zeros. That does not seem right.

> +}
> +
>  static inline u32 pmc_core_reg_read(struct pmc_dev *pmcdev, int reg_offset)
>  {
>  	return readl(pmcdev->regbase + reg_offset);
> @@ -1312,10 +1357,14 @@ static int pmc_core_probe(struct platform_device *pdev)
>  		return -ENOMEM;
>  
>  	mutex_init(&pmcdev->lock);
> +
>  	pmcdev->pmc_xram_read_bit = pmc_core_check_read_lock_bit(pmcdev);
>  	pmc_core_get_low_power_modes(pmcdev);
>  	pmc_core_do_dmi_quirks(pmcdev);
>  
> +	if (pmcdev->map == &tgl_reg_map)
> +		pmc_core_get_tgl_lpm_reqs(pdev);
> +
>  	/*
>  	 * On TGL, due to a hardware limitation, the GBE LTR blocks PC10 when
>  	 * a cable is attached. Tell the PMC to ignore it.
> diff --git a/drivers/platform/x86/intel_pmc_core.h b/drivers/platform/x86/intel_pmc_core.h
> index 3800c1ba6fb7..81d797feed33 100644
> --- a/drivers/platform/x86/intel_pmc_core.h
> +++ b/drivers/platform/x86/intel_pmc_core.h
> @@ -288,6 +288,7 @@ struct pmc_reg_map {
>   * @s0ix_counter:	S0ix residency (step adjusted)
>   * @num_modes:		Count of enabled modes
>   * @lpm_en_modes:	Array of enabled modes from lowest to highest priority
> + * @lpm_req_regs:	List of substate requirements
>   *
>   * pmc_dev contains info about power management controller device.
>   */
> @@ -304,6 +305,7 @@ struct pmc_dev {
>  	u64 s0ix_counter;
>  	int num_modes;
>  	int lpm_en_modes[LPM_MAX_NUM_MODES];
> +	u32 *lpm_req_regs;
>  };
>  
>  #define pmc_for_each_mode(i, mode, pmcdev)		\
> 

Regards,

Hams