Re: [PATCH v2 2/5] x86,fs/resctrl: Remove unnecessary non-linear MBA configuration check

[Jinjie Ruan]

在 2026/10/8 5:06, Reinette Chatre 写道:
> Writing a bandwidth value from user space is rejected if the MBA resource
> has non-linear delay values while the architecture indicates (via
> resctrl_membw::arch_needs_linear) it can only configure resources with
> linear delay values. Intel indicates this for MBA while AMD and MPAM do not.
> 
> This check can never fail. No Intel SKU supports non-linear delay values,
> and the placeholder for such support, rdt_get_mb_table(), is only called
> during enumeration and always fails. A non-linear Intel MBA resource is
> thus never marked as capable of allocation, and resctrl never exposes it
> to user space or attempts to configure it. Every Intel MBA resource capable
> of allocation has linear delay values, and AMD and MPAM never indicate a
> need for them.
> 
> Fail enumeration of a non-linear Intel MBA resource directly to make clear
> that it is not supported, and remove the check together with the
> resctrl_membw::arch_needs_linear property it relies on.
> 
> Keep the informational message printed when such a system is encountered.
> While moving it, reword it to no longer refer to the removed mapping table,
> add the missing newline, and label the printed family value "family" instead
> of "cpu".
> 
> Signed-off-by: Reinette Chatre <reinette.chatre@xxxxxxxxx>
> ---
> Changes since RFC v1:
> - Fix typo in changelog
> 
> Changes since RFC v2:
> - Rework changelog.
> - Group all r->membw assignments together.
> 
> Changes since v1:
> - Keep (and fix) existing pr_info() message if a non-linear platform
>   appears. (Chenyu)
> - Rework changelog.
> - Group rdt_resource::membw assignments together.
> ---
>  arch/x86/kernel/cpu/resctrl/core.c | 42 +++++++-----------------------
>  fs/resctrl/ctrlmondata.c           |  8 ------
>  include/linux/resctrl.h            |  2 --
>  3 files changed, 10 insertions(+), 42 deletions(-)
> 
> diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
> index f9677e52790e..f96f7cf255d0 100644
> --- a/arch/x86/kernel/cpu/resctrl/core.c
> +++ b/arch/x86/kernel/cpu/resctrl/core.c
> @@ -181,27 +181,6 @@ static inline void cache_alloc_hsw_probe(void)
>  	rdt_alloc_capable = true;
>  }
>  
> -/*
> - * rdt_get_mb_table() - get a mapping of bandwidth(b/w) percentage values
> - * exposed to user interface and the h/w understandable delay values.
> - *
> - * The non-linear delay values have the granularity of power of two
> - * and also the h/w does not guarantee a curve for configured delay
> - * values vs. actual b/w enforced.
> - * Hence we need a mapping that is pre calibrated so the user can
> - * express the memory b/w as a percentage value.
> - */
> -static inline bool rdt_get_mb_table(struct rdt_resource *r)
> -{
> -	/*
> -	 * There are no Intel SKUs as of now to support non-linear delay.
> -	 */
> -	pr_info("MBA b/w map not implemented for cpu:%d, model:%d",
> -		boot_cpu_data.x86, boot_cpu_data.x86_model);
> -
> -	return false;
> -}
> -
>  static __init bool __get_mem_config_intel(struct rdt_resource *r)
>  {
>  	struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
> @@ -212,18 +191,18 @@ static __init bool __get_mem_config_intel(struct rdt_resource *r)
>  	cpuid_count(0x00000010, 3, &eax.full, &ebx, &ecx, &edx.full);
>  	hw_res->num_closid = edx.split.cos_max + 1;
>  	max_delay = eax.split.max_delay + 1;
> -	r->membw.max_bw = MAX_MBA_BW;
> -	r->membw.arch_needs_linear = true;
> -	if (ecx & MBA_IS_LINEAR) {
> -		r->membw.delay_linear = true;
> -		r->membw.min_bw = MAX_MBA_BW - max_delay;
> -		r->membw.bw_gran = MAX_MBA_BW - max_delay;
> -	} else {
> -		if (!rdt_get_mb_table(r))
> -			return false;
> -		r->membw.arch_needs_linear = false;
> +
> +	if (!(ecx & MBA_IS_LINEAR)) {
> +		pr_info("Non-linear MBA delay values not supported for family:%d, model:%d\n",
> +			boot_cpu_data.x86, boot_cpu_data.x86_model);
> +		return false;
>  	}

Acked-by: Jinjie Ruan <ruanjinjie@xxxxxxxxxx>

>  
> +	r->membw.delay_linear = true;
> +	r->membw.max_bw = MAX_MBA_BW;
> +	r->membw.min_bw = MAX_MBA_BW - max_delay;
> +	r->membw.bw_gran = MAX_MBA_BW - max_delay;
> +
>  	if (boot_cpu_has(X86_FEATURE_PER_THREAD_MBA))
>  		r->membw.throttle_mode = THREAD_THROTTLE_PER_THREAD;
>  	else
> @@ -255,7 +234,6 @@ static __init bool __rdt_get_mem_config_amd(struct rdt_resource *r)
>  
>  	/* AMD does not use delay */
>  	r->membw.delay_linear = false;
> -	r->membw.arch_needs_linear = false;
>  
>  	/*
>  	 * AMD does not use memory delay throttle model to control
> diff --git a/fs/resctrl/ctrlmondata.c b/fs/resctrl/ctrlmondata.c
> index cafebdff70dc..4562072b8ba6 100644
> --- a/fs/resctrl/ctrlmondata.c
> +++ b/fs/resctrl/ctrlmondata.c
> @@ -45,14 +45,6 @@ static bool bw_validate(char *buf, u32 *data, struct rdt_resource *r)
>  	int ret;
>  	u32 bw;
>  
> -	/*
> -	 * Only linear delay values is supported for current Intel SKUs.
> -	 */
> -	if (!r->membw.delay_linear && r->membw.arch_needs_linear) {
> -		rdt_last_cmd_puts("No support for non-linear MB domains\n");
> -		return false;
> -	}
> -
>  	ret = kstrtou32(buf, 10, &bw);
>  	if (ret) {
>  		rdt_last_cmd_printf("Invalid MB value %s\n", buf);
> diff --git a/include/linux/resctrl.h b/include/linux/resctrl.h
> index b63dc62690ca..66dc9813c888 100644
> --- a/include/linux/resctrl.h
> +++ b/include/linux/resctrl.h
> @@ -249,7 +249,6 @@ enum membw_throttle_mode {
>   * @max_bw:		Maximum memory bandwidth value, used as the reset value
>   * @bw_gran:		Granularity at which the memory bandwidth is allocated
>   * @delay_linear:	True if memory B/W delay is in linear scale
> - * @arch_needs_linear:	True if we can't configure non-linear resources
>   * @throttle_mode:	Bandwidth throttling mode when threads request
>   *			different memory bandwidths
>   * @mba_sc:		True if MBA software controller(mba_sc) is enabled
> @@ -259,7 +258,6 @@ struct resctrl_membw {
>  	u32				max_bw;
>  	u32				bw_gran;
>  	u32				delay_linear;
> -	bool				arch_needs_linear;
>  	enum membw_throttle_mode	throttle_mode;
>  	bool				mba_sc;
>  };

-- 
Best regards,
Jinjie