Re: [PATCH v6 06/40] arm64: mpam: Context switch the MPAM registers

[Jinjie Ruan <ruanjinjie@huawei.com>]

在 2026/3/13 22:45, Ben Horgan 写道:
> From: James Morse <james.morse@arm.com>
> 
> MPAM allows traffic in the SoC to be labeled by the OS, these labels are
> used to apply policy in caches and bandwidth regulators, and to monitor
> traffic in the SoC. The label is made up of a PARTID and PMG value. The x86
> equivalent calls these CLOSID and RMID, but they don't map precisely.
> 
> MPAM has two CPU system registers that is used to hold the PARTID and PMG
> values that traffic generated at each exception level will use. These can
> be set per-task by the resctrl file system. (resctrl is the defacto
> interface for controlling this stuff).
> 
> Add a helper to switch this.
> 
> struct task_struct's separate CLOSID and RMID fields are insufficient to
> implement resctrl using MPAM, as resctrl can change the PARTID (CLOSID) and
> PMG (sort of like the RMID) separately. On x86, the rmid is an independent
> number, so a race that writes a mismatched closid and rmid into hardware is
> benign. On arm64, the pmg bits extend the partid.
> (i.e. partid-5 has a pmg-0 that is not the same as partid-6's pmg-0).  In
> this case, mismatching the values will 'dirty' a pmg value that resctrl
> believes is clean, and is not tracking with its 'limbo' code.
> 
> To avoid this, the partid and pmg are always read and written as a
> pair. This requires a new u64 field. In struct task_struct there are two
> u32, rmid and closid for the x86 case, but as we can't use them here do
> something else. Add this new field, mpam_partid_pmg, to struct thread_info
> to avoid adding more architecture specific code to struct task_struct.
> Always use READ_ONCE()/WRITE_ONCE() when accessing this field.
> 
> Resctrl allows a per-cpu 'default' value to be set, this overrides the
> values when scheduling a task in the default control-group, which has
> PARTID 0. The way 'code data prioritisation' gets emulated means the
> register value for the default group needs to be a variable.
> 
> The current system register value is kept in a per-cpu variable to avoid
> writing to the system register if the value isn't going to change.  Writes
> to this register may reset the hardware state for regulating bandwidth.
> 
> Finally, there is no reason to context switch these registers unless there
> is a driver changing the values in struct task_struct. Hide the whole thing
> behind a static key. This also allows the driver to disable MPAM in
> response to errors reported by hardware. Move the existing static key to
> belong to the arch code, as in the future the MPAM driver may become a
> loadable module.
> 
> All this should depend on whether there is an MPAM driver, hide it behind
> CONFIG_ARM64_MPAM.
> 
> Tested-by: Gavin Shan <gshan@redhat.com>
> Tested-by: Shaopeng Tan <tan.shaopeng@jp.fujitsu.com>
> Tested-by: Peter Newman <peternewman@google.com>
> Tested-by: Zeng Heng <zengheng4@huawei.com>
> Tested-by: Punit Agrawal <punit.agrawal@oss.qualcomm.com>
> CC: Amit Singh Tomar <amitsinght@marvell.com>
> Reviewed-by: Zeng Heng <zengheng4@huawei.com>
> Reviewed-by: Shaopeng Tan <tan.shaopeng@jp.fujitsu.com>
> Reviewed-by: Jonathan Cameron <jonathan.cameron@huawei.com>
> Reviewed-by: Gavin Shan <gshan@redhat.com>
> Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
> Signed-off-by: James Morse <james.morse@arm.com>
> Signed-off-by: Ben Horgan <ben.horgan@arm.com>
> ---
> Changes since rfc:
> CONFIG_MPAM -> CONFIG_ARM64_MPAM in commit message
> Remove extra DECLARE_STATIC_KEY_FALSE
> Function name in comment, __mpam_sched_in() -> mpam_thread_switch()
> Remove unused headers
> Expand comment (Jonathan)
> 
> Changes since v2:
> Tidy up ifdefs
> 
> Changes since v3:
> Always set MPAMEN for MPAM1_EL1 rather than relying on it being read only.
> ---
>  arch/arm64/Kconfig                   |  2 +
>  arch/arm64/include/asm/mpam.h        | 67 ++++++++++++++++++++++++++++
>  arch/arm64/include/asm/thread_info.h |  3 ++
>  arch/arm64/kernel/Makefile           |  1 +
>  arch/arm64/kernel/mpam.c             | 13 ++++++
>  arch/arm64/kernel/process.c          |  7 +++
>  drivers/resctrl/mpam_devices.c       |  2 -
>  drivers/resctrl/mpam_internal.h      |  4 +-
>  8 files changed, 95 insertions(+), 4 deletions(-)
>  create mode 100644 arch/arm64/include/asm/mpam.h
>  create mode 100644 arch/arm64/kernel/mpam.c
> 
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 38dba5f7e4d2..ecaaca13a969 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -2039,6 +2039,8 @@ config ARM64_MPAM
>  
>  	  MPAM is exposed to user-space via the resctrl pseudo filesystem.
>  
> +	  This option enables the extra context switch code.
> +
>  endmenu # "ARMv8.4 architectural features"
>  
>  menu "ARMv8.5 architectural features"
> diff --git a/arch/arm64/include/asm/mpam.h b/arch/arm64/include/asm/mpam.h
> new file mode 100644
> index 000000000000..0747e0526927
> --- /dev/null
> +++ b/arch/arm64/include/asm/mpam.h
> @@ -0,0 +1,67 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/* Copyright (C) 2025 Arm Ltd. */
> +
> +#ifndef __ASM__MPAM_H
> +#define __ASM__MPAM_H
> +
> +#include <linux/jump_label.h>
> +#include <linux/percpu.h>
> +#include <linux/sched.h>
> +
> +#include <asm/sysreg.h>
> +
> +DECLARE_STATIC_KEY_FALSE(mpam_enabled);
> +DECLARE_PER_CPU(u64, arm64_mpam_default);
> +DECLARE_PER_CPU(u64, arm64_mpam_current);
> +
> +/*
> + * The value of the MPAM0_EL1 sysreg when a task is in resctrl's default group.
> + * This is used by the context switch code to use the resctrl CPU property
> + * instead. The value is modified when CDP is enabled/disabled by mounting
> + * the resctrl filesystem.
> + */
> +extern u64 arm64_mpam_global_default;
> +
> +/*
> + * The resctrl filesystem writes to the partid/pmg values for threads and CPUs,
> + * which may race with reads in mpam_thread_switch(). Ensure only one of the old
> + * or new values are used. Particular care should be taken with the pmg field as
> + * mpam_thread_switch() may read a partid and pmg that don't match, causing this
> + * value to be stored with cache allocations, despite being considered 'free' by
> + * resctrl.
> + */
> +#ifdef CONFIG_ARM64_MPAM
> +static inline u64 mpam_get_regval(struct task_struct *tsk)
> +{
> +	return READ_ONCE(task_thread_info(tsk)->mpam_partid_pmg);
> +}
> +
> +static inline void mpam_thread_switch(struct task_struct *tsk)
> +{
> +	u64 oldregval;
> +	int cpu = smp_processor_id();
> +	u64 regval = mpam_get_regval(tsk);
> +
> +	if (!static_branch_likely(&mpam_enabled))
> +		return;
> +
> +	if (regval == READ_ONCE(arm64_mpam_global_default))
> +		regval = READ_ONCE(per_cpu(arm64_mpam_default, cpu));
> +
> +	oldregval = READ_ONCE(per_cpu(arm64_mpam_current, cpu));
> +	if (oldregval == regval)
> +		return;
> +
> +	write_sysreg_s(regval | MPAM1_EL1_MPAMEN, SYS_MPAM1_EL1);
> +	isb();

Hi, everyone,

We found a problem as below:

In an offline/online co-located setup, MPAM gives the offline group a
smaller MBW share, a softirq serving an online service runs with the
PARTID of the throttled offline task it interrupted and is throttled
with it -- a priority inversion.

Is this a known limitation?

I see the following patchset uses a separate closid for kernel.

https://lore.kernel.org/all/cover.1790867441.git.babu.moger@amd.com/

> +
> +	/* Synchronising the EL0 write is left until the ERET to EL0 */
> +	write_sysreg_s(regval, SYS_MPAM0_EL1);
> +
> +	WRITE_ONCE(per_cpu(arm64_mpam_current, cpu), regval);
> +}
> +#else
> +static inline void mpam_thread_switch(struct task_struct *tsk) {}
> +#endif /* CONFIG_ARM64_MPAM */
> +
> +#endif /* __ASM__MPAM_H */
> diff --git a/arch/arm64/include/asm/thread_info.h b/arch/arm64/include/asm/thread_info.h
> index 7942478e4065..5d7fe3e153c8 100644
> --- a/arch/arm64/include/asm/thread_info.h
> +++ b/arch/arm64/include/asm/thread_info.h
> @@ -41,6 +41,9 @@ struct thread_info {
>  #ifdef CONFIG_SHADOW_CALL_STACK
>  	void			*scs_base;
>  	void			*scs_sp;
> +#endif
> +#ifdef CONFIG_ARM64_MPAM
> +	u64			mpam_partid_pmg;
>  #endif
>  	u32			cpu;
>  };
> diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile
> index 76f32e424065..15979f366519 100644
> --- a/arch/arm64/kernel/Makefile
> +++ b/arch/arm64/kernel/Makefile
> @@ -67,6 +67,7 @@ obj-$(CONFIG_CRASH_DUMP)		+= crash_dump.o
>  obj-$(CONFIG_VMCORE_INFO)		+= vmcore_info.o
>  obj-$(CONFIG_ARM_SDE_INTERFACE)		+= sdei.o
>  obj-$(CONFIG_ARM64_PTR_AUTH)		+= pointer_auth.o
> +obj-$(CONFIG_ARM64_MPAM)		+= mpam.o
>  obj-$(CONFIG_ARM64_MTE)			+= mte.o
>  obj-y					+= vdso-wrap.o
>  obj-$(CONFIG_COMPAT_VDSO)		+= vdso32-wrap.o
> diff --git a/arch/arm64/kernel/mpam.c b/arch/arm64/kernel/mpam.c
> new file mode 100644
> index 000000000000..9866d2ca0faa
> --- /dev/null
> +++ b/arch/arm64/kernel/mpam.c
> @@ -0,0 +1,13 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/* Copyright (C) 2025 Arm Ltd. */
> +
> +#include <asm/mpam.h>
> +
> +#include <linux/jump_label.h>
> +#include <linux/percpu.h>
> +
> +DEFINE_STATIC_KEY_FALSE(mpam_enabled);
> +DEFINE_PER_CPU(u64, arm64_mpam_default);
> +DEFINE_PER_CPU(u64, arm64_mpam_current);
> +
> +u64 arm64_mpam_global_default;
> diff --git a/arch/arm64/kernel/process.c b/arch/arm64/kernel/process.c
> index 489554931231..47698955fa1e 100644
> --- a/arch/arm64/kernel/process.c
> +++ b/arch/arm64/kernel/process.c
> @@ -51,6 +51,7 @@
>  #include <asm/fpsimd.h>
>  #include <asm/gcs.h>
>  #include <asm/mmu_context.h>
> +#include <asm/mpam.h>
>  #include <asm/mte.h>
>  #include <asm/processor.h>
>  #include <asm/pointer_auth.h>
> @@ -738,6 +739,12 @@ struct task_struct *__switch_to(struct task_struct *prev,
>  	if (prev->thread.sctlr_user != next->thread.sctlr_user)
>  		update_sctlr_el1(next->thread.sctlr_user);
>  
> +	/*
> +	 * MPAM thread switch happens after the DSB to ensure prev's accesses
> +	 * use prev's MPAM settings.
> +	 */
> +	mpam_thread_switch(next);
> +
>  	/* the actual thread switch */
>  	last = cpu_switch_to(prev, next);
>  
> diff --git a/drivers/resctrl/mpam_devices.c b/drivers/resctrl/mpam_devices.c
> index ff861291bd4e..d50461d6ff3f 100644
> --- a/drivers/resctrl/mpam_devices.c
> +++ b/drivers/resctrl/mpam_devices.c
> @@ -29,8 +29,6 @@
>  
>  #include "mpam_internal.h"
>  
> -DEFINE_STATIC_KEY_FALSE(mpam_enabled); /* This moves to arch code */
> -
>  /*
>   * mpam_list_lock protects the SRCU lists when writing. Once the
>   * mpam_enabled key is enabled these lists are read-only,
> diff --git a/drivers/resctrl/mpam_internal.h b/drivers/resctrl/mpam_internal.h
> index e8971842b124..4632985bcca6 100644
> --- a/drivers/resctrl/mpam_internal.h
> +++ b/drivers/resctrl/mpam_internal.h
> @@ -16,12 +16,12 @@
>  #include <linux/srcu.h>
>  #include <linux/types.h>
>  
> +#include <asm/mpam.h>
> +
>  #define MPAM_MSC_MAX_NUM_RIS	16
>  
>  struct platform_device;
>  
> -DECLARE_STATIC_KEY_FALSE(mpam_enabled);
> -
>  #ifdef CONFIG_MPAM_KUNIT_TEST
>  #define PACKED_FOR_KUNIT __packed
>  #else

-- 
Best regards,
Jinjie