Re: [PATCH v4 2/3] arm_mpam: resctrl: Only enable CDP on a resource that supports it

From: Jinjie Ruan

Date: Tue Nov 17 2026 - 07:06:28 EST




在 2026/11/17 19:46, Ben Horgan 写道:
> Hi Jinjie,
>
> On 17/11/2026 11:41, Jinjie Ruan wrote:
>>
>>
>> 在 2026/11/17 18:35, Ben Horgan 写道:
>>> Hi Gavin, Jinjie,
>>
>> Hi Ben,
>>
>>>
>>> On 17/11/2026 03:31, Jinjie Ruan wrote:
>>>> Hi Gavin,
>>>>
>>>> 在 2026/11/13 7:42, Gavin Shan 写道:
>>>>> On 11/11/26 6:13 PM, Jinjie Ruan wrote:
>>>>>> resctrl_arch_set_cdp_enabled() enables CDP for any cache level that is
>>>>>> requested, even one that has no MPAM cache controls. Mounting resctrl
>>>>>> with -o cdpl2 when only L3 can allocate still turns CDP emulation on for
>>>>>> the whole system, so every CLOSID takes two PARTIDs while resctrl keeps
>>>>>> its CLOSID count at the number of PARTIDs. Control groups in the upper
>>>>>> half then map to PARTIDs beyond mpam_partid_max, which index past the
>>>>>> configuration and monitor arrays and program out of range PARTIDs into
>>>>>> the MSCs. On QEMU with 5 PARTIDs the first monitor read after creating
>>>>>> 2 control groups raises MSMONCFG_ID_RANGE for PARTID 5, and MPAM is
>>>>>> disabled.
>>>>>>
>>>>>> Do not enable CDP when a resource cannot allocate or is not CDP capable,
>>>>>> as x86 does, so the mount fails instead. Only the enable path is checked,
>>>>>> since unmount disables CDP on both cache levels to restore the MBA
>>>>>> control.
>>>>>>
>>>>>> A system where only MBA can allocate, or that only has monitors, now
>>>>>> fails a mount with CDP as well: CDP hides the MBA control because
>>>>>> mbw_max is a per-PARTID maximum, leaving nothing for CDP to apply to.
>>>>>>
>>>>>> Fixes: 6789fb99282c ("arm_mpam: resctrl: Add CDP emulation")
>>>>>> Link: https://sashiko.dev/#/bug/linux-05c19551-0bba-4a06-
>>>>>> aa2c-3c90fab5853d
>>>>>> Suggested-by: Ben Horgan <ben.horgan@xxxxxxx>
>>>>>> Signed-off-by: Jinjie Ruan <ruanjinjie@xxxxxxxxxx>
>>>>>> Cc: <stable@xxxxxxxxxxxxxxx> # 7.1+
>>>>>> ---
>>>>>>   drivers/resctrl/mpam_resctrl.c | 4 ++++
>>>>>>   1 file changed, 4 insertions(+)
>>>>>>
>>>>>> diff --git a/drivers/resctrl/mpam_resctrl.c b/drivers/resctrl/
>>>>>> mpam_resctrl.c
>>>>>> index 573617ba09df..afa3c084c475 100644
>>>>>> --- a/drivers/resctrl/mpam_resctrl.c
>>>>>> +++ b/drivers/resctrl/mpam_resctrl.c
>>>>>> @@ -222,6 +222,10 @@ int resctrl_arch_set_cdp_enabled(enum
>>>>>> resctrl_res_level rid, bool enable)
>>>>>>       if (enable && mpam_partid_max < 1)
>>>>>>           return -EINVAL;
>>>>>>   +    if (enable && (!
>>>>>> mpam_resctrl_controls[rid].resctrl_res.alloc_capable ||
>>>>>> +               !mpam_resctrl_controls[rid].resctrl_res.cdp_capable))
>>>>>> +        return -EINVAL;
>>>>>> +
>>>>>
>>>>> x86 only checks 'cdp_capable', so this check isn't exactly same as to
>>>>> what's
>>>>> done on x86, as claimed in the commit log?
>>>>
>>>> Sorry, the commit message is somewhat misleading and need updated.
>>>>
>>>> In my opinion, CDP is a feature related to resource allocation, and it
>>>> is always reasonable to check alloc_capable.
>>>
>>> Hmm, I know the patch came from my suggestion but I think the reason
>>> that x86 only checks cdp_capable is that alloc_capable is that the
>>> alloc_capable flag is also used to indicate that the resource is fully
>>> setup and can be used.
>>
>> Thank you for your explanation.
>>
>>>
>>> In resctrl_arch_set_cdp_enabled() we manipulate alloc_capable for
>>> re-enabling the MBA resource and the added !alloc_capable check would
>>
>> I think you mean the following code:
>>
>> if (cdp_enabled &&
>> !mpam_resctrl_controls[RDT_RESOURCE_MBA].cdp_enabled)
>>
>> mpam_resctrl_controls[RDT_RESOURCE_MBA].resctrl_res.alloc_capable = false;
>>
>> if (mpam_resctrl_controls[RDT_RESOURCE_MBA].cdp_enabled &&
>> mpam_resctrl_controls[RDT_RESOURCE_MBA].class)
>>
>> mpam_resctrl_controls[RDT_RESOURCE_MBA].resctrl_res.alloc_capable = true;
>>
>>
>>> mean that we'd never be able to do that manipulation. resctrl doesn't
>>
>> Does you mean that we enable CDP for L2 or L3, but not for MBA, which
>> causes aclloc_capable to be cleared to hide MBA. Then, when we try to
>> enable CDP for MBA, it gets intercepted by this newly added
>> alloc_capable() check, resulting in the enabling process failing?
>
> Yes
>
>>
>> Sounds we should remove the alloc_capable().
>
> If you mean the check for !alloc_capable, that this patch adds, then I
> agree.

Yes. That's exactly what it means, and it will fix this problem.

>
> Thanks,
>
> Ben
>
>>
>>> call resctrl_arch_set_cdp_enabled() for the MBA resource but to be more
>>> resilient the MPAM interface allows it.
>>
>> Yes.
>>
>>>
>>> In the future, once we start adding more controls for resources we'll
>>> have to be careful to do the cdp enabling and hiding of features at the
>>> control level rather than the feature level.
>>
>> I agree, in the future there will more control such as
>> MBW_MAX、MBW_MIN、INTPRI、CMAX etc.
>>
>>>
>>> Thanks,
>>>
>>> Ben
>>>
>>>>
>>>> Best regards,
>>>> Jinjie
>>>>
>>>>>
>>>>> int resctrl_arch_set_cdp_enabled(enum resctrl_res_level l, bool enable)
>>>>> {
>>>>>         struct rdt_hw_resource *hw_res = &rdt_resources_all[l];
>>>>>
>>>>>         if (!hw_res->r_resctrl.cdp_capable)
>>>>>                 return -EINVAL;
>>>>>
>>>>>         :
>>>>> }
>>>>>
>>>>>
>>>>>>       /*
>>>>>>        * resctrl_arch_set_cdp_enabled() is only called with enable set to
>>>>>>        * false on error and unmount.
>>>>>
>>>>> Thanks,
>>>>> Gavin
>>>>
>>>
>>
>