Re: [PATCH mpam mpam/snapshot/v6.14-rc1 3/5] arm_mpam: Provide conversion method for new closid/rmid pairs

From: Zeng Heng
Date: Tue Feb 18 2025 - 01:32:26 EST


Hi,

On 2025/2/17 21:36, Dave Martin wrote:
Hi,

On Mon, Feb 17, 2025 at 02:18:44PM +0800, Zeng Heng wrote:
Hi Martin,

On 2025/2/17 11:18, Zeng Heng wrote:
The MPAM driver statically assigns all reqPARTIDs to respective intPARTIDs.
For the new rmid allocation strategy, it will check if there is an
available rmid of any reqPARTID which belongs to the input closid, not just
the rmids belonging to the closid.

For a mixture of MSCs system, for MSCs that do not support narrow-partid,
we use the PARTIDs exceeding the number of closids as reqPARTIDs for
expanding the monitoring groups.

In order to keep the existing resctrl API interface, the rmid contains both
req_idx and PMG information instead of PMG only under the MPAM driver. The
req_idx represents the req_idx-th sub-monitoring group under the control
group. The new rmid would be like:

rmid = (req_idx << shift | pmg).


To consider future compatibility with dynamically allocated reqpartid,
should I refactor the rmid?

Instead of defining rmid.req_idx, we could place the entire reqpartid
directly within rmid. In This way, the allocation of reqpartid will no
longer be constrained by the static allocation of closid, facilitating
future compatibility with dynamic allocation mechanisms.

In this case, it needs to refactor the resctrl_arch_rmid_idx_encode()
and resctrl_arch_rmid_idx_decode(), and we can simplify
closid_rmid2reqpartid() to rmid2reqpartid().

What are your thoughts on this idea? Thank you in advance for your
reply.

Best regards,
Zeng Heng

Does this mean that the RMID must be expanded to cover all possible
(reqPARTID, PMG) combinations?

A single reqPARTID cannot be allocated to two different resctrl control
groups at the same time, even though a PMG value can be reused across
multiple control groups -- so it sounds like your proposal would
require changes in the resctrl core code as well as possibly requiring
a larger rmid_ptrs table.

But I might have misunderstood what you are proposing here...

Can you illustrate with one or two examples?


Yes, as you have written. It might be worth trying this approach:
rmid = (reqpartid, pmg).

But to avoid the reuse of rmid across multiple control groups, we will
check the incoming closid in resctrl_arch_rmid_idx_encode() to prevent
rmid from being reallocated by resctrl_find_free_rmid().

We don't need a larger rmid_ptrs table, nor do we need to modify the
existing resctrl common code by continuing with the current static
allocation method.

Below are the implementations of several key functions for reference:

~~~
u32 rmid2reqpartid(u32 rmid)
{
u8 pmg_shift = fls(mpam_pmg_max);

WARN_ON_ONCE(pmg_shift > 8);

return rmid >> pmg_shift;
}

/*
* If the closid and rmid do not match upon inspection,
* immediately return an invalid idx value.
*/
u32 resctrl_arch_rmid_idx_encode(u32 closid, u32 rmid)
{
u32 reqpartid = rmid2reqpartid(rmid);
u32 intpartid = req2intpartid(reqpartid);

if (cdp_enabled)
intpartid >>= 1;

if (closid != intpartid)
return U32_MAX;

return rmid;
}

void resctrl_arch_rmid_idx_decode(u32 idx, u32 *closid, u32 *rmid)
{
u32 reqpartid = rmid2reqpartid(idx);
u32 intpartid = req2intpartid(reqpartid);

if (rmid)
*rmid = idx;
if (closid) {
if (cdp_enabled)
intpartid >>= 1;
*closid = intpartid;
}
}
~~~

Best regards,
Zeng Heng