Re: [PATCH v2] x86/resctrl: Implement new mba_MBps throttling heuristic

[Tony Luck <tony.luck@intel.com>]

On Tue, Jan 23, 2024 at 04:29:42PM -0800, Tony Luck wrote:
> On Tue, Jan 23, 2024 at 05:07:50PM +0000, Luck, Tony wrote:
> > >> I need to grab an SNC system to re-check that everything still works when re-based.
> > >> But right now, this looks like adding the SNC series will be easy (famous last words!).
> > >
> > > Once you're ready - can you point me at something I can use as a stable branch to rebase onto?
> > 
> > I'll be running the tests this afternoon. But I'm somewhat confident that they will pass.
> > 
> > Tree is here: git://git.kernel.org/pub/scm/linux/kernel/git/aegl/linux.git branch "hopeful_snc"
> > 
> > Commits are:
> > 
> > 793635f10aeb x86/resctrl: Update documentation with Sub-NUMA cluster changes
> > dca7ba785d6f x86/resctrl: Sub NUMA Cluster detection and enable
> > e41bd88101c8 x86/resctrl: Introduce snc_nodes_per_l3_cache
> > ccbab7875197 x86/resctrl: Add node-scope to the options for feature scope
> > 2f0db8c7072b x86/resctrl: Split the rdt_domain and rdt_hw_domain structures
> > 22697627cc5f x86/resctrl: Prepare for different scope for control/monitor operations
> > f3ff831a9042 x86/resctrl: Prepare to split rdt_domain structure
> > c0679868ee78 x86/resctrl: Prepare for new domain scope
> > 327b4394f309 x86/resctrl: Implement new mba_MBps throttling heuristic
> > 5b14817cf87e x86/resctrl: Read supported bandwidth sources using CPUID command
> > 5699cd082e1f x86/resctrl: Remove hard-coded memory bandwidth limit
> > 1b908debf53f x86/resctrl: Fix unused variable warning in cache_alloc_hsw_probe()
> > 
> > Only the last of these (1b908debf53f) had been accepted by Boris into TIP x86/cache
> > when I cut this tree yesterday.
> > 
> > Two more applied by Boris today without any changes from what is in my tree, but the
> > commit IDs in TIP are obviously different from in my tree.
> > 
> > 54e35eb8611c x86/resctrl: Read supported bandwidth sources from CPUID
> > 0976783bb123 x86/resctrl: Remove hard-coded memory bandwidth limit
> > 
> > All the others have the normal risk that Boris will find some way to make them better.
> 
> All my SNC tests pass. So this tree is as good as I can make it.
> 
> Boris: After you take/reject "x86/resctrl: Implement new mba_MBps throttling heuristic"
> in TIP x86/cache I will post v14 of the SNC series rebased to that TIP
> branch.

Boris,

I got the tip bot message that you have applied:

  c2427e70c163 ("x86/resctrl: Implement new mba_MBps throttling heuristic")

I see that you also applied:

  fc747eebef73 ("x86/resctrl: Remove redundant variable in mbm_config_write_domain()")

Does that empty your resctrl queue except for the SNC series? If it
does, then I'll rebase SNC patches onto TIP x86/cache.

-Tony