Re: [PATCH 4/5] KVM: Check for empty mask of harvested dirty ring entries in caller

[Yan Zhao <yan.y.zhao@intel.com>]

On Mon, Jan 13, 2025 at 08:48:28AM -0800, Sean Christopherson wrote:
> On Mon, Jan 13, 2025, Yan Zhao wrote:
> > On Fri, Jan 10, 2025 at 05:04:08PM -0800, Sean Christopherson wrote:
> > > @@ -163,14 +157,31 @@ int kvm_dirty_ring_reset(struct kvm *kvm, struct kvm_dirty_ring *ring,
> > >  				continue;
> > >  			}
> > >  		}
> > > -		kvm_reset_dirty_gfn(kvm, cur_slot, cur_offset, mask);
> > > +
> > > +		/*
> > > +		 * Reset the slot for all the harvested entries that have been
> > > +		 * gathered, but not yet fully processed.
> > > +		 */
> > I really like the logs as it took me quite a while figuring out how this part of
> > the code works :)
> > 
> > Does "processed" mean the entries have been reset, and "gathered" means they've
> > been read from the ring?
> 
> Yeah.
> 
> > I'm not sure, but do you like this version? e.g.
> > "Combined reset of the harvested entries that can be identified by curr_slot
> > plus cur_offset+mask" ?
> 
> I have no objection to documenting the mechanics *and* the high level intent,
> but I definitely want to document the "what", not just the "how".
> 
> > > +		if (mask)
> > > +			kvm_reset_dirty_gfn(kvm, cur_slot, cur_offset, mask);
> > > +
> > > +		/*
> > > +		 * The current slot was reset or this is the first harvested
> > > +		 * entry, (re)initialize the metadata.
> > > +		 */
> > What about
> > "Save the current slot and cur_offset (with mask initialized to 1) to check if
> > any future entries can be found for a combined reset." ?
> 
> Hmm, what if I add a comment at the top to document the overall behavior and the
> variables,
> 
> 	/*
> 	 * To minimize mmu_lock contention, batch resets for harvested entries
> 	 * whose gfns are in the same slot, and are within N frame numbers of
> 	 * each other, where N is the number of bits in an unsigned long.  For
> 	 * simplicity, process the current set of entries when the next entry
> 	 * can't be included in the batch.
> 	 *
> 	 * Track the current batch slot, the gfn offset into the slot for the
> 	 * batch, and the bitmask of gfns that need to be reset (relative to
> 	 * offset).  Note, the offset may be adjusted backwards, e.g. so that
> 	 * a sequence of gfns X, X-1, ... X-N can be batched.
> 	 */
Nice log.
> 	u32 cur_slot, next_slot;
> 	u64 cur_offset, next_offset;
> 	unsigned long mask = 0;
> 	struct kvm_dirty_gfn *entry;
> 
> and then keep this as:
> 
> 		/*
> 		 * The current slot was reset or this is the first harvested
> 		 * entry, (re)initialize the batching metadata.
> 		 */
> 
> > 
> > >  		cur_slot = next_slot;
> > >  		cur_offset = next_offset;
> > >  		mask = 1;
> > >  		first_round = false;
> > >  	}
> > >  
> > > -	kvm_reset_dirty_gfn(kvm, cur_slot, cur_offset, mask);
> > > +	/*
> > > +	 * Perform a final reset if there are harvested entries that haven't
> > > +	 * been processed. The loop only performs a reset when an entry can't
> > > +	 * be coalesced, i.e. always leaves at least one entry pending.
> > The loop only performs a reset when an entry can be coalesced?
> 
> No, if an entry can be coalesced then the loop doesn't perform a reset.  Does
> this read better?
> 
> 	/*
> 	 * Perform a final reset if there are harvested entries that haven't
> 	 * been processed, which is guaranteed if at least one harvested was
> 	 * found.  The loop only performs a reset when the "next" entry can't
> 	 * be batched with "current" the entry(s), and that reset processes the
> 	 * _current_ entry(s), i.e. the last harvested entry, a.k.a. next, will
> 	 * will always be left pending.
> 	 */
LGTM!

> > > +	 */
> > > +	if (mask)
> > > +		kvm_reset_dirty_gfn(kvm, cur_slot, cur_offset, mask);
> > >  
> > >  	/*
> > >  	 * The request KVM_REQ_DIRTY_RING_SOFT_FULL will be cleared
> > > -- 
> > > 2.47.1.613.gc27f4b7a9f-goog
> > >