Date:   Wed, 7 Sep 2022 07:59:16 -0700
From:   Mike Tipton <quic_mdtipton@...cinc.com>
To:     Georgi Djakov <djakov@...nel.org>
CC:     <linux-pm@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <quic_okukatla@...cinc.com>
Subject: Re: [PATCH] interconnect: Replace mutex with rt_mutex

On Wed, Sep 07, 2022 at 10:35:26AM +0300, Georgi Djakov wrote:
> Hi Mike,
> 
> Thanks for the patch!
> 
> On 6.09.22 22:14, Mike Tipton wrote:
> > Replace mutex with rt_mutex to prevent priority inversion between
> > clients, which can cause unacceptable delays in some cases.
> 
> It would be nice if you have any numbers to share in the commit text.

I can try to dig up some numbers, but mileage will vary tremendously of
course. Improvement is really only seen in certain high-concurrency
scenarios.

> 
> > Signed-off-by: Mike Tipton <quic_mdtipton@...cinc.com>
> > ---
> > 
> > We've run into a number of cases internally and from customers where
> > high priority, RT clients (typically display) become blocked for long
> > periods of time on lower priority, non-RT clients. Switching to rt_mutex
> > has proven to help performance in these cases.
> 
> I am wondering if avoiding the inversion on this specific lock is the right
> solution, as there could be other locks that may cause similar issues. Do we
> see similar issue with clocks for example or is it just with interconnects?

I raised these same concerns internally, since some of the clients
experiencing delays also request clocks and regulators. However, I
believe they primarily request interconnects in these critical paths and
not clocks/regulators. At least not as frequently as they request
interconnect. Additionally, I suspect the average interconnect latencies
are higher than clock on our platforms, since interconnect will always
result in blocking calls to RPM/RPMh. I also wouldn't be surprised if we
have more consistent contention on interconnect, since certain clients
update DDR BW quite frequently. I suppose at some point the same
rt_mutex argument could be made for clock and regulator as well, but
to-date we've only needed to change interconnect to see improvement.

I'm not sure what an alternative, generic solution would be. We have
many clients requesting many different paths. Some are more
latency-sensitive and higher priority than others. If these use cases
overlap, then we're subject to these sorts of priority inversion issues.
Bumping the priority of all clients to match the highest priority one
isn't really possible.

> 
> >   drivers/interconnect/core.c | 80 ++++++++++++++++++-------------------
> >   1 file changed, 40 insertions(+), 40 deletions(-)
> > 
> > diff --git a/drivers/interconnect/core.c b/drivers/interconnect/core.c
> > index 25debded65a8..a536c013d9ca 100644
> > --- a/drivers/interconnect/core.c
> > +++ b/drivers/interconnect/core.c
> > @@ -14,7 +14,7 @@
> >   #include <linux/interconnect-provider.h>
> >   #include <linux/list.h>
> >   #include <linux/module.h>
> > -#include <linux/mutex.h>
> > +#include <linux/rtmutex.h>
> >   #include <linux/slab.h>
> >   #include <linux/of.h>
> >   #include <linux/overflow.h>
> > @@ -28,7 +28,7 @@ static DEFINE_IDR(icc_idr);
> >   static LIST_HEAD(icc_providers);
> >   static int providers_count;
> >   static bool synced_state;
> > -static DEFINE_MUTEX(icc_lock);
> > +static DEFINE_RT_MUTEX(icc_lock);
> >   static struct dentry *icc_debugfs_dir;
> >   static void icc_summary_show_one(struct seq_file *s, struct icc_node *n)
> > @@ -47,7 +47,7 @@ static int icc_summary_show(struct seq_file *s, void *data)
> >   	seq_puts(s, " node                                  tag          avg         peak\n");
> >   	seq_puts(s, "--------------------------------------------------------------------\n");
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	list_for_each_entry(provider, &icc_providers, provider_list) {
> >   		struct icc_node *n;
> > @@ -73,7 +73,7 @@ static int icc_summary_show(struct seq_file *s, void *data)
> >   		}
> >   	}
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	return 0;
> >   }
> > @@ -104,7 +104,7 @@ static int icc_graph_show(struct seq_file *s, void *data)
> >   	int i;
> >   	seq_puts(s, "digraph {\n\trankdir = LR\n\tnode [shape = record]\n");
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	/* draw providers as cluster subgraphs */
> >   	cluster_index = 0;
> > @@ -136,7 +136,7 @@ static int icc_graph_show(struct seq_file *s, void *data)
> >   					icc_graph_show_link(s, 1, n,
> >   							    n->links[i]);
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	seq_puts(s, "}");
> >   	return 0;
> > @@ -362,7 +362,7 @@ struct icc_node_data *of_icc_get_from_provider(struct of_phandle_args *spec)
> >   	if (!spec)
> >   		return ERR_PTR(-EINVAL);
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	list_for_each_entry(provider, &icc_providers, provider_list) {
> >   		if (provider->dev->of_node == spec->np) {
> >   			if (provider->xlate_extended) {
> > @@ -378,7 +378,7 @@ struct icc_node_data *of_icc_get_from_provider(struct of_phandle_args *spec)
> >   			}
> >   		}
> >   	}
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	if (IS_ERR(node))
> >   		return ERR_CAST(node);
> > @@ -490,9 +490,9 @@ struct icc_path *of_icc_get_by_index(struct device *dev, int idx)
> >   		return ERR_CAST(dst_data);
> >   	}
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	path = path_find(dev, src_data->node, dst_data->node);
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	if (IS_ERR(path)) {
> >   		dev_err(dev, "%s: invalid path=%ld\n", __func__, PTR_ERR(path));
> >   		goto free_icc_data;
> > @@ -577,12 +577,12 @@ void icc_set_tag(struct icc_path *path, u32 tag)
> >   	if (!path)
> >   		return;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	for (i = 0; i < path->num_nodes; i++)
> >   		path->reqs[i].tag = tag;
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   }
> >   EXPORT_SYMBOL_GPL(icc_set_tag);
> > @@ -632,7 +632,7 @@ int icc_set_bw(struct icc_path *path, u32 avg_bw, u32 peak_bw)
> >   	if (WARN_ON(IS_ERR(path) || !path->num_nodes))
> >   		return -EINVAL;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	old_avg = path->reqs[0].avg_bw;
> >   	old_peak = path->reqs[0].peak_bw;
> > @@ -664,7 +664,7 @@ int icc_set_bw(struct icc_path *path, u32 avg_bw, u32 peak_bw)
> >   		apply_constraints(path);
> >   	}
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	trace_icc_set_bw_end(path, ret);
> > @@ -682,12 +682,12 @@ static int __icc_enable(struct icc_path *path, bool enable)
> >   	if (WARN_ON(IS_ERR(path) || !path->num_nodes))
> >   		return -EINVAL;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	for (i = 0; i < path->num_nodes; i++)
> >   		path->reqs[i].enabled = enable;
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	return icc_set_bw(path, path->reqs[0].avg_bw,
> >   			  path->reqs[0].peak_bw);
> > @@ -726,7 +726,7 @@ struct icc_path *icc_get(struct device *dev, const int src_id, const int dst_id)
> >   	struct icc_node *src, *dst;
> >   	struct icc_path *path = ERR_PTR(-EPROBE_DEFER);
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	src = node_find(src_id);
> >   	if (!src)
> > @@ -748,7 +748,7 @@ struct icc_path *icc_get(struct device *dev, const int src_id, const int dst_id)
> >   		path = ERR_PTR(-ENOMEM);
> >   	}
> >   out:
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	return path;
> >   }
> >   EXPORT_SYMBOL_GPL(icc_get);
> > @@ -773,14 +773,14 @@ void icc_put(struct icc_path *path)
> >   	if (ret)
> >   		pr_err("%s: error (%d)\n", __func__, ret);
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	for (i = 0; i < path->num_nodes; i++) {
> >   		node = path->reqs[i].node;
> >   		hlist_del(&path->reqs[i].req_node);
> >   		if (!WARN_ON(!node->provider->users))
> >   			node->provider->users--;
> >   	}
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	kfree_const(path->name);
> >   	kfree(path);
> > @@ -822,11 +822,11 @@ struct icc_node *icc_node_create(int id)
> >   {
> >   	struct icc_node *node;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	node = icc_node_create_nolock(id);
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	return node;
> >   }
> > @@ -840,7 +840,7 @@ void icc_node_destroy(int id)
> >   {
> >   	struct icc_node *node;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	node = node_find(id);
> >   	if (node) {
> > @@ -848,7 +848,7 @@ void icc_node_destroy(int id)
> >   		WARN_ON(!hlist_empty(&node->req_list));
> >   	}
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	kfree(node);
> >   }
> > @@ -876,7 +876,7 @@ int icc_link_create(struct icc_node *node, const int dst_id)
> >   	if (!node->provider)
> >   		return -EINVAL;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	dst = node_find(dst_id);
> >   	if (!dst) {
> > @@ -900,7 +900,7 @@ int icc_link_create(struct icc_node *node, const int dst_id)
> >   	node->links[node->num_links++] = dst;
> >   out:
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	return ret;
> >   }
> > @@ -925,7 +925,7 @@ int icc_link_destroy(struct icc_node *src, struct icc_node *dst)
> >   	if (IS_ERR_OR_NULL(dst))
> >   		return -EINVAL;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	for (slot = 0; slot < src->num_links; slot++)
> >   		if (src->links[slot] == dst)
> > @@ -946,7 +946,7 @@ int icc_link_destroy(struct icc_node *src, struct icc_node *dst)
> >   		ret = -ENOMEM;
> >   out:
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	return ret;
> >   }
> > @@ -962,7 +962,7 @@ void icc_node_add(struct icc_node *node, struct icc_provider *provider)
> >   	if (WARN_ON(node->provider))
> >   		return;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	node->provider = provider;
> >   	list_add_tail(&node->node_list, &provider->nodes);
> > @@ -988,7 +988,7 @@ void icc_node_add(struct icc_node *node, struct icc_provider *provider)
> >   	node->avg_bw = 0;
> >   	node->peak_bw = 0;
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   }
> >   EXPORT_SYMBOL_GPL(icc_node_add);
> > @@ -998,11 +998,11 @@ EXPORT_SYMBOL_GPL(icc_node_add);
> >    */
> >   void icc_node_del(struct icc_node *node)
> >   {
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	list_del(&node->node_list);
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   }
> >   EXPORT_SYMBOL_GPL(icc_node_del);
> > @@ -1041,12 +1041,12 @@ int icc_provider_add(struct icc_provider *provider)
> >   	if (WARN_ON(!provider->xlate && !provider->xlate_extended))
> >   		return -EINVAL;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	INIT_LIST_HEAD(&provider->nodes);
> >   	list_add_tail(&provider->provider_list, &icc_providers);
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   	dev_dbg(provider->dev, "interconnect provider added to topology\n");
> > @@ -1060,22 +1060,22 @@ EXPORT_SYMBOL_GPL(icc_provider_add);
> >    */
> >   void icc_provider_del(struct icc_provider *provider)
> >   {
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	if (provider->users) {
> >   		pr_warn("interconnect provider still has %d users\n",
> >   			provider->users);
> > -		mutex_unlock(&icc_lock);
> > +		rt_mutex_unlock(&icc_lock);
> >   		return;
> >   	}
> >   	if (!list_empty(&provider->nodes)) {
> >   		pr_warn("interconnect provider still has nodes\n");
> > -		mutex_unlock(&icc_lock);
> > +		rt_mutex_unlock(&icc_lock);
> >   		return;
> >   	}
> >   	list_del(&provider->provider_list);
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   }
> >   EXPORT_SYMBOL_GPL(icc_provider_del);
> > @@ -1110,7 +1110,7 @@ void icc_sync_state(struct device *dev)
> >   	if (count < providers_count)
> >   		return;
> > -	mutex_lock(&icc_lock);
> > +	rt_mutex_lock(&icc_lock);
> >   	synced_state = true;
> >   	list_for_each_entry(p, &icc_providers, provider_list) {
> >   		dev_dbg(p->dev, "interconnect provider is in synced state\n");
> > @@ -1123,7 +1123,7 @@ void icc_sync_state(struct device *dev)
> >   			}
> >   		}
> >   	}
> > -	mutex_unlock(&icc_lock);
> > +	rt_mutex_unlock(&icc_lock);
> >   }
> >   EXPORT_SYMBOL_GPL(icc_sync_state);
> 
> We should also update Kconfig, otherwise i am expecting that the build will fail if
> CONFIG_RT_MUTEXES is not enabled.

You're right, it will.

> 
> Thanks,
> Georgi

Powered by blists - more mailing lists