2020-07-21 12:08:27

by Georgi Djakov

[permalink] [raw]
Subject: [PATCH] interconnect: Do not skip aggregation for disabled paths

When an interconnect path is being disabled, currently we don't aggregate
the requests for it afterwards. But the re-aggregation step shouldn't be
skipped, as it may leave the nodes with outdated bandwidth data. This
outdated data may actually keep the path still enabled and prevent the
device from going into lower power states.

Reported-by: Atul Dhudase <[email protected]>
Fixes: 7d374b209083 ("interconnect: Add helpers for enabling/disabling a path")
Signed-off-by: Georgi Djakov <[email protected]>
---
drivers/interconnect/core.c | 12 +++++++++---
1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/interconnect/core.c b/drivers/interconnect/core.c
index 37d5ec970cc1..5174dcb31ab7 100644
--- a/drivers/interconnect/core.c
+++ b/drivers/interconnect/core.c
@@ -243,6 +243,7 @@ static int aggregate_requests(struct icc_node *node)
{
struct icc_provider *p = node->provider;
struct icc_req *r;
+ u32 avg_bw, peak_bw;

node->avg_bw = 0;
node->peak_bw = 0;
@@ -251,9 +252,14 @@ static int aggregate_requests(struct icc_node *node)
p->pre_aggregate(node);

hlist_for_each_entry(r, &node->req_list, req_node) {
- if (!r->enabled)
- continue;
- p->aggregate(node, r->tag, r->avg_bw, r->peak_bw,
+ if (r->enabled) {
+ avg_bw = r->avg_bw;
+ peak_bw = r->peak_bw;
+ } else {
+ avg_bw = 0;
+ peak_bw = 0;
+ }
+ p->aggregate(node, r->tag, avg_bw, peak_bw,
&node->avg_bw, &node->peak_bw);
}


2020-07-21 13:33:45

by Sibi Sankar

[permalink] [raw]
Subject: Re: [PATCH] interconnect: Do not skip aggregation for disabled paths

Hey Georgi,
Thanks for the patch!

On 2020-07-21 17:37, Georgi Djakov wrote:
> When an interconnect path is being disabled, currently we don't
> aggregate
> the requests for it afterwards. But the re-aggregation step shouldn't
> be
> skipped, as it may leave the nodes with outdated bandwidth data. This
> outdated data may actually keep the path still enabled and prevent the
> device from going into lower power states.
>
> Reported-by: Atul Dhudase <[email protected]>
> Fixes: 7d374b209083 ("interconnect: Add helpers for enabling/disabling
> a path")
> Signed-off-by: Georgi Djakov <[email protected]>
> ---

Reviewed-by: Sibi Sankar <[email protected]>

> drivers/interconnect/core.c | 12 +++++++++---
> 1 file changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/interconnect/core.c b/drivers/interconnect/core.c
> index 37d5ec970cc1..5174dcb31ab7 100644
> --- a/drivers/interconnect/core.c
> +++ b/drivers/interconnect/core.c
> @@ -243,6 +243,7 @@ static int aggregate_requests(struct icc_node
> *node)
> {
> struct icc_provider *p = node->provider;
> struct icc_req *r;
> + u32 avg_bw, peak_bw;
>
> node->avg_bw = 0;
> node->peak_bw = 0;
> @@ -251,9 +252,14 @@ static int aggregate_requests(struct icc_node
> *node)
> p->pre_aggregate(node);
>
> hlist_for_each_entry(r, &node->req_list, req_node) {
> - if (!r->enabled)
> - continue;
> - p->aggregate(node, r->tag, r->avg_bw, r->peak_bw,
> + if (r->enabled) {
> + avg_bw = r->avg_bw;
> + peak_bw = r->peak_bw;
> + } else {
> + avg_bw = 0;
> + peak_bw = 0;
> + }
> + p->aggregate(node, r->tag, avg_bw, peak_bw,
> &node->avg_bw, &node->peak_bw);
> }

--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project.

2020-07-21 15:21:17

by adhudase

[permalink] [raw]
Subject: Re: [PATCH] interconnect: Do not skip aggregation for disabled paths

On 2020-07-21 17:37, Georgi Djakov wrote:
> When an interconnect path is being disabled, currently we don't
> aggregate
> the requests for it afterwards. But the re-aggregation step shouldn't
> be
> skipped, as it may leave the nodes with outdated bandwidth data. This
> outdated data may actually keep the path still enabled and prevent the
> device from going into lower power states.
>
> Reported-by: Atul Dhudase <[email protected]>
> Fixes: 7d374b209083 ("interconnect: Add helpers for enabling/disabling
> a path")
> Signed-off-by: Georgi Djakov <[email protected]>
> ---

Tested-by: Atul Dhudase <[email protected]>
Reviewed-by: Atul Dhudase <[email protected]>

> drivers/interconnect/core.c | 12 +++++++++---
> 1 file changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/interconnect/core.c b/drivers/interconnect/core.c
> index 37d5ec970cc1..5174dcb31ab7 100644
> --- a/drivers/interconnect/core.c
> +++ b/drivers/interconnect/core.c
> @@ -243,6 +243,7 @@ static int aggregate_requests(struct icc_node
> *node)
> {
> struct icc_provider *p = node->provider;
> struct icc_req *r;
> + u32 avg_bw, peak_bw;
>
> node->avg_bw = 0;
> node->peak_bw = 0;
> @@ -251,9 +252,14 @@ static int aggregate_requests(struct icc_node
> *node)
> p->pre_aggregate(node);
>
> hlist_for_each_entry(r, &node->req_list, req_node) {
> - if (!r->enabled)
> - continue;
> - p->aggregate(node, r->tag, r->avg_bw, r->peak_bw,
> + if (r->enabled) {
> + avg_bw = r->avg_bw;
> + peak_bw = r->peak_bw;
> + } else {
> + avg_bw = 0;
> + peak_bw = 0;
> + }
> + p->aggregate(node, r->tag, avg_bw, peak_bw,
> &node->avg_bw, &node->peak_bw);
> }