interconnect: Fix sync-state issues
Subsystems (2)
cover interconnect: Fix sync-state issues
These patches fix a couple of sync-state bugs that either cause the initial BW floors to be ignored entirely, or to be never removed after sync-state is called.
Mike Tipton (4):
interconnect: Zero initial BW after sync-state
interconnect: Always call pre_aggregate before aggregate
interconnect: qcom: icc-rpmh: Ensure floor BW is enforced for all nodes
interconnect: qcom: icc-rpmh: Add BCMs to commit list in pre_aggregate
drivers/interconnect/core.c | 7 ++++++- drivers/interconnect/qcom/icc-rpmh.c | 20 ++++++++++---------- 2 files changed, 16 insertions(+), 11 deletions(-)
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
PATCH 1/4 interconnect: Zero initial BW after sync-state
The initial BW values may be used by providers to enforce floors. Zero these values after sync-state so that providers know when to stop enforcing them.
--- drivers/interconnect/core.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/interconnect/core.c +2 -0
@@ -1106,6 +1106,8 @@ void icc_sync_state(struct device *dev)dev_dbg (p->dev,"interconnect provider is in synced state\n" );list_for_each_entry (n, &p->nodes, node_list) {if (n->init_avg || n->init_peak) { + n->init_avg =0 ; + n->init_peak =0 ;aggregate_requests (n); p->set (n, n); }
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
Re: [PATCH 1/4] interconnect: Zero initial BW after sync-state
Quoted 15 lines · Mike Tipton
The initial BW values may be used by providers to enforce floors. Zero these values after sync-state so that providers know when to stop enforcing them.
Fixes b1d681d8d324 ("interconnect: Add sync state support")Signed-off-by Mike Tipton <[email protected]>--- drivers/interconnect/core.c | 2 ++ 1 file changed, 2 insertions(+)drivers/interconnect/core.c +0
-0
drivers/interconnect/core.c void icc_sync_state(struct device *dev) · line 1110
Quoted 5 lines
drivers/interconnect/core.c +2
-0@@ -1106,6 +1106,8 @@ void icc_sync_state(struct device *dev)dev_dbg (p->dev,"interconnect provider is in synced state\n" );list_for_each_entry (n, &p->nodes, node_list) {if (n->init_avg || n->init_peak) { + n->init_avg =0 ; + n->init_peak =0 ;
nit: It is good to reset init/floor levels back to zero, but we don't need to do this as we have sync_state flag to let providers know when to stop enforcing.
Quoted 3 lines
drivers/interconnect/core.c +0
-0aggregate_requests (n); p->set (n, n); }
Re: [PATCH 1/4] interconnect: Zero initial BW after sync-state
Quoted 11 lines · On 7/1/2021 9:56 AM, [email protected] wrote:
Quoted 10 lines · Mike Tipton
The initial BW values may be used by providers to enforce floors. Zero these values after sync-state so that providers know when to stop enforcing them.
Fixes b1d681d8d324 ("interconnect: Add sync state support")Signed-off-by Mike Tipton <[email protected]>---
drivers/interconnect/core.c | 2 ++1 file changed, 2 insertions(+)
Quoted 11 lines
drivers/interconnect/core.c void icc_sync_state(struct device *dev) · line 1110
Quoted 10 lines
drivers/interconnect/core.c +2
-0@@ -1106,6 +1106,8 @@ void icc_sync_state(struct device *dev) dev_dbg(p->dev, "interconnect provider is in synced state\n"); list_for_each_entry(n, &p->nodes, node_list) { if (n->init_avg || n->init_peak) { + n->init_avg = 0; + n->init_peak = 0;nit: It is good to reset init/floor levels back to zero, but we don't need to do this as we have
sync_stateflag to let providers know when to stop enforcing.
The synced_state variable is static to this file. It's not exposed to providers. In fact, we could entirely remove synced_state with this patch since it's unnecessary after zeroing the initial floors.
Quoted 3 lines
Quoted 3 lines
drivers/interconnect/core.c +0
-0aggregate_requests(n); p->set(n, n); }
PATCH 2/4 interconnect: Always call pre_aggregate before aggregate
In reply to Mike TiptonThe pre_aggregate callback isn't called in all cases before calling aggregate. Add the missing calls so providers can rely on consistent framework behavior.
--- drivers/interconnect/core.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/interconnect/core.c +4 -1
@@ -973,9 +973,12 @@ void icc_node_add(struct icc_node *node, struct icc_provider *provider) } node->avg_bw = node->init_avg; node->peak_bw = node->init_peak; -if (provider->aggregate) +if (provider->aggregate) { +if (provider->pre_aggregate) + provider->pre_aggregate (node); provider->aggregate (node,0 , node->init_avg, node->init_peak, &node->avg_bw, &node->peak_bw); + } provider->set (node, node); node->avg_bw =0 ; node->peak_bw =0 ;
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
Re: [PATCH 2/4] interconnect: Always call pre_aggregate before aggregate
Quoted 15 lines · Mike Tipton
The
pre_aggregatecallback isn't called in all cases before calling aggregate. Add the missing calls so providers can rely on consistent framework behavior.Fixes d3703b3e255f ("interconnect: Aggregate before setting initial bandwidth")Signed-off-by Mike Tipton <[email protected]>--- drivers/interconnect/core.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-)drivers/interconnect/core.c +0
-0
drivers/interconnect/core.c void icc_node_add(struct icc_node *node, struct · line 979
Quoted 8 lines
drivers/interconnect/core.c +3
-1@@ -973,9 +973,12 @@ void icc_node_add(struct icc_node *node, struct icc_provider *provider) } node->avg_bw = node->init_avg; node->peak_bw = node->init_peak; -if (provider->aggregate) +if (provider->aggregate) { +if (provider->pre_aggregate) + provider->pre_aggregate (node);
nit: we can invoke pre_aggregate() out side of if (qcom_icc_aggregate).
Quoted 6 lines
drivers/interconnect/core.c +1
-0provider->aggregate (node,0 , node->init_avg, node->init_peak, &node->avg_bw, &node->peak_bw); + } provider->set (node, node); node->avg_bw =0 ;
node->peak_bw= 0;
Re: [PATCH 2/4] interconnect: Always call pre_aggregate before aggregate
Quoted 11 lines · On 7/1/2021 11:55 AM, [email protected] wrote:
Quoted 10 lines · Mike Tipton
The
pre_aggregatecallback isn't called in all cases before calling aggregate. Add the missing calls so providers can rely on consistent framework behavior.Fixes d3703b3e255f ("interconnect: Aggregate before setting initial bandwidth")Signed-off-by Mike Tipton <[email protected]>---
drivers/interconnect/core.c | 5 ++++-1 file changed, 4 insertions(+), 1 deletion(-)
Quoted 14 lines
drivers/interconnect/core.c void icc_node_add(struct icc_node *node, struct · line 979
Quoted 13 lines
drivers/interconnect/core.c +3
-1@@ -973,9 +973,12 @@ void icc_node_add(struct icc_node *node, struct icc_provider *provider) } node->avg_bw = node->init_avg; node->peak_bw = node->init_peak; - if (provider->aggregate) + if (provider->aggregate) { + if (provider->pre_aggregate) + provider->pre_aggregate(node);nit: we can invoke
pre_aggregate()out side of if (qcom_icc_aggregate).
Sure, will update this.
Quoted 7 lines
Quoted 7 lines
drivers/interconnect/core.c +1
-0provider->aggregate(node, 0, node->init_avg, node->init_peak, &node->avg_bw, &node->peak_bw); + } provider->set(node, node); node->avg_bw = 0;
node->peak_bw= 0;
PATCH 3/4 interconnect: qcom: icc-rpmh: Ensure floor BW is enforced for all nodes
In reply to Mike TiptonWe currently only enforce BW floors for a subset of nodes in a path. All BCMs that need updating are queued in the pre_aggregate/aggregate phase. The first set() commits all queued BCMs and subsequent set() calls short-circuit without committing anything. Since the floor BW isn't set in sum_avg/max_peak until set(), then some BCMs are committed before their associated nodes reflect the floor.
Set the floor as each node is being aggregated. This ensures that all all relevant floors are set before the BCMs are committed.
--- drivers/interconnect/qcom/icc-rpmh.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/interconnect/qcom/icc-rpmh.c +5 -5
@@ -57,6 +57,11 @@ int qcom_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw, qn->sum_avg[i] += avg_bw; qn->max_peak[i] =max_t (u32, qn->max_peak[i], peak_bw); } + +if (node->init_avg || node->init_peak) { + qn->sum_avg[i] =max_t (u64, qn->sum_avg[i], node->init_avg); + qn->max_peak[i] =max_t (u64, qn->max_peak[i], node->init_peak); + } } *agg_avg += avg_bw;
@@ -90,11 +95,6 @@ int qcom_icc_set(struct icc_node *src, struct icc_node *dst) qp =to_qcom_provider (node->provider); qn = node->data; - qn->sum_avg[QCOM_ICC_BUCKET_AMC ] =max_t (u64, qn->sum_avg[QCOM_ICC_BUCKET_AMC ], - node->avg_bw); - qn->max_peak[QCOM_ICC_BUCKET_AMC ] =max_t (u64, qn->max_peak[QCOM_ICC_BUCKET_AMC ], - node->peak_bw); -qcom_icc_bcm_voter_commit (qp->voter);return 0 ;
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
kernel test robot
build botRe: [PATCH 3/4] interconnect: qcom: icc-rpmh: Ensure floor BW is enforced for all nodes
Thank you for the patch! Perhaps something to improve:
Hi Mike,
Thank you for the patch! Perhaps something to improve:
[auto build test WARNING on linus/master]
[also build test WARNING on v5.13-rc7 next-20210625]
[If your patch is applied to the wrong git tree, kindly drop us a note. And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch]
url: https://github.com/0day-ci/linux/commits/Mike-Tipton/interconnect-Fix-sync-state-issues/20210626-053132 base: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 55fcd4493da5ac8a0f7a0b3b5ae8448aee2041bb config: arm64-allyesconfig (attached as .config) compiler: aarch64-linux-gcc (GCC) 9.3.0
reproduce (this is a W=1 build):
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# https://github.com/0day-ci/linux/commit/ad01b68a345bdae6c24bd45720f4a01a2f9be49b
git remote add linux-review https://github.com/0day-ci/linux
git fetch --no-tags linux-review Mike-Tipton/interconnect-Fix-sync-state-issues/20210626-053132 git checkout ad01b68a345bdae6c24bd45720f4a01a2f9be49b
# save the attached .config to linux build tree
COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=arm64
If you fix the issue, kindly add following tag as appropriate
All warnings (new ones prefixed by >>):
drivers/interconnect/qcom/icc-rpmh.c: In function 'qcom_icc_set':
>> drivers/interconnect/qcom/icc-rpmh.c:87:24: warning: variable 'qn' set but not used [-Wunused-but-set-variable]
87 | struct qcom_icc_node *qn;
| ^~vim +/qn +87 drivers/interconnect/qcom/icc-rpmh.c
976daac4a1c581 David Dai 2020-02-28 76
976daac4a1c581 David Dai 2020-02-28 77 /**
976daac4a1c581 David Dai 2020-02-28 78 * qcom_icc_set - set the constraints based on path
976daac4a1c581 David Dai 2020-02-28 79 * @src: source node for the path to set constraints on
976daac4a1c581 David Dai 2020-02-28 80 * @dst: destination node for the path to set constraints on
976daac4a1c581 David Dai 2020-02-28 81 *
976daac4a1c581 David Dai 2020-02-28 82 * Return: 0 on success, or an error code otherwise
976daac4a1c581 David Dai 2020-02-28 83 */
976daac4a1c581 David Dai 2020-02-28 84 int qcom_icc_set(struct icc_node *src, struct icc_node *dst)
976daac4a1c581 David Dai 2020-02-28 85 {
976daac4a1c581 David Dai 2020-02-28 86 struct qcom_icc_provider *qp;
266cd33b591385 Georgi Djakov 2020-10-22 @87 struct qcom_icc_node *qn;
976daac4a1c581 David Dai 2020-02-28 88 struct icc_node *node;
976daac4a1c581 David Dai 2020-02-28 89
976daac4a1c581 David Dai 2020-02-28 90 if (!src)
976daac4a1c581 David Dai 2020-02-28 91 node = dst;
976daac4a1c581 David Dai 2020-02-28 92 else
976daac4a1c581 David Dai 2020-02-28 93 node = src;
976daac4a1c581 David Dai 2020-02-28 94
976daac4a1c581 David Dai 2020-02-28 95 qp = to_qcom_provider(node->provider);
266cd33b591385 Georgi Djakov 2020-10-22 96 qn = node->data;
266cd33b591385 Georgi Djakov 2020-10-22 97
976daac4a1c581 David Dai 2020-02-28 98 qcom_icc_bcm_voter_commit(qp->voter);
976daac4a1c581 David Dai 2020-02-28 99
976daac4a1c581 David Dai 2020-02-28 100 return 0;
976daac4a1c581 David Dai 2020-02-28 101 }
976daac4a1c581 David Dai 2020-02-28 102 EXPORT_SYMBOL_GPL(qcom_icc_set);
976daac4a1c581 David Dai 2020-02-28 103 ---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
Re: [PATCH 3/4] interconnect: qcom: icc-rpmh: Ensure floor BW is enforced for all nodes
In reply to Mike TiptonQuoted 23 lines · Mike Tipton
We currently only enforce BW floors for a subset of nodes in a path. All BCMs that need updating are queued in the pre_aggregate/aggregate phase. The first
set()commits all queued BCMs and subsequentset()calls short-circuit without committing anything. Since the floor BW isn't set in sum_avg/max_peak untilset(), then some BCMs are committed before their associated nodes reflect the floor.Set the floor as each node is being aggregated. This ensures that all all relevant floors are set before the BCMs are committed.
Fixes 266cd33b5913 ("interconnect: qcom: Ensure that the floorbandwidth value is enforced")
Signed-off-by Mike Tipton <[email protected]>--- drivers/interconnect/qcom/icc-rpmh.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)drivers/interconnect/qcom/icc-rpmh.c +0
-0
drivers/interconnect/qcom/icc-rpmh.c int qcom_icc_aggregate(struct icc_node *node, u32 · line 65
Quoted 9 lines
drivers/interconnect/qcom/icc-rpmh.c +5
-0@@ -57,6 +57,11 @@ int qcom_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw, qn->sum_avg[i] += avg_bw; qn->max_peak[i] =max_t (u32, qn->max_peak[i], peak_bw); } + +if (node->init_avg || node->init_peak) { + qn->sum_avg[i] =max_t (u64, qn->sum_avg[i], node->init_avg); + qn->max_peak[i] =max_t (u64, qn->max_peak[i], node->init_peak); + }
Hi Mike,
Original problem is BCMs not getting added to commit_list for unused nodes, right? that is solved by moving *_bcm_voter_add() to pre_aggregate(). I could not get why we need to do above change, we are enforcing node votes with floor votes in framework + below code snippet that you removed. How would adding this code in qcom_icc_aggregate() make difference? Is there any other issue that i am not to able to get?
Quoted 18 lines
drivers/interconnect/qcom/icc-rpmh.c +0
-0}drivers/interconnect/qcom/icc-rpmh.c +0
-5*agg_avg += avg_bw;@@ -90,11 +95,6 @@ int qcom_icc_set(struct icc_node *src, struct icc_node *dst) qp =to_qcom_provider (node->provider); qn = node->data; - qn->sum_avg[QCOM_ICC_BUCKET_AMC ] =max_t (u64, qn->sum_avg[QCOM_ICC_BUCKET_AMC ], - node->avg_bw); - qn->max_peak[QCOM_ICC_BUCKET_AMC ] =max_t (u64, qn->max_peak[QCOM_ICC_BUCKET_AMC], - node->peak_bw); -drivers/interconnect/qcom/icc-rpmh.c +0
-0qcom_icc_bcm_voter_commit (qp->voter);return 0 ;
Re: [PATCH 3/4] interconnect: qcom: icc-rpmh: Ensure floor BW is enforced for all nodes
Quoted 30 lines · On 7/1/2021 11:48 AM, [email protected] wrote:
Quoted 29 lines · Mike Tipton
We currently only enforce BW floors for a subset of nodes in a path. All BCMs that need updating are queued in the pre_aggregate/aggregate phase. The first
set()commits all queued BCMs and subsequentset()calls short-circuit without committing anything. Since the floor BW isn't set in sum_avg/max_peak untilset(), then some BCMs are committed before their associated nodes reflect the floor.Set the floor as each node is being aggregated. This ensures that all all relevant floors are set before the BCMs are committed.
Fixes 266cd33b5913 ("interconnect: qcom: Ensure that the floorbandwidth value is enforced")
Signed-off-by Mike Tipton <[email protected]>---
drivers/interconnect/qcom/icc-rpmh.c | 10 +++++-----1 file changed, 5 insertions(+), 5 deletions(-)drivers/interconnect/qcom/icc-rpmh.c +2
-0@@ -57,6 +57,11 @@ int qcom_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw, qn->sum_avg[i] += avg_bw; qn->max_peak[i] = max_t(u32, qn->max_peak[i], peak_bw); } + + if (node->init_avg || node->init_peak) {
Quoted 8 lines
drivers/interconnect/qcom/icc-rpmh.c int qcom_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw, · line 64
Quoted 4 lines
drivers/interconnect/qcom/icc-rpmh.c +3
-0+ qn->sum_avg[i] = max_t(u64, qn->sum_avg[i], node->init_avg); + qn->max_peak[i] = max_t(u64, qn->max_peak[i], node->init_peak); + }Hi Mike,
Original problem is BCMs not getting added tocommit_listfor unused nodes, right? that is solved by moving*_bcm_voter_add()topre_aggregate(). I could not get why we need to do above change, we are enforcing node votes with floor votes in framework + below code snippet that you removed. How would adding this code inqcom_icc_aggregate()make difference? Is there any other issue that i am not to able to get?
This series fixes a couple of separate issues. One is not removing the initial floor after sync_state for unvoted paths, which the change to add BCMs to the commit list in pre_aggregate addresses. The other issue is not properly enforcing the initial floor *before* sync_state, which this patch addresses.
We only commit to HW what we've aggregated in our internal, provider-specific buckets (AMC, WAKE, SLEEP) during pre_aggregate/aggregate. All BCMs in the path are added to the commit list during this stage. Everything in the commit list is voted on the *first* set() callback for the path. The commit list is empty for every subsequent set() callback, so nothing actually happens except in the first set().
The original snippet below in qcom_icc_set() partially re-aggregates AMC with BW values from the icc_node struct. This is to capture the floor enforced by the framework itself. However, this is only for a single node in the path, which means it's also only for a single BCM in the path. Since we commit all BCMs on the first set(), this means most BCMs in the path don't see the floor BW at the time we commit them. This can lead to some BCMs being under-voted, or in some cases completely off.
So, instead of aggregating the floor for each node in qcom_icc_set(), this patch moves it to qcom_icc_aggregate(). This means all floors are captured internally before committing the BCMs.
Quoted 18 lines
Quoted 18 lines
drivers/interconnect/qcom/icc-rpmh.c +0
-0}
*agg_avg+=avg_bw;drivers/interconnect/qcom/icc-rpmh.c +0
-5@@ -90,11 +95,6 @@ int qcom_icc_set(struct icc_node *src, struct icc_node *dst) qp = to_qcom_provider(node->provider); qn = node->data; - qn->sum_avg[QCOM_ICC_BUCKET_AMC] = max_t(u64, qn->sum_avg[QCOM_ICC_BUCKET_AMC ], - node->avg_bw); - qn->max_peak[QCOM_ICC_BUCKET_AMC] = max_t(u64, qn->max_peak[QCOM_ICC_BUCKET_AMC], - node->peak_bw); -
qcom_icc_bcm_voter_commit(qp->voter);return 0;
PATCH 4/4 interconnect: qcom: icc-rpmh: Add BCMs to commit list in pre_aggregate
In reply to Mike TiptonWe're only adding BCMs to the commit list in aggregate(), but there are cases where pre_aggregate() is called without subsequently calling aggregate(). In particular, in icc_sync_state() when a node with initial BW has zero requests. Since BCMs aren't added to the commit list in these cases, we don't actually send the zero BW request to HW. So the resources remain on unnecessarily.
Add BCMs to the commit list in pre_aggregate() instead, which is always called even when there are no requests.
--- drivers/interconnect/qcom/icc-rpmh.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/interconnect/qcom/icc-rpmh.c +5 -5
@@ -20,13 +20,18 @@ void qcom_icc_pre_aggregate(struct icc_node *node) {size_t i;struct qcom_icc_node *qn; +struct qcom_icc_provider *qp; qn = node->data; + qp =to_qcom_provider (node->provider);for (i =0 ; i <QCOM_ICC_NUM_BUCKETS ; i++) { qn->sum_avg[i] =0 ; qn->max_peak[i] =0 ; } + +for (i =0 ; i < qn->num_bcms; i++) +qcom_icc_bcm_voter_add (qp->voter, qn->bcms[i]); }EXPORT_SYMBOL_GPL (qcom_icc_pre_aggregate);
@@ -44,10 +49,8 @@ int qcom_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw, {size_t i;struct qcom_icc_node *qn; -struct qcom_icc_provider *qp; qn = node->data; - qp =to_qcom_provider (node->provider);if (!tag) tag =QCOM_ICC_TAG_ALWAYS ;
@@ -67,9 +70,6 @@ int qcom_icc_aggregate(struct icc_node *node, u32 tag, u32 avg_bw, *agg_avg += avg_bw; *agg_peak =max_t (u32, *agg_peak, peak_bw); -for (i =0 ; i < qn->num_bcms; i++) -qcom_icc_bcm_voter_add (qp->voter, qn->bcms[i]); -return 0 ; }EXPORT_SYMBOL_GPL (qcom_icc_aggregate);
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project