- Kernel - mailweb.openeuler.org

[openeuler:openEuler-1.0-LTS 1291/1291] drivers/char/svm.c:312:17-38: WARNING: atomic_dec_and_test variation before object free at line 335.
by kernel test robot 14 Nov '24

14 Nov '24

tree: https://gitee.com/openeuler/kernel.git openEuler-1.0-LTS head: 5769bea33a24c5a60de09ceaeb12f349d1161f3b commit: 4cdb60f56180d6ef7c10bd0e50ddda96ebc9f783 [1291/1291] svm: implement unpin pages to dec refcount config: arm64-randconfig-r063-20241113 (https://download.01.org/0day-ci/archive/20241114/202411142318.PmiGB1Ht-lkp@…) compiler: aarch64-linux-gcc (GCC) 14.2.0 If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp(a)intel.com> | Closes: https://lore.kernel.org/oe-kbuild-all/202411142318.PmiGB1Ht-lkp@intel.com/ cocci warnings: (new ones prefixed by >>) >> drivers/char/svm.c:312:17-38: WARNING: atomic_dec_and_test variation before object free at line 335. vim +312 drivers/char/svm.c 304 305 static void svm_remove_sdma(struct svm_process *process, 306 struct svm_sdma *sdma, bool try_rm) 307 { 308 int null_count = 0; 309 310 mutex_lock(&process->mutex); 311 > 312 if (try_rm && (!atomic64_dec_and_test(&sdma->ref))) { 313 mutex_unlock(&process->mutex); 314 return; 315 } 316 317 rb_erase(&sdma->node, &process->sdma_list); 318 RB_CLEAR_NODE(&sdma->node); 319 mutex_unlock(&process->mutex); 320 321 while (sdma->nr_pages--) { 322 if (sdma->pages[sdma->nr_pages] == NULL) { 323 pr_err("null pointer, nr_pages:%d.\n", sdma->nr_pages); 324 null_count++; 325 continue; 326 } 327 328 put_page(sdma->pages[sdma->nr_pages]); 329 } 330 331 if (null_count) 332 dump_stack(); 333 334 kfree(sdma->pages); > 335 kfree(sdma); 336 } 337 -- 0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki

1 0

[PATCH OLK-6.6 0/5] IMA RoT
by GONG Ruiqi 14 Nov '24

14 Nov '24

This patch set introduces IMA's Root of Trust (RoT), which is to support new RoT implementations such as Intel TDX and VirtCCA, as well as the classic TPM. Reference for Intel TDX with IMA: https://www.intel.cn/content/www/cn/zh/developer/articles/community/runtime… Reference for VirtCCA: https://gitee.com/openeuler/kernel/blob/OLK-6.6/Documentation/virtcca/virtc… GONG Ruiqi (5): ima: rot: Introduce basic framework ima: rot: Prepare TPM as an RoT ima: rot: Make RoT kick in ima: Rename ima_cvm to ima_virtcca ima: rot: Adapt VirtCCA into Rot security/integrity/ima/Makefile | 5 +- security/integrity/ima/ima.h | 5 +- security/integrity/ima/ima_api.c | 2 +- security/integrity/ima/ima_crypto.c | 48 +++++------- security/integrity/ima/ima_cvm.c | 77 ------------------ security/integrity/ima/ima_cvm.h | 36 --------- security/integrity/ima/ima_init.c | 31 ++------ security/integrity/ima/ima_queue.c | 51 ++++-------- security/integrity/ima/ima_rot.c | 86 ++++++++++++++++++++ security/integrity/ima/ima_rot.h | 28 +++++++ security/integrity/ima/ima_template.c | 2 +- security/integrity/ima/ima_template_lib.c | 2 +- security/integrity/ima/ima_tpm.c | 63 +++++++++++++++ security/integrity/ima/ima_tpm.h | 21 +++++ security/integrity/ima/ima_virtcca.c | 95 +++++++++++++++++++++++ security/integrity/ima/ima_virtcca.h | 13 ++++ 16 files changed, 355 insertions(+), 210 deletions(-) delete mode 100644 security/integrity/ima/ima_cvm.c delete mode 100644 security/integrity/ima/ima_cvm.h create mode 100644 security/integrity/ima/ima_rot.c create mode 100644 security/integrity/ima/ima_rot.h create mode 100644 security/integrity/ima/ima_tpm.c create mode 100644 security/integrity/ima/ima_tpm.h create mode 100644 security/integrity/ima/ima_virtcca.c create mode 100644 security/integrity/ima/ima_virtcca.h -- 2.25.1

2 6

[PATCH OLK-5.10] RDMA/hns: Fix the valid QP bank set to improve performance
by Chengchang Tang 14 Nov '24

14 Nov '24

From: wenglianfa <wenglianfa(a)huawei.com> driver inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/IB2HJE ---------------------------------------------------------------------- QP banks 0, 1, 4, and 5 use one HW FIFO, while QP banks 2, 3, 6, and 7 use another. Currently, if HNS_ROCE_CAP_FLAG_LIMIT_BANK is set, only QP banks 0, 1, 4, and 5 are used, resulting in a single HW FIFO being utilized. Instead, it is proposed to use QP banks 0, 1, 6, and 7 to leverage different FIFOs and improve performance. Fixes: 1c16701634e4 ("RDMA/hns: Fix RoCEE hang when multiple QP banks use EXT_SGE") Signed-off-by: wenglianfa <wenglianfa(a)huawei.com> Signed-off-by: Xinghai Cen <cenxinghai(a)h-partners.com> --- drivers/infiniband/hw/hns/hns_roce_device.h | 4 ++-- drivers/infiniband/hw/hns/hns_roce_qp.c | 3 ++- 2 files changed, 4 insertions(+), 3 deletions(-) diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h index e4d602bbeaec..a6c4ef631bb7 100644 --- a/drivers/infiniband/hw/hns/hns_roce_device.h +++ b/drivers/infiniband/hw/hns/hns_roce_device.h @@ -113,9 +113,9 @@ #define CQ_BANKID_MASK GENMASK(1, 0) #define VALID_CQ_BANK_MASK_DEFAULT 0xF -#define VALID_CQ_BANK_MASK_LIMIT 0x5 +#define VALID_CQ_BANK_MASK_LIMIT 0x9 -#define QP_HARDEN_MASK GENMASK(1, 0) +#define VALID_EXT_SGE_QP_BANK_MASK_LIMIT 0x41 #define HNS_ROCE_MAX_CQ_COUNT 0xFFFF #define HNS_ROCE_MAX_CQ_PERIOD 0xFFFF diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c index f90bff2ff0c0..98d9f3a77997 100644 --- a/drivers/infiniband/hw/hns/hns_roce_qp.c +++ b/drivers/infiniband/hw/hns/hns_roce_qp.c @@ -244,7 +244,8 @@ static u8 select_qp_bankid(struct hns_roce_dev *hr_dev, continue; if ((hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_LIMIT_BANK) && - use_ext_sge(init_attr) && (QP_HARDEN_MASK & i)) + use_ext_sge(init_attr) && + !(VALID_EXT_SGE_QP_BANK_MASK_LIMIT & BIT(i))) continue; valid_qp_bank_mask |= BIT(i); -- 2.33.0

2 1

[PATCH openEuler-1.0-LTS] Revert "sched/dynamic_affinity: fix preffered_cpu offline problem"
by He Yujie 14 Nov '24

14 Nov '24

hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/IB42TC CVE: NA -------------------------------- This reverts commit 3b2c79ffe971f042f1cdfbd42e2cf52ab0caf6e7. This patch 3b2c79ffe971 ("sched/dynamic_affinity: fix preffered_cpu offline problem") want to fix preffer_cpu offline issue but The requirement that causes the problem is not merge in this version. Therefore, this patch is not required. Fixes: 3b2c79ffe971 ("sched/dynamic_affinity: fix preffered_cpu offline problem") Signed-off-by: He Yujie <coka.heyujie(a)huawei.com> --- kernel/sched/fair.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 41b339c5b671..63f4344ac344 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7248,7 +7248,6 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, struct task_group *tg; long spare; int cpu, mode; - int nr_cpus_valid = 0; rcu_read_lock(); mode = dynamic_affinity_mode(p); @@ -7266,7 +7265,7 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, /* manual mode */ tg = task_group(p); - for_each_cpu_and(cpu, p->prefer_cpus, cpu_online_mask) { + for_each_cpu(cpu, p->prefer_cpus) { if (unlikely(!tg->se[cpu])) continue; @@ -7290,18 +7289,10 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, util_avg_sum += tg->se[cpu]->avg.util_avg; tg_capacity += capacity_of(cpu); - nr_cpus_valid++; } rcu_read_unlock(); - /* - * Follow cases should select cpus_ptr, checking by condition of - * tg_capacity > nr_cpus_valid: - * 1. all prefer_cpus offline; - * 2. all prefer_cpus has no cfs capaicity(tg_capacity = - * nr_cpus_valid * 1) - */ - if (tg_capacity > nr_cpus_valid && + if (tg_capacity > cpumask_weight(p->prefer_cpus) && util_avg_sum * 100 <= tg_capacity * sysctl_sched_util_low_pct) { p->select_cpus = p->prefer_cpus; if (sd_flag & SD_BALANCE_WAKE) -- 2.34.1

2 1

[PATCH openEuler-1.0-LTS] Revert "sched/dynamic_affinity: fix preffered_cpu offline problem"
by He Yujie 14 Nov '24

14 Nov '24

hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/IB42TC CVE: NA -------------------------------- This reverts commit 3b2c79ffe971f042f1cdfbd42e2cf52ab0caf6e7. This patch 3b2c79ffe971 ("sched/dynamic_affinity: fix preffered_cpu offline problem") want to fix preffer_cpu offline issue but The requirement that causes the problem is not merge in this version. Therefore, this patch is not required. Signed-off-by: He Yujie <coka.heyujie(a)huawei.com> --- kernel/sched/fair.c | 13 ++----------- 1 file changed, 2 insertions(+), 11 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 41b339c5b671..63f4344ac344 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7248,7 +7248,6 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, struct task_group *tg; long spare; int cpu, mode; - int nr_cpus_valid = 0; rcu_read_lock(); mode = dynamic_affinity_mode(p); @@ -7266,7 +7265,7 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, /* manual mode */ tg = task_group(p); - for_each_cpu_and(cpu, p->prefer_cpus, cpu_online_mask) { + for_each_cpu(cpu, p->prefer_cpus) { if (unlikely(!tg->se[cpu])) continue; @@ -7290,18 +7289,10 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, util_avg_sum += tg->se[cpu]->avg.util_avg; tg_capacity += capacity_of(cpu); - nr_cpus_valid++; } rcu_read_unlock(); - /* - * Follow cases should select cpus_ptr, checking by condition of - * tg_capacity > nr_cpus_valid: - * 1. all prefer_cpus offline; - * 2. all prefer_cpus has no cfs capaicity(tg_capacity = - * nr_cpus_valid * 1) - */ - if (tg_capacity > nr_cpus_valid && + if (tg_capacity > cpumask_weight(p->prefer_cpus) && util_avg_sum * 100 <= tg_capacity * sysctl_sched_util_low_pct) { p->select_cpus = p->prefer_cpus; if (sd_flag & SD_BALANCE_WAKE) -- 2.34.1

2 1

[OLK-6.6 PATCH 0/2] Some bug fix patches for OLK-6.6 perf
by Yushan Wang 14 Nov '24

14 Nov '24

From: zhangqz <14294317+zhangqizhi3(a)user.noreply.gitee.com> Some bug fix patches for OLK-5.10 perf John Garry (2): perf metricgroup: Fix for metrics containing duration_time perf metricgroup: Fix system PMU metrics tools/perf/util/metricgroup.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) -- 2.33.0

1 5

[openeuler:openEuler-1.0-LTS 1291/1291] drivers/char/svm.c:312:17-38: WARNING: atomic_dec_and_test variation before object free at line 335.
by kernel test robot 14 Nov '24

14 Nov '24

tree: https://gitee.com/openeuler/kernel.git openEuler-1.0-LTS head: d5c3ef07e625af7ba8aded2f507f092bceab3cbc commit: 4cdb60f56180d6ef7c10bd0e50ddda96ebc9f783 [1291/1291] svm: implement unpin pages to dec refcount config: arm64-randconfig-r063-20241113 (https://download.01.org/0day-ci/archive/20241114/202411142057.pcqq5uGT-lkp@…) compiler: aarch64-linux-gcc (GCC) 14.2.0 If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp(a)intel.com> | Closes: https://lore.kernel.org/oe-kbuild-all/202411142057.pcqq5uGT-lkp@intel.com/ cocci warnings: (new ones prefixed by >>) >> drivers/char/svm.c:312:17-38: WARNING: atomic_dec_and_test variation before object free at line 335. vim +312 drivers/char/svm.c 304 305 static void svm_remove_sdma(struct svm_process *process, 306 struct svm_sdma *sdma, bool try_rm) 307 { 308 int null_count = 0; 309 310 mutex_lock(&process->mutex); 311 > 312 if (try_rm && (!atomic64_dec_and_test(&sdma->ref))) { 313 mutex_unlock(&process->mutex); 314 return; 315 } 316 317 rb_erase(&sdma->node, &process->sdma_list); 318 RB_CLEAR_NODE(&sdma->node); 319 mutex_unlock(&process->mutex); 320 321 while (sdma->nr_pages--) { 322 if (sdma->pages[sdma->nr_pages] == NULL) { 323 pr_err("null pointer, nr_pages:%d.\n", sdma->nr_pages); 324 null_count++; 325 continue; 326 } 327 328 put_page(sdma->pages[sdma->nr_pages]); 329 } 330 331 if (null_count) 332 dump_stack(); 333 334 kfree(sdma->pages); > 335 kfree(sdma); 336 } 337 -- 0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki

1 0

[PATCH openEuler-22.03-LTS-SP1] scsi: libsas: Add smp_ata_check_ready_type()
by Li Nan 14 Nov '24

14 Nov '24

From: Jie Zhan <zhanjie9(a)hisilicon.com> mainline inclusion from mainline-v6.2-rc1 commit 9181ce3cb5d9 category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/I6V6I1 Reference: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?… ---------------------------------------------------------------------- Create function smp_ata_check_ready_type() for LLDDs to wait for SATA devices to come up after a link reset. Signed-off-by: Jie Zhan <zhanjie9(a)hisilicon.com> Link: https://lore.kernel.org/r/20221118083714.4034612-4-zhanjie9@hisilicon.com Signed-off-by: Martin K. Petersen <martin.petersen(a)oracle.com> Signed-off-by: xiabing <xiabing12(a)h-partners.com> Signed-off-by: Li Nan <linan122(a)huawei.com> --- drivers/scsi/libsas/sas_internal.h | 2 ++ include/scsi/sas_ata.h | 6 ++++++ drivers/scsi/libsas/sas_ata.c | 25 +++++++++++++++++++++++++ drivers/scsi/libsas/sas_expander.c | 4 ++-- 4 files changed, 35 insertions(+), 2 deletions(-) diff --git a/drivers/scsi/libsas/sas_internal.h b/drivers/scsi/libsas/sas_internal.h index acd515c01861..18749b2a6502 100644 --- a/drivers/scsi/libsas/sas_internal.h +++ b/drivers/scsi/libsas/sas_internal.h @@ -86,6 +86,8 @@ struct domain_device *sas_ex_to_ata(struct domain_device *ex_dev, int phy_id); int sas_ex_phy_discover(struct domain_device *dev, int single); int sas_get_report_phy_sata(struct domain_device *dev, int phy_id, struct smp_resp *rps_resp); +int sas_get_phy_attached_dev(struct domain_device *dev, int phy_id, + u8 *sas_addr, enum sas_device_type *type); int sas_try_ata_reset(struct asd_sas_phy *phy); void sas_hae_reset(struct work_struct *work); diff --git a/include/scsi/sas_ata.h b/include/scsi/sas_ata.h index e0d2d4915257..74e8feea85b4 100644 --- a/include/scsi/sas_ata.h +++ b/include/scsi/sas_ata.h @@ -34,6 +34,7 @@ void sas_suspend_sata(struct asd_sas_port *port); void sas_resume_sata(struct asd_sas_port *port); void sas_ata_end_eh(struct ata_port *ap); int sas_ata_wait_after_reset(struct domain_device *dev, unsigned long deadline); +int smp_ata_check_ready_type(struct ata_link *link); #else @@ -92,6 +93,11 @@ static inline int sas_ata_wait_after_reset(struct domain_device *dev, { return -ETIMEDOUT; } + +static inline int smp_ata_check_ready_type(struct ata_link *link) +{ + return 0; +} #endif #endif /* _SAS_ATA_H_ */ diff --git a/drivers/scsi/libsas/sas_ata.c b/drivers/scsi/libsas/sas_ata.c index d54f91636ad1..ff09e0ff00d2 100644 --- a/drivers/scsi/libsas/sas_ata.c +++ b/drivers/scsi/libsas/sas_ata.c @@ -301,6 +301,31 @@ static int sas_ata_clear_pending(struct domain_device *dev, struct ex_phy *phy) return 1; } +int smp_ata_check_ready_type(struct ata_link *link) +{ + struct domain_device *dev = link->ap->private_data; + struct sas_phy *phy = sas_get_local_phy(dev); + struct domain_device *ex_dev = dev->parent; + enum sas_device_type type = SAS_PHY_UNUSED; + u8 sas_addr[SAS_ADDR_SIZE]; + int res; + + res = sas_get_phy_attached_dev(ex_dev, phy->number, sas_addr, &type); + sas_put_local_phy(phy); + if (res) + return res; + + switch (type) { + case SAS_SATA_PENDING: + return 0; + case SAS_END_DEVICE: + return 1; + default: + return -ENODEV; + } +} +EXPORT_SYMBOL_GPL(smp_ata_check_ready_type); + static int smp_ata_check_ready(struct ata_link *link) { int res; diff --git a/drivers/scsi/libsas/sas_expander.c b/drivers/scsi/libsas/sas_expander.c index dd22c7a32059..2287f8e07968 100644 --- a/drivers/scsi/libsas/sas_expander.c +++ b/drivers/scsi/libsas/sas_expander.c @@ -1727,8 +1727,8 @@ static int sas_get_phy_change_count(struct domain_device *dev, return res; } -static int sas_get_phy_attached_dev(struct domain_device *dev, int phy_id, - u8 *sas_addr, enum sas_device_type *type) +int sas_get_phy_attached_dev(struct domain_device *dev, int phy_id, + u8 *sas_addr, enum sas_device_type *type) { int res; struct smp_disc_resp *disc_resp; -- 2.39.2

2 1

[openeuler:openEuler-1.0-LTS 1291/1291] drivers/char/svm.c:312:17-38: WARNING: atomic_dec_and_test variation before object free at line 335.
by kernel test robot 14 Nov '24

14 Nov '24

tree: https://gitee.com/openeuler/kernel.git openEuler-1.0-LTS head: d5c3ef07e625af7ba8aded2f507f092bceab3cbc commit: 4cdb60f56180d6ef7c10bd0e50ddda96ebc9f783 [1291/1291] svm: implement unpin pages to dec refcount config: arm64-randconfig-r063-20241113 (https://download.01.org/0day-ci/archive/20241114/202411141631.Gm0CdUvq-lkp@…) compiler: aarch64-linux-gcc (GCC) 14.2.0 If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags | Reported-by: kernel test robot <lkp(a)intel.com> | Closes: https://lore.kernel.org/oe-kbuild-all/202411141631.Gm0CdUvq-lkp@intel.com/ cocci warnings: (new ones prefixed by >>) >> drivers/char/svm.c:312:17-38: WARNING: atomic_dec_and_test variation before object free at line 335. vim +312 drivers/char/svm.c 304 305 static void svm_remove_sdma(struct svm_process *process, 306 struct svm_sdma *sdma, bool try_rm) 307 { 308 int null_count = 0; 309 310 mutex_lock(&process->mutex); 311 > 312 if (try_rm && (!atomic64_dec_and_test(&sdma->ref))) { 313 mutex_unlock(&process->mutex); 314 return; 315 } 316 317 rb_erase(&sdma->node, &process->sdma_list); 318 RB_CLEAR_NODE(&sdma->node); 319 mutex_unlock(&process->mutex); 320 321 while (sdma->nr_pages--) { 322 if (sdma->pages[sdma->nr_pages] == NULL) { 323 pr_err("null pointer, nr_pages:%d.\n", sdma->nr_pages); 324 null_count++; 325 continue; 326 } 327 328 put_page(sdma->pages[sdma->nr_pages]); 329 } 330 331 if (null_count) 332 dump_stack(); 333 334 kfree(sdma->pages); > 335 kfree(sdma); 336 } 337 -- 0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki

1 0

[PATCH openEuler-1.0-LTS] sched/dynamic_affinity: fix preffered_cpu offline problem
by He Yujie 14 Nov '24

14 Nov '24

hulk inclusion category: bugfix bugzilla: https://gitee.com/openeuler/kernel/issues/IB42TC CVE: NA -------------------------------- After the preferred_cpu goes offline, the core selection of dynamic affinity does not check whether the preferred_cpu is valid. As a result, related dynamic affinity processes are concentrated on a shared_cpu. This patch resolves this problem to checks whether the preferred_cpu is valid and compares only the usage threshold of the valid preferred_cpu. Fixes: 70a232a564cf ("sched: Adjust wakeup cpu range according CPU util dynamicly") Signed-off-by: He Yujie <coka.heyujie(a)huawei.com> --- kernel/sched/fair.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index 63f4344ac344..41b339c5b671 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -7248,6 +7248,7 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, struct task_group *tg; long spare; int cpu, mode; + int nr_cpus_valid = 0; rcu_read_lock(); mode = dynamic_affinity_mode(p); @@ -7265,7 +7266,7 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, /* manual mode */ tg = task_group(p); - for_each_cpu(cpu, p->prefer_cpus) { + for_each_cpu_and(cpu, p->prefer_cpus, cpu_online_mask) { if (unlikely(!tg->se[cpu])) continue; @@ -7289,10 +7290,18 @@ static void set_task_select_cpus(struct task_struct *p, int *idlest_cpu, util_avg_sum += tg->se[cpu]->avg.util_avg; tg_capacity += capacity_of(cpu); + nr_cpus_valid++; } rcu_read_unlock(); - if (tg_capacity > cpumask_weight(p->prefer_cpus) && + /* + * Follow cases should select cpus_ptr, checking by condition of + * tg_capacity > nr_cpus_valid: + * 1. all prefer_cpus offline; + * 2. all prefer_cpus has no cfs capaicity(tg_capacity = + * nr_cpus_valid * 1) + */ + if (tg_capacity > nr_cpus_valid && util_avg_sum * 100 <= tg_capacity * sysctl_sched_util_low_pct) { p->select_cpus = p->prefer_cpus; if (sd_flag & SD_BALANCE_WAKE) -- 2.34.1

2 1