* Re: [PATCH] rcu/cpu_stall_cputime: fix the hardirq count for x86 architecture
2025-01-08 6:57 [PATCH] rcu/cpu_stall_cputime: fix the hardirq count for x86 architecture Yongliang Gao
@ 2025-01-09 0:51 ` kernel test robot
2025-01-09 2:32 ` kernel test robot
1 sibling, 0 replies; 3+ messages in thread
From: kernel test robot @ 2025-01-09 0:51 UTC (permalink / raw)
To: Yongliang Gao, paulmck, frederic, thunder.leizhen
Cc: oe-kbuild-all, frankjpliu, rcu, linux-kernel, Yongliang Gao
Hi Yongliang,
kernel test robot noticed the following build errors:
[auto build test ERROR on paulmck-rcu/dev]
[also build test ERROR on linus/master v6.13-rc6 next-20250108]
[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#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Yongliang-Gao/rcu-cpu_stall_cputime-fix-the-hardirq-count-for-x86-architecture/20250108-145810
base: https://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git dev
patch link: https://lore.kernel.org/r/20250108065716.2888148-1-leonylgao%40gmail.com
patch subject: [PATCH] rcu/cpu_stall_cputime: fix the hardirq count for x86 architecture
config: arm64-allnoconfig (https://download.01.org/0day-ci/archive/20250109/202501090842.SfI6QPGS-lkp@intel.com/config)
compiler: aarch64-linux-gcc (GCC) 14.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250109/202501090842.SfI6QPGS-lkp@intel.com/reproduce)
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@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202501090842.SfI6QPGS-lkp@intel.com/
All errors (new ones prefixed by >>):
kernel/rcu/tree.c: In function 'rcu_watching_snap_recheck':
>> kernel/rcu/tree.c:960:71: error: implicit declaration of function 'arch_irq_stat_cpu' [-Wimplicit-function-declaration]
960 | rsrp->nr_hardirqs = kstat_cpu_irqs_sum(cpu) + arch_irq_stat_cpu(cpu);
| ^~~~~~~~~~~~~~~~~
vim +/arch_irq_stat_cpu +960 kernel/rcu/tree.c
824
825 /*
826 * Returns positive if the specified CPU has passed through a quiescent state
827 * by virtue of being in or having passed through an dynticks idle state since
828 * the last call to rcu_watching_snap_save() for this same CPU, or by
829 * virtue of having been offline.
830 *
831 * Returns negative if the specified CPU needs a force resched.
832 *
833 * Returns zero otherwise.
834 */
835 static int rcu_watching_snap_recheck(struct rcu_data *rdp)
836 {
837 unsigned long jtsq;
838 int ret = 0;
839 struct rcu_node *rnp = rdp->mynode;
840
841 /*
842 * If the CPU passed through or entered a dynticks idle phase with
843 * no active irq/NMI handlers, then we can safely pretend that the CPU
844 * already acknowledged the request to pass through a quiescent
845 * state. Either way, that CPU cannot possibly be in an RCU
846 * read-side critical section that started before the beginning
847 * of the current RCU grace period.
848 */
849 if (rcu_watching_snap_stopped_since(rdp, rdp->watching_snap)) {
850 trace_rcu_fqs(rcu_state.name, rdp->gp_seq, rdp->cpu, TPS("dti"));
851 rcu_gpnum_ovf(rnp, rdp);
852 return 1;
853 }
854
855 /*
856 * Complain if a CPU that is considered to be offline from RCU's
857 * perspective has not yet reported a quiescent state. After all,
858 * the offline CPU should have reported a quiescent state during
859 * the CPU-offline process, or, failing that, by rcu_gp_init()
860 * if it ran concurrently with either the CPU going offline or the
861 * last task on a leaf rcu_node structure exiting its RCU read-side
862 * critical section while all CPUs corresponding to that structure
863 * are offline. This added warning detects bugs in any of these
864 * code paths.
865 *
866 * The rcu_node structure's ->lock is held here, which excludes
867 * the relevant portions the CPU-hotplug code, the grace-period
868 * initialization code, and the rcu_read_unlock() code paths.
869 *
870 * For more detail, please refer to the "Hotplug CPU" section
871 * of RCU's Requirements documentation.
872 */
873 if (WARN_ON_ONCE(!rcu_rdp_cpu_online(rdp))) {
874 struct rcu_node *rnp1;
875
876 pr_info("%s: grp: %d-%d level: %d ->gp_seq %ld ->completedqs %ld\n",
877 __func__, rnp->grplo, rnp->grphi, rnp->level,
878 (long)rnp->gp_seq, (long)rnp->completedqs);
879 for (rnp1 = rnp; rnp1; rnp1 = rnp1->parent)
880 pr_info("%s: %d:%d ->qsmask %#lx ->qsmaskinit %#lx ->qsmaskinitnext %#lx ->rcu_gp_init_mask %#lx\n",
881 __func__, rnp1->grplo, rnp1->grphi, rnp1->qsmask, rnp1->qsmaskinit, rnp1->qsmaskinitnext, rnp1->rcu_gp_init_mask);
882 pr_info("%s %d: %c online: %ld(%d) offline: %ld(%d)\n",
883 __func__, rdp->cpu, ".o"[rcu_rdp_cpu_online(rdp)],
884 (long)rdp->rcu_onl_gp_seq, rdp->rcu_onl_gp_state,
885 (long)rdp->rcu_ofl_gp_seq, rdp->rcu_ofl_gp_state);
886 return 1; /* Break things loose after complaining. */
887 }
888
889 /*
890 * A CPU running for an extended time within the kernel can
891 * delay RCU grace periods: (1) At age jiffies_to_sched_qs,
892 * set .rcu_urgent_qs, (2) At age 2*jiffies_to_sched_qs, set
893 * both .rcu_need_heavy_qs and .rcu_urgent_qs. Note that the
894 * unsynchronized assignments to the per-CPU rcu_need_heavy_qs
895 * variable are safe because the assignments are repeated if this
896 * CPU failed to pass through a quiescent state. This code
897 * also checks .jiffies_resched in case jiffies_to_sched_qs
898 * is set way high.
899 */
900 jtsq = READ_ONCE(jiffies_to_sched_qs);
901 if (!READ_ONCE(rdp->rcu_need_heavy_qs) &&
902 (time_after(jiffies, rcu_state.gp_start + jtsq * 2) ||
903 time_after(jiffies, rcu_state.jiffies_resched) ||
904 rcu_state.cbovld)) {
905 WRITE_ONCE(rdp->rcu_need_heavy_qs, true);
906 /* Store rcu_need_heavy_qs before rcu_urgent_qs. */
907 smp_store_release(&rdp->rcu_urgent_qs, true);
908 } else if (time_after(jiffies, rcu_state.gp_start + jtsq)) {
909 WRITE_ONCE(rdp->rcu_urgent_qs, true);
910 }
911
912 /*
913 * NO_HZ_FULL CPUs can run in-kernel without rcu_sched_clock_irq!
914 * The above code handles this, but only for straight cond_resched().
915 * And some in-kernel loops check need_resched() before calling
916 * cond_resched(), which defeats the above code for CPUs that are
917 * running in-kernel with scheduling-clock interrupts disabled.
918 * So hit them over the head with the resched_cpu() hammer!
919 */
920 if (tick_nohz_full_cpu(rdp->cpu) &&
921 (time_after(jiffies, READ_ONCE(rdp->last_fqs_resched) + jtsq * 3) ||
922 rcu_state.cbovld)) {
923 WRITE_ONCE(rdp->rcu_urgent_qs, true);
924 WRITE_ONCE(rdp->last_fqs_resched, jiffies);
925 ret = -1;
926 }
927
928 /*
929 * If more than halfway to RCU CPU stall-warning time, invoke
930 * resched_cpu() more frequently to try to loosen things up a bit.
931 * Also check to see if the CPU is getting hammered with interrupts,
932 * but only once per grace period, just to keep the IPIs down to
933 * a dull roar.
934 */
935 if (time_after(jiffies, rcu_state.jiffies_resched)) {
936 if (time_after(jiffies,
937 READ_ONCE(rdp->last_fqs_resched) + jtsq)) {
938 WRITE_ONCE(rdp->last_fqs_resched, jiffies);
939 ret = -1;
940 }
941 if (IS_ENABLED(CONFIG_IRQ_WORK) &&
942 !rdp->rcu_iw_pending && rdp->rcu_iw_gp_seq != rnp->gp_seq &&
943 (rnp->ffmask & rdp->grpmask)) {
944 rdp->rcu_iw_pending = true;
945 rdp->rcu_iw_gp_seq = rnp->gp_seq;
946 irq_work_queue_on(&rdp->rcu_iw, rdp->cpu);
947 }
948
949 if (rcu_cpu_stall_cputime && rdp->snap_record.gp_seq != rdp->gp_seq) {
950 int cpu = rdp->cpu;
951 struct rcu_snap_record *rsrp;
952 struct kernel_cpustat *kcsp;
953
954 kcsp = &kcpustat_cpu(cpu);
955
956 rsrp = &rdp->snap_record;
957 rsrp->cputime_irq = kcpustat_field(kcsp, CPUTIME_IRQ, cpu);
958 rsrp->cputime_softirq = kcpustat_field(kcsp, CPUTIME_SOFTIRQ, cpu);
959 rsrp->cputime_system = kcpustat_field(kcsp, CPUTIME_SYSTEM, cpu);
> 960 rsrp->nr_hardirqs = kstat_cpu_irqs_sum(cpu) + arch_irq_stat_cpu(cpu);
961 rsrp->nr_softirqs = kstat_cpu_softirqs_sum(cpu);
962 rsrp->nr_csw = nr_context_switches_cpu(cpu);
963 rsrp->jiffies = jiffies;
964 rsrp->gp_seq = rdp->gp_seq;
965 }
966 }
967
968 return ret;
969 }
970
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] rcu/cpu_stall_cputime: fix the hardirq count for x86 architecture
2025-01-08 6:57 [PATCH] rcu/cpu_stall_cputime: fix the hardirq count for x86 architecture Yongliang Gao
2025-01-09 0:51 ` kernel test robot
@ 2025-01-09 2:32 ` kernel test robot
1 sibling, 0 replies; 3+ messages in thread
From: kernel test robot @ 2025-01-09 2:32 UTC (permalink / raw)
To: Yongliang Gao, paulmck, frederic, thunder.leizhen
Cc: llvm, oe-kbuild-all, frankjpliu, rcu, linux-kernel, Yongliang Gao
Hi Yongliang,
kernel test robot noticed the following build errors:
[auto build test ERROR on paulmck-rcu/dev]
[also build test ERROR on linus/master v6.13-rc6 next-20250108]
[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#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Yongliang-Gao/rcu-cpu_stall_cputime-fix-the-hardirq-count-for-x86-architecture/20250108-145810
base: https://git.kernel.org/pub/scm/linux/kernel/git/paulmck/linux-rcu.git dev
patch link: https://lore.kernel.org/r/20250108065716.2888148-1-leonylgao%40gmail.com
patch subject: [PATCH] rcu/cpu_stall_cputime: fix the hardirq count for x86 architecture
config: riscv-randconfig-002-20250109 (https://download.01.org/0day-ci/archive/20250109/202501091016.QtqNWWih-lkp@intel.com/config)
compiler: clang version 15.0.7 (https://github.com/llvm/llvm-project 8dfdcc7b7bf66834a761bd8de445840ef68e4d1a)
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250109/202501091016.QtqNWWih-lkp@intel.com/reproduce)
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@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202501091016.QtqNWWih-lkp@intel.com/
All errors (new ones prefixed by >>):
>> kernel/rcu/tree.c:960:50: error: call to undeclared function 'arch_irq_stat_cpu'; ISO C99 and later do not support implicit function declarations [-Werror,-Wimplicit-function-declaration]
rsrp->nr_hardirqs = kstat_cpu_irqs_sum(cpu) + arch_irq_stat_cpu(cpu);
^
In file included from kernel/rcu/tree.c:5743:
>> kernel/rcu/tree_stall.h:439:29: error: call to undeclared function 'arch_irq_stat_cpu'; ISO C99 and later do not support implicit function declarations [-Werror,-Wimplicit-function-declaration]
kstat_cpu_irqs_sum(cpu) + arch_irq_stat_cpu(cpu) - rsrp->nr_hardirqs,
^
2 errors generated.
vim +/arch_irq_stat_cpu +960 kernel/rcu/tree.c
824
825 /*
826 * Returns positive if the specified CPU has passed through a quiescent state
827 * by virtue of being in or having passed through an dynticks idle state since
828 * the last call to rcu_watching_snap_save() for this same CPU, or by
829 * virtue of having been offline.
830 *
831 * Returns negative if the specified CPU needs a force resched.
832 *
833 * Returns zero otherwise.
834 */
835 static int rcu_watching_snap_recheck(struct rcu_data *rdp)
836 {
837 unsigned long jtsq;
838 int ret = 0;
839 struct rcu_node *rnp = rdp->mynode;
840
841 /*
842 * If the CPU passed through or entered a dynticks idle phase with
843 * no active irq/NMI handlers, then we can safely pretend that the CPU
844 * already acknowledged the request to pass through a quiescent
845 * state. Either way, that CPU cannot possibly be in an RCU
846 * read-side critical section that started before the beginning
847 * of the current RCU grace period.
848 */
849 if (rcu_watching_snap_stopped_since(rdp, rdp->watching_snap)) {
850 trace_rcu_fqs(rcu_state.name, rdp->gp_seq, rdp->cpu, TPS("dti"));
851 rcu_gpnum_ovf(rnp, rdp);
852 return 1;
853 }
854
855 /*
856 * Complain if a CPU that is considered to be offline from RCU's
857 * perspective has not yet reported a quiescent state. After all,
858 * the offline CPU should have reported a quiescent state during
859 * the CPU-offline process, or, failing that, by rcu_gp_init()
860 * if it ran concurrently with either the CPU going offline or the
861 * last task on a leaf rcu_node structure exiting its RCU read-side
862 * critical section while all CPUs corresponding to that structure
863 * are offline. This added warning detects bugs in any of these
864 * code paths.
865 *
866 * The rcu_node structure's ->lock is held here, which excludes
867 * the relevant portions the CPU-hotplug code, the grace-period
868 * initialization code, and the rcu_read_unlock() code paths.
869 *
870 * For more detail, please refer to the "Hotplug CPU" section
871 * of RCU's Requirements documentation.
872 */
873 if (WARN_ON_ONCE(!rcu_rdp_cpu_online(rdp))) {
874 struct rcu_node *rnp1;
875
876 pr_info("%s: grp: %d-%d level: %d ->gp_seq %ld ->completedqs %ld\n",
877 __func__, rnp->grplo, rnp->grphi, rnp->level,
878 (long)rnp->gp_seq, (long)rnp->completedqs);
879 for (rnp1 = rnp; rnp1; rnp1 = rnp1->parent)
880 pr_info("%s: %d:%d ->qsmask %#lx ->qsmaskinit %#lx ->qsmaskinitnext %#lx ->rcu_gp_init_mask %#lx\n",
881 __func__, rnp1->grplo, rnp1->grphi, rnp1->qsmask, rnp1->qsmaskinit, rnp1->qsmaskinitnext, rnp1->rcu_gp_init_mask);
882 pr_info("%s %d: %c online: %ld(%d) offline: %ld(%d)\n",
883 __func__, rdp->cpu, ".o"[rcu_rdp_cpu_online(rdp)],
884 (long)rdp->rcu_onl_gp_seq, rdp->rcu_onl_gp_state,
885 (long)rdp->rcu_ofl_gp_seq, rdp->rcu_ofl_gp_state);
886 return 1; /* Break things loose after complaining. */
887 }
888
889 /*
890 * A CPU running for an extended time within the kernel can
891 * delay RCU grace periods: (1) At age jiffies_to_sched_qs,
892 * set .rcu_urgent_qs, (2) At age 2*jiffies_to_sched_qs, set
893 * both .rcu_need_heavy_qs and .rcu_urgent_qs. Note that the
894 * unsynchronized assignments to the per-CPU rcu_need_heavy_qs
895 * variable are safe because the assignments are repeated if this
896 * CPU failed to pass through a quiescent state. This code
897 * also checks .jiffies_resched in case jiffies_to_sched_qs
898 * is set way high.
899 */
900 jtsq = READ_ONCE(jiffies_to_sched_qs);
901 if (!READ_ONCE(rdp->rcu_need_heavy_qs) &&
902 (time_after(jiffies, rcu_state.gp_start + jtsq * 2) ||
903 time_after(jiffies, rcu_state.jiffies_resched) ||
904 rcu_state.cbovld)) {
905 WRITE_ONCE(rdp->rcu_need_heavy_qs, true);
906 /* Store rcu_need_heavy_qs before rcu_urgent_qs. */
907 smp_store_release(&rdp->rcu_urgent_qs, true);
908 } else if (time_after(jiffies, rcu_state.gp_start + jtsq)) {
909 WRITE_ONCE(rdp->rcu_urgent_qs, true);
910 }
911
912 /*
913 * NO_HZ_FULL CPUs can run in-kernel without rcu_sched_clock_irq!
914 * The above code handles this, but only for straight cond_resched().
915 * And some in-kernel loops check need_resched() before calling
916 * cond_resched(), which defeats the above code for CPUs that are
917 * running in-kernel with scheduling-clock interrupts disabled.
918 * So hit them over the head with the resched_cpu() hammer!
919 */
920 if (tick_nohz_full_cpu(rdp->cpu) &&
921 (time_after(jiffies, READ_ONCE(rdp->last_fqs_resched) + jtsq * 3) ||
922 rcu_state.cbovld)) {
923 WRITE_ONCE(rdp->rcu_urgent_qs, true);
924 WRITE_ONCE(rdp->last_fqs_resched, jiffies);
925 ret = -1;
926 }
927
928 /*
929 * If more than halfway to RCU CPU stall-warning time, invoke
930 * resched_cpu() more frequently to try to loosen things up a bit.
931 * Also check to see if the CPU is getting hammered with interrupts,
932 * but only once per grace period, just to keep the IPIs down to
933 * a dull roar.
934 */
935 if (time_after(jiffies, rcu_state.jiffies_resched)) {
936 if (time_after(jiffies,
937 READ_ONCE(rdp->last_fqs_resched) + jtsq)) {
938 WRITE_ONCE(rdp->last_fqs_resched, jiffies);
939 ret = -1;
940 }
941 if (IS_ENABLED(CONFIG_IRQ_WORK) &&
942 !rdp->rcu_iw_pending && rdp->rcu_iw_gp_seq != rnp->gp_seq &&
943 (rnp->ffmask & rdp->grpmask)) {
944 rdp->rcu_iw_pending = true;
945 rdp->rcu_iw_gp_seq = rnp->gp_seq;
946 irq_work_queue_on(&rdp->rcu_iw, rdp->cpu);
947 }
948
949 if (rcu_cpu_stall_cputime && rdp->snap_record.gp_seq != rdp->gp_seq) {
950 int cpu = rdp->cpu;
951 struct rcu_snap_record *rsrp;
952 struct kernel_cpustat *kcsp;
953
954 kcsp = &kcpustat_cpu(cpu);
955
956 rsrp = &rdp->snap_record;
957 rsrp->cputime_irq = kcpustat_field(kcsp, CPUTIME_IRQ, cpu);
958 rsrp->cputime_softirq = kcpustat_field(kcsp, CPUTIME_SOFTIRQ, cpu);
959 rsrp->cputime_system = kcpustat_field(kcsp, CPUTIME_SYSTEM, cpu);
> 960 rsrp->nr_hardirqs = kstat_cpu_irqs_sum(cpu) + arch_irq_stat_cpu(cpu);
961 rsrp->nr_softirqs = kstat_cpu_softirqs_sum(cpu);
962 rsrp->nr_csw = nr_context_switches_cpu(cpu);
963 rsrp->jiffies = jiffies;
964 rsrp->gp_seq = rdp->gp_seq;
965 }
966 }
967
968 return ret;
969 }
970
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
^ permalink raw reply [flat|nested] 3+ messages in thread