diff --git a/linux-ramdump-parser-v2/ramparse.py b/linux-ramdump-parser-v2/ramparse.py
index 9223dc2ff548154e34e43427866277035089db88..fea60a19b63c8774e56eb4c4854fd80ff5aba8b9 100644
--- a/linux-ramdump-parser-v2/ramparse.py
+++ b/linux-ramdump-parser-v2/ramparse.py
@@ -27,7 +27,7 @@ from optparse import OptionParser
import parser_util
from ramdump import RamDump
from print_out import print_out_str, set_outfile, print_out_section, print_out_exception, flush_outfile
-
+from sched_info import verify_active_cpus
# Please update version when something is changed!'
VERSION = '2.0'
@@ -166,6 +166,7 @@ if __name__ == '__main__':
(options, args) = parser.parse_args()
if options.minidump:
default_list = []
+ default_list.append("Schedinfo")
default_list.append("Dmesg")
default_list.append("RTB")
default_list.append("DebugImage")
@@ -378,6 +379,9 @@ if __name__ == '__main__':
get_wdog_timing(dump)
print_out_str('---------- end watchdog time-----')
+ # Always verify Scheduler requirement for active_cpus
+ verify_active_cpus(dump)
+
# we called parser.add_option with dest=p.cls.__name__ above,
# so if the user passed that option then `options' will have a
# p.cls.__name__ attribute.
@@ -389,6 +393,7 @@ if __name__ == '__main__':
parsers_to_run = [p for p in parser_util.get_parsers()
if getattr(options, p.cls.__name__)
or (options.everything and not p.optional)]
+
for i,p in enumerate(parsers_to_run):
if i == 0:
sys.stderr.write("\n")
diff --git a/linux-ramdump-parser-v2/sched_info.py b/linux-ramdump-parser-v2/sched_info.py
new file mode 100644
index 0000000000000000000000000000000000000000..2c55daf10a217cc5c96ab0811a87e8d5e659e7ac
--- /dev/null
+++ b/linux-ramdump-parser-v2/sched_info.py
@@ -0,0 +1,103 @@
+# Copyright (c) 2019, The Linux Foundation. All rights reserved.
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License version 2 and
+# only version 2 as published by the Free Software Foundation.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+
+from parser_util import register_parser, RamParser
+from print_out import print_out_str
+
+DEFAULT_MIGRATION_NR=32
+DEFAULT_MIGRATION_COST=500000
+
+def verify_active_cpus(ramdump):
+ core_sib_off = ramdump.field_offset('struct cpu_topology', 'core_sibling')
+ cpu_topology_addr = ramdump.address_of('cpu_topology')
+ cpu_topology_size = ramdump.sizeof('struct cpu_topology')
+
+ if (ramdump.kernel_version >= (4, 19, 0)):
+ cluster_id_off = ramdump.field_offset('struct cpu_topology', 'package_id')
+ else:
+ cluster_id_off = ramdump.field_offset('struct cpu_topology', 'cluster_id')
+
+ if (ramdump.kernel_version >= (4, 9, 0)):
+ cpu_online_bits = ramdump.read_word('__cpu_online_mask')
+ else:
+ cpu_online_bits = ramdump.read_word('cpu_online_bits')
+ if (ramdump.kernel_version >= (4, 9, 0)):
+ cpu_isolated_bits = ramdump.read_word('__cpu_isolated_mask')
+ elif (ramdump.kernel_version >= (4, 4, 0)):
+ cpu_isolated_bits = ramdump.read_word('cpu_isolated_bits')
+ else:
+ cpu_isolated_bits = 0
+
+ nr_cpus = ramdump.get_num_cpus()
+
+ if (cluster_id_off is None):
+ print_out_str("\n Invalid cluster topology detected\n")
+
+ clusters = ramdump.read_int(cpu_topology_addr + cluster_id_off + ((nr_cpus - 1) * cpu_topology_size))
+ clusters += 1
+
+ # INFO: from 4.19 onwards, core_sibling mask contains only online cpus,
+ # find out cluster cpus dynamically.
+
+ cluster_nrcpus = [0] * (clusters)
+ for j in range(0, nr_cpus):
+ c_id = ramdump.read_int(cpu_topology_addr + (j * cpu_topology_size) + cluster_id_off)
+ cluster_nrcpus[c_id] += 1
+
+ next_cluster_cpu = 0
+ for i in range(0, clusters):
+ cluster_cpus = ramdump.read_word(cpu_topology_addr +
+ (next_cluster_cpu * cpu_topology_size) + core_sib_off)
+ cluster_online_cpus = cpu_online_bits & cluster_cpus
+ cluster_nr_oncpus = bin(cluster_online_cpus).count('1')
+ cluster_isolated_cpus = cpu_isolated_bits & cluster_cpus
+ cluster_nr_isocpus = bin(cluster_isolated_cpus).count('1')
+
+ #print_out_str("Cluster fist cpu {0} cpu_mask {1:b}".format(next_cluster_cpu , cluster_cpus))
+ next_cluster_cpu += cluster_nrcpus[i]
+
+ if (cluster_nrcpus[i] > 2):
+ min_req_cpus = 2
+ else:
+ min_req_cpus = 1
+
+ if (cluster_nr_oncpus < min_req_cpus) or (cluster_nr_isocpus > min_req_cpus):
+ print_out_str("\n************ WARNING **************\n")
+ print_out_str("\tMinimum active cpus are not available in the cluster {0} \n".format(i))
+ print_out_str("\tCluster cpus: {0:b} Online cpus: {1:b} Isolated cpus: {2:b}\n".format(
+ cluster_cpus, cluster_online_cpus, cluster_isolated_cpus))
+
+@register_parser('--sched-info', 'Verify scheduler\'s various parameter status')
+class Schedinfo(RamParser):
+ def parse(self):
+ # Active cpu check verified by default!
+ #verify_active_cpus(self.ramdump)
+
+ # verify nr_migrates
+ sched_nr_migrate = self.ramdump.read_u32('sysctl_sched_nr_migrate')
+ if (sched_nr_migrate != DEFAULT_MIGRATION_NR):
+ print_out_str("\n************ WARNING **************\n")
+ print_out_str("\t sysctl_sched_nr_migrate has changed!!\n")
+ print_out_str("\t If it is single digit, scheduler's load balancer has broken in the dump\n")
+
+ # verify migration cost
+ sched_migration_cost = self.ramdump.read_u32('sysctl_sched_migration_cost')
+ if (sched_migration_cost != DEFAULT_MIGRATION_COST):
+ print_out_str("\n************ WARNING **************\n")
+ print_out_str("\t sysctl_sched_migration_cost has changed!!\n")
+ print_out_str("\t\tDefault: 500000 and Value in dump:{0}\n".format(sched_migration_cost))
+
+ # verify CFS BANDWIDTH enabled
+ cfs_bandwidth_enabled = self.ramdump.read_u32('sysctl_sched_cfs_bandwidth_slice')
+ if cfs_bandwidth_enabled is not None:
+ print_out_str("\n************ INFORMATION **************\n")
+ print_out_str("\tCFS_BANDWIDTH is enabled in the dump!!\n")
+ print_out_str("\tBandwidth slice: {0}\n".format(cfs_bandwidth_enabled))