Hi Paul,
On a multicore interAptiv system, I noticed that booting with "nosmp"
generated a warning:
[ 0.116000] ------------[ cut here ]------------
[ 0.116000] WARNING: CPU: 0 PID: 1 at
/mnt/host/source/src/third_party/kernel/v3.14/arch/mips/kernel/pm-cps.c:270
cps_pm_init+0x798/0xd68()
[ 0.116000] pm-cps: FSB flush unsupported for this CPU
[ 0.116000] Modules linked in:
[ 0.116000] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 3.14.0 #23
[ 0.116000] Stack : 00000000 00000034 00000006 8048afdc 00000000
00000005 00000006 00000000
[ 0.116000] 00000000 00000000 00000000 00000000 80c3054a 00000034
00000000 80ba0000
[ 0.116000] 80b00000 8f8683c8 80a5ec48 00000001 00000000 80ba3604
80afea00 80b9b980
[ 0.116000] 00000001 809b4c30 80ba3604 80432538 00000000 80b00000
80a64efc 8f863d0c
[ 0.116000] 8f863d0c 0085b507 00000000 00000000 00000000 00000000
00000000 00000000
[ 0.116000] ...
[ 0.116000] Call Trace:
[ 0.116000] [<80409e44>] show_stack+0x6c/0xac
[ 0.120000] [<809b54c4>] dump_stack+0xb0/0xd4
[ 0.120000] [<804326c8>] warn_slowpath_common+0x98/0xc8
[ 0.120000] [<80432744>] warn_slowpath_fmt+0x4c/0x70
[ 0.120000] [<80b4193c>] cps_pm_init+0x798/0xd68
[ 0.120000] [<80400540>] do_one_initcall+0x7c/0x1cc
[ 0.120000] [<80b3bbcc>] kernel_init_freeable+0x168/0x220
[ 0.120000] [<809b3a04>] kernel_init+0x24/0x114
[ 0.120000] [<804034f8>] ret_from_kernel_thread+0x14/0x1c
[ 0.120000]
[ 0.120000] ---[ end trace 69ce861cf7aa6d6e ]---
[ 0.120000] Failed to generate core 1 state 1 entry
This happened because cps_gen_core_entries() was being called for CPUs
that have not been booted:
for_each_present_cpu(cpu) {
err = cps_gen_core_entries(cpu);
if (err)
return err;
}
and so fields like cpu_info->cputype and cpu_info->dcache.linesz were still 0.
I worked around it temporarily by using for_each_online_cpu instead,
but I'm not sure this will do the right thing if the CPUs are brought
up later.
Another possibility is to copy over the CPU type / cache info from
CPU0 if the secondary CPUs are not up yet.
Your thoughts?
