sched: smart_grid: fix potential NULL pointer dereference

hulk inclusion
category: bugfix
bugzilla: https://gitee.com/openeuler/kernel/issues/IA6J1H



----------------------------------------

There is a low probability that kernel panic will occur when we test
with smart_grid.

The log show below:

[65160.746953] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000008
[65160.746990] Unable to handle kernel NULL pointer dereference at virtual address 000000000000000
[65160.756974] Mem abort info:
[65160.766849] Mem abort info:
[65160.770660]   ESR = 0x96000004
[65160.770663]   EC = 0x25: DABT (current EL), IL = 32 bits
[65160.774374]   ESR = 0x96000004
[65160.774377]   EC = 0x25: DABT (current EL), IL = 32 bits
[65160.778428]   SET = 0, FnV = 0
[65160.778430]   EA = 0, S1PTW = 0
[65160.784728]   SET = 0, FnV = 0
[65160.784731]   EA = 0, S1PTW = 0
[65160.786018] Detected VIPT I-cache on CPU104
[65160.786070] GICv3: CPU104: found redistributor 3a0000 region 104: 0x00002000aa300000
[65160.786240] CPU104: Booted secondary processor 0x00003a0000 [0x481fd010]
[65160.788696] Data abort info:
[65160.788699]   ISV = 0, ISS = 0x00000004
[65160.794911] Data abort info:
[65160.794913]   ISV = 0, ISS = 0x00000004
[65160.798965]   CM = 0, WnR = 0
[65160.798967] user pgtable: 4k pages, 48-bit VAs, pgdp=00000030059b2000
[65160.803102]   CM = 0, WnR = 0
[65160.803104] user pgtable: 4k pages, 48-bit VAs, pgdp=00000020ab07c000
[65160.807066] [0000000000000008] pgd=0000000000000000, p4d=0000000000000000
[65160.811113] [0000000000000008] pgd=0000000000000000, p4d=0000000000000000
[65160.816199] Internal error: Oops: 0000000096000004 [#1] SMP
[65160.832723] Modules linked in:
[65161.006841] CPU: 39 PID: 195931 Comm: (hrottler) Kdump: loaded Nottainted 5.10.0 #51
[65161.016660] Hardware name: Huawei TaiShan 200 (Model 5280)/BC82AMDD, BIOS 1.08 12/14/2019
[65161.026146] pstate: 80400089 (Nzcv daIf +PAN -UAO -TCO BTYPE=--)
[65161.033199] pc : set_task_select_cpus+0x8c/0x3d0
[65161.038865] lr : select_task_rq_fair+0x1c8/0x5cc
[65161.044528] sp : ffff800172673ba0
[65161.048888] x29: ffff800172673ba0 x28: ffff00303f22cd00
[65161.055237] x27: 0000000000000027 x26: 0000000000000027
[65161.061583] x25: 0000aaaafd32b180 x24: 0000000000000002
[65161.067925] x23: ffff00303f22cd00 x22: 0000000000000000
[65161.074264] x21: ffff800172673cc4 x20: ffff00303f22cd00
[65161.080603] x19: ffff00303f22d7f4 x18: 0000000000000000
[65161.086938] x17: 0000000000000000 x16: 0000000000000000
[65161.093266] x15: 0000aaaafd365130 x14: 0000000000000000
[65161.099584] x13: 0000000000000000 x12: 0000000000000000
[65161.105892] x11: 0000000000000000 x10: 0000000000000000
[65161.112194] x9 : ffff800010129e80 x8 : 0000000000000000
[65161.118489] x7 : ffff00303f22cd00 x6 : 0000000000000001
[65161.124776] x5 : 0000000000000000 x4 : ffff8000118f5008
[65161.131058] x3 : 0000000000000000 x2 : 0000000000000002
[65161.137331] x1 : ffff800172673cc4 x0 : 0000000000000000
[65161.143598] Call trace:
[65161.147005]  set_task_select_cpus+0x8c/0x3d0
[65161.152225]  select_task_rq_fair+0x1c8/0x5cc
[65161.157439]  sched_exec+0x94/0x1bc
[65161.161782]  bprm_execve.part.0+0x60/0x164
[65161.166813]  bprm_execve+0x78/0xc0
[65161.171143]  do_execveat_common+0x1c4/0x250
[65161.176244]  __arm64_sys_execve+0x48/0x70
[65161.181167]  invoke_syscall+0x50/0x130
[65161.185824]  el0_svc_common.constprop.0+0x158/0x180
[65161.191601]  do_el0_svc+0x34/0xe0
[65161.195816]  el0_svc+0x20/0x30
[65161.199773]  el0_sync_handler+0xb8/0xc0
[65161.204502]  el0_sync+0x1e8/0x200
[65161.208712] Code: d50323bf d65f03c0 f941ac00 f941c400 (f9400400)

It's panic on task_group(current)->auto_affinity dereference.

In the scenario like:

	CPU0			CPU1
	rmdir cgroup
	free auto_affinity
				try to wake up
				select_task_rq_fair
				auto_affinity(NULL) dereference
				panic

Because there is no protect when remove task from one cgroup with
task wakeup, so we need to check auto_affinity is NULL in
task_prefer_cpus.

Fixes: 21e5d85e ("sched: Fix possible deadlock in tg_set_dynamic_affinity_mode")
Signed-off-by: Yipeng Zou <zouyipeng@huawei.com>