s390/gmap: voluntarily schedule during key setting
authorChristian Borntraeger <borntraeger@linux.ibm.com>
Mon, 30 May 2022 09:27:05 +0000 (11:27 +0200)
committerHeiko Carstens <hca@linux.ibm.com>
Wed, 1 Jun 2022 10:03:16 +0000 (12:03 +0200)
commit6d5946274df1fff539a7eece458a43be733d1db8
treee7d3215f840d3b1ffad53ba7ff186abe542ff12e
parent315945a43e91c81ebfab0ef011d1dcab419c6d9b
s390/gmap: voluntarily schedule during key setting

With large and many guest with storage keys it is possible to create
large latencies or stalls during initial key setting:

rcu: INFO: rcu_sched self-detected stall on CPU
rcu:   18-....: (2099 ticks this GP) idle=54e/1/0x4000000000000002 softirq=35598716/35598716 fqs=998
       (t=2100 jiffies g=155867385 q=20879)
Task dump for CPU 18:
CPU 1/KVM       R  running task        0 1030947 256019 0x06000004
Call Trace:
sched_show_task
rcu_dump_cpu_stacks
rcu_sched_clock_irq
update_process_times
tick_sched_handle
tick_sched_timer
__hrtimer_run_queues
hrtimer_interrupt
do_IRQ
ext_int_handler
ptep_zap_key

The mmap lock is held during the page walking but since this is a
semaphore scheduling is still possible. Same for the kvm srcu.
To minimize overhead do this on every segment table entry or large page.

Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Reviewed-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Link: https://lore.kernel.org/r/20220530092706.11637-2-borntraeger@linux.ibm.com
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
arch/s390/mm/gmap.c