Unify all our cpu hotplug notifiers to mask out the CPU_TASKS_FROZEN
bit, so we don't have to add all the *_FROZEN variant cases to the
notifiers.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The current virtual timer interface is inherently per-cpu and hard to
use. The sole user of the interface is appldata which uses it to execute
a function after a specific amount of cputime has been used over all cpus.
Rework the virtual timer interface to hook into the cputime accounting.
This makes the interface independent from the CPU timer interrupts, and
makes the virtual timers global as opposed to per-cpu.
Overall the code is greatly simplified. The downside is that the accuracy
is not as good as the original implementation, but it is still good enough
for appldata.
Reviewed-by: Jan Glauber <jang@linux.vnet.ibm.com>
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The external interrupt handlers have a parameter called ext_int_code.
Besides the name this paramter does not only contain the ext_int_code
but in addition also the "cpu address" (POP) which caused the external
interrupt.
To make the code a bit more obvious pass a struct instead so the called
function can easily distinguish between external interrupt code and
cpu address. The cpu address field however is named "subcode" since
some external interrupt sources do not pass a cpu address but a
different parameter (or none at all).
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Whenever the cpu loads an enabled wait PSW it will appear as idle to the
underlying host system. The code in default_idle calls vtime_stop_cpu
which does the necessary voodoo to get the cpu time accounting right.
The udelay code just loads an enabled wait PSW. To correct this rework
the vtime_stop_cpu/vtime_start_cpu logic and move the difficult parts
to entry[64].S, vtime_stop_cpu can now be called from anywhere and
vtime_start_cpu is gone. The correction of the cpu time during wakeup
from an enabled wait PSW is done with a critical section in entry[64].S.
As vtime_start_cpu is gone, s390_idle_check can be removed as well.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Define struct pcpu and merge some of the NR_CPUS arrays into it, including
__cpu_logical_map, current_set and smp_cpu_state. Split smp related
functions to those operating on physical cpus and the functions operating
on a logical cpu number. Make the functions for physical cpus use a
pointer to a struct pcpu. This hides the knowledge about cpu addresses in
smp.c, entry[64].S and swsusp_asm64.S, thus remove the sigp.h header.
The PSW restart mechanism is used to start secondary cpus, calling a
function on an online cpu, calling a function on the ipl cpu, and for
the nmi signal. Replace the different assembler functions with a
single function restart_int_handler. The new entry point calls a function
whose pointer is stored in the lowcore of the target cpu and it can wait
for the source cpu to stop. This covers all existing use cases.
Overall the code is now simpler and there are ~380 lines less code.
Reviewed-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Split out addressing mode bits from PSW_BASE_BITS, rename PSW_BASE_BITS
to PSW_MASK_BASE, get rid of psw_user32_bits, remove unused function
enabled_wait(), introduce PSW_MASK_USER, and drop PSW_MASK_MERGE macros.
Change psw_kernel_bits / psw_user_bits to contain only the bits that
are always set in the respective mode.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Merge irq.c and s390_ext.c into irq.c. That way all external interrupt
related functions are together.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Up to now /proc/interrupts only has statistics for external and i/o
interrupts but doesn't split up them any further.
This patch adds a line for every single interrupt source so that it
is possible to easier tell what the machine is/was doing.
Part of the output now looks like this;
CPU0 CPU2 CPU4
EXT: 3898 4232 2305
I/O: 782 315 245
CLK: 1029 1964 727 [EXT] Clock Comparator
IPI: 2868 2267 1577 [EXT] Signal Processor
TMR: 0 0 0 [EXT] CPU Timer
TAL: 0 0 0 [EXT] Timing Alert
PFL: 0 0 0 [EXT] Pseudo Page Fault
[...]
NMI: 0 1 1 [NMI] Machine Checks
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Add kprobes annotations to get the massive 'probe kernel.function("*") {}'
stress test working.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This fixes the same problem as described in the patch "nohz: fix
printk_needs_cpu() return value on offline cpus" for the arch_needs_cpu()
primitive:
arch_needs_cpu() may return 1 if called on offline cpus. When a cpu gets
offlined it schedules the idle process which, before killing its own cpu,
will call tick_nohz_stop_sched_tick().
That function in turn will call arch_needs_cpu() in order to check if the
local tick can be disabled. On offline cpus this function should naturally
return 0 since regardless if the tick gets disabled or not the cpu will be
dead short after. That is besides the fact that __cpu_disable() should already
have made sure that no interrupts on the offlined cpu will be delivered anyway.
In this case it prevents tick_nohz_stop_sched_tick() to call
select_nohz_load_balancer(). No idea if that really is a problem. However what
made me debug this is that on 2.6.32 the function get_nohz_load_balancer() is
used within __mod_timer() to select a cpu on which a timer gets enqueued.
If arch_needs_cpu() returns 1 then the nohz_load_balancer cpu doesn't get
updated when a cpu gets offlined. It may contain the cpu number of an offline
cpu. In turn timers get enqueued on an offline cpu and not very surprisingly
they never expire and cause system hangs.
This has been observed 2.6.32 kernels. On current kernels __mod_timer() uses
get_nohz_timer_target() which doesn't have that problem. However there might
be other problems because of the too early exit tick_nohz_stop_sched_tick()
in case a cpu goes offline.
This specific bug was indrocuded with 3c5d92a0 "nohz: Introduce
arch_needs_cpu".
In this case a cpu hotplug notifier is used to fix the issue in order to keep
the normal/fast path small. All we need to do is to clear the condition that
makes arch_needs_cpu() return 1 since it is just a performance improvement
which is supposed to keep the local tick running for a short period if a cpu
goes idle. Nothing special needs to be done except for clearing the condition.
Cc: stable@kernel.org
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Read external interrupts parameters from the lowcore in the first
level interrupt handler in entry[64].S.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
A machine check can interrupt the i/o and external interrupt handler
anytime. If the machine check occurs while the interrupt handler is
waking up from idle vtime_start_cpu can get executed a second time
and the int_clock / async_enter_timer values in the lowcore get
clobbered. This can confuse the cpu time accounting.
To fix this problem two changes are needed. First the machine check
handler has to use its own copies of int_clock and async_enter_timer,
named mcck_clock and mcck_enter_timer. Second the nested execution
of vtime_start_cpu has to be prevented. This is done in s390_idle_check
by checking the wait bit in the program status word.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Allow the architecture to request a normal jiffy tick when the system
goes idle and tick_nohz_stop_sched_tick is called . On s390 the hook is
used to prevent the system going fully idle if there has been an
interrupt other than a clock comparator interrupt since the last wakeup.
On s390 the HiperSockets response time for 1 connection ping-pong goes
down from 42 to 34 microseconds. The CPU cost decreases by 27%.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
LKML-Reference: <20090929122533.402715150@de.ibm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Replace the spinlock used in the idle time accounting with a sequence
counter mechanism analog to seqlock.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The slab allocator is earlier available so convert the
bootmem allocations to slab/gfp allocations.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
All definition in cpu.h have to do with cputime accounting. Move
them to cputime.h and remove the header file.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The cpu idle field in the output of /proc/stat is too small for cpus
that have been idle for more than a tick. Add the architecture hook
arch_idle_time that allows to add the not accounted idle time of a
sleeping cpu without waking the cpu.
The s390 implementation of arch_idle_time uses the already existing
s390_idle_data per_cpu variable to find the sleep time of a neighboring
idle cpu.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Start the cpu time accounting very early to catch the cpu time spent
for the initial kernel setup. To make the output of /proc/uptime
match the sum of all cpu accounting values of the boot cpu reset
xtime and wall_to_monotonic to sane values based on the TOD clock.
The values set by timekeeping_init are off by up to a second.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The steal time is calculated by subtracting the time the virtual cpu
has been running on a physical cpu from the wall clock time. To make
that work all wall time needs to be added to the steal time field first
before the virtual cpu time is subtracted.
The time between the last clock update and the load of the enabled wait
psw needs to be added to the steal_time field as well to make the sum
over all cpu accounting numbers match the wall clock.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
In case mod_virt_timer is used to add a non pending timer the timer
is always added as a one-shot timer. If mod_virt_timer is used for
periodic timers they may therfore be degraded to one-shot timers.
Add mod_virt_timer_periodic to the interface to allow safe re-programming
of the interval value.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
On (initial) cpu hotplug the lowcore values for user_timer and
system_timer don't get initialized like they would get on each
process schedule.
On initial start of secondary cpus this leads to the situation
where per thread user/system_timer values are larger than the
corresponding contents of the lowcore. When later calculating
time spent in user/system context the result can be negative.
So for cpu hotplug we should manually initialize lowcore values.
Fixes this bug:
Kernel BUG at 000ec080 [verbose debug info unavailable]
fixpoint divide exception: 0009 [#1] PREEMPT SMP
Modules linked in:
CPU: 10 Not tainted 2.6.28 #4
Process sysctl (pid: 975, task: 3fa752e0, ksp: 3fbebca0)
Krnl PSW : 070c1000 800ec080 (show_stat+0x390/0x5fc)
R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:0 CC:1 PM:0
Krnl GPRS: 7fffffff fefc7ce5 3faec080 003879ae
00000001 01388000 7fffffff 01388000
00000000 00000000 0049ad50 3fbebcf8
01388000 002f51a8 800ec1fe 3fbebcf8
Krnl Code: 800ec076: 9001b188 stm %r0,%r1,392(%r11)
800ec07a: 9801b0c0 lm %r0,%r1,192(%r11)
800ec07e: 1d05 dr %r0,%r5
>800ec080: 9001b0c0 stm %r0,%r1,192(%r11)
800ec084: 5860b0c4 l %r6,196(%r11)
800ec088: 1806 lr %r0,%r6
800ec08a: 8c800001 srdl %r8,1
800ec08e: 1d87 dr %r8,%r7
Call Trace:
([<00000000000ec1ee>] show_stat+0x4fe/0x5fc)
[<00000000000c13e8>] seq_read+0xc4/0x3ac
[<00000000000e4796>] proc_reg_read+0x6e/0x9c
[<00000000000a6a44>] vfs_read+0x78/0x100
[<00000000000a6ba8>] sys_read+0x40/0x80
[<00000000000234a8>] sysc_do_restart+0x1a/0x1e
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Distinguish the cputime of the idle process where idle is actually using
cpu cycles from the cputime where idle is sleeping on an enabled wait psw.
The former is accounted as system time, the later as idle time.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Increase the precision of the idle time calculation that is exported
to user space via /sys/devices/system/cpu/cpu<x>/idle_time_us
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The unit of the cputime accouting values that are stored per process is
currently a microsecond. The CPU timer has a maximum granularity of
2**-12 microseconds. There is no benefit in storing the per process values
in the lesser precision and there is the disadvantage that the backend
has to do the rounding to microseconds. The better solution is to use
the maximum granularity of the CPU timer as cputime unit.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The cpu time spent by the idle process actually doing something is
currently accounted as idle time. This is plain wrong, the architectures
that support VIRT_CPU_ACCOUNTING=y can do better: distinguish between the
time spent doing nothing and the time spent by idle doing work. The first
is accounted with account_idle_time and the second with account_system_time.
The architectures that use the account_xxx_time interface directly and not
the account_xxx_ticks interface now need to do the check for the idle
process in their arch code. In particular to improve the system vs true
idle time accounting the arch code needs to measure the true idle time
instead of just testing for the idle process.
To improve the tick based accounting as well we would need an architecture
primitive that can tell us if the pt_regs of the interrupted context
points to the magic instruction that halts the cpu.
In addition idle time is no more added to the stime of the idle process.
This field now contains the system time of the idle process as it should
be. On systems without VIRT_CPU_ACCOUNTING this will always be zero as
every tick that occurs while idle is running will be accounted as idle
time.
This patch contains the necessary common code changes to be able to
distinguish idle system time and true idle time. The architectures with
support for VIRT_CPU_ACCOUNTING need some changes to exploit this.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The utimescaled / stimescaled fields in the task structure and the
global cpustat should be set on all architectures. On s390 the calls
to account_user_time_scaled and account_system_time_scaled never have
been added. In addition system time that is accounted as guest time
to the user time of a process is accounted to the scaled system time
instead of the scaled user time.
To fix the bugs and to prevent future forgetfulness this patch merges
account_system_time_scaled into account_system_time and
account_user_time_scaled into account_user_time.
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Chris Wright <chrisw@sous-sol.org>
Cc: Michael Neuling <mikey@neuling.org>
Acked-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
On s390 we always want to run with precise cputime accounting.
Remove the config options VIRT_TIMER and VIRT_CPU_ACCOUNTING.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The idle notifier chain consists of at most one element. So there's
no point in having a notifier chain. Remove it and directly call the
function.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch contains the port of Qumranet's kvm kernel module to IBM zSeries
(aka s390x, mainframe) architecture. It uses the mainframe's virtualization
instruction SIE to run virtual machines with up to 64 virtual CPUs each.
This port is only usable on 64bit host kernels, and can only run 64bit guest
kernels. However, running 31bit applications in guest userspace is possible.
The following source files are introduced by this patch
arch/s390/kvm/kvm-s390.c similar to arch/x86/kvm/x86.c, this implements all
arch callbacks for kvm. __vcpu_run calls back into
sie64a to enter the guest machine context
arch/s390/kvm/sie64a.S assembler function sie64a, which enters guest
context via SIE, and switches world before and after that
include/asm-s390/kvm_host.h contains all vital data structures needed to run
virtual machines on the mainframe
include/asm-s390/kvm.h defines kvm_regs and friends for user access to
guest register content
arch/s390/kvm/gaccess.h functions similar to uaccess to access guest memory
arch/s390/kvm/kvm-s390.h header file for kvm-s390 internals, extended by
later patches
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
Since powerpc started using CONFIG_GENERIC_CLOCKEVENTS, the
deterministic CPU accounting (CONFIG_VIRT_CPU_ACCOUNTING) has been
broken on powerpc, because we end up counting user time twice: once in
timer_interrupt() and once in update_process_times().
This fixes the problem by pulling the code in update_process_times
that updates utime and stime into a separate function called
account_process_tick. If CONFIG_VIRT_CPU_ACCOUNTING is not defined,
there is a version of account_process_tick in kernel/timer.c that
simply accounts a whole tick to either utime or stime as before. If
CONFIG_VIRT_CPU_ACCOUNTING is defined, then arch code gets to
implement account_process_tick.
This also lets us simplify the s390 code a bit; it means that the s390
timer interrupt can now call update_process_times even when
CONFIG_VIRT_CPU_ACCOUNTING is turned on, and can just implement a
suitable account_process_tick().
account_process_tick() now takes the task_struct * as an argument.
Tested both with and without CONFIG_VIRT_CPU_ACCOUNTING.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
smp_call_function_single now has the same semantics as s390's
smp_call_function_on. Therefore convert to the *single variant
and get rid of some architecture specific code.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
sched-cfs-v2.6.22-git-v18.patch introduces CPU_IDLE in sched.h.
This conflict with the already existing define in
include/asm-s390/processor.h
Just rename the s390 defines, since they will go away as soon as
we support CONFIG_NO_HZ instead of our own CONFIG_NO_IDLE_HZ.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch adds support for clock synchronization to an external time
reference (ETR). The external time reference sends an oscillator
signal and a synchronization signal every 2^20 microseconds to keep
the TOD clocks of all connected servers in sync. For availability
two ETR units can be connected to a machine. If the clock deviates
for more than the sync-check tolerance all cpus get a machine check
that indicates that the clock is out of sync. For the lovely details
how to get the clock back in sync see the code below.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove the last few places where a pointer to pt_regs gets passed.
Also make sure we call set_irq_regs() before irq_enter() and after
irq_exit(). This doesn't fix anything but makes sure s390 looks the
same like all other architectures.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
finish_arch_switch needs to update the user cpu time as well, not just the
system cpu time. Otherwise the partial user cpu time of a process that is
stored in the lowcore will be (mis-)accounted to the next process.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Remove timer_list.magic and associated debugging code.
I originally added this when a spinlock was added to timer_list - this meant
that an all-zeroes timer became illegal and init_timer() was required.
That spinlock isn't even there any more, although timer.base must now be
initialised.
I'll keep this debugging code in -mm.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Fix overflow in calculation of the new tod value in stop_hz_timer and fix
wrong virtual timer list idle time in case the virtual timer is already
expired in stop_cpu_timer.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Heiko Carstens