kernel.git/arch/s390/kernel/vtime.c, branch linux-4.10.y

s390/vtime: correct system time accounting

2016-12-20T14:22:56Z

There is a slight misaccounting of system time in vtime_account_user. This function is called once per HZ tick in interrupt context. The irq_enter function already accounted the system time up to the point of the irq_enter call. The system time from irq_enter until vtime_account_user/do_account_vtime is reached is irq time but it is accounted to the previous context. Just drop the hardirq offset from arch/s390/kernel/vtime.c. Reported-by: Frederic Weisbecker Signed-off-by: Martin Schwidefsky

Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux

2016-12-14T00:33:33Z

Pull s390 updates from Martin Schwidefsky: "The main bulk of the s390 patches for the 4.10 merge window: - Add support for the contiguous memory allocator. - The recovery for I/O errors in the dasd device driver is improved, the driver will now remove channel paths that are not working properly. - Additional fields are added to /proc/sysinfo, the extended partition name and the partition UUID. - New naming for PCI devices with system defined UIDs. - The last few remaining alloc_bootmem calls are converted to memblock. - The thread_info structure is stripped down and moved to the task_struct. The only field left in thread_info is the flags field. - Rework of the arch topology code to fix a fake numa issue. - Refactoring of the atomic primitives and add a new preempt_count implementation. - Clocksource steering for the STP sync check offsets. - The s390 specific headers are changed to make them usable with CLANG. - Bug fixes and cleanup" * 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux: (70 commits) s390/cpumf: Use configuration level indication for sampling data s390: provide memmove implementation s390: cleanup arch/s390/kernel Makefile s390: fix initrd corruptions with gcov/kcov instrumented kernels s390: exclude early C code from gcov profiling s390/dasd: channel path aware error recovery s390/dasd: extend dasd path handling s390: remove unused labels from entry.S s390/vmlogrdr: fix IUCV buffer allocation s390/crypto: unlock on error in prng_tdes_read() s390/sysinfo: show partition extended name and UUID if available s390/numa: pin all possible cpus to nodes early s390/numa: establish cpu to node mapping early s390/topology: use cpu_topology array instead of per cpu variable s390/smp: initialize cpu_present_mask in setup_arch s390/topology: always use s390 specific sched_domain_topology_level s390/smp: use smp_get_base_cpu() helper function s390/numa: always use logical cpu and core ids s390: Remove VLAIS in ptff() and clear_table() s390: fix machine check panic stack switch ...

sched/cputime, powerpc, s390: Make scaled cputime arch specific

2016-11-15T08:51:05Z

Only s390 and powerpc have hardware facilities allowing to measure cputimes scaled by frequency. On all other architectures utimescaled/stimescaled are equal to utime/stime (however they are accounted separately). Remove {u,s}timescaled accounting on all architectures except powerpc and s390, where those values are explicitly accounted in the proper places. Signed-off-by: Stanislaw Gruszka Signed-off-by: Frederic Weisbecker Cc: Benjamin Herrenschmidt Cc: Heiko Carstens Cc: Linus Torvalds Cc: Martin Schwidefsky Cc: Michael Neuling Cc: Paul Mackerras Cc: Paul Mackerras Cc: Peter Zijlstra Cc: Thomas Gleixner Link: http://lkml.kernel.org/r/20161031162143.GB12646@redhat.com Signed-off-by: Ingo Molnar

s390: move cputime accounting fields from thread_info to thread_struct

2016-11-11T15:37:43Z

The user_timer and system_timer fields are used for the per-thread cputime accounting code. The access to these values is simpler if they are moved to the thread_struct as the task_thread_info(tsk) indirection is not needed anymore. Reviewed-by: Heiko Carstens Signed-off-by: Martin Schwidefsky

s390: add missing declarations

2016-05-10T11:24:51Z

arch_dup_task_struct and the per cpu variable mt_cycles are globally visible, but do not have any header file with a declaration. Therefore add it so we have proper type checking in place. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky

s390/vtime: correct scaled cputime of partially idle CPUs

2015-09-30T14:22:38Z

The calculation for the SMT scaling factor for a hardware thread which has been partially idle needs to disregard the cycles spent by the other threads of the core while the thread is idle. Signed-off-by: Martin Schwidefsky

s390/vtime: correct scaled cputime for SMT

2015-09-17T11:43:43Z

The scaled cputime is supposed to be derived from the normal per-thread cputime by dividing it with the average thread density in the last interval. The calculation of the scaling values for the average thread density is incorrect. The current, incorrect calculation: Ci = cycle count with i active threads T = unscaled cputime, sT = scaled cputime sT = T * (C1 + C2 + ... + Cn) / (1*C1 + 2*C2 + ... + n*Cn) The calculation happens to yield the correct numbers for the simple cases with only one Ci value not zero. But for cases with multiple Ci values not zero it fails. E.g. on a SMT-2 system with one thread active half the time and two threads active for the other half of the time it fails, the scaling factor should be 3/4 but the formula gives 2/3. The correct formula is sT = T * (C1/1 + C2/2 + ... + Cn/n) / (C1 + C2 + ... + Cn) Signed-off-by: Martin Schwidefsky

s390/vtime: limit MT scaling value updates

2015-08-04T12:26:17Z

The MT scaling values are updated on each calll to do_account_vtime. This function is called for each HZ interrupt and for each context switch. Context switch can happen often, the STCCTM instruction on this path is noticeable. Limit the updates to once per jiffy. Signed-off-by: Martin Schwidefsky

s390: add SMT support

2015-01-22T11:16:01Z

The multi-threading facility is introduced with the z13 processor family. This patch adds code to detect the multi-threading facility. With the facility enabled each core will surface multiple hardware threads to the system. Each hardware threads looks like a normal CPU to the operating system with all its registers and properties. The SCLP interface reports the SMT topology indirectly via the maximum thread id. Each reported CPU in the result of a read-scp-information is a core representing a number of hardware threads. To reflect the reduced CPU capacity if two hardware threads run on a single core the MT utilization counter set is used to normalize the raw cputime obtained by the CPU timer deltas. This scaled cputime is reported via the taskstats interface. The normal /proc/stat numbers are based on the raw cputime and are not affected by the normalization. Signed-off-by: Martin Schwidefsky

s390/vtime: Get rid of redundant WARN_ON

2014-12-18T08:55:56Z

in the cpu time accounting function vtime_account_irq_enter (vtime_account_system) we use a WARN_ON_ONCE(!irqs_disabled()). This is redundant as the function virt_timer_forward is always called and has a BUG_ON(!irqs_disabled()). This saves several nanoseconds in my specific testcase (KVM entry/exit) and probably all other callers like (soft)irq entry/exit. Signed-off-by: Christian Borntraeger Reviewed-by: Heiko Carstens Signed-off-by: Martin Schwidefsky