On Fri, May 29, 2020 at 2:00 PM Peter Zijlstra <pet...@infradead.org> wrote: > > On Thu, May 14, 2020 at 12:26:36PM +0300, john mathew wrote: > > > +============= > > +CFS Overview > > +============= > > + > > +Linux 2.6.23 introduced a modular scheduler core and a Completely Fair > > +Scheduler (CFS) implemented as a scheduling module. A brief overview of the > > +CFS design is provided in :doc:`sched-design-CFS` > > + > > +In addition there have been many improvements to the CFS, a few of which > > are > > + > > +**Thermal Pressure**: > > I find these attached headers really hard to read. And what's with the > ** stuff ? > > Other files in this same patch use a different style: > > Header > ------ > test goes here, > > Which I find a lot more readable. Use it here too? > > > +Scale CPU capacity mechanism for CFS so it knows how much CPU capacity is > > left > > +for its use after higher priority sched classes (RT, DL), IRQs and > > +'Thermal Pressure' have reduced the 'original' CPU capacity. > > +Thermal pressure on a CPU means the maximum possible capacity is > > +unavailable due to thermal events. > > + > > +** Optimizations to NUMA balancing**: > ^ iconsistent spacing (although I think it's more readable without > the ** crap attached). > > > +When gathering NUMA statistics, information about whether a core is Idle > > +is also cached. In case of an imbalance, instead of doing a second scan of > > +the node runqueues, the idle core is used as the migration target. When > > +doing so multiple tasks can attempt to select an idle CPU but fail, because > > +a NUMA balance is active on that CPU. In this case an alternative idle CPU > > +scanned. Another optimization is to terminate the search for swap candidate > > +when a reasonable one is found instead of searching all the CPUs on the > > +target domain. > > ^^ that makes no sense to me. That's very much not what numa balancing > is about. > > > + > > +**Asymmetric CPU capacity wakeup scan**: > > +Previous assumption that CPU capacities within an SD_SHARE_PKG_RESOURCES > > +domain (sd_llc) are homogeneous didn't hold for newer generations of > > big.LITTLE > > +systems (DynamIQ) which can accommodate CPUs of different compute capacity > > +within a single LLC domain. A new idle sibling helper function was added > > +which took CPU capacity into account. The policy is to pick the first idle > > +CPU which is big enough for the task (task_util * margin < cpu_capacity). > > +If no idle CPU is big enough, the idle CPU with the highest capacity is > > +picked. > > + > > +**Optimized idle core selection**: > > +Skipped looping through all the threads of a core to evaluate if the > > +core is idle or not. If a thread of a core is not idle, evaluation of > > +other threads of the core can be skipped. > > + > > +**Load balance aggressively for SCHED_IDLE CPUs**: > > +Newly-woken task is preferred to be enqueued on a SCHED_IDLE CPU instead > > +of other busy or idle CPUs. Also load balancer is made to migrate tasks > > more > > +aggressively to a SCHED_IDLE CPU. Fair scheduler now does the next > > +load balance soon after the last non-SCHED_IDLE task is dequeued from a > > +runqueue, i.e. making the CPU SCHED_IDLE. Also the the busy_factor > > +used with the balance interval to prevent frequent load balancing > > +is ignored for such CPU's. > > + > > +**Load balancing algorithm Reworked**: > > +Some heuristics in the load balancing algorithm became meaningless because > > +of the rework of the scheduler's metrics like the introduction of PELT. > > +Those heuristics were removed. The new load balancing algorithm also fixes > > +several pending wrong tasks placement > > + > > + * the 1 task per CPU case with asymmetric system > > + * the case of CFS task preempted by other class > > + * the case of tasks not evenly spread on groups with spare capacity > > + > > +Also the load balance decisions have been consolidated in the 3 separate > > +functions. > > +* update_sd_pick_busiest() select the busiest sched_group. > > +* find_busiest_group() checks if there is an imbalance between local and > > +busiest group. > > +* calculate_imbalance() decides what have to be moved. > > This all reads like a changelog; why do we care about the old stuff? > That is, rephrase it to describe the current situation. > > > + > > +**Energy-aware wake-ups speeded up**: > > +Algorithmic complexity of the EAS was reduced from O(n^2) to O(n). > > +Previous algorithm resulted in prohibitively high wake-up latencies on > > +systems with complex energy models, such as systems with per-CPU DVFS. > > +The EAS wake-up path was re-factored to compute the energy 'delta' on a > > +per-performance domain basis, rather than the whole system. > > Idem; describe what EAS does and how. Nobody cares about what it once > might have been. > > > +**Selection of an energy-efficient CPU on task wake-up**: > > +An Energy efficient CPU is found by estimating the impact on system-level > > +active energy resulting from the placement of the task on the CPU with the > > +highest spare capacity in each performance domain. Energy Model (EM) is > > +used for this. This strategy spreads tasks in a performance domain and > > avoids overly > > +aggressive task packing. The best CPU energy-wise is then selected if it > > +saves a large enough amount of energy with respect to prev_cpu. > > That's EAS, not a separate thing. > > > + > > +**Consider misfit tasks when load-balancing**: > > +A task which ends up on a CPU which doesn't suit its compute demand is > > +identified as a misfit task in asymmetric CPU capacity systems. These > > +'misfit' tasks are migrated to CPUs with higher compute capacity to ensure > > +better throughput. A new group_type: group_misfit_task is added and > > indicates this > > +scenario. Tweaks to the load-balance code are done to make the migrations > > +happen. Misfit balancing is done between a source group of lower per-CPU > > +capacity and destination group of higher compute capacity. Otherwise, > > misfit > > +balancing is ignored. > > That's with the assymetric capacity thing, weird to be separate. > > > + > > + > > +**Make schedstats a runtime tunable that is disabled by default**: > > +A kernel command-line and sysctl tunable was added to enable or disable > > +schedstats on demand (when it's built in). It is disabled by default. > > +The benefits are dependent on how scheduler-intensive the workload is. > > So while I like the idea of an overview; this isn't one. An overview is > where we list current features, and explain (in short) why and what. > > > + > > diff --git a/Documentation/scheduler/index.rst > > b/Documentation/scheduler/index.rst > > index 9bdccea74af9..f311abe5b711 100644 > > --- a/Documentation/scheduler/index.rst > > +++ b/Documentation/scheduler/index.rst > > @@ -17,6 +17,8 @@ specific implementation differences. > > :maxdepth: 2 > > > > overview > > + sched-data-structs > > + cfs-overview > > sched-design-CFS > > sched-features > > arch-specific > > diff --git a/Documentation/scheduler/overview.rst > > b/Documentation/scheduler/overview.rst > > index aee16feefc61..7536bec6afce 100644 > > --- a/Documentation/scheduler/overview.rst > > +++ b/Documentation/scheduler/overview.rst > > @@ -3,3 +3,291 @@ > > ==================== > > Scheduler overview > > ==================== > > + > > +Linux kernel implements priority-based scheduling. More than one process > > are > > +allowed to run at any given time and each process is allowed to run as if > > it > > +were the only process on the system. The process scheduler coordinates > > which > > +process runs when. In that context, it has the following tasks: > > + > > +* share CPU cores equally among all currently running processes. > > +* pick appropriate process to run next if required, considering scheduling > > + class/policy and process priorities. > > +* balance processes between multiple cores in SMP systems. > > indent the bullets at least one space, like: > > * share CPU cores... > * pick .. > > Write it like you want to read this as a text document. Ignore all that > RST bullshit. > > Also, your terminology is ambiguous, what is a core? > > > +The scheduler attempts to be responsive for I/O bound processes and > > efficient > > +for CPU bound processes. The scheduler also applies different scheduling > > +policies for real time and normal processes based on their respective > > +priorities. > > > > Higher priorities in the kernel have a numerical smaller > > +value. Real time priorities range from 1 (highest) – 99 whereas normal > > +priorities range from 100 – 139 (lowest). > > The whole priorities thing is a mess; and you've missed some of it. -1 > is actually the highest (static) priority. But even that doesn't > adequately describe things, since we have a dynamic priority scheduling > class these days. > > Most everything that looks at the static priority of tasks these days; > and doesn't: > > - use it to distinguish classes > - is the RR/FIFO static priority scheduler > > is doing it wrong. Yes we have heaps of legacy, but it's not a main > feature anymore. Slowly but surely the ->prio field becomes less and > less relevant. > > The fair class uses the static priority field to encode the nice level, > but is nice a priority? I think not. > > > Scheduler implements many scheduling > > 1, 2, many, right? ;-) _5_ is the number: stop, deadline, rt, fair, > idle. > > > +classes which encapsulate a particular scheduling policy. Each scheduling > > +policy implements scheduler handling of tasks that belong to a particular > > +priority. > > policy enumeration: > > SCHED_DEADLINE goes here.. > > > SCHED_FIFO and SCHED_RR policies handle real time priorities tasks > > They're both a static priority scheduling class. They're the only ones > for which the term priority actually has a sane meaning. > > > +while SCHED_NORMAL and SCHED_BATCH policies handle tasks with normal > > priorities. > > What's an abnormal priority? Both these are weighted proportionally fair > and encode the weight, as nice value, in the prio field, in a range not > overlapping the static prio range. But that doesn't make it a priority. > > > +SCHED_IDLE is also a normal scheduling policy when means its priority can > > +be set between 100 – 139 range too but they are treated as priority 139. > > Priority for SCHED_IDLE is meaningless, the only reason it 'has' one is > so that code that looks to do a prio->class mapping works. > > > +Their priority doesn't matter since they get minimal weight > > WEIGHT_IDLEPRI=3. > > This. > > > +SCHED_DEADLINE policy tasks have negative priorities, reflecting > > +the fact that any of them has higher priority than RT and NORMAL/BATCH > > tasks. > > Tada, you did find the -1! > > > +And then there are the maintenance scheduler classes: idle sched class and > > +stop sched class. Idle class doesn't manage any user tasks and so doesn't > > Ah, so you do want to treat those too; so perhaps then present it like a > double iteration: > > - stop_class: > > - dl_class: > > * SCHED_DEADLINE: > > - rt_class > > * SCHED_RR / SCHED_FIFO: > > - fair_class: > > * SCHED_NORMAL/SCHED_BATCH: > * SCHED_IDLE: > > - idle_class: > > That's far easier to read than a blob of words. > > > +implement a policy. Its idle tasks 'swapper/X' has priority 120 and and > > aren't > > +visible to user space. Idle tasks are responsible for by putting the CPUs > > +into deep idle states when there is no work to do. > > Priority for idle task is irrelevant, if they have a prio it is purely > by accident. Looking at ->prio for idle task would be a stright bug. > The "swapper" name is a historical accident. We do not in fact swap from > it. > > > +Stop sched class is also used internally by the kernel doesn't implement > > any > > +scheduling policy. Stopper tasks 'migration/X' disguise as as a SCHED_FIFO > > +task with priority 139. > > Really? I thought we exposed it as a FIFO-99 (userpsace 99, not kernel > 99) task. Then again, I haven't actually looked at it recently. The > reason we disguise it is to present a 'known' class to userspace, to > avoid growing the ABI for this. > > > Stopper tasks are a mechanism to force a CPU to stop > > +running everything else and perform a specific task. As this is the > > +highest-priority class, it can preempt everything else and nothing ever > > +preempts it. It is used by one CPU to stop another in order to run a > > specific > > +function, so it is only available on SMP systems. This class is used by the > > +kernel for task migration. > > > > > + > > + > > +Process Management > > +================== > > + > > +Each process in the system is represented by struct task_struct. When a > > +process/thread is created, the kernel allocates a new task_struct for it. > > +The kernel then stores this task_struct in an RCU list. Macro next_task() > > +allows a process to obtain its next task and for_each_process() macro > > enables > > +traversal of the list. > > + > > +Frequently used fields of the task struct are: > > + > > +*state:* The running state of the task. The possible states are: > > + > > +* TASK_RUNNING: The task is currently running or in a run queue waiting > > + to run. > > +* TASK_INTERRUPTIBLE: The task is sleeping waiting for some event to occur. > > + This task can be interrupted by signals. On waking up the task > > transitions > > + to TASK_RUNNING. > > +* TASK_UNINTERRUPTIBLE: Similar to TASK_INTERRUPTIBLE but does not wake > > + up on signals. Needs an explicit wake-up call to be woken up. Contributes > > + to loadavg. > > +* __TASK_TRACED: Task is being traced by another task like a debugger. > > +* __TASK_STOPPED: Task execution has stopped and not eligible to run. > > + SIGSTOP, SIGTSTP etc causes this state. The task can be continued by > > + the signal SIGCONT. > > +* TASK_PARKED: State to support kthread parking/unparking. > > +* TASK_DEAD: If a task dies, then it sets TASK_DEAD in tsk->state and calls > > + schedule one last time. The schedule call will never return. > > return to this task; obviously the system keeps running so it must do > something. > > Perhaps its clearer to state that the task will never be ran again. > > > +* TASK_WAKEKILL: It works like TASK_UNINTERRUPTIBLE with the bonus that it > > + can respond to fatal signals. > > +* TASK_WAKING: To handle concurrent waking of the same task for SMP. > > + Indicates that someone is already waking the task. > > +* TASK_NOLOAD: To be used along with TASK_UNINTERRUPTIBLE to indicate > > + an idle task which does not contribute to loadavg. > > +* TASK_NEW: Set during fork(), to guarantee that no one will run the task, > > + a signal or any other wake event cannot wake it up and insert it on > > + the runqueue. > > + > > +*exit_state* : The exiting state of the task. The possible states are: > > + > > +* EXIT_ZOMBIE: The task is terminated and waiting for parent to collect > > + the exit information of the task. > > +* EXIT_DEAD: After collecting the exit information the task is put to > > + this state and removed from the system. > > + > > +*static_prio:* Nice value of a task. The value of this field does > > + not change. Value ranges from -20 to 19. This value is mapped to nice > > + value and used in the scheduler. > > + > > +*prio:* Dynamic priority of a task. Previously a function of static > > + priority and tasks interactivity. Value not used by CFS scheduler but used > > + by the RT scheduler. Might be boosted by interactivity modifiers. Changes > > Again, no point in mentioning things that aren't there. Those can only > serve to confuse. > > > + upon fork, setprio syscalls, and whenever the interactivity estimator > > + recalculates. > > There is no interactivity estimator. > > > + > > +*normal_prio:* Expected priority of a task. The value of static_prio > > + and normal_prio are the same for non-real-time processes. For real time > > + processes value of prio is used. > > + > > +*rt_priority:* Field used by real time tasks. Real time tasks are > > + prioritized based on this value. > > + > > +*sched_class:* Pointer to sched_class CFS structure. > > + > > +*sched_entity:* Pointer to sched_entity CFS structure. > > + > > +*policy:* Value for scheduling policy. The possible values are: > > + > > +* SCHED_NORMAL: Regular tasks use this policy. > > +* SCHED_BATCH: Tasks which need to run longer without preemption > > + use this policy. Suitable for batch jobs. > > +* SCHED_IDLE: Policy used by background tasks. > > +* SCHED_FIFO & SCHED_RR: These policies for real time tasks. Handled by > > + real time scheduler. > > +* SCHED_DEADLINE: Tasks which are activated on a periodic or sporadic > > fashion > > + use this policy. This policy implements the Earliest Deadline First (EDF) > > + scheduling algorithm. This policy is explained in detail in the > > + :doc:`sched-deadline` documentation. > > + > > +*nr_cpus_allowed:* Bit field containing tasks affinity towards a set of > > + CPU cores. Set using sched_setaffinity() system call. > > nr_cpus_allowed it not a bitfield, it is the hamming weight of a > bitmap. The actual bitmap is found through cpus_ptr. > > > + > > +New processes are created using the fork() system call which is described > > +at manpage :manpage:`FORK(2)` or the clone system call described at > > +:manpage:`CLONE(2)`. > > > > > +Users can create threads within a process to achieve parallelism. Threads > > +share address space, open files and other resources of the process. Threads > > +are created like normal tasks with their unique task_struct, but clone() > > +is provided with flags that enable the sharing of resources such as address > > +space :: > > + > > + clone(CLONE_VM | CLONE_FS | CLONE_FILES | CLONE_SIGHAND, 0); > > + > > +The scheduler schedules task_structs so from scheduler perspective there is > > +no difference between threads and processes. Threads are created using > > +the system call pthread_create described at :manpage:`PTHREAD_CREATE(3)` > > +POSIX threads creation is described at :manpage:`PTHREADS(7)` > > + > > +The Scheduler Entry Point > > +========================= > > + > > +The main scheduler entry point is an architecture independent schedule() > > +function defined in kernel/sched/core.c. Its objective is to find a > > process in > > +the runqueue list and then assign the CPU to it. It is invoked, directly > > +or in a lazy (deferred) way from many different places in the kernel. A > > lazy > > +invocation does not call the function by its name, but gives the kernel a > > +hint by setting a flag TIF_NEED_RESCHED. The flag is a message to the > > kernel > > +that the scheduler should be invoked as soon as possible because another > > +process deserves to run. > > Perhaps add a warning that direct manipulation of TIF_NEED_RESCHED is > unwise. You make it sound like a simple thing -- which I understand from > the PoV of explaining how it sort-of works, but might give people the > wrong impression. > > > + > > +Following are some places that notify the kernel to schedule: > > + > > +* scheduler_tick() > > + > > +* Running task goes to sleep state : Right before a task goes to sleep, > > + schedule() will be called to pick the next task to run and the change > > + its state to either TASK_INTERRUPTIBLE or TASK_UNINTERRUPTIBLE. For > > + instance, prepare_to_wait() is one of the functions that makes the > > + task go to the sleep state. > > + > > +* try_to_wake_up() > > + > > +* yield() > > it is likely that every single user of yield() is a bug. > > > +* wait_event() > > + > > +* cond_resched() : It gives the scheduler a chance to run a higher-priority > > + process. > > + > > +* cond_resched_lock() : If a reschedule is pending, drop the given lock, > > + call schedule, and on return reacquire the lock. > > voluntary preemption points > > > +* do_task_dead() > > + > > +* preempt_schedule() : The function checks whether local interrupts are > > + enabled and the preempt_count field of current is zero; if both > > + conditions are true, it invokes schedule() to select another process > > + to run. > > + > > +* preempt_schedule_irq() > > + > > +Calling functions mentioned above leads to a call to __schedule(). Note > > +that preemption must be disabled before it is called and enabled after > > +the call using preempt_disable and preempt_enable functions family. > > It might be less confusing if you classify those: > > - blocking operations: > > * mutex_lock() / wait_event() / etc.. > > - co-operative / voluntary preemption: > > * cond_resched*() > * yield() > * preempt_enable() > > - involuntary preemption: > > * scheduler_tick() > * wake_up_process() > > The blocking oeprations will suspend the current task and directly call > into the scheduler to find something else to do. > > The co-operative/voluntary crud will allow another task to run at that > point (subject to preemption model). > > The involuntary preemption things will mark TIF_NEED_RESCHED and wait > for action (again depending on preemption model). > > > + > > +The steps during invocation are: > > +-------------------------------- > > +1. Disable preemption to avoid another task preempting the scheduling > > + thread itself. > > +2. Retrieve the runqueue of current processor and its lock is obtained to > > + allow only one thread to modify the runqueue at a time. > > +3. The state of the previously executed task when the schedule() > > + was called is examined. If it is not runnable and has not been > > + preempted in kernel mode, it is removed from the runqueue. If the > > + previous task has non-blocked pending signals, its state is set to > > + TASK_RUNNING and left in the runqueue. > > +4. Scheduler classes are iterated and the corresponding class hook to > > + pick the next suitable task to be scheduled on the CPU is called. > > + Since most tasks are handled by the sched_fair class, a shortcut to this > > + class is implemented in the beginning of the function. > > +5. TIF_NEED_RESCHED and architecture specific need_resched flags are > > cleared. > > +6. If the scheduler class picks a different task from what was running > > + before, a context switch is performed by calling context_switch(). > > + Internally, context_switch() switches to the new task's memory map and > > + swaps the register state and stack. If scheduler class picked the same > > + task as the previous task, no task switch is performed and the current > > + task keeps running. > > +7. Balance callback list is processed. Each scheduling class can migrate > > tasks > > + between CPUs to balance load. These load balancing operations are queued > > + on a Balance callback list which get executed when balance_callback() is > > + called. > > +8. The runqueue is unlocked and preemption is re-enabled. In case > > + preemption was requested during the time in which it was disabled, > > + schedule() is run again right away. > > + > > +Scheduler State Transition > > +========================== > > + > > +A very high level scheduler state transition flow with a few states can > > +be depicted as follows. :: > > + > > + * > > + | > > + | task > > + | forks > > + v > > + +------------------------------+ > > + | TASK_NEW | > > + | (Ready to run) | > > + +------------------------------+ > > + | > > + | > > + v > > + +------------------------------------+ > > + | TASK_RUNNING | > > + +---------------> | (Ready to run) | <--+ > > + | +------------------------------------+ | > > + | | | > > + | | schedule() calls context_switch() | task is > > preempted > > + | v | > > + | +------------------------------------+ | > > + | | TASK_RUNNING | | > > + | | (Running) | ---+ > > + | event occurred +------------------------------------+ > > + | | > > + | | task needs to wait for event > > + | v > > + | +------------------------------------+ > > + | | TASK_INTERRUPTIBLE | > > + | | TASK_UNINTERRUPTIBLE | > > + +-----------------| TASK_WAKEKILL | > > + +------------------------------------+ > > + | > > + | task exits via do_exit() > > + v > > + +------------------------------+ > > + | TASK_DEAD | > > + | EXIT_ZOMBIE | > > + +------------------------------+ > > + > > + > > +Scheduler provides trace events tracing all major events of the scheduler. > > +The trace events are defined in :: > > + > > + include/trace/events/sched.h > > + > > +Using these trace events it is possible to model the scheduler state > > transition > > +in an automata model. The following journal paper discusses such modeling: > > + > > +Daniel B. de Oliveira, Rômulo S. de Oliveira, Tommaso Cucinotta, **A thread > > +synchronization model for the PREEMPT_RT Linux kernel**, *Journal of > > Systems > > +Architecture*, Volume 107, 2020, 101729, ISSN 1383-7621, > > +https://doi.org/10.1016/j.sysarc.2020.101729. > > + > > +To model the scheduler efficiently the system was divided in to generators > > +and specifications. Some of the generators used were "need_resched", > > +"sleepable" and "runnable", "thread_context" and "scheduling context". > > +The specifications are the necessary and sufficient conditions to call > > +the scheduler. New trace events were added to specify the generators > > +and specifications. In case a kernel event referred to more than one > > +event, extra fields of the kernel event was used to distinguish between > > +automation events. The final model was generated from parallel composition > > +of all generators and specifications which composed of 34 events, > > +12 generators and 33 specifications. This resulted in 9017 states, and > > +20103 transitions. > > diff --git a/Documentation/scheduler/sched-cas.rst > > b/Documentation/scheduler/sched-cas.rst > > new file mode 100644 > > index 000000000000..fcebc5770803 > > --- /dev/null > > +++ b/Documentation/scheduler/sched-cas.rst > > @@ -0,0 +1,92 @@ > > +.. SPDX-License-Identifier: GPL-2.0+ > > + > > +========================= > > +Capacity-Aware Scheduling > > +========================= > > + > > +Scheduling load balancing on Asymmetric Multiprocessor systems was improved > > +through the introduction of Capacity-Aware Scheduling. It identifies the > > +most efficient CPU to assign a task based on its capacity. This capacity > > +may be asymmetric due to heterogeneous computing architecture such > > +as ARM big.LITTLE. Scheduler gets information about asymmetric capacities > > +when the scheduler domain hierarchy is built using build_sched_domains(). > > +CPU capacities are provided to the scheduler topology code through the > > +architecture specific implementation of the arch_scale_cpu_capacity(). > > +The SD_ASYM_CPUCAPACITY flag is set by the scheduler topology for a domain > > +in the hierarchy where all CPU capacities are visible for any cpu's point > > +of view on asymmetric CPU capacity systems. The scheduler can then take > > +capacity asymmetry into account when load balancing. > > + > > +Initial CPU capacities are derived from the Device Tree and CPU frequency. > > +For RISC-V & ARM64 it is done in drivers/base/arch_topology.c. A cpu-map > > +device tree is parsed to obtain the cpu topology and the initial CPU > > capacity > > +is set using the CPUFreq subsystem. A callback is registered to the CPUFreq > > +subsystem to rebuild sched_domains once the CPUFreq is loaded, which is > > when > > +a complete view of the capacities of the CPUs (which is a mix of µarch and > > +frequencies) is available. > > + > > +Asymmetric CPU capacity information is used in > > + > > +* Energy Aware Scheduling: The scheduler is able to predict the impact of > > + its decisions on the energy consumed by CPUs. Described in > > :doc:`sched-energy` . > > +* Optimized task wakeup load balancing by finding idle CPU with enough > > capacity. > > + > > +The different scheduler classes asymmetric use the Asymmetric CPU capacity > > +information differently. > > + > > +CFS Capacity Awareness > > +====================== > > + > > +Used to identify misfit tasks: > > +A load intensive task on a CPU which doesn't meet its compute demand is > > +identified as a misfit task. 'Misfit' tasks are migrated to CPUs with > > +higher compute capacity to ensure better throughput. CFS frequently updates > > +the misfit status of the current task by comparing its utilization vs the > > +CPU capacity using task_fits_capacity(). If the utilization is more than > > the > > +CPU capacity the calculated misfit load is updated to the runqueue > > +rq->misfit_task_load. This misfit load is then checked by the load > > +balancing operations to migrate the task to a CPU of higher capacity. > > + > > +Modified wakeup logic to support DynamIQ systems: > > +When the scheduler class calls select_task_rq_fair to select a runqueue for > > +a waking task, load balancing is performed by selecting the idlest CPU in > > +the idlest group, or under certain conditions an idle sibling CPU if the > > +domain has SD_WAKE_AFFINE set. In DynamIQ systems Last Level Cache (LLC) > > +domain of a CPU spans all CPUs in the system. This may include CPU's of > > +different capacities. So in select_idle_sibling() an idle sibling is picked > > +based on CPU capacity for asymmetric CPU capacity systems and for symmetric > > +systems use LLC domain is used. The policy is to pick the first idle CPU > > +which is big enough for the task (task_util * margin < cpu_capacity). > > +If no idle CPU is big enough, the idle CPU with the highest capacity is > > +picked. For asymmetric CPU capacity systems select_idle_sibling() operates > > +on the sd_asym_cpucapacity sched_domain pointer, which is guaranteed to > > span > > +all known CPU capacities in the system. This works for both "legacy" > > +big.LITTLE (LITTLEs & bigs split at MC, joined at DIE) and for newer > > +DynamIQ systems (e.g. LITTLEs and bigs in the same MC domain). > > + > > + > > +RT Capacity Awareness > > +===================== > > + > > +Since RT tasks doesn't have a per task utilization signal RT tasks uses > > uclamp > > +to guarantee a minimum performance point. Utilization clamping is a > > mechanism > > +which allows to "clamp" (i.e. filter) the utilization generated by RT and > > +FAIR tasks within a range defined by user-space. It exposes to user-space a > > +new set of per-task attributes the scheduler can use as hints about the > > +expected/required utilization for a task. RT is made capacity aware > > +by ensuring that the capacity of the CPU is >= uclamp_min value. This check > > +is done in the rt_task_fits_capacity() > > + > > +DL Capacity Awareness > > +===================== > > + > > +TBD > > + > > + > > + > > + > > + > > + > > + > > + > > + > > diff --git a/Documentation/scheduler/sched-data-structs.rst > > b/Documentation/scheduler/sched-data-structs.rst > > new file mode 100644 > > index 000000000000..a16408676b71 > > --- /dev/null > > +++ b/Documentation/scheduler/sched-data-structs.rst > > @@ -0,0 +1,182 @@ > > +.. SPDX-License-Identifier: GPL-2.0+ > > + > > +========================= > > +Scheduler Data Structures > > +========================= > > + > > +The main parts of the Linux scheduler are: > > + > > +Runqueue > > +~~~~~~~~ > > + > > +:c:type:`struct rq <rq>` is the central data structure of process > > I so hate that rst crap; John, can't we teach the thing that anything > called 'struct foo' or 'foo_t' is in fact a C type, just like we did > with foo() being a function? > > > +scheduling. It keeps track of tasks that are in a runnable state assigned > > +for a particular processor. Each CPU has its own run queue and stored in a > > +per CPU array:: > > + > > + DEFINE_PER_CPU(struct rq, runqueues); > > + > > +Access to the queue requires locking and lock acquire operations must be > > +ordered by ascending runqueue. Macros for accessing and locking the > > runqueue > > +are provided in:: > > + > > + kernel/sched/sched.h > > + > > +The runqueue contains scheduling class specific queues and several > > scheduling > > +statistics. > > + > > +Scheduling entity > > +~~~~~~~~~~~~~~~~~ > > +Scheduler uses scheduling entities which contain sufficient information to > > +actually accomplish the scheduling job of a task or a task-group. The > > +scheduling entity may be a group of tasks or a single task. Every task is > > +associated with a sched_entity structure. CFS adds support for nesting of > > +tasks and task groups. Each scheduling entity may be run from its parents > > +runqueue. The scheduler traverses the sched_entity hierarchy to pick the > > +next task to run on the CPU. The entity gets picked up from the cfs_rq on > > +which it is queued and its time slice is divided among all the tasks on > > its my_q. > > + > > +Scheduler classes > > +~~~~~~~~~~~~~~~~~ > > +It is an extensible hierarchy of scheduler modules. The modules encapsulate > > +scheduling policy details. They are called from the core code which is > > +independent. Scheduling classes are implemented through the sched_class > > +structure. dl_sched_class for deadline scheduler, fair_sched_class for CFS > > +and rt_sched_class for RT are implementations of this class. > > wrong order; also perhaps a reference to the earlier exposition of the > same. > > > +The important methods of scheduler class are: > > + > > +enqueue_task and dequeue_task > > You started an enumeration with ':' above, but have no bullet here, > perhaps: > > - sched_class::enqueue_task() > > or somesuch? > > > + These functions are used to put and remove tasks from the runqueue > > + respectively to change a property of a task. This is referred to as > > + change pattern. Change is defined as the following sequence of calls:: > > + > > + * dequeue_task() > > + * put_prev_task() > > + * change a property > > + * enqueue_task() > > + * set_next_task() > > + > > + The enqueue_task function takes the runqueue, the task which needs to > > + be enqueued/dequeued and a bit mask of flags as parameters. The main > > + purpose of the flags is to describe why the enqueue or dequeue is being > > + called. The different flags used are described in :: > > + > > + kernel/sched/sched.h > > + > > + Some places where the enqueue_task and dequeue_task are called for > > + changing task properties are > > + > > + * When migrating a task from one CPU's runqueue to another. > > + * When changing a tasks CPU affinity. > > + * When changing the priority of a task. > > + * When changing the nice value of the task. > > + * When changing the scheduling policy and/or RT priority of a thread. > > + > > +pick_next_task > > + Called by the scheduler to pick the next best task to run. The > > scheduler > > + iterates through the corresponding functions of the scheduler classes > > + in priority order to pick up the next best task to run. Since tasks > > + belonging to the idle class and fair class are frequent, the scheduler > > + optimizes the picking of next task to call the pick_next_task_fair() > > + if the previous task was of the similar scheduling class. > > implies set_next_task() having been called on the task returned. > > (the core-sched patches have some variants here, just in case you're > curious) > > > + > > +put_prev_task > > + Called by the scheduler when a running task is being taken off a CPU. > > + The behavior of this function depends on individual scheduling classes. > > + In CFS class this function is used to put the currently running task > > back > > + into the CFS RB tree. When a task is running it is dequeued from the > > tree. > > + This is to prevent redundant enqueue's and dequeue's for updating its > > + vruntime. vruntime of tasks on the tree needs to be updated by > > update_curr() > > + to keep the tree in sync. In SCHED_DEADLINE and RT classes additional > > tree > > + is maintained to push tasks from the current CPU to another CPU where > > the > > + task can preempt and start executing. Task will be added to this queue > > + if it is present on the scheduling class rq and the task has affinity > > + to more than one CPU. > > + > > +set_next_task > > + Pairs with the put_prev_task(), this function is called when the next > > + task is set to run on the CPU. This function is called in all the > > places > > + where put_prev_task is called to complete the 'change pattern'. In case > > + of CFS scheduling class, it will set current scheduling entity to the > > + picked task and accounts bandwidth usage on the cfs_rq. In addition it > > + will also remove the current entity from the CFS runqueue for the > > vruntime > > + update optimization, opposite to what was done in put_prev_task. > > + For the SCHED_DEADLINE and RT classes it will remove the task from the > > + tree of pushable tasks trigger the balance callback to push another > > task > > + which is non running on the current CPU for execution on another CPU. > > + > > + * dequeue the picked task from the tree of pushable tasks. > > + * update the load average in case the previous task belonged to another > > + class. > > + * queues the function to push tasks from current runqueue to other CPUs > > + which can preempt and start execution. Balance callback list is used. > > + > > +task_tick > > + Called from scheduler_tick(), hrtick() and sched_tick_remote() to > > update > > + the current task statistics and load averages. Also restarting the high > > + resolution tick timer is done if high resolution timers are enabled. > > + scheduler_tick() runs at 1/HZ and is called from the timer interrupt > > + handler of the Kernel internal timers. > > + hrtick() is called from high resolution timers to deliver an accurate > > + preemption tick as the regular scheduler tick that runs at 1/HZ can be > > + too coarse when nice levels are used. > > + sched_tick_remote() gets called by the offloaded residual 1Hz scheduler > > + tick. In order to reduce interruptions to bare metal tasks, it is > > possible > > + to outsource these scheduler ticks to the global workqueue so that a > > + housekeeping CPU handles those remotely. > > + > > +select_task_rq > > + Called by scheduler to get the CPU to assign a task to and migrating > > + tasks between CPUs. Flags describe the reason the function was called. > > + Called by try_to_wake_up() with SD_BALANCE_WAKE flag which wakes up a > > + sleeping task. > > + Called by wake_up_new_task() with SD_BALANCE_FORK flag which wakes up a > > + newly forked task. > > + Called by sched_exec() with SD_BALANCE_EXEC which is called from execv > > + syscall. > > + SCHED_DEADLINE class decides the CPU on which the task should be woken > > + up based on the deadline. RT class decides based on the RT priority. > > Fair > > + scheduling class balances load by selecting the idlest CPU in the > > + idlest group, or under certain conditions an idle sibling CPU if the > > + domain has SD_WAKE_AFFINE set. > > + > > +balance > > + Called by pick_next_task() from scheduler to enable scheduling classes > > + to pull tasks from runqueues of other CPUs for balancing task execution > > + between the CPUs. > > + > > +task_fork > > + Called from sched_fork() of scheduler which assigns a task to a CPU. > > + Fair scheduling class updates runqueue clock, runtime statistics and > > + vruntime for the scheduling entity. > > + > > +yield_task > > + Called from SYSCALL sched_yield to yield the CPU to other tasks. > > + SCHED_DEADLINE class forces the runtime of the task to zero using a > > special > > + flag and dequeues the task from its trees. RT class requeues the task > > + entities to the end of the run list. Fair scheduling class implements > > + the buddy mechanism. This allows skipping onto the next highest > > priority > > + scheduling entity at every level in the CFS tree, unless doing so would > > + introduce gross unfairness in CPU time distribution. > > + > > +check_preempt_curr > > + Check whether the task that woke up should preempt the currently > > + running task. Called by scheduler, > > + > > + * when moving queued task to new runqueue > > + * ttwu() > > + * when waking up newly created task for the first time. > > + > > + SCHED_DEADLINE class compares the deadlines of the tasks and calls > > + scheduler function resched_curr() if the preemption is needed. In case > > + the deadlines are equal, migratability of the tasks is used a criteria > > + for preemption. > > + RT class behaves the same except it uses RT priority for comparison. > > + Fair class sets the buddy hints before calling resched_curr() to > > preempt. > > + > > +Scheduler sets the scheduler class for each task based on its priority. > > +Tasks assigned with SCHED_NORMAL, SCHED_IDLE and SCHED_BATCH call > > +fair_sched_class hooks and tasks assigned with SCHED_RR and > > +SCHED_FIFO call rt_sched_class hooks. Tasks assigned with SCHED_DEADLINE > > +policy calls dl_sched_class hooks. > > Nice! > > > diff --git a/Documentation/scheduler/sched-features.rst > > b/Documentation/scheduler/sched-features.rst > > index 1afbd9cc8d52..e576c7d9e556 100644 > > --- a/Documentation/scheduler/sched-features.rst > > +++ b/Documentation/scheduler/sched-features.rst > > @@ -17,4 +17,5 @@ Scheduler Features > > sched-energy > > sched-nice-design > > sched-rt-group > > + sched-cas > > completion > > diff --git a/Documentation/scheduler/scheduler-api.rst > > b/Documentation/scheduler/scheduler-api.rst > > new file mode 100644 > > index 000000000000..1fc6bd4c2908 > > --- /dev/null > > +++ b/Documentation/scheduler/scheduler-api.rst > > @@ -0,0 +1,31 @@ > > +.. SPDX-License-Identifier: GPL-2.0+ > > + > > +============================= > > +Scheduler related functions > > +============================= > > + > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: __schedule > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: scheduler_tick > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: try_to_wake_up > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: do_task_dead > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: preempt_schedule_irq > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: prepare_task_switch > > + > > +.. kernel-doc:: kernel/sched/core.c > > + :functions: finish_task_switch > > + > > +.. kernel-doc:: kernel/sched/sched.h > > + :functions: rq > > + > > diff --git a/kernel/sched/core.c b/kernel/sched/core.c > > index 9a2fbf98fd6f..b349ed9b4d92 100644 > > --- a/kernel/sched/core.c > > +++ b/kernel/sched/core.c > > @@ -3576,9 +3576,13 @@ void arch_set_thermal_pressure(struct cpumask *cpus, > > WRITE_ONCE(per_cpu(thermal_pressure, cpu), th_pressure); > > } > > > > -/* > > +/** > > + * scheduler_tick - sched tick timer handler > > + * > > * This function gets called by the timer code, with HZ frequency. > > * We call it with interrupts disabled. > > + * > > + * Return: 0. > > */ > > void scheduler_tick(void) > > { > > @@ -3959,8 +3963,10 @@ pick_next_task(struct rq *rq, struct task_struct > > *prev, struct rq_flags *rf) > > BUG(); > > } > > > > -/* > > - * __schedule() is the main scheduler function. > > +/** > > + * __schedule() - the main scheduler function. > > + * > > + * @preempt: preemption enabled/disabled > > * > > * The main means of driving the scheduler and thus entering this function > > are: > > * > > @@ -4089,6 +4095,12 @@ static void __sched notrace __schedule(bool preempt) > > balance_callback(rq); > > } > > > > +/** > > + * do_task_dead - handle task exit > > + * > > + * Changes the the task state to TASK_DEAD and calls > > + * schedule to pick next task to run. > > + */ > > void __noreturn do_task_dead(void) > > { > > /* Causes final put_task_struct in finish_task_switch(): */ > > @@ -4320,7 +4332,8 @@ EXPORT_SYMBOL_GPL(preempt_schedule_notrace); > > > > #endif /* CONFIG_PREEMPTION */ > > > > -/* > > +/** > > + * preempt_schedule_irq - schedule from irq context > > * This is the entry point to schedule() from kernel preemption > > * off of irq context. > > * Note, that this is called and return with irqs disabled. This will > > @@ -5618,6 +5631,13 @@ SYSCALL_DEFINE0(sched_yield) > > } > > > > #ifndef CONFIG_PREEMPTION > > +/** > > + * _cond_resched - explicit rescheduling > > + * > > + * gives the scheduler a chance to run a higher-priority process > > + * > > + * Return: 1 if reschedule was done, 0 if reschedule not done. > > + */ > > This isn't a general API and should not have the kerneldoc on. Put it on > cond_resched*() in linux/sched.h instead. > > > int __sched _cond_resched(void) > > { > > if (should_resched(0)) { > > diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h > > index db3a57675ccf..21f2953b72c7 100644 > > --- a/kernel/sched/sched.h > > +++ b/kernel/sched/sched.h > > @@ -865,12 +865,175 @@ struct uclamp_rq { > > }; > > #endif /* CONFIG_UCLAMP_TASK */ > > > > -/* > > - * This is the main, per-CPU runqueue data structure. > > +/** > > + * struct rq - This is the main, per-CPU runqueue data structure. > > * > > * Locking rule: those places that want to lock multiple runqueues > > * (such as the load balancing or the thread migration code), lock > > * acquire operations must be ordered by ascending &runqueue. > > + * > > + * @lock: > > + * lock to be acquired while modifying the runqueue > > + * @nr_running: > > + * number of runnable tasks on this queue > > + * @nr_numa_running: > > + * number of tasks running that care about their placement > > + * @nr_preferred_running: > > + * number of tasks that are optimally NUMA placed > > + * @numa_migrate_on: > > + * per run-queue variable to check if NUMA-balance is > > + * active on the run-queue > > + * @last_blocked_load_update_tick: > > + * tick stamp for decay of blocked load > > + * @has_blocked_load: > > + * idle CPU has blocked load > > + * @nohz_tick_stopped: > > + * CPU is going idle with tick stopped > > + * @nohz_flags: > > + * flags indicating NOHZ idle balancer actions > > + * @nr_load_updates: > > + * unused > > + * @nr_switches: > > + * number of context switches > > + * @uclamp: > > + * utilization clamp values based on CPU's RUNNABLE tasks > > + * @uclamp_flags: > > + * flags for uclamp actions, currently one flag for idle. > > + * @cfs: > > + * fair scheduling class runqueue > > + * @rt: > > + * rt scheduling class runqueue > > + * @dl: > > + * dl scheduing class runqueue > > + * @leaf_cfs_rq_list: > > + * list of leaf cfs_rq on this CPU > > + * @tmp_alone_branch: > > + * reference to add child before its parent in leaf_cfs_rq_list > > + * @nr_uninterruptible: > > + * global counter where the total sum over all CPUs matters. A task > > + * can increase this counter on one CPU and if it got migrated > > + * afterwards it may decrease it on another CPU. Always updated under > > + * the runqueue lock > > + * @curr: > > + * points to the currently running task of this rq. > > + * @idle: > > + * points to the idle task of this rq > > + * @stop: > > + * points to the stop task of this rq > > + * @next_balance: > > + * shortest next balance before updating nohz.next_balance > > + * @prev_mm: > > + * real address space of the previous task > > + * @clock_update_flags: > > + * RQCF clock_update_flags bits > > + * @clock: > > + * sched_clock() value for the queue > > + * @clock_task: > > + * clock value minus irq handling time > > + * @clock_pelt: > > + * clock which scales with current capacity when something is > > + * running on rq and synchronizes with clock_task when rq is idle > > + * @lost_idle_time: > > + * idle time lost when utilization of a rq has reached the > > + * maximum value > > + * @nr_iowait: > > + * account the idle time that we could have spend running if it > > + * were not for IO > > + * @membarrier_state: > > + * copy of membarrier_state from the mm_struct > > + * @rd: > > + * root domain, each exclusive cpuset essentially defines an island > > + * domain by fully partitioning the member CPUs from any other cpuset > > + * @sd: > > + * a domain heirarchy of CPU groups to balance process load among them > > + * @cpu_capacity: > > + * information about CPUs heterogeneity used for CPU performance > > + * scaling > > + * @cpu_capacity_orig: > > + * original capacity of a CPU before being altered by > > + * rt tasks and/or IRQ > > + * @balance_callback: > > + * queue to hold load balancing push and pull operations > > + * @idle_balance: > > + * flag to do the nohz idle load balance > > + * @misfit_task_load: > > + * set whenever the current running task has a utilization > > + * greater than 80% of rq->cpu_capacity. A non-zero value > > + * in this field enables misfit load balancing > > + * @active_balance: > > + * synchronizes accesses to ->active_balance_work > > + * @push_cpu: > > + * idle cpu to push the running task on to during active load > > + * balancing. > > + * @active_balance_work: > > + * callback scheduled to run on one or multiple cpus > > + * with maximum priority monopolozing those cpus. > > + * @cpu: > > + * CPU of this runqueue > > + * @online: > > + * Used by scheduling classes to support CPU hotplug > > + * @cfs_tasks: > > + * an MRU list used for load balancing, sorted (except > > + * woken tasks) starting from recently given CPU time tasks > > + * toward tasks with max wait time in a run-queue > > + * @avg_rt: > > + * track the utilization of RT tasks for a more accurate > > + * view of the utilization of the CPU when overloaded by CFS and > > + * RT tasks > > + * @avg_dl: > > + * track the utilization of DL tasks as CFS tasks can be preempted > > + * by DL tasks and the CFS's utilization might no longer describe > > + * the real utilization level > > + * @avg_irq: > > + * track the the utilization of interrupt to give a more accurate > > + * level of utilization of CPU taking into account the time spent > > + * under interrupt context when rqs' clock is updated > > + * @avg_thermal: > > + * tracks thermal pressure which is the reduction in maximum > > + * possible capacity due to thermal events > > + * @idle_stamp: > > + * time stamp at which idle load balance started for this rq. > > + * Used to find the idlest CPU, when multiple idle CPUs are in > > + * the same state > > + * @avg_idle: > > + * average idle time for this rq > > + * @max_idle_balance_cost: > > + * used to determine avg_idle's max value > > + * @prev_irq_time: > > + * updated to account time consumed when a previous > > + * update_rq_clock() happened inside a {soft,}irq region > > + * @prev_steal_time: > > + * to account how much elapsed time was spent in steal > > + * @prev_steal_time_rq: > > + * for fine granularity task steal time accounting by > > + * making update_rq_clock() aware of steal time > > + * @calc_load_update: > > + * sample window for global load-average calculations > > + * @calc_load_active: > > + * fold any nr_active delta into a global accumulate > > + * @hrtick_csd: > > + * call_single_data used to set hrtick timer state on a specific CPU > > + * @hrtick_timer: > > + * HR-timer to deliver an accurate preemption tick > > + * @rq_sched_info: > > + * runqueue specific latency stats > > + * @rq_cpu_time: > > + * runqueue specific accumulated per-task cpu runtime > > + * @yld_count: > > + * runqueue specific sys_sched_yield() stats > > + * @sched_count: > > + * runqueue specific __schedule() stats > > + * @sched_goidle: > > + * runqueue specific idle scheduling class stats > > + * @ttwu_count: > > + * runqueue specific idle ttwu stats , both remote and local > > + * @ttwu_local: > > + * ttwu count for the CPU of the rq > > + * @wake_list: > > + * list which stores tasks being woken up remotely by ttwu > > + * @idle_state: > > + * cpuidle state pointer of the CPU of this rq used to make a > > + * better decision when balancing tasks > > */ > > OMG... I suppose I appreciate the effort, but that's unwieldy and 100% > likely to get bitrotten real quick. > > Also, rq really isn't an exported API. So perhaps, if you really feel > the need, expand the comment on some of the fields inside the structure, > but I don't see this as anything other than a giant blob to ignore. > > > struct rq { > > /* runqueue lock: */ > > @@ -1136,7 +1299,7 @@ static inline u64 rq_clock_task(struct rq *rq) > > return rq->clock_task; > > } > > > > -/** > > +/* > > * By default the decay is the default pelt decay period. > > * The decay shift can change the decay period in > > * multiples of 32. > > -- > > 2.17.1 > >
Hi Peter, I addressed the your comments in https://lore.kernel.org/lkml/20200605092906.29478-1-john.mat...@unikie.com/ Can you take a look and ack if ok? -John
