aboutsummaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAge
* Bias task scheduling towards previous CPU.Glenn Elliott2011-04-09
| | | | | | | | | Added the option to allow tasks to be scheduled on their previous CPU if that CPU is not currently executing real-time work. Helps preserve cache affinity and reduction in migration costs. See CONFIG_CPU_SCHED_BIAS help for more information. Solution developed by Jonathan Herman.
* Pfair: remove sporadic_release flagBjoern B. Brandenburg2011-02-01
| | | | | Instead of having an extra flag, Pfair should just infer sporadic release based on deadlines like other plugins, too.
* Pfair: support clustered schedulingBjoern B. Brandenburg2011-02-01
| | | | | | | Just like C-EDF is a global scheduler that is split across several clusters, Pfair can be applied on a per-cluster basis. This patch changes the Pfair implementation to enable clustering based on the recently added generic clustering support.
* Litmus core: add generic clustering supportBjoern B. Brandenburg2011-02-01
| | | | | Inspired by the existing C-EDF code, this generic version will build clusters of CPUs based on a given cache level.
* Litmus core: extract userspace interface from C-EDFBjoern B. Brandenburg2011-02-01
| | | | | Make the cluster size configuration in C-EDF generic so that it can be used by other clustered schedulers.
* Feather-Trace: rename locking trace pointsBjoern B. Brandenburg2011-02-01
| | | | | Since we don't expect to trace more than one lock type at a time, having protocol-specific trace points is not required.
* fdso: pass userpsace config argument to object constructorBjoern B. Brandenburg2011-02-01
| | | | | | | | | | | | As Glenn pointed out, it is useful for some protocols (e.g., k-exclusion protocols) to know the userspace configuration at object creation time. This patch changes the fdso API to pass the parameter to the object constructor, which is then in turn passed to the lock allocater. The return code from the lock allocater is passed to userspace in return. This also fixes some null pointer dereferences in the FDSO code found by the test suite in liblitmus.
* GSN-EDF: re-implement FMLP supportBjoern B. Brandenburg2011-02-01
| | | | This introduces the global FMLP based on the generic locking layer.
* PSN-EDF: re-implement FMLP supportBjoern B. Brandenburg2011-02-01
| | | | | Implement the partitioned FMLP with priority boosting based on the generic lock API.
* EDF: support priority boostingBjoern B. Brandenburg2011-02-01
| | | | While we are at it, simplify edf_higher_prio() a bit.
* FMLP: remove old implementationBjoern B. Brandenburg2011-02-01
|
* SRP: port to new generic locking APIBjoern B. Brandenburg2011-02-01
| | | | | This re-enables SRP support under PSN-EDF and demonstrates how the new locking API should be used.
* Litmus core: replace FMLP & SRP system calls with generic syscallsBjoern B. Brandenburg2011-02-01
| | | | | This renders the FMLP and SRP unfunctional until they are ported to the new locking API.
* Litmus core: add generic locking APIBjoern B. Brandenburg2011-02-01
| | | | | Provide a unified userspace interface for plugin-specific locking protocols.
* Litmus core: change plugin locking interface to generic 'allocate_lock()'Bjoern B. Brandenburg2011-02-01
| | | | | | | | | As the number of supported locking protocols is expected to rise, hard-coding things like priority inheritance in the plugin interface doesn't scale. Instead, use a new generic lock-ops approach. With this approach, each plugin can define its own protocol implementation (or use a generic one), and plugins can support multiple protocols without having to change the plugin interface for each protocol.
* fdso: supply object type to constructor and destructor methodsBjoern B. Brandenburg2011-02-01
| | | | Passing the object type explicitly will enable generic lock constructors.
* Fixed is_hrt, is_srt, and is_be macros.Jonathan Herman2011-02-01
|
* Added task class to feather trace param record.Jonathan Herman2011-02-01
|
* bugfix: don't let children stay Litmus real-time tasksBjoern B. Brandenburg2011-01-30
| | | | | | | | It has always been LITMUS^RT policy that children of real-time tasks may not skip the admissions test, etc. This used to be enforced, but was apparently dropped during some port. This commit re-introduces this policy. This fixes a kernel panic that occurred when "real-time children" exited without proper initilization.
* Litmus core: add copy_and_chomp() helperBjoern B. Brandenburg2011-01-30
| | | | | | We read in a line from userspace and remove the trailing newline in a number of places. This function extracts the common code to avoid future duplication.
* Added support for tracing arbitrary actions.wip-eventsJonathan Herman2011-01-26
|
* Feather-Trace: dynamic memory allocation and clean exit2011.1Christopher Kenna2011-01-12
| | | | | | This patch changes Feather-Trace to allocate memory for the minor devices dynamically, which addresses a long-standing FIXME. It also provides clean module exit and error conditions for Feather-Trace.
* Feather-Trace: register devices with sysfsChristopher Kenna2011-01-12
| | | | | | | | | | This patch implements support for Feather-Trace devices to use the sysfs file system and, consequently, udev support. This allows us to allocate major/minor numbers for Feather-Trace devices dynamically, which is desirable because our old static allocations tend to create conflicts on modern distributions and/or when there are many cores.
* cleanup C-EDF cluster size configurationBjoern B. Brandenburg2011-01-05
| | | | | | | Refactor the code that determines the C-EDF cluster size. - Use an enum with symbolic constants instead of magic int values. - Complain and fail to switch if an unsupported cluster size is requested. - Default to ALL as suggested by Glenn and Andrea.
* bugfix: clear scheduled field of the correct CPU upon task_exit in C-EDFAndrea Bastoni2011-01-03
| | | | | | | | | Do not use the "scheduled_on" field to address the cpus structure within a cluster. cpus may contain less items than num_online_cpus and we may cause an out-of-bound access. Instead, use "scheduled_on" to directly access the per-cpu cpu_entry_t structure. Reported-by: Jonathan Herman <hermanjl@cs.unc.edu>
* bugfix: avoid underflow in budget_remaining()Bjoern B. Brandenburg2010-11-29
| | | | | | | budget_remaining() reports incorrect values due to the operands being switched, which leads to an integer underflow. Reported-by: Chris Kenna <cjk@cs.unc.edu>
* bugfix: fix out-of-bound array access in cedf_activate_plugin()Andrea Bastoni2010-11-29
| | | | | | | Make sure to check for maximum index value when accessing cedf_domain_t array in cedf_activate_plugin(). Reported-by: Jeremy Erickson <jerickso@cs.unc.edu>
* add optional [function@file:line] tag to TRACE() logBjoern B. Brandenburg2010-11-22
| | | | | | | | | | | Add information to each trace message that makes it easier to locate where it came from. It is disabled by default since this adds a lot of clutter. Example: 81281 P1 [gsnedf_schedule@litmus/sched_gsn_edf.c:406]: (rtspin/1483:1) blocks:0 out_of_time:0 np:0 sleep:1 preempt:0 state:0 sig:0 81282 P1 [job_completion@litmus/sched_gsn_edf.c:303]: (rtspin/1483:1) job_completion(). 81283 P1 [__add_release@litmus/rt_domain.c:344]: (rtspin/1483:2) add_release(), rel=41941764351 81284 P1 [gsnedf_schedule@litmus/sched_gsn_edf.c:453]: (rtspin/1483:2) scheduled_on = NO_CPU
* log job number in TRACE_TASK() and TRACE_CUR()Bjoern B. Brandenburg2010-11-22
| | | | | | | | | | | | | | | | For some problems it can be helpful to know which job of a task generated a log message. This patch changes TRACE_TASK to add :<jobno> to the existing (<comm>/<pid>) tag. The result is a trace such as the following, in which the third job of rtspin/1511 completes and the fourth job is added to the release queue. 137615 P0: (rtspin/1511:3) job_completion(). 137616 P0: (rtspin/1511:4) add_release(), rel=262013223089 137617 P0: (rtspin/1511:4) scheduled_on = NO_CPU The job number for non-real-time tasks is always zero.
* Bugfix: synchronize with all other CPUs before switching pluginAndrea Bastoni2010-11-20
| | | | | | | | | | | The CPU triggering the plugin switch should wait until all other CPUs are in a proper state (synch_on_plugin_switch()) before performing the actual switch. Based on the original patch from Jeremy Erickson <jerickso@cs.unc.edu>. This should solve (for most practical cases) the C-EDF-related plugin-switch problem reported on the ML.
* Improve help message for TRACE() bufferBjoern B. Brandenburg2010-11-16
| | | | | It's not being allocated per cpu anymore. Further, provide a hint to the user where to find the data in userspace.
* Make TRACE() buffer size configurableBjoern B. Brandenburg2010-11-16
| | | | | | | | | | | Let the user choose an appropriate buffer size (instead of scaling with NR_CPUS). The kfifo api requires the buffer to be a power of two, so enforce this constraint in the configuration. This fixes a previously-existing compile-time error for values of NR_CPU that are not a power of two. Based on a patch by Mac Mollison <mollison@cs.unc.edu>.
* Workaround: do not set rq->skip_clock_updateBjoern B. Brandenburg2010-11-11
| | | | | | | | | | | | | Disabling the clock update seems to be causing problems even in normal Linux, and causes major bugs under LITMUS^RT. As a workaround, just disable this "optimization" for now. Details: the idle load balancer causes tasks that suspsend to be marked with set_tsk_need_resched(). When such a task resumes, it may wrongly trigger the setting of skip_clock_update. However, a corresponding rescheduling event may not happen immediately, such that the currently-scheduled task is no longer charged for its execution time.
* Remove LITMUS^RT TRACE_BUG_ON macroBjoern B. Brandenburg2010-11-11
| | | | | | Linux now has a macro of the same name, which causes namespace collisions. Since our version is only being used in two places that haven't triggered in several years, let's just remove it.
* Avoid warning on 64bit buildsBjoern B. Brandenburg2010-11-11
| | | | | The specifier %u doesn't match sizeof() if sizeof() returns a 64bit quantity on x86_64. Always cast it to int to avoid the warning.
* Cleanup TRACE() implementationBjoern B. Brandenburg2010-11-11
| | | | | | | | | | | | | | Since the intial rebased from .24 to .32, the TRACE() implementation was a hybrid between our old ringbuffer implementation and the new generic kfifo API. This was a) ugly and b) not save for TRACE() invoctations during early boot. This patch rips out the old parts and replaces the actual buffer with a static kfifo. This also increases TRACE() buffer size considerably. As we avoid a dynamic allocation, this a larger size is less problematic for debug builds. This helps a bit with holes in the debug log if the buffer-flushing task is starved.
* Hook up LITMUS^RT remote preemption support on ARMBjoern B. Brandenburg2010-11-11
| | | | Call into scheduler state machine in the IPI handler.
* Hook up LITMUS^RT remote preemption support on x86Bjoern B. Brandenburg2010-11-11
| | | | Call into scheduler state machine in the IPI handler.
* Implement proper remote preemption supportBjoern B. Brandenburg2010-11-11
| | | | | | | | | | | | To date, Litmus has just hooked into the smp_send_reschedule() IPI handler and marked tasks as having to reschedule to implement remote preemptions. This was never particularly clean, but so far we got away with it. However, changes in the underlying Linux, and peculartities of the ARM code (interrupts enabled before context switch) break this naive approach. This patch introduces new state-machine based remote preemption support. By examining the local state before calling set_tsk_need_resched(), we avoid confusing the underlying Linux scheduler. Further, this patch avoids sending unncessary IPIs.
* hook litmus tick function into hrtimer-driven ticksBjoern B. Brandenburg2010-11-11
| | | | | Litmus plugins should also be activated if ticks are triggered by hrtimer.
* Split out TRACE() from litmus.h and cleanup some includesBjoern B. Brandenburg2010-11-11
| | | | | | | | | | The TRACE() functionality doesn't need all of litmus.h. Currently, it's impossible to use TRACE() in sched.h due to a circular dependency. This patch moves TRACE() and friends to litmus/sched_debug.h, which can be included in sched.h. While at it, also fix some minor include ugliness that was revealed by this change.
* sched_trace: make buffer size configurableBjoern B. Brandenburg2010-11-11
| | | | | Large sched_trace buffers cause boot problems on the ARM box. Allow the user to specify smaller buffers.
* ARM: hookup LITMUS^RT system callsBjoern B. Brandenburg2010-11-11
| | | | | Includes the LITMUS^RT-specifc unistd.h extension and modifies the actual syscall table.
* ARM: Include LITMUS^RT KConfigBjoern B. Brandenburg2010-11-11
| | | | Make the ARM built aware of the LITMUS^RT-specific options.
* ARM: provide get_cycles() for RealView PB11{MP,76} and Cortex-A8Bjoern B. Brandenburg2010-11-11
| | | | | Use the CCNT register to override the default get_cycles() implementation in arch/arm/asm/timex.h. This is useful for overhead measurements and debugging.
* ARM: allow mach/timex.h to define get_cycles()Bjoern B. Brandenburg2010-11-11
| | | | | | Some platforms have access to a cycle coutner (CCNT) register in the CP15 coprocessor. This trivial change will allow such platforms to provide specialized implementations.
* C-EDF: move /proc/litmus/cluster_cache to /proc/litmus/plugins/C-EDF/clusterBjoern B. Brandenburg2010-11-11
| | | | | | | Make use of the new per-plugin proc file infrastructure to avoid littering the global namespace. While at it, also move all the relevant bits to sched_cedf.c. In the future, each plugin's parameters should be handled in the respective plugin file.
* Litmus core: refactor the implementation of /procChristopher Kenna2010-10-23
|
* Litmus core: per-plugin proc directoriesChristopher Kenna2010-10-23
| | | | | | Change the Litmus proc layout so that loaded plugins are visible in /proc/litmus/plugins/loaded and add Litmus functions make_plugin_proc_dir() and remove_plugin_proc_dir() to add per-plugin proc directories.
* Merge commit 'v2.6.36' into wip-merge-2.6.36Andrea Bastoni2010-10-23
|\ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Conflicts: Makefile arch/x86/include/asm/unistd_32.h arch/x86/kernel/syscall_table_32.S kernel/sched.c kernel/time/tick-sched.c Relevant API and functions changes (solved in this commit): - (API) .enqueue_task() (enqueue_task_litmus), dequeue_task() (dequeue_task_litmus), [litmus/sched_litmus.c] - (API) .select_task_rq() (select_task_rq_litmus) [litmus/sched_litmus.c] - (API) sysrq_dump_trace_buffer() and sysrq_handle_kill_rt_tasks() [litmus/sched_trace.c] - struct kfifo internal buffer name changed (buffer -> buf) [litmus/sched_trace.c] - add_wait_queue_exclusive_locked -> __add_wait_queue_tail_exclusive [litmus/fmlp.c] - syscall numbers for both x86_32 and x86_64
generated by cgit v1.2.2 (git 2.25.0) at 2024-11-10 01:42:12 -0500