aboutsummaryrefslogtreecommitdiffstats
path: root/litmus
Commit message (Collapse)AuthorAge
* make it compilepgmGlenn Elliott2013-07-05
|
* GSN-EDF: link tasks to local CPU if it is idleBjoern Brandenburg2013-06-25
| | | | | | | If a task is released on a CPU that is idle, then it is pointless to send an IPI to another CPU just to schedule the newly released task. This patch changes check_for_preemptions() to prefer the local CPU first if it is idle.
* Use vmalloc() to allocate Feather-Trace buffersBjoern Brandenburg2013-06-25
| | | | | | | | | | | | Allocating contiguous buffers with kmalloc() is subject to severe maximum size limits. For Feather-Trace to work, we don't actually need *physically* contiguous memory; virtually contiguous memory is sufficient. By switching to vmalloc(), the code gets simpler and we can support much larger buffers. For convenience, this patch also adds a configuration option for the desired Feather-Trace buffer size and increases the default size.
* Clean up 'time_slice' and 'completed' usageBjoern Brandenburg2013-06-08
| | | | | | The 'time_slice' parameter has not been relevant in a couple of years. The 'completed' flag was misused in a few places. This clean-up patch removes these outdated remnants.
* PFAIR: fix admission of suspended tasksBjoern Brandenburg2013-06-08
| | | | | | | The PFAIR would not properly schedule tasks that were admitted while being suspended. This patch changes the task_new() handler to set the "needs to be requeued" flag on suspended tasks, which is required to ensure that they are processed correctly by the wake_up() handler.
* C-EDF: fix admission of suspended tasksBjoern Brandenburg2013-06-08
| | | | | This patch fixes admission of suspended tasks under the P-FP plugin analogously to commit 3088111567f334cccff1123f9bd97f2ba52af944.
* GSN-EDF: fix admission of suspended tasksBjoern Brandenburg2013-06-08
| | | | | | In the task_new() handler, don't process a job arrival event if the task is actually suspended. This fixes the admission of suspended tasks.
* P-FP: fix admission of suspended taskBjoern Brandenburg2013-06-08
| | | | | This commit fixes admission of suspended tasks under the P-FP plugin analogously to commit 78ae3306ff8b63a4592044aa28c2f2cbc1d36b20.
* PSN-EDF: fix admission of suspended taskBjoern Brandenburg2013-06-08
| | | | | The PSN-EDF plugin used to crash when trying to admit a suspended task. This patch fixes the task_new() handler to tolerate this case.
* Fix synchronous task system release supportBjoern Brandenburg2013-06-08
| | | | | | | | | | | | | | | | | ...and also finally fix the requeue() race (hopefully). Commit 4ffefb822b9d65d4efbedb60e3c9a0e76895cc5b broke the synchronous task system release code. The problem is that release_at() clears the tsk_rt(t)->completed flag. The patch thus only made the race window smaller, but didn't close it entirely. Further, it broke the synchronous task set release, which relies on schedule() being called with completed == 1. This patch fixes both problems by introducing a new flag dont_requeue, which indicates unambiguously that a task should not be requeued by other processors. This flag is cleared only when the job completion is being processed, which always happens while holding the appropriate scheduler locks, which closes the race window.
* Use do_div() for integer division in litmus/sync.cBjoern Brandenburg2013-05-04
| | | | Breaks without do_div() on ARM.
* Changed completion flag use to remove race condition in Cedf,Gedf.Jonathan Herman2013-04-12
|
* Differentiate between PERIODIC and SPORADIC tasks.Glenn Elliott2013-03-12
| | | | | | | | | | | | | Tasks can now be PERIODIC or SPORADIC. PERIODIC tasks do not have their job number incremented when they wake up and are tardy. PERIODIC jobs must end with a call to sys_complete_job() to set up their next release. (Not currently supported by pfair.) SPORADIC tasks _do_ have their job number incremented when they wake up and are tardy. SPORADIC is the default task behavior, carrying forward Litmus's current behavior.
* EDF schedulers: Support early job releasing.Glenn Elliott2013-03-12
| | | | | | | | | | | | | This patch allows a task to request early releasing via rt_task parameters to sys_set_task_rt_param(). Note that early releasing can easily peg your CPUs since early-releasing tasks never suspend to wait for their next job. As such, early releasing is really only useful in the context of implementing bandwidth servers, interrupt handling threads (or any thread that spends most of its time waiting for an event), or short-lived computations. If early releasing pegs your CPUs, then you probably shouldn't be using it.
* C-EDF: Cluster-aware admission test.Glenn Elliott2013-03-12
| | | | | | | | cedf_admit_task() is too restrictive in that the task to be admitted must be executing on the same CPU as is set in the task's task_params. This patch allows the task to be admitted to be executing on the same cluter as the CPU set in the task's task_params.
* uncachedev: mmap memory that is not cached by CPUsGlenn Elliott2013-03-12
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch creates a new character device, uncachedev. Pages of RAM allocated by this device are not cached by CPUs. Uses for such pages: 1) Determining *very* pessimistic emperical worst- case execution times. 2) Compare against performance with caches (quantify the avg. case benefit). 3) Deterministic memory accesses (access cannot cause a cache eviction.) 4) Theoretically, increased performance can be achieved by storing infrequently accessed data in uncache pages. uncachedev allocates pages with the pgprot_noncached() page attribute for user applications. Since pages allocated by uncachedev are not locked in memory by default, applications with any access level may mmap pages with uncachedev. Limitations: 1) Uncache pages must be MAP_PRIVATE. 2) Remapping not supported. Usage (user level): int size = NR_PAGES*PAGE_SIZE; int fd = open("/dev/litmus/uncache", O_RDWR); char *data = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0); <...do stuff...> munmap(data, size);
* Disallow nesting of LITMUS^RT locksBjoern Brandenburg2013-02-15
| | | | | | | | | | | | | | | | | | | | | Nesting of locks was never supported in LITMUS^RT since the required analysis does not exist anyway. That is, as defined in the literature, the protocols implemented in LITMUS^RT have not been studied in conjunction with nested critical sections. In LITMUS^RT, attempting to nest locks could lead to silent or not-so-silent bugs. This patch makes this restriction explicit and returns EBUSY when a process attempts to nest resources. This is enforced on a protocol-by-protocol basis, which means that adding protocols with support for nesting in future versions is not affected by this change. Exception: PCP and SRP resources may be nested, but not within global critical sections.
* PCP: allow NULL config parameterBjoern Brandenburg2013-02-15
| | | | | When the config parameter is NULL, just default to the local CPU, which is what we want in 99% of the cases anyway.
* Fix tracepoint compilation errorstagingFelipe Cerqueira2013-02-14
| | | | | | Linux tracepoints for real time tasks were causing compilation errors. For the tracing data structures to be properly created, it's necessary to add a #define.
* P-FP: fix BUG_ON releated to priority inheritancearchive/unc-master-3.0Bjoern Brandenburg2013-02-04
| | | | | | | In some cases, the PCP priority inheritance code triggered a (defensive) BUG_ON() in fp_common.c. This shuffles the order of operations a bit to be compliant with the restriction that tasks are never compared against themselves.
* FDSO: fail to build if some ops are missingBjoern Brandenburg2013-02-04
| | | | Prevents out-of-bounds lookups.
* Reimplement plugin switching using stop_machine()Bjoern Brandenburg2013-01-10
| | | | | | stop_machine() does exactly what we want (avoid all concurrent scheduling activity) and much simpler than rolling our own (buggy) implementation.
* Avoid race in do_release_ts()Bjoern Brandenburg2013-01-10
| | | | | | | The list is concurrently being modified by the waking processes. This requires the use of the list_for_each_safe() iterator. Reported by Glenn Elliott.
* Config: PREEMPT_STATE_TRACE depends on DEBUG_KERNEL.Christopher Kenna2012-12-11
| | | | | | | | | | This patch makes CONFIG_PREEMPT_STATE_TRACE depend on CONFIG_DEBUG_KERNEL. Prior to this patch, selecting PREEMPT_STATE_TRACE resulted in linker errors (see below), because sched_state_name is not built unless DEBUG_KERNEL is selected. kernel/built-in.o: In function `schedule': (.sched.text+0x3d2): undefined reference to `sched_state_name'
* Send reboot notifications to LitmusGlenn Elliott2012-12-07
| | | | | | | | This patch causes reboot notifications to be send to Litmus. With this patch, Litmus attempts to switch back to the Linux-plugin before the reboot proceeds. Any failures to switch back are reported via printk() (the reboot is not halted).
* litmus: get rid of unused RT_F_EXIT_SEM flagManohar Vanga2012-11-27
| | | | | | | | | | | | | This patch removes the RT_F_EXIT_SEM flag. All code paths depending on it being true are assumed to be unreachable and removed. The 'flags' field in struct rt_params is left as-is for use by specific schedulers. For example, sched_pfair defines a custom flag RT_F_REQUEUE within the 'flags' field. Signed-off-by: Manohar Vanga <mvanga@mpi-sws.org>
* litmus: use is_present() instead of using rt_param->presentManohar Vanga2012-11-27
| | | | Signed-off-by: Manohar Vanga <mvanga@mpi-sws.org>
* litmus: get rid of RT_F_SLEEP and RT_F_RUNNINGManohar Vanga2012-11-27
| | | | | | | | This patch removes the flags RT_F_SLEEP and RT_F_RUNNING as their name is misleading. This patch replaces them with a 'completed' field in struct rt_param. Signed-off-by: Manohar Vanga <mvanga@mpi-sws.org>
* litmus: fix warning about unused label in sched_pfp.cManohar Vanga2012-11-27
| | | | | | | This patch fixes a warning about an unused label in sched_pfp.c when CONFIG_LITMUS_LOCKING is not set. Signed-off-by: Manohar Vanga <mvanga@mpi-sws.org>
* P-FP: be less permissive about non-running tasks in queueBjoern Brandenburg2012-10-18
| | | | No suspended task should ever be queued in this plugin.
* P-FP: be paranoid in the preemption caseBjoern Brandenburg2012-10-18
| | | | | Crash and burn if an expected preemption didn't happen. This is useful to flag any bugs in the queue management code...
* P-FP: use proper CONFIG_ symbolBjoern Brandenburg2012-10-18
| | | | It's CONFIG_LITMUS_LOCKING, not just CONFIG_LOCKING...
* P-FP: simplify priority comparisonBjoern Brandenburg2012-10-18
| | | | | | Make the priority comparison easier to read. Also, remove the "equal PID" clause and insert a corresponding BUG_ON() instead; this should really never happen.
* P-FP: simplify boost_priority()Bjoern Brandenburg2012-10-18
| | | | | | boost_priority() is only applied to already-scheduled tasks. Remove the (untested and unneeded) case handling unscheduled tasks, which was likely not correct anyway.
* P-FP: record job release records for tardy tasksBjoern Brandenburg2012-10-18
| | | | | When a job was tardy, the plugin failed to invoke sched_trace. This caused ugly "holes" in the visualized schedule.
* Move SEND_RESCHED tracing to preempt.cBjoern Brandenburg2012-10-18
| | | | | | | The SEND_RESCHED is really only interesting if the IPI was generated by LITMUS^RT. Therefore, we don't need to trace in Linux's architecture-specific code. Instead, we hook into the preemption state machine, which is informed about incoming IPIs anyway.
* Move SCHED2 trace points to post_schedule()Bjoern Brandenburg2012-10-18
| | | | | | | | | | Linux's post_schedule() scheduling class hook more closely matches what SCHED2 is supposed to trace, namely any scheduling overhead after the context switch. The prior trace points caught timers being armed from finish_switch(), which is already included in the context switch cost CXS. (This patch essentially reverts 8fe2fb8bb1c1cd0194608bc783d0ce7029e8d869).
* Feather-Trace: hide irq in SEND_RESCHED_ENDBjoern Brandenburg2012-10-18
| | | | | | | SEND_RESCHED_END is necessarily preceded by an interrupt. We don't want to filter events based on this expected interrupts, but we still want to detect samples disturbed by other interrupts. Hence, subtract one off the interrupt count.
* Make Feather-Trace timestamp collection non-preemptiveBjoern Brandenburg2012-10-18
| | | | | | | | | | | If the tracing code is interrupted / preempted inbetween the time that a sequence number is drawn and the time that the trace recorded is allocated, then the trace file will contain "out of order" events. These are difficult to detect during post-processing and can create artificial "outliers". This patch briefly disables local interrutps to avoid this. While at it, de-duplicate the timestamp recording code.
* Use user-provided irq count for TS_SYSCALL_IN_ENDBjoern Brandenburg2012-10-18
| | | | | | To detect interrupts that interfered after the initial time stamp was recorded, this patch changes sched_trace to also record the IRQ count as observed by userspace.
* Export IRQ count to control pageBjoern Brandenburg2012-10-18
| | | | Also add some compile-time checks to detect unexpected offsets.
* Feather-Trace: add support for locking-related syscall overheadsBjoern Brandenburg2012-10-18
| | | | | Support recording timestamps that allow tracing the entry and exit costs of locking-related system calls.
* Feather-Trace: update locking timestampsBjoern Brandenburg2012-10-18
| | | | | Reassing locking timestamps and prepare support for tracing system call overheads.
* Feather-Trace: keep track of PID in trace recordBjoern Brandenburg2012-10-18
| | | | | | | | To properly trace locking overheads, it is required to tell apart samples from different tasks, which requires keeping track of their PIDs. The timestamp is shortened to 48 bits to make room for the PID.
* ftdev: don't hold lock while sleepingBjoern B. Brandenburg2012-10-18
| | | | Holding the lock prevents other tasks from opening the device.
* FDSO: make sure fdso_close() is called during task teardownBjoern Brandenburg2012-10-18
| | | | | | | When a task exists and its FDSO table is released, the locking protocol's 'close()' callback should be invoked to do proper protocol-specific cleanup (such as unlocking the resource, if required).
* Re-implement synchronous task set release supportBjoern Brandenburg2012-10-18
| | | | | | | | | The old implementation had the problem that the first job release of all tasks was handled on the *same* CPU (that triggered the synchronous task set release). This could result in significant latency spikes. The new implementation uses a completion per task. Futher, each task programs its own release, which ensures that it will be on the proper processor.
* EDF priority tie-breaks.Glenn Elliott2012-09-21
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Instead of tie-breaking by PID (which is a static priority tie-break), we can tie-break by other job-level-unique parameters. This is desirable because tasks are equaly affected by tardiness since static priority tie-breaks cause tasks with greater PID values to experience the most tardiness. There are four tie-break methods: 1) Lateness. If two jobs, J_{1,i} and J_{2,j} of tasks T_1 and T_2, respectively, have equal deadlines, we favor the job of the task that had the worst lateness for jobs J_{1,i-1} and J_{2,j-1}. Note: Unlike tardiness, lateness may be less than zero. This occurs when a job finishes before its deadline. 2) Normalized Lateness. The same as #1, except lateness is first normalized by each task's relative deadline. This prevents tasks with short relative deadlines and small execution requirements from always losing tie-breaks. 3) Hash. The job tuple (PID, Job#) is used to generate a hash. Hash values are then compared. A job has ~50% chance of winning a tie-break with respect to another job. Note: Emperical testing shows that some jobs can have +/- ~1.5% advantage in tie-breaks. Linux's built-in hash function is not totally a uniform hash. 4) PIDs. PID-based tie-break used in prior versions of Litmus.
* Improve readability of EDF comparisons.Glenn Elliott2012-09-21
| | | | | | Restructured the EDF task comparison code to improve readability. Recoded chained logical expression embedded in return statement into a series of if/else blocks.
* Fix changed task_struct field name.Christopher Kenna2012-09-20
|
generated by cgit v1.2.2 (git 2.25.0) at 2025-04-11 14:54:30 -0400