Abstract: A technique for low overhead contention-based switching between ticket locking and queued locking to access shared data may include establishing a ticket lock, establishing a queue lock, operating in ticket lock mode using the ticket lock to access the shared data during periods of relatively low data contention, and operating in queue lock mode using the queue lock to access the shared data during periods of relatively high data contention.

Abstract: A technique for implementing read-copy update in a shared-memory computing system having two or more processors operatively coupled to a shared memory and to associated incoherent caches that cache copies of data stored in the memory. According to example embodiments disclosed herein, cacheline information for data that has been rendered obsolete due to a data update being performed by one of the processors is recorded. The recorded cacheline information is communicated to one or more of the other processors. The one or more other processors use the communicated cacheline information to flush the obsolete data from all incoherent caches that may be caching such data.

Abstract: A technique for implementing read-copy update in a shared-memory computing system having two or more processors operatively coupled to a shared memory and to associated incoherent caches that cache copies of data stored in the memory. According to example embodiments disclosed herein, cacheline information for data that has been rendered obsolete due to a data update being performed by one of the processors is recorded. The recorded cacheline information is communicated to one or more of the other processors. The one or more other processors use the communicated cacheline information to flush the obsolete data from all incoherent caches that may be caching such data.

Abstract: A technique for improving the performance of RCU-based searches and updates to a shared data element group where readers must see consistent data with respect to the group as a whole. An updater creates one or more new group data elements and assigns each element a new generation number that is different than a global generation number associated with the data element group, allowing readers to track update versions. The updater links the new data elements into the data element group and then updates the global generation number so that referential integrity is maintained. This is done using a generation number element that is referenced by a header pointer for the data element group, and which in turn references or forms part of one of the data elements. After a grace period has elapsed, the any prior version of the generation number element may be freed.

Abstract: A technique for achieving low grace-period latencies in an energy efficient environment in which processors with Read-Copy Update (RCU) callbacks are allowed to enter low power states. In an example embodiment, for each processor that has RCU callbacks, different grace period numbers are assigned to different groups of the processor's RCU callbacks. New grace periods are periodically started and old grace periods are periodically ended. As old grace periods end, groups of RCU callbacks having corresponding assigned grace period numbers are invoked.

Abstract: A technique for resizing a first RCU-protected hash table stored in a memory. A second RCU-protected hash table is allocated in the memory as a resized version of the first hash table having a different number of hash buckets, with the hash buckets being defined but initially having no hash table elements. The second hash table is populated by linking each hash bucket thereof to all hash buckets of the first hash table containing elements that hash to the second hash bucket. The second hash table is then published so that it is available for searching by hash table readers. The first table is freed from memory after waiting for a grace period which guarantees that no readers searching the first hash table will be affected by the freeing.

Abstract: A technique for resizing a first RCU-protected hash table stored in a memory. A second RCU-protected hash table is allocated in the memory as a resized version of the first hash table having a different number of hash buckets, with the hash buckets being defined but initially having no hash table elements. The second hash table is populated by linking each hash bucket thereof to all hash buckets of the first hash table containing elements that hash to the second hash bucket. The second hash table is then published so that it is available for searching by hash table readers. The first table is freed from memory after waiting for a grace period which guarantees that no readers searching the first hash table will be affected by the freeing.

Abstract: A technique for resolving deadlocks between an RCU subsystem and an operating system scheduler. An RCU reader manipulates a counter when entering and exiting an RCU read-side critical section. At the entry, the counter is incremented. At the exit, the counter is manipulated differently depending on the counter value. A first counter manipulation path is taken when the counter indicates a task-context RCU reader is exiting an outermost RCU read-side critical section. This path includes condition-based processing that may result in invocation of the operating system scheduler. The first path further includes a deadlock protection operation that manipulates the counter to prevent an intervening RCU reader from taking the same path. The second manipulation path is taken when the counter value indicates a task-context RCU reader is exiting a non-outermost RCU read-side critical section, or an RCU reader is nested within the first path. This path bypasses the condition-based processing.

Abstract: A technique for resolving deadlocks between an RCU subsystem and an operating system scheduler. An RCU reader manipulates a counter when entering and exiting an RCU read-side critical section. At the entry, the counter is incremented. At the exit, the counter is manipulated differently depending on the counter value. A first counter manipulation path is taken when the counter indicates a task-context RCU reader is exiting an outermost RCU read-side critical section. This path includes condition-based processing that may result in invocation of the operating system scheduler. The first path further includes a deadlock protection operation that manipulates the counter to prevent an intervening RCU reader from taking the same path. The second manipulation path is taken when the counter value indicates a task-context RCU reader is exiting a non-outermost RCU read-side critical section, or an RCU reader is nested within the first path. This path bypasses the condition-based processing.

Abstract: A method, system and computer program product for managing requests for deferred updates to shared data elements while minimizing grace period detection overhead associated with determining whether pre-existing references to the data elements have been removed. Plural update requests that are eligible for grace period detection are buffered without performing grace period detection processing. One or more conditions that could warrant commencement of grace period detection processing are monitored while the update requests are buffered. If warranted by such a condition, grace period detection is performed relative to the update requests so that they can be processed. In this way, grace period detection overhead can be amortized over plural update requests while being sensitive to conditions warranting prompt grace period detection.

Abstract: In some embodiments, an apparatus includes a processor that is configured to execute computer usable program code to perform operations. The operations include executing an atomic transaction in a system having a transactional memory. The operations include receiving a signal interrupt during executing of the atomic transaction. The operations include storing a state of the signal interrupt to enable subsequent execution of the signal interrupt. The operations include returning to executing the atomic transaction until the atomic transaction is at least one of completed and aborted. The operations include after executing the atomic transaction is at least one of completed and aborted, determining whether the signal interrupt is received during executing of the atomic transaction. The operations include after determining that the signal interrupt is received during executing of the atomic transaction, retrieving the state of the signal interrupt.

Abstract: In some embodiments, a method includes executing an atomic transaction in a system having a transactional memory. The method includes receiving a signal interrupt during executing of the atomic transaction. The method includes storing a state of the signal interrupt to enable subsequent execution of the signal interrupt. The method includes returning to executing the atomic transaction until the atomic transaction is at least one of completed and aborted. The method includes after executing the atomic transaction is at least one of completed and aborted, determining whether the signal interrupt is received during executing of the atomic transaction. The method includes after determining that the signal interrupt is received during executing of the atomic transaction, retrieving the state of the signal interrupt. The method includes executing an interrupt handler for processing the signal interrupt and returning from executing of the atomic transaction.

Abstract: A technique for motivating lazy RCU callbacks under out-of-memory conditions. In response to detecting an actual or potential OOM condition, non-lazy callback processing is performed for all processors whose RCU callback lists are non-empty due to at least one callback permitting lazy callback processing being present.

Abstract: A technique for determining if a processor in a multiprocessor system implementing a read-copy update (RCU) subsystem may be placed in a low power state. The technique may include performing a first predictive query of the RCU subsystem to request permission for the processor to enter the low power state. If permission is denied, the processor is not placed in the low power state. If permission is granted, the processor is placed in the low power state for a non-fixed duration. Regardless whether permission is denied or granted, a second confirming query of the RCU subsystem is performed to redetermined whether it is permissible for the processor to be in the low power state.

Abstract: A technique for implementing SRCU with reduced OS jitter may include: (1) providing a pair of critical section counters for each CPU; (2) when entering an SRCU read-side critical section, incrementing one of the critical section counters associated with a first grace period; (3) when exiting an SRCU read-side critical section, decrementing one of the critical section counters associated with the first grace period; (4) when performing a data update, initiating the second grace period and performing a counter summation operation that sums the critical section counters associated with the first grace period to generate a critical section counter sum; (5) storing a snapshot value for each critical section counter during the summing; and (6) if the critical section counter sum indicates there are no active SRCU read-side critical sections for the first grace period, rechecking by comparing the snapshot values to current values of the critical section counters.

Abstract: A technique for detecting full-system idle state in an adaptive-tick kernel includes detecting non-timekeeping CPU idle state, initiating a hysteresis period, waiting for the hysteresis period to end, manipulating a data structure whose state indicates whether a scheduling clock tick may be disabled on all CPUs, and disabling the scheduling clock tick if the data structure is in an appropriate state. In a first embodiment, non-timekeeping CPUs manipulate a global counter when entering an idle state, but add hysteresis to avoid thrashing the counter. Timekeeping is turned off based on the count maintained on the global counter. In a second embodiment, a Read-Copy Update (RCU) dynticks-idle subsystem running on a timekeeping CPU manipulates a global state variable whose states indicate whether all non-timekeeping CPUs are in an idle state, and if so, for how long. Timekeeping is turned off based on the state of the global state variable.

Abstract: A tree-based trylock technique for reducing contention on a root trylock includes attempting to acquire a trylock at each node of a tree-based hierarchical node structure while following a traversal path that begins at a leaf node, passes through one or more of internal nodes, and ends at a root node having the root trylock. The trylock acquisition operation succeeds if each trylock on the traversal path is acquired, and fails if any trylock on the traversal path cannot be acquired. A trylock housekeeping operation releases all non-root trylocks visited by the trylock acquisition operation, such that if the trylock acquisition operation succeeds, only the root trylock will be remain acquired at the end of the operation, and if the trylock acquisition operation fails, none of the trylocks will be remain acquired at the end of the operation.

Abstract: A tree-based trylock technique for reducing contention on a root trylock includes attempting to acquire a trylock at each node of a tree-based hierarchical node structure while following a traversal path that begins at a leaf node, passes through one or more of internal nodes, and ends at a root node having the root trylock. The trylock acquisition operation succeeds if each trylock on the traversal path is acquired, and fails if any trylock on the traversal path cannot be acquired. A trylock housekeeping operation releases all non-root trylocks visited by the trylock acquisition operation, such that if the trylock acquisition operation succeeds, only the root trylock will be remain acquired at the end of the operation, and if the trylock acquisition operation fails, none of the trylocks will be remain acquired at the end of the operation.

Abstract: A technique for low overhead contention-based switching between ticket locking and queued locking to access shared data may include establishing a ticket lock, establishing a queue lock, operating in ticket lock mode using the ticket lock to access the shared data during periods of relatively low data contention, and operating in queue lock mode using the queue lock to access the shared data during periods of relatively high data contention.

Abstract: A technique for improving the performance of RCU-based searches and updates to a shared data element group where readers must see consistent data with respect to the group as a whole. An updater creates one or more new group data elements and assigns each element a new generation number that is different than a global generation number associated with the data element group, allowing readers to track update versions. The updater links the new data elements into the data element group and then updates the global generation number so that referential integrity is maintained. This is done using a generation number element that is referenced by a header pointer for the data element group, and which in turn references or forms part of one of the data elements. After a grace period has elapsed, the any prior version of the generation number element may be freed.