65 lines
2.6 KiB
C
65 lines
2.6 KiB
C
//
|
|
// Copyright 2020 Staysail Systems, Inc. <info@staysail.tech>
|
|
// Copyright 2018 Capitar IT Group BV <info@capitar.com>
|
|
//
|
|
// This software is supplied under the terms of the MIT License, a
|
|
// copy of which should be located in the distribution where this
|
|
// file was obtained (LICENSE.txt). A copy of the license may also be
|
|
// found online at https://opensource.org/licenses/MIT.
|
|
//
|
|
|
|
#ifndef CORE_REAP_H
|
|
#define CORE_REAP_H
|
|
|
|
#include "core/defs.h"
|
|
#include "core/list.h"
|
|
|
|
// nni_reap_node is to be inserted inline into structures
|
|
// for subsystems that wish to support deferred reaping.
|
|
// It should be zeroed at object initialization, but apart
|
|
// from that it must not be touched directly except by the
|
|
// reap subsystem.
|
|
typedef struct nni_reap_node nni_reap_node;
|
|
struct nni_reap_node {
|
|
nni_reap_node *rn_next;
|
|
};
|
|
|
|
// nni_reap_list is for subsystems to define their own reap lists.
|
|
// This allows for the reap linkage to be restricted to a single
|
|
// pointer. Subsystems should initialize rl_offset and rl_func,
|
|
// and leave the rest zeroed. The intention is that this is a global
|
|
// static member for each subsystem.
|
|
typedef struct nni_reap_list nni_reap_list;
|
|
struct nni_reap_list {
|
|
nni_reap_list *rl_next; // linkage in global reap list
|
|
nni_reap_node *rl_nodes; // list of nodes to reap
|
|
size_t rl_offset; // offset of reap_node within member.
|
|
nni_cb rl_func; // function called to reap the item
|
|
bool rl_inited; // initialized means it is linked in the list
|
|
};
|
|
|
|
// nni_reap performs an asynchronous reap of an item. This allows functions
|
|
// it calls to acquire locks or resources without worrying about deadlocks
|
|
// (such as from a completion callback.) The called function should avoid
|
|
// blocking for too long if possible, since only one reap thread is present
|
|
// in the system. The intended usage is for an nni_reap_node to be a member
|
|
// of the structure to be reaped, and and then this function is called to
|
|
// finalize it.
|
|
//
|
|
// Note that is is possible to re-queue an item to reap on the reap list.
|
|
// This is useful if, for example, a reference count indicates that the item
|
|
// is busy. These will be queued at the end of the reap list. This will
|
|
// allow a dependency to defer reaping until its dependents have first been
|
|
// reaped. HOWEVER, it is important that the item in question actually be
|
|
// part of a fully reap-able graph; otherwise this can lead to an infinite
|
|
// loop in the reap thread.
|
|
|
|
extern void nni_reap(nni_reap_list *, void *);
|
|
|
|
// nni_reap_drain waits for the reap queue to be drained.
|
|
extern void nni_reap_drain(void);
|
|
|
|
extern int nni_reap_sys_init(void);
|
|
extern void nni_reap_sys_fini(void);
|
|
|
|
#endif // CORE_REAP_H
|