2 .\" Copyright (C) 2018 Matthew Macy <mmacy@FreeBSD.org>.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice(s), this list of conditions and the following disclaimer as
9 .\" the first lines of this file unmodified other than the possible
10 .\" addition of one or more copyright notices.
11 .\" 2. Redistributions in binary form must reproduce the above copyright
12 .\" notice(s), this list of conditions and the following disclaimer in the
13 .\" documentation and/or other materials provided with the distribution.
15 .\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY
16 .\" EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
17 .\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
18 .\" DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY
19 .\" DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
20 .\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
21 .\" SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
22 .\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
41 .Nm epoch_drain_callbacks ,
43 .Nd kernel epoch based reclamation
49 .Fn epoch_alloc "int flags"
51 .Fn epoch_enter "epoch_t epoch"
53 .Fn epoch_enter_preempt "epoch_t epoch" "epoch_tracker_t et"
55 .Fn epoch_exit "epoch_t epoch"
57 .Fn epoch_exit_preempt "epoch_t epoch" "epoch_tracker_t et"
59 .Fn epoch_wait "epoch_t epoch"
61 .Fn epoch_wait_preempt "epoch_t epoch"
63 .Fn epoch_call "epoch_t epoch" "epoch_context_t ctx" "void (*callback) (epoch_context_t)"
65 .Fn epoch_drain_callbacks "epoch_t epoch"
67 .Fn in_epoch "epoch_t epoch"
69 Epochs are used to guarantee liveness and immutability of data by
70 deferring reclamation and mutation until a grace period has elapsed.
71 Epochs do not have any lock ordering issues.
72 Entering and leaving an epoch section will never block.
74 Epochs are allocated with
78 The flags passed to epoch_alloc determine whether preemption is
79 allowed during a section or not (the default), as specified by
81 Threads indicate the start of an epoch critical section by calling
83 The end of a critical section is indicated by calling
85 The _preempt variants can be used around code which requires preemption.
86 A thread can wait until a grace period has elapsed
87 since any threads have entered
91 .Fn epoch_wait_preempt ,
92 depending on the epoch_type.
93 The use of a default epoch type allows one to use
95 which is guaranteed to have much shorter completion times since
96 we know that none of the threads in an epoch section will be preempted
97 before completing its section.
98 If the thread can't sleep or is otherwise in a performance sensitive
99 path it can ensure that a grace period has elapsed by calling
101 with a callback with any work that needs to wait for an epoch to elapse.
102 Only non-sleepable locks can be acquired during a section protected by
103 .Fn epoch_enter_preempt
105 .Fn epoch_exit_preempt .
106 INVARIANTS can assert that a thread is in an epoch by using
109 The epoch API currently does not support sleeping in epoch_preempt sections.
110 A caller should never call
112 in the middle of an epoch section for the same epoch as this will lead to a deadlock.
114 By default mutexes cannot be held across
115 .Fn epoch_wait_preempt .
116 To permit this the epoch must be allocated with
118 When doing this one must be cautious of creating a situation where a deadlock is
119 possible. Note that epochs are not a straight replacement for read locks.
120 Callers must use safe list and tailq traversal routines in an epoch (see ck_queue).
121 When modifying a list referenced from an epoch section safe removal
122 routines must be used and the caller can no longer modify a list entry
124 An item to be modified must be handled with copy on write
125 and frees must be deferred until after a grace period has elapsed.
128 .Fn epoch_drain_callbacks
129 function is used to drain all pending callbacks which have been invoked by prior
131 function calls on the same epoch.
132 This function is useful when there are shared memory structure(s)
133 referred to by the epoch callback(s) which are not refcounted and are
135 The typical place for calling this function is right before freeing or
136 invalidating the shared resource(s) used by the epoch callback(s).
137 This function can sleep and is not optimized for performance.
139 .Fn in_epoch curepoch
140 will return 1 if curthread is in curepoch, 0 otherwise.
142 One must be cautious when using
143 .Fn epoch_wait_preempt
144 threads are pinned during epoch sections so if a thread in a section is then
145 preempted by a higher priority compute bound thread on that CPU it can be
146 prevented from leaving the section.
147 Thus the wait time for the waiter is
148 potentially unbounded.
154 in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_laddr *laddr,
158 epoch_enter(net_epoch);
159 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
161 if (sa->sa_family != AF_INET)
163 sin = (struct sockaddr_in *)sa;
164 if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
165 ia = (struct in_ifaddr *)ifa;
169 epoch_exit(net_epoch);
176 ifa_free(struct ifaddr *ifa)
179 if (refcount_release(&ifa->ifa_refcnt))
180 epoch_call(net_epoch, &ifa->ifa_epoch_ctx, ifa_destroy);
184 if_purgeaddrs(struct ifnet *ifp)
189 CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
190 IF_ADDR_WUNLOCK(ifp);
195 Thread 1 traverses the ifaddr list in an epoch.
196 Thread 2 unlinks with the corresponding epoch safe macro, marks as logically free,
197 and then defers deletion.
198 More general mutation or a synchronous
199 free would have to follow a call to
206 kernel programming interface is under development and is subject to change.