rcu.rst (3811B)
1.. _rcu_doc: 2 3RCU Concepts 4============ 5 6The basic idea behind RCU (read-copy update) is to split destructive 7operations into two parts, one that prevents anyone from seeing the data 8item being destroyed, and one that actually carries out the destruction. 9A "grace period" must elapse between the two parts, and this grace period 10must be long enough that any readers accessing the item being deleted have 11since dropped their references. For example, an RCU-protected deletion 12from a linked list would first remove the item from the list, wait for 13a grace period to elapse, then free the element. See listRCU.rst for more 14information on using RCU with linked lists. 15 16Frequently Asked Questions 17-------------------------- 18 19- Why would anyone want to use RCU? 20 21 The advantage of RCU's two-part approach is that RCU readers need 22 not acquire any locks, perform any atomic instructions, write to 23 shared memory, or (on CPUs other than Alpha) execute any memory 24 barriers. The fact that these operations are quite expensive 25 on modern CPUs is what gives RCU its performance advantages 26 in read-mostly situations. The fact that RCU readers need not 27 acquire locks can also greatly simplify deadlock-avoidance code. 28 29- How can the updater tell when a grace period has completed 30 if the RCU readers give no indication when they are done? 31 32 Just as with spinlocks, RCU readers are not permitted to 33 block, switch to user-mode execution, or enter the idle loop. 34 Therefore, as soon as a CPU is seen passing through any of these 35 three states, we know that that CPU has exited any previous RCU 36 read-side critical sections. So, if we remove an item from a 37 linked list, and then wait until all CPUs have switched context, 38 executed in user mode, or executed in the idle loop, we can 39 safely free up that item. 40 41 Preemptible variants of RCU (CONFIG_PREEMPT_RCU) get the 42 same effect, but require that the readers manipulate CPU-local 43 counters. These counters allow limited types of blocking within 44 RCU read-side critical sections. SRCU also uses CPU-local 45 counters, and permits general blocking within RCU read-side 46 critical sections. These variants of RCU detect grace periods 47 by sampling these counters. 48 49- If I am running on a uniprocessor kernel, which can only do one 50 thing at a time, why should I wait for a grace period? 51 52 See UP.rst for more information. 53 54- How can I see where RCU is currently used in the Linux kernel? 55 56 Search for "rcu_read_lock", "rcu_read_unlock", "call_rcu", 57 "rcu_read_lock_bh", "rcu_read_unlock_bh", "srcu_read_lock", 58 "srcu_read_unlock", "synchronize_rcu", "synchronize_net", 59 "synchronize_srcu", and the other RCU primitives. Or grab one 60 of the cscope databases from: 61 62 (http://www.rdrop.com/users/paulmck/RCU/linuxusage/rculocktab.html). 63 64- What guidelines should I follow when writing code that uses RCU? 65 66 See checklist.rst. 67 68- Why the name "RCU"? 69 70 "RCU" stands for "read-copy update". 71 listRCU.rst has more information on where this name came from, search 72 for "read-copy update" to find it. 73 74- I hear that RCU is patented? What is with that? 75 76 Yes, it is. There are several known patents related to RCU, 77 search for the string "Patent" in Documentation/RCU/RTFP.txt to find them. 78 Of these, one was allowed to lapse by the assignee, and the 79 others have been contributed to the Linux kernel under GPL. 80 There are now also LGPL implementations of user-level RCU 81 available (https://liburcu.org/). 82 83- I hear that RCU needs work in order to support realtime kernels? 84 85 Realtime-friendly RCU can be enabled via the CONFIG_PREEMPT_RCU 86 kernel configuration parameter. 87 88- Where can I find more information on RCU? 89 90 See the Documentation/RCU/RTFP.txt file. 91 Or point your browser at (http://www.rdrop.com/users/paulmck/RCU/).