1This directory contains the following litmus tests: 2 3CoRR+poonceonce+Once.litmus 4 Test of read-read coherence, that is, whether or not two 5 successive reads from the same variable are ordered. 6 7CoRW+poonceonce+Once.litmus 8 Test of read-write coherence, that is, whether or not a read 9 from a given variable followed by a write to that same variable 10 are ordered. 11 12CoWR+poonceonce+Once.litmus 13 Test of write-read coherence, that is, whether or not a write 14 to a given variable followed by a read from that same variable 15 are ordered. 16 17CoWW+poonceonce.litmus 18 Test of write-write coherence, that is, whether or not two 19 successive writes to the same variable are ordered. 20 21IRIW+mbonceonces+OnceOnce.litmus 22 Test of independent reads from independent writes with smp_mb() 23 between each pairs of reads. In other words, is smp_mb() 24 sufficient to cause two different reading processes to agree on 25 the order of a pair of writes, where each write is to a different 26 variable by a different process? This litmus test is forbidden 27 by LKMM's propagation rule. 28 29IRIW+poonceonces+OnceOnce.litmus 30 Test of independent reads from independent writes with nothing 31 between each pairs of reads. In other words, is anything at all 32 needed to cause two different reading processes to agree on the 33 order of a pair of writes, where each write is to a different 34 variable by a different process? 35 36ISA2+pooncelock+pooncelock+pombonce.litmus 37 Tests whether the ordering provided by a lock-protected S 38 litmus test is visible to an external process whose accesses are 39 separated by smp_mb(). This addition of an external process to 40 S is otherwise known as ISA2. 41 42ISA2+poonceonces.litmus 43 As below, but with store-release replaced with WRITE_ONCE() 44 and load-acquire replaced with READ_ONCE(). 45 46ISA2+pooncerelease+poacquirerelease+poacquireonce.litmus 47 Can a release-acquire chain order a prior store against 48 a later load? 49 50LB+ctrlonceonce+mbonceonce.litmus 51 Does a control dependency and an smp_mb() suffice for the 52 load-buffering litmus test, where each process reads from one 53 of two variables then writes to the other? 54 55LB+poacquireonce+pooncerelease.litmus 56 Does a release-acquire pair suffice for the load-buffering 57 litmus test, where each process reads from one of two variables then 58 writes to the other? 59 60LB+poonceonces.litmus 61 As above, but with store-release replaced with WRITE_ONCE() 62 and load-acquire replaced with READ_ONCE(). 63 64MP+onceassign+derefonce.litmus 65 As below, but with rcu_assign_pointer() and an rcu_dereference(). 66 67MP+polockmbonce+poacquiresilsil.litmus 68 Protect the access with a lock and an smp_mb__after_spinlock() 69 in one process, and use an acquire load followed by a pair of 70 spin_is_locked() calls in the other process. 71 72MP+polockonce+poacquiresilsil.litmus 73 Protect the access with a lock in one process, and use an 74 acquire load followed by a pair of spin_is_locked() calls 75 in the other process. 76 77MP+polocks.litmus 78 As below, but with the second access of the writer process 79 and the first access of reader process protected by a lock. 80 81MP+poonceonces.litmus 82 As below, but without the smp_rmb() and smp_wmb(). 83 84MP+pooncerelease+poacquireonce.litmus 85 As below, but with a release-acquire chain. 86 87MP+porevlocks.litmus 88 As below, but with the first access of the writer process 89 and the second access of reader process protected by a lock. 90 91MP+wmbonceonce+rmbonceonce.litmus 92 Does a smp_wmb() (between the stores) and an smp_rmb() (between 93 the loads) suffice for the message-passing litmus test, where one 94 process writes data and then a flag, and the other process reads 95 the flag and then the data. (This is similar to the ISA2 tests, 96 but with two processes instead of three.) 97 98R+mbonceonces.litmus 99 This is the fully ordered (via smp_mb()) version of one of 100 the classic counterintuitive litmus tests that illustrates the 101 effects of store propagation delays. 102 103R+poonceonces.litmus 104 As above, but without the smp_mb() invocations. 105 106SB+mbonceonces.litmus 107 This is the fully ordered (again, via smp_mb() version of store 108 buffering, which forms the core of Dekker's mutual-exclusion 109 algorithm. 110 111SB+poonceonces.litmus 112 As above, but without the smp_mb() invocations. 113 114S+poonceonces.litmus 115 As below, but without the smp_wmb() and acquire load. 116 117S+wmbonceonce+poacquireonce.litmus 118 Can a smp_wmb(), instead of a release, and an acquire order 119 a prior store against a subsequent store? 120 121WRC+poonceonces+Once.litmus 122WRC+pooncerelease+rmbonceonce+Once.litmus 123 These two are members of an extension of the MP litmus-test 124 class in which the first write is moved to a separate process. 125 The second is forbidden because smp_store_release() is 126 A-cumulative in LKMM. 127 128Z6.0+pooncelock+pooncelock+pombonce.litmus 129 Is the ordering provided by a spin_unlock() and a subsequent 130 spin_lock() sufficient to make ordering apparent to accesses 131 by a process not holding the lock? 132 133Z6.0+pooncelock+poonceLock+pombonce.litmus 134 As above, but with smp_mb__after_spinlock() immediately 135 following the spin_lock(). 136 137Z6.0+pooncerelease+poacquirerelease+mbonceonce.litmus 138 Is the ordering provided by a release-acquire chain sufficient 139 to make ordering apparent to accesses by a process that does 140 not participate in that release-acquire chain? 141 142A great many more litmus tests are available here: 143 144 https://github.com/paulmckrcu/litmus 145