The exact definition and implementation of a safepoint
changes from one VM implementation to another, but considering Hotspot VM, you can find a nice definition in: Safepoints in HotSpot JVM.
HotSpot glossary says:
A point during program execution at which all GC roots are known and all heap object contents are consistent. From a global point of view, all threads must block at a safepoint before the GC can run.
Typically, the safepoint is implemented by JVM injecting a safepoint check into a method, most call sites qualify as safepoints - when reaching the safepoint check, the thread will check if the safepoint is required (e.g. FullGC is scheduled), if yes, then the thread blocks. When all the threads in the VM block, you have reached the safepoint where all the the objects in the VM are perfectly reachable. Then, the VM operation that requested the safepoint is performed (e.g. a FullGC), after that the threads are resumed.
Check for the list of VM operations requiring a safepoint : Safety First: Safepoints.
You can study safepoint behavior in Hotspot by using -XX:+PrintSafepointStatistics –XX:PrintSafepointStatisticsCount=1
.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…