diff options
Diffstat (limited to '')
-rw-r--r-- | lib/base/namespace.hpp | 105 |
1 files changed, 105 insertions, 0 deletions
diff --git a/lib/base/namespace.hpp b/lib/base/namespace.hpp new file mode 100644 index 0000000..94f2055 --- /dev/null +++ b/lib/base/namespace.hpp @@ -0,0 +1,105 @@ +/* Icinga 2 | (c) 2012 Icinga GmbH | GPLv2+ */ + +#ifndef NAMESPACE_H +#define NAMESPACE_H + +#include "base/i2-base.hpp" +#include "base/object.hpp" +#include "base/shared-object.hpp" +#include "base/value.hpp" +#include "base/debuginfo.hpp" +#include <atomic> +#include <map> +#include <vector> +#include <memory> +#include <shared_mutex> + +namespace icinga +{ + +struct NamespaceValue +{ + Value Val; + bool Const; +}; + + +/** + * A namespace. + * + * ## External Locking + * + * Synchronization is handled internally, so almost all functions are safe for concurrent use without external locking. + * The only exception to this are functions returning an iterator. To use these, the caller has to acquire an ObjectLock + * on the namespace. The iterators only remain valid for as long as that ObjectLock is held. Note that this also + * includes range-based for loops. + * + * If consistency across multiple operations is required, an ObjectLock must also be acquired to prevent concurrent + * modifications. + * + * ## Internal Locking + * + * Two mutex objects are involved in locking a namespace: the recursive mutex inherited from the Object class that is + * acquired and released using the ObjectLock class and the m_DataMutex shared mutex contained directly in the + * Namespace class. The ObjectLock is used to synchronize multiple write operations against each other. The shared mutex + * is only used to ensure the consistency of the m_Data data structure. + * + * Read operations must acquire a shared lock on m_DataMutex. This prevents concurrent writes to that data structure + * but still allows concurrent reads. + * + * Write operations must first obtain an ObjectLock and then a shared lock on m_DataMutex. This order is important for + * preventing deadlocks. The ObjectLock prevents concurrent write operations while the shared lock prevents concurrent + * read operations. + * + * External read access to iterators is synchronized by the caller holding an ObjectLock. This ensures no concurrent + * write operations as these require the ObjectLock but still allows concurrent reads as m_DataMutex is not locked. + * + * @ingroup base + */ +class Namespace final : public Object +{ +public: + DECLARE_OBJECT(Namespace); + + typedef std::map<String, NamespaceValue>::iterator Iterator; + + typedef std::map<String, NamespaceValue>::value_type Pair; + + explicit Namespace(bool constValues = false); + + Value Get(const String& field) const; + bool Get(const String& field, Value *value) const; + void Set(const String& field, const Value& value, bool isConst = false, const DebugInfo& debugInfo = DebugInfo()); + bool Contains(const String& field) const; + void Remove(const String& field); + void Freeze(); + + Iterator Begin(); + Iterator End(); + + size_t GetLength() const; + + Value GetFieldByName(const String& field, bool sandboxed, const DebugInfo& debugInfo) const override; + void SetFieldByName(const String& field, const Value& value, bool overrideFrozen, const DebugInfo& debugInfo) override; + bool HasOwnField(const String& field) const override; + bool GetOwnField(const String& field, Value *result) const override; + + static Object::Ptr GetPrototype(); + +private: + std::shared_lock<std::shared_timed_mutex> ReadLockUnlessFrozen() const; + + std::map<String, NamespaceValue> m_Data; + mutable std::shared_timed_mutex m_DataMutex; + bool m_ConstValues; + std::atomic<bool> m_Frozen; +}; + +Namespace::Iterator begin(const Namespace::Ptr& x); +Namespace::Iterator end(const Namespace::Ptr& x); + +} + +extern template class std::map<icinga::String, std::shared_ptr<icinga::NamespaceValue> >; + +#endif /* NAMESPACE_H */ |