A cache for physical memory indexed by page number. More...

#include <memcache.h>

Inheritance diagram for MemCache:

Classes
struct	Buffers
	Cache MMIO registers and pending writes to them.
struct	CacheEntry
Public Types
enum	Type { TYPE_R = 1u << 0, TYPE_W = 1u << 1, TYPE_RMW = TYPE_R \| TYPE_W, TYPE_U = 1u << 2, TYPE_RES = 1u << 3, TYPE_X = 1u << 4 }
Public Member Functions
CacheEntry *	get (unsigned long phys1, unsigned long phys2, unsigned len, Type type)
	Get an entry from the cache or fetch one from memory.
void	invalidate (bool writeback)
	Invalidate the cache, thus writeback the buffers.
	MemCache (DBus< MessageMem > &mem, DBus< MessageMemRegion > &memregion)
Public Attributes
bool	debug
CacheEntry	_values [ASSOZ]
unsigned	_newest
Protected Attributes
DBus< MessageMem > &	_mem
DBus< MessageMemRegion > &	_memregion
unsigned	_fault
unsigned	_error_code
unsigned	_debug_fault_line
unsigned	_mtr_in
unsigned	_mtr_read
unsigned	_mtr_out

Detailed Description

A cache for physical memory indexed by page number.

Member Enumeration Documentation

enum MemCache::Type

Enumerator:

TYPE_R
TYPE_W
TYPE_RMW
TYPE_U
TYPE_RES
TYPE_X

Constructor & Destructor Documentation

MemCache::MemCache	(	DBus< MessageMem > &	mem,
		DBus< MessageMemRegion > &	memregion
	)			`[inline]`

Member Function Documentation

CacheEntry* MemCache::get	(	unsigned long	phys1,
		unsigned long	phys2,
		unsigned	len,
		Type	type
	)			`[inline]`

Get an entry from the cache or fetch one from memory.

What should we do if two different pages are referenced?

We could fallback to dword mode but there is this strange corner case where somebody does an locked operation crossing two non-adjunct pages, where we have to map the two pages into an 8k region and unmap them later on....

Invalidate a dirty entry, as we entry points to the last used buffer it must be the oldest write.

void MemCache::invalidate ( bool writeback ) [inline]

Invalidate the cache, thus writeback the buffers.