amd64/amd64/pmap.c

96  *	mappings at almost any time.  However, invalidations
97  *	of virtual-to-physical mappings must be done as
362     &pg_ps_enabled, 0, "Are large page mappings enabled?");
585  * restrict all mappings to m will wait for the current thread's DI
616 #define	PMAP_ENTER_NOREPLACE	0x2000000	/* Don't replace mappings. */
972 	 * memory, pmap_change_attr() will demote any 2MB or 1GB page mappings
1046 	 * addresses to superpage mappings.
1232     "Allow executable superpage mappings in EPT");
1357 	 * Are large page mappings enabled?
1424 SYSCTL_ULONG(_vm_pmap_pde, OID_AUTO, mappings, CTLFLAG_RD,
1425     &pmap_pde_mappings, 0, "2MB page mappings");
1560 		 * to make sure that any mappings using the old entry are
1607 		 * including any global (PG_G) mappings.
1636  * This will force the vcpu to exit and the cached EPT mappings
1649 	 * The TLB mappings associated with a vcpu context are not
1652 	 * This is in contrast with a process's vtop mappings that
1661 	 * invept will be done to flush stale mappings from the TLB.
2129 	 * by a promotion that did not invalidate the 512 4KB page mappings
2131 	 * may hold both 4KB and 2MB page mappings for the address range [va,
2134 	 * 4KB page mappings for the address range [va, va + NBPDR), and so a
2425  * kernel mappings that do not need to have
2427  * Note that old mappings are simply written
2457  * This routine tears out page mappings from the
2458  * kernel -- it is meant only for temporary mappings.
2712 	 * Do not install the host kernel mappings in the nested page
2713 	 * tables. These mappings are meaningless in the guest physical
2715 	 * Install minimal kernel mappings in PTI case.
2831 			 * PTI: Make all user-space mappings in the
3017  * Should only be called if the map contains no valid mappings.
3653  * 2MB page mappings.
3671  * After demotion from a 2MB page mapping to 512 4KB page mappings,
3673  * entries for each of the 4KB page mappings.
3740  * After promotion from 512 4KB page mappings to a single 2MB page mapping,
3741  * replace the many pv entries for the 4KB page mappings by a single pv entry
3762 	 * mappings that is being promoted.
3784  * page mappings.
3846  * Fills a page table page with mappings to consecutive physical pages.
4281 		 * Weed out invalid mappings.
4467 		 * lingering 4KB page mappings from the TLB.
4514 	 * ordinarily destroy mappings.  Stale TLB entries from
4518 	 * a physical page's PV list of mappings, like
4556 		 * Weed out invalid mappings.
4635  * Tries to promote the 512, contiguous 4KB page mappings that are within a
4637  * to occur, two conditions must be met: (1) the 4KB page mappings must map
4638  * aligned, contiguous physical memory and (2) the 4KB page mappings must have
4833 	 * Set modified bit gratuitously for writeable mappings if
4835 	 * to do the dirty bit accounting for these mappings.
4893 		 * are valid mappings in them. Hence, if a user page is wired,
4935 		 * For consistency, handle unmanaged mappings the same way.
5189 	 * be any lingering 4KB page mappings in the TLB.)
5456 				 * The creation of mappings below is only an
5486  *	Clear the wired attribute from the mappings for the specified range of
5488  *	must have the wired attribute set.  In contrast, invalid mappings
5600 	 * mappings we skip this function entirely for emulated pmaps.
5881  *	Return the number of managed mappings to the given physical page
5965  * Destroy all managed, non-wired mappings in the given user-space
5973  * destroy mappings by iterating over the pmap's collection of PV
5981  * non-wired mappings, it can delay and batch the invalidation of TLB
6315  * Clear the write and modified bits in each of the given page's mappings.
6726 			 * Unless the page mappings are wired, remove the
6962 			panic("%s: too many preinit mappings", __func__);
7152  * behavior of their processors as "undefined" if two or more mappings to the
7670  *	different alignment might result in more superpage mappings.
7905 				 * while mappings are active don't pin the