Aren/NetBSD
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

- pmap_extract(): This needs to take the pmap's lock, to allow for

Browse Source 
concurrent removal of pages (a new requirement).

- pmap_remove_pv(): Keep hold time of pp_lock as short as possible.

- pmap_get_ptp(): Don't re-init struct pmap_page for PD PTPs.  Would
  have no ill effects but is wrong regardless.
This commit is contained in:
ad 2020-03-20 19:06:14 +00:00
parent 067e9296fc
commit 1ffecadffe
1 changed files with 13 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

45
sys/arch/x86/x86/pmap.c
View File

@ -1,4 +1,4 @@
/* $NetBSD: pmap.c,v 1.378 2020/03/19 18:58:14 ad Exp $ */
/* $NetBSD: pmap.c,v 1.379 2020/03/20 19:06:14 ad Exp $ */
/*
* Copyright (c) 2008, 2010, 2016, 2017, 2019, 2020 The NetBSD Foundation, Inc.
@ -130,7 +130,7 @@
*/
#include <sys/cdefs.h>
__KERNEL_RCSID(0, "$NetBSD: pmap.c,v 1.378 2020/03/19 18:58:14 ad Exp $");
__KERNEL_RCSID(0, "$NetBSD: pmap.c,v 1.379 2020/03/20 19:06:14 ad Exp $");
#include "opt_user_ldt.h"
#include "opt_lockdebug.h"
@ -2163,20 +2163,22 @@ pmap_remove_pv(struct pmap *pmap, struct pmap_page *pp, struct vm_page *ptp,
pmap_check_pv(pmap, ptp, pp, va, true);
mutex_spin_enter(&pp->pp_lock);
pp->pp_attrs |= oattrs;
if (pve == NULL) {
mutex_spin_enter(&pp->pp_lock);
KASSERT(pp->pp_pte.pte_ptp == ptp);
KASSERT(pp->pp_pte.pte_va == va);
pp->pp_attrs |= oattrs;
pp->pp_pte.pte_ptp = NULL;
pp->pp_pte.pte_va = 0;
mutex_spin_exit(&pp->pp_lock);
} else {
mutex_spin_enter(&pp->pp_lock);
KASSERT(pp->pp_pte.pte_ptp != ptp ||
pp->pp_pte.pte_va != va);
KASSERT(pve->pve_pte.pte_ptp == ptp);
KASSERT(pve->pve_pte.pte_va == va);
KASSERT(pve->pve_pp == pp);
pp->pp_attrs |= oattrs;
LIST_REMOVE(pve, pve_list);
mutex_spin_exit(&pp->pp_lock);
@ -2347,7 +2349,7 @@ pmap_get_ptp(struct pmap *pmap, struct pmap_ptparray *pt, vaddr_t va,
if (pt->pg[i] == NULL) {
pmap_unget_ptp(pmap, pt);
return ENOMEM;
} else {
} else if (pt->alloced[i]) {
pt->pg[i]->uanon = (struct vm_anon *)(vaddr_t)~0L;
rb_tree_init(&VM_PAGE_TO_PP(pt->pg[i])->pp_rb,
&pmap_rbtree_ops);
@ -3427,10 +3429,8 @@ pmap_extract(struct pmap *pmap, vaddr_t va, paddr_t *pap)
pd_entry_t pde;
pd_entry_t * const *pdes;
struct pmap *pmap2;
struct cpu_info *ci;
paddr_t pa;
lwp_t *l;
bool hard, rv;
bool rv;
int lvl;
if (__predict_false(pmap->pm_extract != NULL)) {
@ -3448,29 +3448,11 @@ pmap_extract(struct pmap *pmap, vaddr_t va, paddr_t *pap)
rv = false;
pa = 0;
l = curlwp;
ci = l->l_cpu;
if (pmap == pmap_kernel() ||
__predict_true(!ci->ci_want_pmapload && ci->ci_pmap == pmap)) {
/*
* no need to lock, because it's pmap_kernel() or our
* own pmap and is active. if a user pmap, the caller
* will hold the vm_map write/read locked and so prevent
* entries from disappearing while we are here. ptps
* can disappear via pmap_remove() and pmap_protect(),
* but they are called with the vm_map write locked.
*/
hard = false;
ptes = PTE_BASE;
pdes = normal_pdes;
kpreempt_disable();
} else {
/* we lose, do it the hard way. */
hard = true;
if (pmap != pmap_kernel()) {
mutex_enter(&pmap->pm_lock);
pmap_map_ptes(pmap, &pmap2, &ptes, &pdes);
}
pmap_map_ptes(pmap, &pmap2, &ptes, &pdes);
if (pmap_pdes_valid(va, pdes, &pde, &lvl)) {
if (lvl == 2) {
pa = (pde & PTE_LGFRAME) | (va & (NBPD_L2 - 1));
@ -3484,15 +3466,14 @@ pmap_extract(struct pmap *pmap, vaddr_t va, paddr_t *pap)
}
}
}
if (__predict_false(hard)) {
pmap_unmap_ptes(pmap, pmap2);
pmap_unmap_ptes(pmap, pmap2);
if (pmap != pmap_kernel()) {
mutex_exit(&pmap->pm_lock);
} else {
kpreempt_enable();
}
if (pap != NULL) {
*pap = pa;
}
return rv;
}