94843b1390
of page interlocks. Require that the page interlock be held over calls to uvm_pageactivate(), uvm_pagewire() and similar. - Solve the concurrency problem with page replacement state. Rather than updating the global state synchronously, set an intended state on individual pages (active, inactive, enqueued, dequeued) while holding the page interlock. After the interlock is released put the pages on a 128 entry per-CPU queue for their state changes to be made real in batch. This results in in a ~400 fold decrease in contention on my test system. Proposed on tech-kern but modified to use the page interlock rather than atomics to synchronise as it's much easier to maintain that way, and cheaper.
279 lines
6.8 KiB
C
279 lines
6.8 KiB
C
/* $NetBSD: uvm_object.c,v 1.19 2019/12/31 22:42:51 ad Exp $ */
|
|
|
|
/*
|
|
* Copyright (c) 2006, 2010, 2019 The NetBSD Foundation, Inc.
|
|
* All rights reserved.
|
|
*
|
|
* This code is derived from software contributed to The NetBSD Foundation
|
|
* by Mindaugas Rasiukevicius.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
|
|
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
|
|
* TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
|
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
|
|
* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
|
|
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
|
|
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
|
|
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
|
|
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
|
|
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
/*
|
|
* uvm_object.c: operate with memory objects
|
|
*
|
|
* TODO:
|
|
* 1. Support PG_RELEASED-using objects
|
|
*/
|
|
|
|
#include <sys/cdefs.h>
|
|
__KERNEL_RCSID(0, "$NetBSD: uvm_object.c,v 1.19 2019/12/31 22:42:51 ad Exp $");
|
|
|
|
#ifdef _KERNEL_OPT
|
|
#include "opt_ddb.h"
|
|
#endif
|
|
|
|
#include <sys/param.h>
|
|
#include <sys/mutex.h>
|
|
#include <sys/queue.h>
|
|
|
|
#include <uvm/uvm.h>
|
|
#include <uvm/uvm_ddb.h>
|
|
#include <uvm/uvm_page_array.h>
|
|
|
|
/* Page count to fetch per single step. */
|
|
#define FETCH_PAGECOUNT 16
|
|
|
|
/*
|
|
* uvm_obj_init: initialize UVM memory object.
|
|
*/
|
|
void
|
|
uvm_obj_init(struct uvm_object *uo, const struct uvm_pagerops *ops,
|
|
bool alock, u_int refs)
|
|
{
|
|
|
|
#if 0 /* notyet */
|
|
KASSERT(ops);
|
|
#endif
|
|
if (alock) {
|
|
/* Allocate and assign a lock. */
|
|
uo->vmobjlock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
|
|
} else {
|
|
/* The lock will need to be set via uvm_obj_setlock(). */
|
|
uo->vmobjlock = NULL;
|
|
}
|
|
uo->pgops = ops;
|
|
LIST_INIT(&uo->uo_ubc);
|
|
uo->uo_npages = 0;
|
|
uo->uo_refs = refs;
|
|
radix_tree_init_tree(&uo->uo_pages);
|
|
}
|
|
|
|
/*
|
|
* uvm_obj_destroy: destroy UVM memory object.
|
|
*/
|
|
void
|
|
uvm_obj_destroy(struct uvm_object *uo, bool dlock)
|
|
{
|
|
|
|
KASSERT(radix_tree_empty_tree_p(&uo->uo_pages));
|
|
|
|
/* Purge any UBC entries associated with this object. */
|
|
ubc_purge(uo);
|
|
|
|
/* Destroy the lock, if requested. */
|
|
if (dlock) {
|
|
mutex_obj_free(uo->vmobjlock);
|
|
}
|
|
radix_tree_fini_tree(&uo->uo_pages);
|
|
}
|
|
|
|
/*
|
|
* uvm_obj_setlock: assign a vmobjlock to the UVM object.
|
|
*
|
|
* => Caller is responsible to ensure that UVM objects is not use.
|
|
* => Only dynamic lock may be previously set. We drop the reference then.
|
|
*/
|
|
void
|
|
uvm_obj_setlock(struct uvm_object *uo, kmutex_t *lockptr)
|
|
{
|
|
kmutex_t *olockptr = uo->vmobjlock;
|
|
|
|
if (olockptr) {
|
|
/* Drop the reference on the old lock. */
|
|
mutex_obj_free(olockptr);
|
|
}
|
|
if (lockptr == NULL) {
|
|
/* If new lock is not passed - allocate default one. */
|
|
lockptr = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
|
|
}
|
|
uo->vmobjlock = lockptr;
|
|
}
|
|
|
|
/*
|
|
* uvm_obj_wirepages: wire the pages of entire UVM object.
|
|
*
|
|
* => NOTE: this function should only be used for types of objects
|
|
* where PG_RELEASED flag is never set (aobj objects)
|
|
* => caller must pass page-aligned start and end values
|
|
*/
|
|
int
|
|
uvm_obj_wirepages(struct uvm_object *uobj, off_t start, off_t end,
|
|
struct pglist *list)
|
|
{
|
|
int i, npages, error;
|
|
struct vm_page *pgs[FETCH_PAGECOUNT], *pg = NULL;
|
|
off_t offset = start, left;
|
|
|
|
left = (end - start) >> PAGE_SHIFT;
|
|
|
|
mutex_enter(uobj->vmobjlock);
|
|
while (left) {
|
|
|
|
npages = MIN(FETCH_PAGECOUNT, left);
|
|
|
|
/* Get the pages */
|
|
memset(pgs, 0, sizeof(pgs));
|
|
error = (*uobj->pgops->pgo_get)(uobj, offset, pgs, &npages, 0,
|
|
VM_PROT_READ | VM_PROT_WRITE, UVM_ADV_SEQUENTIAL,
|
|
PGO_ALLPAGES | PGO_SYNCIO);
|
|
|
|
if (error)
|
|
goto error;
|
|
|
|
mutex_enter(uobj->vmobjlock);
|
|
for (i = 0; i < npages; i++) {
|
|
|
|
KASSERT(pgs[i] != NULL);
|
|
KASSERT(!(pgs[i]->flags & PG_RELEASED));
|
|
|
|
/*
|
|
* Loan break
|
|
*/
|
|
if (pgs[i]->loan_count) {
|
|
while (pgs[i]->loan_count) {
|
|
pg = uvm_loanbreak(pgs[i]);
|
|
if (!pg) {
|
|
mutex_exit(uobj->vmobjlock);
|
|
uvm_wait("uobjwirepg");
|
|
mutex_enter(uobj->vmobjlock);
|
|
continue;
|
|
}
|
|
}
|
|
pgs[i] = pg;
|
|
}
|
|
|
|
if (pgs[i]->flags & PG_AOBJ) {
|
|
pgs[i]->flags &= ~(PG_CLEAN);
|
|
uao_dropswap(uobj, i);
|
|
}
|
|
}
|
|
|
|
/* Wire the pages */
|
|
for (i = 0; i < npages; i++) {
|
|
uvm_pagelock(pgs[i]);
|
|
uvm_pagewire(pgs[i]);
|
|
uvm_pageunlock(pgs[i]);
|
|
if (list != NULL)
|
|
TAILQ_INSERT_TAIL(list, pgs[i], pageq.queue);
|
|
}
|
|
|
|
/* Unbusy the pages */
|
|
uvm_page_unbusy(pgs, npages);
|
|
|
|
left -= npages;
|
|
offset += npages << PAGE_SHIFT;
|
|
}
|
|
mutex_exit(uobj->vmobjlock);
|
|
|
|
return 0;
|
|
|
|
error:
|
|
/* Unwire the pages which has been wired */
|
|
uvm_obj_unwirepages(uobj, start, offset);
|
|
|
|
return error;
|
|
}
|
|
|
|
/*
|
|
* uvm_obj_unwirepages: unwire the pages of entire UVM object.
|
|
*
|
|
* => NOTE: this function should only be used for types of objects
|
|
* where PG_RELEASED flag is never set
|
|
* => caller must pass page-aligned start and end values
|
|
*/
|
|
void
|
|
uvm_obj_unwirepages(struct uvm_object *uobj, off_t start, off_t end)
|
|
{
|
|
struct vm_page *pg;
|
|
off_t offset;
|
|
|
|
mutex_enter(uobj->vmobjlock);
|
|
for (offset = start; offset < end; offset += PAGE_SIZE) {
|
|
pg = uvm_pagelookup(uobj, offset);
|
|
|
|
KASSERT(pg != NULL);
|
|
KASSERT(!(pg->flags & PG_RELEASED));
|
|
|
|
uvm_pagelock(pg);
|
|
uvm_pageunwire(pg);
|
|
uvm_pageunlock(pg);
|
|
}
|
|
mutex_exit(uobj->vmobjlock);
|
|
}
|
|
|
|
#if defined(DDB) || defined(DEBUGPRINT)
|
|
|
|
/*
|
|
* uvm_object_printit: actually prints the object
|
|
*/
|
|
void
|
|
uvm_object_printit(struct uvm_object *uobj, bool full,
|
|
void (*pr)(const char *, ...))
|
|
{
|
|
struct uvm_page_array a;
|
|
struct vm_page *pg;
|
|
int cnt = 0;
|
|
voff_t off;
|
|
|
|
(*pr)("OBJECT %p: locked=%d, pgops=%p, npages=%d, ",
|
|
uobj, mutex_owned(uobj->vmobjlock), uobj->pgops, uobj->uo_npages);
|
|
if (UVM_OBJ_IS_KERN_OBJECT(uobj))
|
|
(*pr)("refs=<SYSTEM>\n");
|
|
else
|
|
(*pr)("refs=%d\n", uobj->uo_refs);
|
|
|
|
if (!full) {
|
|
return;
|
|
}
|
|
(*pr)(" PAGES <pg,offset>:\n ");
|
|
uvm_page_array_init(&a);
|
|
off = 0;
|
|
while ((pg = uvm_page_array_fill_and_peek(&a, uobj, off, 0, 0))
|
|
!= NULL) {
|
|
cnt++;
|
|
(*pr)("<%p,0x%llx> ", pg, (long long)pg->offset);
|
|
if ((cnt % 3) == 0) {
|
|
(*pr)("\n ");
|
|
}
|
|
off = pg->offset + PAGE_SIZE;
|
|
uvm_page_array_advance(&a);
|
|
}
|
|
if ((cnt % 3) != 0) {
|
|
(*pr)("\n");
|
|
}
|
|
uvm_page_array_fini(&a);
|
|
}
|
|
|
|
#endif /* DDB || DEBUGPRINT */
|