2009-07-24 13:50:29 +00:00
|
|
|
/*-
|
2017-11-27 15:23:17 +00:00
|
|
|
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
|
|
|
|
*
|
2015-04-23 14:22:20 +00:00
|
|
|
* Copyright (c) 2009 Hudson River Trading LLC
|
2009-07-24 13:50:29 +00:00
|
|
|
* Written by: John H. Baldwin <jhb@FreeBSD.org>
|
|
|
|
* All rights reserved.
|
|
|
|
*
|
|
|
|
* 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 AUTHOR 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 AUTHOR 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.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <sys/cdefs.h>
|
|
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This pager manages OBJT_SG objects. These objects are backed by
|
|
|
|
* a scatter/gather list of physical address ranges.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <sys/param.h>
|
|
|
|
#include <sys/lock.h>
|
|
|
|
#include <sys/mutex.h>
|
2013-02-20 10:38:34 +00:00
|
|
|
#include <sys/rwlock.h>
|
2009-07-24 13:50:29 +00:00
|
|
|
#include <sys/sglist.h>
|
2017-04-17 17:07:00 +00:00
|
|
|
#include <sys/vmmeter.h>
|
|
|
|
|
2009-07-24 13:50:29 +00:00
|
|
|
#include <vm/vm.h>
|
2012-08-05 14:11:42 +00:00
|
|
|
#include <vm/vm_param.h>
|
2009-07-24 13:50:29 +00:00
|
|
|
#include <vm/vm_object.h>
|
|
|
|
#include <vm/vm_page.h>
|
|
|
|
#include <vm/vm_pager.h>
|
2012-11-16 05:55:56 +00:00
|
|
|
#include <vm/vm_phys.h>
|
2009-07-24 13:50:29 +00:00
|
|
|
#include <vm/uma.h>
|
|
|
|
|
|
|
|
static vm_object_t sg_pager_alloc(void *, vm_ooffset_t, vm_prot_t,
|
|
|
|
vm_ooffset_t, struct ucred *);
|
|
|
|
static void sg_pager_dealloc(vm_object_t);
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
static int sg_pager_getpages(vm_object_t, vm_page_t *, int, int *, int *);
|
2009-07-24 13:50:29 +00:00
|
|
|
static void sg_pager_putpages(vm_object_t, vm_page_t *, int,
|
|
|
|
boolean_t, int *);
|
|
|
|
static boolean_t sg_pager_haspage(vm_object_t, vm_pindex_t, int *,
|
|
|
|
int *);
|
|
|
|
|
|
|
|
struct pagerops sgpagerops = {
|
|
|
|
.pgo_alloc = sg_pager_alloc,
|
|
|
|
.pgo_dealloc = sg_pager_dealloc,
|
|
|
|
.pgo_getpages = sg_pager_getpages,
|
|
|
|
.pgo_putpages = sg_pager_putpages,
|
|
|
|
.pgo_haspage = sg_pager_haspage,
|
|
|
|
};
|
|
|
|
|
|
|
|
static vm_object_t
|
|
|
|
sg_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot,
|
|
|
|
vm_ooffset_t foff, struct ucred *cred)
|
|
|
|
{
|
|
|
|
struct sglist *sg;
|
|
|
|
vm_object_t object;
|
|
|
|
vm_pindex_t npages, pindex;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Offset should be page aligned.
|
|
|
|
*/
|
|
|
|
if (foff & PAGE_MASK)
|
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The scatter/gather list must only include page-aligned
|
|
|
|
* ranges.
|
|
|
|
*/
|
|
|
|
npages = 0;
|
|
|
|
sg = handle;
|
|
|
|
for (i = 0; i < sg->sg_nseg; i++) {
|
|
|
|
if ((sg->sg_segs[i].ss_paddr % PAGE_SIZE) != 0 ||
|
|
|
|
(sg->sg_segs[i].ss_len % PAGE_SIZE) != 0)
|
|
|
|
return (NULL);
|
|
|
|
npages += sg->sg_segs[i].ss_len / PAGE_SIZE;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The scatter/gather list has a fixed size. Refuse requests
|
|
|
|
* to map beyond that.
|
|
|
|
*/
|
|
|
|
size = round_page(size);
|
2018-12-02 13:16:46 +00:00
|
|
|
pindex = OFF_TO_IDX(foff) + OFF_TO_IDX(size);
|
|
|
|
if (pindex > npages || pindex < OFF_TO_IDX(foff) ||
|
|
|
|
pindex < OFF_TO_IDX(size))
|
2009-07-24 13:50:29 +00:00
|
|
|
return (NULL);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Allocate a new object and associate it with the
|
|
|
|
* scatter/gather list. It is ok for our purposes to have
|
|
|
|
* multiple VM objects associated with the same scatter/gather
|
|
|
|
* list because scatter/gather lists are static. This is also
|
|
|
|
* simpler than ensuring a unique object per scatter/gather
|
|
|
|
* list.
|
|
|
|
*/
|
|
|
|
object = vm_object_allocate(OBJT_SG, npages);
|
|
|
|
object->handle = sglist_hold(sg);
|
|
|
|
TAILQ_INIT(&object->un_pager.sgp.sgp_pglist);
|
|
|
|
return (object);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
sg_pager_dealloc(vm_object_t object)
|
|
|
|
{
|
|
|
|
struct sglist *sg;
|
|
|
|
vm_page_t m;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Free up our fake pages.
|
|
|
|
*/
|
|
|
|
while ((m = TAILQ_FIRST(&object->un_pager.sgp.sgp_pglist)) != 0) {
|
2013-08-10 17:36:42 +00:00
|
|
|
TAILQ_REMOVE(&object->un_pager.sgp.sgp_pglist, m, plinks.q);
|
2011-03-11 07:07:48 +00:00
|
|
|
vm_page_putfake(m);
|
2009-07-24 13:50:29 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
sg = object->handle;
|
|
|
|
sglist_free(sg);
|
2015-05-08 19:43:37 +00:00
|
|
|
object->handle = NULL;
|
|
|
|
object->type = OBJT_DEAD;
|
2009-07-24 13:50:29 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
sg_pager_getpages(vm_object_t object, vm_page_t *m, int count, int *rbehind,
|
|
|
|
int *rahead)
|
2009-07-24 13:50:29 +00:00
|
|
|
{
|
|
|
|
struct sglist *sg;
|
|
|
|
vm_page_t m_paddr, page;
|
|
|
|
vm_pindex_t offset;
|
|
|
|
vm_paddr_t paddr;
|
|
|
|
vm_memattr_t memattr;
|
|
|
|
size_t space;
|
|
|
|
int i;
|
|
|
|
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
/* Since our haspage reports zero after/before, the count is 1. */
|
|
|
|
KASSERT(count == 1, ("%s: count %d", __func__, count));
|
2013-02-21 21:54:53 +00:00
|
|
|
VM_OBJECT_ASSERT_WLOCKED(object);
|
2009-07-24 13:50:29 +00:00
|
|
|
sg = object->handle;
|
|
|
|
memattr = object->memattr;
|
2013-02-20 12:03:20 +00:00
|
|
|
VM_OBJECT_WUNLOCK(object);
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
offset = m[0]->pindex;
|
2009-07-24 13:50:29 +00:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Lookup the physical address of the requested page. An initial
|
|
|
|
* value of '1' instead of '0' is used so we can assert that the
|
|
|
|
* page is found since '0' can be a valid page-aligned physical
|
|
|
|
* address.
|
|
|
|
*/
|
|
|
|
space = 0;
|
|
|
|
paddr = 1;
|
|
|
|
for (i = 0; i < sg->sg_nseg; i++) {
|
|
|
|
if (space + sg->sg_segs[i].ss_len <= (offset * PAGE_SIZE)) {
|
|
|
|
space += sg->sg_segs[i].ss_len;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
paddr = sg->sg_segs[i].ss_paddr + offset * PAGE_SIZE - space;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
KASSERT(paddr != 1, ("invalid SG page index"));
|
|
|
|
|
|
|
|
/* If "paddr" is a real page, perform a sanity check on "memattr". */
|
|
|
|
if ((m_paddr = vm_phys_paddr_to_vm_page(paddr)) != NULL &&
|
|
|
|
pmap_page_get_memattr(m_paddr) != memattr) {
|
|
|
|
memattr = pmap_page_get_memattr(m_paddr);
|
|
|
|
printf(
|
|
|
|
"WARNING: A device driver has set \"memattr\" inconsistently.\n");
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Return a fake page for the requested page. */
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
KASSERT(!(m[0]->flags & PG_FICTITIOUS),
|
2009-07-24 13:50:29 +00:00
|
|
|
("backing page for SG is fake"));
|
|
|
|
|
|
|
|
/* Construct a new fake page. */
|
2011-03-11 07:07:48 +00:00
|
|
|
page = vm_page_getfake(paddr, memattr);
|
2013-02-20 12:03:20 +00:00
|
|
|
VM_OBJECT_WLOCK(object);
|
2013-08-10 17:36:42 +00:00
|
|
|
TAILQ_INSERT_TAIL(&object->un_pager.sgp.sgp_pglist, page, plinks.q);
|
2015-12-17 17:48:57 +00:00
|
|
|
vm_page_replace_checked(page, object, offset, m[0]);
|
2016-02-05 19:35:53 +00:00
|
|
|
vm_page_lock(m[0]);
|
|
|
|
vm_page_free(m[0]);
|
|
|
|
vm_page_unlock(m[0]);
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
m[0] = page;
|
2019-10-15 03:45:41 +00:00
|
|
|
vm_page_valid(page);
|
2009-07-24 13:50:29 +00:00
|
|
|
|
A change to KPI of vm_pager_get_pages() and underlying VOP_GETPAGES().
o With new KPI consumers can request contiguous ranges of pages, and
unlike before, all pages will be kept busied on return, like it was
done before with the 'reqpage' only. Now the reqpage goes away. With
new interface it is easier to implement code protected from race
conditions.
Such arrayed requests for now should be preceeded by a call to
vm_pager_haspage() to make sure that request is possible. This
could be improved later, making vm_pager_haspage() obsolete.
Strenghtening the promises on the business of the array of pages
allows us to remove such hacks as swp_pager_free_nrpage() and
vm_pager_free_nonreq().
o New KPI accepts two integer pointers that may optionally point at
values for read ahead and read behind, that a pager may do, if it
can. These pages are completely owned by pager, and not controlled
by the caller.
This shifts the UFS-specific readahead logic from vm_fault.c, which
should be file system agnostic, into vnode_pager.c. It also removes
one VOP_BMAP() request per hard fault.
Discussed with: kib, alc, jeff, scottl
Sponsored by: Nginx, Inc.
Sponsored by: Netflix
2015-12-16 21:30:45 +00:00
|
|
|
if (rbehind)
|
|
|
|
*rbehind = 0;
|
|
|
|
if (rahead)
|
|
|
|
*rahead = 0;
|
|
|
|
|
2009-07-24 13:50:29 +00:00
|
|
|
return (VM_PAGER_OK);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
sg_pager_putpages(vm_object_t object, vm_page_t *m, int count,
|
|
|
|
boolean_t sync, int *rtvals)
|
|
|
|
{
|
|
|
|
|
|
|
|
panic("sg_pager_putpage called");
|
|
|
|
}
|
|
|
|
|
|
|
|
static boolean_t
|
|
|
|
sg_pager_haspage(vm_object_t object, vm_pindex_t pindex, int *before,
|
|
|
|
int *after)
|
|
|
|
{
|
|
|
|
|
|
|
|
if (before != NULL)
|
|
|
|
*before = 0;
|
|
|
|
if (after != NULL)
|
|
|
|
*after = 0;
|
|
|
|
return (TRUE);
|
|
|
|
}
|