freebsd-dev/sys/arm/nvidia/drm2/tegra_bo.c
Mark Johnston fee2a2fa39 Change synchonization rules for vm_page reference counting.
There are several mechanisms by which a vm_page reference is held,
preventing the page from being freed back to the page allocator.  In
particular, holding the page's object lock is sufficient to prevent the
page from being freed; holding the busy lock or a wiring is sufficent as
well.  These references are protected by the page lock, which must
therefore be acquired for many per-page operations.  This results in
false sharing since the page locks are external to the vm_page
structures themselves and each lock protects multiple structures.

Transition to using an atomically updated per-page reference counter.
The object's reference is counted using a flag bit in the counter.  A
second flag bit is used to atomically block new references via
pmap_extract_and_hold() while removing managed mappings of a page.
Thus, the reference count of a page is guaranteed not to increase if the
page is unbusied, unmapped, and the object's write lock is held.  As
a consequence of this, the page lock no longer protects a page's
identity; operations which move pages between objects are now
synchronized solely by the objects' locks.

The vm_page_wire() and vm_page_unwire() KPIs are changed.  The former
requires that either the object lock or the busy lock is held.  The
latter no longer has a return value and may free the page if it releases
the last reference to that page.  vm_page_unwire_noq() behaves the same
as before; the caller is responsible for checking its return value and
freeing or enqueuing the page as appropriate.  vm_page_wire_mapped() is
introduced for use in pmap_extract_and_hold().  It fails if the page is
concurrently being unmapped, typically triggering a fallback to the
fault handler.  vm_page_wire() no longer requires the page lock and
vm_page_unwire() now internally acquires the page lock when releasing
the last wiring of a page (since the page lock still protects a page's
queue state).  In particular, synchronization details are no longer
leaked into the caller.

The change excises the page lock from several frequently executed code
paths.  In particular, vm_object_terminate() no longer bounces between
page locks as it releases an object's pages, and direct I/O and
sendfile(SF_NOCACHE) completions no longer require the page lock.  In
these latter cases we now get linear scalability in the common scenario
where different threads are operating on different files.

__FreeBSD_version is bumped.  The DRM ports have been updated to
accomodate the KPI changes.

Reviewed by:	jeff (earlier version)
Tested by:	gallatin (earlier version), pho
Sponsored by:	Netflix
Differential Revision:	https://reviews.freebsd.org/D20486
2019-09-09 21:32:42 +00:00

368 lines
8.7 KiB
C

/*-
* Copyright (c) 2015 Michal Meloun
* 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("$FreeBSD$");
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bus.h>
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <machine/bus.h>
#include <dev/extres/clk/clk.h>
#include <dev/drm2/drmP.h>
#include <dev/drm2/drm_crtc_helper.h>
#include <dev/drm2/drm_fb_helper.h>
#include <arm/nvidia/drm2/tegra_drm.h>
#include <sys/vmem.h>
#include <sys/vmem.h>
#include <vm/vm.h>
#include <vm/vm_pageout.h>
static void
tegra_bo_destruct(struct tegra_bo *bo)
{
vm_page_t m;
size_t size;
int i;
if (bo->cdev_pager == NULL)
return;
size = round_page(bo->gem_obj.size);
if (bo->vbase != 0)
pmap_qremove(bo->vbase, bo->npages);
VM_OBJECT_WLOCK(bo->cdev_pager);
for (i = 0; i < bo->npages; i++) {
m = bo->m[i];
cdev_pager_free_page(bo->cdev_pager, m);
m->flags &= ~PG_FICTITIOUS;
vm_page_unwire_noq(m);
vm_page_free(m);
}
VM_OBJECT_WUNLOCK(bo->cdev_pager);
vm_object_deallocate(bo->cdev_pager);
if (bo->vbase != 0)
vmem_free(kmem_arena, bo->vbase, size);
}
static void
tegra_bo_free_object(struct drm_gem_object *gem_obj)
{
struct tegra_bo *bo;
bo = container_of(gem_obj, struct tegra_bo, gem_obj);
drm_gem_free_mmap_offset(gem_obj);
drm_gem_object_release(gem_obj);
tegra_bo_destruct(bo);
free(bo->m, DRM_MEM_DRIVER);
free(bo, DRM_MEM_DRIVER);
}
static int
tegra_bo_alloc_contig(size_t npages, u_long alignment, vm_memattr_t memattr,
vm_page_t **ret_page)
{
vm_page_t m;
int pflags, tries, i;
vm_paddr_t low, high, boundary;
low = 0;
high = -1UL;
boundary = 0;
pflags = VM_ALLOC_NORMAL | VM_ALLOC_NOOBJ | VM_ALLOC_NOBUSY |
VM_ALLOC_WIRED | VM_ALLOC_ZERO;
tries = 0;
retry:
m = vm_page_alloc_contig(NULL, 0, pflags, npages, low, high, alignment,
boundary, memattr);
if (m == NULL) {
if (tries < 3) {
if (!vm_page_reclaim_contig(pflags, npages, low, high,
alignment, boundary))
vm_wait(NULL);
tries++;
goto retry;
}
return (ENOMEM);
}
for (i = 0; i < npages; i++, m++) {
if ((m->flags & PG_ZERO) == 0)
pmap_zero_page(m);
m->valid = VM_PAGE_BITS_ALL;
(*ret_page)[i] = m;
}
return (0);
}
/* Initialize pager and insert all object pages to it*/
static int
tegra_bo_init_pager(struct tegra_bo *bo)
{
vm_page_t m;
size_t size;
int i;
size = round_page(bo->gem_obj.size);
bo->pbase = VM_PAGE_TO_PHYS(bo->m[0]);
if (vmem_alloc(kmem_arena, size, M_WAITOK | M_BESTFIT, &bo->vbase))
return (ENOMEM);
VM_OBJECT_WLOCK(bo->cdev_pager);
for (i = 0; i < bo->npages; i++) {
m = bo->m[i];
/*
* XXX This is a temporary hack.
* We need pager suitable for paging (mmap) managed
* real (non-fictitious) pages.
* - managed pages are needed for clean module unload.
* - aliasing fictitious page to real one is bad,
* pmap cannot handle this situation without issues
* It expects that
* paddr = PHYS_TO_VM_PAGE(VM_PAGE_TO_PHYS(paddr))
* for every single page passed to pmap.
*/
m->oflags &= ~VPO_UNMANAGED;
m->flags |= PG_FICTITIOUS;
if (vm_page_insert(m, bo->cdev_pager, i) != 0)
return (EINVAL);
}
VM_OBJECT_WUNLOCK(bo->cdev_pager);
pmap_qenter(bo->vbase, bo->m, bo->npages);
return (0);
}
/* Allocate memory for frame buffer */
static int
tegra_bo_alloc(struct drm_device *drm, struct tegra_bo *bo)
{
size_t size;
int rv;
size = bo->gem_obj.size;
bo->npages = atop(size);
bo->m = malloc(sizeof(vm_page_t *) * bo->npages, DRM_MEM_DRIVER,
M_WAITOK | M_ZERO);
rv = tegra_bo_alloc_contig(bo->npages, PAGE_SIZE,
VM_MEMATTR_WRITE_COMBINING, &(bo->m));
if (rv != 0) {
DRM_WARNING("Cannot allocate memory for gem object.\n");
return (rv);
}
rv = tegra_bo_init_pager(bo);
if (rv != 0) {
DRM_WARNING("Cannot initialize gem object pager.\n");
return (rv);
}
return (0);
}
int
tegra_bo_create(struct drm_device *drm, size_t size, struct tegra_bo **res_bo)
{
struct tegra_bo *bo;
int rv;
if (size <= 0)
return (-EINVAL);
bo = malloc(sizeof(*bo), DRM_MEM_DRIVER, M_WAITOK | M_ZERO);
size = round_page(size);
rv = drm_gem_object_init(drm, &bo->gem_obj, size);
if (rv != 0) {
free(bo, DRM_MEM_DRIVER);
return (rv);
}
rv = drm_gem_create_mmap_offset(&bo->gem_obj);
if (rv != 0) {
drm_gem_object_release(&bo->gem_obj);
free(bo, DRM_MEM_DRIVER);
return (rv);
}
bo->cdev_pager = cdev_pager_allocate(&bo->gem_obj, OBJT_MGTDEVICE,
drm->driver->gem_pager_ops, size, 0, 0, NULL);
rv = tegra_bo_alloc(drm, bo);
if (rv != 0) {
tegra_bo_free_object(&bo->gem_obj);
return (rv);
}
*res_bo = bo;
return (0);
}
static int
tegra_bo_create_with_handle(struct drm_file *file, struct drm_device *drm,
size_t size, uint32_t *handle, struct tegra_bo **res_bo)
{
int rv;
struct tegra_bo *bo;
rv = tegra_bo_create(drm, size, &bo);
if (rv != 0)
return (rv);
rv = drm_gem_handle_create(file, &bo->gem_obj, handle);
if (rv != 0) {
tegra_bo_free_object(&bo->gem_obj);
drm_gem_object_release(&bo->gem_obj);
return (rv);
}
drm_gem_object_unreference_unlocked(&bo->gem_obj);
*res_bo = bo;
return (0);
}
static int
tegra_bo_dumb_create(struct drm_file *file, struct drm_device *drm_dev,
struct drm_mode_create_dumb *args)
{
struct tegra_drm *drm;
struct tegra_bo *bo;
int rv;
drm = container_of(drm_dev, struct tegra_drm, drm_dev);
args->pitch= (args->width * args->bpp + 7) / 8;
args->pitch = roundup(args->pitch, drm->pitch_align);
args->size = args->pitch * args->height;
rv = tegra_bo_create_with_handle(file, drm_dev, args->size,
&args->handle, &bo);
return (rv);
}
static int
tegra_bo_dumb_map_offset(struct drm_file *file_priv,
struct drm_device *drm_dev, uint32_t handle, uint64_t *offset)
{
struct drm_gem_object *gem_obj;
int rv;
DRM_LOCK(drm_dev);
gem_obj = drm_gem_object_lookup(drm_dev, file_priv, handle);
if (gem_obj == NULL) {
device_printf(drm_dev->dev, "Object not found\n");
DRM_UNLOCK(drm_dev);
return (-EINVAL);
}
rv = drm_gem_create_mmap_offset(gem_obj);
if (rv != 0)
goto fail;
*offset = DRM_GEM_MAPPING_OFF(gem_obj->map_list.key) |
DRM_GEM_MAPPING_KEY;
drm_gem_object_unreference(gem_obj);
DRM_UNLOCK(drm_dev);
return (0);
fail:
drm_gem_object_unreference(gem_obj);
DRM_UNLOCK(drm_dev);
return (rv);
}
static int
tegra_bo_dumb_destroy(struct drm_file *file_priv, struct drm_device *drm_dev,
unsigned int handle)
{
int rv;
rv = drm_gem_handle_delete(file_priv, handle);
return (rv);
}
/*
* mmap support
*/
static int
tegra_gem_pager_fault(vm_object_t vm_obj, vm_ooffset_t offset, int prot,
vm_page_t *mres)
{
#ifdef DRM_PAGER_DEBUG
DRM_DEBUG("object %p offset %jd prot %d mres %p\n",
vm_obj, (intmax_t)offset, prot, mres);
#endif
return (VM_PAGER_FAIL);
}
static int
tegra_gem_pager_ctor(void *handle, vm_ooffset_t size, vm_prot_t prot,
vm_ooffset_t foff, struct ucred *cred, u_short *color)
{
if (color != NULL)
*color = 0;
return (0);
}
static void
tegra_gem_pager_dtor(void *handle)
{
}
static struct cdev_pager_ops tegra_gem_pager_ops = {
.cdev_pg_fault = tegra_gem_pager_fault,
.cdev_pg_ctor = tegra_gem_pager_ctor,
.cdev_pg_dtor = tegra_gem_pager_dtor
};
/* Fill up relevant fields in drm_driver ops */
void
tegra_bo_driver_register(struct drm_driver *drm_drv)
{
drm_drv->gem_free_object = tegra_bo_free_object;
drm_drv->gem_pager_ops = &tegra_gem_pager_ops;
drm_drv->dumb_create = tegra_bo_dumb_create;
drm_drv->dumb_map_offset = tegra_bo_dumb_map_offset;
drm_drv->dumb_destroy = tegra_bo_dumb_destroy;
}