coreboot-kgpe-d16/src/commonlib/region.c

530 lines
12 KiB
C

/* SPDX-License-Identifier: GPL-2.0-only */
#include <commonlib/helpers.h>
#include <commonlib/region.h>
#include <string.h>
int region_is_subregion(const struct region *p, const struct region *c)
{
if (region_offset(c) < region_offset(p))
return 0;
if (region_end(c) > region_end(p))
return 0;
if (region_end(c) < region_offset(c))
return 0;
return 1;
}
static int normalize_and_ok(const struct region *outer, struct region *inner)
{
inner->offset += region_offset(outer);
return region_is_subregion(outer, inner);
}
static const struct region_device *rdev_root(const struct region_device *rdev)
{
if (rdev->root == NULL)
return rdev;
return rdev->root;
}
ssize_t rdev_relative_offset(const struct region_device *p,
const struct region_device *c)
{
if (rdev_root(p) != rdev_root(c))
return -1;
if (!region_is_subregion(&p->region, &c->region))
return -1;
return region_device_offset(c) - region_device_offset(p);
}
void *rdev_mmap(const struct region_device *rd, size_t offset, size_t size)
{
const struct region_device *rdev;
struct region req = {
.offset = offset,
.size = size,
};
if (!normalize_and_ok(&rd->region, &req))
return NULL;
rdev = rdev_root(rd);
if (rdev->ops->mmap == NULL)
return NULL;
return rdev->ops->mmap(rdev, req.offset, req.size);
}
int rdev_munmap(const struct region_device *rd, void *mapping)
{
const struct region_device *rdev;
rdev = rdev_root(rd);
if (rdev->ops->munmap == NULL)
return -1;
return rdev->ops->munmap(rdev, mapping);
}
ssize_t rdev_readat(const struct region_device *rd, void *b, size_t offset,
size_t size)
{
const struct region_device *rdev;
struct region req = {
.offset = offset,
.size = size,
};
if (!normalize_and_ok(&rd->region, &req))
return -1;
rdev = rdev_root(rd);
return rdev->ops->readat(rdev, b, req.offset, req.size);
}
ssize_t rdev_writeat(const struct region_device *rd, const void *b,
size_t offset, size_t size)
{
const struct region_device *rdev;
struct region req = {
.offset = offset,
.size = size,
};
if (!normalize_and_ok(&rd->region, &req))
return -1;
rdev = rdev_root(rd);
if (rdev->ops->writeat == NULL)
return -1;
return rdev->ops->writeat(rdev, b, req.offset, req.size);
}
ssize_t rdev_eraseat(const struct region_device *rd, size_t offset,
size_t size)
{
const struct region_device *rdev;
struct region req = {
.offset = offset,
.size = size,
};
if (!normalize_and_ok(&rd->region, &req))
return -1;
rdev = rdev_root(rd);
/* If the eraseat ptr is NULL we assume that the erase
* function was completed successfully. */
if (rdev->ops->eraseat == NULL)
return size;
return rdev->ops->eraseat(rdev, req.offset, req.size);
}
int rdev_chain(struct region_device *child, const struct region_device *parent,
size_t offset, size_t size)
{
struct region req = {
.offset = offset,
.size = size,
};
if (!normalize_and_ok(&parent->region, &req))
return -1;
/* Keep track of root region device. Note the offsets are relative
* to the root device. */
child->root = rdev_root(parent);
child->ops = NULL;
child->region.offset = req.offset;
child->region.size = req.size;
return 0;
}
static void mem_region_device_init(struct mem_region_device *mdev,
const struct region_device_ops *ops, void *base, size_t size)
{
memset(mdev, 0, sizeof(*mdev));
mdev->base = base;
mdev->rdev.ops = ops;
mdev->rdev.region.size = size;
}
void mem_region_device_ro_init(struct mem_region_device *mdev, void *base,
size_t size)
{
return mem_region_device_init(mdev, &mem_rdev_ro_ops, base, size);
}
void mem_region_device_rw_init(struct mem_region_device *mdev, void *base,
size_t size)
{
return mem_region_device_init(mdev, &mem_rdev_rw_ops, base, size);
}
void region_device_init(struct region_device *rdev,
const struct region_device_ops *ops, size_t offset,
size_t size)
{
memset(rdev, 0, sizeof(*rdev));
rdev->root = NULL;
rdev->ops = ops;
rdev->region.offset = offset;
rdev->region.size = size;
}
static void xlate_region_device_init(struct xlate_region_device *xdev,
const struct region_device_ops *ops,
size_t window_count, const struct xlate_window *window_arr,
size_t parent_size)
{
memset(xdev, 0, sizeof(*xdev));
xdev->window_count = window_count;
xdev->window_arr = window_arr;
region_device_init(&xdev->rdev, ops, 0, parent_size);
}
void xlate_region_device_ro_init(struct xlate_region_device *xdev,
size_t window_count, const struct xlate_window *window_arr,
size_t parent_size)
{
xlate_region_device_init(xdev, &xlate_rdev_ro_ops, window_count, window_arr,
parent_size);
}
void xlate_region_device_rw_init(struct xlate_region_device *xdev,
size_t window_count, const struct xlate_window *window_arr,
size_t parent_size)
{
xlate_region_device_init(xdev, &xlate_rdev_rw_ops, window_count, window_arr,
parent_size);
}
void xlate_window_init(struct xlate_window *window, const struct region_device *access_dev,
size_t sub_region_offset, size_t sub_region_size)
{
window->access_dev = access_dev;
window->sub_region.offset = sub_region_offset;
window->sub_region.size = sub_region_size;
}
static void *mdev_mmap(const struct region_device *rd, size_t offset,
size_t size __unused)
{
const struct mem_region_device *mdev;
mdev = container_of(rd, __typeof__(*mdev), rdev);
return &mdev->base[offset];
}
static int mdev_munmap(const struct region_device *rd __unused,
void *mapping __unused)
{
return 0;
}
static ssize_t mdev_readat(const struct region_device *rd, void *b,
size_t offset, size_t size)
{
const struct mem_region_device *mdev;
mdev = container_of(rd, __typeof__(*mdev), rdev);
memcpy(b, &mdev->base[offset], size);
return size;
}
static ssize_t mdev_writeat(const struct region_device *rd, const void *b,
size_t offset, size_t size)
{
const struct mem_region_device *mdev;
mdev = container_of(rd, __typeof__(*mdev), rdev);
memcpy(&mdev->base[offset], b, size);
return size;
}
static ssize_t mdev_eraseat(const struct region_device *rd, size_t offset,
size_t size)
{
const struct mem_region_device *mdev;
mdev = container_of(rd, __typeof__(*mdev), rdev);
memset(&mdev->base[offset], 0, size);
return size;
}
const struct region_device_ops mem_rdev_ro_ops = {
.mmap = mdev_mmap,
.munmap = mdev_munmap,
.readat = mdev_readat,
};
const struct region_device_ops mem_rdev_rw_ops = {
.mmap = mdev_mmap,
.munmap = mdev_munmap,
.readat = mdev_readat,
.writeat = mdev_writeat,
.eraseat = mdev_eraseat,
};
void mmap_helper_device_init(struct mmap_helper_region_device *mdev,
void *cache, size_t cache_size)
{
mem_pool_init(&mdev->pool, cache, cache_size);
}
void *mmap_helper_rdev_mmap(const struct region_device *rd, size_t offset,
size_t size)
{
struct mmap_helper_region_device *mdev;
void *mapping;
mdev = container_of((void *)rd, __typeof__(*mdev), rdev);
mapping = mem_pool_alloc(&mdev->pool, size);
if (mapping == NULL)
return NULL;
if (rd->ops->readat(rd, mapping, offset, size) != size) {
mem_pool_free(&mdev->pool, mapping);
return NULL;
}
return mapping;
}
int mmap_helper_rdev_munmap(const struct region_device *rd, void *mapping)
{
struct mmap_helper_region_device *mdev;
mdev = container_of((void *)rd, __typeof__(*mdev), rdev);
mem_pool_free(&mdev->pool, mapping);
return 0;
}
static const struct xlate_window *xlate_find_window(const struct xlate_region_device *xldev,
const struct region *req)
{
size_t i;
const struct xlate_window *xlwindow;
for (i = 0; i < xldev->window_count; i++) {
xlwindow = &xldev->window_arr[i];
if (region_is_subregion(&xlwindow->sub_region, req))
return xlwindow;
}
return NULL;
}
static void *xlate_mmap(const struct region_device *rd, size_t offset,
size_t size)
{
const struct xlate_region_device *xldev;
struct region req = {
.offset = offset,
.size = size,
};
const struct xlate_window *xlwindow;
xldev = container_of(rd, __typeof__(*xldev), rdev);
xlwindow = xlate_find_window(xldev, &req);
if (!xlwindow)
return NULL;
offset -= region_offset(&xlwindow->sub_region);
return rdev_mmap(xlwindow->access_dev, offset, size);
}
static int xlate_munmap(const struct region_device *rd __unused, void *mapping __unused)
{
/*
* xlate_region_device does not keep track of the access device that was used to service
* a mmap request. So, munmap does not do anything. If munmap functionality is required,
* then xlate_region_device will have to be updated to accept some pre-allocated space
* from caller to keep track of the mapping requests. Since xlate_region_device is only
* used for memory mapped boot media on the backend right now, skipping munmap is fine.
*/
return 0;
}
static ssize_t xlate_readat(const struct region_device *rd, void *b,
size_t offset, size_t size)
{
struct region req = {
.offset = offset,
.size = size,
};
const struct xlate_window *xlwindow;
const struct xlate_region_device *xldev;
xldev = container_of(rd, __typeof__(*xldev), rdev);
xlwindow = xlate_find_window(xldev, &req);
if (!xlwindow)
return -1;
offset -= region_offset(&xlwindow->sub_region);
return rdev_readat(xlwindow->access_dev, b, offset, size);
}
static ssize_t xlate_writeat(const struct region_device *rd, const void *b,
size_t offset, size_t size)
{
struct region req = {
.offset = offset,
.size = size,
};
const struct xlate_window *xlwindow;
const struct xlate_region_device *xldev;
xldev = container_of(rd, __typeof__(*xldev), rdev);
xlwindow = xlate_find_window(xldev, &req);
if (!xlwindow)
return -1;
offset -= region_offset(&xlwindow->sub_region);
return rdev_writeat(xlwindow->access_dev, b, offset, size);
}
static ssize_t xlate_eraseat(const struct region_device *rd,
size_t offset, size_t size)
{
struct region req = {
.offset = offset,
.size = size,
};
const struct xlate_window *xlwindow;
const struct xlate_region_device *xldev;
xldev = container_of(rd, __typeof__(*xldev), rdev);
xlwindow = xlate_find_window(xldev, &req);
if (!xlwindow)
return -1;
offset -= region_offset(&xlwindow->sub_region);
return rdev_eraseat(xlwindow->access_dev, offset, size);
}
const struct region_device_ops xlate_rdev_ro_ops = {
.mmap = xlate_mmap,
.munmap = xlate_munmap,
.readat = xlate_readat,
};
const struct region_device_ops xlate_rdev_rw_ops = {
.mmap = xlate_mmap,
.munmap = xlate_munmap,
.readat = xlate_readat,
.writeat = xlate_writeat,
.eraseat = xlate_eraseat,
};
static void *incoherent_mmap(const struct region_device *rd, size_t offset,
size_t size)
{
const struct incoherent_rdev *irdev;
irdev = container_of(rd, const struct incoherent_rdev, rdev);
return rdev_mmap(irdev->read, offset, size);
}
static int incoherent_munmap(const struct region_device *rd, void *mapping)
{
const struct incoherent_rdev *irdev;
irdev = container_of(rd, const struct incoherent_rdev, rdev);
return rdev_munmap(irdev->read, mapping);
}
static ssize_t incoherent_readat(const struct region_device *rd, void *b,
size_t offset, size_t size)
{
const struct incoherent_rdev *irdev;
irdev = container_of(rd, const struct incoherent_rdev, rdev);
return rdev_readat(irdev->read, b, offset, size);
}
static ssize_t incoherent_writeat(const struct region_device *rd, const void *b,
size_t offset, size_t size)
{
const struct incoherent_rdev *irdev;
irdev = container_of(rd, const struct incoherent_rdev, rdev);
return rdev_writeat(irdev->write, b, offset, size);
}
static ssize_t incoherent_eraseat(const struct region_device *rd, size_t offset,
size_t size)
{
const struct incoherent_rdev *irdev;
irdev = container_of(rd, const struct incoherent_rdev, rdev);
return rdev_eraseat(irdev->write, offset, size);
}
static const struct region_device_ops incoherent_rdev_ops = {
.mmap = incoherent_mmap,
.munmap = incoherent_munmap,
.readat = incoherent_readat,
.writeat = incoherent_writeat,
.eraseat = incoherent_eraseat,
};
const struct region_device *incoherent_rdev_init(struct incoherent_rdev *irdev,
const struct region *r,
const struct region_device *read,
const struct region_device *write)
{
const size_t size = region_sz(r);
if (size != region_device_sz(read) || size != region_device_sz(write))
return NULL;
/* The region is represented as offset 0 to size. That way, the generic
* rdev operations can be called on the read or write implementation
* without any unnecessary translation because the offsets all start
* at 0. */
region_device_init(&irdev->rdev, &incoherent_rdev_ops, 0, size);
irdev->read = read;
irdev->write = write;
return &irdev->rdev;
}