mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-28 11:18:45 +07:00
315786ebbf
Mapping and unmapping are more often than not in the critical path. map_sg allows IOMMU driver implementations to optimize the process of mapping buffers into the IOMMU page tables. Instead of mapping a buffer one page at a time and requiring potentially expensive TLB operations for each page, this function allows the driver to map all pages in one go and defer TLB maintenance until after all pages have been mapped. Additionally, the mapping operation would be faster in general since clients does not have to keep calling map API over and over again for each physically contiguous chunk of memory that needs to be mapped to a virtually contiguous region. Signed-off-by: Olav Haugan <ohaugan@codeaurora.org> Signed-off-by: Joerg Roedel <jroedel@suse.de>
396 lines
10 KiB
C
396 lines
10 KiB
C
/*
|
|
* IOMMU for IPMMU/IPMMUI
|
|
* Copyright (C) 2012 Hideki EIRAKU
|
|
*
|
|
* This program is free software; you can redistribute it and/or modify
|
|
* it under the terms of the GNU General Public License as published by
|
|
* the Free Software Foundation; version 2 of the License.
|
|
*/
|
|
|
|
#include <linux/dma-mapping.h>
|
|
#include <linux/io.h>
|
|
#include <linux/iommu.h>
|
|
#include <linux/platform_device.h>
|
|
#include <linux/sizes.h>
|
|
#include <linux/slab.h>
|
|
#include <asm/dma-iommu.h>
|
|
#include "shmobile-ipmmu.h"
|
|
|
|
#define L1_SIZE CONFIG_SHMOBILE_IOMMU_L1SIZE
|
|
#define L1_LEN (L1_SIZE / 4)
|
|
#define L1_ALIGN L1_SIZE
|
|
#define L2_SIZE SZ_1K
|
|
#define L2_LEN (L2_SIZE / 4)
|
|
#define L2_ALIGN L2_SIZE
|
|
|
|
struct shmobile_iommu_domain_pgtable {
|
|
uint32_t *pgtable;
|
|
dma_addr_t handle;
|
|
};
|
|
|
|
struct shmobile_iommu_archdata {
|
|
struct list_head attached_list;
|
|
struct dma_iommu_mapping *iommu_mapping;
|
|
spinlock_t attach_lock;
|
|
struct shmobile_iommu_domain *attached;
|
|
int num_attached_devices;
|
|
struct shmobile_ipmmu *ipmmu;
|
|
};
|
|
|
|
struct shmobile_iommu_domain {
|
|
struct shmobile_iommu_domain_pgtable l1, l2[L1_LEN];
|
|
spinlock_t map_lock;
|
|
spinlock_t attached_list_lock;
|
|
struct list_head attached_list;
|
|
};
|
|
|
|
static struct shmobile_iommu_archdata *ipmmu_archdata;
|
|
static struct kmem_cache *l1cache, *l2cache;
|
|
|
|
static int pgtable_alloc(struct shmobile_iommu_domain_pgtable *pgtable,
|
|
struct kmem_cache *cache, size_t size)
|
|
{
|
|
pgtable->pgtable = kmem_cache_zalloc(cache, GFP_ATOMIC);
|
|
if (!pgtable->pgtable)
|
|
return -ENOMEM;
|
|
pgtable->handle = dma_map_single(NULL, pgtable->pgtable, size,
|
|
DMA_TO_DEVICE);
|
|
return 0;
|
|
}
|
|
|
|
static void pgtable_free(struct shmobile_iommu_domain_pgtable *pgtable,
|
|
struct kmem_cache *cache, size_t size)
|
|
{
|
|
dma_unmap_single(NULL, pgtable->handle, size, DMA_TO_DEVICE);
|
|
kmem_cache_free(cache, pgtable->pgtable);
|
|
}
|
|
|
|
static uint32_t pgtable_read(struct shmobile_iommu_domain_pgtable *pgtable,
|
|
unsigned int index)
|
|
{
|
|
return pgtable->pgtable[index];
|
|
}
|
|
|
|
static void pgtable_write(struct shmobile_iommu_domain_pgtable *pgtable,
|
|
unsigned int index, unsigned int count, uint32_t val)
|
|
{
|
|
unsigned int i;
|
|
|
|
for (i = 0; i < count; i++)
|
|
pgtable->pgtable[index + i] = val;
|
|
dma_sync_single_for_device(NULL, pgtable->handle + index * sizeof(val),
|
|
sizeof(val) * count, DMA_TO_DEVICE);
|
|
}
|
|
|
|
static int shmobile_iommu_domain_init(struct iommu_domain *domain)
|
|
{
|
|
struct shmobile_iommu_domain *sh_domain;
|
|
int i, ret;
|
|
|
|
sh_domain = kmalloc(sizeof(*sh_domain), GFP_KERNEL);
|
|
if (!sh_domain)
|
|
return -ENOMEM;
|
|
ret = pgtable_alloc(&sh_domain->l1, l1cache, L1_SIZE);
|
|
if (ret < 0) {
|
|
kfree(sh_domain);
|
|
return ret;
|
|
}
|
|
for (i = 0; i < L1_LEN; i++)
|
|
sh_domain->l2[i].pgtable = NULL;
|
|
spin_lock_init(&sh_domain->map_lock);
|
|
spin_lock_init(&sh_domain->attached_list_lock);
|
|
INIT_LIST_HEAD(&sh_domain->attached_list);
|
|
domain->priv = sh_domain;
|
|
return 0;
|
|
}
|
|
|
|
static void shmobile_iommu_domain_destroy(struct iommu_domain *domain)
|
|
{
|
|
struct shmobile_iommu_domain *sh_domain = domain->priv;
|
|
int i;
|
|
|
|
for (i = 0; i < L1_LEN; i++) {
|
|
if (sh_domain->l2[i].pgtable)
|
|
pgtable_free(&sh_domain->l2[i], l2cache, L2_SIZE);
|
|
}
|
|
pgtable_free(&sh_domain->l1, l1cache, L1_SIZE);
|
|
kfree(sh_domain);
|
|
domain->priv = NULL;
|
|
}
|
|
|
|
static int shmobile_iommu_attach_device(struct iommu_domain *domain,
|
|
struct device *dev)
|
|
{
|
|
struct shmobile_iommu_archdata *archdata = dev->archdata.iommu;
|
|
struct shmobile_iommu_domain *sh_domain = domain->priv;
|
|
int ret = -EBUSY;
|
|
|
|
if (!archdata)
|
|
return -ENODEV;
|
|
spin_lock(&sh_domain->attached_list_lock);
|
|
spin_lock(&archdata->attach_lock);
|
|
if (archdata->attached != sh_domain) {
|
|
if (archdata->attached)
|
|
goto err;
|
|
ipmmu_tlb_set(archdata->ipmmu, sh_domain->l1.handle, L1_SIZE,
|
|
0);
|
|
ipmmu_tlb_flush(archdata->ipmmu);
|
|
archdata->attached = sh_domain;
|
|
archdata->num_attached_devices = 0;
|
|
list_add(&archdata->attached_list, &sh_domain->attached_list);
|
|
}
|
|
archdata->num_attached_devices++;
|
|
ret = 0;
|
|
err:
|
|
spin_unlock(&archdata->attach_lock);
|
|
spin_unlock(&sh_domain->attached_list_lock);
|
|
return ret;
|
|
}
|
|
|
|
static void shmobile_iommu_detach_device(struct iommu_domain *domain,
|
|
struct device *dev)
|
|
{
|
|
struct shmobile_iommu_archdata *archdata = dev->archdata.iommu;
|
|
struct shmobile_iommu_domain *sh_domain = domain->priv;
|
|
|
|
if (!archdata)
|
|
return;
|
|
spin_lock(&sh_domain->attached_list_lock);
|
|
spin_lock(&archdata->attach_lock);
|
|
archdata->num_attached_devices--;
|
|
if (!archdata->num_attached_devices) {
|
|
ipmmu_tlb_set(archdata->ipmmu, 0, 0, 0);
|
|
ipmmu_tlb_flush(archdata->ipmmu);
|
|
archdata->attached = NULL;
|
|
list_del(&archdata->attached_list);
|
|
}
|
|
spin_unlock(&archdata->attach_lock);
|
|
spin_unlock(&sh_domain->attached_list_lock);
|
|
}
|
|
|
|
static void domain_tlb_flush(struct shmobile_iommu_domain *sh_domain)
|
|
{
|
|
struct shmobile_iommu_archdata *archdata;
|
|
|
|
spin_lock(&sh_domain->attached_list_lock);
|
|
list_for_each_entry(archdata, &sh_domain->attached_list, attached_list)
|
|
ipmmu_tlb_flush(archdata->ipmmu);
|
|
spin_unlock(&sh_domain->attached_list_lock);
|
|
}
|
|
|
|
static int l2alloc(struct shmobile_iommu_domain *sh_domain,
|
|
unsigned int l1index)
|
|
{
|
|
int ret;
|
|
|
|
if (!sh_domain->l2[l1index].pgtable) {
|
|
ret = pgtable_alloc(&sh_domain->l2[l1index], l2cache, L2_SIZE);
|
|
if (ret < 0)
|
|
return ret;
|
|
}
|
|
pgtable_write(&sh_domain->l1, l1index, 1,
|
|
sh_domain->l2[l1index].handle | 0x1);
|
|
return 0;
|
|
}
|
|
|
|
static void l2realfree(struct shmobile_iommu_domain_pgtable *l2)
|
|
{
|
|
if (l2->pgtable)
|
|
pgtable_free(l2, l2cache, L2_SIZE);
|
|
}
|
|
|
|
static void l2free(struct shmobile_iommu_domain *sh_domain,
|
|
unsigned int l1index,
|
|
struct shmobile_iommu_domain_pgtable *l2)
|
|
{
|
|
pgtable_write(&sh_domain->l1, l1index, 1, 0);
|
|
if (sh_domain->l2[l1index].pgtable) {
|
|
*l2 = sh_domain->l2[l1index];
|
|
sh_domain->l2[l1index].pgtable = NULL;
|
|
}
|
|
}
|
|
|
|
static int shmobile_iommu_map(struct iommu_domain *domain, unsigned long iova,
|
|
phys_addr_t paddr, size_t size, int prot)
|
|
{
|
|
struct shmobile_iommu_domain_pgtable l2 = { .pgtable = NULL };
|
|
struct shmobile_iommu_domain *sh_domain = domain->priv;
|
|
unsigned int l1index, l2index;
|
|
int ret;
|
|
|
|
l1index = iova >> 20;
|
|
switch (size) {
|
|
case SZ_4K:
|
|
l2index = (iova >> 12) & 0xff;
|
|
spin_lock(&sh_domain->map_lock);
|
|
ret = l2alloc(sh_domain, l1index);
|
|
if (!ret)
|
|
pgtable_write(&sh_domain->l2[l1index], l2index, 1,
|
|
paddr | 0xff2);
|
|
spin_unlock(&sh_domain->map_lock);
|
|
break;
|
|
case SZ_64K:
|
|
l2index = (iova >> 12) & 0xf0;
|
|
spin_lock(&sh_domain->map_lock);
|
|
ret = l2alloc(sh_domain, l1index);
|
|
if (!ret)
|
|
pgtable_write(&sh_domain->l2[l1index], l2index, 0x10,
|
|
paddr | 0xff1);
|
|
spin_unlock(&sh_domain->map_lock);
|
|
break;
|
|
case SZ_1M:
|
|
spin_lock(&sh_domain->map_lock);
|
|
l2free(sh_domain, l1index, &l2);
|
|
pgtable_write(&sh_domain->l1, l1index, 1, paddr | 0xc02);
|
|
spin_unlock(&sh_domain->map_lock);
|
|
ret = 0;
|
|
break;
|
|
default:
|
|
ret = -EINVAL;
|
|
}
|
|
if (!ret)
|
|
domain_tlb_flush(sh_domain);
|
|
l2realfree(&l2);
|
|
return ret;
|
|
}
|
|
|
|
static size_t shmobile_iommu_unmap(struct iommu_domain *domain,
|
|
unsigned long iova, size_t size)
|
|
{
|
|
struct shmobile_iommu_domain_pgtable l2 = { .pgtable = NULL };
|
|
struct shmobile_iommu_domain *sh_domain = domain->priv;
|
|
unsigned int l1index, l2index;
|
|
uint32_t l2entry = 0;
|
|
size_t ret = 0;
|
|
|
|
l1index = iova >> 20;
|
|
if (!(iova & 0xfffff) && size >= SZ_1M) {
|
|
spin_lock(&sh_domain->map_lock);
|
|
l2free(sh_domain, l1index, &l2);
|
|
spin_unlock(&sh_domain->map_lock);
|
|
ret = SZ_1M;
|
|
goto done;
|
|
}
|
|
l2index = (iova >> 12) & 0xff;
|
|
spin_lock(&sh_domain->map_lock);
|
|
if (sh_domain->l2[l1index].pgtable)
|
|
l2entry = pgtable_read(&sh_domain->l2[l1index], l2index);
|
|
switch (l2entry & 3) {
|
|
case 1:
|
|
if (l2index & 0xf)
|
|
break;
|
|
pgtable_write(&sh_domain->l2[l1index], l2index, 0x10, 0);
|
|
ret = SZ_64K;
|
|
break;
|
|
case 2:
|
|
pgtable_write(&sh_domain->l2[l1index], l2index, 1, 0);
|
|
ret = SZ_4K;
|
|
break;
|
|
}
|
|
spin_unlock(&sh_domain->map_lock);
|
|
done:
|
|
if (ret)
|
|
domain_tlb_flush(sh_domain);
|
|
l2realfree(&l2);
|
|
return ret;
|
|
}
|
|
|
|
static phys_addr_t shmobile_iommu_iova_to_phys(struct iommu_domain *domain,
|
|
dma_addr_t iova)
|
|
{
|
|
struct shmobile_iommu_domain *sh_domain = domain->priv;
|
|
uint32_t l1entry = 0, l2entry = 0;
|
|
unsigned int l1index, l2index;
|
|
|
|
l1index = iova >> 20;
|
|
l2index = (iova >> 12) & 0xff;
|
|
spin_lock(&sh_domain->map_lock);
|
|
if (sh_domain->l2[l1index].pgtable)
|
|
l2entry = pgtable_read(&sh_domain->l2[l1index], l2index);
|
|
else
|
|
l1entry = pgtable_read(&sh_domain->l1, l1index);
|
|
spin_unlock(&sh_domain->map_lock);
|
|
switch (l2entry & 3) {
|
|
case 1:
|
|
return (l2entry & ~0xffff) | (iova & 0xffff);
|
|
case 2:
|
|
return (l2entry & ~0xfff) | (iova & 0xfff);
|
|
default:
|
|
if ((l1entry & 3) == 2)
|
|
return (l1entry & ~0xfffff) | (iova & 0xfffff);
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
static int find_dev_name(struct shmobile_ipmmu *ipmmu, const char *dev_name)
|
|
{
|
|
unsigned int i, n = ipmmu->num_dev_names;
|
|
|
|
for (i = 0; i < n; i++) {
|
|
if (strcmp(ipmmu->dev_names[i], dev_name) == 0)
|
|
return 1;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
static int shmobile_iommu_add_device(struct device *dev)
|
|
{
|
|
struct shmobile_iommu_archdata *archdata = ipmmu_archdata;
|
|
struct dma_iommu_mapping *mapping;
|
|
|
|
if (!find_dev_name(archdata->ipmmu, dev_name(dev)))
|
|
return 0;
|
|
mapping = archdata->iommu_mapping;
|
|
if (!mapping) {
|
|
mapping = arm_iommu_create_mapping(&platform_bus_type, 0,
|
|
L1_LEN << 20);
|
|
if (IS_ERR(mapping))
|
|
return PTR_ERR(mapping);
|
|
archdata->iommu_mapping = mapping;
|
|
}
|
|
dev->archdata.iommu = archdata;
|
|
if (arm_iommu_attach_device(dev, mapping))
|
|
pr_err("arm_iommu_attach_device failed\n");
|
|
return 0;
|
|
}
|
|
|
|
static const struct iommu_ops shmobile_iommu_ops = {
|
|
.domain_init = shmobile_iommu_domain_init,
|
|
.domain_destroy = shmobile_iommu_domain_destroy,
|
|
.attach_dev = shmobile_iommu_attach_device,
|
|
.detach_dev = shmobile_iommu_detach_device,
|
|
.map = shmobile_iommu_map,
|
|
.unmap = shmobile_iommu_unmap,
|
|
.map_sg = default_iommu_map_sg,
|
|
.iova_to_phys = shmobile_iommu_iova_to_phys,
|
|
.add_device = shmobile_iommu_add_device,
|
|
.pgsize_bitmap = SZ_1M | SZ_64K | SZ_4K,
|
|
};
|
|
|
|
int ipmmu_iommu_init(struct shmobile_ipmmu *ipmmu)
|
|
{
|
|
static struct shmobile_iommu_archdata *archdata;
|
|
|
|
l1cache = kmem_cache_create("shmobile-iommu-pgtable1", L1_SIZE,
|
|
L1_ALIGN, SLAB_HWCACHE_ALIGN, NULL);
|
|
if (!l1cache)
|
|
return -ENOMEM;
|
|
l2cache = kmem_cache_create("shmobile-iommu-pgtable2", L2_SIZE,
|
|
L2_ALIGN, SLAB_HWCACHE_ALIGN, NULL);
|
|
if (!l2cache) {
|
|
kmem_cache_destroy(l1cache);
|
|
return -ENOMEM;
|
|
}
|
|
archdata = kzalloc(sizeof(*archdata), GFP_KERNEL);
|
|
if (!archdata) {
|
|
kmem_cache_destroy(l1cache);
|
|
kmem_cache_destroy(l2cache);
|
|
return -ENOMEM;
|
|
}
|
|
spin_lock_init(&archdata->attach_lock);
|
|
archdata->ipmmu = ipmmu;
|
|
ipmmu_archdata = archdata;
|
|
bus_set_iommu(&platform_bus_type, &shmobile_iommu_ops);
|
|
return 0;
|
|
}
|