tegrakernel/kernel/kernel-4.9/sound/pci/ctxfi/ctvmem.c

246 lines
5.6 KiB
C

/**
* Copyright (C) 2008, Creative Technology Ltd. All Rights Reserved.
*
* This source file is released under GPL v2 license (no other versions).
* See the COPYING file included in the main directory of this source
* distribution for the license terms and conditions.
*
* @File ctvmem.c
*
* @Brief
* This file contains the implementation of virtual memory management object
* for card device.
*
* @Author Liu Chun
* @Date Apr 1 2008
*/
#include "ctvmem.h"
#include "ctatc.h"
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/io.h>
#include <sound/pcm.h>
#define CT_PTES_PER_PAGE (CT_PAGE_SIZE / sizeof(void *))
#define CT_ADDRS_PER_PAGE (CT_PTES_PER_PAGE * CT_PAGE_SIZE)
/* *
* Find or create vm block based on requested @size.
* @size must be page aligned.
* */
static struct ct_vm_block *
get_vm_block(struct ct_vm *vm, unsigned int size, struct ct_atc *atc)
{
struct ct_vm_block *block = NULL, *entry;
struct list_head *pos;
size = CT_PAGE_ALIGN(size);
if (size > vm->size) {
dev_err(atc->card->dev,
"Fail! No sufficient device virtual memory space available!\n");
return NULL;
}
mutex_lock(&vm->lock);
list_for_each(pos, &vm->unused) {
entry = list_entry(pos, struct ct_vm_block, list);
if (entry->size >= size)
break; /* found a block that is big enough */
}
if (pos == &vm->unused)
goto out;
if (entry->size == size) {
/* Move the vm node from unused list to used list directly */
list_move(&entry->list, &vm->used);
vm->size -= size;
block = entry;
goto out;
}
block = kzalloc(sizeof(*block), GFP_KERNEL);
if (!block)
goto out;
block->addr = entry->addr;
block->size = size;
list_add(&block->list, &vm->used);
entry->addr += size;
entry->size -= size;
vm->size -= size;
out:
mutex_unlock(&vm->lock);
return block;
}
static void put_vm_block(struct ct_vm *vm, struct ct_vm_block *block)
{
struct ct_vm_block *entry, *pre_ent;
struct list_head *pos, *pre;
block->size = CT_PAGE_ALIGN(block->size);
mutex_lock(&vm->lock);
list_del(&block->list);
vm->size += block->size;
list_for_each(pos, &vm->unused) {
entry = list_entry(pos, struct ct_vm_block, list);
if (entry->addr >= (block->addr + block->size))
break; /* found a position */
}
if (pos == &vm->unused) {
list_add_tail(&block->list, &vm->unused);
entry = block;
} else {
if ((block->addr + block->size) == entry->addr) {
entry->addr = block->addr;
entry->size += block->size;
kfree(block);
} else {
__list_add(&block->list, pos->prev, pos);
entry = block;
}
}
pos = &entry->list;
pre = pos->prev;
while (pre != &vm->unused) {
entry = list_entry(pos, struct ct_vm_block, list);
pre_ent = list_entry(pre, struct ct_vm_block, list);
if ((pre_ent->addr + pre_ent->size) > entry->addr)
break;
pre_ent->size += entry->size;
list_del(pos);
kfree(entry);
pos = pre;
pre = pos->prev;
}
mutex_unlock(&vm->lock);
}
/* Map host addr (kmalloced/vmalloced) to device logical addr. */
static struct ct_vm_block *
ct_vm_map(struct ct_vm *vm, struct snd_pcm_substream *substream, int size)
{
struct ct_vm_block *block;
unsigned int pte_start;
unsigned i, pages;
unsigned long *ptp;
struct ct_atc *atc = snd_pcm_substream_chip(substream);
block = get_vm_block(vm, size, atc);
if (block == NULL) {
dev_err(atc->card->dev,
"No virtual memory block that is big enough to allocate!\n");
return NULL;
}
ptp = (unsigned long *)vm->ptp[0].area;
pte_start = (block->addr >> CT_PAGE_SHIFT);
pages = block->size >> CT_PAGE_SHIFT;
for (i = 0; i < pages; i++) {
unsigned long addr;
addr = snd_pcm_sgbuf_get_addr(substream, i << CT_PAGE_SHIFT);
ptp[pte_start + i] = addr;
}
block->size = size;
return block;
}
static void ct_vm_unmap(struct ct_vm *vm, struct ct_vm_block *block)
{
/* do unmapping */
put_vm_block(vm, block);
}
/* *
* return the host physical addr of the @index-th device
* page table page on success, or ~0UL on failure.
* The first returned ~0UL indicates the termination.
* */
static dma_addr_t
ct_get_ptp_phys(struct ct_vm *vm, int index)
{
return (index >= CT_PTP_NUM) ? ~0UL : vm->ptp[index].addr;
}
int ct_vm_create(struct ct_vm **rvm, struct pci_dev *pci)
{
struct ct_vm *vm;
struct ct_vm_block *block;
int i, err = 0;
*rvm = NULL;
vm = kzalloc(sizeof(*vm), GFP_KERNEL);
if (!vm)
return -ENOMEM;
mutex_init(&vm->lock);
/* Allocate page table pages */
for (i = 0; i < CT_PTP_NUM; i++) {
err = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV,
snd_dma_pci_data(pci),
PAGE_SIZE, &vm->ptp[i]);
if (err < 0)
break;
}
if (err < 0) {
/* no page table pages are allocated */
ct_vm_destroy(vm);
return -ENOMEM;
}
vm->size = CT_ADDRS_PER_PAGE * i;
vm->map = ct_vm_map;
vm->unmap = ct_vm_unmap;
vm->get_ptp_phys = ct_get_ptp_phys;
INIT_LIST_HEAD(&vm->unused);
INIT_LIST_HEAD(&vm->used);
block = kzalloc(sizeof(*block), GFP_KERNEL);
if (NULL != block) {
block->addr = 0;
block->size = vm->size;
list_add(&block->list, &vm->unused);
}
*rvm = vm;
return 0;
}
/* The caller must ensure no mapping pages are being used
* by hardware before calling this function */
void ct_vm_destroy(struct ct_vm *vm)
{
int i;
struct list_head *pos;
struct ct_vm_block *entry;
/* free used and unused list nodes */
while (!list_empty(&vm->used)) {
pos = vm->used.next;
list_del(pos);
entry = list_entry(pos, struct ct_vm_block, list);
kfree(entry);
}
while (!list_empty(&vm->unused)) {
pos = vm->unused.next;
list_del(pos);
entry = list_entry(pos, struct ct_vm_block, list);
kfree(entry);
}
/* free allocated page table pages */
for (i = 0; i < CT_PTP_NUM; i++)
snd_dma_free_pages(&vm->ptp[i]);
vm->size = 0;
kfree(vm);
}