alistair23-linux/sound/pci/emu10k1/memory.c
Maciej S. Szmigiero 04f8773a3e ALSA: emu10k1: add a IOMMU workaround
The Audigy 2 CA0102 chip (but most likely others from the emu10k1 family,
too) has a problem that from time to time it likes to do few DMA reads a
bit beyond its normal allocation and gets very confused if these reads get
blocked by a IOMMU.

For the first (reserved) page this happens multiple times at every
playback, for various synth pages it happens randomly, rarely for PCM
playback buffers and the page table memory itself.
All these reads seem to follow a similar pattern, observed read offsets
beyond the allocation end were 0x00, 0x40, 0x80 and 0xc0 (PCI cache line
multiples), so it looks like the device tries to accesses up to 256 extra
bytes.

As a workaround let's widen these DMA allocations by an extra page if we
detect that the device is behind a non-passthrough IOMMU (the DMA memory
should be relatively plenty on IOMMU systems).

Signed-off-by: Maciej S. Szmigiero <mail@maciej.szmigiero.name>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
2018-02-14 07:46:55 +01:00

637 lines
17 KiB
C

/*
* Copyright (c) by Jaroslav Kysela <perex@perex.cz>
* Copyright (c) by Takashi Iwai <tiwai@suse.de>
*
* EMU10K1 memory page allocation (PTB area)
*
*
* 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; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
*/
#include <linux/pci.h>
#include <linux/gfp.h>
#include <linux/time.h>
#include <linux/mutex.h>
#include <linux/export.h>
#include <sound/core.h>
#include <sound/emu10k1.h>
/* page arguments of these two macros are Emu page (4096 bytes), not like
* aligned pages in others
*/
#define __set_ptb_entry(emu,page,addr) \
(((__le32 *)(emu)->ptb_pages.area)[page] = \
cpu_to_le32(((addr) << (emu->address_mode)) | (page)))
#define __get_ptb_entry(emu, page) \
(le32_to_cpu(((__le32 *)(emu)->ptb_pages.area)[page]))
#define UNIT_PAGES (PAGE_SIZE / EMUPAGESIZE)
#define MAX_ALIGN_PAGES0 (MAXPAGES0 / UNIT_PAGES)
#define MAX_ALIGN_PAGES1 (MAXPAGES1 / UNIT_PAGES)
/* get aligned page from offset address */
#define get_aligned_page(offset) ((offset) >> PAGE_SHIFT)
/* get offset address from aligned page */
#define aligned_page_offset(page) ((page) << PAGE_SHIFT)
#if PAGE_SIZE == EMUPAGESIZE && !IS_ENABLED(CONFIG_DYNAMIC_DEBUG)
/* fill PTB entrie(s) corresponding to page with addr */
#define set_ptb_entry(emu,page,addr) __set_ptb_entry(emu,page,addr)
/* fill PTB entrie(s) corresponding to page with silence pointer */
#define set_silent_ptb(emu,page) __set_ptb_entry(emu,page,emu->silent_page.addr)
#else
/* fill PTB entries -- we need to fill UNIT_PAGES entries */
static inline void set_ptb_entry(struct snd_emu10k1 *emu, int page, dma_addr_t addr)
{
int i;
page *= UNIT_PAGES;
for (i = 0; i < UNIT_PAGES; i++, page++) {
__set_ptb_entry(emu, page, addr);
dev_dbg(emu->card->dev, "mapped page %d to entry %.8x\n", page,
(unsigned int)__get_ptb_entry(emu, page));
addr += EMUPAGESIZE;
}
}
static inline void set_silent_ptb(struct snd_emu10k1 *emu, int page)
{
int i;
page *= UNIT_PAGES;
for (i = 0; i < UNIT_PAGES; i++, page++) {
/* do not increment ptr */
__set_ptb_entry(emu, page, emu->silent_page.addr);
dev_dbg(emu->card->dev, "mapped silent page %d to entry %.8x\n",
page, (unsigned int)__get_ptb_entry(emu, page));
}
}
#endif /* PAGE_SIZE */
/*
*/
static int synth_alloc_pages(struct snd_emu10k1 *hw, struct snd_emu10k1_memblk *blk);
static int synth_free_pages(struct snd_emu10k1 *hw, struct snd_emu10k1_memblk *blk);
#define get_emu10k1_memblk(l,member) list_entry(l, struct snd_emu10k1_memblk, member)
/* initialize emu10k1 part */
static void emu10k1_memblk_init(struct snd_emu10k1_memblk *blk)
{
blk->mapped_page = -1;
INIT_LIST_HEAD(&blk->mapped_link);
INIT_LIST_HEAD(&blk->mapped_order_link);
blk->map_locked = 0;
blk->first_page = get_aligned_page(blk->mem.offset);
blk->last_page = get_aligned_page(blk->mem.offset + blk->mem.size - 1);
blk->pages = blk->last_page - blk->first_page + 1;
}
/*
* search empty region on PTB with the given size
*
* if an empty region is found, return the page and store the next mapped block
* in nextp
* if not found, return a negative error code.
*/
static int search_empty_map_area(struct snd_emu10k1 *emu, int npages, struct list_head **nextp)
{
int page = 1, found_page = -ENOMEM;
int max_size = npages;
int size;
struct list_head *candidate = &emu->mapped_link_head;
struct list_head *pos;
list_for_each (pos, &emu->mapped_link_head) {
struct snd_emu10k1_memblk *blk = get_emu10k1_memblk(pos, mapped_link);
if (blk->mapped_page < 0)
continue;
size = blk->mapped_page - page;
if (size == npages) {
*nextp = pos;
return page;
}
else if (size > max_size) {
/* we look for the maximum empty hole */
max_size = size;
candidate = pos;
found_page = page;
}
page = blk->mapped_page + blk->pages;
}
size = (emu->address_mode ? MAX_ALIGN_PAGES1 : MAX_ALIGN_PAGES0) - page;
if (size >= max_size) {
*nextp = pos;
return page;
}
*nextp = candidate;
return found_page;
}
/*
* map a memory block onto emu10k1's PTB
*
* call with memblk_lock held
*/
static int map_memblk(struct snd_emu10k1 *emu, struct snd_emu10k1_memblk *blk)
{
int page, pg;
struct list_head *next;
page = search_empty_map_area(emu, blk->pages, &next);
if (page < 0) /* not found */
return page;
if (page == 0) {
dev_err(emu->card->dev, "trying to map zero (reserved) page\n");
return -EINVAL;
}
/* insert this block in the proper position of mapped list */
list_add_tail(&blk->mapped_link, next);
/* append this as a newest block in order list */
list_add_tail(&blk->mapped_order_link, &emu->mapped_order_link_head);
blk->mapped_page = page;
/* fill PTB */
for (pg = blk->first_page; pg <= blk->last_page; pg++) {
set_ptb_entry(emu, page, emu->page_addr_table[pg]);
page++;
}
return 0;
}
/*
* unmap the block
* return the size of resultant empty pages
*
* call with memblk_lock held
*/
static int unmap_memblk(struct snd_emu10k1 *emu, struct snd_emu10k1_memblk *blk)
{
int start_page, end_page, mpage, pg;
struct list_head *p;
struct snd_emu10k1_memblk *q;
/* calculate the expected size of empty region */
if ((p = blk->mapped_link.prev) != &emu->mapped_link_head) {
q = get_emu10k1_memblk(p, mapped_link);
start_page = q->mapped_page + q->pages;
} else
start_page = 1;
if ((p = blk->mapped_link.next) != &emu->mapped_link_head) {
q = get_emu10k1_memblk(p, mapped_link);
end_page = q->mapped_page;
} else
end_page = (emu->address_mode ? MAX_ALIGN_PAGES1 : MAX_ALIGN_PAGES0);
/* remove links */
list_del(&blk->mapped_link);
list_del(&blk->mapped_order_link);
/* clear PTB */
mpage = blk->mapped_page;
for (pg = blk->first_page; pg <= blk->last_page; pg++) {
set_silent_ptb(emu, mpage);
mpage++;
}
blk->mapped_page = -1;
return end_page - start_page; /* return the new empty size */
}
/*
* search empty pages with the given size, and create a memory block
*
* unlike synth_alloc the memory block is aligned to the page start
*/
static struct snd_emu10k1_memblk *
search_empty(struct snd_emu10k1 *emu, int size)
{
struct list_head *p;
struct snd_emu10k1_memblk *blk;
int page, psize;
psize = get_aligned_page(size + PAGE_SIZE -1);
page = 0;
list_for_each(p, &emu->memhdr->block) {
blk = get_emu10k1_memblk(p, mem.list);
if (page + psize <= blk->first_page)
goto __found_pages;
page = blk->last_page + 1;
}
if (page + psize > emu->max_cache_pages)
return NULL;
__found_pages:
/* create a new memory block */
blk = (struct snd_emu10k1_memblk *)__snd_util_memblk_new(emu->memhdr, psize << PAGE_SHIFT, p->prev);
if (blk == NULL)
return NULL;
blk->mem.offset = aligned_page_offset(page); /* set aligned offset */
emu10k1_memblk_init(blk);
return blk;
}
/*
* check if the given pointer is valid for pages
*/
static int is_valid_page(struct snd_emu10k1 *emu, dma_addr_t addr)
{
if (addr & ~emu->dma_mask) {
dev_err(emu->card->dev,
"max memory size is 0x%lx (addr = 0x%lx)!!\n",
emu->dma_mask, (unsigned long)addr);
return 0;
}
if (addr & (EMUPAGESIZE-1)) {
dev_err(emu->card->dev, "page is not aligned\n");
return 0;
}
return 1;
}
/*
* map the given memory block on PTB.
* if the block is already mapped, update the link order.
* if no empty pages are found, tries to release unused memory blocks
* and retry the mapping.
*/
int snd_emu10k1_memblk_map(struct snd_emu10k1 *emu, struct snd_emu10k1_memblk *blk)
{
int err;
int size;
struct list_head *p, *nextp;
struct snd_emu10k1_memblk *deleted;
unsigned long flags;
spin_lock_irqsave(&emu->memblk_lock, flags);
if (blk->mapped_page >= 0) {
/* update order link */
list_move_tail(&blk->mapped_order_link,
&emu->mapped_order_link_head);
spin_unlock_irqrestore(&emu->memblk_lock, flags);
return 0;
}
if ((err = map_memblk(emu, blk)) < 0) {
/* no enough page - try to unmap some blocks */
/* starting from the oldest block */
p = emu->mapped_order_link_head.next;
for (; p != &emu->mapped_order_link_head; p = nextp) {
nextp = p->next;
deleted = get_emu10k1_memblk(p, mapped_order_link);
if (deleted->map_locked)
continue;
size = unmap_memblk(emu, deleted);
if (size >= blk->pages) {
/* ok the empty region is enough large */
err = map_memblk(emu, blk);
break;
}
}
}
spin_unlock_irqrestore(&emu->memblk_lock, flags);
return err;
}
EXPORT_SYMBOL(snd_emu10k1_memblk_map);
/*
* page allocation for DMA
*/
struct snd_util_memblk *
snd_emu10k1_alloc_pages(struct snd_emu10k1 *emu, struct snd_pcm_substream *substream)
{
struct snd_pcm_runtime *runtime = substream->runtime;
struct snd_util_memhdr *hdr;
struct snd_emu10k1_memblk *blk;
int page, err, idx;
if (snd_BUG_ON(!emu))
return NULL;
if (snd_BUG_ON(runtime->dma_bytes <= 0 ||
runtime->dma_bytes >= (emu->address_mode ? MAXPAGES1 : MAXPAGES0) * EMUPAGESIZE))
return NULL;
hdr = emu->memhdr;
if (snd_BUG_ON(!hdr))
return NULL;
idx = runtime->period_size >= runtime->buffer_size ?
(emu->delay_pcm_irq * 2) : 0;
mutex_lock(&hdr->block_mutex);
blk = search_empty(emu, runtime->dma_bytes + idx);
if (blk == NULL) {
mutex_unlock(&hdr->block_mutex);
return NULL;
}
/* fill buffer addresses but pointers are not stored so that
* snd_free_pci_page() is not called in in synth_free()
*/
idx = 0;
for (page = blk->first_page; page <= blk->last_page; page++, idx++) {
unsigned long ofs = idx << PAGE_SHIFT;
dma_addr_t addr;
if (ofs >= runtime->dma_bytes)
addr = emu->silent_page.addr;
else
addr = snd_pcm_sgbuf_get_addr(substream, ofs);
if (! is_valid_page(emu, addr)) {
dev_err(emu->card->dev,
"emu: failure page = %d\n", idx);
mutex_unlock(&hdr->block_mutex);
return NULL;
}
emu->page_addr_table[page] = addr;
emu->page_ptr_table[page] = NULL;
}
/* set PTB entries */
blk->map_locked = 1; /* do not unmap this block! */
err = snd_emu10k1_memblk_map(emu, blk);
if (err < 0) {
__snd_util_mem_free(hdr, (struct snd_util_memblk *)blk);
mutex_unlock(&hdr->block_mutex);
return NULL;
}
mutex_unlock(&hdr->block_mutex);
return (struct snd_util_memblk *)blk;
}
/*
* release DMA buffer from page table
*/
int snd_emu10k1_free_pages(struct snd_emu10k1 *emu, struct snd_util_memblk *blk)
{
if (snd_BUG_ON(!emu || !blk))
return -EINVAL;
return snd_emu10k1_synth_free(emu, blk);
}
/*
* allocate DMA pages, widening the allocation if necessary
*
* See the comment above snd_emu10k1_detect_iommu() in emu10k1_main.c why
* this might be needed.
*
* If you modify this function check whether __synth_free_pages() also needs
* changes.
*/
int snd_emu10k1_alloc_pages_maybe_wider(struct snd_emu10k1 *emu, size_t size,
struct snd_dma_buffer *dmab)
{
if (emu->iommu_workaround) {
size_t npages = (size + PAGE_SIZE - 1) / PAGE_SIZE;
size_t size_real = npages * PAGE_SIZE;
/*
* The device has been observed to accesses up to 256 extra
* bytes, but use 1k to be safe.
*/
if (size_real < size + 1024)
size += PAGE_SIZE;
}
return snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV,
snd_dma_pci_data(emu->pci), size, dmab);
}
/*
* memory allocation using multiple pages (for synth)
* Unlike the DMA allocation above, non-contiguous pages are assined.
*/
/*
* allocate a synth sample area
*/
struct snd_util_memblk *
snd_emu10k1_synth_alloc(struct snd_emu10k1 *hw, unsigned int size)
{
struct snd_emu10k1_memblk *blk;
struct snd_util_memhdr *hdr = hw->memhdr;
mutex_lock(&hdr->block_mutex);
blk = (struct snd_emu10k1_memblk *)__snd_util_mem_alloc(hdr, size);
if (blk == NULL) {
mutex_unlock(&hdr->block_mutex);
return NULL;
}
if (synth_alloc_pages(hw, blk)) {
__snd_util_mem_free(hdr, (struct snd_util_memblk *)blk);
mutex_unlock(&hdr->block_mutex);
return NULL;
}
snd_emu10k1_memblk_map(hw, blk);
mutex_unlock(&hdr->block_mutex);
return (struct snd_util_memblk *)blk;
}
EXPORT_SYMBOL(snd_emu10k1_synth_alloc);
/*
* free a synth sample area
*/
int
snd_emu10k1_synth_free(struct snd_emu10k1 *emu, struct snd_util_memblk *memblk)
{
struct snd_util_memhdr *hdr = emu->memhdr;
struct snd_emu10k1_memblk *blk = (struct snd_emu10k1_memblk *)memblk;
unsigned long flags;
mutex_lock(&hdr->block_mutex);
spin_lock_irqsave(&emu->memblk_lock, flags);
if (blk->mapped_page >= 0)
unmap_memblk(emu, blk);
spin_unlock_irqrestore(&emu->memblk_lock, flags);
synth_free_pages(emu, blk);
__snd_util_mem_free(hdr, memblk);
mutex_unlock(&hdr->block_mutex);
return 0;
}
EXPORT_SYMBOL(snd_emu10k1_synth_free);
/* check new allocation range */
static void get_single_page_range(struct snd_util_memhdr *hdr,
struct snd_emu10k1_memblk *blk,
int *first_page_ret, int *last_page_ret)
{
struct list_head *p;
struct snd_emu10k1_memblk *q;
int first_page, last_page;
first_page = blk->first_page;
if ((p = blk->mem.list.prev) != &hdr->block) {
q = get_emu10k1_memblk(p, mem.list);
if (q->last_page == first_page)
first_page++; /* first page was already allocated */
}
last_page = blk->last_page;
if ((p = blk->mem.list.next) != &hdr->block) {
q = get_emu10k1_memblk(p, mem.list);
if (q->first_page == last_page)
last_page--; /* last page was already allocated */
}
*first_page_ret = first_page;
*last_page_ret = last_page;
}
/* release allocated pages */
static void __synth_free_pages(struct snd_emu10k1 *emu, int first_page,
int last_page)
{
struct snd_dma_buffer dmab;
int page;
dmab.dev.type = SNDRV_DMA_TYPE_DEV;
dmab.dev.dev = snd_dma_pci_data(emu->pci);
for (page = first_page; page <= last_page; page++) {
if (emu->page_ptr_table[page] == NULL)
continue;
dmab.area = emu->page_ptr_table[page];
dmab.addr = emu->page_addr_table[page];
/*
* please keep me in sync with logic in
* snd_emu10k1_alloc_pages_maybe_wider()
*/
dmab.bytes = PAGE_SIZE;
if (emu->iommu_workaround)
dmab.bytes *= 2;
snd_dma_free_pages(&dmab);
emu->page_addr_table[page] = 0;
emu->page_ptr_table[page] = NULL;
}
}
/*
* allocate kernel pages
*/
static int synth_alloc_pages(struct snd_emu10k1 *emu, struct snd_emu10k1_memblk *blk)
{
int page, first_page, last_page;
struct snd_dma_buffer dmab;
emu10k1_memblk_init(blk);
get_single_page_range(emu->memhdr, blk, &first_page, &last_page);
/* allocate kernel pages */
for (page = first_page; page <= last_page; page++) {
if (snd_emu10k1_alloc_pages_maybe_wider(emu, PAGE_SIZE,
&dmab) < 0)
goto __fail;
if (!is_valid_page(emu, dmab.addr)) {
snd_dma_free_pages(&dmab);
goto __fail;
}
emu->page_addr_table[page] = dmab.addr;
emu->page_ptr_table[page] = dmab.area;
}
return 0;
__fail:
/* release allocated pages */
last_page = page - 1;
__synth_free_pages(emu, first_page, last_page);
return -ENOMEM;
}
/*
* free pages
*/
static int synth_free_pages(struct snd_emu10k1 *emu, struct snd_emu10k1_memblk *blk)
{
int first_page, last_page;
get_single_page_range(emu->memhdr, blk, &first_page, &last_page);
__synth_free_pages(emu, first_page, last_page);
return 0;
}
/* calculate buffer pointer from offset address */
static inline void *offset_ptr(struct snd_emu10k1 *emu, int page, int offset)
{
char *ptr;
if (snd_BUG_ON(page < 0 || page >= emu->max_cache_pages))
return NULL;
ptr = emu->page_ptr_table[page];
if (! ptr) {
dev_err(emu->card->dev,
"access to NULL ptr: page = %d\n", page);
return NULL;
}
ptr += offset & (PAGE_SIZE - 1);
return (void*)ptr;
}
/*
* bzero(blk + offset, size)
*/
int snd_emu10k1_synth_bzero(struct snd_emu10k1 *emu, struct snd_util_memblk *blk,
int offset, int size)
{
int page, nextofs, end_offset, temp, temp1;
void *ptr;
struct snd_emu10k1_memblk *p = (struct snd_emu10k1_memblk *)blk;
offset += blk->offset & (PAGE_SIZE - 1);
end_offset = offset + size;
page = get_aligned_page(offset);
do {
nextofs = aligned_page_offset(page + 1);
temp = nextofs - offset;
temp1 = end_offset - offset;
if (temp1 < temp)
temp = temp1;
ptr = offset_ptr(emu, page + p->first_page, offset);
if (ptr)
memset(ptr, 0, temp);
offset = nextofs;
page++;
} while (offset < end_offset);
return 0;
}
EXPORT_SYMBOL(snd_emu10k1_synth_bzero);
/*
* copy_from_user(blk + offset, data, size)
*/
int snd_emu10k1_synth_copy_from_user(struct snd_emu10k1 *emu, struct snd_util_memblk *blk,
int offset, const char __user *data, int size)
{
int page, nextofs, end_offset, temp, temp1;
void *ptr;
struct snd_emu10k1_memblk *p = (struct snd_emu10k1_memblk *)blk;
offset += blk->offset & (PAGE_SIZE - 1);
end_offset = offset + size;
page = get_aligned_page(offset);
do {
nextofs = aligned_page_offset(page + 1);
temp = nextofs - offset;
temp1 = end_offset - offset;
if (temp1 < temp)
temp = temp1;
ptr = offset_ptr(emu, page + p->first_page, offset);
if (ptr && copy_from_user(ptr, data, temp))
return -EFAULT;
offset = nextofs;
data += temp;
page++;
} while (offset < end_offset);
return 0;
}
EXPORT_SYMBOL(snd_emu10k1_synth_copy_from_user);