The size passed to dma_alloc_coherent() doesn't have to be aligned with power-of-two, rather it should be the raw size. As a minor optimization, remove the size adjustment in the current code.

Signed-off-by: Takashi Iwai tiwai@suse.de --- sound/core/memalloc.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-)

diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c index 753d5fc4b284..d85df01bf055 100644 --- a/sound/core/memalloc.c +++ b/sound/core/memalloc.c @@ -84,29 +84,24 @@ EXPORT_SYMBOL(snd_free_pages); /* allocate the coherent DMA pages */ static void *snd_malloc_dev_pages(struct device *dev, size_t size, dma_addr_t *dma) { - int pg; gfp_t gfp_flags;

if (WARN_ON(!dma)) return NULL; - pg = get_order(size); gfp_flags = GFP_KERNEL | __GFP_COMP /* compound page lets parts be mapped */ | __GFP_NORETRY /* don't trigger OOM-killer */ | __GFP_NOWARN; /* no stack trace print - this call is non-critical */ - return dma_alloc_coherent(dev, PAGE_SIZE << pg, dma, gfp_flags); + return dma_alloc_coherent(dev, size, dma, gfp_flags); }

/* free the coherent DMA pages */ static void snd_free_dev_pages(struct device *dev, size_t size, void *ptr, dma_addr_t dma) { - int pg; - if (ptr == NULL) return; - pg = get_order(size); - dma_free_coherent(dev, PAGE_SIZE << pg, ptr, dma); + dma_free_coherent(dev, size, ptr, dma); }

#ifdef CONFIG_GENERIC_ALLOCATOR

In some cases (mainly for x86), we need the DMA coherent buffer with non-cached pages. Although this has been done in each driver side like HD-audio and intel8x0, it can be done cleaner in the core memory allocator.

This patch adds the new types, SNDRV_DMA_TYPE_DEV_UC and SNDRV_DMA_TYPE_DEV_UC_SG, for allocating such non-cached buffer pages. On non-x86 architectures, they work as same as the standard SNDRV_DMA_TYPE_DEV and *_SG.

One additional change by this move is that we can assure to pass the non-cached pgprot to the vmapped buffer, too. It eventually fixes the case like non-snoop mode without mmap access on HD-audio.

Signed-off-by: Takashi Iwai tiwai@suse.de --- include/sound/memalloc.h | 3 +++ sound/core/memalloc.c | 17 +++++++++++++++++ sound/core/sgbuf.c | 15 +++++++++++++-- 3 files changed, 33 insertions(+), 2 deletions(-)

diff --git a/include/sound/memalloc.h b/include/sound/memalloc.h index 67561b997915..af3fa577fa06 100644 --- a/include/sound/memalloc.h +++ b/include/sound/memalloc.h @@ -47,10 +47,13 @@ struct snd_dma_device { #define SNDRV_DMA_TYPE_UNKNOWN 0 /* not defined */ #define SNDRV_DMA_TYPE_CONTINUOUS 1 /* continuous no-DMA memory */ #define SNDRV_DMA_TYPE_DEV 2 /* generic device continuous */ +#define SNDRV_DMA_TYPE_DEV_UC 5 /* continuous non-cahced */ #ifdef CONFIG_SND_DMA_SGBUF #define SNDRV_DMA_TYPE_DEV_SG 3 /* generic device SG-buffer */ +#define SNDRV_DMA_TYPE_DEV_UC_SG 6 /* SG non-cached */ #else #define SNDRV_DMA_TYPE_DEV_SG SNDRV_DMA_TYPE_DEV /* no SG-buf support */ +#define SNDRV_DMA_TYPE_DEV_UC_SG SNDRV_DMA_TYPE_DEV_UC #endif #ifdef CONFIG_GENERIC_ALLOCATOR #define SNDRV_DMA_TYPE_DEV_IRAM 4 /* generic device iram-buffer */ diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c index cc051bfe2f6f..aa266907ec9b 100644 --- a/sound/core/memalloc.c +++ b/sound/core/memalloc.c @@ -25,6 +25,9 @@ #include <linux/mm.h> #include <linux/dma-mapping.h> #include <linux/genalloc.h> +#ifdef CONFIG_X86 +#include <asm/set_memory.h> +#endif #include <sound/memalloc.h>

/* @@ -92,11 +95,21 @@ static void snd_malloc_dev_pages(struct snd_dma_buffer *dmab, size_t size) | __GFP_NOWARN; /* no stack trace print - this call is non-critical */ dmab->area = dma_alloc_coherent(dmab->dev.dev, size, &dmab->addr, gfp_flags); +#ifdef CONFIG_X86 + if (dmab->area && dmab->dev.type == SNDRV_DMA_TYPE_DEV_UC) + set_memory_wc((unsigned long)dmab->area, + PAGE_ALIGN(size) >> PAGE_SHIFT); +#endif }

/* free the coherent DMA pages */ static void snd_free_dev_pages(struct snd_dma_buffer *dmab) { +#ifdef CONFIG_X86 + if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_UC) + set_memory_wb((unsigned long)dmab->area, + PAGE_ALIGN(dmab->bytes) >> PAGE_SHIFT); +#endif dma_free_coherent(dmab->dev.dev, dmab->bytes, dmab->area, dmab->addr); }

@@ -191,11 +204,13 @@ int snd_dma_alloc_pages(int type, struct device *device, size_t size, dmab->dev.type = SNDRV_DMA_TYPE_DEV; #endif /* CONFIG_GENERIC_ALLOCATOR */ case SNDRV_DMA_TYPE_DEV: + case SNDRV_DMA_TYPE_DEV_UC: snd_malloc_dev_pages(dmab, size); break; #endif #ifdef CONFIG_SND_DMA_SGBUF case SNDRV_DMA_TYPE_DEV_SG: + case SNDRV_DMA_TYPE_DEV_UC_SG: snd_malloc_sgbuf_pages(device, size, dmab, NULL); break; #endif @@ -266,11 +281,13 @@ void snd_dma_free_pages(struct snd_dma_buffer *dmab) break; #endif /* CONFIG_GENERIC_ALLOCATOR */ case SNDRV_DMA_TYPE_DEV: + case SNDRV_DMA_TYPE_DEV_UC: snd_free_dev_pages(dmab); break; #endif #ifdef CONFIG_SND_DMA_SGBUF case SNDRV_DMA_TYPE_DEV_SG: + case SNDRV_DMA_TYPE_DEV_UC_SG: snd_free_sgbuf_pages(dmab); break; #endif diff --git a/sound/core/sgbuf.c b/sound/core/sgbuf.c index 84fffabdd129..c1cfaa01a5cb 100644 --- a/sound/core/sgbuf.c +++ b/sound/core/sgbuf.c @@ -23,6 +23,7 @@ #include <linux/mm.h> #include <linux/vmalloc.h> #include <linux/export.h> +#include <asm/pgtable.h> #include <sound/memalloc.h>

@@ -43,6 +44,8 @@ int snd_free_sgbuf_pages(struct snd_dma_buffer *dmab) dmab->area = NULL;

tmpb.dev.type = SNDRV_DMA_TYPE_DEV; + if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_UC_SG) + tmpb.dev.type = SNDRV_DMA_TYPE_DEV_UC; tmpb.dev.dev = sgbuf->dev; for (i = 0; i < sgbuf->pages; i++) { if (!(sgbuf->table[i].addr & ~PAGE_MASK)) @@ -72,12 +75,20 @@ void *snd_malloc_sgbuf_pages(struct device *device, struct snd_dma_buffer tmpb; struct snd_sg_page *table; struct page **pgtable; + int type = SNDRV_DMA_TYPE_DEV; + pgprot_t prot = PAGE_KERNEL;

dmab->area = NULL; dmab->addr = 0; dmab->private_data = sgbuf = kzalloc(sizeof(*sgbuf), GFP_KERNEL); if (! sgbuf) return NULL; + if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_UC_SG) { + type = SNDRV_DMA_TYPE_DEV_UC; +#ifdef pgprot_noncached + prot = pgprot_noncached(PAGE_KERNEL); +#endif + } sgbuf->dev = device; pages = snd_sgbuf_aligned_pages(size); sgbuf->tblsize = sgbuf_align_table(pages); @@ -98,7 +109,7 @@ void *snd_malloc_sgbuf_pages(struct device *device, if (chunk > maxpages) chunk = maxpages; chunk <<= PAGE_SHIFT; - if (snd_dma_alloc_pages_fallback(SNDRV_DMA_TYPE_DEV, device, + if (snd_dma_alloc_pages_fallback(type, device, chunk, &tmpb) < 0) { if (!sgbuf->pages) goto _failed; @@ -125,7 +136,7 @@ void *snd_malloc_sgbuf_pages(struct device *device, }

sgbuf->size = size; - dmab->area = vmap(sgbuf->page_table, sgbuf->pages, VM_MAP, PAGE_KERNEL); + dmab->area = vmap(sgbuf->page_table, sgbuf->pages, VM_MAP, prot); if (! dmab->area) goto _failed; if (res_size)

Now the ALSA memory allocator helper supports the new non-cached pages, let's use the new type, SNDRV_DMA_TYPE_DEV_UC_SG, for HD-audio driver. This allows us to reduce lots of codes.

As another positive side-effect by this patch, the long-standing issue with non-snoop mode playing in the non-mmap mode is fixed. The core memalloc helper does the proper pgprot setup for non-cached pages for vmap(), which was missing in the past.

Reported-and-tested-by: Hans Hu HansHu@zhaoxin.com Signed-off-by: Takashi Iwai tiwai@suse.de --- sound/pci/hda/hda_controller.c | 5 ++- sound/pci/hda/hda_controller.h | 1 - sound/pci/hda/hda_intel.c | 81 ++-------------------------------- 3 files changed, 8 insertions(+), 79 deletions(-)

diff --git a/sound/pci/hda/hda_controller.c b/sound/pci/hda/hda_controller.c index a12e594d4e3b..8bc46676c278 100644 --- a/sound/pci/hda/hda_controller.c +++ b/sound/pci/hda/hda_controller.c @@ -732,6 +732,7 @@ int snd_hda_attach_pcm_stream(struct hda_bus *_bus, struct hda_codec *codec, int pcm_dev = cpcm->device; unsigned int size; int s, err; + int type = SNDRV_DMA_TYPE_DEV_SG;

list_for_each_entry(apcm, &chip->pcm_list, list) { if (apcm->pcm->device == pcm_dev) { @@ -770,7 +771,9 @@ int snd_hda_attach_pcm_stream(struct hda_bus *_bus, struct hda_codec *codec, size = CONFIG_SND_HDA_PREALLOC_SIZE * 1024; if (size > MAX_PREALLOC_SIZE) size = MAX_PREALLOC_SIZE; - snd_pcm_lib_preallocate_pages_for_all(pcm, SNDRV_DMA_TYPE_DEV_SG, + if (chip->uc_buffer) + type = SNDRV_DMA_TYPE_DEV_UC_SG; + snd_pcm_lib_preallocate_pages_for_all(pcm, type, chip->card->dev, size, MAX_PREALLOC_SIZE); return 0; diff --git a/sound/pci/hda/hda_controller.h b/sound/pci/hda/hda_controller.h index 53c3cd28bc99..c4b3de6c42b8 100644 --- a/sound/pci/hda/hda_controller.h +++ b/sound/pci/hda/hda_controller.h @@ -76,7 +76,6 @@ struct azx_dev { * when link position is not greater than FIFO size */ unsigned int insufficient:1; - unsigned int wc_marked:1; };

#define azx_stream(dev) (&(dev)->core) diff --git a/sound/pci/hda/hda_intel.c b/sound/pci/hda/hda_intel.c index e5482fd21568..03ddd1abb44d 100644 --- a/sound/pci/hda/hda_intel.c +++ b/sound/pci/hda/hda_intel.c @@ -397,61 +397,6 @@ static char *driver_short_names[] = { [AZX_DRIVER_GENERIC] = "HD-Audio Generic", };

-#ifdef CONFIG_X86 -static void __mark_pages_wc(struct azx *chip, struct snd_dma_buffer *dmab, bool on) -{ - int pages; - - if (azx_snoop(chip)) - return; - if (!dmab || !dmab->area || !dmab->bytes) - return; - -#ifdef CONFIG_SND_DMA_SGBUF - if (dmab->dev.type == SNDRV_DMA_TYPE_DEV_SG) { - struct snd_sg_buf *sgbuf = dmab->private_data; - if (!chip->uc_buffer) - return; /* deal with only CORB/RIRB buffers */ - if (on) - set_pages_array_wc(sgbuf->page_table, sgbuf->pages); - else - set_pages_array_wb(sgbuf->page_table, sgbuf->pages); - return; - } -#endif - - pages = (dmab->bytes + PAGE_SIZE - 1) >> PAGE_SHIFT; - if (on) - set_memory_wc((unsigned long)dmab->area, pages); - else - set_memory_wb((unsigned long)dmab->area, pages); -} - -static inline void mark_pages_wc(struct azx *chip, struct snd_dma_buffer *buf, - bool on) -{ - __mark_pages_wc(chip, buf, on); -} -static inline void mark_runtime_wc(struct azx *chip, struct azx_dev *azx_dev, - struct snd_pcm_substream *substream, bool on) -{ - if (azx_dev->wc_marked != on) { - __mark_pages_wc(chip, snd_pcm_get_dma_buf(substream), on); - azx_dev->wc_marked = on; - } -} -#else -/* NOP for other archs */ -static inline void mark_pages_wc(struct azx *chip, struct snd_dma_buffer *buf, - bool on) -{ -} -static inline void mark_runtime_wc(struct azx *chip, struct azx_dev *azx_dev, - struct snd_pcm_substream *substream, bool on) -{ -} -#endif - static int azx_acquire_irq(struct azx *chip, int do_disconnect);

/* @@ -2054,22 +1999,14 @@ static int dma_alloc_pages(struct hdac_bus *bus, struct snd_dma_buffer *buf) { struct azx *chip = bus_to_azx(bus); - int err;

- err = snd_dma_alloc_pages(type, - bus->dev, - size, buf); - if (err < 0) - return err; - mark_pages_wc(chip, buf, true); - return 0; + if (!azx_snoop(chip) && type == SNDRV_DMA_TYPE_DEV) + type = SNDRV_DMA_TYPE_DEV_UC; + return snd_dma_alloc_pages(type, bus->dev, size, buf); }

static void dma_free_pages(struct hdac_bus *bus, struct snd_dma_buffer *buf) { - struct azx *chip = bus_to_azx(bus); - - mark_pages_wc(chip, buf, false); snd_dma_free_pages(buf); }

@@ -2077,22 +2014,12 @@ static int substream_alloc_pages(struct azx *chip, struct snd_pcm_substream *substream, size_t size) { - struct azx_dev *azx_dev = get_azx_dev(substream); - int ret; - - mark_runtime_wc(chip, azx_dev, substream, false); - ret = snd_pcm_lib_malloc_pages(substream, size); - if (ret < 0) - return ret; - mark_runtime_wc(chip, azx_dev, substream, true); - return 0; + return snd_pcm_lib_malloc_pages(substream, size); }

static int substream_free_pages(struct azx *chip, struct snd_pcm_substream *substream) { - struct azx_dev *azx_dev = get_azx_dev(substream); - mark_runtime_wc(chip, azx_dev, substream, false); return snd_pcm_lib_free_pages(substream); }

intel8x0 driver requires the non-cached pages for 440MX workaround, and this can be implemented more easily with the new memalloc type, SNDRV_DMA_TYPE_DEV_UC. This allows us to reduce lots of code.

Signed-off-by: Takashi Iwai tiwai@suse.de --- sound/pci/intel8x0.c | 89 ++++++-------------------------------------- 1 file changed, 12 insertions(+), 77 deletions(-)

diff --git a/sound/pci/intel8x0.c b/sound/pci/intel8x0.c index 5ee468d1aefe..9517f9b8f1d4 100644 --- a/sound/pci/intel8x0.c +++ b/sound/pci/intel8x0.c @@ -38,11 +38,6 @@ #include <sound/ac97_codec.h> #include <sound/info.h> #include <sound/initval.h> -/* for 440MX workaround */ -#include <asm/pgtable.h> -#ifdef CONFIG_X86 -#include <asm/set_memory.h> -#endif

MODULE_AUTHOR("Jaroslav Kysela perex@perex.cz"); MODULE_DESCRIPTION("Intel 82801AA,82901AB,i810,i820,i830,i840,i845,MX440; SiS 7012; Ali 5455"); @@ -374,7 +369,6 @@ struct ichdev { unsigned int ali_slot; /* ALI DMA slot */ struct ac97_pcm *pcm; int pcm_open_flag; - unsigned int page_attr_changed: 1; unsigned int suspended: 1; };

@@ -724,25 +718,6 @@ static void snd_intel8x0_setup_periods(struct intel8x0 *chip, struct ichdev *ich iputbyte(chip, port + ichdev->roff_sr, ICH_FIFOE | ICH_BCIS | ICH_LVBCI); }

-#ifdef __i386__ -/* - * Intel 82443MX running a 100MHz processor system bus has a hardware bug, - * which aborts PCI busmaster for audio transfer. A workaround is to set - * the pages as non-cached. For details, see the errata in - * http://download.intel.com/design/chipsets/specupdt/24505108.pdf - */ -static void fill_nocache(void *buf, int size, int nocache) -{ - size = (size + PAGE_SIZE - 1) >> PAGE_SHIFT; - if (nocache) - set_pages_uc(virt_to_page(buf), size); - else - set_pages_wb(virt_to_page(buf), size); -} -#else -#define fill_nocache(buf, size, nocache) do { ; } while (0) -#endif - /* * Interrupt handler */ @@ -938,23 +913,12 @@ static int snd_intel8x0_hw_params(struct snd_pcm_substream *substream, { struct intel8x0 *chip = snd_pcm_substream_chip(substream); struct ichdev *ichdev = get_ichdev(substream); - struct snd_pcm_runtime *runtime = substream->runtime; int dbl = params_rate(hw_params) > 48000; int err;

- if (chip->fix_nocache && ichdev->page_attr_changed) { - fill_nocache(runtime->dma_area, runtime->dma_bytes, 0); /* clear */ - ichdev->page_attr_changed = 0; - } err = snd_pcm_lib_malloc_pages(substream, params_buffer_bytes(hw_params)); if (err < 0) return err; - if (chip->fix_nocache) { - if (runtime->dma_area && ! ichdev->page_attr_changed) { - fill_nocache(runtime->dma_area, runtime->dma_bytes, 1); - ichdev->page_attr_changed = 1; - } - } if (ichdev->pcm_open_flag) { snd_ac97_pcm_close(ichdev->pcm); ichdev->pcm_open_flag = 0; @@ -974,17 +938,12 @@ static int snd_intel8x0_hw_params(struct snd_pcm_substream *substream,

static int snd_intel8x0_hw_free(struct snd_pcm_substream *substream) { - struct intel8x0 *chip = snd_pcm_substream_chip(substream); struct ichdev *ichdev = get_ichdev(substream);

if (ichdev->pcm_open_flag) { snd_ac97_pcm_close(ichdev->pcm); ichdev->pcm_open_flag = 0; } - if (chip->fix_nocache && ichdev->page_attr_changed) { - fill_nocache(substream->runtime->dma_area, substream->runtime->dma_bytes, 0); - ichdev->page_attr_changed = 0; - } return snd_pcm_lib_free_pages(substream); }

@@ -1510,6 +1469,9 @@ struct ich_pcm_table { int ac97_idx; };

+#define intel8x0_dma_type(chip) \ + ((chip)->fix_nocache ? SNDRV_DMA_TYPE_DEV_UC : SNDRV_DMA_TYPE_DEV) + static int snd_intel8x0_pcm1(struct intel8x0 *chip, int device, struct ich_pcm_table *rec) { @@ -1540,7 +1502,7 @@ static int snd_intel8x0_pcm1(struct intel8x0 *chip, int device, strcpy(pcm->name, chip->card->shortname); chip->pcm[device] = pcm;

- snd_pcm_lib_preallocate_pages_for_all(pcm, SNDRV_DMA_TYPE_DEV, + snd_pcm_lib_preallocate_pages_for_all(pcm, intel8x0_dma_type(chip), snd_dma_pci_data(chip->pci), rec->prealloc_size, rec->prealloc_max_size);

@@ -2629,11 +2591,8 @@ static int snd_intel8x0_free(struct intel8x0 *chip) __hw_end: if (chip->irq >= 0) free_irq(chip->irq, chip); - if (chip->bdbars.area) { - if (chip->fix_nocache) - fill_nocache(chip->bdbars.area, chip->bdbars.bytes, 0); + if (chip->bdbars.area) snd_dma_free_pages(&chip->bdbars); - } if (chip->addr) pci_iounmap(chip->pci, chip->addr); if (chip->bmaddr) @@ -2657,17 +2616,6 @@ static int intel8x0_suspend(struct device *dev) snd_power_change_state(card, SNDRV_CTL_POWER_D3hot); for (i = 0; i < chip->pcm_devs; i++) snd_pcm_suspend_all(chip->pcm[i]); - /* clear nocache */ - if (chip->fix_nocache) { - for (i = 0; i < chip->bdbars_count; i++) { - struct ichdev *ichdev = &chip->ichd[i]; - if (ichdev->substream && ichdev->page_attr_changed) { - struct snd_pcm_runtime *runtime = ichdev->substream->runtime; - if (runtime->dma_area) - fill_nocache(runtime->dma_area, runtime->dma_bytes, 0); - } - } - } for (i = 0; i < chip->ncodecs; i++) snd_ac97_suspend(chip->ac97[i]); if (chip->device_type == DEVICE_INTEL_ICH4) @@ -2708,25 +2656,9 @@ static int intel8x0_resume(struct device *dev) ICH_PCM_SPDIF_1011); }

- /* refill nocache */ - if (chip->fix_nocache) - fill_nocache(chip->bdbars.area, chip->bdbars.bytes, 1); - for (i = 0; i < chip->ncodecs; i++) snd_ac97_resume(chip->ac97[i]);

- /* refill nocache */ - if (chip->fix_nocache) { - for (i = 0; i < chip->bdbars_count; i++) { - struct ichdev *ichdev = &chip->ichd[i]; - if (ichdev->substream && ichdev->page_attr_changed) { - struct snd_pcm_runtime *runtime = ichdev->substream->runtime; - if (runtime->dma_area) - fill_nocache(runtime->dma_area, runtime->dma_bytes, 1); - } - } - } - /* resume status */ for (i = 0; i < chip->bdbars_count; i++) { struct ichdev *ichdev = &chip->ichd[i]; @@ -3057,6 +2989,12 @@ static int snd_intel8x0_create(struct snd_card *card,

chip->inside_vm = snd_intel8x0_inside_vm(pci);

+ /* + * Intel 82443MX running a 100MHz processor system bus has a hardware + * bug, which aborts PCI busmaster for audio transfer. A workaround + * is to set the pages as non-cached. For details, see the errata in + * http://download.intel.com/design/chipsets/specupdt/24505108.pdf + */ if (pci->vendor == PCI_VENDOR_ID_INTEL && pci->device == PCI_DEVICE_ID_INTEL_440MX) chip->fix_nocache = 1; /* enable workaround */ @@ -3128,7 +3066,7 @@ static int snd_intel8x0_create(struct snd_card *card,

/* allocate buffer descriptor lists */ /* the start of each lists must be aligned to 8 bytes */ - if (snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, snd_dma_pci_data(pci), + if (snd_dma_alloc_pages(intel8x0_dma_type(chip), snd_dma_pci_data(pci), chip->bdbars_count * sizeof(u32) * ICH_MAX_FRAGS * 2, &chip->bdbars) < 0) { snd_intel8x0_free(chip); @@ -3137,9 +3075,6 @@ static int snd_intel8x0_create(struct snd_card *card, } /* tables must be aligned to 8 bytes here, but the kernel pages are much bigger, so we don't care (on i386) */ - /* workaround for 440MX */ - if (chip->fix_nocache) - fill_nocache(chip->bdbars.area, chip->bdbars.bytes, 1); int_sta_masks = 0; for (i = 0; i < chip->bdbars_count; i++) { ichdev = &chip->ichd[i];