Skip to contentSkip to conversation

ASoC: fsl_dma: Pass the proper device for dma mapping routines

PATCH Applied Anton Vorontsov 5 participants 8 messages Apr 4, 2009 – Apr 6, 2009

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ASoC: fsl_dma: Pass the proper device for dma mapping routines

The driver should pass a device that specifies internal DMA ops, but substream->pcm is just a logical device, and thus doesn't have arch- specific dma callbacks, therefore following bug appears:

Freescale Synchronous Serial Interface (SSI) ASoC Driver

  ------------[ cut here ]------------
  kernel BUG at arch/powerpc/include/asm/dma-mapping.h:237!
  Oops: Exception in kernel mode, sig: 5 [#1]
  ...
  NIP [c02259c4] snd_malloc_dev_pages+0x58/0xac
  LR [c0225c74] snd_dma_alloc_pages+0xf8/0x108
  Call Trace:
  [df02bde0] [df02be2c] 0xdf02be2c (unreliable)
  [df02bdf0] [c0225c74] snd_dma_alloc_pages+0xf8/0x108
  [df02be10] [c023a100] fsl_dma_new+0x68/0x124
  [df02be20] [c02342ac] soc_new_pcm+0x1bc/0x234
  [df02bea0] [c02343dc] snd_soc_new_pcms+0xb8/0x148
  [df02bed0] [c023824c] cs4270_probe+0x34/0x124
  [df02bef0] [c0232fe8] snd_soc_instantiate_card+0x1a4/0x2f4
  [df02bf20] [c0233164] snd_soc_instantiate_cards+0x2c/0x68
  [df02bf30] [c0234704] snd_soc_register_platform+0x60/0x80
  [df02bf50] [c03d5664] fsl_soc_platform_init+0x18/0x28
  ...

This patch fixes the issue by using card's device instead.

---
 sound/soc/fsl/fsl_dma.c |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)
sound/soc/fsl/fsl_dma.c +6 -6
index b3eb857..835e43c 100644
@@ -300,7 +300,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
 	if (!card->dev->coherent_dma_mask)
 		card->dev->coherent_dma_mask = fsl_dma_dmamask;
 
-	ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+	ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
 		fsl_dma_hardware.buffer_bytes_max,
 		&pcm->streams[0].substream->dma_buffer);
 	if (ret) {
@@ -310,7 +310,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
 		return -ENOMEM;
 	}
 
-	ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+	ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
 		fsl_dma_hardware.buffer_bytes_max,
 		&pcm->streams[1].substream->dma_buffer);
 	if (ret) {
@@ -418,7 +418,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
 		return -EBUSY;
 	}
 
-	dma_private = dma_alloc_coherent(substream->pcm->dev,
+	dma_private = dma_alloc_coherent(substream->pcm->card->dev,
 		sizeof(struct fsl_dma_private), &ld_buf_phys, GFP_KERNEL);
 	if (!dma_private) {
 		dev_err(substream->pcm->card->dev,
@@ -445,7 +445,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
 		dev_err(substream->pcm->card->dev,
 			"can't register ISR for IRQ %u (ret=%i)\n",
 			dma_private->irq, ret);
-		dma_free_coherent(substream->pcm->dev,
+		dma_free_coherent(substream->pcm->card->dev,
 			sizeof(struct fsl_dma_private),
 			dma_private, dma_private->ld_buf_phys);
 		return ret;
@@ -761,13 +761,13 @@ static int fsl_dma_close(struct snd_pcm_substream *substream)
 			free_irq(dma_private->irq, dma_private);
 
 		if (dma_private->ld_buf_phys) {
-			dma_unmap_single(substream->pcm->dev,
+			dma_unmap_single(substream->pcm->card->dev,
 				dma_private->ld_buf_phys,
 				sizeof(dma_private->link), DMA_TO_DEVICE);
 		}
 
 		/* Deallocate the fsl_dma_private structure */
-		dma_free_coherent(substream->pcm->dev,
+		dma_free_coherent(substream->pcm->card->dev,
 			sizeof(struct fsl_dma_private),
 			dma_private, dma_private->ld_buf_phys);
 		substream->runtime->private_data = NULL;
Signature · 1 line
1.5.6.5
#1 lore raw

Becky & Kumar,

Considering these fixups, would it be advisable for the dma functions to climb up the dev->parent linkage when dma_ops are NULL?

g.

Quoted 94 lines · Anton Vorontsov

The driver should pass a device that specifies internal DMA ops, but substream->pcm is just a logical device, and thus doesn't have arch- specific dma callbacks, therefore following bug appears:

 Freescale Synchronous Serial Interface (SSI) ASoC Driver

 ------------[ cut here ]------------
 kernel BUG at arch/powerpc/include/asm/dma-mapping.h:237!
 Oops: Exception in kernel mode, sig: 5 [#1]
 ...
 NIP [c02259c4] snd_malloc_dev_pages+0x58/0xac
 LR [c0225c74] snd_dma_alloc_pages+0xf8/0x108
 Call Trace:

 [df02bde0] [df02be2c] 0xdf02be2c (unreliable) [df02bdf0] [c0225c74] snd_dma_alloc_pages+0xf8/0x108 [df02be10] [c023a100] fsl_dma_new+0x68/0x124 [df02be20] [c02342ac] soc_new_pcm+0x1bc/0x234 [df02bea0] [c02343dc] snd_soc_new_pcms+0xb8/0x148 [df02bed0] [c023824c] cs4270_probe+0x34/0x124 [df02bef0] [c0232fe8] snd_soc_instantiate_card+0x1a4/0x2f4 [df02bf20] [c0233164] snd_soc_instantiate_cards+0x2c/0x68 [df02bf30] [c0234704] snd_soc_register_platform+0x60/0x80 [df02bf50] [c03d5664] fsl_soc_platform_init+0x18/0x28
 ...

This patch fixes the issue by using card's device instead.

---
 sound/soc/fsl/fsl_dma.c |   12 ++++++------

 1 files changed, 6 insertions(+), 6 deletions(-)
sound/soc/fsl/fsl_dma.c +6 -7
index b3eb857..835e43c 100644
@@ -300,7 +300,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
        if (!card->dev->coherent_dma_mask)
                card->dev->coherent_dma_mask = fsl_dma_dmamask;
 
-       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
                fsl_dma_hardware.buffer_bytes_max,
                &pcm->streams[0].substream->dma_buffer);
        if (ret) {
@@ -310,7 +310,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
                return -ENOMEM;
        }
 
-       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
                fsl_dma_hardware.buffer_bytes_max,
                &pcm->streams[1].substream->dma_buffer);
        if (ret) {
@@ -418,7 +418,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
                return -EBUSY;
        }
 
-       dma_private = dma_alloc_coherent(substream->pcm->dev,
+       dma_private = dma_alloc_coherent(substream->pcm->card->dev,
                sizeof(struct fsl_dma_private), &ld_buf_phys, GFP_KERNEL);
        if (!dma_private) {
                dev_err(substream->pcm->card->dev,
@@ -445,7 +445,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
                dev_err(substream->pcm->card->dev,
                        "can't register ISR for IRQ %u (ret=%i)\n",
                        dma_private->irq, ret);
-               dma_free_coherent(substream->pcm->dev,
+               dma_free_coherent(substream->pcm->card->dev,
                        sizeof(struct fsl_dma_private),
                        dma_private, dma_private->ld_buf_phys);
                return ret;
@@ -761,13 +761,13 @@ static int fsl_dma_close(struct snd_pcm_substream *substream)
                        free_irq(dma_private->irq, dma_private);
 
                if (dma_private->ld_buf_phys) {
-                       dma_unmap_single(substream->pcm->dev,
+                       dma_unmap_single(substream->pcm->card->dev,
                                dma_private->ld_buf_phys,
                                sizeof(dma_private->link), DMA_TO_DEVICE);
                }
 
                /* Deallocate the fsl_dma_private structure */
-               dma_free_coherent(substream->pcm->dev,
+               dma_free_coherent(substream->pcm->card->dev,
                        sizeof(struct fsl_dma_private),
                        dma_private, dma_private->ld_buf_phys);
                substream->runtime->private_data = NULL;
--

1.5.6.5
_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev

Signature · 2 lines
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.
#2 lore raw
Quoted 4 lines · Grant Likely

Becky & Kumar,

Considering these fixups, would it be advisable for the dma functions to climb up the dev->parent linkage when dma_ops are NULL?

pcm->dev->parent doesn't point to the proper device, so this won't work.

Quoted 102 lines
Quoted 94 lines · Anton Vorontsov

The driver should pass a device that specifies internal DMA ops, but substream->pcm is just a logical device, and thus doesn't have arch- specific dma callbacks, therefore following bug appears:

 Freescale Synchronous Serial Interface (SSI) ASoC Driver

 ------------[ cut here ]------------
 kernel BUG at arch/powerpc/include/asm/dma-mapping.h:237!
 Oops: Exception in kernel mode, sig: 5 [#1]
 ...
 NIP [c02259c4] snd_malloc_dev_pages+0x58/0xac
 LR [c0225c74] snd_dma_alloc_pages+0xf8/0x108
 Call Trace:

 [df02bde0] [df02be2c] 0xdf02be2c (unreliable) [df02bdf0] [c0225c74] snd_dma_alloc_pages+0xf8/0x108 [df02be10] [c023a100] fsl_dma_new+0x68/0x124 [df02be20] [c02342ac] soc_new_pcm+0x1bc/0x234 [df02bea0] [c02343dc] snd_soc_new_pcms+0xb8/0x148 [df02bed0] [c023824c] cs4270_probe+0x34/0x124 [df02bef0] [c0232fe8] snd_soc_instantiate_card+0x1a4/0x2f4 [df02bf20] [c0233164] snd_soc_instantiate_cards+0x2c/0x68 [df02bf30] [c0234704] snd_soc_register_platform+0x60/0x80 [df02bf50] [c03d5664] fsl_soc_platform_init+0x18/0x28
 ...

This patch fixes the issue by using card's device instead.

---
 sound/soc/fsl/fsl_dma.c |   12 ++++++------

 1 files changed, 6 insertions(+), 6 deletions(-)
sound/soc/fsl/fsl_dma.c +6 -7
index b3eb857..835e43c 100644
@@ -300,7 +300,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
        if (!card->dev->coherent_dma_mask)
                card->dev->coherent_dma_mask = fsl_dma_dmamask;
 
-       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
                fsl_dma_hardware.buffer_bytes_max,
                &pcm->streams[0].substream->dma_buffer);
        if (ret) {
@@ -310,7 +310,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
                return -ENOMEM;
        }
 
-       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
                fsl_dma_hardware.buffer_bytes_max,
                &pcm->streams[1].substream->dma_buffer);
        if (ret) {
@@ -418,7 +418,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
                return -EBUSY;
        }
 
-       dma_private = dma_alloc_coherent(substream->pcm->dev,
+       dma_private = dma_alloc_coherent(substream->pcm->card->dev,
                sizeof(struct fsl_dma_private), &ld_buf_phys, GFP_KERNEL);
        if (!dma_private) {
                dev_err(substream->pcm->card->dev,
@@ -445,7 +445,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
                dev_err(substream->pcm->card->dev,
                        "can't register ISR for IRQ %u (ret=%i)\n",
                        dma_private->irq, ret);
-               dma_free_coherent(substream->pcm->dev,
+               dma_free_coherent(substream->pcm->card->dev,
                        sizeof(struct fsl_dma_private),
                        dma_private, dma_private->ld_buf_phys);
                return ret;
@@ -761,13 +761,13 @@ static int fsl_dma_close(struct snd_pcm_substream *substream)
                        free_irq(dma_private->irq, dma_private);
 
                if (dma_private->ld_buf_phys) {
-                       dma_unmap_single(substream->pcm->dev,
+                       dma_unmap_single(substream->pcm->card->dev,
                                dma_private->ld_buf_phys,
                                sizeof(dma_private->link), DMA_TO_DEVICE);
                }
 
                /* Deallocate the fsl_dma_private structure */
-               dma_free_coherent(substream->pcm->dev,
+               dma_free_coherent(substream->pcm->card->dev,
                        sizeof(struct fsl_dma_private),
                        dma_private, dma_private->ld_buf_phys);
                substream->runtime->private_data = NULL;
--

1.5.6.5
_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev

Signature · 2 lines
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.
Signature · 3 lines
Anton Vorontsov
email: [email protected]
irc://irc.freenode.net/bd2
#3 lore raw
Quoted 8 lines · Anton Vorontsov
Quoted 4 lines · Grant Likely

Becky & Kumar,

Considering these fixups, would it be advisable for the dma functions to climb up the dev->parent linkage when dma_ops are NULL?

pcm->dev->parent doesn't point to the proper device, so this won't work.

What does it point to then? IIUC, following the ->parent links should eventually get to a struct device that represents the platform or of_platform device for purposes of power management.

BTW, I'm not NAKing your patch. The patch looks fine to me.

g.

Quoted 109 lines
Quoted 95 lines
Quoted 92 lines · Anton Vorontsov

The driver should pass a device that specifies internal DMA ops, but substream->pcm is just a logical device, and thus doesn't have arch- specific dma callbacks, therefore following bug appears:

 Freescale Synchronous Serial Interface (SSI) ASoC Driver

 ------------[ cut here ]------------
 kernel BUG at arch/powerpc/include/asm/dma-mapping.h:237!
 Oops: Exception in kernel mode, sig: 5 [#1]
 ...
 NIP [c02259c4] snd_malloc_dev_pages+0x58/0xac
 LR [c0225c74] snd_dma_alloc_pages+0xf8/0x108
 Call Trace:

 [df02bde0] [df02be2c] 0xdf02be2c (unreliable) [df02bdf0] [c0225c74] snd_dma_alloc_pages+0xf8/0x108 [df02be10] [c023a100] fsl_dma_new+0x68/0x124 [df02be20] [c02342ac] soc_new_pcm+0x1bc/0x234 [df02bea0] [c02343dc] snd_soc_new_pcms+0xb8/0x148 [df02bed0] [c023824c] cs4270_probe+0x34/0x124 [df02bef0] [c0232fe8] snd_soc_instantiate_card+0x1a4/0x2f4 [df02bf20] [c0233164] snd_soc_instantiate_cards+0x2c/0x68 [df02bf30] [c0234704] snd_soc_register_platform+0x60/0x80 [df02bf50] [c03d5664] fsl_soc_platform_init+0x18/0x28
 ...

This patch fixes the issue by using card's device instead.

---
 sound/soc/fsl/fsl_dma.c |   12 ++++++------

 1 files changed, 6 insertions(+), 6 deletions(-)
sound/soc/fsl/fsl_dma.c +6 -7
index b3eb857..835e43c 100644
@@ -300,7 +300,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
        if (!card->dev->coherent_dma_mask)
                card->dev->coherent_dma_mask = fsl_dma_dmamask;
 
-       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
                fsl_dma_hardware.buffer_bytes_max,
                &pcm->streams[0].substream->dma_buffer);
        if (ret) {
@@ -310,7 +310,7 @@ static int fsl_dma_new(struct snd_card *card, struct snd_soc_dai *dai,
                return -ENOMEM;
        }
 
-       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, pcm->dev,
+       ret = snd_dma_alloc_pages(SNDRV_DMA_TYPE_DEV, card->dev,
                fsl_dma_hardware.buffer_bytes_max,
                &pcm->streams[1].substream->dma_buffer);
        if (ret) {
@@ -418,7 +418,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
                return -EBUSY;
        }
 
-       dma_private = dma_alloc_coherent(substream->pcm->dev,
+       dma_private = dma_alloc_coherent(substream->pcm->card->dev,
                sizeof(struct fsl_dma_private), &ld_buf_phys, GFP_KERNEL);
        if (!dma_private) {
                dev_err(substream->pcm->card->dev,
@@ -445,7 +445,7 @@ static int fsl_dma_open(struct snd_pcm_substream *substream)
                dev_err(substream->pcm->card->dev,
                        "can't register ISR for IRQ %u (ret=%i)\n",
                        dma_private->irq, ret);
-               dma_free_coherent(substream->pcm->dev,
+               dma_free_coherent(substream->pcm->card->dev,
                        sizeof(struct fsl_dma_private),
                        dma_private, dma_private->ld_buf_phys);
                return ret;
@@ -761,13 +761,13 @@ static int fsl_dma_close(struct snd_pcm_substream *substream)
                        free_irq(dma_private->irq, dma_private);
 
                if (dma_private->ld_buf_phys) {
-                       dma_unmap_single(substream->pcm->dev,
+                       dma_unmap_single(substream->pcm->card->dev,
                                dma_private->ld_buf_phys,
                                sizeof(dma_private->link), DMA_TO_DEVICE);
                }
 
                /* Deallocate the fsl_dma_private structure */
-               dma_free_coherent(substream->pcm->dev,
+               dma_free_coherent(substream->pcm->card->dev,
                        sizeof(struct fsl_dma_private),
                        dma_private, dma_private->ld_buf_phys);
                substream->runtime->private_data = NULL;
--

1.5.6.5
_______________________________________________
Linuxppc-dev mailing list
[email protected]

Quoted 8 lines
Quoted 2 lines

https://ozlabs.org/mailman/listinfo/linuxppc-dev

--
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.

--
Anton Vorontsov
email: [email protected]
irc://irc.freenode.net/bd2

Signature · 2 lines
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.
#4 lore raw
Quoted 11 lines · Grant Likely
Quoted 8 lines · Anton Vorontsov
Quoted 4 lines · Grant Likely

Becky & Kumar,

Considering these fixups, would it be advisable for the dma functions to climb up the dev->parent linkage when dma_ops are NULL?

pcm->dev->parent doesn't point to the proper device, so this won't work.

Quoted 7 lines

What does it point to then? IIUC, following the ->parent links should eventually get to a struct device that represents the platform or of_platform device for purposes of power management.

BTW, I'm not NAKing your patch. The patch looks fine to me.

g.

No, I think we should just fix drivers. There aren't that many SoC drivers and once they are fixed this will resolve itself.

- k

#5 lore raw
Quoted 2 lines · Kumar Gala

No, I think we should just fix drivers.  There aren't that many SoC drivers and once they are fixed this will resolve itself.

I agree. The only reason I passed NULL when I wrote the drivers was because I was too lazy to figure out what the right dev pointer was. ASoC is kinda screwy in that many fields are not initialized when you think they should be. More than once I passed a pointer to a wrong object without realizing it, even though the code worked.

I'll check out these two patches on Monday.

Signature · 2 lines
Timur Tabi
Linux kernel developer at Freescale
Acked-by on v1 1/1
#6 lore raw
In reply to Anton Vorontsov
Quoted 24 lines · Anton Vorontsov

The driver should pass a device that specifies internal DMA ops, but substream->pcm is just a logical device, and thus doesn't have arch- specific dma callbacks, therefore following bug appears:

Freescale Synchronous Serial Interface (SSI) ASoC Driver

  ------------[ cut here ]------------
  kernel BUG at arch/powerpc/include/asm/dma-mapping.h:237!
  Oops: Exception in kernel mode, sig: 5 [#1]
  ...
  NIP [c02259c4] snd_malloc_dev_pages+0x58/0xac
  LR [c0225c74] snd_dma_alloc_pages+0xf8/0x108
  Call Trace:
  [df02bde0] [df02be2c] 0xdf02be2c (unreliable)
  [df02bdf0] [c0225c74] snd_dma_alloc_pages+0xf8/0x108
  [df02be10] [c023a100] fsl_dma_new+0x68/0x124
  [df02be20] [c02342ac] soc_new_pcm+0x1bc/0x234
  [df02bea0] [c02343dc] snd_soc_new_pcms+0xb8/0x148
  [df02bed0] [c023824c] cs4270_probe+0x34/0x124
  [df02bef0] [c0232fe8] snd_soc_instantiate_card+0x1a4/0x2f4
  [df02bf20] [c0233164] snd_soc_instantiate_cards+0x2c/0x68
  [df02bf30] [c0234704] snd_soc_register_platform+0x60/0x80
  [df02bf50] [c03d5664] fsl_soc_platform_init+0x18/0x28
  ...
Quoted 4 lines

This patch fixes the issue by using card's device instead.

---

Mark and Takashi: this patch is a must-fix for 2.6.30

Signature · 2 lines
Timur Tabi
Linux kernel developer at Freescale