Skip to contentSkip to conversation

ASoC: wm_adsp: Avoid narrow race condition on compr pointer

PATCH In review Charles Keepax 2 participants 8 messages Feb 19, 2016 – Feb 20, 2016

Subsystems (3)
Sound ASoC Wolfson microelectronics drivers
lore
Share & tools
lore

Charles Keepax

author
#0 lore raw

PATCH 1/4 ASoC: wm_adsp: Avoid narrow race condition on compr pointer

There is a very small window between then wm_adsp_compr_free gets call and when the DSP is actually powered down. If we get an IRQ from the DSP in this window then the wm_adsp_compr pointer will be NULL. This patch adds a check for this into the IRQ handler to avoid any issues when this happens.

Signed-off-by Charles Keepax <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/soc/codecs/wm_adsp.c +1 -1
index ed90e12..00a7d68 100644
@@ -2857,7 +2857,7 @@ int wm_adsp_compr_handle_irq(struct wm_adsp *dsp)
 		goto out;
 	}
 
-	if (compr->stream)
+	if (compr && compr->stream)
 		snd_compr_fragment_elapsed(compr->stream);
 
 out:
Signature · 1 line
2.1.4
author
#1 lore raw

PATCH 2/4 ASoC: wm_adsp: Explicitly set the control access flags in all cases

In the case where the firmware does not tell us the access flags for the control, we let ALSA select a default (READWRITE). But really we should be applying the volatile flag in this case, as we will read the control from the DSP if it is on in this case. This patch explicitly sets the access flags in all cases.

Signed-off-by Charles Keepax <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 3 +++
 1 file changed, 3 insertions(+)
sound/soc/codecs/wm_adsp.c +3 -0
index 00a7d68..c97b3a1 100644
@@ -877,6 +877,9 @@ static int wmfw_add_ctl(struct wm_adsp *dsp, struct wm_coeff_ctl *ctl)
 			kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_READ;
 		if (ctl->flags & WMFW_CTL_FLAG_VOLATILE)
 			kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_VOLATILE;
+	} else {
+		kcontrol->access = SNDRV_CTL_ELEM_ACCESS_READWRITE;
+		kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_VOLATILE;
 	}
 
 	ret = snd_soc_add_card_controls(dsp->card,
Signature · 1 line
2.1.4
maintainer
#2 lore raw

Applied "ASoC: wm_adsp: Explicitly set the control access flags in all cases" to the asoc tree

The patch

ASoC: wm_adsp: Explicitly set the control access flags in all cases

has been applied to the asoc tree at

git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark

Quoted 1 line

From cc815c4b86d14630a7d4ca4f4c7d2c2acb88cb52 Mon Sep 17 00:00:00 2001

From: Charles Keepax <[email protected]>
Date: Fri, 19 Feb 2016 14:44:42 +0000
Subject: [PATCH] ASoC: wm_adsp: Explicitly set the control access flags in all
cases

In the case where the firmware does not tell us the access flags for the control, we let ALSA select a default (READWRITE). But really we should be applying the volatile flag in this case, as we will read the control from the DSP if it is on in this case. This patch explicitly sets the access flags in all cases.

Signed-off-by Charles Keepax <[email protected]>
Signed-off-by Mark Brown <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 3 +++
 1 file changed, 3 insertions(+)
sound/soc/codecs/wm_adsp.c +3 -0
index cb49a63..ff0ce6ba 100644
@@ -857,6 +857,9 @@ static int wmfw_add_ctl(struct wm_adsp *dsp, struct wm_coeff_ctl *ctl)
 			kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_READ;
 		if (ctl->flags & WMFW_CTL_FLAG_VOLATILE)
 			kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_VOLATILE;
+	} else {
+		kcontrol->access = SNDRV_CTL_ELEM_ACCESS_READWRITE;
+		kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_VOLATILE;
 	}
 
 	ret = snd_soc_add_card_controls(dsp->card,
Signature · 1 line
2.7.0
author
#3 lore raw

PATCH 3/4 ASoC: wm_adsp: Fixup some odd line wrapping

In reply to Charles Keepax
Signed-off-by Charles Keepax <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 23 +++++++----------------
 1 file changed, 7 insertions(+), 16 deletions(-)
sound/soc/codecs/wm_adsp.c +7 -16
index c97b3a1..1c6d5fb 100644
@@ -882,15 +882,13 @@ static int wmfw_add_ctl(struct wm_adsp *dsp, struct wm_coeff_ctl *ctl)
 		kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_VOLATILE;
 	}
 
-	ret = snd_soc_add_card_controls(dsp->card,
-					kcontrol, 1);
+	ret = snd_soc_add_card_controls(dsp->card, kcontrol, 1);
 	if (ret < 0)
 		goto err_kcontrol;
 
 	kfree(kcontrol);
 
-	ctl->kcontrol = snd_soc_card_get_kcontrol(dsp->card,
-						  ctl->name);
+	ctl->kcontrol = snd_soc_card_get_kcontrol(dsp->card, ctl->name);
 
 	return 0;
 

@@ -910,9 +908,7 @@ static int wm_coeff_init_control_caches(struct wm_adsp *dsp)
 		if (ctl->flags & WMFW_CTL_FLAG_VOLATILE)
 			continue;
 
-		ret = wm_coeff_read_control(ctl,
-					    ctl->cache,
-					    ctl->len);
+		ret = wm_coeff_read_control(ctl, ctl->cache, ctl->len);
 		if (ret < 0)
 			return ret;
 	}

@@ -929,9 +925,7 @@ static int wm_coeff_sync_controls(struct wm_adsp *dsp)
 		if (!ctl->enabled)
 			continue;
 		if (ctl->set && !(ctl->flags & WMFW_CTL_FLAG_VOLATILE)) {
-			ret = wm_coeff_write_control(ctl,
-						     ctl->cache,
-						     ctl->len);
+			ret = wm_coeff_write_control(ctl, ctl->cache, ctl->len);
 			if (ret < 0)
 				return ret;
 		}

@@ -1527,8 +1521,7 @@ static void *wm_adsp_read_algs(struct wm_adsp *dsp, size_t n_algs,
 
 	ret = regmap_raw_read(dsp->regmap, pos, alg, len * 2);
 	if (ret != 0) {
-		adsp_err(dsp, "Failed to read algorithm list: %d\n",
-			ret);
+		adsp_err(dsp, "Failed to read algorithm list: %d\n", ret);
 		kfree(alg);
 		return ERR_PTR(ret);
 	}

@@ -2027,8 +2020,7 @@ int wm_adsp1_event(struct snd_soc_dapm_widget *w,
 				goto err_mutex;
 			}
 
-			val = (val & dsp->sysclk_mask)
-				>> dsp->sysclk_shift;
+			val = (val & dsp->sysclk_mask) >> dsp->sysclk_shift;
 
 			ret = regmap_update_bits(dsp->regmap,
 						 dsp->base + ADSP1_CONTROL_31,

@@ -2121,8 +2113,7 @@ static int wm_adsp2_ena(struct wm_adsp *dsp)
 
 	/* Wait for the RAM to start, should be near instantaneous */
 	for (count = 0; count < 10; ++count) {
-		ret = regmap_read(dsp->regmap, dsp->base + ADSP2_STATUS1,
-				  &val);
+		ret = regmap_read(dsp->regmap, dsp->base + ADSP2_STATUS1, &val);
 		if (ret != 0)
 			return ret;
Signature · 1 line
2.1.4
maintainer
#4 lore raw

Applied "ASoC: wm_adsp: Fixup some odd line wrapping" to the asoc tree

The patch

ASoC: wm_adsp: Fixup some odd line wrapping

has been applied to the asoc tree at

git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark

Quoted 1 line

From 7d00cd975a52e637da19548c1cc7d87a0a32b49c Mon Sep 17 00:00:00 2001

From: Charles Keepax <[email protected]>
Date: Fri, 19 Feb 2016 14:44:43 +0000
Subject: [PATCH] ASoC: wm_adsp: Fixup some odd line wrapping

Signed-off-by Charles Keepax <[email protected]>
Signed-off-by Mark Brown <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 23 +++++++----------------
 1 file changed, 7 insertions(+), 16 deletions(-)
sound/soc/codecs/wm_adsp.c +7 -16
index ff0ce6ba..0f6eeb1 100644
@@ -862,15 +862,13 @@ static int wmfw_add_ctl(struct wm_adsp *dsp, struct wm_coeff_ctl *ctl)
 		kcontrol->access |= SNDRV_CTL_ELEM_ACCESS_VOLATILE;
 	}
 
-	ret = snd_soc_add_card_controls(dsp->card,
-					kcontrol, 1);
+	ret = snd_soc_add_card_controls(dsp->card, kcontrol, 1);
 	if (ret < 0)
 		goto err_kcontrol;
 
 	kfree(kcontrol);
 
-	ctl->kcontrol = snd_soc_card_get_kcontrol(dsp->card,
-						  ctl->name);
+	ctl->kcontrol = snd_soc_card_get_kcontrol(dsp->card, ctl->name);
 
 	return 0;
 

@@ -890,9 +888,7 @@ static int wm_coeff_init_control_caches(struct wm_adsp *dsp)
 		if (ctl->flags & WMFW_CTL_FLAG_VOLATILE)
 			continue;
 
-		ret = wm_coeff_read_control(ctl,
-					    ctl->cache,
-					    ctl->len);
+		ret = wm_coeff_read_control(ctl, ctl->cache, ctl->len);
 		if (ret < 0)
 			return ret;
 	}

@@ -909,9 +905,7 @@ static int wm_coeff_sync_controls(struct wm_adsp *dsp)
 		if (!ctl->enabled)
 			continue;
 		if (ctl->set && !(ctl->flags & WMFW_CTL_FLAG_VOLATILE)) {
-			ret = wm_coeff_write_control(ctl,
-						     ctl->cache,
-						     ctl->len);
+			ret = wm_coeff_write_control(ctl, ctl->cache, ctl->len);
 			if (ret < 0)
 				return ret;
 		}

@@ -1507,8 +1501,7 @@ static void *wm_adsp_read_algs(struct wm_adsp *dsp, size_t n_algs,
 
 	ret = regmap_raw_read(dsp->regmap, pos, alg, len * 2);
 	if (ret != 0) {
-		adsp_err(dsp, "Failed to read algorithm list: %d\n",
-			ret);
+		adsp_err(dsp, "Failed to read algorithm list: %d\n", ret);
 		kfree(alg);
 		return ERR_PTR(ret);
 	}

@@ -2007,8 +2000,7 @@ int wm_adsp1_event(struct snd_soc_dapm_widget *w,
 				goto err_mutex;
 			}
 
-			val = (val & dsp->sysclk_mask)
-				>> dsp->sysclk_shift;
+			val = (val & dsp->sysclk_mask) >> dsp->sysclk_shift;
 
 			ret = regmap_update_bits(dsp->regmap,
 						 dsp->base + ADSP1_CONTROL_31,

@@ -2101,8 +2093,7 @@ static int wm_adsp2_ena(struct wm_adsp *dsp)
 
 	/* Wait for the RAM to start, should be near instantaneous */
 	for (count = 0; count < 10; ++count) {
-		ret = regmap_read(dsp->regmap, dsp->base + ADSP2_STATUS1,
-				  &val);
+		ret = regmap_read(dsp->regmap, dsp->base + ADSP2_STATUS1, &val);
 		if (ret != 0)
 			return ret;
Signature · 1 line
2.7.0
author
#5 lore raw

PATCH 4/4 ASoC: wm_adsp: wm_coeff_{read|write}_control should use passed length

In reply to Charles Keepax

wm_coeff_{read|write}_control were using the control length rather than the length parameter passed to them. This is not causing any issues as the two values are currently always the same, but this needs fixed to allow future improvements.

Signed-off-by Charles Keepax <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
sound/soc/codecs/wm_adsp.c +8 -8
index 1c6d5fb..6d9552a 100644
@@ -741,19 +741,19 @@ static int wm_coeff_write_control(struct wm_coeff_ctl *ctl,
 	reg = ctl->alg_region.base + ctl->offset;
 	reg = wm_adsp_region_to_reg(mem, reg);
 
-	scratch = kmemdup(buf, ctl->len, GFP_KERNEL | GFP_DMA);
+	scratch = kmemdup(buf, len, GFP_KERNEL | GFP_DMA);
 	if (!scratch)
 		return -ENOMEM;
 
 	ret = regmap_raw_write(dsp->regmap, reg, scratch,
-			       ctl->len);
+			       len);
 	if (ret) {
 		adsp_err(dsp, "Failed to write %zu bytes to %x: %d\n",
-			 ctl->len, reg, ret);
+			 len, reg, ret);
 		kfree(scratch);
 		return ret;
 	}
-	adsp_dbg(dsp, "Wrote %zu bytes to %x\n", ctl->len, reg);
+	adsp_dbg(dsp, "Wrote %zu bytes to %x\n", len, reg);
 
 	kfree(scratch);
 

@@ -800,20 +800,20 @@ static int wm_coeff_read_control(struct wm_coeff_ctl *ctl,
 	reg = ctl->alg_region.base + ctl->offset;
 	reg = wm_adsp_region_to_reg(mem, reg);
 
-	scratch = kmalloc(ctl->len, GFP_KERNEL | GFP_DMA);
+	scratch = kmalloc(len, GFP_KERNEL | GFP_DMA);
 	if (!scratch)
 		return -ENOMEM;
 
-	ret = regmap_raw_read(dsp->regmap, reg, scratch, ctl->len);
+	ret = regmap_raw_read(dsp->regmap, reg, scratch, len);
 	if (ret) {
 		adsp_err(dsp, "Failed to read %zu bytes from %x: %d\n",
 			 ctl->len, reg, ret);
 		kfree(scratch);
 		return ret;
 	}
-	adsp_dbg(dsp, "Read %zu bytes from %x\n", ctl->len, reg);
+	adsp_dbg(dsp, "Read %zu bytes from %x\n", len, reg);
 
-	memcpy(buf, scratch, ctl->len);
+	memcpy(buf, scratch, len);
 	kfree(scratch);
 
 	return 0;
Signature · 1 line
2.1.4
maintainer
#6 lore raw

Applied "ASoC: wm_adsp: wm_coeff_{read|write}_control should use passed length" to the asoc tree

The patch

ASoC: wm_adsp: wm_coeff_{read|write}_control should use passed length

has been applied to the asoc tree at

git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark

Quoted 1 line

From 4f8ea6d759f6ed5d448472143cef7681e319c958 Mon Sep 17 00:00:00 2001

From: Charles Keepax <[email protected]>
Date: Fri, 19 Feb 2016 14:44:44 +0000
Subject: [PATCH] ASoC: wm_adsp: wm_coeff_{read|write}_control should use
passed length

wm_coeff_{read|write}_control were using the control length rather than the length parameter passed to them. This is not causing any issues as the two values are currently always the same, but this needs fixed to allow future improvements.

Signed-off-by Charles Keepax <[email protected]>
Signed-off-by Mark Brown <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
sound/soc/codecs/wm_adsp.c +8 -8
index 0f6eeb1..36687db 100644
@@ -721,19 +721,19 @@ static int wm_coeff_write_control(struct wm_coeff_ctl *ctl,
 	reg = ctl->alg_region.base + ctl->offset;
 	reg = wm_adsp_region_to_reg(mem, reg);
 
-	scratch = kmemdup(buf, ctl->len, GFP_KERNEL | GFP_DMA);
+	scratch = kmemdup(buf, len, GFP_KERNEL | GFP_DMA);
 	if (!scratch)
 		return -ENOMEM;
 
 	ret = regmap_raw_write(dsp->regmap, reg, scratch,
-			       ctl->len);
+			       len);
 	if (ret) {
 		adsp_err(dsp, "Failed to write %zu bytes to %x: %d\n",
-			 ctl->len, reg, ret);
+			 len, reg, ret);
 		kfree(scratch);
 		return ret;
 	}
-	adsp_dbg(dsp, "Wrote %zu bytes to %x\n", ctl->len, reg);
+	adsp_dbg(dsp, "Wrote %zu bytes to %x\n", len, reg);
 
 	kfree(scratch);
 

@@ -780,20 +780,20 @@ static int wm_coeff_read_control(struct wm_coeff_ctl *ctl,
 	reg = ctl->alg_region.base + ctl->offset;
 	reg = wm_adsp_region_to_reg(mem, reg);
 
-	scratch = kmalloc(ctl->len, GFP_KERNEL | GFP_DMA);
+	scratch = kmalloc(len, GFP_KERNEL | GFP_DMA);
 	if (!scratch)
 		return -ENOMEM;
 
-	ret = regmap_raw_read(dsp->regmap, reg, scratch, ctl->len);
+	ret = regmap_raw_read(dsp->regmap, reg, scratch, len);
 	if (ret) {
 		adsp_err(dsp, "Failed to read %zu bytes from %x: %d\n",
 			 ctl->len, reg, ret);
 		kfree(scratch);
 		return ret;
 	}
-	adsp_dbg(dsp, "Read %zu bytes from %x\n", ctl->len, reg);
+	adsp_dbg(dsp, "Read %zu bytes from %x\n", len, reg);
 
-	memcpy(buf, scratch, ctl->len);
+	memcpy(buf, scratch, len);
 	kfree(scratch);
 
 	return 0;
Signature · 1 line
2.7.0
maintainer
#7 lore raw

Applied "ASoC: wm_adsp: Avoid narrow race condition on compr pointer" to the asoc tree

In reply to Charles Keepax

The patch

ASoC: wm_adsp: Avoid narrow race condition on compr pointer

has been applied to the asoc tree at

git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound.git

All being well this means that it will be integrated into the linux-next tree (usually sometime in the next 24 hours) and sent to Linus during the next merge window (or sooner if it is a bug fix), however if problems are discovered then the patch may be dropped or reverted.

You may get further e-mails resulting from automated or manual testing and review of the tree, please engage with people reporting problems and send followup patches addressing any issues that are reported if needed.

If any updates are required or you are submitting further changes they should be sent as incremental updates against current git, existing patches will not be replaced.

Please add any relevant lists and maintainers to the CCs when replying to this mail.

Thanks,
Mark

Quoted 1 line

From c7dae7c4c60693286298e300b3d140c8cff46daa Mon Sep 17 00:00:00 2001

From: Charles Keepax <[email protected]>
Date: Fri, 19 Feb 2016 14:44:41 +0000
Subject: [PATCH] ASoC: wm_adsp: Avoid narrow race condition on compr pointer

There is a very small window between then wm_adsp_compr_free gets call and when the DSP is actually powered down. If we get an IRQ from the DSP in this window then the wm_adsp_compr pointer will be NULL. This patch adds a check for this into the IRQ handler to avoid any issues when this happens.

Signed-off-by Charles Keepax <[email protected]>
Signed-off-by Mark Brown <[email protected]>
---
 sound/soc/codecs/wm_adsp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
sound/soc/codecs/wm_adsp.c +1 -1
index 68b85ee..cb49a63 100644
@@ -2845,7 +2845,7 @@ int wm_adsp_compr_handle_irq(struct wm_adsp *dsp)
 		goto out;
 	}
 
-	if (compr->stream)
+	if (compr && compr->stream)
 		snd_compr_fragment_elapsed(compr->stream);
 
 out:
Signature · 1 line
2.7.0