Skip to contentSkip to conversation

ASoC: SOF: sof-client-probes: fix error codes in sof_probes_compr_copy()

PATCH In review Dan Carpenter 3 participants 11 messages Jul 6, 2022

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/2 ASoC: SOF: sof-client-probes: fix error codes in sof_probes_compr_copy()

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

Fixes 3dc0d7091778 ("ASoC: SOF: Convert the generic probe support to SOF client")
---
 sound/soc/sof/sof-client-probes.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
sound/soc/sof/sof-client-probes.c +6 -6
index 1f1ea93a7fbf..679bc7d371fc 100644
@@ -385,7 +385,6 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 	struct snd_compr_runtime *rtd = cstream->runtime;
 	unsigned int offset, n;
 	void *ptr;
-	int ret;
 
 	if (count > rtd->buffer_size)
 		count = rtd->buffer_size;
@@ -395,14 +394,15 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 	n = rtd->buffer_size - offset;
 
 	if (count < n) {
-		ret = copy_to_user(buf, ptr, count);
+		if (copy_to_user(buf, ptr, count))
+			return -EFAULT;
 	} else {
-		ret = copy_to_user(buf, ptr, n);
-		ret += copy_to_user(buf + n, rtd->dma_area, count - n);
+		if (copy_to_user(buf, ptr, n))
+			return -EFAULT;
+		if (copy_to_user(buf + n, rtd->dma_area, count - n))
+			return -EFAULT;
 	}
 
-	if (ret)
-		return count - ret;
 	return count;
 }
 
Signature · 1 line
2.35.1
#1 lore raw

PATCH 2/2 ASoC: SOF: sof-client-probes: cleanup tokenize_input()

The tokenize_input() function is cleaner if it uses strndup_user() instead of simple_write_to_buffer(). The way it's written now, if *ppos is non-zero then it returns -EIO but normally we would return 0 in that case. It's easier to handle that in the callers.

---
 sound/soc/sof/sof-client-probes.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)
sound/soc/sof/sof-client-probes.c +13 -14
index 679bc7d371fc..6c922b683f67 100644
@@ -461,24 +461,17 @@ static int strsplit_u32(char *buf, const char *delim, u32 **tkns, size_t *num_tk
 }
 
 static int tokenize_input(const char __user *from, size_t count,
-			  loff_t *ppos, u32 **tkns, size_t *num_tkns)
+			  u32 **tkns, size_t *num_tkns)
 {
 	char *buf;
 	int ret;
 
-	buf = kmalloc(count + 1, GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
-	ret = simple_write_to_buffer(buf, count, ppos, from, count);
-	if (ret != count) {
-		ret = ret >= 0 ? -EIO : ret;
-		goto exit;
-	}
+	buf = strndup_user(from, count + 1);
+	if (IS_ERR(buf))
+		return PTR_ERR(buf);
 
-	buf[count] = '\0';
 	ret = strsplit_u32(buf, ",", tkns, num_tkns);
-exit:
+
 	kfree(buf);
 	return ret;
 }
@@ -552,12 +545,15 @@ sof_probes_dfs_points_write(struct file *file, const char __user *from,
 	u32 *tkns;
 	int ret, err;
 
+	if (*ppos)
+		return 0;
+
 	if (priv->extractor_stream_tag == SOF_PROBES_INVALID_NODE_ID) {
 		dev_warn(dev, "no extractor stream running\n");
 		return -ENOENT;
 	}
 
-	ret = tokenize_input(from, count, ppos, &tkns, &num_tkns);
+	ret = tokenize_input(from, count, &tkns, &num_tkns);
 	if (ret < 0)
 		return ret;
 	bytes = sizeof(*tkns) * num_tkns;
@@ -607,12 +603,15 @@ sof_probes_dfs_points_remove_write(struct file *file, const char __user *from,
 	u32 *tkns;
 	int ret, err;
 
+	if (*ppos)
+		return 0;
+
 	if (priv->extractor_stream_tag == SOF_PROBES_INVALID_NODE_ID) {
 		dev_warn(dev, "no extractor stream running\n");
 		return -ENOENT;
 	}
 
-	ret = tokenize_input(from, count, ppos, &tkns, &num_tkns);
+	ret = tokenize_input(from, count, &tkns, &num_tkns);
 	if (ret < 0)
 		return ret;
 	if (!num_tkns) {
Signature · 1 line
2.35.1
maintainer
#2 lore raw

Re: [PATCH 2/2] ASoC: SOF: sof-client-probes: cleanup tokenize_input()

Quoted 4 lines · Dan Carpenter

The tokenize_input() function is cleaner if it uses strndup_user() instead of simple_write_to_buffer(). The way it's written now, if *ppos is non-zero then it returns -EIO but normally we would return 0 in that case. It's easier to handle that in the callers.

This patch breaks the probe point settings:

# echo 52,1,0 > /sys/kernel/debug/sof/probe_points
-bash: echo: write error: Invalid argument

I did not looked for the exact reason, but something is not correct.

Quoted 73 lines
---
 sound/soc/sof/sof-client-probes.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)
sound/soc/sof/sof-client-probes.c +13 -14
index 679bc7d371fc..6c922b683f67 100644
@@ -461,24 +461,17 @@ static int strsplit_u32(char *buf, const char *delim, u32 **tkns, size_t *num_tk
 }
 
 static int tokenize_input(const char __user *from, size_t count,
-			  loff_t *ppos, u32 **tkns, size_t *num_tkns)
+			  u32 **tkns, size_t *num_tkns)
 {
 	char *buf;
 	int ret;
 
-	buf = kmalloc(count + 1, GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-
-	ret = simple_write_to_buffer(buf, count, ppos, from, count);
-	if (ret != count) {
-		ret = ret >= 0 ? -EIO : ret;
-		goto exit;
-	}
+	buf = strndup_user(from, count + 1);
+	if (IS_ERR(buf))
+		return PTR_ERR(buf);
 
-	buf[count] = '\0';
 	ret = strsplit_u32(buf, ",", tkns, num_tkns);
-exit:
+
 	kfree(buf);
 	return ret;
 }
@@ -552,12 +545,15 @@ sof_probes_dfs_points_write(struct file *file, const char __user *from,
 	u32 *tkns;
 	int ret, err;
 
+	if (*ppos)
+		return 0;
+
 	if (priv->extractor_stream_tag == SOF_PROBES_INVALID_NODE_ID) {
 		dev_warn(dev, "no extractor stream running\n");
 		return -ENOENT;
 	}
 
-	ret = tokenize_input(from, count, ppos, &tkns, &num_tkns);
+	ret = tokenize_input(from, count, &tkns, &num_tkns);
 	if (ret < 0)
 		return ret;
 	bytes = sizeof(*tkns) * num_tkns;
@@ -607,12 +603,15 @@ sof_probes_dfs_points_remove_write(struct file *file, const char __user *from,
 	u32 *tkns;
 	int ret, err;
 
+	if (*ppos)
+		return 0;
+
 	if (priv->extractor_stream_tag == SOF_PROBES_INVALID_NODE_ID) {
 		dev_warn(dev, "no extractor stream running\n");
 		return -ENOENT;
 	}
 
-	ret = tokenize_input(from, count, ppos, &tkns, &num_tkns);
+	ret = tokenize_input(from, count, &tkns, &num_tkns);
 	if (ret < 0)
 		return ret;
 	if (!num_tkns) {
Signature · 1 line
Péter
#3 lore raw

Re: [PATCH 2/2] ASoC: SOF: sof-client-probes: cleanup tokenize_input()

Quoted 8 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

The tokenize_input() function is cleaner if it uses strndup_user() instead of simple_write_to_buffer(). The way it's written now, if *ppos is non-zero then it returns -EIO but normally we would return 0 in that case. It's easier to handle that in the callers.

Quoted 7 lines

This patch breaks the probe point settings:

# echo 52,1,0 > /sys/kernel/debug/sof/probe_points
-bash: echo: write error: Invalid argument

I did not looked for the exact reason, but something is not correct.

Crud...

Thanks for testing.

I used strndup_user() in a couple other patches today and I didn't realize how strict it was. I've NAKed my patches which used strndup_user(). One of the patches was an infoleak patch so I'm going to resend that using memdup_user() instead but let's just drop this one.

I guess another safer option would be to just always zero the buffers going into simple_write_to_buffer()...

regards,
dan carpenter

#4 lore raw

Re: [PATCH 2/2] ASoC: SOF: sof-client-probes: cleanup tokenize_input()

Quoted 26 lines · Dan Carpenter
Quoted 8 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

The tokenize_input() function is cleaner if it uses strndup_user() instead of simple_write_to_buffer(). The way it's written now, if *ppos is non-zero then it returns -EIO but normally we would return 0 in that case. It's easier to handle that in the callers.

Quoted 7 lines

This patch breaks the probe point settings:

# echo 52,1,0 > /sys/kernel/debug/sof/probe_points
-bash: echo: write error: Invalid argument

I did not looked for the exact reason, but something is not correct.

Crud...

Thanks for testing.

I used strndup_user() in a couple other patches today and I didn't realize how strict it was. I've NAKed my patches which used strndup_user(). One of the patches was an infoleak patch so I'm going to resend that using memdup_user() instead but let's just drop this one.

Quoted 6 lines

I guess another safer option would be to just always zero the buffers going into simple_write_to_buffer()...

regards,
dan carpenter

Hello,

Indeed the strsplit_u32() contains some bugs - tokenize_input() needs no fixes if I'm not mistaken though.
It seems I did not realize the bugs were not fixed. As the avs-driver makes use of probes too and these are being tested there regularly the team did notice the problems. Below is the implementation. I'm saying this as the plan is to move both strsplit_u32() and tokenize_input() into the common code so it can be re-used by both drivers. Will send the patches soon :)

Regards,
Czarek

static int
strsplit_u32(const char *str, const char *delim, u32 **tkns, size_t *num_tkns)

{
        size_t max_count = 32;
        size_t count = 0;
        char *s, **p;
        u32 *buf, *tmp;
        int ret = 0;

        p = (char **)&str;
        *tkns = NULL;
        *num_tkns = 0;

        buf = kcalloc(max_count, sizeof(*buf), GFP_KERNEL);
        if (!buf)
                return -ENOMEM;

        while ((s = strsep(p, delim)) != NULL) {
                ret = kstrtouint(s, 0, buf + count);
                if (ret)
                        goto free_buf;

                if (++count > max_count) {
                        max_count *= 2;
                        tmp = krealloc(buf, max_count * sizeof(*buf), 
GFP_KERNEL);
                        if (!tmp) {
                                ret = -ENOMEM;
                                goto free_buf;
                        }
                        buf = tmp;
                }
        }

        if (!count)
                goto free_buf;
        *tkns = kmemdup(buf, count * sizeof(*buf), GFP_KERNEL);
        if (*tkns == NULL) {
                ret = -ENOMEM;
                goto free_buf;
        }
        *num_tkns = count;

free_buf:

        kfree(buf);
        return ret;
}
maintainer
#5 lore raw
In reply to Dan Carpenter
Quoted 4 lines · Dan Carpenter

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

The function is copying data from a circular buffer to a use buffer. The single copy_to_user() is used when we don't have wrapping, the 'double' copy_to_user() is when we wrap, so first copy is from the end of the buffer then we copy the data from the start of the buffer to get all data.

Quoted 39 lines
Fixes 3dc0d7091778 ("ASoC: SOF: Convert the generic probe support to SOF client")
---
 sound/soc/sof/sof-client-probes.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
sound/soc/sof/sof-client-probes.c +6 -6
index 1f1ea93a7fbf..679bc7d371fc 100644
@@ -385,7 +385,6 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 	struct snd_compr_runtime *rtd = cstream->runtime;
 	unsigned int offset, n;
 	void *ptr;
-	int ret;
 
 	if (count > rtd->buffer_size)
 		count = rtd->buffer_size;
@@ -395,14 +394,15 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 	n = rtd->buffer_size - offset;
 
 	if (count < n) {
-		ret = copy_to_user(buf, ptr, count);
+		if (copy_to_user(buf, ptr, count))
+			return -EFAULT;
 	} else {
-		ret = copy_to_user(buf, ptr, n);
-		ret += copy_to_user(buf + n, rtd->dma_area, count - n);
+		if (copy_to_user(buf, ptr, n))
+			return -EFAULT;
+		if (copy_to_user(buf + n, rtd->dma_area, count - n))
+			return -EFAULT;
 	}
 
-	if (ret)
-		return count - ret;
 	return count;
 }
 
Signature · 1 line
Péter
maintainer
#6 lore raw
Quoted 8 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

Quoted 5 lines

The function is copying data from a circular buffer to a use buffer. The single copy_to_user() is used when we don't have wrapping, the 'double' copy_to_user() is when we wrap, so first copy is from the end of the buffer then we copy the data from the start of the buffer to get all data.

What I wanted to say is that the original code is correct, this patch would break the functionality.

Quoted 41 lines
Quoted 40 lines
Fixes 3dc0d7091778 ("ASoC: SOF: Convert the generic probe support to SOF client")
---
 sound/soc/sof/sof-client-probes.c | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)
sound/soc/sof/sof-client-probes.c +6 -6
index 1f1ea93a7fbf..679bc7d371fc 100644
@@ -385,7 +385,6 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 	struct snd_compr_runtime *rtd = cstream->runtime;
 	unsigned int offset, n;
 	void *ptr;
-	int ret;
 
 	if (count > rtd->buffer_size)
 		count = rtd->buffer_size;
@@ -395,14 +394,15 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 	n = rtd->buffer_size - offset;
 
 	if (count < n) {
-		ret = copy_to_user(buf, ptr, count);
+		if (copy_to_user(buf, ptr, count))
+			return -EFAULT;
 	} else {
-		ret = copy_to_user(buf, ptr, n);
-		ret += copy_to_user(buf + n, rtd->dma_area, count - n);
+		if (copy_to_user(buf, ptr, n))
+			return -EFAULT;
+		if (copy_to_user(buf + n, rtd->dma_area, count - n))
+			return -EFAULT;
 	}
 
-	if (ret)
-		return count - ret;
 	return count;
 }
 
Signature · 1 line
Péter
#7 lore raw
In reply to Péter Ujfalusi
Quoted 8 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

Quoted 5 lines

The function is copying data from a circular buffer to a use buffer. The single copy_to_user() is used when we don't have wrapping, the 'double' copy_to_user() is when we wrap, so first copy is from the end of the buffer then we copy the data from the start of the buffer to get all data.

Ok. But the bugs in the original code are real. I will resend.

regards,
dan carpenter

#8 lore raw
Quoted 9 lines · Dan Carpenter
Quoted 8 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

Quoted 7 lines
Quoted 5 lines

The function is copying data from a circular buffer to a use buffer. The single copy_to_user() is used when we don't have wrapping, the 'double' copy_to_user() is when we wrap, so first copy is from the end of the buffer then we copy the data from the start of the buffer to get all data.

Ok. But the bugs in the original code are real. I will resend.

Actually that's not true. The bugs in the original code are something that only affect users who deserve it? I might not resend. A fix would look something like below?

regards,
dan carpenter

sound/soc/sof/sof-client-probes.c +6 -1
index 1f1ea93a7fbf..32fa3186c295 100644
@@ -398,9 +398,14 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 		ret = copy_to_user(buf, ptr, count);
 	} else {
 		ret = copy_to_user(buf, ptr, n);
-		ret += copy_to_user(buf + n, rtd->dma_area, count - n);
+		if (ret) {
+			ret += count - n;
+			goto done;
+		}
+		ret = copy_to_user(buf + n, rtd->dma_area, count - n);
 	}
 
+done:
 	if (ret)
 		return count - ret;
 	return count;
#9 lore raw
Quoted 16 lines · Dan Carpenter
Quoted 15 lines · Dan Carpenter
Quoted 13 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

The function is copying data from a circular buffer to a use buffer. The single copy_to_user() is used when we don't have wrapping, the 'double' copy_to_user() is when we wrap, so first copy is from the end of the buffer then we copy the data from the start of the buffer to get all data.

Quoted 4 lines
Quoted 1 line

Ok. But the bugs in the original code are real. I will resend.

Actually that's not true. The bugs in the original code are something that only affect users who deserve it?

Yeah. Never mind. If you set up user space so the first copy fails and the second succeeds then you deserve it.

regards,
dan carpenter

maintainer
#10 lore raw
In reply to Dan Carpenter
Quoted 30 lines · Dan Carpenter
Quoted 9 lines · Dan Carpenter
Quoted 8 lines · Péter Ujfalusi
Quoted 4 lines · Dan Carpenter

This function tries to return the number of bytes that it was able to copy to the user. However, because there are multiple calls to copy_to_user() in a row that means the bytes are not necessarily consecutive so it's not useful. Just return -EFAULT instead.

Quoted 7 lines
Quoted 5 lines

The function is copying data from a circular buffer to a use buffer. The single copy_to_user() is used when we don't have wrapping, the 'double' copy_to_user() is when we wrap, so first copy is from the end of the buffer then we copy the data from the start of the buffer to get all data.

Ok. But the bugs in the original code are real. I will resend.

Actually that's not true. The bugs in the original code are something that only affect users who deserve it? I might not resend. A fix would look something like below?

regards,
dan carpenter

sound/soc/sof/sof-client-probes.c +0 -0
index 1f1ea93a7fbf..32fa3186c295 100644

sound/soc/sof/sof-client-probes.c static int sof_probes_compr_copy(struct snd_soc_component *component, · line 405

Quoted 9 lines
sound/soc/sof/sof-client-probes.c +5 -1
@@ -398,9 +398,14 @@ static int sof_probes_compr_copy(struct snd_soc_component *component,
 		ret = copy_to_user(buf, ptr, count);
 	} else {
 		ret = copy_to_user(buf, ptr, n);
-		ret += copy_to_user(buf + n, rtd->dma_area, count - n);
+		if (ret) {
+			ret += count - n;
+			goto done;
+		}
+		ret = copy_to_user(buf + n, rtd->dma_area, count - n);

I think this should work, can you please resend it?

Quoted 6 lines
sound/soc/sof/sof-client-probes.c +1 -0
 	}
 
+done:
 	if (ret)
 		return count - ret;
 	return count;
Signature · 1 line
Péter