2022-08-25 16:44:15

by Cezary Rojewski

[permalink] [raw]
Subject: [PATCH v2 0/2] libfs: Introduce tokenize_user_input()

Continuation of recent upstream discussion [1] regarding user string
tokenization.

First, tokenize_user_input() is introduced to allow for splitting
specified user string into a sequence of integers. Makes use of
get_options() internally so the parsing logic is not duplicated.

With that done, redundant parts of the sound driver are removed.

Originally similar functionality was added for the SOF sound driver. As
more users are on the horizon, it is desirable to update existing fs
code and provide a unified solution.


Changes in v2:
- reused get_options() so no parsing logic is duplicated
- simplified __user variant with help of memdup_user_nul()
Both suggested by Andy, thanks for thourough review


[1]: https://lore.kernel.org/alsa-devel/[email protected]/


Cezary Rojewski (2):
libfs: Introduce tokenize_user_input()
ASoC: SOF: Remove strsplit_u32() and tokenize_input()

fs/libfs.c | 45 +++++++++++++++
include/linux/fs.h | 1 +
sound/soc/sof/sof-client-probes.c | 92 ++++---------------------------
3 files changed, 57 insertions(+), 81 deletions(-)

--
2.25.1


2022-08-25 16:44:18

by Cezary Rojewski

[permalink] [raw]
Subject: [PATCH v2 1/2] libfs: Introduce tokenize_user_input()

Add new helper function to allow for splitting specified user string
into a sequence of integers. Internally it makes use of get_options() so
the returned sequence contains the integers extracted plus an additional
element that begins the sequence and specifies the integers count.

Suggested-by: Andy Shevchenko <[email protected]>
Signed-off-by: Cezary Rojewski <[email protected]>
---
fs/libfs.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
include/linux/fs.h | 1 +
2 files changed, 46 insertions(+)

diff --git a/fs/libfs.c b/fs/libfs.c
index 31b0ddf01c31..078b23e26741 100644
--- a/fs/libfs.c
+++ b/fs/libfs.c
@@ -809,6 +809,51 @@ ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
}
EXPORT_SYMBOL(simple_write_to_buffer);

+/**
+ * tokenize_user_input - Split string into a sequence of integers
+ * @from: The user space buffer to read from
+ * @ppos: The current position in the buffer
+ * @count: The maximum number of bytes to read
+ * @tkns: Returned pointer to sequence of integers
+ *
+ * On success @tkns is allocated and initialized with a sequence of
+ * integers extracted from the @from plus an additional element that
+ * begins the sequence and specifies the integers count.
+ *
+ * Caller takes responsibility for freeing @tkns when it is no longer
+ * needed.
+ */
+int tokenize_user_input(const char __user *from, size_t count, int **tkns)
+{
+ int *ints, nints;
+ char *buf;
+ int ret = 0;
+
+ buf = memdup_user_nul(from, count);
+ if (IS_ERR(buf))
+ return PTR_ERR(buf);
+
+ get_options(buf, 0, &nints);
+ if (!nints) {
+ ret = -ENOENT;
+ goto free_buf;
+ }
+
+ ints = kcalloc(nints + 1, sizeof(*ints), GFP_KERNEL);
+ if (!ints) {
+ ret = -ENOMEM;
+ goto free_buf;
+ }
+
+ get_options(buf, nints + 1, ints);
+ *tkns = ints;
+
+free_buf:
+ kfree(buf);
+ return ret;
+}
+EXPORT_SYMBOL(tokenize_user_input);
+
/**
* memory_read_from_buffer - copy data from the buffer
* @to: the kernel space buffer to read to
diff --git a/include/linux/fs.h b/include/linux/fs.h
index 9eced4cc286e..ab04cc7f9efa 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -3345,6 +3345,7 @@ extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
loff_t *ppos, const void *from, size_t available);
extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
const void __user *from, size_t count);
+extern int tokenize_user_input(const char __user *from, size_t count, int **tkns);

extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
--
2.25.1

2022-08-25 17:01:46

by Cezary Rojewski

[permalink] [raw]
Subject: [PATCH v2 2/2] ASoC: SOF: Remove strsplit_u32() and tokenize_input()

Make use of global user input tokenization helper instead of the
internal one as both serve same purpose. With that, both strsplit_u32()
and tokenize_input() become unused so remove them.

Signed-off-by: Cezary Rojewski <[email protected]>
---
sound/soc/sof/sof-client-probes.c | 92 ++++---------------------------
1 file changed, 11 insertions(+), 81 deletions(-)

diff --git a/sound/soc/sof/sof-client-probes.c b/sound/soc/sof/sof-client-probes.c
index eb246b823461..3bab54cac07b 100644
--- a/sound/soc/sof/sof-client-probes.c
+++ b/sound/soc/sof/sof-client-probes.c
@@ -410,79 +410,6 @@ static const struct snd_compress_ops sof_probes_compressed_ops = {
.copy = sof_probes_compr_copy,
};

-/**
- * strsplit_u32 - Split string into sequence of u32 tokens
- * @buf: String to split into tokens.
- * @delim: String containing delimiter characters.
- * @tkns: Returned u32 sequence pointer.
- * @num_tkns: Returned number of tokens obtained.
- */
-static int strsplit_u32(char *buf, const char *delim, u32 **tkns, size_t *num_tkns)
-{
- char *s;
- u32 *data, *tmp;
- size_t count = 0;
- size_t cap = 32;
- int ret = 0;
-
- *tkns = NULL;
- *num_tkns = 0;
- data = kcalloc(cap, sizeof(*data), GFP_KERNEL);
- if (!data)
- return -ENOMEM;
-
- while ((s = strsep(&buf, delim)) != NULL) {
- ret = kstrtouint(s, 0, data + count);
- if (ret)
- goto exit;
- if (++count >= cap) {
- cap *= 2;
- tmp = krealloc(data, cap * sizeof(*data), GFP_KERNEL);
- if (!tmp) {
- ret = -ENOMEM;
- goto exit;
- }
- data = tmp;
- }
- }
-
- if (!count)
- goto exit;
- *tkns = kmemdup(data, count * sizeof(*data), GFP_KERNEL);
- if (!(*tkns)) {
- ret = -ENOMEM;
- goto exit;
- }
- *num_tkns = count;
-
-exit:
- kfree(data);
- return ret;
-}
-
-static int tokenize_input(const char __user *from, size_t count,
- loff_t *ppos, 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[count] = '\0';
- ret = strsplit_u32(buf, ",", tkns, num_tkns);
-exit:
- kfree(buf);
- return ret;
-}
-
static ssize_t sof_probes_dfs_points_read(struct file *file, char __user *to,
size_t count, loff_t *ppos)
{
@@ -548,8 +475,8 @@ sof_probes_dfs_points_write(struct file *file, const char __user *from,
struct sof_probes_priv *priv = cdev->data;
struct device *dev = &cdev->auxdev.dev;
struct sof_probe_point_desc *desc;
- size_t num_tkns, bytes;
- u32 *tkns;
+ size_t bytes;
+ u32 num_tkns, *tkns;
int ret, err;

if (priv->extractor_stream_tag == SOF_PROBES_INVALID_NODE_ID) {
@@ -557,16 +484,18 @@ sof_probes_dfs_points_write(struct file *file, const char __user *from,
return -ENOENT;
}

- ret = tokenize_input(from, count, ppos, &tkns, &num_tkns);
+ ret = tokenize_user_input(from, count, (int **)&tkns);
if (ret < 0)
return ret;
+
+ num_tkns = *tkns;
bytes = sizeof(*tkns) * num_tkns;
if (!num_tkns || (bytes % sizeof(*desc))) {
ret = -EINVAL;
goto exit;
}

- desc = (struct sof_probe_point_desc *)tkns;
+ desc = (struct sof_probe_point_desc *)&tkns[1];

ret = pm_runtime_resume_and_get(dev);
if (ret < 0 && ret != -EACCES) {
@@ -603,8 +532,7 @@ sof_probes_dfs_points_remove_write(struct file *file, const char __user *from,
struct sof_client_dev *cdev = file->private_data;
struct sof_probes_priv *priv = cdev->data;
struct device *dev = &cdev->auxdev.dev;
- size_t num_tkns;
- u32 *tkns;
+ u32 num_tkns, *tkns;
int ret, err;

if (priv->extractor_stream_tag == SOF_PROBES_INVALID_NODE_ID) {
@@ -612,9 +540,11 @@ sof_probes_dfs_points_remove_write(struct file *file, const char __user *from,
return -ENOENT;
}

- ret = tokenize_input(from, count, ppos, &tkns, &num_tkns);
+ ret = tokenize_user_input(from, count, (int **)&tkns);
if (ret < 0)
return ret;
+
+ num_tkns = *tkns;
if (!num_tkns) {
ret = -EINVAL;
goto exit;
@@ -626,7 +556,7 @@ sof_probes_dfs_points_remove_write(struct file *file, const char __user *from,
goto exit;
}

- ret = sof_probes_points_remove(cdev, tkns, num_tkns);
+ ret = sof_probes_points_remove(cdev, &tkns[1], num_tkns);
if (!ret)
ret = count;

--
2.25.1

2022-08-30 19:17:04

by Matthew Wilcox

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] libfs: Introduce tokenize_user_input()

On Thu, Aug 25, 2022 at 06:48:32PM +0200, Cezary Rojewski wrote:
> Add new helper function to allow for splitting specified user string
> into a sequence of integers. Internally it makes use of get_options() so
> the returned sequence contains the integers extracted plus an additional
> element that begins the sequence and specifies the integers count.
>
> Suggested-by: Andy Shevchenko <[email protected]>
> Signed-off-by: Cezary Rojewski <[email protected]>
> ---
> fs/libfs.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
> include/linux/fs.h | 1 +

This really has nothing to do with filesystems. Surely
string_helpers.[ch] is the appropriate place for this code?
Also get_options() should probably move its prototype from kernel.h to
string_helpers.h.

2022-08-31 16:18:49

by Cezary Rojewski

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] libfs: Introduce tokenize_user_input()

On 2022-08-30 8:33 PM, Matthew Wilcox wrote:
> On Thu, Aug 25, 2022 at 06:48:32PM +0200, Cezary Rojewski wrote:
>> Add new helper function to allow for splitting specified user string
>> into a sequence of integers. Internally it makes use of get_options() so
>> the returned sequence contains the integers extracted plus an additional
>> element that begins the sequence and specifies the integers count.
>>
>> Suggested-by: Andy Shevchenko <[email protected]>
>> Signed-off-by: Cezary Rojewski <[email protected]>
>> ---
>> fs/libfs.c | 45 +++++++++++++++++++++++++++++++++++++++++++++
>> include/linux/fs.h | 1 +
>
> This really has nothing to do with filesystems. Surely
> string_helpers.[ch] is the appropriate place for this code?
> Also get_options() should probably move its prototype from kernel.h to
> string_helpers.h.


Hello Matthew,

Thanks for your input. The initial version of the change was located in
the string_helpers.[ch] just like you propose and I have no problem
moving it back again. string_helpers are devoid of __user content though
and that's why in v2 it's part of libfs instead.

I'll give a day or two to see if there are other suggestions and then
send v3 relocating the implementation.


Regards,
Czarek