Skip to contentSkip to conversation

drivers/iio: Remove all strcpy() uses in favor of strscpy()

PATCH Superseded Len Baker 5 participants 9 messages Aug 7, 2021 – Aug 10, 2021

Subsystems (2)

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Len Baker

author
#0 lore raw

PATCH v2 1/1 drivers/iio: Remove all strcpy() uses in favor of strscpy()

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy().

This patch is an effort to clean up the proliferation of str*() functions in the kernel and a previous step in the path to remove the strcpy function from the kernel entirely [1].

[1] https://github.com/KSPP/linux/issues/88

---
Changelog v1 -> v2
- Modify the commit changelog to inform that the motivation of this patch is to remove the strcpy() function from the kernel entirely (Jonathan Cameron).

 drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +7 -5
index f282e9cc34c5..3a6aa1c4bf6c 100644
@@ -264,6 +264,7 @@ int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st)
 	const char *orient;
 	char *str;
 	int i;
+	size_t n;
 
 	/* fill magnetometer orientation */
 	switch (st->chip_type) {
@@ -282,17 +283,18 @@ int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st)
 		for (i = 0; i < 3; ++i) {
 			orient = st->orientation.rotation[6 + i];
 			/* use length + 2 for adding minus sign if needed */
-			str = devm_kzalloc(regmap_get_device(st->map),
-					   strlen(orient) + 2, GFP_KERNEL);
+			n = strlen(orient) + 2;
+			str = devm_kzalloc(regmap_get_device(st->map), n,
+					   GFP_KERNEL);
 			if (str == NULL)
 				return -ENOMEM;
 			if (strcmp(orient, "0") == 0) {
-				strcpy(str, orient);
+				strscpy(str, orient, n);
 			} else if (orient[0] == '-') {
-				strcpy(str, &orient[1]);
+				strscpy(str, &orient[1], n);
 			} else {
 				str[0] = '-';
-				strcpy(&str[1], orient);
+				strscpy(&str[1], orient, n - 1);
 			}
 			st->magn_orient.rotation[6 + i] = str;
 		}

--
2.25.1

maintainer
#1 lore raw
Quoted 11 lines · On Sat, 7 Aug 2021 17:22:25 +0200 Len Baker <[email protected]> wrote:

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy().

This patch is an effort to clean up the proliferation of str*() functions in the kernel and a previous step in the path to remove the strcpy function from the kernel entirely [1].

[1] https://github.com/KSPP/linux/issues/88

Applied to the togreg branch of iio.git and pushed out as testing so 0-day can poke at it and see if we missed anything.

Thanks,

Jonathan

Quoted 48 lines

---
Changelog v1 -> v2
- Modify the commit changelog to inform that the motivation of this patch is to remove the strcpy() function from the kernel entirely (Jonathan Cameron).

 drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +7 -6
index f282e9cc34c5..3a6aa1c4bf6c 100644
@@ -264,6 +264,7 @@ int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st)
 	const char *orient;
 	char *str;
 	int i;
+	size_t n;
 
 	/* fill magnetometer orientation */
 	switch (st->chip_type) {
@@ -282,17 +283,18 @@ int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st)
 		for (i = 0; i < 3; ++i) {
 			orient = st->orientation.rotation[6 + i];
 			/* use length + 2 for adding minus sign if needed */
-			str = devm_kzalloc(regmap_get_device(st->map),
-					   strlen(orient) + 2, GFP_KERNEL);
+			n = strlen(orient) + 2;
+			str = devm_kzalloc(regmap_get_device(st->map), n,
+					   GFP_KERNEL);
 			if (str == NULL)
 				return -ENOMEM;
 			if (strcmp(orient, "0") == 0) {
-				strcpy(str, orient);
+				strscpy(str, orient, n);
 			} else if (orient[0] == '-') {
-				strcpy(str, &orient[1]);
+				strscpy(str, &orient[1], n);
 			} else {
 				str[0] = '-';
-				strcpy(&str[1], orient);
+				strscpy(&str[1], orient, n - 1);
 			}
 			st->magn_orient.rotation[6 + i] = str;
 		}
--

2.25.1

#2 lore raw
Quoted 12 lines · Jonathan Cameron
Quoted 8 lines · On Sat, 7 Aug 2021 17:22:25 +0200 Len Baker <[email protected]> wrote:

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy().

This patch is an effort to clean up the proliferation of str*() functions in the kernel and a previous step in the path to remove the strcpy function from the kernel entirely [1].

Quoted 5 lines
Quoted 3 lines

[1] https://github.com/KSPP/linux/issues/88

Applied to the togreg branch of iio.git and pushed out as testing so 0-day can poke at it and see if we missed anything.

Isn't it too early? Or am I missing something (see below)?

...

Quoted 10 lines
Quoted 10 lines
                      /* use length + 2 for adding minus sign if needed */
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -3
-                     str = devm_kzalloc(regmap_get_device(st->map),
-                                        strlen(orient) + 2, GFP_KERNEL);
+                     n = strlen(orient) + 2;
+                     str = devm_kzalloc(regmap_get_device(st->map), n,
+                                        GFP_KERNEL);
                      if (str == NULL)
                              return -ENOMEM;
                      if (strcmp(orient, "0") == 0) {
-                             strcpy(str, orient);
Quoted 8 lines

drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st) · line 297

Quoted 8 lines
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -2
+                             strscpy(str, orient, n);
                      } else if (orient[0] == '-') {
-                             strcpy(str, &orient[1]);
+                             strscpy(str, &orient[1], n);
                      } else {
                              str[0] = '-';
-                             strcpy(&str[1], orient);
+                             strscpy(&str[1], orient, n - 1);

Why n-1?

Quoted 1 line
Quoted 1 line
                      }

As far as I understood the logic, it inverts the sign except the case when it equals 0.

I have a question here, why can't we always use +/-? Why can't 0 be prefixed with a sign?

If the above can be used, we may simplify this code.

Len, I think this task may be considered simple, but I recommend thinking about each case and finding a way to simplify it more.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw
Quoted 42 lines · On Sun, 8 Aug 2021 22:00:34 +0300 Andy Shevchenko <[email protected]>…
Quoted 12 lines · Jonathan Cameron
Quoted 8 lines · On Sat, 7 Aug 2021 17:22:25 +0200 Len Baker <[email protected]> wrote:

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy().

This patch is an effort to clean up the proliferation of str*() functions in the kernel and a previous step in the path to remove the strcpy function from the kernel entirely [1].

Quoted 5 lines
Quoted 3 lines

[1] https://github.com/KSPP/linux/issues/88

Applied to the togreg branch of iio.git and pushed out as testing so 0-day can poke at it and see if we missed anything.

Isn't it too early? Or am I missing something (see below)?

...

Quoted 10 lines
Quoted 10 lines
                      /* use length + 2 for adding minus sign if needed */
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -3
-                     str = devm_kzalloc(regmap_get_device(st->map),
-                                        strlen(orient) + 2, GFP_KERNEL);
+                     n = strlen(orient) + 2;
+                     str = devm_kzalloc(regmap_get_device(st->map), n,
+                                        GFP_KERNEL);
                      if (str == NULL)
                              return -ENOMEM;
                      if (strcmp(orient, "0") == 0) {
-                             strcpy(str, orient);
Quoted 8 lines

drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st) · line 297

Quoted 8 lines
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -2
+                             strscpy(str, orient, n);
                      } else if (orient[0] == '-') {
-                             strcpy(str, &orient[1]);
+                             strscpy(str, &orient[1], n);
                      } else {
                              str[0] = '-';
-                             strcpy(&str[1], orient);
+                             strscpy(&str[1], orient, n - 1);  
Quoted 1 line

Why n-1?

n is the total length and this is printing from [1], so n - 1 is remaining space.

Quoted 8 lines
Quoted 2 lines
Quoted 2 lines
                      }  

As far as I understood the logic, it inverts the sign except the case when it equals 0.

I have a question here, why can't we always use +/-? Why can't 0 be prefixed with a sign?

Technically a userspace ABI change, but I agree it should be fairly harmless unless someone is rolling their own string handling routines.

Quoted 6 lines

If the above can be used, we may simplify this code.

Len, I think this task may be considered simple, but I recommend thinking about each case and finding a way to simplify it more.

It could be a little simpler doing this, so 'maybe' worth doing.

Jonathan

author
#4 lore raw

Hi,

Quoted 53 lines · Jonathan Cameron
Quoted 42 lines · On Sun, 8 Aug 2021 22:00:34 +0300 Andy Shevchenko <[email protected]>…
Quoted 12 lines · Jonathan Cameron
Quoted 8 lines · On Sat, 7 Aug 2021 17:22:25 +0200 Len Baker <[email protected]> wrote:

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy().

This patch is an effort to clean up the proliferation of str*() functions in the kernel and a previous step in the path to remove the strcpy function from the kernel entirely [1].

Quoted 5 lines
Quoted 3 lines

[1] https://github.com/KSPP/linux/issues/88

Applied to the togreg branch of iio.git and pushed out as testing so 0-day can poke at it and see if we missed anything.

Isn't it too early? Or am I missing something (see below)?

...

Quoted 10 lines
Quoted 10 lines
                      /* use length + 2 for adding minus sign if needed */
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -3
-                     str = devm_kzalloc(regmap_get_device(st->map),
-                                        strlen(orient) + 2, GFP_KERNEL);
+                     n = strlen(orient) + 2;
+                     str = devm_kzalloc(regmap_get_device(st->map), n,
+                                        GFP_KERNEL);
                      if (str == NULL)
                              return -ENOMEM;
                      if (strcmp(orient, "0") == 0) {
-                             strcpy(str, orient);
Quoted 8 lines

drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st) · line 297

Quoted 8 lines
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -2
+                             strscpy(str, orient, n);
                      } else if (orient[0] == '-') {
-                             strcpy(str, &orient[1]);
+                             strscpy(str, &orient[1], n);
                      } else {
                              str[0] = '-';
-                             strcpy(&str[1], orient);
+                             strscpy(&str[1], orient, n - 1);
Quoted 1 line

Why n-1?

n is the total length and this is printing from [1], so n - 1 is remaining space.

Quoted 3 lines
Quoted 2 lines
Quoted 2 lines
                      }
Quoted 8 lines
Quoted 5 lines

As far as I understood the logic, it inverts the sign except the case when it equals 0.

I have a question here, why can't we always use +/-? Why can't 0 be prefixed with a sign?

Technically a userspace ABI change, but I agree it should be fairly harmless unless someone is rolling their own string handling routines.

I personally don't like the idea of zero having a sign. It's my opinion. But if you prefer it I have no problem.

Quoted 6 lines
Quoted 5 lines

If the above can be used, we may simplify this code.

Len, I think this task may be considered simple, but I recommend thinking about each case and finding a way to simplify it more.

Andy, is what I try to do ;) Thanks for the advise.

Quoted 1 line

It could be a little simpler doing this, so 'maybe' worth doing.

Ok, I will send a new version if there are no objections.

Regards,
Len

#5 lore raw
Quoted 6 lines · Len Baker
Quoted 5 lines · Jonathan Cameron
Quoted 3 lines · On Sun, 8 Aug 2021 22:00:34 +0300 Andy Shevchenko <[email protected]>…
Quoted 2 lines · Jonathan Cameron

On Sat, 7 Aug 2021 17:22:25 +0200
Len Baker <[email protected]> wrote:

...

Quoted 12 lines
Quoted 12 lines
Quoted 12 lines

Isn't it too early? Or am I missing something (see below)?

Quoted 10 lines
Quoted 10 lines
                      /* use length + 2 for adding minus sign if needed */
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -3
-                     str = devm_kzalloc(regmap_get_device(st->map),
-                                        strlen(orient) + 2, GFP_KERNEL);
+                     n = strlen(orient) + 2;
+                     str = devm_kzalloc(regmap_get_device(st->map), n,
+                                        GFP_KERNEL);
                      if (str == NULL)
                              return -ENOMEM;
                      if (strcmp(orient, "0") == 0) {
-                             strcpy(str, orient);
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines

drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c int inv_mpu_magn_set_orient(struct inv_mpu6050_state *st) · line 297

Quoted 8 lines
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +3 -2
+                             strscpy(str, orient, n);
                      } else if (orient[0] == '-') {
-                             strcpy(str, &orient[1]);
+                             strscpy(str, &orient[1], n);
                      } else {
                              str[0] = '-';
-                             strcpy(&str[1], orient);
+                             strscpy(&str[1], orient, n - 1);

Even if we leave the logic as is, this might be better

if (orient[0] == '-')
str = devm_kstrdup(dev, orient + 1, GFP_KERNEL);
else if (orient[0] != '0' || orient[1] != '\0')
str = devm_kasprintf(dev, GFP_KERNEL, "-%s", orient);
else
str = devm_kstrdup(dev, orient, GFP_KERNEL);
if (!str)
return -ENOMEM;

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#6 lore raw
Quoted 6 lines · Andy Shevchenko

On Mon, Aug 9, 2021 at 7:14 PM Len Baker <[email protected]> wrote:

Even if we leave the logic as is, this might be better

Quoted 7 lines

if (orient[0] == '-')
str = devm_kstrdup(dev, orient + 1, GFP_KERNEL);

else if (orient[0] != '0' || orient[1] != '\0')
str = devm_kasprintf(dev, GFP_KERNEL, "-%s", orient);
else
str = devm_kstrdup(dev, orient, GFP_KERNEL);

Or these two swapped for better reading

else if (orient[0] == '0' **&& orient[1] == '\0')

    str = devm_kstrdup(dev, orient, GFP_KERNEL);
else
    str = devm_kasprintf(dev, GFP_KERNEL, "-%s", orient);

And with a comment added that we negate the result according to the
rules: 1) drop leading '-' (minus); 2) leave 0 as is; add leading '-' (minus).

Quoted 2 lines

if (!str)
return -ENOMEM;

Signature · 2 lines
With Best Regards,
Andy Shevchenko
author
#7 lore raw

Hi Andy, David and Jonathan

First of all thanks you very much for your code snippets ;) More below.

Quoted 8 lines · Andy Shevchenko
Quoted 6 lines · Andy Shevchenko

On Mon, Aug 9, 2021 at 7:14 PM Len Baker <[email protected]> wrote:

Even if we leave the logic as is, this might be better

Quoted 7 lines
Quoted 7 lines

if (orient[0] == '-')
str = devm_kstrdup(dev, orient + 1, GFP_KERNEL);

else if (orient[0] != '0' || orient[1] != '\0')
str = devm_kasprintf(dev, GFP_KERNEL, "-%s", orient);
else
str = devm_kstrdup(dev, orient, GFP_KERNEL);

I think the above snippet is the right one: Is compact and easy to understand.

Quoted 14 lines

Or these two swapped for better reading

else if (orient[0] == '0' **&& orient[1] == '\0')

    str = devm_kstrdup(dev, orient, GFP_KERNEL);
else
    str = devm_kasprintf(dev, GFP_KERNEL, "-%s", orient);

And with a comment added that we negate the result according to the
rules: 1) drop leading '-' (minus); 2) leave 0 as is; add leading '-' (minus).

Quoted 2 lines

if (!str)
return -ENOMEM;

Also, I think that it is better to leave the logic as is and don't try to use always the +/- sign.

Again thanks to all for the feedback.

Regards,
Len

#8 lore raw
In reply to Jonathan Cameron

From: Jonathan Cameron

Quoted 47 lines

Sent: 09 August 2021 10:22

Quoted 42 lines · On Sun, 8 Aug 2021 22:00:34 +0300 Andy Shevchenko <[email protected]>…
Quoted 12 lines · Jonathan Cameron
Quoted 8 lines · On Sat, 7 Aug 2021 17:22:25 +0200 Len Baker <[email protected]> wrote:

strcpy() performs no bounds checking on the destination buffer. This could result in linear overflows beyond the end of the buffer, leading to all kinds of misbehaviors. The safe replacement is strscpy().

This patch is an effort to clean up the proliferation of str*() functions in the kernel and a previous step in the path to remove the strcpy function from the kernel entirely [1].

Quoted 5 lines
Quoted 3 lines

[1] https://github.com/KSPP/linux/issues/88

Applied to the togreg branch of iio.git and pushed out as testing so 0-day can poke at it and see if we missed anything.

Isn't it too early? Or am I missing something (see below)?

...

Quoted 18 lines
Quoted 18 lines
                      /* use length + 2 for adding minus sign if needed */
drivers/iio/imu/inv_mpu6050/inv_mpu_magn.c +6 -5
-                     str = devm_kzalloc(regmap_get_device(st->map),
-                                        strlen(orient) + 2, GFP_KERNEL);
+                     n = strlen(orient) + 2;
+                     str = devm_kzalloc(regmap_get_device(st->map), n,
+                                        GFP_KERNEL);
                      if (str == NULL)
                              return -ENOMEM;
                      if (strcmp(orient, "0") == 0) {
-                             strcpy(str, orient);
+                             strscpy(str, orient, n);
                      } else if (orient[0] == '-') {
-                             strcpy(str, &orient[1]);
+                             strscpy(str, &orient[1], n);
                      } else {
                              str[0] = '-';
-                             strcpy(&str[1], orient);
+                             strscpy(&str[1], orient, n - 1);
Quoted 4 lines
Quoted 1 line

Why n-1?

n is the total length and this is printing from [1], so n - 1 is remaining space.

If you do:

	/* negate 'orient' */
	n = strlen(orient) + 1;
	str = malloc(n + 1);
	if (n == 2 && orient[0] == '0')
		memcpy(str, orient, n);
	else if (orient[0] == '-')
		memcpy(str, orient + 1, n - 1);
	else {
		str[0] = '-';
		memcpy(str + 1, orient, n);
	}

Then it is probably less confusing.

David

-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)