Skip to contentSkip to conversation

pinctrl: pinmux: Add pinmux-select debugfs file

PATCH Superseded Drew Fustini 5 participants 20 messages Feb 10, 2021

Subsystems (2)

v2 of 9 Superseded Latest is v9 →

lore
Share & tools
lore
#0 lore raw

v2 cover pinctrl: pinmux: Add pinmux-select debugfs file

This series first converts the debugfs files in the pinctrl subsystem to octal permissions and then adds a new debug file "pinmux-select". Function name and group name can be written to "pinmux-select" which will cause the function and group to be activated on the pin controller.

Notes for PATCH v2:
- create patch series that includes patch to switch all the debugfs files in pinctrl subsystem over to octal permission
- write function name and group name, instead of error-prone selector numbers, to the 'pinmux-select' file
- switch from static to dynamic allocation for the kernel buffer filled by strncpy_from_user()
- look up function selector from function name using pinmux_func_name_to_selector()
- validate group name with get_function_groups() and match_string()
- look up selector for group name with pinctrl_get_group_selector()

Notes for PATCH v1:
- posted seperate patch to switch all the debugfs files in pinctrl subsystem over to octal permission [1]
- there is no existing documentation for any of the debugfs enteries for pinctrl, so it seemed to have a bigger scope than just this patch. I also noticed that rst documentation is confusingly named "pinctl" (no 'r') and started thread about that [2]. Linus suggested chaning that to 'pin-control'. Thus I am planning a seperate documentation patch series where the file is renamed, references changed and a section on the pinctrl debugfs files is added.

Notes for RFC v2 [3]:
- rename debugfs file "pinmux-set" to "pinmux-select"
- renmae pinmux_set_write() to pinmux_select()
- switch from memdup_user_nul() to strncpy_from_user()
- switch from pr_warn() to dev_err()

[1] https://lore.kernel.org/linux-gpio/[email protected]/
[2] https://lore.kernel.org/linux-gpio/20210126050817.GA187797@x1/
[3] https://lore.kernel.org/linux-gpio/[email protected]/

Drew Fustini (2):
pinctrl: use to octal permissions for debugfs files
pinctrl: pinmux: Add pinmux-select debugfs file

 drivers/pinctrl/core.c    |   6 +--
 drivers/pinctrl/pinconf.c |   4 +-
 drivers/pinctrl/pinmux.c  | 110 +++++++++++++++++++++++++++++++++++++-
 3 files changed, 113 insertions(+), 7 deletions(-)
Signature · 1 line
2.25.1
#1 lore raw

PATCH v2 1/2 pinctrl: use to octal permissions for debugfs files

Switch over pinctrl debugfs files to use octal permissions as they are preferred over symbolic permissions. Refer to commit f90774e1fd27 ("checkpatch: look for symbolic permissions and suggest octal instead").

---
 drivers/pinctrl/core.c    | 6 +++---
 drivers/pinctrl/pinconf.c | 4 ++--
 drivers/pinctrl/pinmux.c  | 4 ++--
 3 files changed, 7 insertions(+), 7 deletions(-)
drivers/pinctrl/core.c +3 -3
index 3663d87f51a0..c9c28f653799 100644
@@ -1914,11 +1914,11 @@ static void pinctrl_init_debugfs(void)
 		return;
 	}
 
-	debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-devices", 0400,
 			    debugfs_root, NULL, &pinctrl_devices_fops);
-	debugfs_create_file("pinctrl-maps", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-maps", 0400,
 			    debugfs_root, NULL, &pinctrl_maps_fops);
-	debugfs_create_file("pinctrl-handles", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-handles", 0400,
 			    debugfs_root, NULL, &pinctrl_fops);
 }
 
drivers/pinctrl/pinconf.c +2 -2
index 02c075cc010b..f005921bb49e 100644
@@ -370,9 +370,9 @@ DEFINE_SHOW_ATTRIBUTE(pinconf_groups);
 void pinconf_init_device_debugfs(struct dentry *devroot,
 			 struct pinctrl_dev *pctldev)
 {
-	debugfs_create_file("pinconf-pins", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinconf-pins", 0400,
 			    devroot, pctldev, &pinconf_pins_fops);
-	debugfs_create_file("pinconf-groups", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinconf-groups", 0400,
 			    devroot, pctldev, &pinconf_groups_fops);
 }
 
drivers/pinctrl/pinmux.c +2 -2
index bab888fe3f8e..7f6190eaedbb 100644
@@ -676,9 +676,9 @@ DEFINE_SHOW_ATTRIBUTE(pinmux_pins);
 void pinmux_init_device_debugfs(struct dentry *devroot,
 			 struct pinctrl_dev *pctldev)
 {
-	debugfs_create_file("pinmux-functions", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinmux-functions", 0400,
 			    devroot, pctldev, &pinmux_functions_fops);
-	debugfs_create_file("pinmux-pins", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinmux-pins", 0400,
 			    devroot, pctldev, &pinmux_pins_fops);
 }
 
Signature · 1 line
2.25.1
#2 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

Quoted 22 lines · Drew Fustini

Switch over pinctrl debugfs files to use octal permissions as they are preferred over symbolic permissions. Refer to commit f90774e1fd27 ("checkpatch: look for symbolic permissions and suggest octal instead").

---

 drivers/pinctrl/core.c    | 6 +++---
 drivers/pinctrl/pinconf.c | 4 ++--
 drivers/pinctrl/pinmux.c  | 4 ++--
 3 files changed, 7 insertions(+), 7 deletions(-)
drivers/pinctrl/core.c +0 -0
index 3663d87f51a0..c9c28f653799 100644
@@ -1914,11 +1914,11 @@ static void pinctrl_init_debugfs(void)
  		return;
  	}

drivers/pinctrl/core.c +1 -1
-	debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-devices", 0400,
Quoted 1 line

  debugfs_root, NULL, &pinctrl_devices_fops);

NAK. You've changed the permission levels.

S_IRUGO is 0444 not 0400.
And you have to keep the S_IFREG or'd along with the octal.

include/linux/stat.h:#define S_IRUGO (S_IRUSR|S_IRGRP|S_IROTH)

checkpatch does this conversion using this command line:

$ ./scripts/checkpatch.pl -f --show-types --terse drivers/pinctrl/*.[ch] --types=SYMBOLIC_PERMS --fix-inplace
drivers/pinctrl/core.c:1893: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/core.c:1895: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/core.c:1897: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/core.c:1919: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/core.c:1921: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/core.c:1923: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
total: 0 errors, 6 warnings, 2302 lines checked
drivers/pinctrl/pinconf.c:373: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/pinconf.c:375: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
total: 0 errors, 2 warnings, 379 lines checked
drivers/pinctrl/pinmux.c:679: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
drivers/pinctrl/pinmux.c:681: WARNING:SYMBOLIC_PERMS: Symbolic permissions 'S_IRUGO' are not preferred. Consider using octal permissions '0444'.
total: 0 errors, 2 warnings, 854 lines checked
$ git diff --stat -p drivers/pinctrl/
 drivers/pinctrl/core.c    | 12 ++++++------
 drivers/pinctrl/pinconf.c |  4 ++--
 drivers/pinctrl/pinmux.c  |  4 ++--
 3 files changed, 10 insertions(+), 10 deletions(-)
drivers/pinctrl/core.c +6 -6
index 7d3370289938..6992b805ae41 100644
@@ -1890,11 +1890,11 @@ static void pinctrl_init_device_debugfs(struct pinctrl_dev *pctldev)
 			dev_name(pctldev->dev));
 		return;
 	}
-	debugfs_create_file("pins", S_IFREG | S_IRUGO,
+	debugfs_create_file("pins", S_IFREG | 0444,
 			    device_root, pctldev, &pinctrl_pins_fops);
-	debugfs_create_file("pingroups", S_IFREG | S_IRUGO,
+	debugfs_create_file("pingroups", S_IFREG | 0444,
 			    device_root, pctldev, &pinctrl_groups_fops);
-	debugfs_create_file("gpio-ranges", S_IFREG | S_IRUGO,
+	debugfs_create_file("gpio-ranges", S_IFREG | 0444,
 			    device_root, pctldev, &pinctrl_gpioranges_fops);
 	if (pctldev->desc->pmxops)
 		pinmux_init_device_debugfs(device_root, pctldev);
@@ -1916,11 +1916,11 @@ static void pinctrl_init_debugfs(void)
 		return;
 	}
 
-	debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-devices", S_IFREG | 0444,
 			    debugfs_root, NULL, &pinctrl_devices_fops);
-	debugfs_create_file("pinctrl-maps", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-maps", S_IFREG | 0444,
 			    debugfs_root, NULL, &pinctrl_maps_fops);
-	debugfs_create_file("pinctrl-handles", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinctrl-handles", S_IFREG | 0444,
 			    debugfs_root, NULL, &pinctrl_fops);
 }
 
drivers/pinctrl/pinconf.c +2 -2
index 02c075cc010b..f9ee12b50428 100644
@@ -370,9 +370,9 @@ DEFINE_SHOW_ATTRIBUTE(pinconf_groups);
 void pinconf_init_device_debugfs(struct dentry *devroot,
 			 struct pinctrl_dev *pctldev)
 {
-	debugfs_create_file("pinconf-pins", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinconf-pins", S_IFREG | 0444,
 			    devroot, pctldev, &pinconf_pins_fops);
-	debugfs_create_file("pinconf-groups", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinconf-groups", S_IFREG | 0444,
 			    devroot, pctldev, &pinconf_groups_fops);
 }
 
drivers/pinctrl/pinmux.c +2 -2
index 36a11c9e893a..ea7559a25fed 100644
@@ -676,9 +676,9 @@ DEFINE_SHOW_ATTRIBUTE(pinmux_pins);
 void pinmux_init_device_debugfs(struct dentry *devroot,
 			 struct pinctrl_dev *pctldev)
 {
-	debugfs_create_file("pinmux-functions", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinmux-functions", S_IFREG | 0444,
 			    devroot, pctldev, &pinmux_functions_fops);
-	debugfs_create_file("pinmux-pins", S_IFREG | S_IRUGO,
+	debugfs_create_file("pinmux-pins", S_IFREG | 0444,
 			    devroot, pctldev, &pinmux_pins_fops);
 }
 
#3 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

Quoted 7 lines · Joe Perches

drivers/pinctrl/core.c static void pinctrl_init_debugfs(void) · line 1920

Quoted 3 lines · Drew Fustini
drivers/pinctrl/core.c +1 -1
-     debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+     debugfs_create_file("pinctrl-devices", 0400,
                          debugfs_root, NULL, &pinctrl_devices_fops);

NAK. You've changed the permission levels.

NAK is usually given when the whole idea is broken. Here is not the case and you may have helped to amend the patch.

...

Quoted 1 line

And you have to keep the S_IFREG or'd along with the octal.

Perhaps time to read the code?
https://elixir.bootlin.com/linux/latest/source/fs/debugfs/inode.c#L387

...

Quoted 3 lines

checkpatch does this conversion using this command line:

$ ./scripts/checkpatch.pl -f --show-types --terse drivers/pinctrl/*.[ch] --types=SYMBOLIC_PERMS --fix-inplace

NAK! See above.

Quoted 3 lines
drivers/pinctrl/core.c +1 -1
-       debugfs_create_file("pins", S_IFREG | S_IRUGO,
+       debugfs_create_file("pins", S_IFREG | 0444,
                            device_root, pctldev, &pinctrl_pins_fops);
Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

Quoted 11 lines · Andy Shevchenko
Quoted 7 lines · Joe Perches
Quoted 3 lines · Drew Fustini
(quoted diff) +1 -1
-     debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+     debugfs_create_file("pinctrl-devices", 0400,

                          debugfs_root, NULL, &pinctrl_devices_fops);

NAK. You've changed the permission levels.

NAK is usually given when the whole idea is broken. Here is not the case and you may have helped to amend the patch.

NAK IMO just means the patch should not be applied, not that the concept is broken.

Quoted 6 lines

...

Quoted 1 line

And you have to keep the S_IFREG or'd along with the octal.

Perhaps time to read the code?
https://elixir.bootlin.com/linux/latest/source/fs/debugfs/inode.c#L387

Then the commit message is also broken.

Quoted 5 lines
Quoted 3 lines

checkpatch does this conversion using this command line:

$ ./scripts/checkpatch.pl -f --show-types --terse drivers/pinctrl/*.[ch] --types=SYMBOLIC_PERMS --fix-inplace

NAK! See above.

The command line above is for octal conversion of the symbolic permissions.

Any other conversion would be for a different purpose and that purpose and should be described in the commit message.

#5 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

Quoted 27 lines · Joe Perches
Quoted 11 lines · Andy Shevchenko
Quoted 7 lines · Joe Perches
Quoted 3 lines · Drew Fustini
(quoted diff) +1 -1
-     debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+     debugfs_create_file("pinctrl-devices", 0400,

                          debugfs_root, NULL, &pinctrl_devices_fops);

NAK. You've changed the permission levels.

NAK is usually given when the whole idea is broken. Here is not the case and you may have helped to amend the patch.

NAK IMO just means the patch should not be applied, not that the concept is broken.

Quoted 6 lines

...

Quoted 1 line

And you have to keep the S_IFREG or'd along with the octal.

Perhaps time to read the code?
https://elixir.bootlin.com/linux/latest/source/fs/debugfs/inode.c#L387

Then the commit message is also broken.

Quoted 2 lines
Quoted 2 lines

checkpatch does this conversion using this command line:

Quoted 10 lines
Quoted 3 lines
Quoted 1 line
$ ./scripts/checkpatch.pl -f --show-types --terse drivers/pinctrl/*.[ch] --types=SYMBOLIC_PERMS --fix-inplace

NAK! See above.

The command line above is for octal conversion of the symbolic permissions.

Any other conversion would be for a different purpose and that purpose and should be described in the commit message.

Thanks for review comments from all.

I will change from the incorrect 0400 to 0444.

As for S_IFREG, it does seem like leaving off S_IFREG is the most common case when using octal permissions with debugfs_create_*():

$ git grep debugfs_create drivers/ |grep 0444 |grep -v S_IFREG | wc -l
302
$ git grep debugfs_create drivers/ |grep 0444 |grep S_IFREG | wc -l
9

As noted by Andy, this is okay as the S_IFREG flag is added to the mode __debugfs_create_file() inside fs/debugfs/inode.c. I will note this in the commit message.

Thank you,
Drew

#6 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

Quoted 1 line · Drew Fustini

I will change from the incorrect 0400 to 0444.

Thanks.

Quoted 7 lines

As for S_IFREG, it does seem like leaving off S_IFREG is the most common case when using octal permissions with debugfs_create_*():

$ git grep debugfs_create drivers/ |grep 0444 |grep -v S_IFREG | wc -l
302
$ git grep debugfs_create drivers/ |grep 0444 |grep S_IFREG | wc -l
9

It's ~2:1 when using S_IRUGO

$ git grep debugfs_create_file drivers/ | grep S_IRUGO | grep -v S_IFREG | wc -l
109
$ git grep debugfs_create_file drivers/ | grep S_IRUGO | grep S_IFREG | wc -l
48
#7 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

In reply to Drew Fustini

Hi Drew,

Quoted 5 lines · Drew Fustini

Switch over pinctrl debugfs files to use octal permissions as they are preferred over symbolic permissions. Refer to commit f90774e1fd27 ("checkpatch: look for symbolic permissions and suggest octal instead").

Thanks for your patch!

drivers/pinctrl/core.c static void pinctrl_init_debugfs(void) · line 1917

Quoted 8 lines
drivers/pinctrl/core.c +1 -1
@@ -1914,11 +1914,11 @@ static void pinctrl_init_debugfs(void)
                return;
        }
 
-       debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinctrl-devices", 0400,

What about the loss of S_IFREG?
S_IRUGO = 0444, not 0400.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
#8 lore raw

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

Quoted 1 line · Geert Uytterhoeven

On Wed, Feb 10, 2021 at 8:50 AM Drew Fustini <[email protected]> wrote:

...

Quoted 4 lines

drivers/pinctrl/core.c static void pinctrl_init_debugfs(void) · line 1917

Quoted 2 lines
drivers/pinctrl/core.c +1 -1
-       debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinctrl-devices", 0400,

What about the loss of S_IFREG?

What do you mean?
https://elixir.bootlin.com/linux/latest/source/fs/debugfs/inode.c#L387

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

Re: [PATCH v2 1/2] pinctrl: use to octal permissions for debugfs files

In reply to Drew Fustini
Quoted 4 lines · Drew Fustini

Switch over pinctrl debugfs files to use octal permissions as they are preferred over symbolic permissions. Refer to commit f90774e1fd27 ("checkpatch: look for symbolic permissions and suggest octal instead").

You forgot:

LGTM after addressing what Geert noticed.

Quoted 59 lines
---
 drivers/pinctrl/core.c    | 6 +++---
 drivers/pinctrl/pinconf.c | 4 ++--
 drivers/pinctrl/pinmux.c  | 4 ++--
 3 files changed, 7 insertions(+), 7 deletions(-)
drivers/pinctrl/core.c +3 -3
index 3663d87f51a0..c9c28f653799 100644
@@ -1914,11 +1914,11 @@ static void pinctrl_init_debugfs(void)
                return;
        }
 
-       debugfs_create_file("pinctrl-devices", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinctrl-devices", 0400,
                            debugfs_root, NULL, &pinctrl_devices_fops);
-       debugfs_create_file("pinctrl-maps", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinctrl-maps", 0400,
                            debugfs_root, NULL, &pinctrl_maps_fops);
-       debugfs_create_file("pinctrl-handles", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinctrl-handles", 0400,
                            debugfs_root, NULL, &pinctrl_fops);
 }
 
drivers/pinctrl/pinconf.c +2 -2
index 02c075cc010b..f005921bb49e 100644
@@ -370,9 +370,9 @@ DEFINE_SHOW_ATTRIBUTE(pinconf_groups);
 void pinconf_init_device_debugfs(struct dentry *devroot,
                         struct pinctrl_dev *pctldev)
 {
-       debugfs_create_file("pinconf-pins", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinconf-pins", 0400,
                            devroot, pctldev, &pinconf_pins_fops);
-       debugfs_create_file("pinconf-groups", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinconf-groups", 0400,
                            devroot, pctldev, &pinconf_groups_fops);
 }
 
drivers/pinctrl/pinmux.c +2 -3
index bab888fe3f8e..7f6190eaedbb 100644
@@ -676,9 +676,9 @@ DEFINE_SHOW_ATTRIBUTE(pinmux_pins);
 void pinmux_init_device_debugfs(struct dentry *devroot,
                         struct pinctrl_dev *pctldev)
 {
-       debugfs_create_file("pinmux-functions", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinmux-functions", 0400,
                            devroot, pctldev, &pinmux_functions_fops);
-       debugfs_create_file("pinmux-pins", S_IFREG | S_IRUGO,
+       debugfs_create_file("pinmux-pins", 0400,
                            devroot, pctldev, &pinmux_pins_fops);
 }
 
--
Quoted 2 lines

2.25.1

--
With Best Regards,
Andy Shevchenko

#10 lore raw

PATCH v2 2/2 pinctrl: pinmux: Add pinmux-select debugfs file

In reply to Drew Fustini

Add "pinmux-select" to debugfs which will activate a function and group when 2 integers "<function-selector> <group-selector>" are written to the file. The write operation pinmux_select() handles this by checking if fsel and gsel are valid selectors and then calling ops->set_mux().

The existing "pinmux-functions" debugfs file lists the pin functions registered for the pin controller. For example:

function: pinmux-uart0, groups = [ pinmux-uart0-pins ]
function: pinmux-mmc0, groups = [ pinmux-mmc0-pins ]
function: pinmux-mmc1, groups = [ pinmux-mmc1-pins ]
function: pinmux-i2c0, groups = [ pinmux-i2c0-pins ]
function: pinmux-i2c1, groups = [ pinmux-i2c1-pins ]
function: pinmux-spi1, groups = [ pinmux-spi1-pins ]

To activate function pinmux-i2c1 (fsel 4) and group pinmux-i2c1-pins (gsel 4):

echo '4 4' > pinmux-select

---
 drivers/pinctrl/pinmux.c | 106 +++++++++++++++++++++++++++++++++++++++
 1 file changed, 106 insertions(+)
drivers/pinctrl/pinmux.c +106 -0
index 7f6190eaedbb..b8cd0c3bedf7 100644
@@ -673,6 +673,110 @@ void pinmux_show_setting(struct seq_file *s,
 DEFINE_SHOW_ATTRIBUTE(pinmux_functions);
 DEFINE_SHOW_ATTRIBUTE(pinmux_pins);
 
+
+#define PINMUX_MAX_NAME 64
+static ssize_t pinmux_select(struct file *file, const char __user *user_buf,
+				   size_t len, loff_t *ppos)
+{
+	struct seq_file *sfile = file->private_data;
+	struct pinctrl_dev *pctldev = sfile->private;
+	const struct pinmux_ops *pmxops = pctldev->desc->pmxops;
+	const char *const *groups;
+	char *buf, *fname, *gname;
+	unsigned int num_groups;
+	int fsel, gsel, ret;
+
+	if (len > (PINMUX_MAX_NAME * 2)) {
+		dev_err(pctldev->dev, "write too big for buffer");
+		return -EINVAL;
+	}
+
+	buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);
+	if (!buf)
+		return -ENOMEM;
+
+	fname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
+	if (!fname) {
+		ret = -ENOMEM;
+		goto free_buf;
+	}
+
+	gname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
+	if (!buf) {
+		ret = -ENOMEM;
+		goto free_fname;
+	}
+
+	ret = strncpy_from_user(buf, user_buf, PINMUX_MAX_NAME * 2);
+	if (ret < 0) {
+		dev_err(pctldev->dev, "failed to copy buffer from userspace");
+		goto free_gname;
+	}
+	buf[len-1] = '\0';
+
+	ret = sscanf(buf, "%s %s", fname, gname);
+	if (ret != 2) {
+		dev_err(pctldev->dev, "expected format: <function-name> <group-name>");
+		goto free_gname;
+	}
+
+	fsel = pinmux_func_name_to_selector(pctldev, fname);
+	if (fsel < 0) {
+		dev_err(pctldev->dev, "invalid function %s in map table\n", fname);
+		ret = -EINVAL;
+		goto free_gname;
+	}
+
+	ret = pmxops->get_function_groups(pctldev, fsel, &groups, &num_groups);
+	if (ret) {
+		dev_err(pctldev->dev, "no groups for function %d (%s)", fsel, fname);
+		goto free_gname;
+
+	}
+
+	ret = match_string(groups, num_groups, gname);
+	if (ret < 0) {
+		dev_err(pctldev->dev, "invalid group %s", gname);
+		goto free_gname;
+	}
+
+	ret = pinctrl_get_group_selector(pctldev, gname);
+	if (ret < 0) {
+		dev_err(pctldev->dev, "failed to get group selectorL %s", gname);
+		goto free_gname;
+	}
+	gsel = ret;
+
+	ret = pmxops->set_mux(pctldev, fsel, gsel);
+	if (ret) {
+		dev_err(pctldev->dev, "set_mux() failed: %d", ret);
+		goto free_gname;
+	}
+
+	return len;
+free_buf:
+	devm_kfree(pctldev->dev, buf);
+free_fname:
+	devm_kfree(pctldev->dev, fname);
+free_gname:
+	devm_kfree(pctldev->dev, gname);
+	return ret;
+}
+
+static int pinmux_select_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, NULL, inode->i_private);
+}
+
+static const struct file_operations pinmux_select_ops = {
+	.owner = THIS_MODULE,
+	.open = pinmux_select_open,
+	.read = seq_read,
+	.write = pinmux_select,
+	.llseek = no_llseek,
+	.release = single_release,
+};
+
 void pinmux_init_device_debugfs(struct dentry *devroot,
 			 struct pinctrl_dev *pctldev)
 {
@@ -680,6 +784,8 @@ void pinmux_init_device_debugfs(struct dentry *devroot,
 			    devroot, pctldev, &pinmux_functions_fops);
 	debugfs_create_file("pinmux-pins", 0400,
 			    devroot, pctldev, &pinmux_pins_fops);
+	debugfs_create_file("pinmux-select", 0200,
+			    devroot, pctldev, &pinmux_select_ops);
 }
 
 #endif /* CONFIG_DEBUG_FS */
Signature · 1 line
2.25.1
#11 lore raw

Hi Drew,

Quoted 4 lines · Drew Fustini

Add "pinmux-select" to debugfs which will activate a function and group when 2 integers "<function-selector> <group-selector>" are written to the file. The write operation pinmux_select() handles this by checking if fsel and gsel are valid selectors and then calling ops->set_mux().

Thanks for your patch!

Quoted 10 lines

The existing "pinmux-functions" debugfs file lists the pin functions registered for the pin controller. For example:

function: pinmux-uart0, groups = [ pinmux-uart0-pins ]
function: pinmux-mmc0, groups = [ pinmux-mmc0-pins ]
function: pinmux-mmc1, groups = [ pinmux-mmc1-pins ]
function: pinmux-i2c0, groups = [ pinmux-i2c0-pins ]
function: pinmux-i2c1, groups = [ pinmux-i2c1-pins ]
function: pinmux-spi1, groups = [ pinmux-spi1-pins ]

Quoted 4 lines

To activate function pinmux-i2c1 (fsel 4) and group pinmux-i2c1-pins (gsel 4):

echo '4 4' > pinmux-select

I think you forgot to update the example.

I assume

echo pinmux-i2c1 pinmux-i2c1-pins > mux-select

?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

#12 lore raw
In reply to Drew Fustini
Quoted 20 lines · Drew Fustini

Add "pinmux-select" to debugfs which will activate a function and group when 2 integers "<function-selector> <group-selector>" are written to the file. The write operation pinmux_select() handles this by checking if fsel and gsel are valid selectors and then calling ops->set_mux().

The existing "pinmux-functions" debugfs file lists the pin functions registered for the pin controller. For example:

function: pinmux-uart0, groups = [ pinmux-uart0-pins ]
function: pinmux-mmc0, groups = [ pinmux-mmc0-pins ]
function: pinmux-mmc1, groups = [ pinmux-mmc1-pins ]
function: pinmux-i2c0, groups = [ pinmux-i2c0-pins ]
function: pinmux-i2c1, groups = [ pinmux-i2c1-pins ]
function: pinmux-spi1, groups = [ pinmux-spi1-pins ]

To activate function pinmux-i2c1 (fsel 4) and group pinmux-i2c1-pins (gsel 4):

echo '4 4' > pinmux-select

...

Quoted 4 lines

DEFINE_SHOW_ATTRIBUTE(pinmux_pins);

+

One blank line (existed) is enough.

Quoted 1 line

+#define PINMUX_MAX_NAME 64

...

Quoted 1 line

+ buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);

You have to (re-)read documentation about Device Managed Resources. Keyword here is *device*! Pay attention to it. TL;DR: misuse of device managed resources here.
Potentially memory exhausting (local DoS attack), but see below.

Quoted 2 lines
drivers/pinctrl/pinmux.c +2 -0
+       if (!buf)
+               return -ENOMEM;

...

Quoted 1 line

+ devm_kfree(pctldev->dev, buf);

Calling devm_kfree() or other devm_*() release kinda APIs is a red flag in 99%. See above.

--
With Best Regards,
Andy Shevchenko

#13 lore raw
Quoted 45 lines · Andy Shevchenko
Quoted 20 lines · Drew Fustini

Add "pinmux-select" to debugfs which will activate a function and group when 2 integers "<function-selector> <group-selector>" are written to the file. The write operation pinmux_select() handles this by checking if fsel and gsel are valid selectors and then calling ops->set_mux().

The existing "pinmux-functions" debugfs file lists the pin functions registered for the pin controller. For example:

function: pinmux-uart0, groups = [ pinmux-uart0-pins ]
function: pinmux-mmc0, groups = [ pinmux-mmc0-pins ]
function: pinmux-mmc1, groups = [ pinmux-mmc1-pins ]
function: pinmux-i2c0, groups = [ pinmux-i2c0-pins ]
function: pinmux-i2c1, groups = [ pinmux-i2c1-pins ]
function: pinmux-spi1, groups = [ pinmux-spi1-pins ]

To activate function pinmux-i2c1 (fsel 4) and group pinmux-i2c1-pins (gsel 4):

echo '4 4' > pinmux-select

...

Quoted 4 lines

DEFINE_SHOW_ATTRIBUTE(pinmux_pins);

+

One blank line (existed) is enough.

Quoted 1 line

+#define PINMUX_MAX_NAME 64

...

Quoted 1 line

+ buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);

You have to (re-)read documentation about Device Managed Resources. Keyword here is *device*! Pay attention to it. TL;DR: misuse of device managed resources here.
Potentially memory exhausting (local DoS attack), but see below.

drivers/pinctrl/pinmux.c void pinmux_show_setting(struct seq_file *s, · line 696

Quoted 2 lines
drivers/pinctrl/pinmux.c +2 -0
+       if (!buf)
+               return -ENOMEM;
Quoted 6 lines

...

Quoted 1 line

+ devm_kfree(pctldev->dev, buf);

Calling devm_kfree() or other devm_*() release kinda APIs is a red flag in 99%. See above.

Thank you for reviewing and pointing out this issue.

Do you mean that I should not be treating these buffers used in the debugfs write op as belonging to the pin controller device?

I have looked through the kernel code and I realize now that I don't see any instances of devm_*() being used inside the read or write op for a debugfs file. As I consider it further, devm_*() does not seem to make sense as I am creating the buffers only for temporary use inside pinmux_select().

I'll get that fixed in v3.

Thank you,
Drew

#14 lore raw
In reply to Drew Fustini

Hi Drew,

url:    https://github.com/0day-ci/linux/commits/Drew-Fustini/pinctrl-pinmux-Add-pinmux-select-debugfs-file/20210210-160108
base:   https://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl.git devel
config: i386-randconfig-m021-20210209 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0

If you fix the issue, kindly add following tag as appropriate

smatch warnings:
drivers/pinctrl/pinmux.c:762 pinmux_select() error: uninitialized symbol 'gname'.

vim +/gname +762 drivers/pinctrl/pinmux.c

99b2f99aa41aa7 Drew Fustini  2021-02-09  678  static ssize_t pinmux_select(struct file *file, const char __user *user_buf,
99b2f99aa41aa7 Drew Fustini  2021-02-09  679  				   size_t len, loff_t *ppos)
99b2f99aa41aa7 Drew Fustini  2021-02-09  680  {
99b2f99aa41aa7 Drew Fustini  2021-02-09  681  	struct seq_file *sfile = file->private_data;
99b2f99aa41aa7 Drew Fustini  2021-02-09  682  	struct pinctrl_dev *pctldev = sfile->private;
99b2f99aa41aa7 Drew Fustini  2021-02-09  683  	const struct pinmux_ops *pmxops = pctldev->desc->pmxops;
99b2f99aa41aa7 Drew Fustini  2021-02-09  684  	const char *const *groups;
99b2f99aa41aa7 Drew Fustini  2021-02-09  685  	char *buf, *fname, *gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  686  	unsigned int num_groups;
99b2f99aa41aa7 Drew Fustini  2021-02-09  687  	int fsel, gsel, ret;
99b2f99aa41aa7 Drew Fustini  2021-02-09  688  
99b2f99aa41aa7 Drew Fustini  2021-02-09  689  	if (len > (PINMUX_MAX_NAME * 2)) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  690  		dev_err(pctldev->dev, "write too big for buffer");
99b2f99aa41aa7 Drew Fustini  2021-02-09  691  		return -EINVAL;
99b2f99aa41aa7 Drew Fustini  2021-02-09  692  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  693  
99b2f99aa41aa7 Drew Fustini  2021-02-09  694  	buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  695  	if (!buf)
99b2f99aa41aa7 Drew Fustini  2021-02-09  696  		return -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  697  
99b2f99aa41aa7 Drew Fustini  2021-02-09  698  	fname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  699  	if (!fname) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  700  		ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  701  		goto free_buf;

The gotos are out of order. They should be in mirror/reverse order of the allocations:

free_gmane:

	devm_kfree(pctldev->dev, gname);
free_fname:
	devm_kfree(pctldev->dev, fname);
free_buf:
	devm_kfree(pctldev->dev, buf);

But also why do we need to use devm_kfree() at all? I thought the whole point of devm_ functions was that they are garbage collected automatically for you. Can we not just delete all error handling and return -ENOMEM here?

99b2f99aa41aa7 Drew Fustini  2021-02-09  702  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  703  
99b2f99aa41aa7 Drew Fustini  2021-02-09  704  	gname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  705  	if (!buf) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  706  		ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  707  		goto free_fname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  708  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  709  
99b2f99aa41aa7 Drew Fustini  2021-02-09  710  	ret = strncpy_from_user(buf, user_buf, PINMUX_MAX_NAME * 2);
99b2f99aa41aa7 Drew Fustini  2021-02-09  711  	if (ret < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  712  		dev_err(pctldev->dev, "failed to copy buffer from userspace");
99b2f99aa41aa7 Drew Fustini  2021-02-09  713  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  714  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  715  	buf[len-1] = '\0';
99b2f99aa41aa7 Drew Fustini  2021-02-09  716  
99b2f99aa41aa7 Drew Fustini  2021-02-09  717  	ret = sscanf(buf, "%s %s", fname, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  718  	if (ret != 2) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  719  		dev_err(pctldev->dev, "expected format: <function-name> <group-name>");
99b2f99aa41aa7 Drew Fustini  2021-02-09  720  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  721  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  722  
99b2f99aa41aa7 Drew Fustini  2021-02-09  723  	fsel = pinmux_func_name_to_selector(pctldev, fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  724  	if (fsel < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  725  		dev_err(pctldev->dev, "invalid function %s in map table\n", fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  726  		ret = -EINVAL;
99b2f99aa41aa7 Drew Fustini  2021-02-09  727  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  728  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  729  
99b2f99aa41aa7 Drew Fustini  2021-02-09  730  	ret = pmxops->get_function_groups(pctldev, fsel, &groups, &num_groups);
99b2f99aa41aa7 Drew Fustini  2021-02-09  731  	if (ret) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  732  		dev_err(pctldev->dev, "no groups for function %d (%s)", fsel, fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  733  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  734  
99b2f99aa41aa7 Drew Fustini  2021-02-09  735  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  736  
99b2f99aa41aa7 Drew Fustini  2021-02-09  737  	ret = match_string(groups, num_groups, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  738  	if (ret < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  739  		dev_err(pctldev->dev, "invalid group %s", gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  740  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  741  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  742  
99b2f99aa41aa7 Drew Fustini  2021-02-09  743  	ret = pinctrl_get_group_selector(pctldev, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  744  	if (ret < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  745  		dev_err(pctldev->dev, "failed to get group selectorL %s", gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  746  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  747  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  748  	gsel = ret;
99b2f99aa41aa7 Drew Fustini  2021-02-09  749  
99b2f99aa41aa7 Drew Fustini  2021-02-09  750  	ret = pmxops->set_mux(pctldev, fsel, gsel);
99b2f99aa41aa7 Drew Fustini  2021-02-09  751  	if (ret) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  752  		dev_err(pctldev->dev, "set_mux() failed: %d", ret);
99b2f99aa41aa7 Drew Fustini  2021-02-09  753  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  754  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  755  
99b2f99aa41aa7 Drew Fustini  2021-02-09  756  	return len;
99b2f99aa41aa7 Drew Fustini  2021-02-09  757  free_buf:
99b2f99aa41aa7 Drew Fustini  2021-02-09  758  	devm_kfree(pctldev->dev, buf);
99b2f99aa41aa7 Drew Fustini  2021-02-09  759  free_fname:
99b2f99aa41aa7 Drew Fustini  2021-02-09  760  	devm_kfree(pctldev->dev, fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  761  free_gname:
99b2f99aa41aa7 Drew Fustini  2021-02-09 @762  	devm_kfree(pctldev->dev, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  763  	return ret;
99b2f99aa41aa7 Drew Fustini  2021-02-09  764  }

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

#15 lore raw

Hi Dan,

Quoted 19 lines · Dan Carpenter
99b2f99aa41aa7 Drew Fustini  2021-02-09  694    buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  695    if (!buf)
99b2f99aa41aa7 Drew Fustini  2021-02-09  696            return -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  697
99b2f99aa41aa7 Drew Fustini  2021-02-09  698    fname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  699    if (!fname) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  700            ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  701            goto free_buf;

The gotos are out of order. They should be in mirror/reverse order of the allocations:

free_gmane:

        devm_kfree(pctldev->dev, gname);
free_fname:
        devm_kfree(pctldev->dev, fname);
free_buf:
        devm_kfree(pctldev->dev, buf);
Quoted 4 lines

But also why do we need to use devm_kfree() at all? I thought the whole point of devm_ functions was that they are garbage collected automatically for you. Can we not just delete all error handling and return -ENOMEM here?

No, because the lifetime of the objects allocated here does not match the lifetime of dev. If they're not freed here, they will only be freed when the device is unbound. As the user can access the sysfs files at will, he can OOM the system.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert(a)linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
#16 lore raw
Quoted 27 lines · Geert Uytterhoeven

Hi Dan,

Quoted 19 lines · Dan Carpenter
99b2f99aa41aa7 Drew Fustini  2021-02-09  694    buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  695    if (!buf)
99b2f99aa41aa7 Drew Fustini  2021-02-09  696            return -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  697
99b2f99aa41aa7 Drew Fustini  2021-02-09  698    fname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  699    if (!fname) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  700            ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  701            goto free_buf;

The gotos are out of order. They should be in mirror/reverse order of the allocations:

free_gmane:

        devm_kfree(pctldev->dev, gname);
free_fname:
        devm_kfree(pctldev->dev, fname);
free_buf:
        devm_kfree(pctldev->dev, buf);
Quoted 4 lines

But also why do we need to use devm_kfree() at all? I thought the whole point of devm_ functions was that they are garbage collected automatically for you. Can we not just delete all error handling and return -ENOMEM here?

Quoted 5 lines

No, because the lifetime of the objects allocated here does not match the lifetime of dev. If they're not freed here, they will only be freed when the device is unbound. As the user can access the sysfs files at will, he can OOM the system.

Then why not use vanilla kmalloc()?

regards,
dan carpenter

#17 lore raw
Quoted 28 lines · Dan Carpenter
Quoted 27 lines · Geert Uytterhoeven

Hi Dan,

Quoted 23 lines · Dan Carpenter
99b2f99aa41aa7 Drew Fustini  2021-02-09  694    buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  695    if (!buf)
99b2f99aa41aa7 Drew Fustini  2021-02-09  696            return -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  697
99b2f99aa41aa7 Drew Fustini  2021-02-09  698    fname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  699    if (!fname) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  700            ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  701            goto free_buf;

The gotos are out of order. They should be in mirror/reverse order of the allocations:

free_gmane:

        devm_kfree(pctldev->dev, gname);
free_fname:
        devm_kfree(pctldev->dev, fname);
free_buf:
        devm_kfree(pctldev->dev, buf);

But also why do we need to use devm_kfree() at all? I thought the whole point of devm_ functions was that they are garbage collected automatically for you. Can we not just delete all error handling and return -ENOMEM here?

Quoted 7 lines
Quoted 5 lines

No, because the lifetime of the objects allocated here does not match the lifetime of dev. If they're not freed here, they will only be freed when the device is unbound. As the user can access the sysfs files at will, he can OOM the system.

Then why not use vanilla kmalloc()?

Yes, I believe that is the correct approach. The problem was due to my misunderstanding of when devm_*() was appropriate. In this case, I should have been using the vanilla allocation as the buffers used in this debugfs write operation are not tied to the lifetime of the pin controller device. The are just allocated for internal use inside the write function.

thanks,
drew

#18 lore raw
In reply to Dan Carpenter
Quoted 10 lines · Dan Carpenter

Hi Drew,

url:    https://github.com/0day-ci/linux/commits/Drew-Fustini/pinctrl-pinmux-Add-pinmux-select-debugfs-file/20210210-160108
base:   https://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl.git devel
config: i386-randconfig-m021-20210209 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0

If you fix the issue, kindly add following tag as appropriate

Does it makes sense for me to include that tag in this patch?

It's not really a fix but rather creating a new feature through debugfs. I am wondering if it might confuse people reading the git log as to whether the Reported-by: was with regards to the addition of 'pinmux-select' to debugfs, rather than it was just reporting that my goto handling was incorrect.

I think it makes sense for me to mention that in the PATCH changelog but that won't be in the git commit message.

Quoted 34 lines

smatch warnings:
drivers/pinctrl/pinmux.c:762 pinmux_select() error: uninitialized symbol 'gname'.

vim +/gname +762 drivers/pinctrl/pinmux.c

99b2f99aa41aa7 Drew Fustini  2021-02-09  678  static ssize_t pinmux_select(struct file *file, const char __user *user_buf,
99b2f99aa41aa7 Drew Fustini  2021-02-09  679  				   size_t len, loff_t *ppos)
99b2f99aa41aa7 Drew Fustini  2021-02-09  680  {
99b2f99aa41aa7 Drew Fustini  2021-02-09  681  	struct seq_file *sfile = file->private_data;
99b2f99aa41aa7 Drew Fustini  2021-02-09  682  	struct pinctrl_dev *pctldev = sfile->private;
99b2f99aa41aa7 Drew Fustini  2021-02-09  683  	const struct pinmux_ops *pmxops = pctldev->desc->pmxops;
99b2f99aa41aa7 Drew Fustini  2021-02-09  684  	const char *const *groups;
99b2f99aa41aa7 Drew Fustini  2021-02-09  685  	char *buf, *fname, *gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  686  	unsigned int num_groups;
99b2f99aa41aa7 Drew Fustini  2021-02-09  687  	int fsel, gsel, ret;
99b2f99aa41aa7 Drew Fustini  2021-02-09  688  
99b2f99aa41aa7 Drew Fustini  2021-02-09  689  	if (len > (PINMUX_MAX_NAME * 2)) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  690  		dev_err(pctldev->dev, "write too big for buffer");
99b2f99aa41aa7 Drew Fustini  2021-02-09  691  		return -EINVAL;
99b2f99aa41aa7 Drew Fustini  2021-02-09  692  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  693  
99b2f99aa41aa7 Drew Fustini  2021-02-09  694  	buf = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME * 2, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  695  	if (!buf)
99b2f99aa41aa7 Drew Fustini  2021-02-09  696  		return -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  697  
99b2f99aa41aa7 Drew Fustini  2021-02-09  698  	fname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  699  	if (!fname) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  700  		ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  701  		goto free_buf;

The gotos are out of order. They should be in mirror/reverse order of the allocations:

Quoted 6 lines

free_gmane:

	devm_kfree(pctldev->dev, gname);
free_fname:
	devm_kfree(pctldev->dev, fname);
free_buf:
	devm_kfree(pctldev->dev, buf);

Thank you, yes, I should have caught this.

Quoted 5 lines

But also why do we need to use devm_kfree() at all? I thought the whole point of devm_ functions was that they are garbage collected automatically for you. Can we not just delete all error handling and return -ENOMEM here?

Andy replied already that it is incorrect for me to be using devm_*() in the debugfs write operation. I'll be changing the code to use normal kzalloc() and thus will need to keep the goto error handling (but with the correct order as you pointed out).

Quoted 65 lines
99b2f99aa41aa7 Drew Fustini  2021-02-09  702  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  703  
99b2f99aa41aa7 Drew Fustini  2021-02-09  704  	gname = devm_kzalloc(pctldev->dev, PINMUX_MAX_NAME, GFP_KERNEL);
99b2f99aa41aa7 Drew Fustini  2021-02-09  705  	if (!buf) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  706  		ret = -ENOMEM;
99b2f99aa41aa7 Drew Fustini  2021-02-09  707  		goto free_fname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  708  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  709  
99b2f99aa41aa7 Drew Fustini  2021-02-09  710  	ret = strncpy_from_user(buf, user_buf, PINMUX_MAX_NAME * 2);
99b2f99aa41aa7 Drew Fustini  2021-02-09  711  	if (ret < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  712  		dev_err(pctldev->dev, "failed to copy buffer from userspace");
99b2f99aa41aa7 Drew Fustini  2021-02-09  713  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  714  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  715  	buf[len-1] = '\0';
99b2f99aa41aa7 Drew Fustini  2021-02-09  716  
99b2f99aa41aa7 Drew Fustini  2021-02-09  717  	ret = sscanf(buf, "%s %s", fname, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  718  	if (ret != 2) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  719  		dev_err(pctldev->dev, "expected format: <function-name> <group-name>");
99b2f99aa41aa7 Drew Fustini  2021-02-09  720  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  721  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  722  
99b2f99aa41aa7 Drew Fustini  2021-02-09  723  	fsel = pinmux_func_name_to_selector(pctldev, fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  724  	if (fsel < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  725  		dev_err(pctldev->dev, "invalid function %s in map table\n", fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  726  		ret = -EINVAL;
99b2f99aa41aa7 Drew Fustini  2021-02-09  727  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  728  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  729  
99b2f99aa41aa7 Drew Fustini  2021-02-09  730  	ret = pmxops->get_function_groups(pctldev, fsel, &groups, &num_groups);
99b2f99aa41aa7 Drew Fustini  2021-02-09  731  	if (ret) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  732  		dev_err(pctldev->dev, "no groups for function %d (%s)", fsel, fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  733  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  734  
99b2f99aa41aa7 Drew Fustini  2021-02-09  735  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  736  
99b2f99aa41aa7 Drew Fustini  2021-02-09  737  	ret = match_string(groups, num_groups, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  738  	if (ret < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  739  		dev_err(pctldev->dev, "invalid group %s", gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  740  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  741  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  742  
99b2f99aa41aa7 Drew Fustini  2021-02-09  743  	ret = pinctrl_get_group_selector(pctldev, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  744  	if (ret < 0) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  745  		dev_err(pctldev->dev, "failed to get group selectorL %s", gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  746  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  747  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  748  	gsel = ret;
99b2f99aa41aa7 Drew Fustini  2021-02-09  749  
99b2f99aa41aa7 Drew Fustini  2021-02-09  750  	ret = pmxops->set_mux(pctldev, fsel, gsel);
99b2f99aa41aa7 Drew Fustini  2021-02-09  751  	if (ret) {
99b2f99aa41aa7 Drew Fustini  2021-02-09  752  		dev_err(pctldev->dev, "set_mux() failed: %d", ret);
99b2f99aa41aa7 Drew Fustini  2021-02-09  753  		goto free_gname;
99b2f99aa41aa7 Drew Fustini  2021-02-09  754  	}
99b2f99aa41aa7 Drew Fustini  2021-02-09  755  
99b2f99aa41aa7 Drew Fustini  2021-02-09  756  	return len;
99b2f99aa41aa7 Drew Fustini  2021-02-09  757  free_buf:
99b2f99aa41aa7 Drew Fustini  2021-02-09  758  	devm_kfree(pctldev->dev, buf);
99b2f99aa41aa7 Drew Fustini  2021-02-09  759  free_fname:
99b2f99aa41aa7 Drew Fustini  2021-02-09  760  	devm_kfree(pctldev->dev, fname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  761  free_gname:
99b2f99aa41aa7 Drew Fustini  2021-02-09 @762  	devm_kfree(pctldev->dev, gname);
99b2f99aa41aa7 Drew Fustini  2021-02-09  763  	return ret;
99b2f99aa41aa7 Drew Fustini  2021-02-09  764  }
Quoted 3 lines

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/[email protected]

Thank you,
Drew

#19 lore raw
Quoted 8 lines · Drew Fustini
Quoted 7 lines · Dan Carpenter

Hi Drew,

url:    https://github.com/0day-ci/linux/commits/Drew-Fustini/pinctrl-pinmux-Add-pinmux-select-debugfs-file/20210210-160108 
base:   https://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl.git  devel
config: i386-randconfig-m021-20210209 (attached as .config)
compiler: gcc-9 (Debian 9.3.0-15) 9.3.0
Quoted 6 lines
Quoted 3 lines

If you fix the issue, kindly add following tag as appropriate

Does it makes sense for me to include that tag in this patch?

This stuff is just in the autogenerated portions of the kbuild bot email. I normally just hit forward, without all the extra pontifying that I included this time. :P

regards,
dan carpenter