Skip to contentSkip to conversation

drivers: base: Warn if driver name is not present

PATCH In review [email protected] 3 participants 5 messages Jun 8, 2020 Driver core

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 drivers: base: Warn if driver name is not present

From: Matthias Brugger <[email protected]>

If we pass a driver without a name, we end up in a NULL pointer derefernce. Check for the name before trying to register the driver. As we don't have a driver name to point to in the error message, we dump the call stack to make it easier to detect the buggy driver.

---
 drivers/base/driver.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/base/driver.c +6 -0
index 57c68769e157..40fba959c140 100644
@@ -149,6 +149,12 @@ int driver_register(struct device_driver *drv)
 	int ret;
 	struct device_driver *other;
 
+	if (!drv->name) {
+		pr_err("Driver has no name.\n");
+		dump_stack();
+		return -EINVAL;
+	}
+
 	if (!drv->bus->p) {
 		pr_err("Driver '%s' was unable to register with bus_type '%s' because the bus was not initialized.\n",
 			   drv->name, drv->bus->name);
Signature · 1 line
2.26.2
#1 lore raw

PATCH 2/2 drivers: base: Convert to printk alias functions

From: Matthias Brugger <[email protected]>

The file mixes printk calls together with calls to pr_*(). Covert to printk alias functions to unify the code.

---
 drivers/base/driver.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/base/driver.c +2 -2
index 40fba959c140..aa3b3a226a02 100644
@@ -164,12 +164,12 @@ int driver_register(struct device_driver *drv)
 	if ((drv->bus->probe && drv->probe) ||
 	    (drv->bus->remove && drv->remove) ||
 	    (drv->bus->shutdown && drv->shutdown))
-		printk(KERN_WARNING "Driver '%s' needs updating - please use "
+		pr_warn("Driver '%s' needs updating - please use "
 			"bus_type methods\n", drv->name);
 
 	other = driver_find(drv->name, drv->bus);
 	if (other) {
-		printk(KERN_ERR "Error: Driver '%s' is already registered, "
+		pr_err("Error: Driver '%s' is already registered, "
 			"aborting...\n", drv->name);
 		return -EBUSY;
 	}
Signature · 1 line
2.26.2
maintainer
#2 lore raw
In reply to [email protected]
Quoted 4 lines · On Mon, Jun 08, 2020 at 11:52:16AM +0200, [email protected] wrote:

From: Matthias Brugger <[email protected]>

If we pass a driver without a name, we end up in a NULL pointer derefernce.

That's a very good reason not to have a driver without a name :)

What in-kernel driver does this?

Quoted 15 lines

Check for the name before trying to register the driver.
As we don't have a driver name to point to in the error message, we dump the call stack to make it easier to detect the buggy driver.

---
 drivers/base/driver.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/base/driver.c +0 -0
index 57c68769e157..40fba959c140 100644

drivers/base/driver.c int driver_register(struct device_driver *drv) · line 155

Quoted 7 lines
drivers/base/driver.c +4 -0
@@ -149,6 +149,12 @@ int driver_register(struct device_driver *drv)
 	int ret;
 	struct device_driver *other;
 
+	if (!drv->name) {
+		pr_err("Driver has no name.\n");
+		dump_stack();
+		return -EINVAL;

Ick, no, an oops-traceback for doing something dumb like this should be all that we need, right?

How "hardened" do we need to make internal apis anyway? What's the odds that if this does trigger, the driver author would even notice it?

thanks,

greg k-h

#3 lore raw
Quoted 36 lines · Greg KH
Quoted 4 lines · On Mon, Jun 08, 2020 at 11:52:16AM +0200, [email protected] wrote:

From: Matthias Brugger <[email protected]>

If we pass a driver without a name, we end up in a NULL pointer derefernce.

That's a very good reason not to have a driver without a name :)

What in-kernel driver does this?

Quoted 15 lines

Check for the name before trying to register the driver.
As we don't have a driver name to point to in the error message, we dump the call stack to make it easier to detect the buggy driver.

---
 drivers/base/driver.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/base/driver.c +0 -0
index 57c68769e157..40fba959c140 100644

drivers/base/driver.c int driver_register(struct device_driver *drv) · line 155

Quoted 7 lines
drivers/base/driver.c +4 -0
@@ -149,6 +149,12 @@ int driver_register(struct device_driver *drv)
 	int ret;
 	struct device_driver *other;
 
+	if (!drv->name) {
+		pr_err("Driver has no name.\n");
+		dump_stack();
+		return -EINVAL;

Ick, no, an oops-traceback for doing something dumb like this should be all that we need, right?

Quoted 3 lines

How "hardened" do we need to make internal apis anyway? What's the odds that if this does trigger, the driver author would even notice it?

We just had the case that a driver got accepted in a maintainer repository without a name. Which got later found by the kernel test robot.

I agree with you that it probably doesn't make much sense to check for this kind of bugs, as it should be discoverable if you test your code, before you submit.

I propose to ignore this patch.

Regards,
Matthias

maintainer
#4 lore raw
Quoted 39 lines · Matthias Brugger
Quoted 36 lines · Greg KH
Quoted 4 lines · On Mon, Jun 08, 2020 at 11:52:16AM +0200, [email protected] wrote:

From: Matthias Brugger <[email protected]>

If we pass a driver without a name, we end up in a NULL pointer derefernce.

That's a very good reason not to have a driver without a name :)

What in-kernel driver does this?

Quoted 15 lines

Check for the name before trying to register the driver.
As we don't have a driver name to point to in the error message, we dump the call stack to make it easier to detect the buggy driver.

---
 drivers/base/driver.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/base/driver.c +0 -0
index 57c68769e157..40fba959c140 100644

drivers/base/driver.c int driver_register(struct device_driver *drv) · line 155

Quoted 7 lines
drivers/base/driver.c +4 -0
@@ -149,6 +149,12 @@ int driver_register(struct device_driver *drv)
 	int ret;
 	struct device_driver *other;
 
+	if (!drv->name) {
+		pr_err("Driver has no name.\n");
+		dump_stack();
+		return -EINVAL;

Ick, no, an oops-traceback for doing something dumb like this should be all that we need, right?

Quoted 6 lines
Quoted 3 lines

How "hardened" do we need to make internal apis anyway? What's the odds that if this does trigger, the driver author would even notice it?

We just had the case that a driver got accepted in a maintainer repository without a name. Which got later found by the kernel test robot.

That driver had obviously never actually been run before :(

Quoted 4 lines

I agree with you that it probably doesn't make much sense to check for this kind of bugs, as it should be discoverable if you test your code, before you submit.

I propose to ignore this patch.

Thanks, now dropped!

greg k-h