Skip to contentSkip to conversation

HID: rmi: Check that the RMI_STARTED bit is set before unregistering the RMI transport device

PATCH In review Andrew Duggan 3 participants 7 messages Oct 23, 2019 – Nov 18, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 HID: rmi: Check that the RMI_STARTED bit is set before unregistering the RMI transport device

In the event that the RMI device is unreachable, the calls to rmi_set_mode() or rmi_set_page() will fail before registering the RMI transport device. When the device is removed, rmi_remove() will call rmi_unregister_transport_device() which will attempt to access the rmi_dev pointer which was not set. This patch adds a check of the RMI_STARTED bit before calling rmi_unregister_transport_device(). The RMI_STARTED bit is only set after rmi_register_transport_device() completes successfully. A subsequent patch in the RMI core will add checks to validate the pointers before accessing them.

The kernel oops was reported in this message:
https://www.spinics.net/lists/linux-input/msg58433.html

---
 drivers/hid/hid-rmi.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/hid/hid-rmi.c +2 -1
index 7c6abd7e0979..9ce22acdfaca 100644
@@ -744,7 +744,8 @@ static void rmi_remove(struct hid_device *hdev)
 {
 	struct rmi_data *hdata = hid_get_drvdata(hdev);
 
-	if (hdata->device_flags & RMI_DEVICE) {
+	if ((hdata->device_flags & RMI_DEVICE)
+	    && test_bit(RMI_STARTED, &hdata->flags)) {
 		clear_bit(RMI_STARTED, &hdata->flags);
 		cancel_work_sync(&hdata->reset_work);
 		rmi_unregister_transport_device(&hdata->xport);
Signature · 1 line
2.20.1
#1 lore raw

PATCH 1/1 Input: synaptics-rmi4 - validate that the rmi_dev pointer is set before dereferencing it

A bug in hid-rmi was causing rmi_unregister_transport_device() to be called even if the call to rmi_register_transport_device() failed to allocate the rmi device. A patch has been submitted to fix the issue in hid-rmi. This patch will ensure that should a simialr situation occur then the rmi driver will not dereference a NULL pointer.

---
 drivers/input/rmi4/rmi_bus.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
drivers/input/rmi4/rmi_bus.c +4 -2
index af706a583656..6c3abae1e159 100644
@@ -118,8 +118,10 @@ void rmi_unregister_transport_device(struct rmi_transport_dev *xport)
 {
 	struct rmi_device *rmi_dev = xport->rmi_dev;
 
-	device_del(&rmi_dev->dev);
-	put_device(&rmi_dev->dev);
+	if (rmi_dev) {
+		device_del(&rmi_dev->dev);
+		put_device(&rmi_dev->dev);
+	}
 }
 EXPORT_SYMBOL(rmi_unregister_transport_device);
 
Signature · 1 line
2.20.1
maintainer
#2 lore raw

Re: [PATCH] Input: synaptics-rmi4 - validate that the rmi_dev pointer is set before dereferencing it

Hi Andrew,

Quoted 5 lines · Andrew Duggan

A bug in hid-rmi was causing rmi_unregister_transport_device() to be called even if the call to rmi_register_transport_device() failed to allocate the rmi device. A patch has been submitted to fix the issue in hid-rmi. This patch will ensure that should a simialr situation occur then the rmi driver will not dereference a NULL pointer.

This looks like "garbage in, garbage out" problem where we should not be calling unregister in the first place. I'd rather not apply this.

Thanks.

Signature · 1 line
Dmitry
#3 lore raw

Re: [PATCH] Input: synaptics-rmi4 - validate that the rmi_dev pointer is set before dereferencing it

Hi Dmitry,

Quoted 5 lines · Dmitry Torokhov

CAUTION: Email originated externally, do not click links or open attachments unless you recognize the sender and know the content is safe.

Hi Andrew,

Quoted 8 lines
Quoted 5 lines · Andrew Duggan

A bug in hid-rmi was causing rmi_unregister_transport_device() to be called even if the call to rmi_register_transport_device() failed to allocate the rmi device. A patch has been submitted to fix the issue in hid-rmi. This patch will ensure that should a simialr situation occur then the rmi driver will not dereference a NULL pointer.

This looks like "garbage in, garbage out" problem where we should not be calling unregister in the first place. I'd rather not apply this.

That's fine, like I said the actual fix to prevent rmi_unregister_transport_device() from being called inappropriately is in the hid-rmi driver.

https://lore.kernel.org/linux-input/[email protected]/

I was on the fence on whether or not it was better to prevent the NULL pointer dereference even at the expense of masking bugs like the one in the hid-rmi driver.

Thanks for the feedback,

Andrew

Quoted 4 lines

Thanks.

--
Dmitry

maintainer
#4 lore raw
In reply to Andrew Duggan
Quoted 9 lines · Andrew Duggan

In the event that the RMI device is unreachable, the calls to rmi_set_mode() or rmi_set_page() will fail before registering the RMI transport device. When the device is removed, rmi_remove() will call rmi_unregister_transport_device() which will attempt to access the rmi_dev pointer which was not set. This patch adds a check of the RMI_STARTED bit before calling rmi_unregister_transport_device(). The RMI_STARTED bit is only set after rmi_register_transport_device() completes successfully. A subsequent patch in the RMI core will add checks to validate the pointers before accessing them.

Andrew,

my mailbox doesn't seem to have that 'subsequent patch' ... was it ever sent and I missed it? If so, could you please resend it?

Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs
#5 lore raw

Hi Jiri,

Quoted 11 lines · Jiri Kosina
Quoted 9 lines · Andrew Duggan

In the event that the RMI device is unreachable, the calls to rmi_set_mode() or rmi_set_page() will fail before registering the RMI transport device. When the device is removed, rmi_remove() will call rmi_unregister_transport_device() which will attempt to access the rmi_dev pointer which was not set. This patch adds a check of the RMI_STARTED bit before calling rmi_unregister_transport_device(). The RMI_STARTED bit is only set after rmi_register_transport_device() completes successfully. A subsequent patch in the RMI core will add checks to validate the pointers before accessing them.

Quoted 4 lines

Andrew,

my mailbox doesn't seem to have that 'subsequent patch' ... was it ever sent and I missed it? If so, could you please resend it?

The subsequent patch which I was referring to is:

https://lore.kernel.org/linux-input/[email protected]/

Since this second patch was for the input subsystem I decided to just make them separate patches instead of creating a series. However, based on Dmitry's feedback it was determined that the second patch wasn't a good idea and it won't be applied. This first patch is enough to fix the issue by preventing the call to rmi_unregister_transport_device() if the subsequent call to register failed. The only change I would make to this patch would be to remove the last sentence of the comment. If you choose to apply that patch then would this be a change you would make? Or would you prefer I submit a v2 with this update?

Thanks,

Andrew

Quoted 6 lines

Thanks,

--
Jiri Kosina
SUSE Labs

maintainer
#6 lore raw
Quoted 9 lines · Andrew Duggan

Since this second patch was for the input subsystem I decided to just make them separate patches instead of creating a series. However, based on Dmitry's feedback it was determined that the second patch wasn't a good idea and it won't be applied. This first patch is enough to fix the issue by preventing the call to rmi_unregister_transport_device() if the subsequent call to register failed. The only change I would make to this patch would be to remove the last sentence of the comment. If you choose to apply that patch then would this be a change you would make? Or would you prefer I submit a v2 with this update?

I've modified the changelog and applied. Thanks,

Signature · 2 lines
Jiri Kosina
SUSE Labs