2015-08-14 18:43:59

by Alexander Kuleshov

Subject: [PATCH] drivers/hid: Fix debugfs_create_dir's error checking method for hid-core

The debugfs_create_dir() function returns NULL if an error occurs. This patch
adds check of the result of the debufs_create_dir() execution in the
hid_debug_register().

Signed-off-by: Alexander Kuleshov <[email protected]>
---
drivers/hid/hid-core.c | 6 ++++--
drivers/hid/hid-debug.c | 3 +++
2 files changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/hid/hid-core.c b/drivers/hid/hid-core.c
index e6fce23..315a980 100644
--- a/drivers/hid/hid-core.c
+++ b/drivers/hid/hid-core.c
@@ -2599,8 +2599,10 @@ int hid_add_device(struct hid_device *hdev)
ret = device_add(&hdev->dev);
if (!ret)
hdev->status |= HID_STAT_ADDED;
- else
- hid_debug_unregister(hdev);
+ else {
+ if (hdev->debug_dir)
+ hid_debug_unregister(hdev);
+ }

return ret;
}
diff --git a/drivers/hid/hid-debug.c b/drivers/hid/hid-debug.c
index 2886b64..80fa03d 100644
--- a/drivers/hid/hid-debug.c
+++ b/drivers/hid/hid-debug.c
@@ -1223,6 +1223,9 @@ static const struct file_operations hid_debug_events_fops = {
void hid_debug_register(struct hid_device *hdev, const char *name)
{
hdev->debug_dir = debugfs_create_dir(name, hid_debug_root);
+ if (!hdev->debug_dir)
+ return -ENOMEM;
+
hdev->debug_rdesc = debugfs_create_file("rdesc", 0400,
hdev->debug_dir, hdev, &hid_debug_rdesc_fops);
hdev->debug_events = debugfs_create_file("events", 0400,
--
2.5.0

2015-08-18 13:12:38

by Jiri Kosina

Subject: Re: [PATCH] drivers/hid: Fix debugfs_create_dir's error checking method for hid-core

On Sat, 15 Aug 2015, Alexander Kuleshov wrote:

> The debugfs_create_dir() function returns NULL if an error occurs. This patch
> adds check of the result of the debufs_create_dir() execution in the
> hid_debug_register().
>
> Signed-off-by: Alexander Kuleshov <[email protected]>
> ---
> drivers/hid/hid-core.c | 6 ++++--
> drivers/hid/hid-debug.c | 3 +++
> 2 files changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/hid/hid-core.c b/drivers/hid/hid-core.c
> index e6fce23..315a980 100644
> --- a/drivers/hid/hid-core.c
> +++ b/drivers/hid/hid-core.c
> @@ -2599,8 +2599,10 @@ int hid_add_device(struct hid_device *hdev)
> ret = device_add(&hdev->dev);
> if (!ret)
> hdev->status |= HID_STAT_ADDED;
> - else
> - hid_debug_unregister(hdev);
> + else {
> + if (hdev->debug_dir)
> + hid_debug_unregister(hdev);
> + }
>
> return ret;
> }
> diff --git a/drivers/hid/hid-debug.c b/drivers/hid/hid-debug.c
> index 2886b64..80fa03d 100644
> --- a/drivers/hid/hid-debug.c
> +++ b/drivers/hid/hid-debug.c
> @@ -1223,6 +1223,9 @@ static const struct file_operations hid_debug_events_fops = {
> void hid_debug_register(struct hid_device *hdev, const char *name)
> {
> hdev->debug_dir = debugfs_create_dir(name, hid_debug_root);
> + if (!hdev->debug_dir)
> + return -ENOMEM;
> +

What sense does it make to return -ENOMEM from void function?

> hdev->debug_rdesc = debugfs_create_file("rdesc", 0400,
> hdev->debug_dir, hdev, &hid_debug_rdesc_fops);
> hdev->debug_events = debugfs_create_file("events", 0400,

Fixing this is ultimately a good thing, but why did you just pick
debugfs_create_dir() and left the other, with the same issue
(debugfs_create_file()) intact?

--
Jiri Kosina
SUSE Labs