[PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Luis de Bethencourt]

The driver is using -1 instead of the -ENOMEM defined macro to specify
that a buffer allocation failed. Since the error number is propagated,
the caller will get a -EPERM which is the wrong error condition.

Also, the smatch tool complains with the following warning:
hiddev_connect() warn: returning -1 instead of -ENOMEM is sloppy

Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
---
 drivers/hid/usbhid/hiddev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/hid/usbhid/hiddev.c b/drivers/hid/usbhid/hiddev.c
index 2f1ddca..c5290ff 100644
--- a/drivers/hid/usbhid/hiddev.c
+++ b/drivers/hid/usbhid/hiddev.c
@@ -894,7 +894,7 @@ int hiddev_connect(struct hid_device *hid, unsigned int force)
 	}
 
 	if (!(hiddev = kzalloc(sizeof(struct hiddev), GFP_KERNEL)))
-		return -1;
+		return -ENOMEM;
 
 	init_waitqueue_head(&hiddev->wait);
 	INIT_LIST_HEAD(&hiddev->list);
-- 
2.5.1

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Luis de Bethencourt]

On 30/09/15 10:52, Luis de Bethencourt wrote:
> The driver is using -1 instead of the -ENOMEM defined macro to specify
> that a buffer allocation failed. Since the error number is propagated,
> the caller will get a -EPERM which is the wrong error condition.
> 
> Also, the smatch tool complains with the following warning:
> hiddev_connect() warn: returning -1 instead of -ENOMEM is sloppy
> 
> Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
> ---
>  drivers/hid/usbhid/hiddev.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/hid/usbhid/hiddev.c b/drivers/hid/usbhid/hiddev.c
> index 2f1ddca..c5290ff 100644
> --- a/drivers/hid/usbhid/hiddev.c
> +++ b/drivers/hid/usbhid/hiddev.c
> @@ -894,7 +894,7 @@ int hiddev_connect(struct hid_device *hid, unsigned int force)
>  	}
>  
>  	if (!(hiddev = kzalloc(sizeof(struct hiddev), GFP_KERNEL)))
> -		return -1;
> +		return -ENOMEM;
>  
>  	init_waitqueue_head(&hiddev->wait);
>  	INIT_LIST_HEAD(&hiddev->list);
> 

Hello,

I got an "Undelivered Mail Returned to Sender" from Jiri Kosina's
jikos@kernel.com email address. This email is listed multiple times
in the MAINTAINERS file, does he have a new address to update this
file?

Thanks,
Luis

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Sudip Mukherjee]

On Wed, Sep 30, 2015 at 10:56:26AM +0100, Luis de Bethencourt wrote:
> On 30/09/15 10:52, Luis de Bethencourt wrote:
> > The driver is using -1 instead of the -ENOMEM defined macro to specify
> > that a buffer allocation failed. Since the error number is propagated,
> > the caller will get a -EPERM which is the wrong error condition.
> > 
> > Also, the smatch tool complains with the following warning:
> > hiddev_connect() warn: returning -1 instead of -ENOMEM is sloppy
> > 
> > Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
> > ---
> >  drivers/hid/usbhid/hiddev.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/drivers/hid/usbhid/hiddev.c b/drivers/hid/usbhid/hiddev.c
> > index 2f1ddca..c5290ff 100644
> > --- a/drivers/hid/usbhid/hiddev.c
> > +++ b/drivers/hid/usbhid/hiddev.c
> > @@ -894,7 +894,7 @@ int hiddev_connect(struct hid_device *hid, unsigned int force)
> >  	}
> >  
> >  	if (!(hiddev = kzalloc(sizeof(struct hiddev), GFP_KERNEL)))
> > -		return -1;
> > +		return -ENOMEM;
> >  
> >  	init_waitqueue_head(&hiddev->wait);
> >  	INIT_LIST_HEAD(&hiddev->list);
> > 
> 
> Hello,
> 
> I got an "Undelivered Mail Returned to Sender" from Jiri Kosina's
> jikos@kernel.com email address. This email is listed multiple times
> in the MAINTAINERS file, does he have a new address to update this
> file?
Its jikos@kernel.org

regards
sudip

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Luis de Bethencourt]

On 30/09/15 11:04, Sudip Mukherjee wrote:
> On Wed, Sep 30, 2015 at 10:56:26AM +0100, Luis de Bethencourt wrote:
>> On 30/09/15 10:52, Luis de Bethencourt wrote:
>>> The driver is using -1 instead of the -ENOMEM defined macro to specify
>>> that a buffer allocation failed. Since the error number is propagated,
>>> the caller will get a -EPERM which is the wrong error condition.
>>>
>>> Also, the smatch tool complains with the following warning:
>>> hiddev_connect() warn: returning -1 instead of -ENOMEM is sloppy
>>>
>>> Signed-off-by: Luis de Bethencourt <luisbg@osg.samsung.com>
>>> ---
>>>  drivers/hid/usbhid/hiddev.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/hid/usbhid/hiddev.c b/drivers/hid/usbhid/hiddev.c
>>> index 2f1ddca..c5290ff 100644
>>> --- a/drivers/hid/usbhid/hiddev.c
>>> +++ b/drivers/hid/usbhid/hiddev.c
>>> @@ -894,7 +894,7 @@ int hiddev_connect(struct hid_device *hid, unsigned int force)
>>>  	}
>>>  
>>>  	if (!(hiddev = kzalloc(sizeof(struct hiddev), GFP_KERNEL)))
>>> -		return -1;
>>> +		return -ENOMEM;
>>>  
>>>  	init_waitqueue_head(&hiddev->wait);
>>>  	INIT_LIST_HEAD(&hiddev->list);
>>>
>>
>> Hello,
>>
>> I got an "Undelivered Mail Returned to Sender" from Jiri Kosina's
>> jikos@kernel.com email address. This email is listed multiple times
>> in the MAINTAINERS file, does he have a new address to update this
>> file?
> Its jikos@kernel.org
> 
> regards
> sudip
> 

Hi Sudip,

*facepalm* My mistake, yes.

Sorry,
Luis

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Jiri Kosina]

On Wed, 30 Sep 2015, Luis de Bethencourt wrote:

> The driver is using -1 instead of the -ENOMEM defined macro to specify
> that a buffer allocation failed. Since the error number is propagated,
> the caller will get a -EPERM which is the wrong error condition.

Generally I agree that the more specific errno, the better.

But I am not really sure where you are seeing the bug (mapping to -EPERM) 
in this case? I think the only caller of hiddev_connect() should be 
hid_connect(), and the only thing that guy cares about whether individual 
callbacks succeed or fail, so that it sets hdev->clamed flags accordingly.

Could you please be more specific about the -EPERM mapping you are talking 
about?

Thanks,

-- 
Jiri Kosina
SUSE Labs

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Luis de Bethencourt]

On 30/09/15 20:40, Jiri Kosina wrote:
> On Wed, 30 Sep 2015, Luis de Bethencourt wrote:
> 
>> The driver is using -1 instead of the -ENOMEM defined macro to specify
>> that a buffer allocation failed. Since the error number is propagated,
>> the caller will get a -EPERM which is the wrong error condition.
> 
> Generally I agree that the more specific errno, the better.
> 
> But I am not really sure where you are seeing the bug (mapping to -EPERM) 
> in this case? I think the only caller of hiddev_connect() should be 
> hid_connect(), and the only thing that guy cares about whether individual 
> callbacks succeed or fail, so that it sets hdev->clamed flags accordingly.
> 
> Could you please be more specific about the -EPERM mapping you are talking 
> about?
> 
> Thanks,
> 

I agree with you. The only caller of hiddev_connect() only checks if the
callback succeded. It checks if the return < 0.
What I meant is that -1 means -EPERM. [0]

This patch is purely about the correctness of using -ENOMEM. The word
"propagated" was not the best way to describe this problem. I could edit
the commit message if you would like.

Thanks for the review,
Luis

[0] http://lxr.free-electrons.com/source/include/uapi/asm-generic/errno-base.h#L15

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Jiri Kosina]

On Sat, 3 Oct 2015, Luis de Bethencourt wrote:

> > But I am not really sure where you are seeing the bug (mapping to 
> > -EPERM) in this case? I think the only caller of hiddev_connect() 
> > should be hid_connect(), and the only thing that guy cares about 
> > whether individual callbacks succeed or fail, so that it sets 
> > hdev->clamed flags accordingly.
> > 
> > Could you please be more specific about the -EPERM mapping you are 
> > talking about?
> > 
> I agree with you. The only caller of hiddev_connect() only checks if the
> callback succeded. It checks if the return < 0.
> What I meant is that -1 means -EPERM. [0]

I still don't understand what problem you are chasing here, sorry. EPERM 
is defined to be 1, yes. So are many other completely unrelated #defines.

> This patch is purely about the correctness of using -ENOMEM. The word 
> "propagated" was not the best way to describe this problem. I could edit 
> the commit message if you would like.

You seem to imply that someone might be interpreting that -1 as a define 
from errno.h. But that's not the case.

Are you going to look at every 'return -1' occurence in the kernel and 
convert it to something else? That can keep you busy for quite some time:

	$ git grep 'return -1' | wc -l
	9167

The only cleanup I'd imagine at least remotely possible in this case would 
be to convert the ->connect() callbacks return bool.

-- 
Jiri Kosina
SUSE Labs

Re: [PATCH] HID: hiddev: fix returned errno code in hiddev_connect()

[Luis de Bethencourt]

On 05/10/15 15:24, Jiri Kosina wrote:
> On Sat, 3 Oct 2015, Luis de Bethencourt wrote:
> 
>>> But I am not really sure where you are seeing the bug (mapping to 
>>> -EPERM) in this case? I think the only caller of hiddev_connect() 
>>> should be hid_connect(), and the only thing that guy cares about 
>>> whether individual callbacks succeed or fail, so that it sets 
>>> hdev->clamed flags accordingly.
>>>
>>> Could you please be more specific about the -EPERM mapping you are 
>>> talking about?
>>>
>> I agree with you. The only caller of hiddev_connect() only checks if the
>> callback succeded. It checks if the return < 0.
>> What I meant is that -1 means -EPERM. [0]
> 
> I still don't understand what problem you are chasing here, sorry. EPERM 
> is defined to be 1, yes. So are many other completely unrelated #defines.
> 
>> This patch is purely about the correctness of using -ENOMEM. The word 
>> "propagated" was not the best way to describe this problem. I could edit 
>> the commit message if you would like.
> 
> You seem to imply that someone might be interpreting that -1 as a define 
> from errno.h. But that's not the case.
> 
> Are you going to look at every 'return -1' occurence in the kernel and 
> convert it to something else? That can keep you busy for quite some time:
> 
> 	$ git grep 'return -1' | wc -l
> 	9167
> 
> The only cleanup I'd imagine at least remotely possible in this case would 
> be to convert the ->connect() callbacks return bool.
> 

This is a very good point. I will write a second version of the patch that
converts the return to a boolean.

Thanks for the review,
Luis