[PATCH] misc: uacce - add the null check for the input pointer and its pointer members

[PATCH] misc: uacce - add the null check for the input pointer and its pointer members

[Chenghai Huang]

The uacce_alloc() is the member of the EXPORT_SYMBOL_GPL. Therefore, null
pointer verification is added on the pointer type input parameter and its
pointer members.

Signed-off-by: Chenghai Huang <huangchenghai2@huawei.com>
---
 drivers/misc/uacce/uacce.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/misc/uacce/uacce.c b/drivers/misc/uacce/uacce.c
index bdc2e6fda782..964f1a6a16e0 100644
--- a/drivers/misc/uacce/uacce.c
+++ b/drivers/misc/uacce/uacce.c
@@ -514,6 +514,9 @@ struct uacce_device *uacce_alloc(struct device *parent,
 	struct uacce_device *uacce;
 	int ret;
 
+	if (!parent || !interface || !interface->ops)
+		return ERR_PTR(-EINVAL);
+
 	uacce = kzalloc(sizeof(struct uacce_device), GFP_KERNEL);
 	if (!uacce)
 		return ERR_PTR(-ENOMEM);
-- 
2.30.0

Re: [PATCH] misc: uacce - add the null check for the input pointer and its pointer members

[Greg KH]

On Fri, Mar 29, 2024 at 02:26:55PM +0800, Chenghai Huang wrote:
> The uacce_alloc() is the member of the EXPORT_SYMBOL_GPL. Therefore, null
> pointer verification is added on the pointer type input parameter and its
> pointer members.

I do not understand, why does the export type matter?  Just fix any
callers to use this properly and send proper parameters.  What in-tree
caller needs this?

thanks,

greg k-h

Re: [PATCH] misc: uacce - add the null check for the input pointer and its pointer members

[Greg KH]

On Sat, Mar 30, 2024 at 11:34:24AM +0800, huangchenghai wrote:
> 
> On Fri, Mar 29, 2024 at2:34PM, Greg KH wrote:
> 
> > On Fri, Mar 29, 2024 at 02:26:55PM +0800, Chenghai Huang wrote:
> > > The uacce_alloc() is the member of the EXPORT_SYMBOL_GPL. Therefore, null
> > > pointer verification is added on the pointer type input parameter and its
> > > pointer members.
> > I do not understand, why does the export type matter?  Just fix any
> > callers to use this properly and send proper parameters.  What in-tree
> > caller needs this?
> > 
> > thanks,
> > 
> > greg k-h
> 
> The interface defined by the export type seems important and the input
> parameters need to be verified.

The export type does not matter at all.

> But I understand from your mail that this is the job of the caller.

Exactly.

> By the way, I still have a confusion. Interfaces like ioctrl, debugfs, read,
> or write require parameter validation. so what kind of kernel interfaces require
> parameter validation? Is there a definition?

Some do, some do not, it depends on the situation.  If data comes from
an untrusted source (i.e. outside the kernel), then it MUST be validated
(remember "all input is evil"), but if it's from within the kernel,
usually it does not.

thanks,

greg k-h

Re: [PATCH] misc: uacce - add the null check for the input pointer and its pointer members

[huangchenghai]

在 2024/3/30 15:01, Greg KH 写道:
> On Sat, Mar 30, 2024 at 11:34:24AM +0800, huangchenghai wrote:
>> On Fri, Mar 29, 2024 at2:34PM, Greg KH wrote:
>>
>>> On Fri, Mar 29, 2024 at 02:26:55PM +0800, Chenghai Huang wrote:
>>>> The uacce_alloc() is the member of the EXPORT_SYMBOL_GPL. Therefore, null
>>>> pointer verification is added on the pointer type input parameter and its
>>>> pointer members.
>>> I do not understand, why does the export type matter?  Just fix any
>>> callers to use this properly and send proper parameters.  What in-tree
>>> caller needs this?
>>>
>>> thanks,
>>>
>>> greg k-h
>> The interface defined by the export type seems important and the input
>> parameters need to be verified.
> The export type does not matter at all.
>
>> But I understand from your mail that this is the job of the caller.
> Exactly.
>
>> By the way, I still have a confusion. Interfaces like ioctrl, debugfs, read,
>> or write require parameter validation. so what kind of kernel interfaces require
>> parameter validation? Is there a definition?
> Some do, some do not, it depends on the situation.  If data comes from
> an untrusted source (i.e. outside the kernel), then it MUST be validated
> (remember "all input is evil"), but if it's from within the kernel,
> usually it does not.
>
> thanks,
>
> greg k-h

I get it, thank you very much.

thanks,

Chenghai