Re: [PATCH] vga_switcheroo: Fix impossible judgment condition

2023-10-26 Thread Dan Carpenter
On Thu, Oct 26, 2023 at 04:46:29PM +0800, Su Hui wrote:
> On 2023/10/26 12:44, Dan Carpenter wrote:
> > On Thu, Oct 26, 2023 at 10:10:57AM +0800, Su Hui wrote:
> > > 'id' is enum type like unsigned int, so it will never be less than zero.
> > > 
> > > Fixes: 4aaf448fa975 ("vga_switcheroo: set audio client id according to 
> > > bound GPU id")
> > > Signed-off-by: Su Hui 
> > > ---
> > >   drivers/gpu/vga/vga_switcheroo.c | 2 +-
> > >   1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/gpu/vga/vga_switcheroo.c 
> > > b/drivers/gpu/vga/vga_switcheroo.c
> > > index 365e6ddbe90f..d3064466fd3a 100644
> > > --- a/drivers/gpu/vga/vga_switcheroo.c
> > > +++ b/drivers/gpu/vga/vga_switcheroo.c
> > > @@ -375,7 +375,7 @@ int vga_switcheroo_register_audio_client(struct 
> > > pci_dev *pdev,
> > >   mutex_lock(_mutex);
> > >   if (vgasr_priv.active) {
> > >   id = vgasr_priv.handler->get_client_id(vga_dev);
> > > - if (id < 0) {
> > > + if ((int)id < 0) {
> > Hi,
> > 
> > I feel like you're using Smatch?  Which is great!  Fantastic!
> Yep, Smatch helps me  a lot to find these bugs! I really like this excellent
> tool!
> > 
> > Have you built the cross function database?  If you have there is a
> > command that's useful.
> Not yet, bu I want to build this.

Yeah.  It's super useful for kernel development.  It helps to understand
how functions are called and where variables are set etc.  The smatch
documentation is crap, I know.  But I did write a short blog about the
cross function DB.

https://staticthinking.wordpress.com/2023/05/02/the-cross-function-db/

It's simple to build, but it takes a long time.  Just run
smatch_scripts/build_kernel_data.sh

regards,
dan carpenter



Re: [PATCH] vga_switcheroo: Fix impossible judgment condition

2023-10-26 Thread Su Hui

On 2023/10/26 12:44, Dan Carpenter wrote:

On Thu, Oct 26, 2023 at 10:10:57AM +0800, Su Hui wrote:

'id' is enum type like unsigned int, so it will never be less than zero.

Fixes: 4aaf448fa975 ("vga_switcheroo: set audio client id according to bound GPU 
id")
Signed-off-by: Su Hui 
---
  drivers/gpu/vga/vga_switcheroo.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/vga/vga_switcheroo.c b/drivers/gpu/vga/vga_switcheroo.c
index 365e6ddbe90f..d3064466fd3a 100644
--- a/drivers/gpu/vga/vga_switcheroo.c
+++ b/drivers/gpu/vga/vga_switcheroo.c
@@ -375,7 +375,7 @@ int vga_switcheroo_register_audio_client(struct pci_dev 
*pdev,
mutex_lock(_mutex);
if (vgasr_priv.active) {
id = vgasr_priv.handler->get_client_id(vga_dev);
-   if (id < 0) {
+   if ((int)id < 0) {

Hi,

I feel like you're using Smatch?  Which is great!  Fantastic!
Yep, Smatch helps me  a lot to find these bugs! I really like this 
excellent tool!


Have you built the cross function database?  If you have there is a
command that's useful.

Not yet, bu I want to build this.

$ ~/smatch/smatch_db/smdb.py functions vga_switcheroo_handler get_client_id | 
tee where
drivers/gpu/drm/nouveau/nouveau_acpi.c | (struct 
vga_switcheroo_handler)->get_client_id | nouveau_dsm_get_client_id | 1
drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c | (struct 
vga_switcheroo_handler)->get_client_id | amdgpu_atpx_get_client_id | 1
drivers/gpu/drm/radeon/radeon_atpx_handler.c | (struct 
vga_switcheroo_handler)->get_client_id | radeon_atpx_get_client_id | 1
drivers/platform/x86/apple-gmux.c | (struct 
vga_switcheroo_handler)->get_client_id | gmux_get_client_id | 1
$ make cscope
$ vim where
Use cscope to jump to each of those four functions.  Move the cursor to
the nouveau_dsm_get_client_id and hit CTRL-].

Sounds great! I must try this!


They never return negatives.  The enum vga_switcheroo_client_id has a
VGA_SWITCHEROO_UNKNOWN_ID define which I guess these functions are
supposed to return on error.  They never do return that, but I bet
that's what we are supposed to check for.  It honestly might be good
to check for both...

if ((int)id < 0 || id == VGA_SWITCHEROO_UNKNOWN_ID) {
mutex_unlock(_mutex);
return -EINVAL;
}

Agreed, I will send v2 patch soon.
Really thanks for your wonderful suggestion! :)

Su Hui


regards,
dan carpenter



Re: [PATCH] vga_switcheroo: Fix impossible judgment condition

2023-10-25 Thread Dan Carpenter
On Thu, Oct 26, 2023 at 10:10:57AM +0800, Su Hui wrote:
> 'id' is enum type like unsigned int, so it will never be less than zero.
> 
> Fixes: 4aaf448fa975 ("vga_switcheroo: set audio client id according to bound 
> GPU id")
> Signed-off-by: Su Hui 
> ---
>  drivers/gpu/vga/vga_switcheroo.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/vga/vga_switcheroo.c 
> b/drivers/gpu/vga/vga_switcheroo.c
> index 365e6ddbe90f..d3064466fd3a 100644
> --- a/drivers/gpu/vga/vga_switcheroo.c
> +++ b/drivers/gpu/vga/vga_switcheroo.c
> @@ -375,7 +375,7 @@ int vga_switcheroo_register_audio_client(struct pci_dev 
> *pdev,
>   mutex_lock(_mutex);
>   if (vgasr_priv.active) {
>   id = vgasr_priv.handler->get_client_id(vga_dev);
> - if (id < 0) {
> + if ((int)id < 0) {

Hi,

I feel like you're using Smatch?  Which is great!  Fantastic!

Have you built the cross function database?  If you have there is a
command that's useful.

$ ~/smatch/smatch_db/smdb.py functions vga_switcheroo_handler get_client_id | 
tee where
drivers/gpu/drm/nouveau/nouveau_acpi.c | (struct 
vga_switcheroo_handler)->get_client_id | nouveau_dsm_get_client_id | 1
drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c | (struct 
vga_switcheroo_handler)->get_client_id | amdgpu_atpx_get_client_id | 1
drivers/gpu/drm/radeon/radeon_atpx_handler.c | (struct 
vga_switcheroo_handler)->get_client_id | radeon_atpx_get_client_id | 1
drivers/platform/x86/apple-gmux.c | (struct 
vga_switcheroo_handler)->get_client_id | gmux_get_client_id | 1
$ make cscope
$ vim where
Use cscope to jump to each of those four functions.  Move the cursor to
the nouveau_dsm_get_client_id and hit CTRL-].

They never return negatives.  The enum vga_switcheroo_client_id has a
VGA_SWITCHEROO_UNKNOWN_ID define which I guess these functions are
supposed to return on error.  They never do return that, but I bet
that's what we are supposed to check for.  It honestly might be good
to check for both...

if ((int)id < 0 || id == VGA_SWITCHEROO_UNKNOWN_ID) {
mutex_unlock(_mutex);
return -EINVAL;
}

regards,
dan carpenter



[PATCH] vga_switcheroo: Fix impossible judgment condition

2023-10-25 Thread Su Hui
'id' is enum type like unsigned int, so it will never be less than zero.

Fixes: 4aaf448fa975 ("vga_switcheroo: set audio client id according to bound 
GPU id")
Signed-off-by: Su Hui 
---
 drivers/gpu/vga/vga_switcheroo.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/vga/vga_switcheroo.c b/drivers/gpu/vga/vga_switcheroo.c
index 365e6ddbe90f..d3064466fd3a 100644
--- a/drivers/gpu/vga/vga_switcheroo.c
+++ b/drivers/gpu/vga/vga_switcheroo.c
@@ -375,7 +375,7 @@ int vga_switcheroo_register_audio_client(struct pci_dev 
*pdev,
mutex_lock(_mutex);
if (vgasr_priv.active) {
id = vgasr_priv.handler->get_client_id(vga_dev);
-   if (id < 0) {
+   if ((int)id < 0) {
mutex_unlock(_mutex);
return -EINVAL;
}
-- 
2.30.2