[PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[Anders Roxell]

GCC-8 and GCC-9 emits a hard error when the value passed to
`u32_encode_bits()`. These versions somehow think that
RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) is an out of bounds
constant.  Open code this calculation using FIELD_PREP() to avoid this
compile error.

error: call to '__field_overflow' declared with attribute error: value
doesn't fit into mask

Fixes: b8eee90f0ba5 ("wifi: iwlwifi: cfg: unify num_rbds config")
Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
Closes: https://lore.kernel.org/all/CA+G9fYssasMnOE36xLH5m7ky4fKxbzN7kX5mEE7icnuu+0hGuQ@mail.gmail.com/
Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
---
 drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
index cb36baac14da..1854d071aff2 100644
--- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
+++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
@@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
 
 	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
 	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
-	control_flags |=
-		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
-				IWL_CTXT_INFO_RB_CB_SIZE);
+	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
+	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
+		RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
+		FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));
 	control_flags |= u32_encode_bits(rb_size, IWL_CTXT_INFO_RB_SIZE);
 	ctxt_info->control.control_flags = cpu_to_le32(control_flags);
 
-- 
2.47.2

Re: [PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[Dan Carpenter]

On Thu, Jun 12, 2025 at 03:07:19PM +0200, Anders Roxell wrote:
> GCC-8 and GCC-9 emits a hard error when the value passed to
> `u32_encode_bits()`. These versions somehow think that
> RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) is an out of bounds
> constant.  Open code this calculation using FIELD_PREP() to avoid this
> compile error.
> 
> error: call to '__field_overflow' declared with attribute error: value
> doesn't fit into mask
> 
> Fixes: b8eee90f0ba5 ("wifi: iwlwifi: cfg: unify num_rbds config")
> Reported-by: Linux Kernel Functional Testing <lkft@linaro.org>
> Closes: https://lore.kernel.org/all/CA+G9fYssasMnOE36xLH5m7ky4fKxbzN7kX5mEE7icnuu+0hGuQ@mail.gmail.com/
> Signed-off-by: Anders Roxell <anders.roxell@linaro.org>

Reviewed-by: Dan Carpenter <dan.carpenter@linaro.org>

regards,
dan carpenter

Re: [PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[Johannes Berg]

> 
> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> index cb36baac14da..1854d071aff2 100644
> --- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> @@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
>  
>  	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
>  	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
> -	control_flags |=
> -		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
> -				IWL_CTXT_INFO_RB_CB_SIZE);
> +	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
> +	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
> +		RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
> +		FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));
> 

The coding style is really confusing now though - what are arguments to
the macro and all that.

johannes

Re: [PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[Anders Roxell]

On 2025-06-12 17:21, Johannes Berg wrote:
> > 
> > diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> > index cb36baac14da..1854d071aff2 100644
> > --- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> > +++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> > @@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
> >  
> >  	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
> >  	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
> > -	control_flags |=
> > -		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
> > -				IWL_CTXT_INFO_RB_CB_SIZE);
> > +	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
> > +	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
> > +		RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
> > +		FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));
> > 
> 
> The coding style is really confusing now though - what are arguments to
> the macro and all that.

Would it help if I indent like this?

diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
index cb36baac14da..5bb81ed7db79 100644
--- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
+++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
@@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
 
 	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
 	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
-	control_flags |=
-		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
-				IWL_CTXT_INFO_RB_CB_SIZE);
+	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
+	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
+				    RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
+				    FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));
 	control_flags |= u32_encode_bits(rb_size, IWL_CTXT_INFO_RB_SIZE);
 	ctxt_info->control.control_flags = cpu_to_le32(control_flags);
 

Cheers,
Anders

Re: [PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[Johannes Berg]

> 
> Would it help if I indent like this?

Yeah, maybe? I actually misread it and thought the & went outside
FIELD_PREP() ...


> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> index cb36baac14da..5bb81ed7db79 100644
> --- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> @@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
>  
>  	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
>  	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
> -	control_flags |=
> -		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
> -				IWL_CTXT_INFO_RB_CB_SIZE);
> +	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
> +	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
> +				    RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
> +				    FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));

Also now that I think more about it, this really just adds the part with
the masking ("& FIELD_MAX()"), is it even necessary to use FIELD_PREP()
rather than u32_encode_bits()?

johannes

Re: [PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[Arnd Bergmann]

On Thu, Jun 12, 2025, at 21:33, Anders Roxell wrote:
> On 2025-06-12 17:21, Johannes Berg wrote:
>
> Would it help if I indent like this?
>
> +	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
> +	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
> +				    RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
> +				    FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));
>  	control_flags |= u32_encode_bits(rb_size, IWL_CTXT_INFO_RB_SIZE);

I still find it odd to mix FIELD_PREP() and u32_encode_bits() here,
as they do the same thing.

      Arnd

Re: [PATCH] wifi: iwlwifi: pcie: ensure RX_QUEUE_CB_SIZE fits bitfield for gcc-8|9

[David Laight]

On Thu, 12 Jun 2025 21:33:44 +0200
Anders Roxell <anders.roxell@linaro.org> wrote:

> On 2025-06-12 17:21, Johannes Berg wrote:
> > > 
> > > diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> > > index cb36baac14da..1854d071aff2 100644
> > > --- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> > > +++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> > > @@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
> > >  
> > >  	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
> > >  	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
> > > -	control_flags |=
> > > -		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
> > > -				IWL_CTXT_INFO_RB_CB_SIZE);
> > > +	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
> > > +	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
> > > +		RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
> > > +		FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));
> > >   
> > 
> > The coding style is really confusing now though - what are arguments to
> > the macro and all that.  
> 
> Would it help if I indent like this?
> 
> diff --git a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> index cb36baac14da..5bb81ed7db79 100644
> --- a/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> +++ b/drivers/net/wireless/intel/iwlwifi/pcie/ctxt-info.c
> @@ -204,9 +204,10 @@ int iwl_pcie_ctxt_info_init(struct iwl_trans *trans,
>  
>  	WARN_ON(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) > 12);
>  	control_flags = IWL_CTXT_INFO_TFD_FORMAT_LONG;
> -	control_flags |=
> -		u32_encode_bits(RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)),
> -				IWL_CTXT_INFO_RB_CB_SIZE);
> +	/* This should just be u32_encode_bits() but gcc-8 and gcc-9 fail to build */
> +	control_flags |= FIELD_PREP(IWL_CTXT_INFO_RB_CB_SIZE,
> +				    RX_QUEUE_CB_SIZE(iwl_trans_get_num_rbds(trans)) &
> +				    FIELD_MAX(IWL_CTXT_INFO_RB_CB_SIZE));

I would always indent the 'continued' part of an arithmetic expression.
Otherwise it looks (in this case) like another argument to FIELD_PREP().
(Not helped by the coding style requiring the '&' on the end of the line
rather than the start of the continuation line.)

	David


>  	control_flags |= u32_encode_bits(rb_size, IWL_CTXT_INFO_RB_SIZE);
>  	ctxt_info->control.control_flags = cpu_to_le32(control_flags);
>  
> 
> Cheers,
> Anders
>