[PATCH] PCI: brcmstb: Add missing if statement

[PATCH] PCI: brcmstb: Add missing if statement

[Alex Dewar]

brcm_pcie_resume() contains a return statement that was presumably
intended to have an "if (ret)" in front of it, otherwise the function
returns prematurely. Fix this.

I don't know if this code was tested or not, but I assume that this bug
means that this driver will not resume properly.

Fixes: ad3d29c77e1e ("PCI: brcmstb: Add control of rescal reset")
Addresses-Coverity: CID 1497099: Control flow issues (UNREACHABLE)
Signed-off-by: Alex Dewar <a.dewar@sussex.ac.uk>
---
 drivers/pci/controller/pcie-brcmstb.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
index 7a3ff4632e7c..cb0c11b7308e 100644
--- a/drivers/pci/controller/pcie-brcmstb.c
+++ b/drivers/pci/controller/pcie-brcmstb.c
@@ -1154,6 +1154,7 @@ static int brcm_pcie_resume(struct device *dev)
 	clk_prepare_enable(pcie->clk);
 
 	ret = brcm_phy_start(pcie);
+	if (ret)
 		return ret;
 
 	/* Take bridge out of reset so we can access the SERDES reg */
-- 
2.28.0

Re: [PATCH] PCI: brcmstb: Add missing if statement

[Jim Quinlan]

[-- Attachment #1: Type: text/plain, Size: 1429 bytes --]

On Mon, Sep 21, 2020 at 4:45 PM Alex Dewar <alex.dewar90@gmail.com> wrote:
>
> brcm_pcie_resume() contains a return statement that was presumably
> intended to have an "if (ret)" in front of it, otherwise the function
> returns prematurely. Fix this.
>
> I don't know if this code was tested or not, but I assume that this bug
> means that this driver will not resume properly.
>
> Fixes: ad3d29c77e1e ("PCI: brcmstb: Add control of rescal reset")
> Addresses-Coverity: CID 1497099: Control flow issues (UNREACHABLE)
> Signed-off-by: Alex Dewar <a.dewar@sussex.ac.uk>
> ---
>  drivers/pci/controller/pcie-brcmstb.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
> index 7a3ff4632e7c..cb0c11b7308e 100644
> --- a/drivers/pci/controller/pcie-brcmstb.c
> +++ b/drivers/pci/controller/pcie-brcmstb.c
> @@ -1154,6 +1154,7 @@ static int brcm_pcie_resume(struct device *dev)
>         clk_prepare_enable(pcie->clk);
>
>         ret = brcm_phy_start(pcie);
> +       if (ret)

Hello,
Florian suggested adding braces to the if clause and inserting a
"clk_disable_unprepare(-cie->clk);" before the return.  I am fine with
either what you have or implementing Florian's additional suggestion.

Thank you,
Jim Quinlan
Broadcom  STB
>                 return ret;
>
>         /* Take bridge out of reset so we can access the SERDES reg */
> --
> 2.28.0
>

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4167 bytes --]

[PATCH v2] PCI: brcmstb: Add missing if statement and error path

[Alex Dewar]

brcm_pcie_resume() contains a return statement that was presumably
intended to have an "if (ret)" in front of it, otherwise the function
returns prematurely. Fix this.

Additionally, redisable the clock on the error path.

I don't know if this code was tested or not, but I assume that this bug
means that this driver will not resume properly.

Fixes: ad3d29c77e1e ("PCI: brcmstb: Add control of rescal reset")
Addresses-Coverity: CID 1497099: Control flow issues (UNREACHABLE)
Signed-off-by: Alex Dewar <alex.dewar90@gmail.com>
---
Hi Jim,

Here's a new version of the patch.

I added the missing clk_disable_unprepare() and assumed that it would
also be required in the case that the call to brcm_pcie_setup() fails.

I didn't follow the suggestion of adding extra braces to the
if-statement as the kernel style is not to add unnecessary braces in
this case: https://www.kernel.org/doc/html/latest/process/coding-style.html?highlight=style#placing-braces-and-spaces

Best,
Alex

 drivers/pci/controller/pcie-brcmstb.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
index 7a3ff4632e7c..8bae3a4f8e49 100644
--- a/drivers/pci/controller/pcie-brcmstb.c
+++ b/drivers/pci/controller/pcie-brcmstb.c
@@ -1154,7 +1154,8 @@ static int brcm_pcie_resume(struct device *dev)
 	clk_prepare_enable(pcie->clk);
 
 	ret = brcm_phy_start(pcie);
-		return ret;
+	if (ret)
+		goto err;
 
 	/* Take bridge out of reset so we can access the SERDES reg */
 	pcie->bridge_sw_init_set(pcie, 0);
@@ -1169,12 +1170,16 @@ static int brcm_pcie_resume(struct device *dev)
 
 	ret = brcm_pcie_setup(pcie);
 	if (ret)
-		return ret;
+		goto err;
 
 	if (pcie->msi)
 		brcm_msi_set_regs(pcie->msi);
 
 	return 0;
+
+err:
+	clk_disable_unprepare(pcie->clk);
+	return ret;
 }
 
 static void __brcm_pcie_remove(struct brcm_pcie *pcie)
-- 
2.28.0

Re: [PATCH v2] PCI: brcmstb: Add missing if statement and error path

[Florian Fainelli]

On 9/21/20 2:16 PM, Alex Dewar wrote:
> brcm_pcie_resume() contains a return statement that was presumably
> intended to have an "if (ret)" in front of it, otherwise the function
> returns prematurely. Fix this.
> 
> Additionally, redisable the clock on the error path.
> 
> I don't know if this code was tested or not, but I assume that this bug
> means that this driver will not resume properly.
> 
> Fixes: ad3d29c77e1e ("PCI: brcmstb: Add control of rescal reset")
> Addresses-Coverity: CID 1497099: Control flow issues (UNREACHABLE)
> Signed-off-by: Alex Dewar <alex.dewar90@gmail.com>
> ---
You need to coordinate with Colin, or with whomever is looking at
Coverity defects, because he just submitted a similar patch minutes ago:

https://lore.kernel.org/linux-arm-kernel/20200921211555.383458-1-colin.king@canonical.com/
-- 
Florian

Re: [PATCH v2] PCI: brcmstb: Add missing if statement and error path

[Lorenzo Pieralisi]

On Mon, Sep 21, 2020 at 10:16:24PM +0100, Alex Dewar wrote:
> brcm_pcie_resume() contains a return statement that was presumably
> intended to have an "if (ret)" in front of it, otherwise the function
> returns prematurely. Fix this.
> 
> Additionally, redisable the clock on the error path.
> 
> I don't know if this code was tested or not, but I assume that this bug
> means that this driver will not resume properly.
> 
> Fixes: ad3d29c77e1e ("PCI: brcmstb: Add control of rescal reset")
> Addresses-Coverity: CID 1497099: Control flow issues (UNREACHABLE)
> Signed-off-by: Alex Dewar <alex.dewar90@gmail.com>
> ---
> Hi Jim,
> 
> Here's a new version of the patch.
> 
> I added the missing clk_disable_unprepare() and assumed that it would
> also be required in the case that the call to brcm_pcie_setup() fails.
> 
> I didn't follow the suggestion of adding extra braces to the
> if-statement as the kernel style is not to add unnecessary braces in
> this case: https://www.kernel.org/doc/html/latest/process/coding-style.html?highlight=style#placing-braces-and-spaces
> 
> Best,
> Alex
> 
>  drivers/pci/controller/pcie-brcmstb.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)

Squashed in the original patch, pushed out in my pci/brcmstb branch,
I will drop Colin's patch following this discussion. If there is
any issue please shout, thanks.

Lorenzo

> diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
> index 7a3ff4632e7c..8bae3a4f8e49 100644
> --- a/drivers/pci/controller/pcie-brcmstb.c
> +++ b/drivers/pci/controller/pcie-brcmstb.c
> @@ -1154,7 +1154,8 @@ static int brcm_pcie_resume(struct device *dev)
>  	clk_prepare_enable(pcie->clk);
>  
>  	ret = brcm_phy_start(pcie);
> -		return ret;
> +	if (ret)
> +		goto err;
>  
>  	/* Take bridge out of reset so we can access the SERDES reg */
>  	pcie->bridge_sw_init_set(pcie, 0);
> @@ -1169,12 +1170,16 @@ static int brcm_pcie_resume(struct device *dev)
>  
>  	ret = brcm_pcie_setup(pcie);
>  	if (ret)
> -		return ret;
> +		goto err;
>  
>  	if (pcie->msi)
>  		brcm_msi_set_regs(pcie->msi);
>  
>  	return 0;
> +
> +err:
> +	clk_disable_unprepare(pcie->clk);
> +	return ret;
>  }
>  
>  static void __brcm_pcie_remove(struct brcm_pcie *pcie)
> -- 
> 2.28.0
>