[PATCH 3/3] net: ravb: Fix error handling in ravb_probe

[Marek Vasut]

On 3/19/25 3:34 AM, Marek Vasut wrote:
> On 3/5/25 9:39 PM, Paul Barker wrote:
>> Hi Marek,
> 
> Hi,
> 
>> On 05/03/2025 19:55, Marek Vasut wrote:
>>> On 3/4/25 9:07 PM, Paul Barker wrote:
>>>> In ravb_probe(), we were missing a couple of things in the error
>>>> handling path:
>>>>
>>>>     * We must unregister the MDIO bus before freeing the corresponding
>>>>       struct mii_dev instance to avoid the potential for use-after-free
>>>>       bugs.
>>>>
>>>>     * We must free the resources acquired by clk_get_bulk() even if the
>>>>       clocks have not yet been enabled.
>>>>
>>>> Fixes: 8ae51b6f324e ("net: ravb: Add Renesas Ethernet RAVB driver")
>>>> Signed-off-by: Paul Barker <paul.barker.ct at bp.renesas.com>
>>>> ---
>>>>    drivers/net/ravb.c | 14 ++++++++------
>>>>    1 file changed, 8 insertions(+), 6 deletions(-)
>>>>
>>>> diff --git a/drivers/net/ravb.c b/drivers/net/ravb.c
>>>> index 6bd94cf6bb1b..83cf6e0d19f4 100644
>>>> --- a/drivers/net/ravb.c
>>>> +++ b/drivers/net/ravb.c
>>>> @@ -591,7 +591,7 @@ static int ravb_probe(struct udevice *dev)
>>>>        ret = clk_get_bulk(dev, &eth->clks);
>>>>        if (ret < 0)
>>>> -        goto err_mdio_alloc;
>>>> +        goto err_clk_get;
>>>>        mdiodev = mdio_alloc();
>>>>        if (!mdiodev) {
>>>> @@ -613,23 +613,25 @@ static int ravb_probe(struct udevice *dev)
>>>>        /* Bring up PHY */
>>>>        ret = clk_enable_bulk(&eth->clks);
>>>>        if (ret)
>>>> -        goto err_mdio_register;
>>>> +        goto err_clk_enable;
>>>>        ret = ravb_reset(dev);
>>>>        if (ret)
>>>> -        goto err_mdio_reset;
>>>> +        goto err_clk_enable;
>>>
>>> Shouldn't this also assure clk_disable_bulk() is called ?
>>
>> Looking at include/clk.h, the comment for clk_release_bulk() says:
>>
>>      For each clock contained in the clock bulk struct, this function 
>> will check
>>      if clock has been previously requested and then will disable and 
>> free it.
>>
>> So we shouldn't need to separately call clk_disable_bulk().
> Understood. Thank you for checking.

Reviewed-by: Marek Vasut <marek.vasut+renesas at mailbox.org>