[PATCH net-next 2/3] net: stmmac: qcom-ethqos: add rgmii set/clear functions

[Konrad Dybcio]

On 11/20/25 10:46 AM, Russell King (Oracle) wrote:
> On Thu, Nov 20, 2025 at 10:42:04AM +0100, Konrad Dybcio wrote:
>> On 11/19/25 12:34 PM, Russell King (Oracle) wrote:
>>> The driver has a lot of bit manipulation of the RGMII registers. Add
>>> a pair of helpers to set bits and clear bits, converting the various
>>> calls to rgmii_updatel() as appropriate.
>>>
>>> Most of the change was done via this sed script:
>>>
>>> /rgmii_updatel/ {
>>> 	N
>>> 	/,$/N
>>> 	/mask, / ! {
>>> 		s|rgmii_updatel\(([^,]*,\s+([^,]*),\s+)\2,\s+|rgmii_setmask(\1|
>>> 		s|rgmii_updatel\(([^,]*,\s+([^,]*),\s+)0,\s+|rgmii_clrmask(\1|
>>> 		s|^\s+$||
>>> 	}
>>> }
>>>
>>> and then formatting tweaked where necessary.
>>>
>>> Signed-off-by: Russell King (Oracle) <rmk+kernel at armlinux.org.uk>
>>> ---
>>>  .../stmicro/stmmac/dwmac-qcom-ethqos.c        | 187 +++++++++---------
>>>  1 file changed, 89 insertions(+), 98 deletions(-)
>>>
>>> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
>>> index ae3cf163005b..cdaf02471d3a 100644
>>> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
>>> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-qcom-ethqos.c
>>> @@ -137,6 +137,18 @@ static void rgmii_updatel(struct qcom_ethqos *ethqos, u32 mask, u32 val,
>>>  	rgmii_writel(ethqos, temp, offset);
>>>  }
>>>  
>>> +static void rgmii_setmask(struct qcom_ethqos *ethqos, u32 mask,
>>> +			  unsigned int offset)
>>> +{
>>> +	rgmii_updatel(ethqos, mask, mask, offset);
>>> +}
>>
>> It's almost unbelieveable there's no set/clr/rmw generics for
>> readl and friends
> 
> Consider what that would mean - such operations can not be atomic, but
> users would likely not realise, which means we get a load of new
> potential bugs. Not having these means that driver authors get to
> code this up, and because they realise they have to do separate read
> and write operations, it's more obvious that there may be races.

Right, I don't think that would show up a lot in practice, but the 1
case it did would be exhaustively painful to debug

Konrad