Re: [PATCH] net: stmmac: set multicast filter to zero if feature is unsupported

[Yanteng Si <si.yanteng@linux.dev>]

在 5/27/25 12:17 AM, Nikunj Kela 写道:
>
> On 5/25/25 7:09 PM, Yanteng Si wrote:
>> 在 5/24/25 6:19 AM, Nikunj Kela 写道:
>>> Hash based multicast filtering is an optional feature. Currently,
>>> driver overrides the value of multicast_filter_bins based on the hash
>>> table size. If the feature is not supported, hash table size reads 0
>>> however the value of multicast_filter_bins remains set to default
>>> HASH_TABLE_SIZE which is incorrect. Let's override it to 0 if the
>>> feature is unsupported.
>>>
>>> Signed-off-by: Nikunj Kela <nikunj.kela@sima.ai>
>>> ---
>>>   drivers/net/ethernet/stmicro/stmmac/stmmac_main.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c 
>>> b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>>> index 085c09039af4..ccea9f811a05 100644
>>> --- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>>> +++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
>>> @@ -7241,6 +7241,9 @@ static int stmmac_hw_init(struct stmmac_priv 
>>> *priv)
>>>                       (BIT(priv->dma_cap.hash_tb_sz) << 5);
>>>               priv->hw->mcast_bits_log2 =
>>> ilog2(priv->hw->multicast_filter_bins);
>>> +        } else {
>>> +            priv->hw->multicast_filter_bins = 0;
>>> +            priv->hw->mcast_bits_log2 = 0;
>>>           }
>>
>> I didn't read the code carefully, just did a simple search：
>>
>> ❯ grep -rn multicast_filter_bins drivers/net/
>> drivers/net/ethernet/stmicro/stmmac/common.h:611:    unsigned int 
>> multicast_filter_bins;
>> drivers/net/ethernet/stmicro/stmmac/dwmac-sophgo.c:26: 
>> plat_dat->multicast_filter_bins = 0;
>> ***
>> drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:512: 
>> plat->multicast_filter_bins = HASH_TABLE_SIZE;
>> ***
>> drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:536: 
>> &plat->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:541: 
>> plat->multicast_filter_bins = dwmac1000_validate_mcast_bins(
>> drivers/net/ethernet/stmicro/stmmac/stmmac_platform.c:542: 
>> &pdev->dev, plat->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c:523: 
>> mac->multicast_filter_bins = priv->plat->multicast_filter_bins;
>> drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c:527:    if 
>> (mac->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/dwmac1000_core.c:528: 
>> mac->mcast_bits_log2 = ilog2(mac->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:516: 
>> (netdev_mc_count(dev) > hw->multicast_filter_bins)) {
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:1527: 
>> mac->multicast_filter_bins = priv->plat->multicast_filter_bins;
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:1531:    if 
>> (mac->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:1532: 
>> mac->mcast_bits_log2 = ilog2(mac->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:1568: 
>> mac->multicast_filter_bins = priv->plat->multicast_filter_bins;
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:1572:    if 
>> (mac->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/dwxgmac2_core.c:1573: 
>> mac->mcast_bits_log2 = ilog2(mac->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/stmmac_selftests.c:543: if 
>> (netdev_mc_count(priv->dev) >= priv->hw->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/stmmac_selftests.c:633: if 
>> (netdev_mc_count(priv->dev) >= priv->hw->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/stmmac_selftests.c:679: if 
>> (netdev_mc_count(priv->dev) >= priv->hw->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c:31: 
>> plat->multicast_filter_bins = HASH_TABLE_SIZE;
>> drivers/net/ethernet/stmicro/stmmac/stmmac_pci.c:84: 
>> plat->multicast_filter_bins = HASH_TABLE_SIZE;
>> drivers/net/ethernet/stmicro/stmmac/dwmac-loongson.c:98: 
>> plat->multicast_filter_bins = 256;
>> drivers/net/ethernet/stmicro/stmmac/dwmac-loongson.c:365: 
>> mac->multicast_filter_bins = priv->plat->multicast_filter_bins;
>> drivers/net/ethernet/stmicro/stmmac/dwmac-loongson.c:369:    if 
>> (mac->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/dwmac-loongson.c:370: 
>> mac->mcast_bits_log2 = ilog2(mac->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/stmmac_main.c:7240: 
>> priv->hw->multicast_filter_bins =
>> drivers/net/ethernet/stmicro/stmmac/stmmac_main.c:7243: 
>> ilog2(priv->hw->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c:570: 
>> plat->multicast_filter_bins = HASH_TABLE_SIZE;
>> drivers/net/ethernet/stmicro/stmmac/dwmac-intel.c:707: 
>> plat->multicast_filter_bins = HASH_TABLE_SIZE;
>> drivers/net/ethernet/stmicro/stmmac/dwmac-ipq806x.c:479: 
>> plat_dat->multicast_filter_bins = 0;
>> drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:456:    int 
>> numhashregs = (hw->multicast_filter_bins >> 5);
>> drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:485: 
>> (netdev_mc_count(dev) > hw->multicast_filter_bins)) {
>> drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:1057: 
>> mac->multicast_filter_bins = priv->plat->multicast_filter_bins;
>> drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:1061:    if 
>> (mac->multicast_filter_bins)
>> drivers/net/ethernet/stmicro/stmmac/dwmac4_core.c:1062: 
>> mac->mcast_bits_log2 = ilog2(mac->multicast_filter_bins);
>> drivers/net/ethernet/stmicro/stmmac/dwmac-generic.c:43: 
>> plat_dat->multicast_filter_bins = HASH_TABLE_SIZE;
>>
>> and
>>
>> drivers/net/ethernet/stmicro/stmmac/common.h:265:#define 
>> HASH_TABLE_SIZE 64
>>
>>
>> From the search results, the default value of multicast_filter_bins
>> may be meaningful. And I think that even if some hardware does not
>> support this feature, it should still be overridden in its own 
>> directory.
>
> There is a DT property 'snps,multicast-filter-bins' available to 
> override the default value for a platform however this property is not 
> taken into consideration in case of xgmac. That being said, 
> stmmac_platform.c logic can be modified to extend the property use for 
> xgmac also however the value will be overridden later based on hash 
> table size read from the HW Feature register. So only zero value will 
> be usable via DT in that case. Hence I thought of setting it to 0 in 
> the else part of the code I 

> modified in this patch. If you think I should extend the DT property 
> for xgmac, I can modify the patch too. 

Thanks for your reply. I think this is better.


Thanks,

Yanteng


>
> Thanks,
>
> -Nikunj
>
>>
>>
>> Thanks,
>> Yanteng
>>