On Fri, Nov 15, 2024 at 02:59:18PM -0800, Jakub Kicinski wrote: > On Fri, 15 Nov 2024 22:59:18 +0100 Christian Marangi wrote: > > On Thu, Nov 14, 2024 at 07:22:02PM -0800, Jakub Kicinski wrote: > > > On Tue, 12 Nov 2024 21:47:26 +0100 Christian Marangi wrote: > > > > + MIB_DESC(1, 0x00, "TxDrop"), > > > > + MIB_DESC(1, 0x04, "TxCrcErr"), > > > > > > What is a CRC Tx error :o > > > Just out of curiosity, not saying its worng. > > > > > > > From Documentation, FCS error frame due to TX FIFO underrun. > > Interesting > Seems it's even supported in stats. > > > > + MIB_DESC(1, 0x08, "TxUnicast"), > > > > + MIB_DESC(1, 0x0c, "TxMulticast"), > > > > + MIB_DESC(1, 0x10, "TxBroadcast"), > > > > + MIB_DESC(1, 0x14, "TxCollision"), > > > > > > Why can't these be rtnl stats, please keep in mind that we ask that > > > people don't duplicate in ethtool -S what can be exposed via standard > > > stats > > > > > > > Ok I will search for this but it does sounds like something new and not > > used by other DSA driver, any hint on where to look for examples? > > It's relatively recent but I think the ops are plumbed thru to DSA. > Take a look at all the *_stats members of struct dsa_switch_ops, most > of them take a fixed format struct to fill in and the struct has some > extra kdoc on which field is what. Thanks for the follow-up, they are the get_stats64 I assume, quite different to the ethtools one as we need a poll logic. Ok I will check what to drop and rework it. -- Ansuel
