November 30, 20232 yr Hi all, I am trying to bring up a new board that uses both Allwinner H616 emacs, the 1Gbps (emac0) and the 10/100Mbps w/internal phy (emac1). I brought in all patches found about that, but it looks like the emac1 does not work properly. It does not reset, this is what I found in the dmesg: [ 2.228695] dwmac-sun8i 5030000.ethernet: EMAC reset timeout As per the datasheet, the reg EMAC_BASIC_CTL1's LSB (SOFT_RST bit) is a bit that must set to 1 to reset, then automatically resets to 0 when the emac has finished the reset procedure. In my case, it remains 1 causing the kernel to unload emac1 driver. Anyone succeded in support the 10/100Mbps internal phy? Thanks in advance!
December 1, 20232 yr it is known that some boards may have defective ethernet PHY chips or possibly shorts on the PCB (e.g. pcb defects during soldering) say where the ethernet PHY are located. it is 'useless' to try to fix it in software if indeed it is a *hardware problem*, i.e. that the ethernet phy simply *won't work*. no amount of codes or fixes will fix that, one may need to 'throw the board away' and get a new one or get a different board with a different chip. https://github.com/MichaIng/DietPi/issues/6594#issuecomment-1827683215 ^ for the record in this test with the defective ethernet PHY chip. tested in 3 different distributions from 3 different sources did not fix the problem, each distribution is gigabytes of binaries in 3 different sd cards. after none stop round the clock days of different sets of tests and fixes, including like using 100 Mbps instead of 1 Gbps none of that fix the problem (of the defective ethernet PHY chip, to get the point across. Edited December 1, 20232 yr by ag123
December 1, 20232 yr the SBC (single board computers) landscape is littered with *undeclared ethernet and wifi chips* (not given in specs when the boards are sold) they only tell you the SOC like H616, and that's it ! after you buy the board *undocumented ethernet and wifi chips* after you made a best effort attempt to find drivers *unsupported* spaghetti codes, that patches *undeclared hexidecimal commands that access *unknown* chip registers* for *unknown purpose* and *unknown functionality* with *unknown binary blobs* of *unknown versions* for *unknown hardware matches* it is a lottery if anything works at all Edited December 1, 20232 yr by ag123
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.