I am also looking into it since it's our change that is causing the issue. The functions modified don't return status codes so that must mean that it is causing the transaction to be configured incorrectly. However I don't have hardware to test these platforms locally, but I might be able to get a coworker who has an am62 platform to help me test (not sure which one though).
Some things I would like to look at first: 1) host->ops->adma_write_desc is expected to be NULL on this platform. If it weren't it'd probably crash instead but it seems like a decent starting point. I believe this corresponds to j721e_4bit_sdhci_ops in this case? 2) Could we compare the contents of the descriptors that are written in sdhci_adma_write_desc between working and broken builds? 3) sdhci_adma_init is only called if the descriptor table is null; is there any chance it has a random value instead of NULL to start off, so it isn't called and adma_addr is not populated correctly? Thanks, Greg -- Greg Malysa Timesys Corporation
