On Wed, Dec 13, 2023 at 5:19 AM Stephen Hemminger <step...@networkplumber.org> wrote: > > The generic RTE_LOGTYPE_PMD is a leftover and should be removed. > As a first step, fix many drivers to not use it, and add a > helper for the RTE_LOG_DP().
I don't see why we need a new rte_log_dp helper. The #define RTE_LOGTYPE_XXX trick should work fine with RTE_LOG_DP(). What else is missing? > > Most of this patchset is boiler plate but there were some > places where use of PMD type snuck in with changes to > original driver and get fixed here. I see odd changes, like a change in PMD_DRV_LOG meaning in the first patch: https://patchwork.dpdk.org/project/dpdk/patch/20231213041920.729403-4-step...@networkplumber.org/ A lot of drivers use this macro to log "driver"/configuration level errors (see below). Switching this macro to the DP version looks wrong to me. $ git grep PMD_DRV_LOG drivers/net/atlantic/ drivers/net/atlantic/atl_ethdev.c: PMD_DRV_LOG(ERR, "rte_eal_alarm_set fail"); drivers/net/atlantic/atl_ethdev.c: PMD_DRV_LOG(INFO, "Port %d: Link Up - speed %u Mbps - %s", drivers/net/atlantic/atl_ethdev.c: PMD_DRV_LOG(INFO, " Port %d: Link Down", drivers/net/atlantic/atl_ethdev.c: PMD_DRV_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT, ... -- David Marchand
