The ?base driver? was not developed for DPDK only. So any code changes specifically for DPDK shouldn?t be part of it. We should think of moving it to other place than that ?base driver?. For the bug fixes, I think it is feasible to report the issue back to the ?base driver? owners, and get it onto the next release later.
We rely on them to maintain the ?base driver?, and support newer hardware. They have more developers and community developers, and the bugs could possibly be found earlier than us. I don?t think we have better choices if we don?t want to maintain such around 30 thousands or more of source lines per NIC type (e.g. e1000, ixgbe, i40e, fm10k). Generally this working model provides us more benefits than the model of we maintaining it. Actually, we did quite well on this working model, I don?t think we must change it till now. Regards, Helin From: Stephen Hemminger [mailto:step...@networkplumber.org] Sent: Wednesday, April 22, 2015 9:46 AM To: Zhang, Helin Cc: dev at dpdk.org; Vlad Zolotarov Subject: Re: [dpdk-dev] [PATCH v8 0/3]: Add LRO support to ixgbe PMD On Apr 21, 2015 6:23 PM, "Zhang, Helin" <helin.zhang at intel.com<mailto:helin.zhang at intel.com>> wrote: > > Hi Vlad > > I have a concern about the code changes you added in ixgbe_type.h. > For ixgbe, all source files in librte_pmd_ixgbe/ixgbe, except ixgbe_osdep.h > were called as "base driver", which was not developed by DPDK developers, and > released by the other team. We never modify any code in those base driver > source files, and just copy those file into DPDK project. I have a concern long term that this model is not going to work. It stifles innovation, prevents bug fixes and discourages contributors.
