Hi All, On 02/09/2016 09:38 PM, Arnd Bergmann wrote: > On Tuesday 09 February 2016 16:55:42 Karicheri, Muralidharan wrote: >> >> The descriptors are usable by different drivers, one driver may use it as >> buf ptr/ len, other for something else. So they should remain as generic >> and it is up to individual drivers to use it in whatever way it requires. >> My suggestion is to rename pad field in struct knav_dma_desc to sw_data >> to avoid confusion. i.e >> >> + __le32 pad[4]; >> >> to >> >> + __le32 sw_data[4]; >> > > If the hardware doesn't access them, they can probably just be u32 > and not do any byte swapping. >
Most of propositions, mentioned in this thread (and not only), sounds reasonable, and there are definitely a lot of optimization/cleanups and refactorings which could be done for this driver (of course it's up to NETCP maintainers to decide). But this particular patch is intended to fix regression and restore Keystone 2 networking functionality, *including NFS boot*, when vanilla's config files are used (keystone_defconfig, multi_v7_defconfig). And this is "not" an optimization or cleanup - this is partial revert of buggy commit. So, could we fix regression first, pleeeaaase? Then other things can be done if someone have time and would like to take the initiative (and would be ready to get at minimum "Tested-by" tag from netcp maintainers :)). -- regards, -grygorii
