On Sat, Jun 06, 2020 at 03:24:22PM +0200, Heiko Thiery wrote: > Hi Michael et all, > > I'm digging in the reason for a failure when building ethtool with > buildroot [1]. > > I see the following error: > --- > data/buildroot/buildroot-test/instance-0/output/host/bin/i686-linux-gcc > -DHAVE_CONFIG_H -I. -I./uapi -D_LARGEFILE_SOURCE > -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Wall -D_LARGEFILE_SOURCE > -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Os -static -c -o > netlink/desc-rtnl.o netlink/desc-rtnl.c > In file included from ./uapi/linux/ethtool_netlink.h:12, > from netlink/desc-ethtool.c:7: > ./uapi/linux/ethtool.h:1294:19: warning: implicit declaration of > function '__KERNEL_DIV_ROUND_UP' [-Wimplicit-function-declaration] > __u32 queue_mask[__KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32)]; > ^~~~~~~~~~~~~~~~~~~~~ > ./uapi/linux/ethtool.h:1294:8: error: variably modified 'queue_mask' > at file scope > __u32 queue_mask[__KERNEL_DIV_ROUND_UP(MAX_NUM_QUEUE, 32)]; > ^~~~~~~~~~ > ---
Thank you for the report. This is fixed by first part of this patch: https://patchwork.ozlabs.org/project/netdev/patch/bb60cbfe99071fca4b0ea9e62d67a2341d8dd652.1590707335.git.mkube...@suse.cz/ I'm going to apply it (with the rest of the series) this weekend. > The problems seems to be injected by the "warning: implicit > declaration of function". > > When I move the __KERNEL_DIV_ROUND_UP macro right beside usage in > "uapi/linux/ethtool.h" the failure is gone. > > --- > diff --git a/uapi/linux/ethtool.h b/uapi/linux/ethtool.h > index d3dcb45..6710fa0 100644 > --- a/uapi/linux/ethtool.h > +++ b/uapi/linux/ethtool.h > @@ -1288,6 +1288,11 @@ enum ethtool_sfeatures_retval_bits { > * @queue_mask: Bitmap of the queues which sub command apply to > * @data: A complete command structure following for each of the > queues addressed > */ > +/* ethtool.h epxects __KERNEL_DIV_ROUND_UP to be defined by <linux/kernel.h> > */ > +#include <linux/kernel.h> > +#ifndef __KERNEL_DIV_ROUND_UP > +#define __KERNEL_DIV_ROUND_UP(n, d) (((n) + (d) - 1) / (d)) > +#endif > struct ethtool_per_queue_op { > __u32 cmd; > __u32 sub_command; > --- This would fix the warning and error too but uapi/linux/ethtool.h is a sanitized copy of a kernel header which we import and do not apply further changes. Moreover, there is no need to have multiple definitions of the same macro and there is already one in internal.h. Michal
