On Wed, Oct 17, 2018 at 11:36:56AM -0400, Doug Ledford wrote: > On Mon, 2018-10-15 at 07:55 +0800, kbuild test robot wrote: [...] > > drivers/infiniband/core/verbs.c: In function 'ib_dealloc_pd': > > drivers/infiniband/core/verbs.c:340:10: error: implicit declaration of function 'rdma_is_user_pd'; did you mean 'rdma_is_zero_gid'? [-Werror=implicit-function-declaration] > > WARN_ON(rdma_is_user_pd(pd)); > > ^ > > include/asm-generic/bug.h:122:25: note: in definition of macro 'WARN_ON' > > int __ret_warn_on = !!(condition); \ > > ^~~~~~~~~ > > cc1: some warnings being treated as errors > > Are you even building this locally before submitting it? Just so you > know, I ignore anything with kbuild errors. Please fix those up and > resubmit. > Sure. It was built and tested. But Jason asked me to split the (long) series of patches to 3 series. This patch is part of series #2. Series #1 has this commit that add the (assumed) missing function: [PATCH v2 2/4] IB/verbs: Add rdma_is_user_pd function We have 3 patch sets that kbuild test need to apply them all by order: 1: [PATCH v2 0/4] figure uverbs/kernel ib_pd w/o using ib_pd uobject 2: [PATCH v2 0/7] convey ib_ucontext via ib_udata 3: [PATCH v2 0/4] remove ib_pd uobject pointer Any way to still split those and not have the kbuild assuming that it can apply each of those independently? Thanks, Shamir > -- > Doug Ledford <dledford@xxxxxxxxxx> > GPG KeyID: B826A3330E572FDD > Key fingerprint = AE6B 1BDA 122B 23B4 265B 1274 B826 A333 0E57 2FDD
