mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH bpf-next 0/3] bpf: Fix builds broken by fatal resolve_btfids warnings
@ 2026-10-19 21:10 Sasha Levin
  2026-10-19 21:10 ` [PATCH bpf-next 1/3] bpf: Drop unused entries from special_kfunc_list again Sasha Levin
                   ` (3 more replies)
  0 siblings, 4 replies; 6+ messages in thread
From: Sasha Levin @ 2026-10-19 21:10 UTC (permalink / raw)
  To: bpf
  Cc: Sasha Levin, Alexei Starovoitov, Daniel Borkmann, John Fastabend,
	Andrii Nakryiko, Eduard Zingerman, Kumar Kartikeya Dwivedi,
	Martin KaFai Lau, Song Liu, Yonghong Song, Jiri Olsa,
	Emil Tsalapatis, Ihor Solodrai, Eric Dumazet, Kuniyuki Iwashima,
	Paolo Abeni, Willem de Bruijn, David S. Miller, Jakub Kicinski,
	Simon Horman, Amery Hung, Puranjay Mohan, Stanislav Fomichev,
	Mahe Tardy, Jiayuan Chen, linux-kernel, netdev, Petr Pavlu,
	Leon Hwang

Commit 201a454df948 ("resolve_btfids: Fix the --fatal_warnings exit
status") makes resolve_btfids warnings fail the build again with
CONFIG_WERROR=y. That exposed BTF ID references to functions some
configs don't build. KernelCI has been seeing i386
i386_defconfig+kselftest builds fail on linux-next since next-20261015:

  WARN: resolve_btfids: unresolved symbol bpf_arena_reserve_pages
  WARN: resolve_btfids: unresolved symbol bpf_arena_free_pages
  WARN: resolve_btfids: unresolved symbol bpf_arena_alloc_pages
  Failed to generate BTF for vmlinux
  Try to disable CONFIG_DEBUG_INFO_BTF

https://files.kernelci.org/kbuild-gcc-14-i386-kselftest-6ad2315ca3f195e1aa930c8a/build.log.gz

Patch 1 drops three special_kfunc_list entries that have no users,
bpf_arena_reserve_pages among them. Patch 2 makes the remaining arena
entries depend on the condition arena.o is built with. Patch 3 fixes the
same problem for bpf_lsm_socket_sendmsg in bpf_ksock.c, which breaks
x86_64 builds with CONFIG_BPF_LSM=y and CONFIG_SECURITY_NETWORK=n.

The entries left in special_kfunc_list resolve to the same IDs as
before, and the IDs that become BTF_ID_UNUSED were already zero.

Tested by building vmlinux with CONFIG_DEBUG_INFO_BTF=y and
CONFIG_WERROR=y for i386, x86_64, and x86_64 with CONFIG_BPF_LSM=y and
both CONFIG_SECURITY_NETWORK=n and =y, and comparing special_kfunc_list
and bpf_lsm_socket_sendmsg_id before and after the series.

Sasha Levin (3):
  bpf: Drop unused entries from special_kfunc_list again
  bpf: Don't reference arena kfuncs when arena isn't built
  bpf: Fix ksock build with BPF_LSM=y and SECURITY_NETWORK=n

 kernel/bpf/verifier.c | 11 +++++------
 net/core/bpf_ksock.c  |  7 ++++++-
 2 files changed, 11 insertions(+), 7 deletions(-)


base-commit: 25bb03f0a47b622b9025a1558bf2095ec1f55d26
-- 
2.53.0


^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2026-10-20  0:20 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-10-19 21:10 [PATCH bpf-next 0/3] bpf: Fix builds broken by fatal resolve_btfids warnings Sasha Levin
2026-10-19 21:10 ` [PATCH bpf-next 1/3] bpf: Drop unused entries from special_kfunc_list again Sasha Levin
2026-10-19 21:10 ` [PATCH bpf-next 2/3] bpf: Don't reference arena kfuncs when arena isn't built Sasha Levin
2026-10-19 21:10 ` [PATCH bpf-next 3/3] bpf: Fix ksock build with BPF_LSM=y and SECURITY_NETWORK=n Sasha Levin
2026-10-19 23:16   ` Mahe Tardy
2026-10-20  0:20 ` [PATCH bpf-next 0/3] bpf: Fix builds broken by fatal resolve_btfids warnings patchwork-bot+netdevbpf

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®