All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH v2 0/2] arm64: use typed entries into copied code for kCFI
@ 2026-11-11 23:26 Ben Cressey
  2026-11-11 23:26 ` [PATCH v2 1/2] arm64: kexec: use a typed entry to fix kCFI oops Ben Cressey
                   ` (2 more replies)
  0 siblings, 3 replies; 7+ messages in thread
From: Ben Cressey @ 2026-11-11 23:26 UTC (permalink / raw)
  To: Catalin Marinas, Will Deacon, Mark Rutland, Pasha Tatashin,
	Rafael J. Wysocki, Pavel Machek
  Cc: linux-arm-kernel, linux-kernel, linux-pm, Kees Cook,
	Sami Tolvanen, kexec, llvm, Nathan Chancellor, Zhaoyang Huang,
	Jeson Gao, stable

An arm64 kernel built with CONFIG_CFI=y oopses after "Bye!" on every
kexec that relocates the image. v1 [1] marked machine_kexec() __nocfi.
This version types the copied relocation code instead, as Kees Cook
suggested [2], and does the same for hibernate, as Mark Rutland asked
[3], so that swsusp_arch_resume() loses its __nocfi. Only kexec faults
today, and its fix does not depend on the hibernate patch. x86 follows
as a separate patch.

Kees, the kexec patch is your diff with two comments and the ASSERT
message reworded. It had no sign-off, so it carries Suggested-by. Send
one if you would rather be co-developer or author.

An LLM diagnosed the oops, did that rewording, wrote the hibernate
patch, and drafted the changelogs and this letter.

Tested on v7.4-rc1 with clang. Under QEMU TCG, so functional only,
kexec_file_load(2), kexec_load(2), crash kexec and hibernate/resume pass
with CFI at EL1 and booted at EL2, and with CFI off at EL1. In KVM
guests at EL1 on Graviton4 the three kexec cases pass, where v7.4-rc1
alone oopses under the two load calls, and hibernate/resume passed one
commit before v7.4-rc1. The KHO selftest passes under TCG, and oopses
at v7.4-rc1. Bare metal, BTI at the typed entries on hardware, GCC's
kCFI and LTO are untested.

Changes in v2:
- Type arm64_relocate_new_kernel() and enter its copy at the symbol's
  offset, instead of marking machine_kexec() __nocfi (Kees Cook, Mark
  Rutland).
- Take back v1's statement that arm64_relocate_new_kernel "cannot be
  given a type hash".
- Add the hibernate patch (Mark Rutland).
- Rebase onto v7.4-rc1.
- Drop the Assisted-by tag. The cover letter now says how tools were
  used.

[1] https://lore.kernel.org/20260924-arm64-kexec-nocfi-v1-1-bbae2e2eadc8@cressey.dev
[2] https://lore.kernel.org/202609241804.BB23BF10@keescook
[3] https://lore.kernel.org/arZSe8hNnRPM5FqY@J2N7QTR9R3.cambridge.arm.com

---
Ben Cressey (2):
      arm64: kexec: use a typed entry to fix kCFI oops
      arm64: hibernate: use a typed entry to drop __nocfi

 arch/arm64/include/asm/kexec.h      |  2 ++
 arch/arm64/include/asm/suspend.h    |  5 +++++
 arch/arm64/kernel/hibernate-asm.S   |  7 +++++--
 arch/arm64/kernel/hibernate.c       | 14 +++++++++-----
 arch/arm64/kernel/machine_kexec.c   |  7 +++++--
 arch/arm64/kernel/relocate_kernel.S | 11 +++++++----
 arch/arm64/kernel/vmlinux.lds.S     | 10 ++++++----
 7 files changed, 39 insertions(+), 17 deletions(-)
---
base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4
change-id: 20260924-arm64-kexec-nocfi-496e643578ce


^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2026-11-18 18:33 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-11 23:26 [PATCH v2 0/2] arm64: use typed entries into copied code for kCFI Ben Cressey
2026-11-11 23:26 ` [PATCH v2 1/2] arm64: kexec: use a typed entry to fix kCFI oops Ben Cressey
2026-11-11 23:42   ` sashiko-bot
2026-11-11 23:26 ` [PATCH v2 2/2] arm64: hibernate: use a typed entry to drop __nocfi Ben Cressey
2026-11-11 23:43   ` sashiko-bot
2026-11-18 18:33   ` Catalin Marinas
2026-11-18 18:26 ` (subset) [PATCH v2 0/2] arm64: use typed entries into copied code for kCFI Catalin Marinas

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.