]> git.ipfire.org Git - thirdparty/kernel/stable.git/commit
um: mark rodata read-only and implement _nofault accesses
authorJohannes Berg <johannes.berg@intel.com>
Mon, 10 Feb 2025 16:09:25 +0000 (17:09 +0100)
committerJohannes Berg <johannes.berg@intel.com>
Tue, 18 Mar 2025 10:03:14 +0000 (11:03 +0100)
commitd1d7f01f7cd35e16c6bcef5a0e31988b5c9980f9
treeac2a7b2c4ccc6a9f1749c48eca519d615d125942
parent5550187c4c21740942c32a9ae56f9f472a104cb4
um: mark rodata read-only and implement _nofault accesses

Mark read-only data actually read-only (simple mprotect), and
to be able to test it also implement _nofault accesses. This
works by setting up a new "segv_continue" pointer in current,
and then when we hit a segfault we change the signal return
context so that we continue at that address. The code using
this sets it up so that it jumps to a label and then aborts
the access that way, returning -EFAULT.

It's possible to optimize the ___backtrack_faulted() thing by
using asm goto (compiler version dependent) and/or gcc's (not
sure if clang has it) &&label extension, but at least in one
attempt I made the && caused the compiler to not load -EFAULT
into the register in case of jumping to the &&label from the
fault handler. So leave it like this for now.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Co-developed-by: Benjamin Berg <benjamin.berg@intel.com>
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20250210160926.420133-2-benjamin@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
15 files changed:
arch/um/Kconfig
arch/um/include/asm/processor-generic.h
arch/um/include/asm/uaccess.h
arch/um/include/shared/arch.h
arch/um/include/shared/as-layout.h
arch/um/include/shared/irq_user.h
arch/um/include/shared/kern_util.h
arch/um/kernel/irq.c
arch/um/kernel/mem.c
arch/um/kernel/trap.c
arch/um/os-Linux/signal.c
arch/um/os-Linux/skas/process.c
arch/x86/um/os-Linux/mcontext.c
arch/x86/um/shared/sysdep/faultinfo_32.h
arch/x86/um/shared/sysdep/faultinfo_64.h