Messages in this thread Patch in this message | | | Date | Mon, 16 Jan 2017 19:22:39 -0500 | From | Jessica Yu <> | Subject | Re: arm: Add livepatch arch specific code |
| |
+++ Miroslav Benes [16/01/17 17:47 +0100]: >On Tue, 6 Dec 2016, Abel Vesa wrote: > >> klp_get_ftrace_location is used by ftrace to get the entry for a >> specific function from the mcount list. klp_arch_set_pc is used >> to set the pc from the regs passed as an argument to the >> ftrace_ops_no_ops function to the starting address of the patched >> function. klp_write_module_reloc is not doing anything at this >> moment. >> >> Signed-off-by: Abel Vesa <abelvesa@linux.com> >> --- >> MAINTAINERS | 3 +++ >> arch/arm/include/asm/livepatch.h | 46 ++++++++++++++++++++++++++++++++++++++++ >> arch/arm/kernel/livepatch.c | 43 +++++++++++++++++++++++++++++++++++++ >> 3 files changed, 92 insertions(+) >> create mode 100644 arch/arm/include/asm/livepatch.h >> create mode 100644 arch/arm/kernel/livepatch.c >> >> diff --git a/MAINTAINERS b/MAINTAINERS >> index bd182a1..d43b790 100644 >> --- a/MAINTAINERS >> +++ b/MAINTAINERS >> @@ -7466,12 +7466,15 @@ M: Josh Poimboeuf <jpoimboe@redhat.com> >> M: Jessica Yu <jeyu@redhat.com> >> M: Jiri Kosina <jikos@kernel.org> >> M: Miroslav Benes <mbenes@suse.cz> >> +M: Abel Vesa <abelvesa@linux.com> >> R: Petr Mladek <pmladek@suse.com> >> S: Maintained >> F: kernel/livepatch/ >> F: include/linux/livepatch.h >> F: arch/x86/include/asm/livepatch.h >> F: arch/x86/kernel/livepatch.c >> +F: arch/arm/include/asm/livepatch.h >> +F: arch/arm/kernel/livepatch.c >> F: Documentation/livepatch/ >> F: Documentation/ABI/testing/sysfs-kernel-livepatch >> F: samples/livepatch/ > >Thanks for the offer but I think we manage quite well ;) > >> diff --git a/arch/arm/include/asm/livepatch.h b/arch/arm/include/asm/livepatch.h >> new file mode 100644 >> index 0000000..d4e3ff0 >> --- /dev/null >> +++ b/arch/arm/include/asm/livepatch.h >> @@ -0,0 +1,46 @@ >> +/* >> + * livepatch.h - arm specific Kernel Live Patching Core >> + * >> + * Copyright (C) 2016 Abel Vesa <abelvesa@linux.com> >> + * >> + * This program is free software; you can redistribute it and/or >> + * modify it under the terms of the GNU General Public License >> + * as published by the Free Software Foundation; either version 2 >> + * of the License, or (at your option) any later version. >> + * >> + * This program is distributed in the hope that it will be useful, >> + * but WITHOUT ANY WARRANTY; without even the implied warranty of >> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the >> + * GNU General Public License for more details. >> + * >> + * You should have received a copy of the GNU General Public License >> + * along with this program; if not, see <http://www.gnu.org/licenses/>. >> + */ >> + >> +#ifndef _ASM_ARM_LIVEPATCH_H >> +#define _ASM_ARM_LIVEPATCH_H >> + >> +#include <asm/setup.h> >> +#include <linux/module.h> >> +#include <linux/ftrace.h> >> + >> +static inline int klp_check_compiler_support(void) >> +{ >> + return 0; >> +} >> + >> +int klp_write_module_reloc(struct module *mod, unsigned long type, >> + unsigned long loc, unsigned long value); >This is not needed. See below. > >> + >> +static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long ip) >> +{ >> + regs->uregs[15] = ip; >> +} >> + >> +#define klp_get_ftrace_location klp_get_ftrace_location >> +static inline unsigned long klp_get_ftrace_location(unsigned long faddr) >> +{ >> + return ftrace_location_range(faddr, faddr + 24); > >Comment here about +24 would be great. See similar in powerpc header file. > >> +} >> + >> +#endif /* _ASM_ARM_LIVEPATCH_H */ >> diff --git a/arch/arm/kernel/livepatch.c b/arch/arm/kernel/livepatch.c >> new file mode 100644 >> index 0000000..0656cd6 >> --- /dev/null >> +++ b/arch/arm/kernel/livepatch.c >> @@ -0,0 +1,43 @@ >> +/* >> + * livepatch.c - arm specific Kernel Live Patching Core >> + * >> + * Copyright (C) 2016 Abel Vesa <abelvesa@linux.com> >> + * >> + * This program is free software; you can redistribute it and/or >> + * modify it under the terms of the GNU General Public License >> + * as published by the Free Software Foundation; either version 2 >> + * of the License, or (at your option) any later version. >> + * >> + * This program is distributed in the hope that it will be useful, >> + * but WITHOUT ANY WARRANTY; without even the implied warranty of >> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the >> + * GNU General Public License for more details. >> + * >> + * You should have received a copy of the GNU General Public License >> + * along with this program; if not, see <http://www.gnu.org/licenses/>. >> + */ >> + >> +#include <linux/module.h> >> +#include <linux/uaccess.h> >> +#include <linux/ftrace.h> >> +#include <asm/elf.h> >> +#include <asm/livepatch.h> >> +#include <asm/insn.h> >> +#include <asm/ftrace.h> >> + >> +/** >> + * klp_write_module_reloc() - write a relocation in a module >> + * @mod: module in which the section to be modified is found >> + * @type: ELF relocation type (see asm/elf.h) >> + * @loc: address that the relocation should be written to >> + * @value: relocation value (sym address + addend) >> + * >> + * This function writes a relocation to the specified location for >> + * a particular module. >> + */ >> +int klp_write_module_reloc(struct module *mod, unsigned long type, >> + unsigned long loc, unsigned long value) >> +{ >> + /* Not implemented yet */ >> + return 0; >> +} > >This whole file is not needed. Livepatching uses kernel's internal >apply_relocate_add() for dealing with relocations. The only thing you need >to do is to make sure that all needed arch-specific info is preserved >during a module loading. Specifically mod_arch_specific structure needs to >be examined in this respect.
Yup. Since we rely on apply_relocate_add() to apply relocations, just make sure that you preserve anything (e.g., plt tables) you need to make that call post-module init. In the case of s390, we needed to make sure we kept mod->arch.syminfo, so that the call to apply_relocate_add() would succeed.
>But apply_relocate_add() is not implemented on arm yet. I guess it would >nice to have it... otherwise we could get to an unpleasant situation. >Livepatch module can rely on its livepatching relocations (that is, there >are some). apply_relocate_add() returns 0 on arm, so everything seems to >be nice and then boom some time later.
Doesn't it return -ENOEXEC? MODULES_USE_ELF_RELA is not defined on arm, I think (see moduleloader.h).
>The question is what happens with normal modules. There are no SHT_RELA >records probably.
For arm, I think there are only SHT_REL relocation sections (MODULES_USE_ELF_REL is set instead of MODULES_USE_ELF_RELA), so during apply_relocations() in load_module(), only apply_relocate() should be called, and not apply_relocate_add().
Hm, I guess that means if we want livepatch support for the architectures that don't support RELA relocations, we would have to check if the section is SHT_REL or SHT_RELA before calling the right apply_relocate* function.
Jessica
| |