+++ 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@xxxxxxxxx>
---
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@xxxxxxxxxx>
M: Jessica Yu <jeyu@xxxxxxxxxx>
M: Jiri Kosina <jikos@xxxxxxxxxx>
M: Miroslav Benes <mbenes@xxxxxxx>
+M: Abel Vesa <abelvesa@xxxxxxxxx>
R: Petr Mladek <pmladek@xxxxxxxx>
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@xxxxxxxxx>
+ *
+ * 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@xxxxxxxxx>
+ *
+ * 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
--
To unsubscribe from this list: send the line "unsubscribe live-patching" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
