| Message ID | 20160627151722.B81C968D29@newverein.lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On Mon, 27 Jun 2016, Torsten Duwe wrote: > diff --git a/arch/arm64/include/asm/livepatch.h b/arch/arm64/include/asm/livepatch.h > new file mode 100644 > index 0000000..6b9a3d1 > --- /dev/null > +++ b/arch/arm64/include/asm/livepatch.h > @@ -0,0 +1,37 @@ > +/* > + * livepatch.h - arm64-specific Kernel Live Patching Core > + * > + * Copyright (C) 2016 SUSE > + * > + * 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_ARM64_LIVEPATCH_H > +#define _ASM_ARM64_LIVEPATCH_H > + > +#include <linux/module.h> > +#include <linux/ftrace.h> > + > +#ifdef CONFIG_LIVEPATCH A nit but we removed such guards in the other header files. > +static inline int klp_check_compiler_support(void) > +{ > + return 0; > +} > + > +static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long ip) > +{ > + regs->pc = ip; > +} > +#endif /* CONFIG_LIVEPATCH */ I also checked mod_arch_specific structure because of the way we deal with relocations. It is defined only if CONFIG_ARM64_MODULE_PLTS is enabled and there is a pointer to 'struct elf64_shdr' called plt. It is used indirectly in apply_relocate_add() so we need it to stay. However it points to an existing Elf section and SHF_ALLOC is added to its sh_flags in module_frob_arch_sections() (arch/arm64/kernel/module-plts.c). Therefore we should be ok. Jessica, could you check it as well, please? Thanks, Miroslav
+++ Miroslav Benes [11/07/16 16:03 +0200]: >On Mon, 27 Jun 2016, Torsten Duwe wrote: > >> diff --git a/arch/arm64/include/asm/livepatch.h b/arch/arm64/include/asm/livepatch.h >> new file mode 100644 >> index 0000000..6b9a3d1 >> --- /dev/null >> +++ b/arch/arm64/include/asm/livepatch.h >> @@ -0,0 +1,37 @@ >> +/* >> + * livepatch.h - arm64-specific Kernel Live Patching Core >> + * >> + * Copyright (C) 2016 SUSE >> + * >> + * 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_ARM64_LIVEPATCH_H >> +#define _ASM_ARM64_LIVEPATCH_H >> + >> +#include <linux/module.h> >> +#include <linux/ftrace.h> >> + >> +#ifdef CONFIG_LIVEPATCH > >A nit but we removed such guards in the other header files. > >> +static inline int klp_check_compiler_support(void) >> +{ >> + return 0; >> +} >> + >> +static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long ip) >> +{ >> + regs->pc = ip; >> +} >> +#endif /* CONFIG_LIVEPATCH */ > >I also checked mod_arch_specific structure because of the way we deal >with relocations. It is defined only if CONFIG_ARM64_MODULE_PLTS is >enabled and there is a pointer to 'struct elf64_shdr' called plt. It is >used indirectly in apply_relocate_add() so we need it to stay. However it >points to an existing Elf section and SHF_ALLOC is added to its sh_flags >in module_frob_arch_sections() (arch/arm64/kernel/module-plts.c). >Therefore we should be ok. > >Jessica, could you check it as well, please? That sounds right, the plt will remain in module core memory, so we are fine there. However I think the plt->sh_size calculation will be incorrect for livepatch modules. In calculating mod->arch.plt_max_entries (see: module-plts.c), count_plts() is called for every rela section. For livepatch modules, this means count_plts() will also be called for our .klp.rela sections, which is correct behavior. However, count_plts() only considers relas referring to SHN_UNDEF symbols, and since every rela in a klp rela section refers to a SHN_LIVEPATCH symbol, these are all ignored. So count_plts() may return an incorrect value for a klp rela section. Miroslav, can you confirm the issue? I think the fix would be easy though; we can just add an additional check for SHN_LIVEPATCH in count_plts(). Jessica
On Mon, 11 Jul 2016, Jessica Yu wrote: > +++ Miroslav Benes [11/07/16 16:03 +0200]: > > On Mon, 27 Jun 2016, Torsten Duwe wrote: > > > > > diff --git a/arch/arm64/include/asm/livepatch.h > > > b/arch/arm64/include/asm/livepatch.h > > > new file mode 100644 > > > index 0000000..6b9a3d1 > > > --- /dev/null > > > +++ b/arch/arm64/include/asm/livepatch.h > > > @@ -0,0 +1,37 @@ > > > +/* > > > + * livepatch.h - arm64-specific Kernel Live Patching Core > > > + * > > > + * Copyright (C) 2016 SUSE > > > + * > > > + * 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_ARM64_LIVEPATCH_H > > > +#define _ASM_ARM64_LIVEPATCH_H > > > + > > > +#include <linux/module.h> > > > +#include <linux/ftrace.h> > > > + > > > +#ifdef CONFIG_LIVEPATCH > > > > A nit but we removed such guards in the other header files. > > > > > +static inline int klp_check_compiler_support(void) > > > +{ > > > + return 0; > > > +} > > > + > > > +static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long > > > ip) > > > +{ > > > + regs->pc = ip; > > > +} > > > +#endif /* CONFIG_LIVEPATCH */ > > > > I also checked mod_arch_specific structure because of the way we deal > > with relocations. It is defined only if CONFIG_ARM64_MODULE_PLTS is > > enabled and there is a pointer to 'struct elf64_shdr' called plt. It is > > used indirectly in apply_relocate_add() so we need it to stay. However it > > points to an existing Elf section and SHF_ALLOC is added to its sh_flags > > in module_frob_arch_sections() (arch/arm64/kernel/module-plts.c). > > Therefore we should be ok. > > > > Jessica, could you check it as well, please? > > That sounds right, the plt will remain in module core memory, so we > are fine there. > > However I think the plt->sh_size calculation will be incorrect for > livepatch modules. In calculating mod->arch.plt_max_entries (see: > module-plts.c), count_plts() is called for every rela section. > For livepatch modules, this means count_plts() will also be called for > our .klp.rela sections, which is correct behavior. However, > count_plts() only considers relas referring to SHN_UNDEF symbols, and > since every rela in a klp rela section refers to a SHN_LIVEPATCH > symbol, these are all ignored. So count_plts() may return an incorrect > value for a klp rela section. You're right. During the patch module creation we basically transform all SHN_UNDEF relas to SHN_LIVEPATCH, right? We must take it into account here. > Miroslav, can you confirm the issue? I think the fix would be easy > though; we can just add an additional check for SHN_LIVEPATCH in > count_plts(). Yes, such a check should be sufficient. Thanks for looking into it. Miroslav
On Mon, Jun 27, 2016 at 11:17 AM, Torsten Duwe <duwe@lst.de> wrote: > On top of FTRACE_WITH_REGS and the klp changes that go into v4.7 > this is straightforward. > > Signed-off-by: Torsten Duwe <duwe@suse.de> > --- > arch/arm64/Kconfig | 3 +++ > arch/arm64/include/asm/livepatch.h | 37 +++++++++++++++++++++++++++++++++++++ > arch/arm64/kernel/entry-ftrace.S | 13 +++++++++++++ > 3 files changed, 53 insertions(+) > create mode 100644 arch/arm64/include/asm/livepatch.h > > diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig > index 36a0e26..cb5adf3 100644 > --- a/arch/arm64/Kconfig > +++ b/arch/arm64/Kconfig > @@ -80,6 +80,7 @@ config ARM64 > select HAVE_GENERIC_DMA_COHERENT > select HAVE_HW_BREAKPOINT if PERF_EVENTS > select HAVE_IRQ_TIME_ACCOUNTING > + select HAVE_LIVEPATCH > select HAVE_MEMBLOCK > select HAVE_MEMBLOCK_NODE_MAP if NUMA > select HAVE_PATA_PLATFORM > @@ -1042,4 +1043,6 @@ if CRYPTO > source "arch/arm64/crypto/Kconfig" > endif > > +source "kernel/livepatch/Kconfig" > + > source "lib/Kconfig" > diff --git a/arch/arm64/include/asm/livepatch.h b/arch/arm64/include/asm/livepatch.h > new file mode 100644 > index 0000000..6b9a3d1 > --- /dev/null > +++ b/arch/arm64/include/asm/livepatch.h > @@ -0,0 +1,37 @@ > +/* > + * livepatch.h - arm64-specific Kernel Live Patching Core > + * > + * Copyright (C) 2016 SUSE > + * > + * 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_ARM64_LIVEPATCH_H > +#define _ASM_ARM64_LIVEPATCH_H > + > +#include <linux/module.h> > +#include <linux/ftrace.h> These includes don't look right. It would seem all you need is the one for struct pt_regs. Paul. -- > + > +#ifdef CONFIG_LIVEPATCH > +static inline int klp_check_compiler_support(void) > +{ > + return 0; > +} > + > +static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long ip) > +{ > + regs->pc = ip; > +} > +#endif /* CONFIG_LIVEPATCH */ > + > +#endif /* _ASM_ARM64_LIVEPATCH_H */ > diff --git a/arch/arm64/kernel/entry-ftrace.S b/arch/arm64/kernel/entry-ftrace.S > index 3ebe791..b166cbf 100644 > --- a/arch/arm64/kernel/entry-ftrace.S > +++ b/arch/arm64/kernel/entry-ftrace.S > @@ -204,6 +204,9 @@ ENTRY(ftrace_caller) > str x9, [sp, #S_LR] > /* The program counter just after the ftrace call site */ > str lr, [sp, #S_PC] > +#if defined(CONFIG_LIVEPATCH) && defined(CONFIG_FUNCTION_GRAPH_TRACER) > + mov x19,lr /* remember old return address */ > +#endif > /* The stack pointer as it was on ftrace_caller entry... */ > add x29, sp, #S_FRAME_SIZE+16 /* ...is also our new FP */ > str x29, [sp, #S_SP] > @@ -219,6 +222,16 @@ ftrace_call: > > bl ftrace_stub > > +#if defined(CONFIG_LIVEPATCH) && defined(CONFIG_FUNCTION_GRAPH_TRACER) > + /* Is the trace function a live patcher an has messed with > + * the return address? > + */ > + ldr x9, [sp, #S_PC] > + cmp x9, x19 /* compare with the value we remembered */ > + /* to not call graph tracer's "call" mechanism twice! */ > + b.eq ftrace_regs_return > +#endif > + > #ifdef CONFIG_FUNCTION_GRAPH_TRACER > .global ftrace_graph_call > ftrace_graph_call: // ftrace_graph_caller(); > -- > 2.6.6 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-arch" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html
On Mon, Jul 11, 2016 at 04:03:08PM +0200, Miroslav Benes wrote: > On Mon, 27 Jun 2016, Torsten Duwe wrote: > > + > > +#ifdef CONFIG_LIVEPATCH > > A nit but we removed such guards in the other header files. I just notice this has fallen between the cracks :-/ Torsten
diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig index 36a0e26..cb5adf3 100644 --- a/arch/arm64/Kconfig +++ b/arch/arm64/Kconfig @@ -80,6 +80,7 @@ config ARM64 select HAVE_GENERIC_DMA_COHERENT select HAVE_HW_BREAKPOINT if PERF_EVENTS select HAVE_IRQ_TIME_ACCOUNTING + select HAVE_LIVEPATCH select HAVE_MEMBLOCK select HAVE_MEMBLOCK_NODE_MAP if NUMA select HAVE_PATA_PLATFORM @@ -1042,4 +1043,6 @@ if CRYPTO source "arch/arm64/crypto/Kconfig" endif +source "kernel/livepatch/Kconfig" + source "lib/Kconfig" diff --git a/arch/arm64/include/asm/livepatch.h b/arch/arm64/include/asm/livepatch.h new file mode 100644 index 0000000..6b9a3d1 --- /dev/null +++ b/arch/arm64/include/asm/livepatch.h @@ -0,0 +1,37 @@ +/* + * livepatch.h - arm64-specific Kernel Live Patching Core + * + * Copyright (C) 2016 SUSE + * + * 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_ARM64_LIVEPATCH_H +#define _ASM_ARM64_LIVEPATCH_H + +#include <linux/module.h> +#include <linux/ftrace.h> + +#ifdef CONFIG_LIVEPATCH +static inline int klp_check_compiler_support(void) +{ + return 0; +} + +static inline void klp_arch_set_pc(struct pt_regs *regs, unsigned long ip) +{ + regs->pc = ip; +} +#endif /* CONFIG_LIVEPATCH */ + +#endif /* _ASM_ARM64_LIVEPATCH_H */ diff --git a/arch/arm64/kernel/entry-ftrace.S b/arch/arm64/kernel/entry-ftrace.S index 3ebe791..b166cbf 100644 --- a/arch/arm64/kernel/entry-ftrace.S +++ b/arch/arm64/kernel/entry-ftrace.S @@ -204,6 +204,9 @@ ENTRY(ftrace_caller) str x9, [sp, #S_LR] /* The program counter just after the ftrace call site */ str lr, [sp, #S_PC] +#if defined(CONFIG_LIVEPATCH) && defined(CONFIG_FUNCTION_GRAPH_TRACER) + mov x19,lr /* remember old return address */ +#endif /* The stack pointer as it was on ftrace_caller entry... */ add x29, sp, #S_FRAME_SIZE+16 /* ...is also our new FP */ str x29, [sp, #S_SP] @@ -219,6 +222,16 @@ ftrace_call: bl ftrace_stub +#if defined(CONFIG_LIVEPATCH) && defined(CONFIG_FUNCTION_GRAPH_TRACER) + /* Is the trace function a live patcher an has messed with + * the return address? + */ + ldr x9, [sp, #S_PC] + cmp x9, x19 /* compare with the value we remembered */ + /* to not call graph tracer's "call" mechanism twice! */ + b.eq ftrace_regs_return +#endif + #ifdef CONFIG_FUNCTION_GRAPH_TRACER .global ftrace_graph_call ftrace_graph_call: // ftrace_graph_caller();
On top of FTRACE_WITH_REGS and the klp changes that go into v4.7 this is straightforward. Signed-off-by: Torsten Duwe <duwe@suse.de> --- arch/arm64/Kconfig | 3 +++ arch/arm64/include/asm/livepatch.h | 37 +++++++++++++++++++++++++++++++++++++ arch/arm64/kernel/entry-ftrace.S | 13 +++++++++++++ 3 files changed, 53 insertions(+) create mode 100644 arch/arm64/include/asm/livepatch.h