Re: [patch 06/11] Text Edit Lock - Alternative code for x86 (updated)
Text Edit Lock - Alternative code for x86 Fix a memcpy that should be a text_poke (in apply_alternatives). Use kernel_wp_save/kernel_wp_restore in text_poke to support DEBUG_RODATA correctly and so the CPU HOTPLUG special case can be removed. clflush all the cachelines touched by text_poke. Add text_poke_early, for alternatives and paravirt boot-time and module load time patching. Notes: - the clflush is left there, even though Andi Kleen says it breaks some architecture. The proper fix is to detect these CPUs and set the cpu_has_clflush flag appropriately. It does not belong here. - we use a macro for kernel_wp_save/restore to mimic local_irq_save/restore: the argument is passed without &. Changelog: - Fix text_set and text_poke alignment check (mixed up bitwise and and or) - Remove text_set - Use the new macro INIT_ARRAY() to stop polluting the C files with ({ }) brackets (which breaks some c parsers in editors). - Export add_nops, so it can be used by others. - Remove x86 test for "wp_works_ok", it will just be ignored by the architecture if not supported. - Document text_poke_early. Signed-off-by: Mathieu Desnoyers <[EMAIL PROTECTED]> CC: Andi Kleen <[EMAIL PROTECTED]> CC: [EMAIL PROTECTED] --- arch/x86/kernel/alternative.c| 78 +-- include/asm-x86/alternative_32.h | 37 ++ include/asm-x86/alternative_64.h | 39 +++ 3 files changed, 132 insertions(+), 22 deletions(-) Index: linux-2.6-lttng/arch/x86/kernel/alternative.c === --- linux-2.6-lttng.orig/arch/x86/kernel/alternative.c 2007-11-13 13:43:20.0 -0500 +++ linux-2.6-lttng/arch/x86/kernel/alternative.c 2007-11-13 20:40:54.0 -0500 @@ -27,6 +27,58 @@ __setup("smp-alt-boot", bootonly); #define smp_alt_once 1 #endif +/* + * Warning: + * When you use this code to patch more than one byte of an instruction + * you need to make sure that other CPUs cannot execute this code in parallel. + * Also no thread must be currently preempted in the middle of these + * instructions. And on the local CPU you need to be protected again NMI or MCE + * handlers seeing an inconsistent instruction while you patch. + * Warning: read_cr0 is modified by paravirt, this is why we have _early + * versions. They are not in the __init section because they can be used at + * module load time. + */ +static inline void text_sync(void *addr, size_t len) +{ + void *faddr; + + sync_core(); + /* FIXME Could also do a CLFLUSH here to speed up CPU recovery; but + that causes hangs on some VIA CPUs. */ + /* Not strictly needed, but can speed CPU recovery up. */ + if (0 && cpu_has_clflush) + for (faddr = addr; faddr < addr + len; + faddr += boot_cpu_data.x86_clflush_size) + asm("clflush (%0) " :: "r" (faddr) : "memory"); +} + +void *text_poke_early(void *addr, const void *opcode, size_t len) +{ + memcpy(addr, opcode, len); + text_sync(addr, len); + return addr; +} + +/* + * Only atomic text poke/set should be allowed when not doing early patching. + * It means the size must be writable atomically and the address must be aligned + * in a way that permits an atomic write. + */ + +void *__kprobes text_poke(void *addr, const void *opcode, size_t len) +{ + unsigned long cr0; + + BUG_ON(len > sizeof(long)); + BUG_ONlong)addr + len - 1) & ~(sizeof(long) - 1)) + - ((long)addr & ~(sizeof(long) - 1))); + kernel_wp_save(cr0); + memcpy(addr, opcode, len); + kernel_wp_restore(cr0); + text_sync(addr, len); + return addr; +} + static int debug_alternative; static int __init debug_alt(char *str) @@ -173,7 +225,7 @@ static const unsigned char*const * find_ #endif /* CONFIG_X86_64 */ /* Use this to add nops to a buffer, then text_poke the whole buffer. */ -static void add_nops(void *insns, unsigned int len) +void add_nops(void *insns, unsigned int len) { const unsigned char *const *noptable = find_nop_table(); @@ -186,6 +238,7 @@ static void add_nops(void *insns, unsign len -= noplen; } } +EXPORT_SYMBOL_GPL(add_nops); extern struct alt_instr __alt_instructions[], __alt_instructions_end[]; extern u8 *__smp_locks[], *__smp_locks_end[]; @@ -219,7 +272,7 @@ void apply_alternatives(struct alt_instr memcpy(insnbuf, a->replacement, a->replacementlen); add_nops(insnbuf + a->replacementlen, a->instrlen - a->replacementlen); - text_poke(instr, insnbuf, a->instrlen); + text_poke_early(instr, insnbuf, a->instrlen); } } @@ -234,7 +287,8 @@ static void alternatives_smp_lock(u8 **s continue; if (*ptr > text_end) continue; -
Re: [patch 06/11] Text Edit Lock - Alternative code for x86 (updated)
Text Edit Lock - Alternative code for x86 Fix a memcpy that should be a text_poke (in apply_alternatives). Use kernel_wp_save/kernel_wp_restore in text_poke to support DEBUG_RODATA correctly and so the CPU HOTPLUG special case can be removed. clflush all the cachelines touched by text_poke. Add text_poke_early, for alternatives and paravirt boot-time and module load time patching. Notes: - the clflush is left there, even though Andi Kleen says it breaks some architecture. The proper fix is to detect these CPUs and set the cpu_has_clflush flag appropriately. It does not belong here. - we use a macro for kernel_wp_save/restore to mimic local_irq_save/restore: the argument is passed without . Changelog: - Fix text_set and text_poke alignment check (mixed up bitwise and and or) - Remove text_set - Use the new macro INIT_ARRAY() to stop polluting the C files with ({ }) brackets (which breaks some c parsers in editors). - Export add_nops, so it can be used by others. - Remove x86 test for wp_works_ok, it will just be ignored by the architecture if not supported. - Document text_poke_early. Signed-off-by: Mathieu Desnoyers [EMAIL PROTECTED] CC: Andi Kleen [EMAIL PROTECTED] CC: [EMAIL PROTECTED] --- arch/x86/kernel/alternative.c| 78 +-- include/asm-x86/alternative_32.h | 37 ++ include/asm-x86/alternative_64.h | 39 +++ 3 files changed, 132 insertions(+), 22 deletions(-) Index: linux-2.6-lttng/arch/x86/kernel/alternative.c === --- linux-2.6-lttng.orig/arch/x86/kernel/alternative.c 2007-11-13 13:43:20.0 -0500 +++ linux-2.6-lttng/arch/x86/kernel/alternative.c 2007-11-13 20:40:54.0 -0500 @@ -27,6 +27,58 @@ __setup(smp-alt-boot, bootonly); #define smp_alt_once 1 #endif +/* + * Warning: + * When you use this code to patch more than one byte of an instruction + * you need to make sure that other CPUs cannot execute this code in parallel. + * Also no thread must be currently preempted in the middle of these + * instructions. And on the local CPU you need to be protected again NMI or MCE + * handlers seeing an inconsistent instruction while you patch. + * Warning: read_cr0 is modified by paravirt, this is why we have _early + * versions. They are not in the __init section because they can be used at + * module load time. + */ +static inline void text_sync(void *addr, size_t len) +{ + void *faddr; + + sync_core(); + /* FIXME Could also do a CLFLUSH here to speed up CPU recovery; but + that causes hangs on some VIA CPUs. */ + /* Not strictly needed, but can speed CPU recovery up. */ + if (0 cpu_has_clflush) + for (faddr = addr; faddr addr + len; + faddr += boot_cpu_data.x86_clflush_size) + asm(clflush (%0) :: r (faddr) : memory); +} + +void *text_poke_early(void *addr, const void *opcode, size_t len) +{ + memcpy(addr, opcode, len); + text_sync(addr, len); + return addr; +} + +/* + * Only atomic text poke/set should be allowed when not doing early patching. + * It means the size must be writable atomically and the address must be aligned + * in a way that permits an atomic write. + */ + +void *__kprobes text_poke(void *addr, const void *opcode, size_t len) +{ + unsigned long cr0; + + BUG_ON(len sizeof(long)); + BUG_ONlong)addr + len - 1) ~(sizeof(long) - 1)) + - ((long)addr ~(sizeof(long) - 1))); + kernel_wp_save(cr0); + memcpy(addr, opcode, len); + kernel_wp_restore(cr0); + text_sync(addr, len); + return addr; +} + static int debug_alternative; static int __init debug_alt(char *str) @@ -173,7 +225,7 @@ static const unsigned char*const * find_ #endif /* CONFIG_X86_64 */ /* Use this to add nops to a buffer, then text_poke the whole buffer. */ -static void add_nops(void *insns, unsigned int len) +void add_nops(void *insns, unsigned int len) { const unsigned char *const *noptable = find_nop_table(); @@ -186,6 +238,7 @@ static void add_nops(void *insns, unsign len -= noplen; } } +EXPORT_SYMBOL_GPL(add_nops); extern struct alt_instr __alt_instructions[], __alt_instructions_end[]; extern u8 *__smp_locks[], *__smp_locks_end[]; @@ -219,7 +272,7 @@ void apply_alternatives(struct alt_instr memcpy(insnbuf, a-replacement, a-replacementlen); add_nops(insnbuf + a-replacementlen, a-instrlen - a-replacementlen); - text_poke(instr, insnbuf, a-instrlen); + text_poke_early(instr, insnbuf, a-instrlen); } } @@ -234,7 +287,8 @@ static void alternatives_smp_lock(u8 **s continue; if (*ptr text_end) continue; - text_poke(*ptr, ((unsigned char