From 0b053c9518292705736329a8fe20ef4686ffc8e9 Mon Sep 17 00:00:00 2001 From: mancha security Date: Wed, 18 Mar 2015 18:47:25 +0100 Subject: lib: memzero_explicit: use barrier instead of OPTIMIZER_HIDE_VAR OPTIMIZER_HIDE_VAR(), as defined when using gcc, is insufficient to ensure protection from dead store optimization. For the random driver and crypto drivers, calls are emitted ... $ gdb vmlinux (gdb) disassemble memzero_explicit Dump of assembler code for function memzero_explicit: 0xffffffff813a18b0 <+0>: push %rbp 0xffffffff813a18b1 <+1>: mov %rsi,%rdx 0xffffffff813a18b4 <+4>: xor %esi,%esi 0xffffffff813a18b6 <+6>: mov %rsp,%rbp 0xffffffff813a18b9 <+9>: callq 0xffffffff813a7120 0xffffffff813a18be <+14>: pop %rbp 0xffffffff813a18bf <+15>: retq End of assembler dump. (gdb) disassemble extract_entropy [...] 0xffffffff814a5009 <+313>: mov %r12,%rdi 0xffffffff814a500c <+316>: mov $0xa,%esi 0xffffffff814a5011 <+321>: callq 0xffffffff813a18b0 0xffffffff814a5016 <+326>: mov -0x48(%rbp),%rax [...] ... but in case in future we might use facilities such as LTO, then OPTIMIZER_HIDE_VAR() is not sufficient to protect gcc from a possible eviction of the memset(). We have to use a compiler barrier instead. Minimal test example when we assume memzero_explicit() would *not* be a call, but would have been *inlined* instead: static inline void memzero_explicit(void *s, size_t count) { memset(s, 0, count); } int main(void) { char buff[20]; snprintf(buff, sizeof(buff) - 1, "test"); printf("%s", buff); memzero_explicit(buff, sizeof(buff)); return 0; } With := OPTIMIZER_HIDE_VAR(): (gdb) disassemble main Dump of assembler code for function main: [...] 0x0000000000400464 <+36>: callq 0x400410 0x0000000000400469 <+41>: xor %eax,%eax 0x000000000040046b <+43>: add $0x28,%rsp 0x000000000040046f <+47>: retq End of assembler dump. With := barrier(): (gdb) disassemble main Dump of assembler code for function main: [...] 0x0000000000400464 <+36>: callq 0x400410 0x0000000000400469 <+41>: movq $0x0,(%rsp) 0x0000000000400471 <+49>: movq $0x0,0x8(%rsp) 0x000000000040047a <+58>: movl $0x0,0x10(%rsp) 0x0000000000400482 <+66>: xor %eax,%eax 0x0000000000400484 <+68>: add $0x28,%rsp 0x0000000000400488 <+72>: retq End of assembler dump. As can be seen, movq, movq, movl are being emitted inlined via memset(). Reference: http://thread.gmane.org/gmane.linux.kernel.cryptoapi/13764/ Fixes: d4c5efdb9777 ("random: add and use memzero_explicit() for clearing data") Cc: Theodore Ts'o Signed-off-by: mancha security Signed-off-by: Daniel Borkmann Acked-by: Hannes Frederic Sowa Acked-by: Stephan Mueller Signed-off-by: Herbert Xu --- lib/string.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'lib/string.c') diff --git a/lib/string.c b/lib/string.c index ce81aaec3839..a5792019193c 100644 --- a/lib/string.c +++ b/lib/string.c @@ -607,7 +607,7 @@ EXPORT_SYMBOL(memset); void memzero_explicit(void *s, size_t count) { memset(s, 0, count); - OPTIMIZER_HIDE_VAR(s); + barrier(); } EXPORT_SYMBOL(memzero_explicit); -- cgit v1.2.3 From 7829fb09a2b4268b30dd9bc782fa5ebee278b137 Mon Sep 17 00:00:00 2001 From: Daniel Borkmann Date: Thu, 30 Apr 2015 04:13:52 +0200 Subject: lib: make memzero_explicit more robust against dead store elimination In commit 0b053c951829 ("lib: memzero_explicit: use barrier instead of OPTIMIZER_HIDE_VAR"), we made memzero_explicit() more robust in case LTO would decide to inline memzero_explicit() and eventually find out it could be elimiated as dead store. While using barrier() works well for the case of gcc, recent efforts from LLVMLinux people suggest to use llvm as an alternative to gcc, and there, Stephan found in a simple stand-alone user space example that llvm could nevertheless optimize and thus elimitate the memset(). A similar issue has been observed in the referenced llvm bug report, which is regarded as not-a-bug. Based on some experiments, icc is a bit special on its own, while it doesn't seem to eliminate the memset(), it could do so with an own implementation, and then result in similar findings as with llvm. The fix in this patch now works for all three compilers (also tested with more aggressive optimization levels). Arguably, in the current kernel tree it's more of a theoretical issue, but imho, it's better to be pedantic about it. It's clearly visible with gcc/llvm though, with the below code: if we would have used barrier() only here, llvm would have omitted clearing, not so with barrier_data() variant: static inline void memzero_explicit(void *s, size_t count) { memset(s, 0, count); barrier_data(s); } int main(void) { char buff[20]; memzero_explicit(buff, sizeof(buff)); return 0; } $ gcc -O2 test.c $ gdb a.out (gdb) disassemble main Dump of assembler code for function main: 0x0000000000400400 <+0>: lea -0x28(%rsp),%rax 0x0000000000400405 <+5>: movq $0x0,-0x28(%rsp) 0x000000000040040e <+14>: movq $0x0,-0x20(%rsp) 0x0000000000400417 <+23>: movl $0x0,-0x18(%rsp) 0x000000000040041f <+31>: xor %eax,%eax 0x0000000000400421 <+33>: retq End of assembler dump. $ clang -O2 test.c $ gdb a.out (gdb) disassemble main Dump of assembler code for function main: 0x00000000004004f0 <+0>: xorps %xmm0,%xmm0 0x00000000004004f3 <+3>: movaps %xmm0,-0x18(%rsp) 0x00000000004004f8 <+8>: movl $0x0,-0x8(%rsp) 0x0000000000400500 <+16>: lea -0x18(%rsp),%rax 0x0000000000400505 <+21>: xor %eax,%eax 0x0000000000400507 <+23>: retq End of assembler dump. As gcc, clang, but also icc defines __GNUC__, it's sufficient to define this in compiler-gcc.h only to be picked up. For a fallback or otherwise unsupported compiler, we define it as a barrier. Similarly, for ecc which does not support gcc inline asm. Reference: https://llvm.org/bugs/show_bug.cgi?id=15495 Reported-by: Stephan Mueller Tested-by: Stephan Mueller Signed-off-by: Daniel Borkmann Cc: Theodore Ts'o Cc: Stephan Mueller Cc: Hannes Frederic Sowa Cc: mancha security Cc: Mark Charlebois Cc: Behan Webster Signed-off-by: Herbert Xu --- include/linux/compiler-gcc.h | 16 +++++++++++++++- include/linux/compiler-intel.h | 3 +++ include/linux/compiler.h | 4 ++++ lib/string.c | 2 +- 4 files changed, 23 insertions(+), 2 deletions(-) (limited to 'lib/string.c') diff --git a/include/linux/compiler-gcc.h b/include/linux/compiler-gcc.h index cdf13ca7cac3..371e560d13cf 100644 --- a/include/linux/compiler-gcc.h +++ b/include/linux/compiler-gcc.h @@ -9,10 +9,24 @@ + __GNUC_MINOR__ * 100 \ + __GNUC_PATCHLEVEL__) - /* Optimization barrier */ + /* The "volatile" is due to gcc bugs */ #define barrier() __asm__ __volatile__("": : :"memory") +/* + * This version is i.e. to prevent dead stores elimination on @ptr + * where gcc and llvm may behave differently when otherwise using + * normal barrier(): while gcc behavior gets along with a normal + * barrier(), llvm needs an explicit input variable to be assumed + * clobbered. The issue is as follows: while the inline asm might + * access any memory it wants, the compiler could have fit all of + * @ptr into memory registers instead, and since @ptr never escaped + * from that, it proofed that the inline asm wasn't touching any of + * it. This version works well with both compilers, i.e. we're telling + * the compiler that the inline asm absolutely may see the contents + * of @ptr. See also: https://llvm.org/bugs/show_bug.cgi?id=15495 + */ +#define barrier_data(ptr) __asm__ __volatile__("": :"r"(ptr) :"memory") /* * This macro obfuscates arithmetic on a variable address so that gcc diff --git a/include/linux/compiler-intel.h b/include/linux/compiler-intel.h index ba147a1727e6..0c9a2f2c2802 100644 --- a/include/linux/compiler-intel.h +++ b/include/linux/compiler-intel.h @@ -13,9 +13,12 @@ /* Intel ECC compiler doesn't support gcc specific asm stmts. * It uses intrinsics to do the equivalent things. */ +#undef barrier_data #undef RELOC_HIDE #undef OPTIMIZER_HIDE_VAR +#define barrier_data(ptr) barrier() + #define RELOC_HIDE(ptr, off) \ ({ unsigned long __ptr; \ __ptr = (unsigned long) (ptr); \ diff --git a/include/linux/compiler.h b/include/linux/compiler.h index 0e41ca0e5927..867722591be2 100644 --- a/include/linux/compiler.h +++ b/include/linux/compiler.h @@ -169,6 +169,10 @@ void ftrace_likely_update(struct ftrace_branch_data *f, int val, int expect); # define barrier() __memory_barrier() #endif +#ifndef barrier_data +# define barrier_data(ptr) barrier() +#endif + /* Unreachable code */ #ifndef unreachable # define unreachable() do { } while (1) diff --git a/lib/string.c b/lib/string.c index a5792019193c..bb3d4b6993c4 100644 --- a/lib/string.c +++ b/lib/string.c @@ -607,7 +607,7 @@ EXPORT_SYMBOL(memset); void memzero_explicit(void *s, size_t count) { memset(s, 0, count); - barrier(); + barrier_data(s); } EXPORT_SYMBOL(memzero_explicit); -- cgit v1.2.3