diff options
author | Meir Shpilraien (Spielrein) <meir@redis.com> | 2022-09-29 08:58:58 +0300 |
---|---|---|
committer | Oran Agra <oran@redislabs.com> | 2022-12-12 17:36:34 +0200 |
commit | b1ede2120ae51d3288aa133348bfda94f8542d4c (patch) | |
tree | d6bfd084e8c8ba65a3f227475e36708f55c76ed7 | |
parent | 5b2119c6395b7eb7bb09f51961911377c610ed26 (diff) | |
download | redis-b1ede2120ae51d3288aa133348bfda94f8542d4c.tar.gz |
Avoid crash on crash report when a bad function pointer was called (#11298)
If Redis crashes due to calling an invalid function pointer,
the `backtrace` function will try to dereference this invalid pointer
which will cause a crash inside the crash report and will kill
the processes without having all the crash report information.
Example:
```
=== REDIS BUG REPORT START: Cut & paste starting from here ===
198672:M 19 Sep 2022 18:06:12.936 # Redis 255.255.255 crashed by signal: 11, si_code: 1
198672:M 19 Sep 2022 18:06:12.936 # Accessing address: 0x1
198672:M 19 Sep 2022 18:06:12.936 # Crashed running the instruction at: 0x1
// here the processes is crashing
```
This PR tries to fix this crash be:
1. Identify the issue when it happened.
2. Replace the invalid pointer with a pointer to some dummy function
so that `backtrace` will not crash.
I identification is done by comparing `eip` to `info->si_addr`, if they
are the same we know that the crash happened on the same address it tries to
accesses and we can conclude that it tries to call and invalid function pointer.
To replace the invalid pointer we introduce a new function, `setMcontextEip`,
which is very similar to `getMcontextEip` and it knows to set the Eip for the
different supported OS's. After printing the trace we retrieve the old `Eip` value.
(cherry picked from commit 0bf90d944313919eb8e63d3588bf63a367f020a3)
-rw-r--r-- | src/debug.c | 80 |
1 files changed, 58 insertions, 22 deletions
diff --git a/src/debug.c b/src/debug.c index 1500eb35b..e60f36bc1 100644 --- a/src/debug.c +++ b/src/debug.c @@ -1123,73 +1123,88 @@ void bugReportStart(void) { } #ifdef HAVE_BACKTRACE -static void *getMcontextEip(ucontext_t *uc) { + +/* Returns the current eip and set it to the given new value (if its not NULL) */ +static void* getAndSetMcontextEip(ucontext_t *uc, void *eip) { #define NOT_SUPPORTED() do {\ UNUSED(uc);\ + UNUSED(eip);\ return NULL;\ } while(0) +#define GET_SET_RETURN(target_var, new_val) do {\ + void *old_val = (void*)target_var; \ + if (new_val) { \ + void **temp = (void**)&target_var; \ + *temp = new_val; \ + } \ + return old_val; \ +} while(0) #if defined(__APPLE__) && !defined(MAC_OS_X_VERSION_10_6) /* OSX < 10.6 */ #if defined(__x86_64__) - return (void*) uc->uc_mcontext->__ss.__rip; + GET_SET_RETURN(uc->uc_mcontext->__ss.__rip, eip); #elif defined(__i386__) - return (void*) uc->uc_mcontext->__ss.__eip; + GET_SET_RETURN(uc->uc_mcontext->__ss.__eip, eip); #else - return (void*) uc->uc_mcontext->__ss.__srr0; + GET_SET_RETURN(uc->uc_mcontext->__ss.__srr0, eip); #endif #elif defined(__APPLE__) && defined(MAC_OS_X_VERSION_10_6) /* OSX >= 10.6 */ #if defined(_STRUCT_X86_THREAD_STATE64) && !defined(__i386__) - return (void*) uc->uc_mcontext->__ss.__rip; + GET_SET_RETURN(uc->uc_mcontext->__ss.__rip, eip); #elif defined(__i386__) - return (void*) uc->uc_mcontext->__ss.__eip; + GET_SET_RETURN(uc->uc_mcontext->__ss.__eip, eip); #else /* OSX ARM64 */ - return (void*) arm_thread_state64_get_pc(uc->uc_mcontext->__ss); + void *old_val = (void*)arm_thread_state64_get_pc(uc->uc_mcontext->__ss); + if (eip) { + arm_thread_state64_set_pc_fptr(uc->uc_mcontext->__ss, eip); + } + return old_val; #endif #elif defined(__linux__) /* Linux */ #if defined(__i386__) || ((defined(__X86_64__) || defined(__x86_64__)) && defined(__ILP32__)) - return (void*) uc->uc_mcontext.gregs[14]; /* Linux 32 */ + GET_SET_RETURN(uc->uc_mcontext.gregs[14], eip); #elif defined(__X86_64__) || defined(__x86_64__) - return (void*) uc->uc_mcontext.gregs[16]; /* Linux 64 */ + GET_SET_RETURN(uc->uc_mcontext.gregs[16], eip); #elif defined(__ia64__) /* Linux IA64 */ - return (void*) uc->uc_mcontext.sc_ip; + GET_SET_RETURN(uc->uc_mcontext.sc_ip, eip); #elif defined(__arm__) /* Linux ARM */ - return (void*) uc->uc_mcontext.arm_pc; + GET_SET_RETURN(uc->uc_mcontext.arm_pc, eip); #elif defined(__aarch64__) /* Linux AArch64 */ - return (void*) uc->uc_mcontext.pc; + GET_SET_RETURN(uc->uc_mcontext.pc, eip); #else NOT_SUPPORTED(); #endif #elif defined(__FreeBSD__) /* FreeBSD */ #if defined(__i386__) - return (void*) uc->uc_mcontext.mc_eip; + GET_SET_RETURN(uc->uc_mcontext.mc_eip, eip); #elif defined(__x86_64__) - return (void*) uc->uc_mcontext.mc_rip; + GET_SET_RETURN(uc->uc_mcontext.mc_rip, eip); #else NOT_SUPPORTED(); #endif #elif defined(__OpenBSD__) /* OpenBSD */ #if defined(__i386__) - return (void*) uc->sc_eip; + GET_SET_RETURN(uc->sc_eip, eip); #elif defined(__x86_64__) - return (void*) uc->sc_rip; + GET_SET_RETURN(uc->sc_rip, eip); #else NOT_SUPPORTED(); #endif #elif defined(__NetBSD__) #if defined(__i386__) - return (void*) uc->uc_mcontext.__gregs[_REG_EIP]; + GET_SET_RETURN(uc->uc_mcontext.__gregs[_REG_EIP], eip); #elif defined(__x86_64__) - return (void*) uc->uc_mcontext.__gregs[_REG_RIP]; + GET_SET_RETURN(uc->uc_mcontext.__gregs[_REG_RIP], eip); #else NOT_SUPPORTED(); #endif #elif defined(__DragonFly__) - return (void*) uc->uc_mcontext.mc_rip; + GET_SET_RETURN(uc->uc_mcontext.mc_rip, eip); #else NOT_SUPPORTED(); #endif @@ -1951,6 +1966,10 @@ void dumpCodeAroundEIP(void *eip) { } } +void invalidFunctionWasCalled() {} + +typedef void (*invalidFunctionWasCalledType)(); + void sigsegvHandler(int sig, siginfo_t *info, void *secret) { UNUSED(secret); UNUSED(info); @@ -1968,13 +1987,30 @@ void sigsegvHandler(int sig, siginfo_t *info, void *secret) { #ifdef HAVE_BACKTRACE ucontext_t *uc = (ucontext_t*) secret; - void *eip = getMcontextEip(uc); + void *eip = getAndSetMcontextEip(uc, NULL); if (eip != NULL) { serverLog(LL_WARNING, "Crashed running the instruction at: %p", eip); } - logStackTrace(getMcontextEip(uc), 1); + if (eip == info->si_addr) { + /* When eip matches the bad address, it's an indication that we crashed when calling a non-mapped + * function pointer. In that case the call to backtrace will crash trying to access that address and we + * won't get a crash report logged. Set it to a valid point to avoid that crash. */ + + /* This trick allow to avoid compiler warning */ + void *ptr; + invalidFunctionWasCalledType *ptr_ptr = (invalidFunctionWasCalledType*)&ptr; + *ptr_ptr = invalidFunctionWasCalled; + getAndSetMcontextEip(uc, ptr); + } + + logStackTrace(eip, 1); + + if (eip == info->si_addr) { + /* Restore old eip */ + getAndSetMcontextEip(uc, eip); + } logRegisters(uc); #endif @@ -2079,7 +2115,7 @@ void watchdogSignalHandler(int sig, siginfo_t *info, void *secret) { serverLogFromHandler(LL_WARNING,"\n--- WATCHDOG TIMER EXPIRED ---"); #ifdef HAVE_BACKTRACE - logStackTrace(getMcontextEip(uc), 1); + logStackTrace(getAndSetMcontextEip(uc, NULL), 1); #else serverLogFromHandler(LL_WARNING,"Sorry: no support for backtrace()."); #endif |