/* Page fault handling library. Copyright (C) 1993-2021 Bruno Haible Copyright (C) 2018 Nylon Chen 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 3 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 . */ #include /* Specification. */ #include "sigsegv.h" #include #include /* declares perror */ #include /* defines uintptr_t */ #include #include #if HAVE_GETRLIMIT # include /* declares struct rlimit */ #endif #ifdef __OpenBSD__ # include /* defines macro OpenBSD */ #endif /* Version number. */ int libsigsegv_version = LIBSIGSEGV_VERSION; /* ======================= Fault handler information ======================= */ /* Define: SIGSEGV_FAULT_HANDLER_ARGLIST is the argument list for the actual fault handler. and if available (optional): SIGSEGV_FAULT_ADDRESS is a macro for fetching the fault address. SIGSEGV_FAULT_CONTEXT is a macro giving a pointer to the entire fault context (i.e. the register set etc.). SIGSEGV_FAULT_STACKPOINTER is a macro for fetching the stackpointer at the moment the fault occurred. */ #if defined __linux__ || defined __ANDROID__ /* Linux */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO # if defined __alpha__ /* See glibc/sysdeps/unix/sysv/linux/alpha/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/alpha/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/alpha/sys/ucontext.h and the 'struct sigcontext' defined in are actually the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.sc_regs[30] # elif defined __arm64__ /* 64-bit */ /* See glibc/sysdeps/unix/sysv/linux/aarch64/sys/ucontext.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/aarch64/sys/ucontext.h and the 'struct sigcontext' defined in are actually the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.sp # elif defined __arm__ || defined __armhf__ /* 32-bit */ /* See glibc/sysdeps/unix/sysv/linux/arm/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/arm/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/arm/sys/ucontext.h and the 'struct sigcontext' defined in are actually the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.arm_sp # elif defined __cris__ /* See glibc-ports/sysdeps/unix/sysv/linux/cris/sys/ucontext.h. Note that the 'mcontext_t' defined in glibc-ports/sysdeps/unix/sysv/linux/cris/sys/ucontext.h and the 'struct sigcontext' defined in are actually the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.usp # elif defined __hppa__ /* See glibc/sysdeps/unix/sysv/linux/hppa/sys/ucontext.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/hppa/sys/ucontext.h and the 'struct sigcontext' defined in are actually the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.sc_gr[30] # elif defined __x86_64__ /* 64 bit registers */ /* See glibc/sysdeps/unix/sysv/linux/x86/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/x86_64/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/x86/sys/ucontext.h and the 'struct sigcontext' defined in glibc/sysdeps/unix/sysv/linux/x86/bits/sigcontext.h (see also ) are effectively the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[REG_RSP] # elif defined __i386__ /* 32 bit registers */ /* See glibc/sysdeps/unix/sysv/linux/x86/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/i386/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/x86/sys/ucontext.h and the 'struct sigcontext_ia32' defined in are effectively the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[REG_ESP] /* same value as ((ucontext_t *) ucp)->uc_mcontext.gregs[REG_UESP] */ # elif defined __ia64__ /* See glibc/sysdeps/unix/sysv/linux/ia64/sys/ucontext.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/ia64/sys/ucontext.h and the 'struct sigcontext' defined in glibc/sysdeps/unix/sysv/linux/ia64/bits/sigcontext.h (see also ) are actually the same. */ /* IA-64 has two stack pointers, one that grows down, called $r12, and one that grows up, called $bsp/$bspstore. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.sc_gr[12] /* It would be better to access $bspstore instead of $bsp but I don't know where to find it in 'struct sigcontext'. Anyway, it doesn't matter because $bsp and $bspstore never differ by more than ca. 1 KB. */ # define SIGSEGV_FAULT_BSP_POINTER ((ucontext_t *) ucp)->uc_mcontext.sc_ar_bsp # elif defined __m68k__ /* See glibc/sysdeps/unix/sysv/linux/m68k/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/m68k/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/m68k/sys/ucontext.h and the 'struct sigcontext' defined in are quite different types. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[R_SP] # elif defined __mips__ || defined __mipsn32__ || defined __mips64__ /* See glibc/sysdeps/unix/sysv/linux/mips/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/mips/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/mips/sys/ucontext.h and the 'struct sigcontext' defined in glibc/sysdeps/unix/sysv/linux/mips/bits/sigcontext.h (see also ) are effectively the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[29] # elif defined __nds32__ /* See glibc/sysdeps/unix/sysv/linux/nds32/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/nds32/sigcontextinfo.h. Both are found in part 08/11 . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.nds32_sp # elif defined __powerpc__ || defined __powerpc64__ || defined __powerpc64_elfv2__ /* See glibc/sysdeps/unix/sysv/linux/powerpc/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/powerpc/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/powerpc/sys/ucontext.h, the 'struct sigcontext' defined in , and the 'struct pt_regs' defined in are quite different types. */ # if defined __powerpc64__ || defined __powerpc64_elfv2__ /* 64-bit */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gp_regs[1] # else /* 32-bit */ /* both should be equivalent */ # if 0 # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.regs->gpr[1] # else # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.uc_regs->gregs[1] # endif # endif # elif defined __riscv32__ || __riscv64__ /* See glibc/sysdeps/unix/sysv/linux/riscv/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/riscv/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/riscv/sys/ucontext.h and the 'struct sigcontext' defined in glibc/sysdeps/unix/sysv/linux/riscv/bits/sigcontext.h start with the same block of 32 general-purpose registers. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.__gregs[REG_SP] # elif defined __s390__ || defined __s390x__ /* See glibc/sysdeps/unix/sysv/linux/s390/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/s390/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/s390/sys/ucontext.h and the '_sigregs' type, indirect part of 'struct sigcontext', defined in , are effectively the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[15] # elif defined __sh__ /* See glibc/sysdeps/unix/sysv/linux/sh/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/sh/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/sh/sys/ucontext.h and the 'struct sigcontext' defined in are effectively the same. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[15] # elif defined __sparc__ || defined __sparc64__ /* See glibc/sysdeps/unix/sysv/linux/sparc/sys/ucontext.h and the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux/sparc/{sparc32,sparc64}/sigcontextinfo.h. Note that the 'mcontext_t' defined in glibc/sysdeps/unix/sysv/linux/sparc/sys/ucontext.h and the 'struct sigcontext' defined in glibc/sysdeps/unix/sysv/linux/sparc/bits/sigcontext.h (see also ) are quite different types. */ # if defined __sparc64__/* 64-bit */ /* From linux-4.8.1/arch/sparc/kernel/signal_64.c, function setup_rt_frame, we see that ucp is not an 'ucontext_t *' but rather a 'struct sigcontext *' that happens to have the same value as sip (which is possible because a 'struct sigcontext' starts with 128 bytes room for the siginfo_t). */ # define SIGSEGV_FAULT_STACKPOINTER (((struct sigcontext *) ucp)->sigc_regs.u_regs[14] + 2047) # else /* 32-bit */ /* From linux-4.8.1/arch/sparc/kernel/signal_32.c, function setup_rt_frame, and linux-4.8.1/arch/sparc/kernel/signal32.c, function setup_rt_frame32, we see that ucp is a 'struct pt_regs *' or 'struct pt_regs32 *', respectively. In userland, this is a 'struct sigcontext *'. */ # define SIGSEGV_FAULT_STACKPOINTER ((struct sigcontext *) ucp)->si_regs.u_regs[14] # endif /* The sip->si_addr field is correct for a normal fault, but unusable in case of a stack overflow. What I observe (when running tests/test-sigsegv-catch-stackoverflow1, with a printf right at the beginning of sigsegv_handler) is that sip->si_addr is near 0: - in 64-bit mode: sip->si_addr = 0x000000000000030F, and gdb shows me that the fault occurs in an instruction 'stx %o3,[%fp+0x30f]' and %fp is 0. In fact, all registers %l0..%l7 and %i0..%i7 are 0. - in 32-bit mode: sip->si_addr = 0xFFFFFA64, and gdb shows me that the fault occurs in an instruction 'st %g2,[%fp-1436]' and %fp is 0. In fact, all registers %l0..%l7 and %i0..%i7 are 0. Apparently when the stack overflow occurred, some trap has tried to move the contents of the registers %l0..%l7 and %i0..%i7 (a "window" in SPARC terminology) to the stack, did not succeed in doing this, replaced all these register values with 0, and resumed execution at the fault location. This time, due to %fp = 0, a different fault was triggered. Now it is impossible to determine the real (previous) fault address because, even if know the faulting instruction, the previous register values have been lost. */ # define BOGUS_FAULT_ADDRESS_UPON_STACK_OVERFLOW # else /* When adding support for other CPUs here: */ /* For SIGSEGV_FAULT_HANDLER_ARGLIST, see the definition of SIGCONTEXT in glibc/sysdeps/unix/sysv/linux//sigcontextinfo.h. */ /* For SIGSEGV_FAULT_STACKPOINTER, see the definition of GET_STACK in glibc/sysdeps/unix/sysv/linux//sigcontextinfo.h. */ # endif #endif #if defined __GNU__ /* Hurd */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, int code, struct sigcontext *scp # define SIGSEGV_FAULT_ADDRESS (unsigned long) code # define SIGSEGV_FAULT_CONTEXT scp # if defined __i386__ /* scp points to a 'struct sigcontext' (defined in glibc/sysdeps/mach/hurd/i386/bits/sigcontext.h). The registers of this struct get pushed on the stack through gnumach/i386/i386/locore.S:trapall. */ /* Both sc_esp and sc_uesp appear to have the same value. It appears more reliable to use sc_uesp because it is labelled as "old esp, if trapped from user". */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_uesp # endif #endif #if defined __FreeBSD_kernel__ || defined __FreeBSD__ || defined __DragonFly__ /* GNU/kFreeBSD, FreeBSD */ # if defined __arm__ || defined __armhf__ || defined __arm64__ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # if defined __arm64__ /* 64-bit */ /* See sys/arm64/include/ucontext.h. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.mc_gpregs.gp_sp # elif defined __arm__ || defined __armhf__ /* 32-bit */ /* See sys/arm/include/ucontext.h. */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.__gregs[_REG_SP] # endif # else /* On FreeBSD 12, both of these approaches work. On FreeBSD derivatives, the first one has more chances to work. */ # if 1 /* Use signal handlers without SA_SIGINFO. */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, int code, struct sigcontext *scp, void *addr # define SIGSEGV_FAULT_ADDRESS addr # define SIGSEGV_FAULT_CONTEXT scp /* See sys/x86/include/signal.h. */ # if defined __x86_64__ /* 64 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_rsp # elif defined __i386__ /* 32 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_esp # endif # else /* Use signal handlers with SA_SIGINFO. */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *scp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((struct sigcontext *) scp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO /* See sys/x86/include/signal.h. */ # if defined __x86_64__ /* 64 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER ((struct sigcontext *) scp)->sc_rsp # elif defined __i386__ /* 32 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER ((struct sigcontext *) scp)->sc_esp # endif # endif # endif #endif #if defined __NetBSD__ /* NetBSD */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO /* _UC_MACHINE_SP is a platform independent macro. Defined in , see http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/arch/$arch/include/mcontext.h Supported on alpha, amd64, i386, ia64, m68k, mips, powerpc, sparc since NetBSD 2.0. On i386, _UC_MACHINE_SP is the same as ->uc_mcontext.__gregs[_REG_UESP], and apparently the same value as ->uc_mcontext.__gregs[_REG_ESP]. */ # ifdef _UC_MACHINE_SP # define SIGSEGV_FAULT_STACKPOINTER _UC_MACHINE_SP ((ucontext_t *) ucp) # endif #endif #if defined __OpenBSD__ /* OpenBSD */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, struct sigcontext *scp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT scp # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO # if defined __alpha__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/alpha/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_regs[30] # elif defined __arm__ || defined __armhf__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/arm/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_usr_sp # elif defined __hppa__ || defined __hppa64__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/hppa/include/signal.h and openbsd-src/sys/arch/hppa64/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_regs[30] # elif defined __x86_64__ /* 64 bit registers */ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/amd64/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_rsp # elif defined __i386__ /* 32 bit registers */ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/i386/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_esp # elif defined __m68k__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/m68k/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_sp # elif defined __m88k__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/m88k/include/signal.h and the definition of 'struct reg' in openbsd-src/sys/arch/m88k/include/reg.h. */ # if OpenBSD >= 201211 /* OpenBSD version >= 5.2 */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_regs[31] # else # define SIGSEGV_FAULT_STACKPOINTER scp->sc_regs.r[31] # endif # elif defined __mips__ || defined __mipsn32__ || defined __mips64__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/mips64/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_regs[29] # elif defined __powerpc__ || defined __powerpc64__ /* See the definition of 'struct sigcontext' and 'struct trapframe' in openbsd-src/sys/arch/powerpc/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_frame.fixreg[1] # elif defined __sh__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/sh/include/signal.h and the definition of 'struct reg' in openbsd-src/sys/arch/sh/include/reg.h. */ # if OpenBSD >= 201211 /* OpenBSD version >= 5.2 */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_reg[20-15] # else # define SIGSEGV_FAULT_STACKPOINTER scp->sc_reg.r_r15 # endif # elif defined __sparc__ || defined __sparc64__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/sparc/include/signal.h and openbsd-src/sys/arch/sparc64/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_sp # elif defined __vax__ /* See the definition of 'struct sigcontext' in openbsd-src/sys/arch/vax/include/signal.h. */ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_sp # endif #endif #if (defined __APPLE__ && defined __MACH__) /* macOS */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO # if defined __x86_64__ /* See the definitions of - 'ucontext_t' and 'struct __darwin_ucontext' in , - 'struct __darwin_mcontext64' in , and - 'struct __darwin_x86_thread_state64' in . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext->__ss.__rsp # elif defined __i386__ /* See the definitions of - 'ucontext_t' and 'struct __darwin_ucontext' in , - 'struct __darwin_mcontext32' in , and - 'struct __darwin_i386_thread_state' in . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext->__ss.__esp # elif defined __arm64__ /* See the definitions of - 'ucontext_t' and 'struct __darwin_ucontext' in , - 'struct __darwin_mcontext64' in , and - 'struct __darwin_arm_thread_state64' in . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext->__ss.__sp # elif defined __powerpc__ /* See the definitions of - 'ucontext_t' and 'struct __darwin_ucontext' in , - 'struct __darwin_mcontext' in , and - 'struct __darwin_ppc_thread_state' in . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext->__ss.__r1 # endif #endif #if defined _AIX /* AIX */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO # if defined __powerpc__ || defined __powerpc64__ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.jmp_context.gpr[1] # endif #endif #if defined __sgi /* IRIX */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, int code, struct sigcontext *scp # define SIGSEGV_FAULT_ADDRESS (unsigned long) scp->sc_badvaddr # define SIGSEGV_FAULT_CONTEXT scp # if defined __mips__ || defined __mipsn32__ || defined __mips64__ # define SIGSEGV_FAULT_STACKPOINTER scp->sc_regs[29] # endif #endif #if defined __sun /* Solaris */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO # if defined __x86_64__ /* 64 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[REG_RSP] # elif defined __i386__ /* 32 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[ESP] # elif defined __sparc__ || defined __sparc64__ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.gregs[REG_O6] # if SOLARIS11 /* On Solaris 11.3/SPARC, both in 32-bit and 64-bit mode, when catching stack overflow, the fault address is correct the first time, but is zero or near zero the second time. 'truss tests/test-sigsegv-catch-stackoverflow1' shows it: In 32-bit mode: Incurred fault #6, FLTBOUNDS %pc = 0x000116E8 siginfo: SIGSEGV SEGV_MAPERR addr=0xFFB00000 Received signal #11, SIGSEGV [caught] siginfo: SIGSEGV SEGV_MAPERR addr=0xFFB00000 then Incurred fault #6, FLTBOUNDS %pc = 0x000116E8 siginfo: SIGSEGV SEGV_MAPERR addr=0x00000008 Received signal #11, SIGSEGV [caught] siginfo: SIGSEGV SEGV_MAPERR addr=0x00000008 In 64-bit mode: Incurred fault #6, FLTBOUNDS %pc = 0x100001C58 siginfo: SIGSEGV SEGV_MAPERR addr=0xFFFFFFFF7FF00000 Received signal #11, SIGSEGV [caught] siginfo: SIGSEGV SEGV_MAPERR addr=0xFFFFFFFF7FF00000 then Incurred fault #6, FLTBOUNDS %pc = 0x100001C58 siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000 Received signal #11, SIGSEGV [caught] siginfo: SIGSEGV SEGV_MAPERR addr=0x00000000 */ # define BOGUS_FAULT_ADDRESS_UPON_STACK_OVERFLOW # endif # endif #endif #if defined __CYGWIN__ /* Cygwin */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO /* See the definition of 'ucontext_t' in and of 'struct __mcontext' in . */ # if defined __x86_64__ /* 64 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.rsp # elif defined __i386__ /* 32 bit registers */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.esp # endif #endif #if defined __HAIKU__ /* Haiku */ # define SIGSEGV_FAULT_HANDLER_ARGLIST int sig, siginfo_t *sip, void *ucp # define SIGSEGV_FAULT_ADDRESS sip->si_addr # define SIGSEGV_FAULT_CONTEXT ((ucontext_t *) ucp) # define SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO # if defined __x86_64__ /* 64 bit registers */ /* See the definition of 'ucontext_t' in and of 'struct vregs' in . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.rsp # elif defined __i386__ /* 32 bit registers */ /* See the definition of 'ucontext_t' in and of 'struct vregs' in . */ # define SIGSEGV_FAULT_STACKPOINTER ((ucontext_t *) ucp)->uc_mcontext.esp # endif #endif /* ========================================================================== */ /* List of signals that are sent when an invalid virtual memory address is accessed, or when the stack overflows. */ #if defined __GNU__ \ || defined __FreeBSD_kernel__ || defined __FreeBSD__ || defined __DragonFly__ \ || defined __NetBSD__ || defined __OpenBSD__ \ || (defined __APPLE__ && defined __MACH__) # define SIGSEGV_FOR_ALL_SIGNALS(var,body) \ { int var; var = SIGSEGV; { body } var = SIGBUS; { body } } #else # define SIGSEGV_FOR_ALL_SIGNALS(var,body) \ { int var; var = SIGSEGV; { body } } #endif /* ========================================================================== */ /* Determine the virtual memory area of a given address. */ #include "stackvma.h" /* ========================================================================== */ /* On the average Unix platform, we define HAVE_SIGSEGV_RECOVERY if there is a fault-*.h include file which defines SIGSEGV_FAULT_HANDLER_ARGLIST and SIGSEGV_FAULT_ADDRESS. HAVE_STACK_OVERFLOW_RECOVERY if HAVE_SIGALTSTACK is set and at least two of the following are true: A) There is a fault-*.h include file which defines SIGSEGV_FAULT_HANDLER_ARGLIST and SIGSEGV_FAULT_ADDRESS. B) There is a fault-*.h include file which defines SIGSEGV_FAULT_HANDLER_ARGLIST and SIGSEGV_FAULT_STACKPOINTER. C) There is a stackvma-*.c, other than stackvma-none.c, which defines sigsegv_get_vma. Why? Obviously, to catch stack overflow, we need an alternate signal stack; this requires kernel support. But we also need to distinguish (with a reasonable confidence) a stack overflow from a regular SIGSEGV. If we have A) and B), we use the Heuristic AB: If the fault address is near the stack pointer, it's a stack overflow. If we have A) and C), we use the Heuristic AC: If the fault address is near and beyond the bottom of the stack's virtual memory area, it's a stack overflow. If we have B) and C), we use the Heuristic BC: If the stack pointer is near the bottom of the stack's virtual memory area, it's a stack overflow. This heuristic comes in two flavours: On OSes which let the stack's VMA grow continuously, we determine the bottom by use of getrlimit(). On OSes which preallocate the stack's VMA with its maximum size (like BeOS), we use the stack's VMA directly. */ #if HAVE_SIGSEGV_RECOVERY \ && !(defined SIGSEGV_FAULT_HANDLER_ARGLIST && defined SIGSEGV_FAULT_ADDRESS) # error "You need to define SIGSEGV_FAULT_HANDLER_ARGLIST and SIGSEGV_FAULT_ADDRESS before you can define HAVE_SIGSEGV_RECOVERY." #endif #if !HAVE_SIGSEGV_RECOVERY \ && (defined SIGSEGV_FAULT_HANDLER_ARGLIST && defined SIGSEGV_FAULT_ADDRESS) \ && !(defined __FreeBSD__ && (defined __sparc__ || defined __sparc64__)) # if __GNUC__ || (__clang_major__ >= 4) # warning "You can define HAVE_SIGSEGV_RECOVERY on this platform." # else # error "You can define HAVE_SIGSEGV_RECOVERY on this platform." # endif #endif #if HAVE_STACK_OVERFLOW_RECOVERY \ && !(defined SIGSEGV_FAULT_ADDRESS + defined SIGSEGV_FAULT_STACKPOINTER + HAVE_STACKVMA >= 2) # error "You need to define two of SIGSEGV_FAULT_ADDRESS, SIGSEGV_FAULT_STACKPOINTER, HAVE_STACKVMA, before you can define HAVE_STACK_OVERFLOW_RECOVERY." #endif #if !HAVE_STACK_OVERFLOW_RECOVERY \ && (defined SIGSEGV_FAULT_ADDRESS + defined SIGSEGV_FAULT_STACKPOINTER + HAVE_STACKVMA >= 2) \ && !(defined __FreeBSD__ && (defined __sparc__ || defined __sparc64__)) \ && !(defined __NetBSD__ && (defined __sparc__ || defined __sparc64__)) # if __GNUC__ || (__clang_major__ >= 4) # warning "You can define HAVE_STACK_OVERFLOW_RECOVERY on this platform." # else # error "You can define HAVE_STACK_OVERFLOW_RECOVERY on this platform." # endif #endif /* ========================================================================== */ #if HAVE_STACK_OVERFLOW_RECOVERY /* ======= Leaving a signal handler executing on the alternate stack ======= */ /* Platform dependent: Leaving a signal handler executing on the alternate stack. */ static void sigsegv_reset_onstack_flag (void); /* -------------------------- leave-sigaltstack.c -------------------------- */ # if defined __GNU__ \ || defined __FreeBSD_kernel__ || defined __FreeBSD__ || defined __DragonFly__ \ || defined __NetBSD__ || defined __OpenBSD__ static void sigsegv_reset_onstack_flag (void) { stack_t ss; if (sigaltstack (NULL, &ss) >= 0) { ss.ss_flags &= ~SS_ONSTACK; sigaltstack (&ss, NULL); } } /* --------------------------- leave-setcontext.c --------------------------- */ # elif defined __sgi || defined __sun /* IRIX, Solaris */ # include static void sigsegv_reset_onstack_flag (void) { ucontext_t uc; if (getcontext (&uc) >= 0) /* getcontext returns twice. We are interested in the returned context only the first time, i.e. when the SS_ONSTACK bit is set. */ if (uc.uc_stack.ss_flags & SS_ONSTACK) { uc.uc_stack.ss_flags &= ~SS_ONSTACK; /* Note that setcontext() does not refill uc. Therefore if setcontext() keeps SS_ONSTACK set in the kernel, either setcontext() will return -1 or getcontext() will return a second time, with the SS_ONSTACK bit being cleared. */ setcontext (&uc); } } /* ------------------------------ leave-nop.c ------------------------------ */ # else static void sigsegv_reset_onstack_flag (void) { /* Nothing to do. sigaltstack() simply looks at the stack pointer, therefore SS_ONSTACK is not sticky. */ } # endif /* ========================================================================== */ # if HAVE_STACKVMA /* Address of the last byte belonging to the stack vma. */ static uintptr_t stack_top = 0; /* Needs to be called once only. */ static void remember_stack_top (void *some_variable_on_stack) { struct vma_struct vma; if (sigsegv_get_vma ((uintptr_t) some_variable_on_stack, &vma) >= 0) stack_top = vma.end - 1; } # endif /* HAVE_STACKVMA */ static stackoverflow_handler_t stk_user_handler = (stackoverflow_handler_t)NULL; static uintptr_t stk_extra_stack; static size_t stk_extra_stack_size; #endif /* HAVE_STACK_OVERFLOW_RECOVERY */ #if HAVE_SIGSEGV_RECOVERY /* User's SIGSEGV handler. */ static sigsegv_handler_t user_handler = (sigsegv_handler_t)NULL; #endif /* HAVE_SIGSEGV_RECOVERY */ /* Our SIGSEGV handler, with OS dependent argument list. */ #if HAVE_SIGSEGV_RECOVERY static void sigsegv_handler (SIGSEGV_FAULT_HANDLER_ARGLIST) { void *address = (void *) (SIGSEGV_FAULT_ADDRESS); # if HAVE_STACK_OVERFLOW_RECOVERY # if !(HAVE_STACKVMA || defined SIGSEGV_FAULT_STACKPOINTER) #error "Insufficient heuristics for detecting a stack overflow. Either define CFG_STACKVMA and HAVE_STACKVMA correctly, or define SIGSEGV_FAULT_STACKPOINTER correctly, or undefine HAVE_STACK_OVERFLOW_RECOVERY!" # endif /* Call user's handler. */ if (user_handler && (*user_handler) (address, 0)) { /* Handler successful. */ } else { /* Handler declined responsibility. */ /* Did the user install a stack overflow handler? */ if (stk_user_handler) { /* See whether it was a stack overflow. If so, longjump away. */ # ifdef SIGSEGV_FAULT_STACKPOINTER uintptr_t old_sp = (uintptr_t) (SIGSEGV_FAULT_STACKPOINTER); # ifdef __ia64 uintptr_t old_bsp = (uintptr_t) (SIGSEGV_FAULT_BSP_POINTER); # endif # endif # if HAVE_STACKVMA /* Were we able to determine the stack top? */ if (stack_top) { /* Determine stack bounds. */ int saved_errno; struct vma_struct vma; int ret; saved_errno = errno; ret = sigsegv_get_vma (stack_top, &vma); errno = saved_errno; if (ret >= 0) { # ifndef BOGUS_FAULT_ADDRESS_UPON_STACK_OVERFLOW /* Heuristic AC: If the fault_address is nearer to the stack segment's [start,end] than to the previous segment, we consider it a stack overflow. In the case of IA-64, we know that the previous segment is the up-growing bsp segment, and either of the two stacks can overflow. */ uintptr_t addr = (uintptr_t) address; # ifdef __ia64 if (addr >= vma.prev_end && addr <= vma.end - 1) # else # if STACK_DIRECTION < 0 if (addr >= vma.start ? (addr <= vma.end - 1) : vma.is_near_this (addr, &vma)) # else if (addr <= vma.end - 1 ? (addr >= vma.start) : vma.is_near_this (addr, &vma)) # endif # endif { # else /* BOGUS_FAULT_ADDRESS_UPON_STACK_OVERFLOW */ # if HAVE_GETRLIMIT && defined RLIMIT_STACK /* Heuristic BC: If the stack size has reached its maximal size, and old_sp is near the low end, we consider it a stack overflow. */ struct rlimit rl; saved_errno = errno; ret = getrlimit (RLIMIT_STACK, &rl); errno = saved_errno; if (ret >= 0) { uintptr_t current_stack_size = vma.end - vma.start; uintptr_t max_stack_size = rl.rlim_cur; if (current_stack_size <= max_stack_size + 4096 && max_stack_size <= current_stack_size + 4096 # else { if (1 # endif # ifdef SIGSEGV_FAULT_STACKPOINTER /* Heuristic BC: If we know old_sp, and it is neither near the low end, nor in the alternate stack, then it's probably not a stack overflow. */ && ((old_sp >= stk_extra_stack && old_sp <= stk_extra_stack + stk_extra_stack_size) # if STACK_DIRECTION < 0 || (old_sp <= vma.start + 4096 && vma.start <= old_sp + 4096)) # else || (old_sp <= vma.end + 4096 && vma.end <= old_sp + 4096)) # endif # endif ) # endif /* BOGUS_FAULT_ADDRESS_UPON_STACK_OVERFLOW */ # else /* !HAVE_STACKVMA */ /* Heuristic AB: If the fault address is near the stack pointer, it's a stack overflow. */ uintptr_t addr = (uintptr_t) address; if ((addr <= old_sp + 4096 && old_sp <= addr + 4096) # ifdef __ia64 || (addr <= old_bsp + 4096 && old_bsp <= addr + 4096) # endif ) { { { # endif /* !HAVE_STACKVMA */ { # ifdef SIGSEGV_FAULT_STACKPOINTER int emergency = (old_sp >= stk_extra_stack && old_sp <= stk_extra_stack + stk_extra_stack_size); stackoverflow_context_t context = (SIGSEGV_FAULT_CONTEXT); # else int emergency = 0; stackoverflow_context_t context = (void *) 0; # endif /* Call user's handler. */ (*stk_user_handler) (emergency, context); } } } } } # endif /* HAVE_STACK_OVERFLOW_RECOVERY */ if (user_handler && (*user_handler) (address, 1)) { /* Handler successful. */ } else { /* Handler declined responsibility for real. */ /* Remove ourselves and dump core. */ SIGSEGV_FOR_ALL_SIGNALS (sig, signal (sig, SIG_DFL);) } # if HAVE_STACK_OVERFLOW_RECOVERY } # endif /* HAVE_STACK_OVERFLOW_RECOVERY */ } #elif HAVE_STACK_OVERFLOW_RECOVERY static void # ifdef SIGSEGV_FAULT_STACKPOINTER sigsegv_handler (SIGSEGV_FAULT_HANDLER_ARGLIST) # else sigsegv_handler (int sig) # endif { # if !((HAVE_GETRLIMIT && defined RLIMIT_STACK) || defined SIGSEGV_FAULT_STACKPOINTER) # error "Insufficient heuristics for detecting a stack overflow. Either define SIGSEGV_FAULT_STACKPOINTER correctly, or undefine HAVE_STACK_OVERFLOW_RECOVERY!" # endif /* Did the user install a handler? */ if (stk_user_handler) { /* See whether it was a stack overflow. If so, longjump away. */ # ifdef SIGSEGV_FAULT_STACKPOINTER uintptr_t old_sp = (uintptr_t) (SIGSEGV_FAULT_STACKPOINTER); # endif /* Were we able to determine the stack top? */ if (stack_top) { /* Determine stack bounds. */ int saved_errno; struct vma_struct vma; int ret; saved_errno = errno; ret = sigsegv_get_vma (stack_top, &vma); errno = saved_errno; if (ret >= 0) { # if HAVE_GETRLIMIT && defined RLIMIT_STACK /* Heuristic BC: If the stack size has reached its maximal size, and old_sp is near the low end, we consider it a stack overflow. */ struct rlimit rl; saved_errno = errno; ret = getrlimit (RLIMIT_STACK, &rl); errno = saved_errno; if (ret >= 0) { uintptr_t current_stack_size = vma.end - vma.start; uintptr_t max_stack_size = rl.rlim_cur; if (current_stack_size <= max_stack_size + 4096 && max_stack_size <= current_stack_size + 4096 # else { if (1 # endif # ifdef SIGSEGV_FAULT_STACKPOINTER /* Heuristic BC: If we know old_sp, and it is neither near the low end, nor in the alternate stack, then it's probably not a stack overflow. */ && ((old_sp >= stk_extra_stack && old_sp <= stk_extra_stack + stk_extra_stack_size) # if STACK_DIRECTION < 0 || (old_sp <= vma.start + 4096 && vma.start <= old_sp + 4096)) # else || (old_sp <= vma.end + 4096 && vma.end <= old_sp + 4096)) # endif # endif ) { # ifdef SIGSEGV_FAULT_STACKPOINTER int emergency = (old_sp >= stk_extra_stack && old_sp <= stk_extra_stack + stk_extra_stack_size); stackoverflow_context_t context = (SIGSEGV_FAULT_CONTEXT); # else int emergency = 0; stackoverflow_context_t context = (void *) 0; # endif /* Call user's handler. */ (*stk_user_handler)(emergency,context); } } } } } /* Remove ourselves and dump core. */ SIGSEGV_FOR_ALL_SIGNALS (sig, signal (sig, SIG_DFL);) } #endif #if HAVE_SIGSEGV_RECOVERY || HAVE_STACK_OVERFLOW_RECOVERY static void install_for (int sig) { struct sigaction action; # ifdef SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO action.sa_sigaction = &sigsegv_handler; # else action.sa_handler = (void (*) (int)) &sigsegv_handler; # endif /* Block most signals while SIGSEGV is being handled. */ /* Signals SIGKILL, SIGSTOP cannot be blocked. */ /* Signals SIGCONT, SIGTSTP, SIGTTIN, SIGTTOU are not blocked because dealing with these signals seems dangerous. */ /* Signals SIGILL, SIGABRT, SIGFPE, SIGSEGV, SIGTRAP, SIGIOT, SIGEMT, SIGBUS, SIGSYS, SIGSTKFLT are not blocked because these are synchronous signals, which may require immediate intervention, otherwise the process may starve. */ sigemptyset (&action.sa_mask); # ifdef SIGHUP sigaddset (&action.sa_mask,SIGHUP); # endif # ifdef SIGINT sigaddset (&action.sa_mask,SIGINT); # endif # ifdef SIGQUIT sigaddset (&action.sa_mask,SIGQUIT); # endif # ifdef SIGPIPE sigaddset (&action.sa_mask,SIGPIPE); # endif # ifdef SIGALRM sigaddset (&action.sa_mask,SIGALRM); # endif # ifdef SIGTERM sigaddset (&action.sa_mask,SIGTERM); # endif # ifdef SIGUSR1 sigaddset (&action.sa_mask,SIGUSR1); # endif # ifdef SIGUSR2 sigaddset (&action.sa_mask,SIGUSR2); # endif # ifdef SIGCHLD sigaddset (&action.sa_mask,SIGCHLD); # endif # ifdef SIGCLD sigaddset (&action.sa_mask,SIGCLD); # endif # ifdef SIGURG sigaddset (&action.sa_mask,SIGURG); # endif # ifdef SIGIO sigaddset (&action.sa_mask,SIGIO); # endif # ifdef SIGPOLL sigaddset (&action.sa_mask,SIGPOLL); # endif # ifdef SIGXCPU sigaddset (&action.sa_mask,SIGXCPU); # endif # ifdef SIGXFSZ sigaddset (&action.sa_mask,SIGXFSZ); # endif # ifdef SIGVTALRM sigaddset (&action.sa_mask,SIGVTALRM); # endif # ifdef SIGPROF sigaddset (&action.sa_mask,SIGPROF); # endif # ifdef SIGPWR sigaddset (&action.sa_mask,SIGPWR); # endif # ifdef SIGLOST sigaddset (&action.sa_mask,SIGLOST); # endif # ifdef SIGWINCH sigaddset (&action.sa_mask,SIGWINCH); # endif /* Note that sigaction() implicitly adds sig itself to action.sa_mask. */ /* Ask the OS to provide a structure siginfo_t to the handler. */ # ifdef SIGSEGV_FAULT_ADDRESS_FROM_SIGINFO action.sa_flags = SA_SIGINFO; # else action.sa_flags = 0; # endif # if HAVE_STACK_OVERFLOW_RECOVERY && HAVE_SIGALTSTACK /* not BeOS */ /* Work around Linux 2.2.5 bug: If SA_ONSTACK is specified but sigaltstack() has not been called, the kernel will busy loop, eating CPU time. So avoid setting SA_ONSTACK until the user has requested stack overflow handling. */ if (stk_user_handler) action.sa_flags |= SA_ONSTACK; # endif sigaction (sig, &action, (struct sigaction *) NULL); } #endif /* HAVE_SIGSEGV_RECOVERY || HAVE_STACK_OVERFLOW_RECOVERY */ int sigsegv_install_handler (sigsegv_handler_t handler) { #if HAVE_SIGSEGV_RECOVERY user_handler = handler; SIGSEGV_FOR_ALL_SIGNALS (sig, install_for (sig);) return 0; #else return -1; #endif } void sigsegv_deinstall_handler (void) { #if HAVE_SIGSEGV_RECOVERY user_handler = (sigsegv_handler_t)NULL; # if HAVE_STACK_OVERFLOW_RECOVERY if (!stk_user_handler) # endif { SIGSEGV_FOR_ALL_SIGNALS (sig, signal (sig, SIG_DFL);) } #endif } int sigsegv_leave_handler (void (*continuation) (void*, void*, void*), void* cont_arg1, void* cont_arg2, void* cont_arg3) { #if HAVE_STACK_OVERFLOW_RECOVERY /* * Reset the system's knowledge that we are executing on the alternate * stack. If we didn't do that, siglongjmp would be needed instead of * longjmp to leave the signal handler. */ sigsegv_reset_onstack_flag (); #endif (*continuation) (cont_arg1, cont_arg2, cont_arg3); return 1; } int stackoverflow_install_handler (stackoverflow_handler_t handler, void *extra_stack, size_t extra_stack_size) { #if HAVE_STACK_OVERFLOW_RECOVERY # if HAVE_STACKVMA if (!stack_top) { int dummy; remember_stack_top (&dummy); if (!stack_top) return -1; } # endif stk_user_handler = handler; stk_extra_stack = (uintptr_t) extra_stack; stk_extra_stack_size = extra_stack_size; { stack_t ss; # if SIGALTSTACK_SS_REVERSED ss.ss_sp = (char *) extra_stack + extra_stack_size - sizeof (void *); ss.ss_size = extra_stack_size - sizeof (void *); # else ss.ss_sp = extra_stack; ss.ss_size = extra_stack_size; # endif ss.ss_flags = 0; /* no SS_DISABLE */ if (sigaltstack (&ss, (stack_t*)0) < 0) return -1; } /* Install the signal handlers with SA_ONSTACK. */ SIGSEGV_FOR_ALL_SIGNALS (sig, install_for (sig);) return 0; #else return -1; #endif } void stackoverflow_deinstall_handler (void) { #if HAVE_STACK_OVERFLOW_RECOVERY stk_user_handler = (stackoverflow_handler_t) NULL; # if HAVE_SIGSEGV_RECOVERY if (user_handler) { /* Reinstall the signal handlers without SA_ONSTACK, to avoid Linux bug. */ SIGSEGV_FOR_ALL_SIGNALS (sig, install_for (sig);) } else # endif { SIGSEGV_FOR_ALL_SIGNALS (sig, signal (sig, SIG_DFL);) } { stack_t ss; ss.ss_flags = SS_DISABLE; if (sigaltstack (&ss, (stack_t *) 0) < 0) perror ("gnulib sigsegv (stackoverflow_deinstall_handler)"); } #endif }