head -r343670 on armv7 (cortexA7): /rescue/WHATEVER gets "Segmentation fault" in my context [backtraces shows handle_static_init as the context]

Mon Feb 4 00:02:44 UTC 2019

On 2019-Feb-3, at 13:11, Mark Millard <marklmi at yahoo.com> wrote:

> I reliably get results like:
> 
> # /rescue/echo
> Segmentation fault (core dumped)
> #
> 
> # /rescue/tar -xpf /usr/obj/clang-armv7-installworld-poud.tar -C /usr/obj/DESTDIRs/clang-armv7-installworld-poud
> Segmentation fault (core dumped)
> #
> 
> That last is what I first tried /rescue with. Using just "tar" worked fine.
> That in turn lead to trying other simple /rescue/WHATEVER commands.
> 
> 
> (I've not yet figured out how to get debug information for rescue.
> This limits the below information.)
> 
> Both fail at the same place:
> 
> # gdb /rescue/echo echo.15825.core 
> GNU gdb (GDB) 8.2 [GDB v8.2 for FreeBSD]
> . . .
> Reading symbols from /rescue/echo...(no debugging symbols found)...done.
> [New LWP 100179]
> Core was generated by `/rescue/echo'.
> Program terminated with signal SIGSEGV, Segmentation fault.
> #0  0x0067b3f0 in ?? ()
> (gdb) bt
> #0  0x0067b3f0 in ?? ()
> #1  0x0066fe28 in ?? ()
> #2  0x001d11f0 in ?? ()
> Backtrace stopped: previous frame identical to this frame (corrupt stack?)
> (gdb) disass
> No function contains program counter for selected frame.
> 
> 
> # gdb /rescue/tar tar.15819.core
> GNU gdb (GDB) 8.2 [GDB v8.2 for FreeBSD]
> Copyright (C) 2018 Free Software Foundation, Inc.
> . . .
> [New LWP 100179]
> Core was generated by `/rescue/tar -xpf /usr/obj/clang-armv7-installworld-poud.tar -C /usr/obj/DESTDIRs'.
> Program terminated with signal SIGSEGV, Segmentation fault.
> #0  0x0067b3f0 in ?? ()
> (gdb) bt
> #0  0x0067b3f0 in ?? ()
> #1  0x0066fe28 in ?? ()
> #2  0x001d11f0 in ?? ()
> Backtrace stopped: previous frame identical to this frame (corrupt stack?)
> 
> 
> The buildworld was a cross build amd64->armv7 using my historical way of building.
> amd64 itself and the amd64->aarch64 cross-build that I've also built and installed
> have rescue working fine. /usr/src/ is the same for all these builds (diff -r -x.svn
> finds no differences).
> 
> 
> # file /rescue/echo
> /rescue/echo: ELF 32-bit LSB executable, ARM, EABI5 version 1 (FreeBSD), statically linked, for FreeBSD 13.0 (1300010), FreeBSD-style, stripped
> 
> # uname -apKU
> FreeBSD OPiP2E 13.0-CURRENT FreeBSD 13.0-CURRENT #4 r343670M: Fri Feb  1 16:45:22 PST 2019     markmi at FBSDFSSD:/usr/obj/armv7_clang/arm.armv7/usr/src/arm.armv7/sys/GENERIC-NODBG  arm armv7 1300010 1300010
> 
> My GENERIC-NODBG includes the standard GENERIC one and forces various debug things to be
> disabled.
> 
> My builds do use -mcpu=cortex-a7 :
> 
> # more ~/src.configs/src.conf.armv7-clang-bootstrap.amd64-host 
> TO_TYPE=armv7
> #
> KERNCONF=GENERIC-NODBG
> TARGET=arm
> .if ${.MAKE.LEVEL} == 0
> TARGET_ARCH=${TO_TYPE}
> .export TARGET_ARCH
> .endif
> #
> #WITH_CROSS_COMPILER=
> WITH_SYSTEM_COMPILER=
> WITH_SYSTEM_LINKER=
> #
> #CPUTYPE=soft
> WITH_LIBCPLUSPLUS=
> #WITH_LLD_BOOTSTRAP=
> WITHOUT_BINUTILS_BOOTSTRAP=
> WITH_ELFTOOLCHAIN_BOOTSTRAP=
> #Disables avoiding bootstrap: WITHOUT_LLVM_TARGET_ALL=
> WITHOUT_LLVM_TARGET_AARCH64=
> WITH_LLVM_TARGET_ARM=
> WITHOUT_LLVM_TARGET_MIPS=
> WITHOUT_LLVM_TARGET_POWERPC=
> WITHOUT_LLVM_TARGET_SPARC=
> WITHOUT_LLVM_TARGET_X86=
> #WITH_CLANG_BOOTSTRAP=
> WITH_CLANG=
> WITH_CLANG_IS_CC=
> WITH_CLANG_FULL=
> WITH_CLANG_EXTRAS=
> WITH_LLD=
> WITH_LLD_IS_LD=
> WITHOUT_BINUTILS=
> #
> # Linking lldb fails for armv7
> WITHOUT_LLDB=
> #
> WITH_BOOT=
> WITHOUT_LIB32=
> #
> WITHOUT_GCC_BOOTSTRAP=
> WITHOUT_GCC=
> WITHOUT_GCC_IS_CC=
> WITHOUT_GNUCXX=
> #
> NO_WERROR=
> #WERROR=
> MALLOC_PRODUCTION=
> #
> WITH_REPRODUCIBLE_BUILD=
> WITH_DEBUG_FILES=
> #
> XCFLAGS+= -mcpu=cortex-a7
> XCXXFLAGS+= -mcpu=cortex-a7
> # There is no XCPPFLAGS but XCPP gets XCFLAGS content.
> 
> I use:
> 
> # more ~/sys_build_scripts.amd64-host/make_armv7_nodebug_clang_bootstrap-amd64-host.sh 
> kldload -n filemon && \
> script ~/sys_typescripts/typescript_make_armv7_nodebug_clang_bootstrap-amd64-host-$(date +%Y-%m-%d:%H:%M:%S) \
> env __MAKE_CONF="/root/src.configs/make.conf" SRCCONF="/dev/null" SRC_ENV_CONF="/root/src.configs/src.conf.armv7-clang-bootstrap.amd64-host" \
> WITH_META_MODE=yes \
> WORLD_FLAGS="${WORLD_FLAGS} UBLDR_LOADADDR=0x42000000" \
> MAKEOBJDIRPREFIX="/usr/obj/armv7_clang/arm.armv7" \
> make $*
> 
> 
> I do not know when I last tried /rescue/SOMETHING so I've no clue
> when this problem started. Plus the prior build was -r341864 or
> some such so this was a fairly large jump even if I used /rescue
> with my prior build/installation.

I got backtraces ( /rescue/tar example first ):

# gdb /rescue/tar tar.16186.core 
GNU gdb (GDB) 8.2 [GDB v8.2 for FreeBSD]
. . .
Reading symbols from /rescue/tar...done.
[New LWP 100201]
Core was generated by `/rescue/tar -xpf /usr/obj/clang-armv7-installworld-poud.tar -C /usr/obj/DESTDIRs'.
Program terminated with signal SIGSEGV, Segmentation fault.
#0  thr_malloc_lock (curthread=0x0) at /usr/src/lib/libthr/thread/thr_malloc.c:63
63              _thr_umutex_lock(&thr_malloc_umtx, TID(curthread));
(gdb) bt
#0  thr_malloc_lock (curthread=0x0) at /usr/src/lib/libthr/thread/thr_malloc.c:63
#1  __thr_calloc (num=1, size=<optimized out>) at /usr/src/lib/libthr/thread/thr_malloc.c:82
#2  0x0066ff38 in mutex_init (mutex=<optimized out>, mutex_attr=<optimized out>, calloc_cb=<optimized out>) at /usr/src/lib/libthr/thread/thr_mutex.c:294
#3  __pthread_mutex_init (mutex=0xc758c8 <atomic_mtx>, mutex_attr=<optimized out>) at /usr/src/lib/libthr/thread/thr_mutex.c:393
#4  0x001d11f0 in handle_static_init (argc=5, argv=<optimized out>, env=<optimized out>) at /usr/src/lib/csu/common/ignore_init.c:124
#5  0x001d10e8 in __start (argc=5, argv=<optimized out>, env=<optimized out>, ps_strings=<optimized out>, obj=0x0, cleanup=0x0) at /usr/src/lib/csu/arm/crt1.c:112
#6  0x001d1000 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

Or, looking at /rescue/echo :

# gdb /rescue/echo
GNU gdb (GDB) 8.2 [GDB v8.2 for FreeBSD]
. . .
Reading symbols from /rescue/echo...done.
(gdb) run
Starting program: /rescue/echo 

Program received signal SIGSEGV, Segmentation fault.
__thr_calloc (num=1, size=<optimized out>) at /usr/src/lib/libthr/thread/thr_malloc.c:82
82              thr_malloc_lock(curthread);
(gdb) bt
#0  __thr_calloc (num=1, size=<optimized out>) at /usr/src/lib/libthr/thread/thr_malloc.c:82
#1  0x0066ff38 in mutex_init (mutex=<optimized out>, mutex_attr=<optimized out>, calloc_cb=<optimized out>) at /usr/src/lib/libthr/thread/thr_mutex.c:294
#2  __pthread_mutex_init (mutex=0xc758c8 <atomic_mtx>, mutex_attr=<optimized out>) at /usr/src/lib/libthr/thread/thr_mutex.c:393
#3  0x001d11f0 in handle_static_init (argc=1, argv=<optimized out>, env=<optimized out>) at /usr/src/lib/csu/common/ignore_init.c:124
#4  0x001d10e8 in __start (argc=1, argv=<optimized out>, env=<optimized out>, ps_strings=<optimized out>, obj=0x0, cleanup=0x0) at /usr/src/lib/csu/arm/crt1.c:112
#5  0x001d1000 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)