calling C function perl_clone of libperl causes segfault in perl 5.22 and and 5.24 but fine on perl 5.20 #15917
Comments
From mitch.capper@gmail.comFor using perl in C++ programs unfortunately a simple call to perl_clone ============Backtrace============== ===========Repro Details=========== You can use these commands in a container like: ============= For 5.24 ============ apt-get update && apt-get install -y lbzip2 zlibc libbz2-dev wget ============= For 5.20 ============ wget ============= perlbug -d for 5.20.0 ============ Flags: Site configuration information for perl 5.20.0: Configured by root at Tue Mar 7 18:44:42 UTC 2017. Summary of my perl5 (revision 5 version 20 subversion 0) configuration: Platform: @INC for perl 5.20.0: Environment for perl 5.20.0: Flags: Site configuration information for perl 5.24.1: Configured by root at Tue Mar 7 19:37:15 UTC 2017. Summary of my perl5 (revision 5 version 24 subversion 1) configuration: Platform: @INC for perl 5.24.1: Environment for perl 5.24.1: --------------1.40.perlbug #include <EXTERN.h> void main(){ // apt-get update && apt-get install build-essential libperl-dev gdb --------------1.40.perlbug-- ~mitch |
From mitch.capper@gmail.com#include <EXTERN.h>
#include <perl.h>
void main(){
PerlInterpreter *my_perl;
if (!(my_perl = perl_alloc())) {
printf("Could not allocate perl interpreter\n");
}
perl_construct(my_perl);
PerlInterpreter *perl2 = perl_clone(my_perl, CLONEf_COPY_STACKS | CLONEf_KEEP_PTR_TABLE);
printf("ALL FINE\n");
}
// apt-get update && apt-get install build-essential libperl-dev gdb
// zesty: gcc -g test.c -I/usr/lib/x86_64-linux-gnu/perl/5.24.1/CORE/ -lperl && ./a.out
// xenial/yakkety (yakkety might be 5.22.2): gcc -g test.c -I/usr/lib/x86_64-linux-gnu/perl/5.22.1/CORE/ -lperl && ./a.out
// trusty: gcc -g test.c -I/usr/lib/perl/5.18.2/CORE/ -lperl && ./a.out |
From @iabynOn Tue, Mar 07, 2017 at 11:51:38AM -0800, Mitch Capper wrote:
Your test program allocates and constructs a perl interpreter, but *doesn't* Some of the stuff that perl_clone() expects to be in place only happens Do you have a persuasive use case for calling perl_clone() on an "empty" For porters: I reproduced this by building blead with threads and $ cc -o test -I . -DPERL_CORE -D_REENTRANT -D_GNU_SOURCE -DDEBUGGING -ggdb -fwrapv -fno-strict-aliasing -pipe -fstack-protector-strong -I/usr/local/include -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -std=c89 -g -Wall -Werror=declaration-after-statement -Wextra -Wc++-compat -Wwrite-strings -L. -lperl ~/tmp/test.c $ LD_LIBRARY_PATH=. ./test where ~/tmp/test.c looks like the following. Uncommenting the perl_parse() #include <EXTERN.h> int main(int argc, char **argv){ /* perl2 = perl_clone(my_perl, CLONEf_COPY_STACKS | CLONEf_KEEP_PTR_TABLE); Perl_croak_nocontext("ALL FINE\n"); -- |
|
The RT System itself - Status changed from 'new' to 'open' |
From mitch.capper+bit@gmail.comI apologize I have oversimplified the use case here from the FreeSWITCH code to try and make the test case more concise. Their use of mod_perl (https://freeswitch.org/stash/projects/FS/repos/freeswitch/browse/src/mod/languages/mod_perl/mod_perl.c) does immediately call perl_parse after perl_construct so this must be trigged by something else. It also only happens on 5.22 and newer. I will work to update the test case with a better implementation of how FS is calling it that still repros shortly. In the mean time, if obvious, here is the BT from FreeSWITCH when it crashes: Thread 34 "freeswitch" received signal SIGSEGV, Segmentation fault. |
From mitch.capper+bit@gmail.comOk tracked down the problem if an empty string is past after -e it segfaults: #include <EXTERN.h> int main(int argc, char **argv){ PL_exit_flags |= PERL_EXIT_DESTRUCT_END; perl2 = perl_clone(my_perl, CLONEf_COPY_STACKS | CLONEf_KEEP_PTR_TABLE); This makes some sense given -e's meaning we should be able to easily work around for FS but not sure if the change of behavior is expected or not. |
From @tonycozOn Wed, Mar 22, 2017 at 10:50:43AM -0700, Mitch Capper via RT wrote:
The argv array requires a NULL at the end: static char *embedding[] = { "", "-e", "", NULL }; though I don't know if this is the cause of the failure you're seeing. Tony |
From mitch.capper+bit@gmail.comOk so technically still segfaults, but does spit out an error: if you do: #include <EXTERN.h> int main(int argc, char **argv){ PL_exit_flags |= PERL_EXIT_DESTRUCT_END; perl2 = perl_clone(my_perl, CLONEf_COPY_STACKS | CLONEf_KEEP_PTR_TABLE); |
From mitch.capper+bit@gmail.comSorry and to clarify similarly: |
From @iabynOn Wed, Mar 22, 2017 at 05:45:37PM -0700, Mitch Capper via RT wrote:
Yes, but you're now calling perl_clone after a parse error. Strictly Again, show me a valid use case for calling perl_clone() in such -- |
From mitch.capper+bit@gmail.comNo problem, we have a work-around for our use and the function changes it seems are a non issue so this can be closed. |
|
@iabyn - Status changed from 'open' to 'rejected' |
Migrated from rt.perl.org#130949 (status was 'rejected')
Searchable as RT130949$
The text was updated successfully, but these errors were encountered: