Dear Community,

> ---
>  resolv/Makefile          |  2 ++
>  resolv/tst-gai_suspend.c | 53
> ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 55
> insertions(+) create mode 100644 resolv/tst-gai_suspend.c
> 
> diff --git a/resolv/Makefile b/resolv/Makefile
> index 1047bb6ae5..e7e66fe903 100644
> --- a/resolv/Makefile
> +++ b/resolv/Makefile
> @@ -60,6 +60,7 @@ tests += \
>    tst-resolv-res_init-multi \
>    tst-resolv-search \
>    tst-resolv-trailing \
> +  tst-gai_suspend \
>  
>  # This test calls __res_context_send directly, which is not exported
>  # from libresolv.
> @@ -209,3 +210,4 @@ $(objpfx)tst-ns_name_compress:
> $(objpfx)libresolv.so $(objpfx)tst-ns_name_pton: $(objpfx)libresolv.so
>  $(objpfx)tst-res_hnok: $(objpfx)libresolv.so
>  $(objpfx)tst-p_secstodate: $(objpfx)libresolv.so
> +$(objpfx)tst-gai_suspend: $(objpfx)libanl.so
> diff --git a/resolv/tst-gai_suspend.c b/resolv/tst-gai_suspend.c
> new file mode 100644
> index 0000000000..bbafd13fc4
> --- /dev/null
> +++ b/resolv/tst-gai_suspend.c
> @@ -0,0 +1,53 @@
> +/* Test for gai_suspend timeout
> +   Copyright (C) 2021 Free Software Foundation, Inc.
> +   This file is part of the GNU C Library.
> +
> +   The GNU C Library is free software; you can redistribute it and/or
> +   modify it under the terms of the GNU Lesser General Public
> +   License as published by the Free Software Foundation; either
> +   version 2.1 of the License, or (at your option) any later version.
> +
> +   The GNU C Library 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
> +   Lesser General Public License for more details.
> +
> +   You should have received a copy of the GNU Lesser General Public
> +   License along with the GNU C Library; if not, see
> +   <https://www.gnu.org/licenses/>.  */
> +
> +#include <time.h>
> +#include <netdb.h>
> +#include <errno.h>
> +#include <string.h>
> +#include <support/check.h>
> +#include <support/xtime.h>
> +#include <support/timespec.h>
> +
> +static int
> +do_test (void)
> +{
> +  struct timespec ts;
> +  xclock_gettime (CLOCK_REALTIME, &ts);
> +  struct timespec timeout = make_timespec (1, 0);
> +  ts = timespec_add (ts, timeout);
> +
> +  /* Ignore gaicb content - just wait for timeout.  */
> +  struct addrinfo result;
> +
> +  struct gaicb gai = { "foo.baz", "0", NULL, &result };
> +  const struct gaicb * const gai_s[] = { &gai };
> +  struct gaicb *gai_p = &gai;
> +
> +  int ret = getaddrinfo_a (GAI_NOWAIT, &gai_p, 1, NULL);
> +  if (ret != 0)
> +    FAIL_EXIT1 ("getaddrinfo_a failed: %m\n");
> +
> +  ret = gai_suspend (gai_s, 1, &timeout);
> +  TEST_COMPARE (ret, EAI_AGAIN);
> +  TEST_TIMESPEC_NOW_OR_AFTER (CLOCK_REALTIME, ts);
> +
> +  return 0;
> +}
> +
> +#include <support/test-driver.c>

I'm a bit puzzled with the above gai_suspend() test.


The idea for Y2038 test is to check if one can timeout when passed
crafted struct gaicb causes the gai_suspend() to exit with timeout.
This approach is similar to one from tst-ppoll.c


1. When I do run it by hand on the remove machine (from glibc test
suite):
--------------------------------------------------

rm -rf ./resolv/*
make PARALLELMFLAGS="-j8"
test-wrapper='/home/lukma/work/yocto/y2038/build/workspace/sources/y2038-glibc/scripts/cross-test-ssh.sh
--allow-time-setting root@192.168.7.2' subdirs=resolv check

I do see the following error:
-----------------------------

cat ./resolv/tst-gai_suspend.out
Didn't expect signal from child: got `Segmentation fault'

This is strange, as I've passed NULL to getaddrinfo_a as the last
parameter, which causes passing no signals to program running it.

What am I missing here? Has the test setup any strict policy to handle
LWPs ?


2. On the VM (armv7):

gai_suspend.c test program:
---------------------------

#define _GNU_SOURCE

#include <stdio.h>
#include <time.h>
#include <netdb.h>

int main (int argc, char **argv)
{
        struct timespec ts, timeout, now;

        clock_gettime (CLOCK_REALTIME, &ts);

        timeout.tv_sec = ts.tv_sec + 3;
        timeout.tv_nsec = ts.tv_nsec;

        struct addrinfo result;
        struct gaicb gai = { "foo.baz", "0", NULL, &result };
        const struct gaicb * const gai_s[] = { &gai };
        struct gaicb *gai_p = &gai;

        int ret = getaddrinfo_a (GAI_NOWAIT, &gai_p, 1, NULL);
        if (ret != 0)
          printf("getaddrinfo_a error");

        ret = gai_suspend (gai_s, 1, &timeout);
        // According to man getaddrinfo_s I shall receive ret == -3
        // when conversion from name to IP credentials is performed
        if (ret != 0)
          printf("gai_suspend error");

        clock_gettime (CLOCK_REALTIME, &now);

        printf("gai_suspend seconds diff: %llu\n", now.tv_sec -
        ts.tv_sec);

        return 0;
}


gcc -Wall -ggdb -D_TIME_BITS=64 -D_FILE_OFFSET_BITS=64 -I/opt/include \
    -I/opt/usr/include -L/opt/usr/lib -Wl,-rpath=/opt/lib \
    -Wl,--dynamic-linker=/opt/lib/ld-linux.so.2  gai_suspend.c -o gs
-lanl \ -lpthread -static

The above line has some manual adjustments for linker, but this works
for other syscalls for some long time now - more info [1].


In this case the program quits without any issues with threads, but the
output and behavior of gai_suspend is not as expected (according at
least to the manual [2]). The return code from gai_suspend shall be
EAI_AGAIN and the result struct has some garbage in it.


The question(s):
================

1. Why I do experience thread error in glibc test suite?

2. Is the code for gai_suspend testing wrong (not much reference in
glibc), or do we have a bug in the the implementation?


Thanks in advance for any help/hints.



Links:
[1] - https://github.com/lmajewski/meta-y2038/
[2] - https://linux.die.net/man/3/gai_suspend

Best regards,

Lukasz Majewski

--

DENX Software Engineering GmbH,      Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma@denx.de