* [RFC PATCH] tst: Add test for gai_suspend
@ 2021-03-12 15:49 Lukasz Majewski
2021-03-12 16:09 ` Lukasz Majewski
0 siblings, 1 reply; 7+ messages in thread
From: Lukasz Majewski @ 2021-03-12 15:49 UTC (permalink / raw)
To: Joseph Myers, Adhemerval Zanella, Florian Weimer, DJ Delorie
Cc: Paul Eggert, Alistair Francis, Arnd Bergmann, Alistair Francis,
GNU C Library, Carlos O'Donell, Florian Weimer,
Zack Weinberg, libc-help, Lukasz Majewski
---
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>
--
2.20.1
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC PATCH] tst: Add test for gai_suspend
2021-03-12 15:49 [RFC PATCH] tst: Add test for gai_suspend Lukasz Majewski
@ 2021-03-12 16:09 ` Lukasz Majewski
2021-03-12 16:41 ` Andreas Schwab
2021-03-23 13:09 ` Adhemerval Zanella
0 siblings, 2 replies; 7+ messages in thread
From: Lukasz Majewski @ 2021-03-12 16:09 UTC (permalink / raw)
To: Joseph Myers, Adhemerval Zanella, Florian Weimer, DJ Delorie
Cc: Paul Eggert, Alistair Francis, Arnd Bergmann, Alistair Francis,
GNU C Library, Carlos O'Donell, Florian Weimer,
Zack Weinberg, libc-help
[-- Attachment #1: Type: text/plain, Size: 6335 bytes --]
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
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC PATCH] tst: Add test for gai_suspend
2021-03-12 16:09 ` Lukasz Majewski
@ 2021-03-12 16:41 ` Andreas Schwab
2021-03-23 13:09 ` Adhemerval Zanella
1 sibling, 0 replies; 7+ messages in thread
From: Andreas Schwab @ 2021-03-12 16:41 UTC (permalink / raw)
To: Lukasz Majewski
Cc: Joseph Myers, Adhemerval Zanella, Florian Weimer, DJ Delorie,
GNU C Library, libc-help, Florian Weimer, Alistair Francis
On Mär 12 2021, Lukasz Majewski wrote:
> 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'
Where does it crash?
Andreas.
--
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510 2552 DF73 E780 A9DA AEC1
"And now for something completely different."
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC PATCH] tst: Add test for gai_suspend
2021-03-12 16:09 ` Lukasz Majewski
2021-03-12 16:41 ` Andreas Schwab
@ 2021-03-23 13:09 ` Adhemerval Zanella
2021-03-23 15:23 ` Lukasz Majewski
2021-03-26 11:30 ` Lukasz Majewski
1 sibling, 2 replies; 7+ messages in thread
From: Adhemerval Zanella @ 2021-03-23 13:09 UTC (permalink / raw)
To: Lukasz Majewski; +Cc: GNU C Library
On 12/03/2021 13:09, Lukasz Majewski wrote:
> 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
I am not sure from where the failure come from, on my environment I am
not seeing the SEGFAULT failure on neither x86_64 nor on i686.
But I am seeing:
$ ./testrun.sh resolv/tst-gai_suspend
tst-gai_suspend.c:47: numeric comparison failure
left: 0 (0x0); from: ret
right: -3 (0xfffffffd); from: EAI_AGAIN
tst-gai_suspend.c:48: 1616504556.995728406s not after
1616504557.971222237s (difference 0.975493831s)
error: 2 test failures
The issue is tests is using the system NSS and possible DNS cache so the
expected name resolution timeout does not happen reliable. I am not
sure how we can accomplish it, maybe through running it on a container
with a custom NSS module to actually emulate the delay.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC PATCH] tst: Add test for gai_suspend
2021-03-23 13:09 ` Adhemerval Zanella
@ 2021-03-23 15:23 ` Lukasz Majewski
2021-03-26 11:30 ` Lukasz Majewski
1 sibling, 0 replies; 7+ messages in thread
From: Lukasz Majewski @ 2021-03-23 15:23 UTC (permalink / raw)
To: Adhemerval Zanella; +Cc: GNU C Library
[-- Attachment #1: Type: text/plain, Size: 8360 bytes --]
Hi Adhemerval,
> On 12/03/2021 13:09, Lukasz Majewski wrote:
> > 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
>
> I am not sure from where the failure come from, on my environment I am
> not seeing the SEGFAULT failure on neither x86_64 nor on i686.
>
> But I am seeing:
>
> $ ./testrun.sh resolv/tst-gai_suspend
> tst-gai_suspend.c:47: numeric comparison failure
> left: 0 (0x0); from: ret
> right: -3 (0xfffffffd); from: EAI_AGAIN
> tst-gai_suspend.c:48: 1616504556.995728406s not after
> 1616504557.971222237s (difference 0.975493831s)
> error: 2 test failures
>
> The issue is tests is using the system NSS and possible DNS cache so
> the expected name resolution timeout does not happen reliable. I am
> not sure how we can accomplish it, maybe through running it on a
> container with a custom NSS module to actually emulate the delay.
I thought that calling:
int ret = getaddrinfo_a (GAI_NOWAIT, &gai_p, 1, NULL);
ret = gai_suspend (gai_s, 1, &timeout);
shall just timeout as struct gaicb is providing data which shall be not
resolved. As a result we will have timeout (as we do in e.g. select or
ppoll).
Anyway, I'm wondering how this can be easily tested? I mean crafting
container with custom NSS to check timeout seems not so easy.
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
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC PATCH] tst: Add test for gai_suspend
2021-03-23 13:09 ` Adhemerval Zanella
2021-03-23 15:23 ` Lukasz Majewski
@ 2021-03-26 11:30 ` Lukasz Majewski
2021-05-06 9:11 ` Lukasz Majewski
1 sibling, 1 reply; 7+ messages in thread
From: Lukasz Majewski @ 2021-03-26 11:30 UTC (permalink / raw)
To: Adhemerval Zanella; +Cc: GNU C Library
[-- Attachment #1: Type: text/plain, Size: 8871 bytes --]
Hi Adhemerval,
> On 12/03/2021 13:09, Lukasz Majewski wrote:
> > 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
>
> I am not sure from where the failure come from, on my environment I am
> not seeing the SEGFAULT failure on neither x86_64 nor on i686.
>
> But I am seeing:
>
> $ ./testrun.sh resolv/tst-gai_suspend
> tst-gai_suspend.c:47: numeric comparison failure
> left: 0 (0x0); from: ret
> right: -3 (0xfffffffd); from: EAI_AGAIN
> tst-gai_suspend.c:48: 1616504556.995728406s not after
> 1616504557.971222237s (difference 0.975493831s)
> error: 2 test failures
>
> The issue is tests is using the system NSS and possible DNS cache so
> the expected name resolution timeout does not happen reliable.
Even more strange (tested on x86_64):
1. For "sourceware.org" + "http"
struct gaicb gai = { "sourceware.org", "http", NULL, &result };
getaddrinfo_a (GAI_NOWAIT, &gai_p, 1, NULL);
gai_suspend (gai_s, 1, &timeout);
ret = gai_error(gai_p);
Here the ret is 0 as expected.
2. For "foo.baz" + "0" (the gai_suspend and getaddrinfo_a are called in
the same way as above)
struct gaicb gai = { "foo.baz", "0", NULL, &result };
Here the ret is -2 as expected (EAI_NONAME - NAME or SERVICE unknown) -
the "0" here is wrong.
3. For "foo.baz" + "http" (the gai_suspend and getaddrinfo_a are called
in the same way as above)
struct gaicb gai = { "foo.baz", "http", NULL, &result };
Here the ret is also -2, which I don't understand. I would expect that
the foo.baz name is going to be resolved as it "looks" correctly and
hence the timeout will occur.
Any hints on how timeout can be forced?
> I am
> not sure how we can accomplish it, maybe through running it on a
> container with a custom NSS module to actually emulate the delay.
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
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC PATCH] tst: Add test for gai_suspend
2021-03-26 11:30 ` Lukasz Majewski
@ 2021-05-06 9:11 ` Lukasz Majewski
0 siblings, 0 replies; 7+ messages in thread
From: Lukasz Majewski @ 2021-05-06 9:11 UTC (permalink / raw)
To: Adhemerval Zanella; +Cc: GNU C Library
[-- Attachment #1: Type: text/plain, Size: 9789 bytes --]
On Fri, 26 Mar 2021 12:30:43 +0100
Lukasz Majewski <lukma@denx.de> wrote:
> Hi Adhemerval,
>
> > On 12/03/2021 13:09, Lukasz Majewski wrote:
> > > 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
> >
> > I am not sure from where the failure come from, on my environment I
> > am not seeing the SEGFAULT failure on neither x86_64 nor on i686.
> >
> > But I am seeing:
> >
> > $ ./testrun.sh resolv/tst-gai_suspend
> > tst-gai_suspend.c:47: numeric comparison failure
> > left: 0 (0x0); from: ret
> > right: -3 (0xfffffffd); from: EAI_AGAIN
> > tst-gai_suspend.c:48: 1616504556.995728406s not after
> > 1616504557.971222237s (difference 0.975493831s)
> > error: 2 test failures
> >
> > The issue is tests is using the system NSS and possible DNS cache so
> > the expected name resolution timeout does not happen reliable.
>
> Even more strange (tested on x86_64):
>
> 1. For "sourceware.org" + "http"
> struct gaicb gai = { "sourceware.org", "http", NULL, &result };
>
> getaddrinfo_a (GAI_NOWAIT, &gai_p, 1, NULL);
> gai_suspend (gai_s, 1, &timeout);
>
> ret = gai_error(gai_p);
>
> Here the ret is 0 as expected.
>
> 2. For "foo.baz" + "0" (the gai_suspend and getaddrinfo_a are called
> in the same way as above)
>
> struct gaicb gai = { "foo.baz", "0", NULL, &result };
>
> Here the ret is -2 as expected (EAI_NONAME - NAME or SERVICE unknown)
> - the "0" here is wrong.
>
>
> 3. For "foo.baz" + "http" (the gai_suspend and getaddrinfo_a are
> called in the same way as above)
>
> struct gaicb gai = { "foo.baz", "http", NULL, &result };
>
> Here the ret is also -2, which I don't understand. I would expect that
> the foo.baz name is going to be resolved as it "looks" correctly and
> hence the timeout will occur.
>
> Any hints on how timeout can be forced?
Gentle ping on this problem.
>
> > I am
> > not sure how we can accomplish it, maybe through running it on a
> > container with a custom NSS module to actually emulate the delay.
>
>
> 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
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
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2021-05-06 9:11 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-12 15:49 [RFC PATCH] tst: Add test for gai_suspend Lukasz Majewski
2021-03-12 16:09 ` Lukasz Majewski
2021-03-12 16:41 ` Andreas Schwab
2021-03-23 13:09 ` Adhemerval Zanella
2021-03-23 15:23 ` Lukasz Majewski
2021-03-26 11:30 ` Lukasz Majewski
2021-05-06 9:11 ` Lukasz Majewski
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).