Re: [PATCH glibc 5/9] glibc: Perform rseq(2) registration at C startup and thread creation (v17)

From: Florian Weimer <fw@deneb.enyo.de>
To: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>,
	libc-alpha <libc-alpha@sourceware.org>,
	carlos <carlos@redhat.com>, Rich Felker <dalias@libc.org>,
	linux-api <linux-api@vger.kernel.org>,
	Boqun Feng <boqun.feng@gmail.com>,
	Will Deacon <will.deacon@arm.com>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Ben Maurer <bmaurer@fb.com>, Dave Watson <davejwatson@fb.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Paul <paulmck@linux.vnet.ibm.com>, Paul Turner <pjt@google.com>,
	Joseph Myers <joseph@codesourcery.com>,
	Szabolcs Nagy <szabolcs.nagy@arm.com>
Subject: Re: [PATCH glibc 5/9] glibc: Perform rseq(2) registration at C startup and thread creation (v17)
Date: Tue, 28 Apr 2020 14:54:04 +0200	[thread overview]
Message-ID: <87tv13py8j.fsf@mid.deneb.enyo.de> (raw)
In-Reply-To: <190402462.72430.1588077816717.JavaMail.zimbra@efficios.com> (Mathieu Desnoyers's message of "Tue, 28 Apr 2020 08:43:36 -0400 (EDT)")

* Mathieu Desnoyers:

> ----- On Apr 28, 2020, at 8:35 AM, Florian Weimer fw@deneb.enyo.de wrote:
>
>> * Mathieu Desnoyers:
>> 
>>> One issue I'm currently facing when running "make check": because
>>> nptl/tst-rseq-nptl.c uses pthread_cancel(), I run into an Abort
>>> with:
>>>
>>> libgcc_s.so.1 must be installed for pthread_cancel to work
>>> Didn't expect signal from child: got `Aborted'
>> 
>> This is really unusual.  Is the affected test statically linked?
>
> I built glibc without specifying anything particular, and ran
> "make check". It indeed seems to be dynamically linked to libc:
>
> ldd tst-rseq-nptl
> ./tst-rseq-nptl: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.32' not found (required by ./tst-rseq-nptl)
> linux-vdso.so.1 (0x00007ffd3a2f3000)
> libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007f0527560000)
> libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f052716f000)
> /home/efficios/glibc-test5/lib/ld-linux-x86-64.so.2 => /lib64/ld-linux-x86-64.so.2 (0x00007f0527986000)

That's expected if the installed glibc is older than the built glibc.

> After make check I have:
>
> cat tst-rseq-nptl.test-result 
> FAIL: nptl/tst-rseq-nptl
> original exit status 134

What's in the tst-rseq-nptl.out file?

> And if I run
>
> ./tst-rseq-nptl
>
> Then I get
>
> libgcc_s.so.1 must be installed for pthread_cancel to work
> Didn't expect signal from child: got `Aborted'
> libgcc_s.so.1 must be installed for pthread_cancel to work
> Aborted (core dumped)

I'm puzzled why you don't get a GLIBC_2.32 version error in this case.
Do you build with --enable-hardcoded-path-in-tests?

> Same result if I do ./testrun.sh nptl/tst-rseq-nptl

That one definitely should work.

I expect you might see this if libgcc_s.so.1 is installed into a
multiarch subdirectory that upstream glibc does not search.  (The
Debian patches are unfortunately not upstream.)

I think on my system, the built glibc can find the system libgcc_s via
/etc/ld.so.cache, so I haven't seen this issue yet.