From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from simark.ca by simark.ca with LMTP id SJbfAL1shF9FawAAWB0awg (envelope-from ) for ; Mon, 12 Oct 2020 10:48:29 -0400 Received: by simark.ca (Postfix, from userid 112) id 0044B1EF6F; Mon, 12 Oct 2020 10:48:28 -0400 (EDT) X-Spam-Checker-Version: SpamAssassin 3.4.2 (2018-09-13) on simark.ca X-Spam-Level: X-Spam-Status: No, score=-1.0 required=5.0 tests=MAILING_LIST_MULTI, URIBL_BLOCKED autolearn=ham autolearn_force=no version=3.4.2 Received: from sourceware.org (server2.sourceware.org [8.43.85.97]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256) (No client certificate requested) by simark.ca (Postfix) with ESMTPS id 6C05E1E58D for ; Mon, 12 Oct 2020 10:48:27 -0400 (EDT) Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 31FCF386F41A; Mon, 12 Oct 2020 14:48:27 +0000 (GMT) Received: from mail-wm1-f65.google.com (mail-wm1-f65.google.com [209.85.128.65]) by sourceware.org (Postfix) with ESMTPS id 4C81C3857C45 for ; Mon, 12 Oct 2020 14:48:24 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.3.2 sourceware.org 4C81C3857C45 Authentication-Results: sourceware.org; dmarc=none (p=none dis=none) header.from=palves.net Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=alves.ped@gmail.com Received: by mail-wm1-f65.google.com with SMTP id f21so17377774wml.3 for ; Mon, 12 Oct 2020 07:48:24 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:subject:to:references:from:message-id:date :user-agent:mime-version:in-reply-to:content-language :content-transfer-encoding; bh=TgdRsGDReh0s02bM7yGPTAsByJ32fDMONNkJY1OMw8s=; b=DBk8KfzSlvNGEVN3GxUyYEoaVzLxPViMcPfA6G0DomwJjytPrR3ZTrpGb0+/Q0evC0 F6KWEO6wq8af1RPRumMJR7xSaHwNGf6rG/1g9KnBTFWH761OYOCEHmop5q272TB83YT2 ESIE+XHH72+aADgHunLEOrf2rN/YVT5/8uj9S6Rhk56ssQ3nVWoPk4yKI9MU599c5zGP eqLwykbpimoEMhb9mnJgOeK9AScw29QCHHzngqHlj1GXU+rHNfn7kUssmTB2XBmpNwgZ Jv5LdfECsrTIeVZ1Gna8SLEjeyeNvaFTV3aAFUc1gkeo9rrh5rH2KZ/z8yTTVZBPRlfJ Io6Q== X-Gm-Message-State: AOAM531FD+egXnOmc6yM9aAfJiSYioYEje490v/CFubucao6ZY1SEbFa EnkNajtUj87f7QPcTZxGkUQKtbnsNaYTIWV+ X-Google-Smtp-Source: ABdhPJzt2t2+P7b/06xTpb3c2MSjAn9OHm/v6k4uwdPsrVeRT771K9MDesLmg0mL854GROIfQ0q4TA== X-Received: by 2002:a1c:111:: with SMTP id 17mr10912375wmb.126.1602514102383; Mon, 12 Oct 2020 07:48:22 -0700 (PDT) Received: from ?IPv6:2001:8a0:f91e:6d00:c80a:ea25:47ef:5f73? ([2001:8a0:f91e:6d00:c80a:ea25:47ef:5f73]) by smtp.gmail.com with ESMTPSA id v9sm24445656wmh.23.2020.10.12.07.48.20 (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Mon, 12 Oct 2020 07:48:21 -0700 (PDT) Subject: Re: [PATCH] gdb: don't pass TARGET_WNOHANG to targets that can't async (PR 26642) To: Simon Marchi , gdb-patches@sourceware.org References: <20201001025656.2561757-1-simon.marchi@polymtl.ca> From: Pedro Alves Message-ID: <2fc59140-0fb4-cab3-a816-2abc95116387@palves.net> Date: Mon, 12 Oct 2020 15:48:19 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.2.1 MIME-Version: 1.0 In-Reply-To: <20201001025656.2561757-1-simon.marchi@polymtl.ca> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit X-BeenThere: gdb-patches@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gdb-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: gdb-patches-bounces@sourceware.org Sender: "Gdb-patches" On 10/1/20 3:56 AM, Simon Marchi via Gdb-patches wrote: > Debugging with "maintenance set target-async off" on Linux has been > broken since 5b6d1e4fa4f ("Multi-target support"). > > The issue is easy to reproduce: > > $ ./gdb -q --data-directory=data-directory -nx ./test > Reading symbols from ./test... > (gdb) maintenance set target-async off > (gdb) start > Temporary breakpoint 1 at 0x1151: file test.c, line 5. > Starting program: /home/simark/build/binutils-gdb/gdb/test > > ... and it hangs there. > > The difference between pre-5b6d1e4fa4f and 5b6d1e4fa4f is that > fetch_inferior_event now calls target_wait with TARGET_WNOHANG for > non-async-capable targets, whereas it didn't before. > > For non-async-capable targets, this is how it's expected to work when > resuming execution: > > 1. we call resume > 2. the infrun async handler is marked in prepare_to_wait, to immediatly immediatly -> immediately > wake up the event loop > 3. fetch_inferior_event calls the target's wait method without > TARGET_WNOHANG, effectively blocking until the target has something > to report > > However, since we call the target's wait method with TARGET_WNOHANG, > this happens: > > 1. we call resume > 2. the infrun async handler is marked in prepare_to_wait, to immediatly Ditto. > wake up the event loop > 3. fetch_inferior_event calls the target's wait method with > TARGET_WNOHANG, the target has nothing to report yet > 4. we go back to blocking on the event loop > 5. SIGCHLD finally arrives, but the event loop is not woken up, because > we are not in async mode. Normally, we should have been stuck in > waitpid the SIGCHLD would have unblocked us. > > We end up in this situation because these two necessary conditions are > met: > > 1. GDB uses the TARGET_WNOHANG option with a target that can't do async. > I don't think this makes sense. I mean, it's technically possible, > the doc for TARGET_WNOHANG is: > > /* Return immediately if there's no event already queued. If this > options is not requested, target_wait blocks waiting for an > event. */ > TARGET_WNOHANG = 1, > > ... which isn't in itself necessarily incompatible with synchronous > targets. But I don't see when it would be useful to ask a sync > target to do a non-blocking wait. I could imagine it being useful to implement async mode for targets that can poll for events, but don't have a native asynchronous "there's a new event!" mechanism (like SIGCHLD or similar). So GDB could poll for events periodically, say, with a timer via the event loop. Windows could gain async support that way. > > Pass TARGET_WNOHANG to sync targets also poses the risk of passing > TARGET_WNOHANG to a target that doesn't implement it. The caller of > target_wait would expect the call not to block, but the call may > indeed block. > > Since supporting TARGET_WNOHANG is a requirement for targets that can > do async (I believe), I propose (as implemented in this patch) that > we add an assertion in target_wait to make sure we don't ask a target > that can't do async to handle TARGET_WNOHANG. > > > A question in my mind is: could we bind the blocking or non-blocking > behavior of wait with can_async_p? In other words: > > - Do we ever need to do a blocking wait on a target that supports > async? Yes, for example in prepare_for_detach, we do a blocking wait (via do_target_wait). I think all such loops could likely be converted to go via a nested event loop, which may be better for handling user input events at the same time. But regardless, I wouldn't be conformable with trying can_async_p with what a specific target_wait call wants. Seems best to me to keep them orthogonal. > - Do we ever need to do a non-blocking wait on a target that does > not support async? Not today, I don't think. > > So, could we make it so that if target's can_async_p method returns > true, its wait method is necessarily non-blocking? And if > can_async_p returns false, its wait method is necessarily blocking? I'd rather not. > > It sounds to me like it would simplify the semantic of > target_ops::wait a little bit. > > > 2. The linux-nat target, even in the mode where it emulates a > synchronous target (with "maintenance set target-async off") respects > TARGET_WNOHANG. > > Non-async targets, such as windows_nat_target, simply don't check / > support TARGET_WNOHANG. Their wait method is always blocking. So, > to properly emulate a non-async target, I believe that linux-nat > should also ignore it when in "maintenance set target-async off" > mode. Its behavior would be closer to a "true" non-async target. > > The problem disappears if we simply fix either of these issues, but I > think it wouldn't hurt to fix both. Unless there's a need otherwise, it would seem better to me to just fix the common code to not request a non-blocking wait out of !async targets. I don't think there's any good reason to complicate (however little) linux-nat.c to handle a scenario that doesn't exist. > > The new test gdb.base/maint-target-async-off.exp is a simple test that > just tries running to main and then to the end of the program, with > "maintenance set target-async off". > > gdb/ChangeLog: > > PR gdb/26642 > * infrun.c (do_target_wait_1): Clear TARGET_WNOHANG if the > target can't do async. > * linux-nat.c (linux_nat_wait_1): Ignore TARGET_WNOHANG if > target_async_permitted is false. > * target.c (target_wait): Assert that we don't pass > TARGET_WNOHANG to a target that can't async. > > gdb/testsuite/ChangeLog: > > PR gdb/26642 > * gdb.base/maint-target-async-off.c: New test. > * gdb.base/maint-target-async-off.exp: New test. > > Change-Id: I69ad3a14598863d21338a8c4e78700a58ce7ad86 > --- > gdb/infrun.c | 5 +++ > gdb/linux-nat.c | 6 +++- > gdb/target.c | 5 +++ > .../gdb.base/maint-target-async-off.c | 22 +++++++++++++ > .../gdb.base/maint-target-async-off.exp | 32 +++++++++++++++++++ > 5 files changed, 69 insertions(+), 1 deletion(-) > create mode 100644 gdb/testsuite/gdb.base/maint-target-async-off.c > create mode 100644 gdb/testsuite/gdb.base/maint-target-async-off.exp > > diff --git a/gdb/infrun.c b/gdb/infrun.c > index daf10417842f..1b88514c9cc7 100644 > --- a/gdb/infrun.c > +++ b/gdb/infrun.c > @@ -3533,6 +3533,11 @@ do_target_wait_1 (inferior *inf, ptid_t ptid, > > /* But if we don't find one, we'll have to wait. */ > > + /* We can't ask a non-async target to do a non-blocking wait, so this will be > + a blocking wait. */ > + if (!target_can_async_p ()) > + options &= ~TARGET_WNOHANG; > + > if (deprecated_target_wait_hook) > event_ptid = deprecated_target_wait_hook (ptid, status, options); > else > diff --git a/gdb/linux-nat.c b/gdb/linux-nat.c > index fbb538859429..1064fc4f8c72 100644 > --- a/gdb/linux-nat.c > +++ b/gdb/linux-nat.c > @@ -3238,7 +3238,11 @@ linux_nat_wait_1 (ptid_t ptid, struct target_waitstatus *ourstatus, > > /* No interesting event to report to the core. */ > > - if (target_options & TARGET_WNOHANG) > + /* If target_async_permitted is false (maintenance set target-async off > + is used), pretend that we don't know about TARGET_WNOHANG and go block > + in wait_for_signal. */ > + if (target_options & TARGET_WNOHANG > + && target_async_permitted) I'd rather go without this hunk. Would it still fix things? > { > linux_nat_debug_printf ("exit (ignore)"); > > diff --git a/gdb/target.c b/gdb/target.c > index dd78a848caec..6f340678b7ca 100644 > --- a/gdb/target.c > +++ b/gdb/target.c > @@ -1997,6 +1997,11 @@ ptid_t > target_wait (ptid_t ptid, struct target_waitstatus *status, > target_wait_flags options) > { > + target_ops *target = current_top_target (); > + > + if (!target->can_async_p ()) Why not call "!target_can_async_p ()" instead? It's exactly the same. > + gdb_assert ((options & TARGET_WNOHANG) == 0); > + > return current_top_target ()->wait (ptid, status, options); Perhaps you meant to adjust this to do "target->wait" instead? > } > > diff --git a/gdb/testsuite/gdb.base/maint-target-async-off.c b/gdb/testsuite/gdb.base/maint-target-async-off.c > new file mode 100644 > index 000000000000..9d7b2f1a4c28 > --- /dev/null > +++ b/gdb/testsuite/gdb.base/maint-target-async-off.c > @@ -0,0 +1,22 @@ > +/* This testcase is part of GDB, the GNU debugger. > + > + Copyright 2020 Free Software Foundation, Inc. > + > + This program is free software; you can redistribute it and/or modify > + it under the terms of the GNU General Public License as published by > + the Free Software Foundation; either version 3 of the License, or > + (at your option) any later version. > + > + This program 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 General Public License for more details. > + > + You should have received a copy of the GNU General Public License > + along with this program. If not, see . */ > + > +int > +main (void) > +{ > + return 0; > +} > diff --git a/gdb/testsuite/gdb.base/maint-target-async-off.exp b/gdb/testsuite/gdb.base/maint-target-async-off.exp > new file mode 100644 > index 000000000000..e77bc79a21e1 > --- /dev/null > +++ b/gdb/testsuite/gdb.base/maint-target-async-off.exp > @@ -0,0 +1,32 @@ > +# Copyright 2020 Free Software Foundation, Inc. > + > +# This program is free software; you can redistribute it and/or modify > +# it under the terms of the GNU General Public License as published by > +# the Free Software Foundation; either version 3 of the License, or > +# (at your option) any later version. > +# > +# This program 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 General Public License for more details. > +# > +# You should have received a copy of the GNU General Public License > +# along with this program. If not, see . > + > +# Verify that debugging with "maintenance target-async off" works somewhat. At > +# least running to main and to the end of the program. > + > +standard_testfile > + > +if { [prepare_for_testing "failed to prepare" ${testfile} ${srcfile}] } { > + return > +} > + > +gdb_test_no_output "maintenance set target-async off" I don't think this works correctly with --target_board=native-extended-gdbserver, since by the time you get here, the remote connection is already set up. I think the best way to handle that is to do the same we do what non-stop testcases do: save_vars { GDBFLAGS } { append GDBFLAGS " -ex \"set non-stop $nonstop\"" clean_restart ${executable} } > + > +if { ![runto_main] } { > + fail "can't run to main" > + return > +} > + > +gdb_continue_to_end >