From owner-svn-src-all@freebsd.org Sun Mar 31 11:28:15 2019 Return-Path: Delivered-To: svn-src-all@mailman.ysv.freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2610:1c1:1:606c::19:1]) by mailman.ysv.freebsd.org (Postfix) with ESMTP id EBCAF155A15E; Sun, 31 Mar 2019 11:28:14 +0000 (UTC) (envelope-from brde@optusnet.com.au) Received: from mail105.syd.optusnet.com.au (mail105.syd.optusnet.com.au [211.29.132.249]) by mx1.freebsd.org (Postfix) with ESMTP id 76BF875490; Sun, 31 Mar 2019 11:28:06 +0000 (UTC) (envelope-from brde@optusnet.com.au) Received: from [192.168.0.102] (c110-21-101-228.carlnfd1.nsw.optusnet.com.au [110.21.101.228]) by mail105.syd.optusnet.com.au (Postfix) with ESMTPS id 0B273105D0D5; Sun, 31 Mar 2019 22:27:55 +1100 (AEDT) Date: Sun, 31 Mar 2019 22:27:54 +1100 (EST) From: Bruce Evans X-X-Sender: bde@besplex.bde.org To: Konstantin Belousov cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r345696 - head/lib/libvgl In-Reply-To: <20190330094558.GA1923@kib.kiev.ua> Message-ID: <20190331214235.K961@besplex.bde.org> References: <201903291557.x2TFv9AW097226@repo.freebsd.org> <20190329182100.GZ1923@kib.kiev.ua> <20190330142319.I1011@besplex.bde.org> <20190330094558.GA1923@kib.kiev.ua> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII; format=flowed X-Optus-CM-Score: 0 X-Optus-CM-Analysis: v=2.2 cv=UJetJGXy c=1 sm=1 tr=0 cx=a_idp_d a=PalzARQSbocsUSjMRkwAPg==:117 a=PalzARQSbocsUSjMRkwAPg==:17 a=jpOVt7BSZ2e4Z31A5e1TngXxSK0=:19 a=kj9zAlcOel0A:10 a=6I5d2MoRAAAA:8 a=fx17AMauYh6YMvl-fBgA:9 a=owp30hQRNWXAININ:21 a=82lPrFZ9h1KNQv0H:21 a=CjuIK1q_8ugA:10 a=IjZwj45LgO3ly-622nXo:22 X-Rspamd-Queue-Id: 76BF875490 X-Spamd-Bar: ----- Authentication-Results: mx1.freebsd.org; spf=pass (mx1.freebsd.org: domain of brde@optusnet.com.au designates 211.29.132.249 as permitted sender) smtp.mailfrom=brde@optusnet.com.au X-Spamd-Result: default: False [-5.55 / 15.00]; ARC_NA(0.00)[]; NEURAL_HAM_MEDIUM(-1.00)[-1.000,0]; RCVD_COUNT_TWO(0.00)[2]; FROM_HAS_DN(0.00)[]; RCPT_COUNT_THREE(0.00)[4]; FREEMAIL_FROM(0.00)[optusnet.com.au]; R_SPF_ALLOW(-0.20)[+ip4:211.29.132.0/23]; MIME_GOOD(-0.10)[text/plain]; DMARC_NA(0.00)[optusnet.com.au]; NEURAL_HAM_LONG(-1.00)[-1.000,0]; TO_DN_SOME(0.00)[]; TO_MATCH_ENVRCPT_SOME(0.00)[]; MX_GOOD(-0.01)[extmail.optusnet.com.au]; NEURAL_HAM_SHORT(-0.86)[-0.859,0]; IP_SCORE(-2.38)[ip: (-5.60), ipnet: 211.28.0.0/14(-3.48), asn: 4804(-2.77), country: AU(-0.04)]; FREEMAIL_TO(0.00)[gmail.com]; RCVD_NO_TLS_LAST(0.10)[]; RCVD_IN_DNSWL_LOW(-0.10)[249.132.29.211.list.dnswl.org : 127.0.5.1]; R_DKIM_NA(0.00)[]; FREEMAIL_ENVFROM(0.00)[optusnet.com.au]; ASN(0.00)[asn:4804, ipnet:211.28.0.0/14, country:AU]; MIME_TRACE(0.00)[0:+]; FROM_EQ_ENVFROM(0.00)[] X-BeenThere: svn-src-all@freebsd.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: "SVN commit messages for the entire src tree \(except for " user" and " projects" \)" List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 31 Mar 2019 11:28:15 -0000 On Sat, 30 Mar 2019, Konstantin Belousov wrote: > On Sat, Mar 30, 2019 at 03:24:40PM +1100, Bruce Evans wrote: >> On Fri, 29 Mar 2019, Konstantin Belousov wrote: >> >>> On Fri, Mar 29, 2019 at 03:57:09PM +0000, Bruce Evans wrote: >>>> Author: bde >>>> Date: Fri Mar 29 15:57:08 2019 >>>> New Revision: 345696 >>>> URL: https://svnweb.freebsd.org/changeset/base/345696 >>>> >>>> Log: >>>> Fix endless loops for handling SIGBUS and SIGSEGV. >>>> >>>> r80270 has the usual wrong fix for unsafe signal handling -- just set >>>> a flag and return to let an event loop check the flag and do safe >>>> handling. This never works for signals like SIGBUS and SIGSEGV that >>>> repeat and works poorly for others unless the application has an event >>>> loop designed to support this. >>>> >>>> For these signals, clean up unsafely as before, except for arranging that >>>> nested signals are fatal and forcing a nested signal if the cleanup doesn't >>>> cause one. >>>> >>>> Modified: >>>> head/lib/libvgl/main.c >>>> >>>> Modified: head/lib/libvgl/main.c >>>> ============================================================================== >>>> --- head/lib/libvgl/main.c Fri Mar 29 15:20:48 2019 (r345695) >>>> +++ head/lib/libvgl/main.c Fri Mar 29 15:57:08 2019 (r345696) >>>> ... >>>> @@ -107,14 +107,22 @@ struct vt_mode smode; >>>> } >>>> >>>> static void >>>> -VGLAbort(int arg __unused) >>>> +VGLAbort(int arg) >>>> { >>>> + sigset_t mask; >>>> + >>>> VGLAbortPending = 1; >>>> signal(SIGINT, SIG_IGN); >>>> signal(SIGTERM, SIG_IGN); >>>> - signal(SIGSEGV, SIG_IGN); >>>> - signal(SIGBUS, SIG_IGN); >>>> signal(SIGUSR2, SIG_IGN); >>>> + if (arg == SIGBUS || arg == SIGSEGV) { >>>> + signal(arg, SIG_DFL); >>>> + sigemptyset(&mask); >>>> + sigaddset(&mask, arg); >>>> + sigprocmask(SIG_UNBLOCK, &mask, NULL); >>>> + VGLEnd(); >>>> + kill(getpid(), arg); >>> This of course misses the siginfo information from the real fault. >> >> It is in the nested signal frame. >> >>> Why SIGBUS/SIGSEGV are caught at all ? >> >> Otherwise, the screen is left in an application mode that the kernel >> doesn't support (except to not write to the screen, so that the >> application has full control). Also, the keyboard may be left in a strange state. Usually this is no worse than termios raw mode, but it may be raw scancodes. In raw scancodes mode, Alt-Fn to switch to another vty to fix the problem. libvgl disables switching to another vty without going through a libvgl handler anyway. The combination is usually enough to break switching to vty0 to run ddb, though that should work for at least breakpoints in the kernel. IIRC, sc_cngrab() switches the keyboard mode, and my version of it restores ignoring of the anti-switch flag. > ... > I am more about not doing kill(2) from the handler, instead do plain > return to the context which caused the original signal. That won't work well for SIGBUS's and SIGSEGV's related to the library. If the above cleanup is not done, then it is too hard to debug the problem, and if the above cleanup is done then it is impossible to debug the original problem if it was for and invalid access in libvgl code. The latter is a problem for the above cleanup too -- the cleanup clobbers the libvgl state. However, I could do a more limited cleanup of just restoring the screen and keyboard state using ioctls. ioctl() isn't async-signal safe in POSIX, but most ioctls are async-signal safe in practice provided they don't use global variables that were set too recently or too non-atomically. VGLEnd() uses main globals established by VGLInit(). It depends on program order which is not guaranteed in signal handlers for testing the flags set by initialization. The cleanup now also does: - screen clearing. This breaks debugging and is dangerous if the bug is in screen clearing. Screen clearing is also very slow, and belongs in the kernel. It is now done 4 times per libvgl use, but its slowness is limited by the kernel not clearing properly and the kernel not allowing the application to mmap() the whole physical frame buffer in some versions of FreeBSD (this breaks panning). - munmap(). This breaks debugging. - free() of backing store buffer and the main VGLDisplay data. This breaks debugging and is unnecessary if we are going to exit. Bruce