<html>
<head>
<style><!--
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
font-size: 12pt;
font-family:Calibri
}
--></style></head>
<body class='hmmessage'><div dir='ltr'>Pavel,<BR> <BR>Thanks for reply. You are right, it is reported as exception, here is the portion of MonitorSignal routine on Linux:<BR> <BR><font face="Courier New,sans-serif" size="2"> case SIGSEGV:<br> case SIGILL:<br> case SIGFPE:<br> case SIGBUS:<br> if (thread_sp)<br> std::static_pointer_cast<NativeThreadLinux> (thread_sp)->SetCrashedWithException (*info);<br> break;<br> default:<br> // This is just a pre-signal-delivery notification of the incoming signal.<br> if (thread_sp)<br> std::static_pointer_cast<NativeThreadLinux> (thread_sp)->SetStoppedBySignal (signo);</font><BR> <BR> <BR>I have no knowledge about LLGS (yet :)) but I'd vote for the second way: current LLDB behavior makes total sense for a "normal" application which is not in the business of managing its address space. The same applies to illegal instruction - the program could be self-modifying (say, JIT compiling). Exceptions are signals on Unix from the day one and we'd better not meddle with this long-established abstraction.<BR> <BR>I think we also need to give the choice to the developer during interactive debugging: i.e. "continue" command better have a flag that tells if to pass signal to the application or not.<BR> <BR>Thanks,<BR>Eugene<BR> <BR><div>> From: labath@google.com<br>> Date: Tue, 19 May 2015 11:35:32 +0100<br>> Subject: Re: [lldb-dev] LLDB blocks progress of program that handles SIGSEGV<br>> To: eugenebi@hotmail.com<br>> CC: lldb-dev@cs.uiuc.edu<br>> <br>> Hello Eugene,<br>> <br>> thanks for the report. What you are describing sounds like a genuine<br>> LLDB problem. I have created a bug<br>> <https://llvm.org/bugs/show_bug.cgi?id=23571> so we don't forget about<br>> this. It looks like LLGS (server component of lldb) is reporting this<br>> segfault to LLDB as an "exception" rather than a signal<br>> (reason:expection in the stop-reply packet). This causes LLDB to not<br>> reinject the signal when continuing the inferior. Now there are two<br>> places we can fix this:<br>> - have LLGS report this event as a signal. Then LLDB should be able to<br>> reinject the signal correctly.<br>> - teach LLDB to reinject the signal even in when the stop was reported<br>> as an exception.<br>> <br>> I'm inclined to do the first thing but I am not sure. Could somebody<br>> with more LLDB knowledge shed some light on this? What is the intended<br>> usage/behavior in case LLGS reports reason:exception?<br>> <br>> cheers,<br>> pl<br>> <br>> <br>> <br>> On 18 May 2015 at 18:27, Eugene Birukov <eugenebi@hotmail.com> wrote:<br>> > Hello,<br>> ><br>> > I am running Linux Ubuntu 14.04 and I tried both LLDB-3.6 (installed<br>> > binaries using apt-get) and LLDB-3.7 (built from sources sync'ed to trunk),<br>> > the result is the same.<br>> ><br>> > I have a simple program (the source code is at the bottom of the message)<br>> > that maps a page of memory as not accessible, installs a SIGSEGV handler<br>> > that remaps this page as read-write, and then tries to read from it. So, the<br>> > expected result is that program initially receives SIGSEGV but happily<br>> > continues after the handler fixes the problem.<br>> ><br>> > The program runs as expected, both standalone and under GDB:<br>> ><br>> ><br>> > $ ./mm<br>> > signal 11 received<br>> > success 777<br>> ><br>> ><br>> > $ gdb --quiet ./mm<br>> > Reading symbols from ./mm...done.<br>> > (gdb) r<br>> > Starting program: /home/eugene/tmp/mm<br>> > Program received signal SIGSEGV, Segmentation fault.<br>> > 0x0000000000400acc in main () at mm.cpp:27<br>> > 27 int x = *(int*)address;<br>> > (gdb) c<br>> > Continuing.<br>> > signal 11 received<br>> > success 777<br>> > [Inferior 1 (process 14155) exited normally]<br>> > (gdb) q<br>> ><br>> > But under LLDB it spins forever on failing instruction without invoking the<br>> > signal handler. Also, setting "process handle" does not have any effect at<br>> > all:<br>> ><br>> ><br>> > $ ~/llvm/bin/lldb ./mm<br>> > (lldb) target create "./mm"<br>> > Current executable set to './mm' (x86_64).<br>> > (lldb) br se -b main<br>> > Breakpoint 1: where = mm`main + 30 at mm.cpp:24, address =<br>> > 0x0000000000400a7e<br>> > (lldb) pr lau<br>> > Process 14194 launched: './mm' (x86_64)<br>> > Process 14194 stopped<br>> > * thread #1: tid = 14194, 0x0000000000400a7e mm`main + 30 at mm.cpp:24, name<br>> > = 'mm', stop reason = breakpoint 1.1<br>> > frame #0: 0x0000000000400a7e mm`main + 30 at mm.cpp:24<br>> > 21<br>> > 22 int main()<br>> > 23 {<br>> > -> 24 sigset(SIGSEGV, handler);<br>> > 25<br>> > 26 address = mmap(NULL, size, PROT_NONE, MAP_ANONYMOUS |<br>> > MAP_PRIVATE, -1, 0);<br>> > 27 int x = *(int*)address;<br>> > (lldb) pr ha SIGSEGV -s false -p true -n false<br>> > NAME PASS STOP NOTIFY<br>> > ========== ===== ===== ======<br>> > SIGSEGV true false false<br>> > (lldb) c<br>> > Process 14194 resuming<br>> > Process 14194 stopped<br>> > * thread #1: tid = 14194, 0x0000000000400acc mm`main + 108 at mm.cpp:27,<br>> > name = 'mm', stop reason = address access protected (fault address:<br>> > 0x7ffff7ff7000)<br>> > frame #0: 0x0000000000400acc mm`main + 108 at mm.cpp:27<br>> > 24 sigset(SIGSEGV, handler);<br>> > 25<br>> > 26 address = mmap(NULL, size, PROT_NONE, MAP_ANONYMOUS |<br>> > MAP_PRIVATE, -1, 0);<br>> > -> 27 int x = *(int*)address;<br>> > 28 std::cout << (signaled ? "success " : "failure ") << x << "\n";<br>> > 29 }<br>> > (lldb) c<br>> > Process 14194 resuming<br>> > Process 14194 stopped<br>> > * thread #1: tid = 14194, 0x0000000000400acc mm`main + 108 at mm.cpp:27,<br>> > name = 'mm', stop reason = address access protected (fault address:<br>> > 0x7ffff7ff7000)<br>> > frame #0: 0x0000000000400acc mm`main + 108 at mm.cpp:27<br>> > 24 sigset(SIGSEGV, handler);<br>> > 25<br>> > 26 address = mmap(NULL, size, PROT_NONE, MAP_ANONYMOUS |<br>> > MAP_PRIVATE, -1, 0);<br>> > -> 27 int x = *(int*)address;<br>> > 28 std::cout << (signaled ? "success " : "failure ") << x << "\n";<br>> > 29 }<br>> > (lldb) c<br>> > Process 14194 resuming<br>> > Process 14194 stopped<br>> > * thread #1: tid = 14194, 0x0000000000400acc mm`main + 108 at mm.cpp:27,<br>> > name = 'mm', stop reason = address access protected (fault address:<br>> > 0x7ffff7ff7000)<br>> > frame #0: 0x0000000000400acc mm`main + 108 at mm.cpp:27<br>> > 24 sigset(SIGSEGV, handler);<br>> > 25<br>> > 26 address = mmap(NULL, size, PROT_NONE, MAP_ANONYMOUS |<br>> > MAP_PRIVATE, -1, 0);<br>> > -> 27 int x = *(int*)address;<br>> > 28 std::cout << (signaled ? "success " : "failure ") << x << "\n";<br>> > 29 }<br>> > (lldb)<br>> ><br>> > So, do I miss some LLDB setting?<br>> > Here is the program code:<br>> ><br>> ><br>> > $ cat mm.cpp<br>> > #include <iostream><br>> > #include <sys/mman.h><br>> > #include <signal.h><br>> > #include <assert.h><br>> > #include <sys/types.h><br>> > #include <unistd.h><br>> ><br>> > void* address;<br>> > size_t size = 0x1000;<br>> > bool signaled = false;<br>> ><br>> > void handler(int sig)<br>> > {<br>> > std::cout << "signal " << sig << " received\n";<br>> > signaled = true;<br>> > munmap(address, size);<br>> > void* newaddr = mmap(address, size, PROT_READ | PROT_WRITE,<br>> > MAP_ANONYMOUS | MAP_FIXED | MAP_PRIVATE, -1, 0);<br>> > assert(newaddr == address);<br>> > *(int*)newaddr = 777;<br>> > }<br>> ><br>> > int main()<br>> > {<br>> > sigset(SIGSEGV, handler);<br>> > address = mmap(NULL, size, PROT_NONE, MAP_ANONYMOUS | MAP_PRIVATE, -1,<br>> > 0);<br>> > int x = *(int*)address;<br>> > std::cout << (signaled ? "success " : "failure ") << x << "\n";<br>> > }<br>> > $<br>> ><br>> > Thanks,<br>> > Eugene<br>> ><br>> ><br>> ><br>> > _______________________________________________<br>> > lldb-dev mailing list<br>> > lldb-dev@cs.uiuc.edu<br>> > http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev<br>> ><br></div> </div></body>
</html>