[lldb-dev] eh_frame or debug_frame

Jason Molenda jmolenda at apple.com
Wed Oct 15 18:16:13 PDT 2014


That one doesn't seem to repo the problem.  The unwinder always comes back with

(lldb) bt
* thread #1: tid = 0x4ff734, 0x000000000000201a test`main.foo(x=1) + 26 at test.go:4, stop reason = breakpoint 1.1
  * #0: 0x000000000000201a test`main.foo(x=1) + 26 at test.go:4
    #1: 0x0000000000002111 test`main.main + 49 at test.go:15
    #2: 0x000000000000d463 test`runtime.main + 243 at proc.go:63
    #3: 0x00000000000259f0 test`runtime.gosched_m + 192 at proc.c:1641
(lldb)

I can si and backtrace again and I get the same backtrace -- lldb sees that the saved pc for frame 4 would be in non-executable memory and stops the stack walk:

    th1/fr4 pc = 0x00000002080b7f98
   th1/fr3 supplying caller's saved reg 6's location using x86_64 default unwind plan UnwindPlan
   th1/fr3 supplying caller's register 6 from the stack, saved at CFA plus offset -16
    th1/fr4 fp = 0x0000000000000000
   th1/fr3 supplying caller's stack pointer (7) value, computed from CFA
    th1/fr4 sp = 0x00000002080c0010
    th1/fr4 using architectural default unwind method
    th1/fr4 pc is in a non-executable section of memory and this isn't the 2nd frame in the stack walk.
    Frame 4 invalid RegisterContext for this frame, stopping stack walk


I needed the patch you appended to http://llvm.org/bugs/show_bug.cgi?id=21118 / http://reviews.llvm.org/D5735 to run the program.  Do you have llvm commit access?  I'll commit the patch if you don't.

J


> On Oct 15, 2014, at 5:33 PM, Ryan Brown <ribrdb at google.com> wrote:
> 
> Yes, I'm using os x. You could try using the binary I uploaded in http://llvm.org/bugs/show_bug.cgi?id=21118
> That will only have one go thread though.
> 
> -- Ryan Brown
> 
> On Wed, Oct 15, 2014 at 5:22 PM, Jason Molenda <jmolenda at apple.com> wrote:
> Yeah, I was afraid of that.
> 
> What I'm trying to do with this code is say "unwind using your super-super smart techniques ... but if you hit a wall, try the simplistic unwind method and see if you can get further."
> 
> The problem here is that lldb is doing the full stack walk as far as it can be walked ... but it thinks maybe switching to the architecture default unwind plan might get it further (which it does not).  The switch to the arch default unwind plan is destructive - it replaces the assembly profile unwind instructions for that function - and is remembered for future stack walks.  That's why your threads get progressively fewer backtraces.
> 
> I'll need to look into this and come up with a fix.  I don't suppose your go binary runs on mac os x, does it?  It would be great if I had a failing test program in front of me while I try to come up with a fix.
> 
> 
> > On Oct 15, 2014, at 5:13 PM, Ryan Brown <ribrdb at google.com> wrote:
> >
> > So adding "return false" to the top of TryFallbackUnwindPlan() fixes the problem.
> > The call at UnwindLLDB:177, when !reg_ctx_sp->IsValid() seems to be the only one I'm hitting.
> >
> > -- Ryan Brown
> >
> > On Wed, Oct 15, 2014 at 3:20 PM, Jason Molenda <jmolenda at apple.com> wrote:
> > urgh, sorry, I wasn't paying attention to the svn log output when I copy & pasted the rev.  It's this change I wanted to mention - r219247.  It's going to be someone calling TryFallbackUnwindPlan(), I just added some new cases where that could be called.  It may not be my most recent change (219247) but it's going to be that method which is causing the problem.
> >
> >
> > > On Oct 15, 2014, at 2:59 PM, Ryan Brown <ribrdb at google.com> wrote:
> > >
> > > rolling back r219772 (Be more consistent about null checks for the Process and ABI in GetFullUnwindPlanForFrame) doesn't seem to have any effect.
> > >
> > > -- Ryan Brown
> > >
> > > On Wed, Oct 15, 2014 at 2:31 PM, Jason Molenda <jmolenda at apple.com> wrote:
> > >
> > > > On Oct 15, 2014, at 1:43 PM, Ryan Brown <ribrdb at google.com> wrote:
> > > >
> > > > Go doesn't have exception handlers, so it doesn't write .eh_frame. Wouldn't it make sense to use .debug_frame if .eh_frame is missing?
> > >
> > >
> > > We could do that.  I'm surprised if go is emitting x86_64 code without eh_frame.  As Joerg points out, debug_frame is great but it may not be available when an analysis tool is examining a binary.  eh_frame has the benefit of always being in the binary.
> > >
> > > >
> > > > With my custom RegisterContext I got backtraces to work for my memory threads. But something strange is going on. I have 10 threads that should have identical traces, but the first has 5 frames, then 4, 3, 2, and the rest only have 1 frame.
> > >
> > >
> > > It's easiest to isolate one thread backtrace in a situation like this.  For instance, looking at thread 7 in your program.  (the unwind algorithms have no cross-thread information passing):
> > >
> > >
> > > th7/fr0 initialized frame current pc is 0xdaef cfa is 0x20809feb8 using assembly insn profiling UnwindPlan
> > >
> > > lldb is using the assembly unwind inspection for frame 0.  You said that all ten threads should have the same backtrace but thread #2 is at 0x2fe8c, #3 is at 0x209a, threads 4-15 are at 0xdaef.  You meant threads 4-15 should all be the same.
> > >
> > >
> > >      th7/fr5 pc = 0x0000000000002078
> > >      th7/fr5 fp = 0xffffffffffffffff
> > >     th7/fr4 supplying caller's stack pointer (7) value, computed from CFA
> > >      th7/fr5 sp = 0x000000020809ffc8
> > >      th7/fr5 active row: 0x0000000000002050: CFA=rbp+16 => rbp=[rbp] rsp=rbp+16 rip=[rbp+8]
> > >
> > > That's the architectural default unwind plan for x86_64 ABIs.  Over in thread 6, it looks like failed to unwind past frame 5 with the assembly unwind, figured the assembly unwind was incorrect, and tried switching over to using the architectural default unwind plan:
> > >
> > > th6/fr0 supplying caller's saved reg 6's location, cached
> > >      th6/fr5 full unwind plan 'assembly insn profiling' has been replaced by architecture default unwind plan 'x86_64 default unwind plan' for this function from now on.
> > >      th6/fr5 supplying caller's saved reg 16's location using x86_64 default unwind plan UnwindPlan
> > >      th6/fr5 supplying caller's register 16 from the stack, saved at CFA plus offset -8
> > >       th6/fr6 could not get pc value
> > >       Frame 6 invalid RegisterContext for this frame, stopping stack walk
> > > th6 Unwind of this thread is complete.
> > >
> > > From this point forward main.okread() will use the arch default unwind plan which isn't going to work.
> > >
> > > Can you try rolling back r219772 and seeing if that helps?  I suspect lldb may be slowly stripping off the last frame of the unwind for each thread as it progresses.
> > >
> > > J
> > >
> > > PS- "bt all" works just as well as "thread backtrace all".
> > >
> > >
> > > >
> > > > There's a log here, thread 6 is the one with the complete backtrace. https://gist.github.com/ribrdb/386fb0e555e82483d21d
> > > >
> > > > Comparing thread 7 with thread 6, things seem fine up to line 627:
> > > >     th7/fr4 supplying caller's stack pointer (7) value, computed from CFA
> > > >      th7/fr5 sp = 0x000000020809ffc8
> > > >      th7/fr5 active row: 0x0000000000002050: CFA=rbp+16 => rbp=[rbp] rsp=rbp+16 rip=[rbp+8]
> > > >
> > > > While thread 6 has:
> > > >      th6/fr4 supplying caller's stack pointer (7) value, computed from CFA
> > > >      th6/fr5 sp = 0x000000020809f7c8
> > > >      th6/fr5 active row: 0x000000000000206a: CFA=rsp+16 => rsp=rsp+16 rip=[rsp+8]
> > > >
> > > > I don't know where rbp came from, it's not in the function at all:
> > > > 0x2050 <main.okread>: movq   %gs:0x8a0, %rcx
> > > > 0x2059 <main.okread+9>: cmpq   0x10(%rcx), %rsp
> > > > 0x205d <main.okread+13>: ja     0x2066                    ; main.okread + 22 at test.go:9
> > > > 0x205f <main.okread+15>: callq  0x2d510                   ; runtime.morestack_noctxt at asm_amd64.s:330
> > > > 0x2064 <main.okread+20>: jmp    0x2050                    ; main.okread at test.go:9
> > > > 0x2066 <main.okread+22>: subq   $0x8, %rsp
> > > > 0x206a <main.okread+26>: movq   0x10(%rsp), %rbx
> > > > 0x206f <main.okread+31>: movq   %rbx, (%rsp)
> > > > 0x2073 <main.okread+35>: callq  0x2000                    ; main.doread at test.go:5
> > > > 0x2078 <main.okread+40>: addq   $0x8, %rsp
> > > > 0x207c <main.okread+44>: retq
> > > > 0x207d <main.okread+45>: addb   %al, (%rax)
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > -- Ryan Brown
> > > >
> > > > On Wed, Oct 15, 2014 at 11:48 AM, Ryan Brown <ribrdb at google.com> wrote:
> > > > Yes, I'm writing a class to do that now. It's just not supported by any of the existing register contexts.
> > > >
> > > > -- Ryan Brown
> > > >
> > > > On Wed, Oct 15, 2014 at 11:37 AM, Jason Molenda <jason at molenda.com> wrote:
> > > > Can't your OS plugin for the goroutines use the same sp and ip register numbers as x86_64 (instead of 0 and 1 like you might be using right now) when it reports them to lldb, and return all the other registers as "unavailable" if they're requested?
> > > >
> > > > The tricky bit about living on eh_frame / debug_frame is that lldb doesn't know what kind of unwind info it is being given.  Is it just for exception handling locations?  Does it contain prologue setup?  epilogue?  Is it fully asynchronous - giving unwind details at all locations?  There aren't any flags in eh_frame/debug_frame that could give us a hint about what we're working with.
> > > >
> > > >
> > > >
> > > > On Oct 15, 2014, at 11:24 AM, Ryan Brown <ribrdb at google.com> wrote:
> > > >
> > > > > I'm actually struggling with this right now. I'm trying to implement an OS plugin so goroutines show up as threads.
> > > > > The go compiler puts instruction accurate unwind info into .debug_frame, I'm not sure what (if anything) goes into eh_frame.
> > > > > However lldb uses the disassembly instead of the dwarf info. The x86 unwinder assumes that all threads have the same LLDB register numbers, but other parts of the code require that the LLDB register number is < (number of registers). Goroutines only store sp and ip, so it seems I'm going to have to create a custom RegisterContext subclass to get the existing unwinder to work for goroutines.
> > > > >
> > > > > On Tue, Oct 14, 2014 at 5:51 PM, Jason Molenda <jmolenda at apple.com> wrote:
> > > > > > On Oct 13, 2014, at 9:55 AM, Greg Clayton <gclayton at apple.com
> > > > > > wrote:
> > > > >
> > > > > >
> > > > >
> > > > >
> > > > > >
> > > > >
> > > > >
> > > > > >> On Oct 10, 2014, at 1:58 PM, Francois Pichet <pichet2000 at gmail.com
> > > > > > wrote:
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >> On Fri, Oct 10, 2014 at 4:20 PM, Greg Clayton <gclayton at apple.com
> > > > > > wrote:
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >>> On Oct 10, 2014, at 1:05 PM, Philippe Lavoie <philippe.lavoie at octasic.com
> > > > > > wrote:
> > > > >
> > > > > >>>
> > > > >
> > > > >
> > > > > >>>
> > > > >  Hi,
> > > > >
> > > > > >>>
> > > > >
> > > > >
> > > > > >>>
> > > > >  I noticed that by default lldb does not read .debug_frame section to unwind frames but relies instead on .eh_frame .
> > > > >
> > > > > >>>
> > > > >
> > > > >
> > > > > >>>
> > > > >  Is there a way to fallback to reading .debug_frame?
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >>
> > > > >  Not currently. Most compilers (gcc _and_ clang) put the same old stuff in .debug_frame as they do in .eh_frame, so we haven't had to use .debug_frame over .eh_frame yet. What compiler are using that is putting different (more complete) info in .debug_frame vs .eh_frame?
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >>
> > > > >
> > > > >
> > > > > >>
> > > > >  What about about C or C++ program compiled with -fno-exceptions?
> > > > >
> > > > > >>
> > > > >  They will fall back to the UnwindAssembly way even if the .debug_frame is present right?
> > > > >
> > > > > >
> > > > >
> > > > >
> > > > > >
> > > > >  If no EH frame exists for a frame, then we will always fall back to UnwindAssembly. We always use UnwindAssembly for the first frame and for any frame that is past an async interrupt (sigtramp). We use the EH frame/.debug_frame for any non-zero frames, but will always use UnwindAssembly if there is no such info.
> > > > >
> > > > >
> > > > >
> > > > > I want to expand on what Greg said earlier about eh_frame versus debug_frame.
> > > > >
> > > > > Ideally, eh_frame will be the minimal unwind instructions necessary to unwind the stack when exceptions are thrown/caught.  eh_frame will not include unwind instructions for the prologue instructions or epilogue instructions -- because we can't throw an exception there, or have an exception thrown from a called function "below" us on the stack.  We call these unwind instructions "synchronous" because they only describe the unwind state from a small set of locations.
> > > > >
> > > > > debug_frame would describe how to unwind the stack at every instruction location.  Every instruction of the prologue and epilogue.  If the code is built without a frame pointer, then it would have unwind instructions at every place where the stack pointer is modified.  We describe these unwind instructions as "asynchronous" because they describe the unwind state at every instruction location.
> > > > >
> > > > >
> > > > > Instead what we have with gcc and clang is eh_frame instructions that describe the prologue (and some versions of gcc, the epilogue) plus the unwind state at synchronous unwind locations (where an exception can be thrown).  We have a half-way blend of asynchronous and synchronous ... it's "pretty good" but not "guaranteed" from a debugger's perspective.  It would be great if eh_frame was genuinely only the unwind instructions for exception handling and debug_frame had the full unwind state at every instruction and we could depend on debug_frame.  But in reality, the same unwind instructions are put in both eh_frame and debug_frame -- so there's little point in ever reading debug_frame.  lldb does not read debug_frame today, although it would be easy to do so.
> > > > >
> > > > >
> > > > > As an experiment starting late August (r216406), lldb is now trying to use eh_frame for the currently-executing frame.  Even though it isn't *guaranteed* to be accurate at all instructions, in practice it's pretty good -- good enough that gdb seems to be able to live on it.  Tong Shen's patch in r216406 does augment the eh_frame unwind instructions with the epilogue unwind... newer gcc's apparently describe the epilogue in eh_frame but few other compilers do.
> > > > >
> > > > > It's an open question how well living off eh_frame unwind instructions will work with a non-gcc/non-clang compiler.  That's why I say this is an "experiment" - we may have to revert to lldb's UnwindAssembly profiling code for the currently-executing function if this breaks with other compilers.
> > > > >
> > > > > J
> > > > >
> > > > >
> > > > > _______________________________________________
> > > > > lldb-dev mailing list
> > > > > lldb-dev at cs.uiuc.edu
> > > > > http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev
> > > >
> > > >
> > > >
> > > > _______________________________________________
> > > > lldb-dev mailing list
> > > > lldb-dev at cs.uiuc.edu
> > > > http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev
> > >
> > >
> >
> >
> 
> 





More information about the lldb-dev mailing list