- Pharo-dev - lists.pharo.org

Fixing crashes on delivery of SIGIO etc [Was Re: Seg Fault Pharo 7.0.3, Was Difficult to debug VM crash with full blocks and Sista V1]
by Eliot Miranda Oct. 9, 2019

Oct. 9, 2019

Hi Nicolas, ClÃ©ment, et al, On Tue, Oct 8, 2019 at 1:40 PM Nicolas Cellier < nicolas.cellier.aka.nice(a)gmail.com> wrote: > More on the problem that Eliot is speaking about: it can happen in these > conditions: > if is SIGIO is delivered while executing any trampoline transiting from > SmalltalkToCStackSwitch. > What happens here? > > We must load 64bits contents of memory (cStackPointerAddress) to the stack > pointer register (SPReg = $rsp) - See genLoadCStackPointer. > This is done by using the CogRTLOpcode abstract instruction MoveAwR > But we have no matching instruction in IA-32 X64... We can only load a > 64bits memory content in $rax > So the idea is to generate this sequence > (CogX64Compiler>>concretizeMoveAwR): > xchgq %rsp, %rax > movabsq 0x10027c338, %rax ; cStackPointerAddress > xchgq %rsp, %rax > That's clever because it preserves $rax which could be in use when we want > to MoveAwR. > Clever, but in the case of %rsp and %rbp, dead wrong :blush:. The constraint is that x86_64 only provides a 64-bit load from an absolute address into %rax; no other register can be used. > But it has an unfortunate side effect: the stack pointer temporarily gets > the contents of $rax and can thus temporarily point anywhere. > Right; which is unacceptable. It's a tiny window, but one we hit all the time. > What happens when performing Squeak SocketTest is that we use some > trampolines (I guess for invoking primitives for example) and we generate > SIGIO (for some reasons, there are a lot of SIGIO generated on my > particular macos machine, so I can trigger the bug more easily than Eliot). > We previously installed a handler for SIGIO via signal(). When we use > signal(), the handler shares the stack pointer with user program. > If the event is delivered in between the two xchgq instructions above, the > signal handler will then use a corrupted stack pointer pointing anywhere > (depending no contents of $rax) when the VM enter the signal handler > function, it uses stack pointer to save some states, and corrupt a memory > zone, segfault or whatever. > In the case described by the opensmalltalk vm-dev thread, $rax was > pointing to the generated code zone (jitted methodZone), so we corrupted > the generated code and soon get punished for that. But it's probable that > there might be other (rare) occurrences of this bug. > > Not sure if it is causing the bugs described by Sean, but it's important > to use the fix from Eliot ASAP and retry. > There might be other occurrence of signal(SIGIO,forceInterruptCheck) in > minheadless flavour, I did not check if Eliot also corrected it, if not it > should also be corrected ASAP, as should every usage of signal() be > replaced by ussage of sigaction() with appropriate flags to use > sigaltstack() - see Eliot's commit details. > I want to discuss potential fixes with ClÃ©ment and Nicolas, and anyone else interested. So... 1. the straight forward fix is to generate different code for setting %rsp and %rbp. I shall do this very soon. On x86_64 we dedicate a register for code generation purposes, this is called RISCTempReg, and is either %r8 (SysV ABI (unix)) or %r11 (Windows). RISCTempReg is never assumed to be live except within a single instruction sequence. We can swap with this before assigning to %rsp, e.g. xchgq %r8, %rax movabsq 0x10027c338, %rax ; cStackPointerAddress xchgq %r8, %rax movq %r8, %rsp this adds a couple of bytes, but is reliable. 2. one could imagine exchanging TempReg and RISCTempReg, i.e. TempReg would be either %r8 or %r11, and RISCTemptReg would be %rax. That would allow movabsq 0x10027c338, %rax ; cStackPointerAddress movq %rax, %rsp this at least worked on Ryan's MIPS32 back end when it was in use (no one has tested it in as while AFAIA) where TempReg is S5 (r21), RISCTempReg is AT (r1) and CResultRegister is V0 (r2). So this is worth investigating. [I had tried something similar with HPS and it completely broke the code generator so my gut is trying to tell me this will never work; Ryan has proved otherwise]. 3. I would much rather implement some form of tracking whether a particular register is live or not. A trivial implementation would only track TempReg and only track being not live up until the first assignment to TempReg. A more sophisticated approach could deal with control flow branching and merging of the liveness info. I like the trivial approach for now. That would simply track TempReg and only up until the first explicit assignment to TempReg. 4. we could simply treat assignments to SPReg and FPReg specially, knowing that these are done only in trampolines and enilopmarts (these are the pieces of code that sit b between JITted code and the rest of the run-time (interpreter, memory manager, primitives, etc), either to call the runtime from JITted code or, as in this case, to enter machine code from the run-time. I suppose this is OK as long as we document it. This would allow us to generate what we would generate for #2 above, i.e. movabsq 0x10027c338, %rax ; cStackPointerAddress movq %rax, %rsp N. are there better ways? ClÃ©ment, Nicolas (et al), what do you think? Le dim. 6 oct. 2019 Ã 12:36, Eliot Miranda <eliot.miranda(a)gmail.com> a Ã©crit : > Hi Sean, Hi All, >> >> this may be because of the issue described here: >> http://forum.world.st/Difficult-to-debug-VM-crash-with-full-blocks-and-Sist… >> >> This issue is characterized by the system crashing soon after start up >> when some significant i/o is done, typically either to files or sockets. >> It affects macOS only and may indeed affect only 64-bits. We have strong >> evidence that it is caused by the dynamic linker being invoked in the >> signal handler for SIGIO when the signal is delivered while the VM is >> executing JITted code. The symptom that causes the crash is corruption of >> a particular jitted methodâs machine code, eg Delay class>>#startEventLoop, >> and we believe that the corruption is caused by the linker when it >> misinterprets a jitted Smalltalk stack frame as an ABI-compliant stack >> frame and attempts to scan code to link it. >> >> Our diagnosis is speculative; this is extremely hard to reproduce. >> Typically in repeating a crashing run SIGIO may no longer be delivered at >> the same point because any remote server has now woken up and delivers >> results sooner, etc. However, Nicolas Cellier and I are both confident >> that we have correctly identified the bug. >> >> The fix is simple; SIGIO should be delivered on a dedicated signal stack >> (see sigaltstack(2)). I committed a fix yesterday evening and we should >> see within a week or so if these crashes have disappeared. >> >> I encourage the Pharo vm maintainers to build and release vms that >> include >> https://github.com/OpenSmalltalk/opensmalltalk-vm/commit/c24970eb2859a47406… >> asap. >> >> >> Cheers, >> Eliot >> _,,,^..^,,,_ (phone) >> >> On Oct 3, 2019, at 1:24 PM, Sean P. DeNigris <sean(a)clipperadams.com> >> wrote: >> >> Segmentation fault Thu Oct 3 15:52:33 2019 >> >> >> VM: 201901051900 https://github.com/OpenSmalltalk/opensmalltalk-vm.git >> Date: Sat Jan 5 20:00:11 2019 CommitHash: 7a3c6b6 >> Plugins: 201901051900 >> https://github.com/OpenSmalltalk/opensmalltalk-vm.git >> >> C stack backtrace & registers: >> rax 0x0000000124380000 rbx 0x00007ffeebd00050 rcx 0x0000000000468260 >> rdx >> 0x0000000000dd6800 >> rdi 0x0000000124cee5a0 rsi 0x0000000124cee5a0 rbp 0x00007ffeebcffe50 >> rsp >> 0x00007ffeebcffe50 >> r8 0x00007fff3f2cefe5 r9 0x0000000000000b00 r10 0x0000000000006000 >> r11 >> 0xfffffffffcd8d5a0 >> r12 0x0000000000000002 r13 0x0000000035800000 r14 0x00007ffeebd00064 >> r15 >> 0x0000000000002800 >> rip 0x00007fff630f7d09 >> 0 libsystem_platform.dylib 0x00007fff630f7d09 >> _platform_memmove$VARIANT$Haswell + 41 >> 1 Pharo 0x0000000103f52642 >> reportStackState >> + 952 >> 2 Pharo 0x0000000103f52987 sigsegv + 174 >> 3 libsystem_platform.dylib 0x00007fff630fab3d _sigtramp + 29 >> 4 ??? 0x0000058900000a00 0x0 + >> 6085968660992 >> 5 libGLImage.dylib 0x00007fff3f2ce29e >> glgProcessPixelsWithProcessor + 2149 >> 6 AMDRadeonX5000GLDriver 0x000000010db16db1 >> glrATIStoreLevels >> + 1600 >> 7 AMDRadeonX5000GLDriver 0x000000010db52c83 >> glrAMD_GFX9_LoadSysTextureStandard + 45 >> 8 AMDRadeonX5000GLDriver 0x000000010db519bb >> glrUpdateTexture >> + 1346 >> 9 libGPUSupportMercury.dylib 0x00007fff5181279d >> gpusLoadCurrentTextures + 591 >> 10 AMDRadeonX5000GLDriver 0x000000010db5a099 >> gldUpdateDispatch >> + 397 >> 11 GLEngine 0x00007fff3ff72078 >> gleDoDrawDispatchCore + 629 >> 12 GLEngine 0x00007fff3ff16369 >> glDrawArraysInstanced_STD_Exec + 264 >> 13 GLEngine 0x00007fff3ff1625a >> glDrawArrays_UnpackThread + 40 >> 14 GLEngine 0x00007fff3ff6dce1 >> gleCmdProcessor + >> 77 >> 15 libdispatch.dylib 0x00007fff62ec2dcf >> _dispatch_client_callout + 8 >> 16 libdispatch.dylib 0x00007fff62ecea2c >> _dispatch_lane_barrier_sync_invoke_and_complete + 60 >> 17 GLEngine 0x00007fff3fec4b85 >> glFlush_ExecThread + 15 >> 18 Pharo 0x0000000103f4cc62 >> -[sqSqueakOSXOpenGLView drawRect:flush:] + 314 >> 19 Pharo 0x0000000103f4cb22 - >> ... >> >> Smalltalk stack dump: >> 0x7ffeebd14238 M DelaySemaphoreScheduler>unscheduleAtTimingPriority >> 0x10fab3ad0: a(n) DelaySemaphoreScheduler >> 0x7ffeebd14270 M [] in >> >> DelaySemaphoreScheduler(DelayBasicScheduler)>runBackendLoopAtTimingPriority >> 0x10fab3ad0: a(n) DelaySemaphoreScheduler >> 0x1125923f8 s BlockClosure>ensure: >> 0x111e88d30 s >> >> DelaySemaphoreScheduler(DelayBasicScheduler)>runBackendLoopAtTimingPriority >> 0x112590a50 s [] in >> DelaySemaphoreScheduler(DelayBasicScheduler)>startTimerEventLoopPriority: >> 0x111e88e08 s [] in BlockClosure>newProcess >> >> Most recent primitives >> @ >> actualScreenSize >> millisecondClockValue >> tempAt: >> >> >> >> ----- >> Cheers, >> Sean >> -- >> Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html >> >> -- _,,,^..^,,,_ best, Eliot

1 0

[Pharo 8.0] Build #850: FAILURE
by ci-pharo-ci-jenkins2＠inria.fr Oct. 9, 2019

Oct. 9, 2019

There is a new Pharo build available! The status of the build #850 was: FAILURE. Could not extract further issue information from commit message: Update stale.yml - both issues and PRs have the same timeout for stale - moved it up to 250 days / 1 week to close - reduces the actions per hour to 2 Build Url: https://ci.inria.fr/pharo-ci-jenkins2/job/Test%20pending%20pull%20request%2…

1 0

[Pharo 8.0] Build #849: Compiler: Document and simplify #lookupVar:
by ci-pharo-ci-jenkins2＠inria.fr Oct. 9, 2019

Oct. 9, 2019

There is a new Pharo build available! The status of the build #849 was: SUCCESS. The Pull Request #4868 was integrated: "Compiler: Document and simplify #lookupVar:" Pull request url: https://github.com/pharo-project/pharo/pull/4868 Issue Url: https://github.com/pharo-project/pharo/issues/backstop Build Url: https://ci.inria.fr/pharo-ci-jenkins2/job/Test%20pending%20pull%20request%2…

1 0

What is a GlobalIdentifier
by StÃ©phane Ducasse Oct. 9, 2019

Oct. 9, 2019

Hi We have a package SystemIdentification and I would like to know if this is still in use. Here is the comment of the class. I keep global IDs that are used for tracking user activity, e.g. computer id. By default, you should access me by calling #uniqueInstance. Responsibility: The IDs that are important to store onto disk (and shared by all images) should be placed in persistedInformation instance variable. On other hand, if you do not want to store it onto disk, create a new instance variable. I can #loadPreferences and #savePreferences onto a disk. I know computer ID and secret ID. Computer ID is a global UUID that is share among all the images. It is stored on a local disk. Secret ID is use for encrypting information, e.g., class names, method names. You can use #hashForText: method. Collaborators: I do not collaborate with other classes. I only offer the basic IDs for other frameworks. Public API and Key Messages - computerUUID - ensureComputerUUID - hashForText: - loadPreferences - savePreferences. Before using #computerUUID, you should call #ensureComputerUUID. It will update UUID from the disk or stores existing one if it is not stored yet. I behave like this as automatic storing data to disk on image start-up leads to errors. It looks like it is used by the GTEventCollector and I wonder if it is used and useful anymore. Stef -------------------------------------------- StÃ©phane Ducasse http://stephane.ducasse.free.fr http://www.synectique.eu / http://www.pharo.org 03 59 35 87 52 Assistant: Julie Jonas FAX 03 59 57 78 50 TEL 03 59 35 86 16 S. Ducasse - Inria 40, avenue Halley, Parc Scientifique de la Haute Borne, BÃ¢t.A, Park Plaza Villeneuve d'Ascq 59650 France

3 2

[Pharo 8.0] Build #848: Simplify TempVector Offset from IR
by ci-pharo-ci-jenkins2＠inria.fr Oct. 9, 2019

Oct. 9, 2019

There is a new Pharo build available! The status of the build #848 was: FAILURE. The Pull Request #4857 was integrated: "Simplify TempVector Offset from IR" Pull request url: https://github.com/pharo-project/pharo/pull/4857 Issue Url: https://github.com/pharo-project/pharo/issues/SimplifyTempVectorOffset Build Url: https://ci.inria.fr/pharo-ci-jenkins2/job/Test%20pending%20pull%20request%2…

1 0

[ANN] Roadmap progress report. September 2019
by Esteban Lorenzano Oct. 9, 2019

Oct. 9, 2019

Hi all, Since some time we wanted to start with this report to keep you all informed on what we are doing to achieve our goals for the release. After ESUG we took some time to organise and we managed to put this in movement. The idea is to post one the first week of each month (but it may vary, as you see with this one who came a bit later :P) Actions reportes are just dots, since we do not have the time to add a big explanation and we hope you are already involved and more or less know what we are talking about. But do not hesitate to ask if you want more details :) Roadmap progress report. September 2019 Image Code Loading Profiling the time spent to load code Reimplementing a better Read/Write Buffered Stream Improving the performance of Source writing and reading, Improving in at least 20% the time of loading code using Metacello. Spec 2 Pass on SpMillerColumnPresenter to made it of general use. Implement a specific SpMillerLayout to handle Miller columns. Pass on regular input ports (transmissions subsystem) to implement activation ports. Pass on SpStyle subsystem (kind of STON based stylesheets for Morphic). Fix SpGridLayout horizontal cell calculous. (It was broken). Fix SpBoxLayout berderWith and padding. (It was broken). Added contextual menus to widgets. This functionality was completely revamped to make it work in all backends. It was generalised to all widgets (even if not all should use it!) Added focus order to presenters. This is also a full revamp of old existing functionality. it accepts you to change the default order, but it has a default order (previous implementations forced you to do the focusing by hand). Added key binding support for all widgets. This is also a full revamp of old exiting functionality. Spec2+GTK3 callbacks can now be chained. This requires some explanation: In Gtk3 you connect to signals. And you can connect just one callback to one signal. Thankfully we control the callbacks in our side, so we modified signal callbacks to accept to chain other callbacks (then we can connect several actions to one single event, for example "keypress" or âhide"). along with this, we reworked the callback hierarchy. implemented inheritance in Pharo of Gtk3 objects and interfaces (this is needed to enhance control of tables and many others). added autocompletion capabilities to SpCodePresenter. added a mechanism to avoid duplicate signalling between the backend and the adaptors. Fix an important leak on GtkFormView. NewTools Inspector it now implements Miller columns Using TaskIt as backend for stepping. extensions migrated (we are at 10% of the total of migrations needed, we will enhance this in iterative way). Debugger Almost complete revamp of UI. Fix context updating (to allow commands to actually work) Integrate new inspector. Work on new layouts for debugger Make an experimental tool: "playbook", to concentrate playgrounds and test some UI ideas. UFFI Adding Tests Support for kind of literals and support type definitions in literals Documentation Measuring performance issues on large images. TaskIT Implement a prototype of a distributed backend to execute tasks. Implement a communication layer using Rabbit MQ Implement the first version of a process manager for handling multiples images and workers. VM Fixing problem with LibGIT and Windows 1903. Fixing the performance issues of Windows 1903 with File accesses when using solid-state 4k physical sector disks. XCode integration Adding an initial version to detect invalid order in the return from concurrent callbacks. Improving the implementation of TFFI to run 50% faster in callouts. Build of StackVM

7 10

Pharo architecture and design
by niepiekm Oct. 8, 2019

Oct. 8, 2019

Hi everybody, I'd like to say that I am very impressed with how Pharo as a system is designed. I'm impressed with multiple of features it has compared to Squeak 5, its functionality as IDE. I'd like to know how was it designed from the start, how did you go about creating its architecture, document it so that developers could implement it so well. Are there any high/detail/low level design documents on which implementation was carried out? Regards, Marek -- Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

1 0

[Pharo 8.0] Build #847: 2874-clampLowHigh
by ci-pharo-ci-jenkins2＠inria.fr Oct. 8, 2019

Oct. 8, 2019

There is a new Pharo build available! The status of the build #847 was: SUCCESS. The Pull Request #4682 was integrated: "2874-clampLowHigh" Pull request url: https://github.com/pharo-project/pharo/pull/4682 Issue Url: https://github.com/pharo-project/pharo/issues/2874 Build Url: https://ci.inria.fr/pharo-ci-jenkins2/job/Test%20pending%20pull%20request%2…

1 0

Seg Fault Pharo 7.0.3
by Sean P. DeNigris Oct. 8, 2019

Oct. 8, 2019

Segmentation fault Thu Oct 3 15:52:33 2019 VM: 201901051900 https://github.com/OpenSmalltalk/opensmalltalk-vm.git Date: Sat Jan 5 20:00:11 2019 CommitHash: 7a3c6b6 Plugins: 201901051900 https://github.com/OpenSmalltalk/opensmalltalk-vm.git C stack backtrace & registers: rax 0x0000000124380000 rbx 0x00007ffeebd00050 rcx 0x0000000000468260 rdx 0x0000000000dd6800 rdi 0x0000000124cee5a0 rsi 0x0000000124cee5a0 rbp 0x00007ffeebcffe50 rsp 0x00007ffeebcffe50 r8 0x00007fff3f2cefe5 r9 0x0000000000000b00 r10 0x0000000000006000 r11 0xfffffffffcd8d5a0 r12 0x0000000000000002 r13 0x0000000035800000 r14 0x00007ffeebd00064 r15 0x0000000000002800 rip 0x00007fff630f7d09 0 libsystem_platform.dylib 0x00007fff630f7d09 _platform_memmove$VARIANT$Haswell + 41 1 Pharo 0x0000000103f52642 reportStackState + 952 2 Pharo 0x0000000103f52987 sigsegv + 174 3 libsystem_platform.dylib 0x00007fff630fab3d _sigtramp + 29 4 ??? 0x0000058900000a00 0x0 + 6085968660992 5 libGLImage.dylib 0x00007fff3f2ce29e glgProcessPixelsWithProcessor + 2149 6 AMDRadeonX5000GLDriver 0x000000010db16db1 glrATIStoreLevels + 1600 7 AMDRadeonX5000GLDriver 0x000000010db52c83 glrAMD_GFX9_LoadSysTextureStandard + 45 8 AMDRadeonX5000GLDriver 0x000000010db519bb glrUpdateTexture + 1346 9 libGPUSupportMercury.dylib 0x00007fff5181279d gpusLoadCurrentTextures + 591 10 AMDRadeonX5000GLDriver 0x000000010db5a099 gldUpdateDispatch + 397 11 GLEngine 0x00007fff3ff72078 gleDoDrawDispatchCore + 629 12 GLEngine 0x00007fff3ff16369 glDrawArraysInstanced_STD_Exec + 264 13 GLEngine 0x00007fff3ff1625a glDrawArrays_UnpackThread + 40 14 GLEngine 0x00007fff3ff6dce1 gleCmdProcessor + 77 15 libdispatch.dylib 0x00007fff62ec2dcf _dispatch_client_callout + 8 16 libdispatch.dylib 0x00007fff62ecea2c _dispatch_lane_barrier_sync_invoke_and_complete + 60 17 GLEngine 0x00007fff3fec4b85 glFlush_ExecThread + 15 18 Pharo 0x0000000103f4cc62 -[sqSqueakOSXOpenGLView drawRect:flush:] + 314 19 Pharo 0x0000000103f4cb22 - ... Smalltalk stack dump: 0x7ffeebd14238 M DelaySemaphoreScheduler>unscheduleAtTimingPriority 0x10fab3ad0: a(n) DelaySemaphoreScheduler 0x7ffeebd14270 M [] in DelaySemaphoreScheduler(DelayBasicScheduler)>runBackendLoopAtTimingPriority 0x10fab3ad0: a(n) DelaySemaphoreScheduler 0x1125923f8 s BlockClosure>ensure: 0x111e88d30 s DelaySemaphoreScheduler(DelayBasicScheduler)>runBackendLoopAtTimingPriority 0x112590a50 s [] in DelaySemaphoreScheduler(DelayBasicScheduler)>startTimerEventLoopPriority: 0x111e88e08 s [] in BlockClosure>newProcess Most recent primitives @ actualScreenSize millisecondClockValue tempAt: ----- Cheers, Sean -- Sent from: http://forum.world.st/Pharo-Smalltalk-Developers-f1294837.html

4 6

Oct. 8, 2019

> Begin forwarded message: > > From: Norm Green <norm.green(a)gemtalksystems.com> > Subject: [Esug-list] Smalltalk Employment > Date: 8 October 2019 at 18:24:28 CEST > To: esug-list(a)lists.esug.org > > GemTalk has a customer in Canada looking for Smalltalkers in the greater Toronto area. Please email if you know someone who would be interested. > > Norm > > > _______________________________________________ > Esug-list mailing list > Esug-list(a)lists.esug.org > http://lists.esug.org/mailman/listinfo/esug-list_lists.esug.org -------------------------------------------- StÃ©phane Ducasse http://stephane.ducasse.free.fr http://www.synectique.eu / http://www.pharo.org 03 59 35 87 52 Assistant: Julie Jonas FAX 03 59 57 78 50 TEL 03 59 35 86 16 S. Ducasse - Inria 40, avenue Halley, Parc Scientifique de la Haute Borne, BÃ¢t.A, Park Plaza Villeneuve d'Ascq 59650 France

1 0