On Tue, Jul 9, 2019 at 7:32 PM Thomson Tan <lilo...@gmail.com> wrote:
> Thanks a lot for the details. For a single isolate, how is the case for > multiple nested JSEntry and CEntry handled? > Stack frames by definition form a sequence, there is no way to nest them. Ranges of C++ frames are bounded by a CEntry frame on one side and a JSEntry from the other, that's how the stack walker can skip over them. > For EntryFrame::GetCallerState, seems it requires to use stack walker > provided by V8, like JavaScriptFrameIterator. How does it work for other > stackwalker which cannot invoke V8 code directly, for example, stack walk > in postmortem dump file and ETW? > I don't know. > On Tuesday, July 9, 2019 at 8:52:34 AM UTC-7, Jakob Kummerow wrote: >> >> On Tue, Jul 9, 2019 at 5:20 PM Thomson Tan <lil...@gmail.com> wrote: >> >>> The stack iterator needs to be able to skip over C++ frames on the stack >>>> >>> Is this a general requirement for V8 generated code for all platforms? >>> >> >> Yes. >> >> >>> Use x64 for example which walks the stack by following frame pointer >>> (rbp) chain, does this mean JS entry function should stop this frame >>> pointer chain, and store the caller's frame pointer in isolate? >>> >> >> The other way round. The caller of JSEntry is a C++ function, and we make >> no assumptions about the stack layout that the C++ compiler generates for >> those. The JSEntry stub creates a new chain of rbp pointers (throughout the >> following invocations of generated code). It stores the last C entry frame >> position, so when walking the stack in the other direction, the stack >> walker can jump from the JSEntry frame to the previous CEntry frame (if >> there is one). >> >> >>> How to usually handle the frame for JS entry function if a full >>> stackwalk is needed? >>> >> >> See class EntryFrame in frames.h, in >> particular EntryFrame::GetCallerState, and compare it with what >> Generate_JSEntryVariant in builtins-x64.cc does. >> >> >>> On Tuesday, July 9, 2019 at 2:09:20 AM UTC-7, Jakob Kummerow wrote: >>>> >>>> The stack iterator needs to be able to skip over C++ frames on the >>>> stack. Information about the topmost C++ section is stored on the isolate >>>> (where else would you store it?), information about additional sections >>>> further down needs to be saved elsewhere. Storing it on the stack makes it >>>> easy for the stack walker to find it. >>>> >>>> On Tue, Jul 9, 2019 at 10:21 AM Thomson Tan <lil...@gmail.com> wrote: >>>> >>>>> Seems the JSEntry frame (Generate_JSEntryVariant) >>>>> saves IsolateAddressId::kCEntryFPAddress from isolate. What is this C >>>>> entry >>>>> FP field for? I guess it saves frame pointer passed from C caller, but >>>>> don't understand why it comes from isolate. >>>>> >>>>> -- >>>> >>>> -- > -- > v8-users mailing list > v8-users@googlegroups.com > http://groups.google.com/group/v8-users > --- > You received this message because you are subscribed to the Google Groups > "v8-users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to v8-users+unsubscr...@googlegroups.com. > To view this discussion on the web visit > https://groups.google.com/d/msgid/v8-users/5b35aac3-dd64-4cb9-a54d-f29df130c364%40googlegroups.com > <https://groups.google.com/d/msgid/v8-users/5b35aac3-dd64-4cb9-a54d-f29df130c364%40googlegroups.com?utm_medium=email&utm_source=footer> > . > For more options, visit https://groups.google.com/d/optout. > -- -- v8-users mailing list v8-users@googlegroups.com http://groups.google.com/group/v8-users --- You received this message because you are subscribed to the Google Groups "v8-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to v8-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/v8-users/CAKSzg3QT%3DAO_%3Dj%2BRD1b%3DTYbu9bGw-v1%3Dqpzye60cO4DGDwfBkg%40mail.gmail.com. For more options, visit https://groups.google.com/d/optout.
