Re: [PATCH v4 00/10] linux-user: Detect and report host crashes

[Richard Henderson]

On 9/19/23 10:26, Michael Tokarev wrote:
> 19.09.2023 11:00, Helge Deller wrote:
> ..
> > > Probably the right solution is to use qemu_abort() (and qemu_assert() etc),
> > > and maybe #define abort(x) qemu_abort(x).  Even if some way to redefine
> > > abort like the above will work on glibc, it does not mean it will work
> > > on *bsd and in other contexts.
> >
> > True. That's probably the better solution.
>
> That wont work, since abort() gets called from a lot of libraries
> (gilbc has 1000s of calls to it)
>
> Sigh.
>
> /mjt

A possible solution that occurs to me is to treat SIGABRT like patch 7 of this patch set 
treats SIGPROF: remap the guest signal to one of the host RT signals.

Then we leave the host SIGABRT as SIG_DFL, producing the expected crash when the signal 
originates from a host abort() (etc).  A guest abort() would use a different signal which 
is caught and emulated.

Things do get confusing across processes, but should be no worse than any of the existing 
signal number swizzling.

Thoughts?


r~