Re: ARM build slaves (tcpdump mirror in Germany)

[Guy Harris via tcpdump-workers <tcpdump-workers () lists tcpdump org>]

> --- Begin Message ---
>
>
> From: Guy Harris <gharris () sonic net>
>
> Date: Mon, 22 Mar 2021 23:06:43 -0700
>
> On Mar 22, 2021, at 5:35 PM, Denis Ovsienko via tcpdump-workers <tcpdump-workers () lists tcpdump org> wrote:
>
> > On Mon, 22 Mar 2021 19:00:31 +0100
> > Harald Welte <laforge () gnumonks org> wrote:
>
>         ...
>
> > > btw: I'm not sure if qemu full system emulation of e.g. ppc on a
> > > x86_64 hardware would be an option, though.  I think
> > > openbuildservice.org is doing that a lot for building packages on
> > > less popular architectures.
> >
> > QEMU was very useful for the NetBSD setup. NetBSD for some reason did
> > not provide binary packages for 9.1/aarch64, and heavy non-default
> > packages (LLVM, Clang, GCC 10) just do not compile on 1GB RAM of RPI3B
> > (NetBSD release does not run on RPI4B), so the only way to compile
> > these was in a QEMU VM with more RAM.
> >
> > That said, on a Linux host with i7-3770 CPU the QEMU guest measured at
> > 64% core-to-core CPU performance of an RPI3B. So after the initial
> > setup a hardware Pi does a better job.
>
> The main PowerPC/Power ISA buildbot we'd want would probably be ppcle64, as the ppcle64 implementation of some crypto 
> library routines, as used by tcpdump, require strict adherence to the API documentation, e.g. 1) don't use the same 
> buffer for encrypted and decrypted data and 2) provide all the necessary padding in the input buffer and leave enough 
> room in the output buffer, as per
>
>         https://github.com/the-tcpdump-group/tcpdump/issues/814
>
> 64% isn't perfect, but it's a lot better than 10%, so if QEMUs' PPC64/64-bit Power ISA emulation supports both 
> big-endian and little-endian mode, and runs with acceptable performance (anything in the range of 50% is probably good 
> enough), and the emulation is faithful enough (which being able to boot ppc64le Linux would probably imply), that would 
> probably be sufficient.
>
> Having *some* big-endian machine would be useful primarily for tcpdump testing, to make sure there's no code that 
> implicitly assumes it's running on a little-endian machine (which most developers probably have); any of SPARC, ppcbe, 
> or s390/s390x would suffice for that.
>
> SPARC has the additional advantage of trapping on unaligned accesses, so it'll also detect code that implicitly assumes 
> that unaligned accesses work.  S/3x0 hasn't required alignment since S/370 came out (unaligned accesses were an 
> optional feature of S/360, but were made a standard feature in S/370), and I'm not sure PPC requires it.  We already 
> have SPARC/Solaris 10 testing with OpenCSW, so that will fail on unaligned accesses; the only thing additional 
> buildbots would do would be to give us Solaris 11 and Linux.
>
> --- End Message ---
_______________________________________________
tcpdump-workers mailing list
tcpdump-workers () lists tcpdump org
https://lists.sandelman.ca/mailman/listinfo/tcpdump-workers