Date: Thu, 2 Mar 2017 00:56:31 -0800 From: Mark Millard <markmi@dsl-only.net> To: FreeBSD PowerPC ML <freebsd-ppc@freebsd.org> Subject: Re: Want info: most recent known-working TARGET_ARCH=powerpc head -r version usable on the likes of the PowerMac G5 so-called "Quad Core"? [ingore] Message-ID: <8DFA5D9B-66F9-4818-97E4-564D914C140D@dsl-only.net> In-Reply-To: <CA8AB47E-CED9-4F42-B67D-0FB6597667F1@dsl-only.net> References: <CA8AB47E-CED9-4F42-B67D-0FB6597667F1@dsl-only.net>
next in thread | previous in thread | raw e-mail | index | archive | help
[Turns out that apparently clang 3.9.x based builds got worse.] On 2017-Mar-1, at 7:48 PM, Mark Millard <markmi at dsl-only.net> wrote: > I'm looking for reports of when TARGET_ARCH=powerpc > builds last were able to boot the likes of PowerMac > G5 so-called "Quad Core"s. Other PowerMac G5's should > be a fit. I may need to do a bisect in part based on > this information. > > > Details of the context for the question follow. > > The only powerpc family member that I currently have > access to is a PowerMac G5 so-called "Quad Core". > > [Note: After -r313254 (or a little more) locking has been > worked on and TARGET_ARCH=powerpc64 has been messed up > until the fixes in -r314474. I've done nothing with > TARGET_ARCH=powerpc over this time but did some of the > powerpc64 testing.] > > My intent was to test the new locking for > TARGET_ARCH=powerpc but I did not get that far. > > Note: My context is clang 3.9.1 experiments for > TARGET_ARCH=powerpc head builds. But I'm happy with > information based on gcc 4.2.1 . I'd be surprised if > many folks are trying clang given its powerpc problems. > > I tried to build and install TARGET_ARCH=powerpc. > > This used to boot (with one oddity) but I've not > attempted it since something like 2016-Nov sometime > --or even before. > > When I tried -r314479 for TARGET_ARCH=powerpc it failed > (manually transcribed): > > > exec /sbin/init: error 13 > > fatal kernel trap: > > exception = 0x600 (alignment) > ssr0 = 0x88a754 > ssr1 = 0x9032 > lr = 0x88a548 > curthread = 0x1477a20 > pid = 1, comm = kernel > > [ thread pid 1 tid 100002 ] > Stopped at sf_bug_alloc+0x24c: stw r5,0x4(r3) > db:0:kdb.enter.default> bt > Tracing pid 1 tid 100002 td 0x1477120 > 0xdf5e06c0: at exec_map_first_page+0x510 > 0xdf5e0750: at kern_execve+0x3e4 > 0xdf5e0990: at sys_execve+0x6c > 0xdf5e09f0: at start_init+0x2d0 > 0xdf5e0a50: at fork_exit+0xcc > 0xdf5e0a80: at fork_trampoline+0xc > db> > > > [Note this is before the PowerMac can accept input to the > db> prompt. I have a compiled-in script automatically run > to produce a bt when ddb is started. No serial console > context is set up.] > > Sometimes it hangs without showing the "fatal kernel trap" > and later text but is does report the "error 13". > > I've tried both a amd64 -> TARGET_ARCH=powerpc cross build > and building on powerpc64 with TARGET_ARCH=powerpc. Same > basic results both ways. The above is from the amd64 based > cross build. I tried building a gcc 4.2.1 based TARGET_ARCH=powerpc build and that worked fine for the same svn revision. Apparently clang now does something new compared my prior TARGET_ARCH=powerpc experiments. === Mark Millard markmi at dsl-only.net
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?8DFA5D9B-66F9-4818-97E4-564D914C140D>