Date: Wed, 13 Feb 2019 23:10:09 +0000 From: John F Carr <jfc@mit.edu> To: "kraileth@elderlinux.org" <kraileth@elderlinux.org>, "freebsd-arm@freebsd.org" <freebsd-arm@freebsd.org> Subject: Re: ThunderX support broken since r336520 Message-ID: <9440FF73-4B4C-4884-B001-CD8A1402F69C@exchange.mit.edu> In-Reply-To: <20190213224058.52402wm0s4ube0ve@webmail.omc.net> References: <20190213224058.52402wm0s4ube0ve@webmail.omc.net>
next in thread | previous in thread | raw e-mail | index | archive | help
> On Feb 13, 2019, at 16:40 , kraileth@elderlinux.org wrote: >=20 > Recently I got my hands on a Cavium ThunderX server for some tests before= it is deployed for production. I wanted to install FreeBSD on it so I grab= bed a 12.0 image and was surprised that it didn't work at all. The beastie = menu is displayed, the kernel loaded - and then it just stops. >=20 > As pointed out by Marcel Flores in December (https://lists.freebsd.org/pi= permail/freebsd-arm/2018-December/019130.html), 11.2 does boot to the insta= ller however. Upgrading the system to 11-STABLE worked fine, too. Building = 12-STABLE or -CURRENT resulted in a kernel that wouldn't boot. >=20 > I decided to find the point where ThunderX support broke in head. It's ro= ughly 40,000 commits from the first 12-CURRENT to today. After a few nights= of checking out various old revisions of HEAD and building kernels over an= d over again, I found out that in r336520 Emmanuel Vadot added vt_efifb to = the GENERIC kernel config on arm64. According to his commit message this wa= s tested on PINE64 hardware - but unfortunately it obviously does not work = on ThunderX. In vt_efifb_probe I see there's a hook to disable: TUNABLE_INT_FETCH("hw.syscons.disable", &disabled); if (disabled !=3D 0) return (CN_DEAD);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9440FF73-4B4C-4884-B001-CD8A1402F69C>