Date: Sun, 21 Jun 2020 11:38:38 +0200 From: =?utf-8?Q?S=C3=B8ren_Schmidt?= <soren.schmidt@gmail.com> To: Emmanuel Vadot <manu@bidouilliste.com> Cc: charlesr@scd-systems.net, freebsd-arm@freebsd.org Subject: Re: Rock64 dwc interface issues Message-ID: <48C101F9-9C1F-46EB-A092-EE903D5070D1@gmail.com> In-Reply-To: <20200619194222.73bc7dd705f6d80e63757796@bidouilliste.com> References: <877d6e2d-6cd8-30f3-08c3-60a2bacb5873@scd-systems.net> <7B4746AD-A8B7-43F0-B5EC-F3812A91F344@gmail.com> <20200619194222.73bc7dd705f6d80e63757796@bidouilliste.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On 19 Jun 2020, at 19.42, Emmanuel Vadot <manu@bidouilliste.com> wrote: >=20 > I've just took the last image for rock64 and rockpro64 to test on my > boards (I usually netboot but testing the image give a common = reference > between everyone). > As a reference the image I've burned on the sd are : > FreeBSD-13.0-CURRENT-arm64-aarch64-ROCK64-20200618-r362292.img.xz > FreeBSD-13.0-CURRENT-arm64-aarch64-ROCKPRO64-20200618-r362292.img.xz >=20 > On Rock64 v2.0 (2017-0713) I see no issues, I got an IP from my dhcp > server and could install packages properly. > Using iperf3 I have ~275Mbits/sec which isn't great but we're aware of > the bad perf of the dwc(4) driver. I checked my build env, and the error was that I didn=E2=80=99t change = the u-boot used (got the pinebookpro version mixed in). Using the correct u-boot makes the dwc net work. However, I did use the correct DTB for the kernel to use (static build = in to be sure), so it should have configured things correctly = shouldn=E2=80=99t it ? I guess we inherit lots of things from u-boot = that does not get initialised by the kernel even if the correct DTB is = in place. Funny thing is that it appeared to work, led=E2=80=99s and all, but no = traffic was coming through. > On RockPro64 v2.1 I also see no issues, iperf3 give the same speed > result for tx but rx is worse (~100Mbits/sec). It the 2.1 version I=E2=80=99ve got btw. > S=C3=B8ren, what are your hack for R40 (which do not have support for = in > term of clock so I'm not surprise that you need "something" for it).=20= I have implemented the clks etc that=E2=80=99s missing from our support, = but instead of mutilating the dwc driver, I=E2=80=99ve hacked up an = older version (before syscon etc went in) and made it work with the R40 = way of things as a new driver. All this is done on a 12-stable tree though, but I guess it could be = moved to -current rather easily. It does make the Bananapi M2 Berry work = quite nicely (except video which I have no use for. -S=C3=B8ren=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?48C101F9-9C1F-46EB-A092-EE903D5070D1>