*** fitzsim <fitzsim!~user@69-165-165-189.dsl.teksavvy.com> has quit IRC | 07:12 | |
openpowerbot | [slack] <Benjamin Herrenschmidt> @joel So your latest buildroot builds a kernel along with the buildroot | 12:07 |
---|---|---|
openpowerbot | [slack] <Benjamin Herrenschmidt> which when flashed on my arty a7-100t with the latest microwatt from git.. dies and panics at boot | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> last messages are | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> ```[ 0.404225] 65536 pages RAM | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> [ 0.407013] 0 pages HighMem/MovableOnly | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> [ 0.410957] 10545 pages reserved | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> [ 0.414213] Kernel panic - not syncing: vīŋŊD | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> [ 0.418432] Rebooting in 10 seconds..``` | 12:08 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I don't have time to dig into it before I come back from Plumbers I think (will be flying out on Tuesday and have a ton of things to finish) | 12:09 |
openpowerbot | [slack] <joel> There's a device tree patch that was merged in 5.19 that fixes it | 12:10 |
openpowerbot | [slack] <joel> I put it in buildroot but I couldn't get it to apply during the build. It works if you rebuild the kernel | 12:10 |
openpowerbot | [slack] <Benjamin Herrenschmidt> upstream is happier indeed | 12:11 |
openpowerbot | [slack] <joel> https://github.com/shenki/buildroot/commit/81056b369bb236e82a343294523820411a729d77 | 12:11 |
openpowerbot | [slack] <Benjamin Herrenschmidt> it's me or litesdram has become faster ? | 12:13 |
openpowerbot | [slack] <Benjamin Herrenschmidt> or Paul just made the core much faster đ | 12:13 |
openpowerbot | [slack] <joel> All of the above | 12:15 |
openpowerbot | [slack] <joel> Litedram was a fair bit faster when Anton regenerated it recently | 12:15 |
openpowerbot | [slack] <joel> But the core has also been getting faster, and using less resources, thanks to Paul | 12:15 |
openpowerbot | [slack] <Benjamin Herrenschmidt> yeah ... I still get one timing violation from Vivado generating for the 100T | 12:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I might try side-installing a newer Vivado just out of curiosity, see if it's better or not | 12:19 |
openpowerbot | [slack] <Benjamin Herrenschmidt> what would be good is if somebody could be bothered turning the L2 cache into a proper writeback one đ | 12:20 |
openpowerbot | [slack] <Benjamin Herrenschmidt> what are we missing to run a distro ? VMX/VSX ? | 12:21 |
openpowerbot | [slack] <Benjamin Herrenschmidt> we should look into a fast SW emu for this using a custom interrupt vector and a bit of HW helpers | 12:22 |
openpowerbot | [slack] <Paul Mackerras> both, the litesdram helped and eliminating the pipeline bubble we used to get between LS and other instructions helped too | 13:08 |
openpowerbot | [slack] <Paul Mackerras> I did have enough vmx/vsx implemented to boot fedora at one stage | 13:09 |
openpowerbot | [slack] <Benjamin Herrenschmidt> in hardware ? fun | 13:14 |
openpowerbot | [slack] <Benjamin Herrenschmidt> hah https://github.com/paulusmack/microwatt/commits/vecvsx | 13:16 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I think if we could implement the basic ops and microcode the rest ... | 13:16 |
openpowerbot | [slack] <Benjamin Herrenschmidt> @joel you've used gpio-mmio successfully with microwatt gpio ? | 13:36 |
openpowerbot | [slack] <Benjamin Herrenschmidt> we should put it in the dts.... | 13:37 |
openpowerbot | [slack] <Benjamin Herrenschmidt> well actually we really need to start having sdram_init populate the DT based on the design | 13:37 |
*** fitzsim <fitzsim!~user@69-165-165-189.dsl.teksavvy.com> has joined #microwatt | 13:37 | |
openpowerbot | [slack] <Matt Johnston> worked for me with an added `compatible` string, not sure if the DT can do without that. https://github.com/mkj/linux/commit/c5e562f572c0ac7b499175d2b6bc5fbe3dbd494a https://github.com/mkj/linux/commit/6f24d1f952878778536ac7fc2b51d0b0e76cccb2 | 13:40 |
openpowerbot | [slack] <Benjamin Herrenschmidt> which means we should probably start adding a SYSCTL bit for GPIO | 13:40 |
openpowerbot | [slack] <Matt Johnston> guess it'll need that when it gets interrupts too | 13:40 |
openpowerbot | [slack] <Benjamin Herrenschmidt> yuck | 13:41 |
openpowerbot | [slack] <Benjamin Herrenschmidt> the driver doesn't have a more generic compatible string... that's ... backwards | 13:41 |
openpowerbot | [slack] <Benjamin Herrenschmidt> yeah and platform_device's created from the device-tree don't have a name that we can control to match | 13:48 |
*** fitzsim <fitzsim!~user@69-165-165-189.dsl.teksavvy.com> has quit IRC | 13:48 | |
openpowerbot | [slack] <Benjamin Herrenschmidt> somebody should submit a patch upstream for this đ | 13:49 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I'm thinking of plonking a simple SPI controller to play with random HW devices in there, probably should do an i2c one too at some point | 13:50 |
openpowerbot | [slack] <Benjamin Herrenschmidt> there are a few that have open designs and drivers upstream ... but they seem to all suck in different ways đ | 13:50 |
openpowerbot | [slack] <Benjamin Herrenschmidt> I might re-implement one that happens to be compatible with an existing driver, we'll see | 13:50 |
openpowerbot | [slack] <Benjamin Herrenschmidt> if I have time đ | 13:50 |
*** fitzsim <fitzsim!~user@69-165-165-189.dsl.teksavvy.com> has joined #microwatt | 14:53 | |
*** lkcl <lkcl!lkcl@freebnc.bnc4you.xyz> has quit IRC | 20:55 | |
*** lkcl <lkcl!lkcl@freebnc.bnc4you.xyz> has joined #microwatt | 21:09 | |
*** toshywoshy <toshywoshy!~toshywosh@ptr-377wf33o3bnthuddmycb.18120a2.ip6.access.telenet.be> has quit IRC | 22:58 | |
*** toshywoshy <toshywoshy!~toshywosh@ptr-377wf33o3bnthuddmycb.18120a2.ip6.access.telenet.be> has joined #microwatt | 22:59 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!