Search
Tuesday, 11th of September 2018, 16:23:12 UTC
16:24:04
housel
Another challenge for raspi support would be the necessity of talking to the GPU over the mailbox interface
16:24:37
froggey
does it not have a framebuffer that can be written to directly?
16:24:56
johnjay
ok my qemu on cygwin is saying it supports raspi2 but not raspi3... odd
16:24:57
housel
It does, but you have to ask the GPU to configure it first
16:25:43
johnjay
housel: i tried going to raspberrypi.org for documentation about that, but i found... not much
16:25:59
housel
The GPU is actually the first processor to boot, then it loads the kernel and starts up the ARM
16:26:11
froggey
I think we can have kboot set it up before passing control over
16:26:16
froggey
kboot will also need to be ported
16:26:19
housel
https://github.com/raspberrypi/firmware/wiki/Accessing-mailboxes
16:26:49
froggey
qemu-system-aarch64 -machine virt -cpu cortex-a53 -kernel tools/kboot/kboot-qemu-arm64.elf -serial stdio -m 1G -device virtio-gpu-device -device virtio-keyboard-device -device virtio-mouse-device -drive if=none,file=path/to/disk.vmdk,id=blk -device virtio-blk-device,drive=blk -netdev user,id=vmnic,hostname=qemu -device virtio-net-device,netdev=vmnic
16:27:00
johnjay
froggey: it says kboot boots a linux kernel, you're using a linux kernel for testing?
16:27:04
froggey
^ the qemu command I use for testing
16:27:35
froggey
https://github.com/froggey/kboot
16:27:49
froggey
most of that is just setting up a bunch of devices
16:28:50
johnjay
hrm so waf build system is based on scons
16:33:22
johnjay
no matter i pick qemu-system-arm or qemu-system-aarch64 there is no rasp3 option
16:33:30
johnjay
maybe i need to update my qemu version, i was sure it had rasp3 support
16:34:18
froggey
qemu 3.0 has an rasp3 machine
16:37:42
johnjay
oh, qemu 2 is the latest in the debian buster repo
16:37:49
johnjay
is qemu 3 in development now?
16:38:38
froggey
3 is the latest release
16:38:53
johnjay
ah ok. so you manually compiled it?
16:39:21
johnjay
nice. i probably need to know more tools that are outdated so i can manually compile them
16:39:31
johnjay
even my chat client was outdated on stretch. o_o
18:07:40
froggey
varjag: changing sys.int::*cold-stream* will catch some system output, but not the really low-level stuff
18:08:49
froggey
you can also boot with the video console enabled, that'll blast all output directly onto the screen. that's not really a great option though
18:10:57
varjag
froggey: i've been meaning more like printing my debug messages into the terminal
18:11:09
varjag
rather then redirecting the whole system output
18:13:08
varjag
i want to tweak network-card code a bit to show how you can e.g. capture packet data on a live system
18:13:27
varjag
but if i print to standard-output it would go to serial console
18:16:46
varjag
i just don't see any pretty solution
18:18:43
froggey
I think my lazy-popup-io-stream thing will get you some of the way there
18:19:03
froggey
https://github.com/froggey/Mezzano/blob/master/ipl.lisp#L172
18:20:01
froggey
it automatically creates a repl for any thread that uses the global *terminal-io*, instead of sending output to the serial port
18:21:23
froggey
can be a bit annoying because you end up getting windows for all kinds of stuff, like the keyboard worker complaining that it doesn't understand a certain key
18:22:47
froggey
or you could create your own popup-io-stream and have your messages write directly to it instead of *standard-output*
18:38:44
varjag
ah it wasn't around in demo3 it seems
18:40:18
froggey
oh right, yeah. you should be fine with the regular one if you're outputting directly to it
18:59:32
ebrasca
froggey: I have exported public functions and update http-demo.
19:21:42
froggey
ebrasca: looks ok to me
19:22:29
froggey
do you want to try updating swank to use the new api? then the old *server-alist* stuff can be removed entirely
19:22:36
froggey
I don't think there's anything else using it
19:31:42
ebrasca
I don't like to change swank/slime.
19:44:51
ebrasca
I like to add server for usocket but I don't know how.
2:24:29
leo_song
froggey: how about adding a flag to the thread data structure, so we can filter out the driver threads when doing snapshots?
Wednesday, 12th of September 2018, 4:23:12 UTC