| *** tpb <[email protected]> has joined #openrisc | 00:00 | |
| *** knz <knz!~kena@2001:41d0:a:f6e9::1> has quit IRC (*.net *.split) | 04:28 | |
| *** knz <knz!~kena@2001:41d0:a:f6e9::1> has joined #openrisc | 04:28 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has quit IRC (Ping timeout: 240 seconds) | 05:05 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has joined #openrisc | 05:15 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has quit IRC (Ping timeout: 240 seconds) | 16:01 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has joined #openrisc | 16:04 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has quit IRC (Ping timeout: 240 seconds) | 16:47 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has joined #openrisc | 17:05 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has quit IRC (Ping timeout: 240 seconds) | 19:17 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has joined #openrisc | 19:18 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has quit IRC (Ping timeout: 240 seconds) | 19:47 | |
| *** littlebobeep <littlebobeep!~alMalsamo@gateway/tor-sasl/almalsamo> has joined #openrisc | 19:47 | |
| shorne | zx2c4: ok, let me look at the math test failing, adding a second uart should be relatively easy, could you use a network interface instead why uart? | 20:32 |
|---|---|---|
| zx2c4 | shorne: network interface not a viable option | 20:33 |
| zx2c4 | also: openrisc is the only platform that doesnt enable me to add a second serial port | 20:33 |
| zx2c4 | (since it doesnt have pci/mmio/virtio/etc) | 20:33 |
| shorne | I was actaully planning to add a virtio platform for the or1k target in qemu, I got side tracked with a verilog bug though | 20:34 |
| shorne | It might be a better option than to add it to the or1ksim platform | 20:38 |
| zx2c4 | so long as it can work with VIRTIO_MMIO, works for me | 20:40 |
| zx2c4 | shorne: specifically, https://א.cc/zB5uLaoi/diff is the change that should be required to use it | 20:41 |
| shorne | I see | 20:43 |
| zx2c4 | alternativly, just add support for a second `-serial ...` -- seems like way less work :) | 20:43 |
| zx2c4 | i can probably write the qemu patch for that myself | 20:43 |
| zx2c4 | maybe ill do that when im back home tonight | 20:44 |
| zx2c4 | the arithmetic bug, otoh, is way more concerning... and i dont know how to fix that | 20:44 |
| shorne | I tried to reproduce it, so far I don't see anything popping up for the curve25519 tests | 20:46 |
| shorne | I see thing like | 20:46 |
| shorne | [ 11.050000] alg: self-tests for drbg_pr_hmac_sha256 (stdrng) failed (rc=-22) | 20:46 |
| zx2c4 | sounds like that's another one? | 20:49 |
| zx2c4 | might want to make sure youre actually running the tests... | 20:49 |
| zx2c4 | shorne: https://א.cc/ijLK3IfJ/diff with something like that | 20:50 |
| shorne | https://gist.github.com/stffrdhrn/a7add7242658679c8b7c943de345fb81 | 20:50 |
| shorne | yeah, it might not be running the curve25519 test, I tried to make sure I enabled that one | 20:51 |
| zx2c4 | shorne: https://א.cc/nQVjoFMu | 20:51 |
| zx2c4 | shorne: enable CONFIG_WIREGUARD=y | 20:51 |
| zx2c4 | https://א.cc/qmm8zCf4 differen tests fail on different runs | 20:52 |
| zx2c4 | usually the failures are exactly a second apart... | 20:52 |
| zx2c4 | so likely interrupt related | 20:52 |
| zx2c4 | https://א.cc/VcD46tal | 20:53 |
| zx2c4 | https://א.cc/oONXmISR | 20:53 |
| zx2c4 | https://א.cc/8yXAsXTX | 20:54 |
| zx2c4 | anyway you get the idea | 20:54 |
| shorne | I see, yes, I am not hitting even the printf, Ill have to look at that | 20:54 |
| shorne | it is compiling the curve25519-selftest.o but not sure why its not running | 20:55 |
| shorne | I need to go | 20:55 |
| zx2c4 | shorne: do not enable CONFIG_CRYPTO_MANAGER_DISABLE_TESTS | 20:56 |
| zx2c4 | and make sure CONFIG_WIREGUARD=y, not =m | 20:56 |
| shorne | ok I see [ 0.230000] SARU running 25519 tests | 20:57 |
| shorne | but no failures | 20:57 |
| zx2c4 | shorne: cool. Now run it a bunch of times | 20:58 |
| zx2c4 | https://xn--4db.cc/JZujg1aZ here's a config file that im using | 20:58 |
| zx2c4 | -cpu or1200 -machine or1k-sim -smp 1 -m 256M | 20:59 |
| zx2c4 | no initramfs, so using https://א.cc/rHI6JaCF/diff | 20:59 |
| zx2c4 | using this compiler https://musl.cc/or1k-linux-musl-cross.tgz | 21:00 |
| shorne | zx2c4: ok, I fixed the sha256 error, it was just because I didnt turn on hmac(sha256) | 22:32 |
| shorne | ok, I will try to reproduce with what you have, then check the diff, I am using this compiler: https://github.com/openrisc/or1k-gcc/releases/tag/or1k-12.0.1-20220210-20220304 | 22:35 |
| shorne | the or1k-linux- build | 22:36 |
| shorne | I tried the musl toolchain, and tried rebooting qemu several times but cannot reproduce | 22:47 |
| shorne | I am using qemu with external initrd | 22:48 |
| shorne | QEMU emulator version 6.2.50 (v6.2.0-1769-g94c71f14e9-dirty) | 22:48 |
| shorne | qemu-system-or1k -cpu or1200 -M or1k-sim -kernel /home/shorne/work/linux/vmlinux -initrd /home/shorne/work/linux/initramfs.cpio.gz -serial mon:stdio -nographic -gdb tcp::10001 -m 64 | 22:48 |
| shorne | let me try with your exact settings, I don't think it should be much different | 22:48 |
| zx2c4 | shorne: im using qemu 7 | 22:53 |
| zx2c4 | i can also repro with master | 22:53 |
| zx2c4 | shorne: just emailed you the shutdown/reboot patch for qemu | 22:53 |
| zx2c4 | now gonna work on serial port | 22:53 |
| shorne | the patch looks fine, ill try to reproduce after updating qemu | 22:58 |
| shorne | I might not have a lot of time left today, have a meeting now, then have to drive | 22:58 |
| shorne | thanks for helping | 22:58 |
| zx2c4 | aw darn | 23:02 |
| zx2c4 | shorne: okay i emailed ya the serial port thing too | 23:21 |
| zx2c4 | https://git.zx2c4.com/wireguard-linux/commit/?h=jd/openrisc | 23:22 |
| tpb | Title: wireguard-linux - WireGuard for the Linux kernel (at git.zx2c4.com) | 23:22 |
| zx2c4 | shorne: i cant reproduce the arithmetic error with CONFIG_PREEMPT | 23:32 |
| zx2c4 | shorne: i cant reproduce the arithmetic error with CONFIG_PREEMPT_NONE=y, i mean | 23:32 |
| zx2c4 | I can reproduce it with CONFIG_PREMPT=y | 23:32 |
| zx2c4 | so maybe the interrupt code is wrong | 23:32 |
| zx2c4 | some registers not being restored right | 23:32 |
| zx2c4 | Another data point, I am unable to reproduce with CONFIG_SMP=y | 23:43 |
| zx2c4 | so, this requires CONFIG_PREEMPTION=y and CONFIG_SMP=n | 23:43 |
| zx2c4 | so this looks like a kernel bug | 23:43 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!