#asahi-dev · 2021-05-28 07:20
arndright#asahi-dev · 2021-05-04 13:30
arndright, makes sense#asahi-dev · 2021-05-04 12:14
arndI just received a thunderbolt drive to install onto, as the internal drive is already full of my coworker's work stuff#asahi-dev · 2021-05-04 12:13
arndand that in turn would mean I could my machine without needing a serial cable first#asahi-dev · 2021-05-04 12:12
arndmarcan: can the hypervisor already process usb data like keyboard input over cdc-acm while the guest is running? If that works, adding drivers/tty/hvc/hvc_dcc.c earlycon support to the hypervisor should be no more than a few lines#asahi-dev · 2021-05-04 11:56
arndmarcan: that looks pretty cool. Do you keep talking to the debug shell over usb while using the serial port for the guest console, or is both of these the serial port?#asahi-dev · 2021-03-06 13:54
arndkettenis: do you need anything beyond the config space access for PCIe in the kernel today?#asahi-dev · 2021-02-15 10:47
arndmaybe check if any of the people that signed your old key are already on the kernel keyring#asahi-dev · 2021-02-15 10:47
arndfor the account, I don't think there is a way around the minimum three signatures#asahi-dev · 2021-02-15 10:47
arndhaving signatures on the does give a better feeling about it of course#asahi-dev · 2021-02-15 10:46
arndand once I pulled from you once, I would generally also assume it's fine if future pull requests are signed by the same key. The 'one connected signature' requirement is specifically if you send a pull request to torvalds yourself#asahi-dev · 2021-02-15 10:44
arndmarcan: no, I'm happy to apply the patches from email, or by applying with the same amount of care when I get a pull request#asahi-dev · 2021-02-15 10:35
arndmarcan: regarding signatures on the key, the document I linked to says you need at least one signature from someone who is connected to torvalds, but IIRC the requirement for getting a kernel.org account is three such signatures. If you have trouble finding kernel folks to sign your kernel locally, we can probably arrange for me to sign your kernel through some video chat#asahi-dev · 2021-02-15 10:27
arndI assume yubikey works the same#asahi-dev · 2021-02-15 10:24
arndj`ey: review is always by email on the list, the pull request is what you'd send once the series is fully reviewed and ready to be merged. I usually do some final checks on the pull request to make sure the contents match up with the description, and the signature is valid then#asahi-dev · 2021-02-15 10:21
arndj`ey: yes, you just send an email with the subject containing [GIT PULL] instead of [PATCH], and the body generated by 'git request pull', and pointing to a tag with gpg signature and a description#asahi-dev · 2021-02-15 10:19
arndif the gpg key is really old, it might be too short, I think the minimum is now 2048 bits#asahi-dev · 2021-02-15 10:19
arndmarcan: ok either one of those is fine for pull requests, or you keep using separate patches until you have a kernel.org account#asahi-dev · 2021-02-15 10:16
arndmarcan: note that you can ask for a kernel.org account that allows you to host git trees and one email alias there once you are listed in the MAINTAINERS file and have a sufficiently connected gpg key#asahi-dev · 2021-02-15 10:15
arndmarcan: do you have a git tree somewhere other than github.com that you can use for sending pull requests? For the initial merge, separate patches will be fine, but in general, I prefer pull requests, and github is slightly annoying: it's often really slow for pulling kernel trees for some reason, and its generally less trusted than an tree on kernel.org or one you host yourself#asahi-dev · 2021-02-15 10:11
arndj`ey: the process is to send an email to sfr to ask him to add a branch. What I did specifically is to have a 'for-next' branch and I merge all of my branches into that#asahi-dev · 2021-02-15 08:12
arndah, that is Russell's address#asahi-dev · 2021-02-15 08:11
arndI've had problems with arnd@arndb.de recently and started using the other one#asahi-dev · 2021-02-15 08:11
arndno, if you get a bounce, please forward that to me at arnd@kernel.org and arnd@arndb.de (in case one of the two bounces as well)#asahi-dev · 2021-02-15 08:10
arndno worries#asahi-dev · 2021-02-15 08:09
arndmarcan: no, don't cc soc@kernel.org on patches unless I they have been reviewed and are ready to get picked up#asahi-dev · 2021-02-15 08:08
arndhttp://phb-crystal-ball.org/ projects 5.13-rc1 for 2021-05-09, and the official 5.13 for 2021-07-04#asahi-dev · 2021-02-15 08:07
arndtarzeau: the most likely path at the moment is that I'll create a separate branch in https://git.kernel.org/pub/scm/linux/kernel/git/soc/soc.git/ and pull in material for 5.13 as soon as it's been sufficiently reviewed and marcan sends it to soc@kernel.org. It will then show up in https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/ a day later, and in torvalds/linux.git during the 5.13 merge window#asahi-dev · 2021-02-11 18:34
arndVia Wondermedia SoCs had Via network and gpu components with existing PCI drivers, and their BSP also faked a bus. Those were likely real PCI devices,just without the host portion#asahi-dev · 2021-02-11 18:31
arndA lot of the server arm chips are like that as well: close enough to pci to almost make it work but then need some pci vendor/device id specific quirks#asahi-dev · 2021-02-11 18:28
arnds/name/nvme/#asahi-dev · 2021-02-11 18:28
arndI have started patching the name pci driver to separate out some of the pci dependencies, but its not done yet#asahi-dev · 2021-02-05 16:30
arndwhich would mostly come from better cache utilization#asahi-dev · 2021-02-05 16:30
arndnot quite, but it's a noticeable reduction nonetheless. I haven't done memory usage measurements, but did see up to 15% difference in performance on kernel builds#asahi-dev · 2021-02-05 16:28
arndx32 uses numbers 512 through 547 for syscalls that require some compat handling#asahi-dev · 2021-02-05 16:28
arndI still hope we can kill it off before we assign system call number #512, which currently conflicts with x32#asahi-dev · 2021-02-05 16:27
arndAh right, I forgot there was a whole debianport, I thought it was just a few packages for some reason. We tried killing it off twice, but chickened out each time because there are a couple of people that have specific applications (gcc mostly) running that way#asahi-dev · 2021-02-05 16:13
arndilp32 does have a noticeable performance advantage (5% to 20% according to numbers I've seen) over both traditional 32-bit binaries and native 64-bit binaries, but a lot of the applications that do care about performance are the same ones that care about addressing lots of memory and want 64-bit anyway#asahi-dev · 2021-02-05 16:10
arndit's a similar niche to big-endian arm64, which was merged early on, but has a similar set of users, and in restrospect we would have perhaps been better off without that as well#asahi-dev · 2021-02-05 16:09
arndfor arm64, there are still ilp32 patches getting ported to every kernel because there are a handful of users, but it never made it into mainline because that would have required maintaining the ABI indefinitely#asahi-dev · 2021-02-05 16:08
arnddhewg, modwizcode: supporting extra ABIs in a distro is a pain, so almost nothing supports x32 applications, and the ones that do (gentoo, mainly) suffer from running into all the bugs that nobody else noticed#asahi-dev · 2021-01-30 14:34
arndmarcan: are you running with 4KB pages when you get the SError faults? I've been wondering whether there is a register that changes the cache between a mode that is faster on 16KB pages and one that works correctly on 4KB. I was mostly thinking of the L1 cache that appears as though they designed it as a non-aliasing VIPT 8-way set-associative L1d, which would make it aliasing for 4K#asahi-dev · 2021-01-29 17:33
arndqemu only knows a53, a57 and a72, as well a lots of older 32-bit cores, and passing through the host cpu in virtualization#asahi-dev · 2021-01-29 16:25
arndI guess it would be similar to how qemu handles e.g. the TI specific arm925 with its custom registers