| + B[] (~Thunderbi@125-239-59-195-fibre.sparkbb.co.nz) | 04:27 | |
| - jacky (QUIT: Quit: ZNC 1.7.1 - https://znc.in) (~ichoquo0A@unaffiliated/webjadmin) | 08:20 | |
| + ichoquo0Aigh9ie (~ichoquo0A@unaffiliated/webjadmin) | 08:39 | |
| + newby (578c1fb7@gateway/web/freenode/ip.87.140.31.183) | 14:15 | |
| newby | Hi everyone, just got my laptop for which thanks :) I am wondering if anyone has got X running not as root? | 14:16 |
|---|---|---|
| * newby -> jsnewby | 14:27 | |
| mntmn | jsnewby: yes kind of. but it needs suid bit on Xorg | 14:43 |
| mntmn | and, welcome! | 14:44 |
| mntmn | jsnewby, chmod +s /usr/local/bin/Xorg | 14:47 |
| jsnewby | lovely, thanks | 14:57 |
| Jookia | the default debian uses a wrapper that's more secure than the suid bit | 14:57 |
| - B[] (QUIT: Ping timeout: 272 seconds) (~Thunderbi@125-239-59-195-fibre.sparkbb.co.nz) | 15:45 | |
| - Jookia (QUIT: Remote host closed the connection) (~Jookia@gateway/tor-sasl/jookia) | 16:03 | |
| + Jookia (~Jookia@gateway/tor-sasl/jookia) | 16:03 | |
| - bkeys (QUIT: Ping timeout: 272 seconds) (~Thunderbi@c-76-17-206-162.hsd1.mn.comcast.net) | 16:31 | |
| mntmn | Jookia, but that works only with a display manager or systemd or? | 17:56 |
| mntmn | hmmmm something recently broke my ath9k pcie card on imx6 | 17:56 |
| mntmn | > [ 2.857215] ath9k 0000:01:00.0: PCI memory region reserve error | 17:56 |
| mntmn | > [ 1.153129] pci 0000:01:00.0: BAR 0: no space for [mem size 0x00020000 64bit] | 17:57 |
| mntmn | i wonder how that can be... there should be 14 MB of pci address space (and it used to work) | 17:57 |
| mntmn | ah i'm realizing only now that one of the USB controllers of imx6qp is connected to PCIe internally | 18:02 |
| - jsnewby (QUIT: Ping timeout: 256 seconds) (578c1fb7@gateway/web/freenode/ip.87.140.31.183) | 18:03 | |
| mntmn | what the hell, riscv chips by nxp? https://open-isa.org/ | 18:39 |
| mntmn | RV32M1, seems to be ARM/RISCV hybrid | 18:42 |
| + erlehmann (~erlehmann@95.91.214.125) | 20:24 | |
| swivel | is Xorg not using the modesetting driver on reform? | 20:47 |
| Jookia | mntmn: i think that's just default, idk | 20:48 |
| mntmn | swivel, it is, why? | 20:52 |
| mntmn | swivel, KMS needs root to work | 20:53 |
| swivel | mntmn: no it doesn't | 20:57 |
| swivel | not generally | 20:57 |
| swivel | mntmn: are you saying KMS on reform requires root for some unique reason? | 20:59 |
| swivel | my project rototiller has a drm backend and I run it regularly as a normal user from the console on i915, it just requires read/write access to the device (/dev/dri/card0) which on debian is just a matter of adding my user to the video group | 21:01 |
| swivel | my Xorg also runs as a regular user | 21:02 |
| - erlehmann (QUIT: Quit: Leaving) (~erlehmann@95.91.214.125) | 21:15 | |
| mntmn | swivel, i looked into this before, i’ll dig out the reason for you | 21:18 |
| swivel | ok, one of the core reasons we have kms is to stop running X as root so I'm rather surprised | 21:21 |
| mntmn | swivel, drmSetMaster | 21:24 |
| swivel | doesn't require root when uncontended | 21:25 |
| mntmn | https://bugs.freedesktop.org/show_bug.cgi?id=90079 | 21:26 |
| mntmn | uncontended? | 21:26 |
| swivel | that bug is talking about multiple drm programs running at once which both want to be master | 21:27 |
| swivel | drm special cases the lone drm master to not require root | 21:27 |
| swivel | since the expected use of drm is console->X or console->wayland then you just live under those multiplexers, the usual case should not require root. | 21:28 |
| swivel | but Xorg is even nice enough to drop the master when you switch to a virtual console while Xorg is still running | 21:29 |
| swivel | so you can run drm programs from the console without even exiting X, but you must be careful to exit them before switching back to Xorg (or if they drop master cooperatively, it should be ok) | 21:30 |
| mntmn | hmm i wonder why it didn’t work then. i’ll look into it after i’ve unbroken pcie _again_ | 21:30 |
| mntmn | it’s a nightmare | 21:30 |
| swivel | yeah i don't mean to make this a priority, it just sounds wrong :) | 21:30 |
| mntmn | yeah, good point | 21:31 |
| swivel | another kernel pcie breakage? | 21:31 |
| mntmn | something broke a part of BAR allocation mechanism | 21:31 |
| mntmn | so it complains about not having space for the device which is laughable, it’s like 128kB | 21:31 |
| swivel | nxp should have you guys on payroll | 21:31 |
| mntmn | (and it worked before) | 21:31 |
| mntmn | yeah for testing ;) | 21:32 |
| andrej235 | *makes a note | 22:20 |
| mntmn | PCIe card BAR cannot allocate because of type_mask mismatch | 22:57 |
| swivel | BTW I just remembered something about root-less drm without systemd - the special case in the kernel is that the *first* attempt to get drm master succeeds without root | 23:08 |
| mntmn | ok... maybe the console already does something there on imx, don’t know | 23:14 |
| swivel | that first-open problem might have been i915-specific, I'm trying to find the email exchange I had with the kernel devs about that. I was experimenting with running drm clients from an interactive initrd, so no systemd-logind etc. Shit would work the first run, but stopped working after every subequent run - not even concurrent, sequential runs | 23:31 |
| swivel | they were kind of like *shrug* it's not a use case we care much about | 23:32 |
| mntmn | yeah | 23:32 |
| swivel | as long as plymouth works for them they don't care | 23:32 |
| mntmn | uhh i found a weird thing | 23:32 |
| mntmn | for some reason this quirk is applied here https://patchwork.kernel.org/patch/10666445/ | 23:33 |
| mntmn | but imx6qp doesn’t have a usb3 controller, the quirk matches on the imx6qp pcie controller | 23:33 |
| mntmn | and converts it wrongly into a usb controller device | 23:34 |
| swivel | nice | 23:34 |
| mntmn | just reverted that... lets see if that was the issue | 23:35 |
| mntmn | omg indeed | 23:35 |
| mntmn | what the actual hell | 23:35 |
| mntmn | how can stuff like this go unnoticed for 6 weeks | 23:36 |
| swivel | lack of use | 23:37 |
| Jookia | mntmn: that fixes pcie? | 23:38 |
| mntmn | Jookia yes | 23:38 |
| Jookia | i'll pass it on to someone that has an issue with pcie on another chipset for the same amount of time you have | 23:39 |
| mntmn | commit comes from synopsys. maker of the pcie controller | 23:39 |
| mntmn | this is the one to revert https://github.com/torvalds/linux/commit/03e6742584af8866ba014874669d0440bed3a623 | 23:43 |
| swivel | mntmn: it's odd that a usb-controller-less imx6qp would match on PCI_DEVICE_ID_SYNOPSYS_HAPSUSB3* | 23:56 |
| mntmn | swivel, maybe they reused the device id | 23:57 |
| mntmn | yes | 23:58 |
| mntmn | 0xabcd | 23:58 |
| mntmn | what a nice id. also the id of the pcie controller in imx6qp | 23:58 |
| swivel | sigh | 23:58 |
| swivel | you feeling like the only one using imx6qp w/mainline yet? | 23:59 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!