alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - https://gitlab.freedesktop.org/panfrost - Logs https://freenode.irclog.whitequark.org/panfrost - Discord Discard
BenG83 has joined #panfrost
anarsoul|2 has quit [Ping timeout: 250 seconds]
BenG83 has quit [Ping timeout: 244 seconds]
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
pH5 has quit [Quit: bey]
cwabbott has quit [Remote host closed the connection]
cwabbott has joined #panfrost
cwabbott has quit [Ping timeout: 252 seconds]
<tomeu> robclark: what do you think? do you have such wide differences within the msm driver?
<robclark> tomeu, what was the question?
robclark has quit [Quit: Coyote finally caught me]
cwabbott has joined #panfrost
NeuroScr has quit [Read error: Connection reset by peer]
NeuroScr has joined #panfrost
yawniek has quit [Ping timeout: 244 seconds]
cwabbott has quit [Ping timeout: 252 seconds]
<Lyude> tomeu: hm, using the devicetree platform still seems to make mali_kbase hang when the GPU fails on this vim2: https://paste.fedoraproject.org/paste/e1EmTHtQxdkoUFLwdcr3yw
<Lyude> although the vmf_page thing works fine, gonna merge that in a sec
<Lyude> [ 705.841825] mali d00c0000.gpu: Continuing without Mali regulator control
<Lyude> wonder if it's got something to do with that
cwabbott has joined #panfrost
<Lyude> daniels: will try in a sec, hitting a random segfault in mali_kbase now when it tries to call kbase_reg_read(), and I'm not sure what's causing it o_O
<daniels> Lyude: 'synchronous external abort' generally means you're trying to reach a peripheral which is off via power or clock
<Lyude> figured
<daniels> Lyude: so it makes sense; you'd need to fix the regulator support so the mali actually exists before you can start to interrogate it :P
<Lyude> yeah-that would explain why loading mali with the meson platform fixes this problem if I try to load this after
<daniels> heh, right
<Lyude> good thing I got this vim2 to reboot really fast now :)
<daniels> https://gitlab.collabora.com/daniels/linux/commit/e72a606fa763f9c082ca30089d67d4b8ee80e4e9 has the platform i was using, with EPROBE_DEFER plumbed through from where we get the regulator/clk
<daniels> (it's a pretty brutally hacked-up mess; i wasn't exactly aiming for 'upstreamable')
<Lyude> yeah, neither am I lol
cwabbott has quit [Ping timeout: 252 seconds]
robclark has joined #panfrost
<urjaman> lol
<Lyude> I am quickly being reminded of how much I dislike makefiles
cwabbott has joined #panfrost
* urjaman doesnt remember typing lol here... i suppose i typed that into the wrong window
<Lyude> a-ha
<Lyude> meson platform code fixed, now to try that fix for the hanging issues
<Lyude> ...also narmstrong ^
cwabbott has quit [Ping timeout: 252 seconds]
cwabbott has joined #panfrost
adjtm_ has joined #panfrost
adjtm has quit [Ping timeout: 246 seconds]
<Lyude> daniels: that patch ( https://gitlab.collabora.com/daniels/linux/commit/a08501251a3e948d62d9e4b1ff0f6c7c974707cb ) looks like it's against a fairly old version of mali_kbase
<daniels> Lyude: some would say ancient
cwabbott has quit [Ping timeout: 252 seconds]
<Lyude> hm, not sure it's what I'm looking for either. I've got clock control working now on mali_kbase but I'm still hitting the hang in kbase_pm_poweroff_wait when the GPU hits an error
<Lyude> (@daniels)
<daniels> Lyude: oh ok, that was just to support -EPROBE_DEFER so you wouldn't lose clock/power control just due to bad ordering
<Lyude> by power control do you mean regulator support?
<Lyude> i think that is still broken so maybe it does apply
<Lyude> daniels: what system did you use this on, was it the same kind of exynos soc you sent to me?
<daniels> Lyude: yeah exactly (to both)
<Lyude> daniels: alright
<Lyude> tomeu: btw; have you been getting the GPU Fault 0x00000088 (SHAREABILITY_FAULT) at 0x0000008518962ee0 error on your soc too?
<Lyude> hm, hold on a second
<Lyude> mali d00c0000.gpu: GPU Fault 0x00000080 (DELAYED_BUS_FAULT) at 0x0000008118d60ee0
<Lyude> i think the bus just isn't on
<narmstrong> Lyude: please note I *never* tested the meson platform code, I only copied the code I used on the ARM utgard driver !
<narmstrong> I we had a libmali for the t820 we would have debugged it a while ago
<Lyude> narmstrong: oh yeah i'm aware!
<Lyude> which is why I figured it might be something as silly as that :p
<Lyude> figured you might just be interested since i'm making progress, slowly but surely
<Lyude> narmstrong: are there any hwdocs on how the mali hw is setup at least? or just general hwdocs for the vim2
cwabbott has joined #panfrost
<Lyude> YES, figured out a workaround the gpu reset issue
<Lyude> just, don't turn the gpu off. ever
<Lyude> that makes sense i guess
<Lyude> [alyssa]: i think i am very close, just dealing with some memory access issues now (which is not terribly surprising considering the kbase change)