#
…

2021-02-22

00:05mxw39 has joined #asahi
00:10Tokamak has quit [Ping timeout: 246 seconds]
00:10skg has quit [Quit: 〜バイバイ〜!]
00:12skg has joined #asahi
00:13Tokamak has joined #asahi
00:17ky0ko has joined #asahi
00:29skg has quit [Quit: 〜バイバイ〜!]
00:30skg has joined #asahi
00:32n1v has joined #asahi
00:33n1v has quit [Client Quit]
00:34n1v has joined #asahi
00:35n1v has quit [Client Quit]
00:36niv has quit [Ping timeout: 258 seconds]
00:36skg has quit [Remote host closed the connection]
00:37niv has joined #asahi
00:38skg has joined #asahi
00:38niv has quit [Client Quit]
00:38skg has quit [Remote host closed the connection]
00:38niv has joined #asahi
00:39skg has joined #asahi
00:39niv has quit [Remote host closed the connection]
00:40niv has joined #asahi
00:42niv has quit [Client Quit]
00:42niv has joined #asahi
00:44BaughnLogBot has quit [Ping timeout: 240 seconds]
00:44BaughnLogBot has joined #asahi
00:44niv has quit [Client Quit]
00:45niv has joined #asahi
00:47niv has quit [Client Quit]
00:47niv has joined #asahi
00:49niv has quit [Client Quit]
00:49niv has joined #asahi
00:51niv has quit [Client Quit]
00:51niv has joined #asahi
00:52niv has quit [Client Quit]
00:52niv has joined #asahi
00:57niv has quit [Client Quit]
00:57niv has joined #asahi
00:58niv has quit [Client Quit]
00:58niv has joined #asahi
01:00niv has quit [Client Quit]
01:03niv has joined #asahi
01:03skg has quit [Quit: 〜バイバイ〜!]
01:03skg has joined #asahi
01:04niv has quit [Client Quit]
01:04niv has joined #asahi
01:05niv has quit [Client Quit]
01:06niv has joined #asahi
01:08niv has quit [Client Quit]
01:09niv has joined #asahi
01:09niv has quit [Client Quit]
01:10niv has joined #asahi
01:11niv has quit [Client Quit]
01:12niv has joined #asahi
01:17choozy has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
03:03inglor has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
03:05phiologe has quit [Ping timeout: 240 seconds]
03:06phiologe has joined #asahi
03:13inglor has joined #asahi
03:19luca020400 has quit [Ping timeout: 272 seconds]
03:19luca020400 has joined #asahi
04:03niv has quit [Ping timeout: 258 seconds]
04:03phiologe has quit [Ping timeout: 260 seconds]
04:03PhilippvK has joined #asahi
04:31BaughnLogBot has quit [Ping timeout: 260 seconds]
04:31BaughnLogBot_ has joined #asahi
04:31BaughnLogBot_ is now known as BaughnLogBot
04:34clayfreeman_ has quit [Ping timeout: 265 seconds]
04:38Tokamak has quit [Ping timeout: 240 seconds]
04:41Tokamak has joined #asahi
04:46marvin24 has quit [Ping timeout: 240 seconds]
04:47marvin24 has joined #asahi
04:51Necrosporus has quit [Read error: Connection reset by peer]
04:52Necrosporus has joined #asahi
05:09Tokamak has quit [Ping timeout: 240 seconds]
05:32Tokamak has joined #asahi
05:55niv has joined #asahi
06:11BaughnLogBot has quit [Ping timeout: 264 seconds]
06:13BaughnLogBot has joined #asahi
06:27VinDuv has joined #asahi
06:39nicolas17 has quit [Ping timeout: 246 seconds]
06:57_whitelogger has joined #asahi
07:29VinDuv has quit [Quit: Leaving.]
07:30odmir has quit [Remote host closed the connection]
07:42BaughnLogBot has quit [Ping timeout: 260 seconds]
07:42BaughnLogBot_ has joined #asahi
07:42BaughnLogBot_ is now known as BaughnLogBot
07:48ephe_meral has joined #asahi
07:52amw has quit [Ping timeout: 264 seconds]
07:54ephe_meral has quit [Ping timeout: 260 seconds]
08:08vimal has joined #asahi
08:32n2o has joined #asahi
08:41Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
08:43ephe_meral has joined #asahi
08:46nonemu has quit [Ping timeout: 240 seconds]
08:52n2o has quit [Quit: Leaving]
09:00iparaskev[m] has quit [Quit: Idle for 30+ days]
09:00assusdan[m] has quit [Quit: Idle for 30+ days]
09:00undvasistas[m] has quit [Quit: Idle for 30+ days]
09:00citruscitrus[m] has quit [Quit: Idle for 30+ days]
09:00randohacker[m] has quit [Quit: Idle for 30+ days]
09:00aimileus has quit [Quit: Idle for 30+ days]
09:00pedrojordao[m] has quit [Quit: Idle for 30+ days]
09:05ephe_meral has quit [Ping timeout: 272 seconds]
09:06mazmarcan: the delayed VHE setup code has gone in mainline, so you probably want to take this patch as part of your next round of patches:
09:13jn__do both CPU types (firestorm, icestorm) in the M1 SoC have this bug?
09:16marcanjn__: it's not a bug, it's a deliberate design choice in violation of the spec
09:16mazjn__: no idea.
09:16jn__marcan: ah, ok
09:16klaus has joined #asahi
09:16marcanI think it is safe to assume that both CPU types will behave identically in details such as this; I would only expect differences in corner cases and errata
09:18mazmarcan: my sentiment as well. you don't have this "bug" by accident, you have to design it as such. and write all your SW accordingly.
09:18marcanso... as for that patch... are you happy with it as it is, or do we need to come up with another solution?
09:19mazmarcan: I'm happy with it for now, and we can always have some discussion on the list. I expect Mark to nitpick at it, but that's the rule of the game!
09:20marcanI'll also base it off of Mark's tree with the FIQ changes
09:20mazthe really ugly part is the interaction with the cpufeature override.
09:21jn__("CPU known as Apple M1" in the patch text is a bit imprecise, because M1 the SoC has two kinds of CPU, that are at least in some regards different — hence my question)
09:22marcanfair point
09:22dhewgon that note "fruity cpu" sounds like raspberry
09:23mazjn__: feel free to point this out on the list in reply to https://lore.kernel.org/kvmarm/20210208095732.3267263-24-maz@kernel.org/
09:23mazdhewg: RPi doesn't design CPUs.
09:23marcanAIUI, the way these big.LITTLE systems are engineered (at least in Apple's case) is that the CPUs basically share the same design, and instead are just largely different configurations and synthesis options
09:23marcanso the power-efficient cores will have less wide dispatch, be built with less leaky / slower cells, have downsized elements, and stuff like that
09:23marcanbut not so much be a completely different design
09:24marcanso you would expect them to match in behavior (which is also pretty important not to run into horribleness with SMP)
09:24mazmarcan: it'd be good if they had a similar uarch. the traditional BL crap is awful to deal with.
09:24dhewgmaz: yeah, I know. I've just seen various projects using "fruit" to refer to rpi or clones like banana pi
09:24marcanpart of this I'm guessing based on the chicken bit names
09:25marcansome of them are the same, some of them have identical function in the E/P registers
09:25mazdhewg: well, fruit had a meaning before dev boards, so I feel perfectly fine with my comment! :-)
09:26dhewghehe alright ;)
09:26marcanmaz: oh yeah, one more thing, since I had a request for documenting ioremap_np... I haven't found any actual arch-agnostic ioremap docs, and I have no idea if I'd dare write one from scratch given all the "fun" details involved... but maybe it's worth giving a shot at writing an arm64 one?
09:26marcanat least listing out the mapping between ioremap types, MAIR settings, and what they mean
09:27dhewgjust as an example, first paragraph of the readme here https://github.com/crust-firmware/crust
09:30mazmarcan: I don't think this should be arm64-specific, as the ioremap API should have a consistent meaning across architectures (after all, that's how we write "portable" code).
09:31mazmarcan: but if you can start with something that also describe the MAIR settings for arm64 as an example of an architectural implementqwtion, that'd be great!
09:31marcanmaybe a new section in Documentation/driver-api/device-io.rst then
09:31marcanthat mentions ioremap, but none of the variants
09:32mazsounds very good to me! cc Jon Corbet for this.
09:32marcanthanks, will do :)
09:34arndmarcan: if you start a Documentation text in a wiki or your favourite web based shared text editing platform, I'll help fill in some blanks for the other functions and architecture specfics
09:35marcanoh, that works for me; I can throw it on the Asahi wiki as a WIP text, or google docs or something like that if you prefer real-time collaboration as a one-off thing?
09:36mazit may be worth looking at what x86 has done in their own corner: Documentation/x86/pat.rst
09:36marcanyup, I saw that one
09:36arndI prefer google docs, but wiki might be easier for others to join in
09:37arndI see we are also lacking proper documentation for the differences between readl()/readl_relaxed()/ioread32()/ioread32_be()/__raw_readl()
09:38arndI can write that section, probably should have done that years ago, because I keep having to explain it in code review ;-)
09:38marcanyeah, honestly I think Documentation/driver-api/device-io.rst probably needs a big revamp to make this all clearer, and also give arches a template to work off of
09:38marcangoogle docs should work for everyone, I can just have by-link access
09:39marcanit should be a short-term thing anyway, not a living doc
09:40marcanhow about I de-wrap Documentation/driver-api/device-io.rst into a doc, we can mess with it there, and then I can take care of re-wrapping it for the submission?
09:40marcan(I'll make sure not to spuriously re-wrap stuff we didn't touch, but it's easier to do all that in one final pass anyway rather than try to keep stuff wrapped)
09:41arndmarcan: I'd prefer to just add one or two sections to the document for the moment and not touch the other bits
09:41marcanoh sure, but I figured I might as well have the whole doc in there for context
09:41marcaneven if we don't touch it
09:41arndah yes, good point
10:01raster has joined #asahi
10:12BaughnLogBot has quit [Ping timeout: 265 seconds]
10:12BaughnLogBot has joined #asahi
10:30Necrosporus is now known as Guest1818
10:30Necrosporus has joined #asahi
10:33Guest1818 has quit [Ping timeout: 264 seconds]
10:38j`eyyou two are making good progress it seems!
10:41marcan(bbiab, need to make a quick trip to the pharmacy)
10:41marcan(comments welcome, I'm no expert on the ioremap_*() sematics :))
10:43plainbits has joined #asahi
10:43marcan(uhh nevermind the pharmacy, mission abort on that)
10:53arndMy brain dump on readl/writel gets fairly long, but that probably means it was indeed necessary do document it properly
10:53marcanabsolutely
10:53marcanit's very hard to wrap your head around this stuff if you're not already familiar with the semantics
10:53marcanI had to think pretty hard about readl() vs _relaxed for AIC and I'm still not 100% on the details
10:59arndmarcan: I think the description of ioremap_wc() is misleading as it sounds like this is cached plus allows write-combining, while I think it is the reverse:
11:00arndioremap_wc() is like ioremap(), plus write-combining
11:00arndioremap_wt() is like ioremap_wc(), plus it is cached (but not write-back)
11:00marcanah, right, I have them in the wrong order then
11:04amw has joined #asahi
11:07arndI have to read up on previous mailing list discussions about ioremap_uc(), I think this is also wrong. It's definitely specific to x86 and ia64
11:09marcanit's used in... two drivers outside of arch/
11:10marcanatyfb and intel-lpss
11:10arndright, there was a larger cleanup to remove it from almost all drivers that previously used it
11:11arndsee "git grep ioremap_nocache v3.0"
11:13arnda8ff78f7f773 ("mfd: intel-lpss: Use devm_ioremap_uc for MMIO") explains one of the two, I think the other is similar but old
11:14marcanwell I just realized something silly perhaps
11:14marcanioremap_wt is not implemented on ARM64, which means it falls back to ioremap
11:14marcanbut ioremap_wc is
11:14marcanthat feels odd, if the order is ioremap > ioremap_wc > ioremap_wt, then ioremap_wt should fall back to ioremap_wc, not ioremap
11:15arndthat sounds correct, yes
11:15amw has quit [Ping timeout: 272 seconds]
11:15arndDon't know why there is no real ioremap_wt() though
11:15marcanyeah, it's a thing anyway, so it should be implemented
11:15arndarm32 has "#define ioremap_wt ioremap_wc"
11:16marcanwe just need #define ioremap_wt(addr, size) __ioremap((addr), (size), __pgprot(PROT_NORMAL_WT))
11:17marcanI hope that doesn't break any broken drivers that wrongly used it :)
11:18marcan(cc maz, any idea why we don't have this yet?)
11:19j`eythere was an ioreamp_wt that was removed d092a87073269677b7ff09e71a8d91912b7f969a but it used PROT_DEVICE_nGnRE
11:19j`eyoh nvm, that's just the same as default ioremap
11:21marcanlooks like that was a multi-arch NFC cleanup patch
11:21marcanthe better question is why was it nGnRE to begin with
11:21arndprobably copied from arm32
11:22j`ey556269c138a8b2d3f5714b8105fa6119ecc505f2 added it
11:22arndyes, found the same
11:23marcanthe only things using ioremap_wt are certain framebuffers, and I bet there are 0 intances of any of those being used on ARM64
11:23arndlooks like I was on Cc, but none of the arm (or other) architecture maintainers were
11:23marcanand z2ram
11:23marcanamiga, atari, powermac...
11:25arndefifb was changed to memremap()
11:28marcanthis makes me wonder why other FBs do not use _wt, but just _wc
11:28arndmy guess is that they predated the ioremap_wt interface, so they used what they could get
11:29arnddrivers/video/fbdev is 95% obsolete
11:30marcanbut it's the other way around; a few old-looking drivers are using ioremap_wt, while many more are using _wc
11:31arndthose are the even older ones that are m68k specific
11:31marcanyeah, so nothing outside of m68k uses ioremap_wt
11:31arndand powermac
11:32marcanbut on the face of it ioremap_wt sounds appropriate for framebuffers? so why does nobody use it for modern hardware?
11:32arndI suspect nobody cares enough about the speed of the text console on fbdev
11:32arndanything else is in user space, which uses a separate mapping
11:33marcanI need to slap some kernel devs then, you have no idea how much slow fbcons frustrate me at the worst times ;) (as recently as a few months ago I had one of those, I still have a video somewhere...)
11:34arndmarcan: be careful, or you might end up as the fbdev subsystem maintainer
11:34marcannooooooo
11:35arndI think this has changed hands more frequently than any other subsystem. Usually it's someone who needs to get something done, but then gets frustrated to moves on to other work and abandons it again
11:35svensounds like the perfect job for marcan! :P
11:36marcansven: I hate you :D
11:36svenI know ;)
11:36aaronsmall99a[m] has left #asahi ["User left"]
11:36amw has joined #asahi
11:38marcanarnd: incidentally, the thing seems to be fully coherent on M1, as we're getting away with normal cached in m1n1 for the FB
11:38marcanyay for unified memory
11:39marcansimplefb still tries to map it with something less efficient though
11:39mazmarcan: I'm not surprised. they have also implemented ARMv8.4-FWB, which is going to be a godsend for KVM.
11:39mazno more silly cache maintenance!
11:40marcanhuh, does it require icache coherence?
11:40marcanI don't think icache is coherent here (i.e. you still need cache maintenance if you touch code)
11:41mazcicache coherency is isn't that big a deal.
11:41mazicache*
11:41mazthe killer is the PoC flush on each page being mapped into the guest.
11:41maz(vecause the guest could run with MMU off or a NC mapping)
11:41mazbecause*
11:42marcanright...
11:42marcan(I'm just going off the commit msg of e48d53a9 )
11:42amw has quit [Ping timeout: 240 seconds]
11:43marcanre slow framebuffers, this is the speed of efifb earlycon on an iMac at 4K resolution: https://photos.app.goo.gl/pyWM4VBdPjWQxQxk8
11:44luca020400looking smooth
11:51j`eymarcan: lol ok, i thought it might be slow, but not that slow
11:53marcan(afk, 10m)
11:54never_releasedmaz: Apple M1 is defined by Apple as "ARMv8.5 minus BTI"
11:54never_releasedwhich is pretty nice
11:55arndI have this vague memory of PCI prefetchable memory BARs being mapped to write-combining page mappings on some architectures, but can't find that now
11:56never_releasedmarcan: yes, Arm designs with coherent icache are quite rare
11:57arndI see that some old x86 fbdev drivers force write-combining with MTRR
11:57never_released(Neoverse N1, but is broken on almost-all-revisions-but-the-recent-ones there, and NVIDIA Denver are the ones that you can easily get access to)
11:59never_releasedI'm wondering if memory types is why Apple didn't support eGPUs
12:05nonemu has joined #asahi
12:07arndAh, I found arch_can_pci_mmap_wc(), which makes the user space mapping for a PCI bar write-combined when it comes from a IORESOURCE_PREFETCH BAR
12:07arndbut it does not affect the in-kernel mapping
12:14never_releaseda thousand-dollar question: what happens if we map the same range with different memory types for kernel and for user-space
12:14never_releasedI'd expect things to blow up essentially
12:14arndand now I read proc_bus_pci_ioctl() and wish I had never learned about that
12:15arndnever_released: on powerpc it blows up spectacularly, on arm you technically get undefined behavior but it will (mostly?) work in practice
12:16never_releasedand on some Arm designs, memory model guarantees vary too depending on the memory type in some designs
12:16arndI have seen multiple reports about broken framebuffer access on arm64, in particular with unaligned cacheable writes to the framebuffer, but I think that is a separate bug
12:16never_releasedugh, I repeated myself above
12:17never_released> For coherent memory types, Carmel cores provide a single, sequentially consistent view of coherent memory. Accordingly if no non-coherent access, Cache maintenance or TLB maintenance instruction has been executed since the last memory barrier, memory barriers behave similarly to a single-cycle NOP.
12:17never_releasedarm64 is everything from that to weakly-ordered for everything
12:17never_released(except some device memory types)
12:20arndnever_released: I thought it was NVIDIA who wanted RISC-V to adopt the weakest possible ordering model, I wonder if I mixed that up or if it was simply a different part of the company with different requirements
12:21never_releasedarnd: NVIDIA's own CPU cores all implement sequential consistency
12:21never_releasedas the memory model
12:21arndnever_released: what about the GPUs?
12:22never_releasedarnd: weakly-ordered there of course
12:24arndok, then I probably remembered correctly, as they have been talking about RISC-V in their GPUs but seem to favor ARM for applications cores
12:25never_releasedfor the firmware cores in the GPUs
12:25never_releasedalready shipped in Ampere as far as I know
12:29never_releasedI wonder what NVIDIA will do after the Arm acquisition happens, could be very interesting
12:30never_releasedespecially as relating to their contributions to RISC-V
12:35maznever_released: ARMv8.5 minus BTI, minus architecture compliance. At which stage, it doesn't mean much...
12:37akemin_dayo has quit [Ping timeout: 264 seconds]
12:38akemin_dayo has joined #asahi
12:41plainbits has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
12:48vimal has quit [Ping timeout: 272 seconds]
12:48never_releasedmaz: full compilance at EL0
12:48never_releasedand at EL1 with the right settings
12:49marcan(well that was longer than 10m)
12:50marcanarnd: IIRC sven had our framebuffer implementation blow up when he tried to use it from non-boot CPUs (which run with the MMU and caches off)
12:51svenyeah, weird things *unrelated* to the frame buffer started happening.
12:52svenI didn't investigate any further so maybe it was just another strange race that was triggered
12:52svenbut it went away the moment I stopped writing to the frame buffer range
12:52marcanoh wait, doesn't the smp code have a pretty tight timeout on smp boot?
12:52marcanmaybe you just blew the timeout with the uncached fb access?
12:52never_releasedusing a modern CPU with the caches and/or MMU off is begging for bad things to happen
12:52sventhat's what I thought at first but I increased that timeout
12:52svenbut maybe not enough ;)
12:52marcanhm, nah, we give it 0.5s
12:53never_released0.5s might as well be infinity, those are equivalent in that scenario
12:53marcan0.5s would not work with that iMac efifb I linked above ;)
12:53svenyeah, I think I pushed it to 5s and the smp still failed
12:54arndis the frame buffer part of the normal LP-DDR4X or separate?
12:54marcanit's part of normal memory, but not part of what we tell the kernel is usable RAM
12:54never_releasedarnd: only DRAM present is LPDDR4X
12:54svenit's part of the normal ram but behind at least one iommu
12:54marcanit's effectively a bootloader-time carveout
12:54never_releasedno other DRAM around
12:54sven(and another thing apple seems to call SMMU which is probably unrelated to the ARM SMMU though)
12:54never_releasedmarcan: that you can release and give back to the OS once a display driver is up
12:55marcan(which is actually required at this point, because if we let the kernel map it, simplefb fails to remap it as io)
12:55never_releasedwhich would allocate its own FB
12:55marcannever_released: this is something that I'll investigate once I write the display driver
12:55marcanfor glitch-free boot you need to reuse the FB to some extent, or at least do an atomic transition to another FB
12:55marcanbut there is also the question of what else, if anything, is in that "vram" area
12:56svenI'm pretty sure there's some kind of command/response/whatever thing in there
12:56never_releasedmarcan: you can smoothly change even the screen resolution on those
12:56marcanyes, I know
12:56sventhere are some more pages mapped in a second master of that display iommu and memsetting that to 0xff sometimes breaks the whole display driver
12:56marcanthough I don't think that's a hardware thing
12:57marcanpretty much any modern GPU should support atomic scanout changes including scaler settings
12:57marcanwhich is all you *really* need for that, plus a cooperative driver
12:57marcanit's just that legacy GPU drivers all suck on that front
12:57marcanand Apple got to write their own shiny new driver that sucks less this time :)
12:58marcanI saw that
12:58marcanI don't buy this being about their hardware "display pipes"
12:58marcanmore like the display pipe code ;)
12:58mechpilotace has quit [Quit: WeeChat 2.8]
13:02marcannever_released: playing with xrandr transforms on my old iMac 2015 retina results in glitchless scaling changes; and of course all modern GPUs must support page flipping
13:02marcanthose are the only two building blocks you need to do glitchless resolution switchin, *if* the software stack is up to par
13:03marcanfor some reason going to 1080p scaled to 4K glitches, but going back from that to plain 4K does not at all, other than windows moving around and stuff because X11 sucks at this
13:03mechpilotace has joined #asahi
13:03marcanso, glitch-free resolution switching is clearly a software problem and has been for many years
13:06vimal has joined #asahi
13:08marcanactually I think there might be 1 frame of black with the xrandr change here, but I'm sure that's the driver being stupid; obviously there is no output mode re-configuration as otherwise you get a very obvious slow black blink including a fade-in at the end (which is part of Apple's display controller)
13:09marcanI should try wayland again one of these days
13:09j`eyis resolution change that big of a deal? maybe I'm just not fussy about these kinda things :P
13:09marcanit's just one of those polish things
13:10marcaneither way, this isn't possible with external displays if you *actually* change the signal resolution (not just GPU-side scaling), because there's no way of stopping external displays from blinking for a whole second or more on mode changes
13:16marcanhttps://hub.marcan.st/t/scaling_change.mp4 <- tested with an external output, which in mirror mode avoids the desktop/WM getting involved; no glitching.
13:17j`eywhat have you subjected my ears to
13:35arndOur mac mini on a 27" screen randomly comes up with text much too small or much too big. Changing the resolution from 4K to 1080p and back usually gets it into the other mode. There is probably a setting somewhere to change the font size to the desired value, but I have never found that. "many years ahead of the industry" ;-)
13:36marcanthe M1 is many years behind on bug fixes ;)
13:36marcanI've seen graphical corruption in the settings menu of all places
13:36marcanthis launch was very, very obviously rushed, and it's rough around the edges
13:37marcanj`ey: you clearly do not know me well enough :P https://twitter.com/i/events/1284652732998008832
13:43j`eymarcan: unmuted *5s passes* muted :P
13:43marcanto be fair, that *is* a loud song... and the remasters are even louder :/
13:44marcanjapan has a real problem with the loudness war
13:44marcanbut I'm guessing you wouldn't be the kind to enjoy my own music then ;)
13:45marcanat least I don't crush it with a limiter though :p
13:45marcanarnd: moved the relevant ioremap_uc() points to ioremap_wc() then
13:46j`eymarcan: listening to Endless Scarlet Night
13:46marcanoriginal or soundflora's remix?
13:46j`eymarcan: Tsuioku Circuit linked from your yt
13:47marcanyeah, but there are two versions
13:47j`eyah, not the soundflora one
13:47marcanhers is much groovier than mine :)
13:48* etalikes the original endless scarlet night
13:48marcan(also turn on the english subs if you haven't yet)
13:50marcanarnd: do you have any idea about what the store buffer story is with _wc, if any?
13:50etare Wayland: it's nice, but be prepared to run bleeding edge if you want stuff like screen sharing to work
13:50marcanAIUI write combining technically means stuff can stick around in the store buffer for a while, possibly forever if no other stores happen? but I'm not sure if we kind of assume that real CPUs will flush that at some point, or there is some barrier that forces that
13:51marcaneta: for a long time I could not use Wayland because there was no standard middle-button-click protocol (and thus KDE did not support it) and I refuse to give that up
13:51marcanfor paste I mean
13:51etamarcan: yeah, this is par for the course with Wayland IME
13:52marcannow it does, but I'm a bit scared because I still have a lot of legacy X11 software doing very X11-ish things
13:52etathere is no "standard" screen sharing protocol
13:52marcanI'm getting the feeling this M1 stuff is also going to become my own personal wayland testbed, good excuse to try it on the new platform
13:52etathere's the undocumented gnome framebuffer thing and then xdg-desktop-portal
13:52etathe latter "works" but only under Firefox with some config flags twiddled and with pipewire running
13:52* etaends up using OBS to screen share mostly
13:53marcanwell, that's also a good excuse to use pipewire :)
13:53eta(which has a hacky pipewire source)
13:53etaI love pipewire
13:53marcanamusingly enough though, I don't actually care about OBS screen sharing any more... since I use HDMI capture with a different PC for that
13:53etaso nice
13:53marcanbut OTOH I need support for scaled mirroring
13:53marcan(which xrandr can do, no idea about wayland)
13:53marcanI really *really* want to replace JACK+PA with PW
13:53* etashrugs
13:53etaprobably depends on your compositor
13:53etamarcan: do it!!
13:53marcanbut first I need to spend a weekend finishing that libffado plugin I started writing for it
13:53etaoh yeah, firewire
13:53marcanI got audio in/out working at least, but it's still broken-ish
13:53arndmarcan: as far as I can tell, the use of a store buffer is what does the write-combining, so you can't have one without the other
13:54etafwiw you can run PW as a JACK client
13:54marcanneed to polish it into something usable
13:54etathat's cursed though
13:54etait's easiest if you just replace libjack with pw
13:54marcanarnd: right, but are there any guarantees as to the lifetime of writes in the store buffer, or operations guaranteed to flush them?
13:54arndso no guarantees about completion with _wc
13:54marcaneta: yeah I don't want to do that, half the reason to run PW is to hopefully avert JACK's brokenness
13:54marcanJACK1 and JACK2 both have *different* problems
13:55marcanarnd: i.e. if there are zero guarantees, when technically _wc is like _cached, but that isn't how it is used in practice
13:56arndmarcan: my understanding is that normal ioremap() being posted already requires a read-back to flush it (as you write there), so that would be the same with _wc
13:56marcanah, so reads are guaranteed to flush the wc buffer?
13:57marcanthough it's still not quite the same thing; ioremap() being posted means you need a read to ensure ordering, but a write via ioremap() *is* guaranteed to hit the device in due time
13:57arndI would assume so, but I don't know for sure
13:57marcanwhile something stuck in the wc buffer isn't
13:58arndisn't there an implementation specific upper bound on the write buffer?
13:58marcantime to pull up some docs
13:58marcanlet me see what PPC750CL says about this (which I happen to have some experience with)
13:59arndI mean for write-back cached mappings there isn't, because nothing necessarily enforces the writeback, but for a normal write-buffer I would assume that it gets flushed out as soon as there is bandwidth avaiable on the bus
13:59odmir has joined #asahi
14:03marcanI don't think that is necessarily the case, because you could end up with somewhat pathological cases
14:03marcanbut it probably depends on the implementation
14:04kettenismy understanding is that you need a memory barrier to guarantee that store buffers have been flushed
14:07kettenisbtw, the reason nobody bothers mapping framebuffers write-through is that most grapphics code these days only writes to the framebuffer
14:08kettenisso it doesn't really offer a speed benefit over write-combining
14:09kettenis(write-combining tends to speed things up a lot though)
14:11marcanok, so the ARM ARM actually says writes must reach the endpoint in finite time for NC/WT modes
14:11marcanso clearly staying forever in a write buffer is not allowable
14:13arndkettenis: good point, no reason to throw away your dcache when filling a 4K framebuffer
14:15arndmarcan: ok, let's assume that is the sensible behavior then. Without this, you could end up never seeing a framebuffer update
14:25marcanI think we can just say nothing in that case
14:30arndmarcan: do you mind if I share the current draft on another channel with more kernel developers? I'd like to find out if anyone else knows why mapping a prefetchable PCI resource from user space uses a _wc mapping while pci_ioremap_bar() doesn't
14:30marcansure, go for it!
14:30marcan(where?)
14:31arndit's an invitation-only channel, I could never figure out who exactly qualifies to get invited
14:31marcanwell, I guess I shouldn't be surprised the kernel cabal has one of those too ;)
14:32kettenishardware that icorrectly tags a BAR as prefetchable exists
14:39arndkettenis: that would explain the in-kernel interface, but not the user space side
15:00ephe_meral has joined #asahi
15:02klaus has quit [Quit: Textual IRC Client: www.textualapp.com]
15:31choozy has joined #asahi
15:38vimal has quit [Read error: Connection reset by peer]
15:38vimal has joined #asahi
15:54marcanmaz: I think after this doc update is in good shape, I just need to review the AIC changes and then I'm good for sending out v3
16:00ryanhrob[m] has quit [Quit: Idle for 30+ days]
16:00blazra has quit [Quit: Idle for 30+ days]
16:00xerpi[m] has quit [Quit: Idle for 30+ days]
16:14linuxgemini has quit [Remote host closed the connection]
16:21the-mentor has quit [Quit: The Lounge - https://thelounge.chat]
16:23Tokamak has joined #asahi
16:23tomhv has joined #asahi
16:25the-mentor has joined #asahi
16:32carlosn has quit [Read error: Connection reset by peer]
16:32carlosn has joined #asahi
16:39carlosn1 has joined #asahi
16:39carlosn has quit [Read error: Connection reset by peer]
16:41Tokamak has quit [Ping timeout: 264 seconds]
16:42vimal has quit [Quit: Leaving]
16:44Tokamak has joined #asahi
17:23nicolas17 has joined #asahi
17:29carlosn1 has quit [Read error: Connection reset by peer]
17:29carlosn has joined #asahi
18:01ephe_meral has quit [Quit: WeeChat 2.9]
18:01ephe_meral has joined #asahi
18:08ephe_meral has quit [Ping timeout: 264 seconds]
18:22VinDuv has joined #asahi
18:35ephe_meral has joined #asahi
18:53raster has quit [Quit: Gettin' stinky!]
19:09linuxgemini has joined #asahi
19:24carlosn1 has joined #asahi
19:24carlosn has quit [Ping timeout: 265 seconds]
19:34brandas has quit [Quit: quit]
19:35DarthCloud has quit [Remote host closed the connection]
19:36DarthCloud has joined #asahi
19:43odmir_ has joined #asahi
19:45odmir has quit [Ping timeout: 240 seconds]
19:46brandas has joined #asahi
19:49brandas has quit [Client Quit]
19:52ephe_meral has quit [Ping timeout: 240 seconds]
20:10tomhv has quit [Read error: Connection reset by peer]
20:13tomhv has joined #asahi
20:20ephe_meral has joined #asahi
20:20tomhv has quit [Ping timeout: 258 seconds]
20:24tomhv has joined #asahi
20:35brandas has joined #asahi
20:36brandas has quit [Client Quit]
20:37brandas has joined #asahi
20:45brandas has quit [Quit: quit]
20:46brandas has joined #asahi
20:48ephe_meral has quit [Ping timeout: 265 seconds]
21:18brandas has quit [Quit: quit]
21:20ryanhrob[m] has joined #asahi
21:21eta has quit [Quit: we're here, we're queer, connection reset by peer]
21:22eta has joined #asahi
21:24eeeeeta has joined #asahi
21:28PendulumSwinger has joined #asahi
21:45VinDuv has quit [Quit: Leaving.]
21:55tomhv has quit [Quit: WeeChat 1.9.1]
21:59brandas has joined #asahi
22:06qyousef has quit [Ping timeout: 264 seconds]
22:08qyousef has joined #asahi
22:32brandas has quit [Quit: quit]
22:33brandas has joined #asahi
22:53brandas has quit [Quit: quit]
23:44KindOne has quit [Ping timeout: 260 seconds]
23:45KindOne has joined #asahi
…