warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
naoki has quit [Quit: naoki]
System_Error has quit [Ping timeout: 260 seconds]
System_Error has joined #linux-rockchip
warpme has joined #linux-rockchip
Daanct12 has quit [Quit: WeeChat 4.2.1]
macromorgan has quit [Read error: Connection reset by peer]
macromorgan has joined #linux-rockchip
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
Stat_headcrabed has joined #linux-rockchip
raster has quit [Quit: Gettin' stinky!]
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
<macromorgan>
mmind00: I noticed that there aren't any boards with overlay support enabled. Is that on purpose/desired? Otherwise there's at least a few boards I'd like to enable it for in mainline... ones with GPIO headers.
<qschulz>
diederik: I would assume that this patch is working? So while there aren't DTC_FLAGS in rockchip's DT makefile, I assume we get it from somewhere else?
<diederik>
I can't (quickly) find a related discussion where setting that globally was still not recommended/done, but enabling it on directory (rockchip) or board level was no longer frowned upon
<qschulz>
macromorgan: we've talked with mmind00 to add support for DTBO mainline for our boards, so DTBO support is very much welcome upstream... Just that the DTBO we want to upstream require a few more fixes (and also because we need to change our documentation/processes/workflows internally to use DTBOs :) )
<qschulz>
diederik: the "issue" is that the DTSO patch I sent is for a rockchip board, and I don't see a DTC_FLAGS in rockchip's Makefile for it to work, so it must work, but through a different mechanism :)
<diederik>
qschulz: I'm completely *un*familiar with the last patch you linked
<qschulz>
diederik: ahah! I think I got it
<qschulz>
+# Base DTB that overlay is applied onto (each first word of $(*-dtbs) expansion)
<qschulz>
so provided your dtso has a filename that starts the same as a a dtb would have -@ flag?
<diederik>
TBF: I don't really/fully understand this material, I just remembered the discussion
System_Error has quit [Remote host closed the connection]
<qschulz>
so I guess a common "root" in the filename would make any dtb with that common root string (but maybe longer) compile with -@?
System_Error has joined #linux-rockchip
<diederik>
Or the alternative, like it's done in Aurelien's patch is to add the `DTC_FLAGS := -@` in an appropriate place in the Makefile.
<diederik>
Doing it for all rockchip devices may not be the best, but you could also specify it on specified boards. At least that's how I understood it
warpme has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<diederik>
qschulz: I think you're correct. If there's a .dtso file then the DTC_FLAGS get set on similarly named .dts file
<diederik>
But this would only work on overlays which are part of the kernel, not external ones
<qschulz>
diederik: absolutely
stikonas has quit [Quit: Konversation terminated!]
stikonas has joined #linux-rockchip
<macromorgan>
yes, I mean like that broadcom commit
<macromorgan>
maybe we simply have something in the top of the Makefile for rockchip's SoCs that enable overlays for certain boards if CONFIG_OF_OVERLAY is also defined?
<macromorgan>
to me the demarcation would be whether or not there are exposed GPIO pins, but that's obviously up to the maintainers
<macromorgan>
for example I'd enable overlays for things like the Odroid Go Advance/Advance BE (which both have exposed pins) as well as some dev boards I'm messing with; but things like the Anbernic devices would not be added to the overlay list since no exposed pins
<macromorgan>
hmm, but then we'd have to completely segregate it within the Makefile
<macromorgan>
note that when tested for the Indiedroid Nova and Rock 5B the DTB files were twice as large
Stat_headcrabed has quit [Quit: Stat_headcrabed]
<qschulz>
I tried to check if this DTC_FLAGS autorequirement based on filename works by decompiling the dtb and dtbo but I clearly don't understand how this works
<qschulz>
macromorgan: I assume the point is to allow out-of-tree DTBOs?
<macromorgan>
yes, correct
<macromorgan>
it only makes sense for "tinker" boards though. Otherwise we're doubling filesizes for no reason.
<macromorgan>
ohh, duh... it can be done for individual boards by doing the following: `DTC_FLAGS_rk3588s-indiedroid-nova := -@`
<qschulz>
macromorgan: not really no, for example, our devkit has a custom connector which accepts whatever you develop for it, we already support two adapters for PX30 Ringneck for example
<qschulz>
we also have an adapter board we connect to the GPIO headers for end of production line testing
<qschulz>
all those could (should) be overlays
<macromorgan>
I guess then what I'm asking is "would it be acceptable to add stuff like `DTC_FLAGS_rk3588s-indiedroid-nova := -@` to the Makefile, or should I do it differently?"
<qschulz>
while it's a devkit and I could agree this is a "tinker" board, we also have an SBC with a proprietary connector with multiple adapter boards today for example
<macromorgan>
that's fine... I guess I'm really saying its up to the board maintainer to enable it. Some devices it makes no sense but other devices that have external connectors it does.
<qschulz>
macromorgan: it's a trivial patch to send, so I'd recommend to send it and see what Heiko thinks of it whenever he has time :)
<macromorgan>
yep, think that's what I'll do. I'll enable it for boards that I own and where it makes sense (so none of those handheld gaming devices excluding the Odroid ones with the external connector).
<macromorgan>
worst they can do is say no :-)
<qschulz>
absolutely :)
vagrantc has joined #linux-rockchip
<macromorgan>
for now I'll just enable it for Rock 5B and Indiedroid Nova. I'll let other board maintainers do as they see fit.
<macromorgan>
filesize for the DTB went from 68k to 134k with overlays enabled
<qschulz>
macromorgan: can you do a diff on the output of dtc -I dtb -O dts rock5b.dtb ?
<qschulz>
that should tell you what actually increase the size
<qschulz>
if I'm not mistaken, what could be increasing the size could be that we have too many labels?
<macromorgan>
the increase is phandle references which are not present in the non-overlay version, a boatload of unused pincfgs (this is the main driver of size growth), and a symbols list
<macromorgan>
given the size growth I'd be loathe to enable it for all rk3588 and rk3588s boards like I was planning on (all but the nanopi r6s which curiously enough is the only rk3588 based board without external expansion connectors)
<macromorgan>
so I'll just push it upstream for the two I can test... the Indiedroid and the Rock 5B
<qschulz>
yeah, so we would likely have this big size increase for all rockchip boards, regardless of the SoC
<qschulz>
since rockchip-pinconf.dtsi is SoC-agnostic
<macromorgan>
I'm okay with that, 60kb isn't much of a concern to me, but I have no idea if someone's bootloader config only has a set amount of memory set aside for the dtb
<macromorgan>
so I'll just do these two for now
<qschulz>
macromorgan: the "issue" is that there are some people who are trying to use the same dtb binary for U-Boot and the kernel, so that may become an issue :)
<qschulz>
in any case, start with your two boards and we'll tackle that whenever we need to :)
<macromorgan>
well that's the other reason why I'm only doing it for these two boards. Rock 5B is already supported in mainline U-Boot and I'm trying my best to figure out the bloblist stuff so I can get the Indiedroid there as well. As long as we don't do something silly and allow overlay support in U-Boot SPL pre relocation we'd probably be okay.
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
vagrantc has quit [Quit: leaving]
vagrantc has joined #linux-rockchip
Livio has joined #linux-rockchip
werdahias has joined #linux-rockchip
werdahias has quit [Quit: werdahias]
werdahias has joined #linux-rockchip
werdahias has quit [Client Quit]
System_Error has quit [Remote host closed the connection]
System_Error has joined #linux-rockchip
<mmind00>
macromorgan: "from 68k to 134k with overlays enabled" ... I guess with board measuring memory in the gigabytes that doesn't really matter that much ;-)
<macromorgan>
the one wild card is how does the BSP U-Boot handle it, since if I recall correctly it's rigged somehow to use the kernel's DTB and we're now doubling it in size
<CounterPillow>
there's an old rockchip ATF floating around where large DTBs (>64k) caused it to crash but they've since fixed that iirc