00:49 <apritzel> ah, looks like the Teclast P50 AI tablet uses the new A733 SoC, with two A76 + six A55 cores, on offer at Aliexpress atm

00:56 <MasterR3C0RD> Isn't that the SoC that supports DDR5?

00:59 <apritzel> at least it claims so, and the tablet seems to come with 6GB also (currently all SoCs are limited to 4GB at most)

01:06 <apritzel> the display is 11 inch @ 90 Hertz, but they don't mention the resolution on Aliexpress - though the Teclast website reveals the sad truth: 1280x800 (!)

01:38 <MasterR3C0RD> That Black Friday price might convince me to pick one up and take a look at its PHY sooner rather than later

01:40 <apritzel> be warned that there is no user manual yet (though some code in github). And it's probably "secure boot", so more painful to work with (FEL will probably not work easily)

01:41 <apritzel> I guess if you open it, you get access to the UART, plus maybe a FEL button, but otherwise you can only use the UART via the SD pins - but then you don't have a SD card ...

01:41 <MasterR3C0RD> User manual isn't very useful for the DRAM controller anyways

01:42 <apritzel> true, but just DRAM isn't very useful on its own, without U-Boot and Linux support ;-)

01:42 <apritzel> and you would need info like the clock details and the memory map

01:44 <MasterR3C0RD> Yeah, I'm mostly curious to see if this one's a "real" (known) IP core of some sort or at least shows more signs of being one, considering LPDDR5/DDR5 is a curious jump

01:44 <apritzel> and it comes with UFS flash, which is a first for Allwinner, so we have no real clue about the IP used - which means no easy mainline access to the storage

01:46 <apritzel> I let myself cool off here, the offer is valid for 10 more days ;-)

01:47 <MasterR3C0RD> True enough, maybe there's a small chance more info will drop

01:48 <apritzel> ah yeah, I am hopeful a manual will appear at some point, but probably only if the chip appears on boards

05:35 <MoeIcenowy> ah I heard about A733 too

05:35 <MoeIcenowy> and a bad news here, it comes with Imagination GPU

05:35 <MoeIcenowy> BXM-4-64

05:36 <MoeIcenowy> apritzel: I start to think this kind of consumer-oriented tablets not interesting to hack at all

05:36 <MoeIcenowy> (especially the Teclast tablet PR says "Widevine L1", which is a big warning

05:37 <MoeIcenowy> (something on betray of its owner must have been implanted

06:36 <MasterR3C0RD> MoeIcenowy: At least on the GPU front, the BXM-4-64 appears to be fairly close to the BXS-4-64 that Imagination's current GPU work has been focused on

06:37 <MasterR3C0RD> Frank Binns has said adding support for it once those patches land should be trivial: https://gitlab.freedesktop.org/imagination/linux-firmware/-/issues/2#note_2643442

11:16 <apritzel> MoeIcenowy: I don't particularly care about GPUs at all, to be honest ;-)

11:18 <apritzel> but it has quite some new features, like UFS, >4 GB RAM, dual USB 3.2, and then the elephant in the room: big cores!

11:19 <apritzel> A76 is quite a step forward performance-wise

11:20 <apritzel> MoeIcenowy: but yeah: this tablet will probably be quite boarded-up, so not ideal for hacking. Had this experience already with my A523 tablet

11:35 <maz> apritzel: working PCIe?

11:47 <apritzel> maz: it says 1xPCIe 3.0, and the situation on the A523 seems much better than on the ill-fated H6 in this regard, so I am hopeful

11:47 <maz> progress!

11:48 <apritzel> the OrangePi 4A (with the T527, sibling of A523) has an M.2 slot, so it should be easy to try

11:49 <apritzel> but from what I read in the manual, they did it right this time: there is a 256MB slot in the memory map for PCIe

11:50 <apritzel> so hint hint: can someone put this A733 on a devboard, please? ;-)

11:51 <apritzel> AW promised quite some decently spec'ed chips in the past (on road maps), but most of them never materialised, so it's good to see some actual silicon with better specs

11:56 <apritzel> maz: since I have you here: can you have a brief look at this GIC patch for TF-A and say whether this makes sense: https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/31141

12:07 <maz> apritzel: the potential infinite loop is a bit sketchy, but, yeah, not supposed to happen.

12:14 <apritzel> maz: thanks, I see. The vendor boot0 sets ProcessorSleep before entering TF-A (without any waits), to work around this assert, but this sounds a bit more sketchy to me

12:15 <maz> yeah, that's terrible. you pray that the GIC disconnects instantaneously...

12:16 <apritzel> and I think we cannot even do this in the SPL, since we return to the BROM, for FEL, which requires interrupts

12:16 <maz> it's a bit odd to require interrupts that early, but hey, why not.

12:17 <apritzel> I agree, but it's for the MUSB USB-OTG controller, maybe that requires interrupts for proper functionality?

15:15 <hlauer> <apritzel> "hlix: what are the differences..." <- Something in the usb subsystem does a reset during boot. So the current 6.12 kernel only starts with "usbcore.nousb". And I noticed u-boot tells about "starting USB... Bus usb@1c1xx00: Port not available."

15:32 <apritzel> hlauer: this might be about a different GPIO for powering on VBUS. Do you have any schematic, or at least some vendor DTB?

15:37 <hlauer> I noticed https://patchwork.kernel.org/project/linux-arm-kernel/patch/ZYqRZev1g_mztff2@debian.cyg/ which is for a A40i board which powers up reg_aldo1, but adding this to the m2ultra dts didn't make a difference

16:23 <apritzel> hlauer: that looks like shooting in the dark ;-) So there are no separate documentation or images for v1.1?

17:14 <apritzel> does any of you graphics buffs ;-) know what's the best way to drive a STV7789 panel, as on the Avaota-A1? I am lost between drm/panel, staging/fbtft and drm_mipi_dbi ...

17:54 <hlauer> On https://docs.banana-pi.org/en/BPI-M2_Ultra/BananaPi_BPI-M2_Ultra there are some links, mostly to google drive and baidu. The most interesting stuff seems all to be behind "ask for permissions" on g.d.

18:07 <apritzel> yeah, I tried some of those links yesterday, but got upset after a while about all this mess

18:30 <hlauer> got a dts from openwrt, which may support V1.1. For what should I look?

19:38 <hlauer> Maybe the log https://gist.github.com/halmartin/59f8056aca5416ab37c47a8123d7015d is of interest: u-boot didn't recognize usb ports, but the openwrt 5.10 kernel loaded looks like running with usb

20:50 <apritzel> is there only one DT in OpenWRT? But in general it's hard and tiresome to debug all of this on old software versions (like kernel 5.10)

20:51 <apritzel> hlauer: so can you use mainline versions of U-Boot and the kernel and see how it goes?

20:52 <apritzel> another thing to check is to diff the DT that apparently works in Linux with the mainline DT, to see if there are any changes

20:59 <hlauer> openwrt ships a uboot vfat partition containing a single dtb and uImage file besides the boot.scr - it's tailored for low space

21:05 <hlauer> Yes, 2024.04-rc4 is working fine. I miscompiled 2024.10 somehow (tried to get rid of all this unneeded efi stuff)