00:13 <apritzel> macromorgan: ah, bummer, I've been misled: the nets are called AXP717-CC[12], but lead to two 5.1K resistors, and on the AXP the pins are not connected :-(

01:14 <macromorgan> yeah... when connecting stuff I think I always detected DCP anyway which made me think the whole thing was a scam. That's why I added the logic in mainline to let you set a hard limit, and I hard limited the device to 1.5A no matter what.

02:08 <kjn260> Hey guys. got a tricky one. Have one of these: https://www.myirtech.com/list.asp?id=722 on a custom carrier

02:09 <kjn260> it comes preinstalled with some uboot and a allwinner Linux on the eMMC

02:09 <kjn260> but I cannot get to USB FEL for the life of me

02:09 <kjn260> based on the datasheet I figured pulling down the BOOT_SEL0 would be sufficient

02:10 <kjn260> but it doesn't seem to work - maybe its set with the effuse

11:22 <apritzel> kjn260: I don't think it's BOOT_SEL0 that triggers FEL mode, that's just one of the two pins selecting the boot *order*

11:24 <apritzel> I am not even sure the T113-s3 has a dedicated FEL pin, at least the data sheet doesn't show it, though the user manual mentions it (maybe copy&paste?)

11:24 <apritzel> the MangoPi MQ-R has a "FEL" button, but that just shorts the SPI CS to disable that boot medium, so it *falls through* to FEL

11:27 <apritzel> I guess you would need to do something similar: sabotage the eMMC briefly at boot time, for instance by shorting PC2 (SDC2-CLK) to GND, so the BROM can't talk to the eMMC

13:04 <IlikeTech> Anyone got suggestions for what kernel compression alg to use for a decent speed / size ratio? Trying to fit the kernel into NOR so it should be pretty small. This is on the F1C100s which is ARM926, so not terribly fast. I've tried xz, and while it's small, it's also very slow to decompress.

13:05 <IlikeTech> gzip is kind of huge

13:07 <apritzel> On a bigger, but still limited machine I ended up with "zstd -10" as a good compromise between extraction speed and size

13:08 <apritzel> when you create it on your build machine, make sure to use "-T n" to make use of multiple cores for the compression part

13:13 <IlikeTech> interestingly, the kernel doesn't give me a zstd compression option

13:17 <apritzel> I see, I used that actually for an initrd, and zstd is available there

13:18 <IlikeTech> Yeah, it's available there

13:21 <apritzel> maybe you have more luck by shrinking your kernel config? That's quite an endeavour, but could give you much smaller kernels

13:25 <IlikeTech> yeah, I'll try to reduce it down more and see how that goes

13:25 <IlikeTech> sometimes it's hard to tell what's safe to turn off lol

13:46 <apritzel> I agree, it's tricky. And I figured that turning off things (so coming from a full-featured config) will only shrink the kernel so much.

13:46 <apritzel> The real savings can be achieved by starting with a minimal config, then just enabling what you need

13:47 <apritzel> but that is even more tedious, and you probably need a few cycles to find what's missing

15:49 <apritzel> wens, MoeIcenowy: the A523 subtly changed the MMC mod clock to use *two* divider fields (clksrc/N/M), not one divider and one shift (clksrc/M>>N) as before. It seems we do not model this kind of clock yet? At least I didn't find it ...

17:33 <wens> I don't think so