<minute>
i figured that i can move /boot to NVMe instead (with g12b) and it works, but it's very slow! > 33499648 bytes read in 37113 ms (880.9 KiB/s)
<minute>
takes over a minute to load kernel and initramfs from nvme
BlueMatt has quit [Server closed connection]
BlueMatt has joined #linux-amlogic
<minute>
ah, this disk also has a similarly slow read rate in linux :D
<xdarklight>
f_: didn't you also have some trouble with CMD6 during your u-boot SPL journey?
<f_>
I think so, but it was on gxbb/gxl, not g12b
<f_>
Need to properly debug and fix this still.. but I do not have my gxbb/gxl boards with me at the moment.
<xdarklight>
minute: ^ so it may indeed be related to the meson_gx_mmc driver in u-boot
<minute>
aha!
<f_>
I was a bit confused about why it happened, TBH, BL2 + proper U-Boot loading stuff from eMMC on gxl seemed to work fine
<f_>
didn't test gxbb though.
<f_>
So maybe U-Boot is doing very weird stuff with eMMC?
<f_>
BTW, on gxbb and gxl, BL2 does no eMMC init whatsoever
<f_>
Just straight up goes loading stuff
<f_>
it's possible that the bootROM did enough of the init for BL2
<minute>
f_: aha, this goes kinda in the direction i'm suspecting. i wonder if i could hack it up so that uboot doesn't touch the emmc configuration at all...
<minute>
but now i can boot from nvme partition (and the speed is very good too, it was a hardware problem here, had an old adapter with pcie reset not wired up)
naoki has joined #linux-amlogic
<f_>
Probably could work, strip down u-boot's eMMC init
naoki has quit [Client Quit]
paulk-bis has quit [Quit: WeeChat 3.0]
paulk has joined #linux-amlogic
paulk has joined #linux-amlogic
luka177 has quit [Read error: Connection reset by peer]
<narmstrong>
minute: which mode are you trying to setup on the eMMC ?
<narmstrong>
hs200 ?
luka177 has joined #linux-amlogic
<minute>
narmstrong: no, just any mode basically
<minute>
narmstrong: hs200 doesn't seem to be supported in the uboot driver
<narmstrong>
yes hs200 is not supported, only basic modes
<minute>
when it fails CMD 6 it tries all the modes and they most of the time all fail and it gives up
<narmstrong>
the mmc controller is not very well designed...
<minute>
sometimes it can be coerced into working by manually typing "mmc dev 1" a few times
<minute>
ah :D
jkl has quit [Server closed connection]
jkl has joined #linux-amlogic
luka177 has quit [Ping timeout: 252 seconds]
f_[xmpp] has quit [Server closed connection]
f_[xmpp] has joined #linux-amlogic
<f_[xmpp]>
I can see that
<f_[xmpp]>
Amlogic my beloved
<f_[xmpp]>
But really, some manufacturers also cut off some corners
gabes22 has quit [Server closed connection]
gabes22 has joined #linux-amlogic
JerryXiao has quit [Server closed connection]
JerryXiao has joined #linux-amlogic
gis has quit [Server closed connection]
gis has joined #linux-amlogic
repk has quit [Server closed connection]
repk has joined #linux-amlogic
<minute>
trying systemctl suspend on g12b for giggles... > bl30 get wakeup sources!
<minute>
hangs after > Big core clk resume rate 50000000
<narmstrong>
it should work, at some point it worked...
<minute>
that sounds promising
<narmstrong>
I recall the cpu core rate from resume is weird and it caused some issues
<minute>
yeah sounds like 50mhz
<minute>
> Little core clk resume rate 1000000000
<minute>
like, little core 1ghz, big core 50mhz? :D
<minute>
not sure what would be a good wakeup source
<narmstrong>
yeah it's expected, we "park" the cores on a slow clock, but the amlogic resume clock f***s the clock state instead of leaving is the same as suspend, so we should re-apply the cpu clock state when resuming