<davidlt>
So there is polling mode to detect sd card in some drivers.
<davidlt>
Have no idea what's inside StarFive eMMC.
<Tenkawa>
yeah its an odd one
<Tenkawa>
I need one in mine now anyway since I plan on migrating to an xfs root so it "mitigates" it.. but its a workaround lol
<Tenkawa>
Since xfs booting still needs chainloading
<Tenkawa>
(Unless its been added in the newest u-boot/OpenSBI)
<davidlt>
Tenkawa, looking at their schematics I see CD pin is wired.
<davidlt>
SD_SDIO0_CD_GPIO41
<davidlt>
I don't see it listed in DTS
<Tenkawa>
intriguing
<davidlt>
There is "cd-gpios", description: The card detection will be done using the GPIO provided.
<davidlt>
ah, but it has "broken-cd;"
<davidlt>
which means "There is no card detection available; polling must be used."
Bluefoxicy has quit [Ping timeout: 255 seconds]
Bluefoxicy has joined #riscv
junaid_ has joined #riscv
esv has quit [Ping timeout: 245 seconds]
esv has joined #riscv
esv has quit [Client Quit]
<Tenkawa>
There we go... much better
<Tenkawa>
root@vf2:~#
<Tenkawa>
/dev/nvme0n1p1 on / type xfs (rw,relatime,attr2,inode64,logbufs=8,logbsize=32k,noquota)
<Tenkawa>
now I have my preferred filesystem
<Tenkawa>
mps: the next interesting part of that microsd discovery... have to eject/reinsert every boot to drop the load
<Tenkawa>
davidlt: that does seem to go with the disconnected polling
<Tenkawa>
now the pivoltal question... how to fix that so that a "physical" reseating is not needed every boot
leah2 has quit [Ping timeout: 245 seconds]
<Tenkawa>
er pivotal
junaid_ has quit [Remote host closed the connection]
cousteau has quit [Quit: ♫ I can't forget the day I shot that network down ♫]
Trifton has quit [Quit: Error: no route to host]
agent314 has joined #riscv
leah2 has joined #riscv
ntwk has quit [Quit: ntwk]
arnd has quit [Server closed connection]
arnd has joined #riscv
billchenchina- has joined #riscv
billchenchina has quit [Read error: Connection reset by peer]
junaid_ has joined #riscv
GenTooMan has quit [Ping timeout: 248 seconds]
GenTooMan has joined #riscv
agent314 has quit [Ping timeout: 250 seconds]
agent314 has joined #riscv
sympt has quit [Server closed connection]
sympt has joined #riscv
jmdaemon has joined #riscv
jmdaemon has quit [Ping timeout: 246 seconds]
davidlt has quit [Ping timeout: 245 seconds]
<Esmil>
ok, finally figured out why i keep getting the 'nvme0: I/O 640 QID 5 timeout, completion polled' errors on my VF2: CONFIG_INIT_ON_ALLOC_DEFAULT_ON
<Esmil>
with that enabled my nvme drive works, so it seems like the starfive pcie driver (or nvme driver) is leaving garbage in some allocations :/
<Tenkawa>
Esmil: did you see the above findings with microsd/nvme and high load?
<Esmil>
no
<Tenkawa>
mps and I have recreated this..
<Tenkawa>
if you look with a nvme only the idle load average is probably running aroumd .5
<Tenkawa>
after you plugin a microsd it drops to 0
<Tenkawa>
until next boot
<Tenkawa>
s/aroumd/around
<Tenkawa>
I had been wondering what was causing the high idle
Jackneill_ has quit [Ping timeout: 246 seconds]
BootLayer has quit [Quit: Leaving]
Narrat has joined #riscv
<Tenkawa>
I've also found that the only NVMe drives that seem to work with low IO errors are ones without 4096 blocksize even available (not just in use)
<Tenkawa>
were you getting the IO errors with CONFIG_INIT_ON_ALLOC_DEFAULT_ON on or off?
billchenchina- has quit [Remote host closed the connection]
JSharp has quit [Server closed connection]
JSharp has joined #riscv
<Esmil>
off fails =y works
<Tenkawa>
Not the same here
<Tenkawa>
I would check your drve's blocksize capabilties... that seems to be the most consistent finding I've found so far. 8 different models... only ones that had no problems could "only" use 512 blocksize
<Tenkawa>
The 4096 ones even if they weren't enabled still had problems
<Esmil>
No it's quite consistent. The same drive, same kernel config fails without it, but works with it on
<Tenkawa>
# CONFIG_INIT_ON_ALLOC_DEFAULT_ON is not set
junaid_ has quit [Remote host closed the connection]
<Tenkawa>
haven't had a IO error since 6.5 (yours and aurels builds came out)
<Esmil>
yeah, but even with it off you might get lucky that the allocations just happens to be zero or something that works
<Tenkawa>
no.. on my other unit I get them all the time with the other testing scenario
<Esmil>
try enabling CONFIG_INIT_ON_ALLOC_DEFAULT_ON and see if it changes anything
<Tenkawa>
Ok let me find one of my other drives
<Tenkawa>
Since this one is not having any errors it wont help here
<Tenkawa>
Time to wait for the compile... afk for a while
Leopold has quit [Remote host closed the connection]
Leopold has joined #riscv
zjason` has joined #riscv
zjason has quit [Ping timeout: 250 seconds]
cwittlut has quit [Server closed connection]
cwittlut has joined #riscv
ahs3_ has quit [Server closed connection]
ahs3 has joined #riscv
<mps>
Tenkawa: I don't have to eject/reinsert mmc on boot, if it is inside load is normal when keep it inserted before booting
<Tenkawa>
mps: not here.. soon as I reboot the load goes back up... We are using different u-boot and opensbi setups too if I remember correctly
<Tenkawa>
Hopefully this next kernel tweak will allow the faster mvmes to run better
<mps>
I test mainline u-boot and mostly use vendor/starfive u-boot, and I think behavior is same
<mps>
which kernel tweak?
<Tenkawa>
Esmil posted it above
<mps>
CONFIG_INIT_ON_ALLOC_DEFAULT_ON=y ?
<Tenkawa>
yeah
<Tenkawa>
Seems an odd fix that a kernel hardening change would affect disk io though
<mps>
(Heisenberg)
<Tenkawa>
but yeah.. a lot of oddities ....
<Tenkawa>
and uncertainty
<mps>
I could now rebuild with this enabled
<Tenkawa>
I am just disappointed the other JH7110's (Pine64) aren't going to get USB mainline support in time for 6.6
<Tenkawa>
So they are effectively going to have to wait.
<mps>
merge window for 6.6 is not yet closed
<gurki>
welp its a pine. ppl have low expectations for these to begin with.
<mps>
at least less patches will be needed with 6.6 kernel
<Tenkawa>
gurki: Noone/entites of resources seem to be there to take it on...
<gurki>
Tenkawa: i dont by pine hardware since it always ends up the same way
<gurki>
buy*
<Tenkawa>
gurki: Yeah
<Tenkawa>
Its got potential.. but so disorganized
ntwk has joined #riscv
EchelonX has quit [Quit: Leaving]
frkzoid has joined #riscv
freakazoid332 has quit [Ping timeout: 248 seconds]
<mps>
Tenkawa: I don't see any change with CONFIG_INIT_ON_ALLOC_DEFAULT_ON=y
<Tenkawa>
Yeah it seemed like a strange option to have any effect to me too
<Tenkawa>
I am still going to try it on a drive though just to be thorough....
<Tenkawa>
At the rate everything is going and what I have scheduled it will probably be tuesday.. I'm leaving in the morning for a camping trip
<mps>
also I need a day without keyboard/screen combo
guerby has quit [Ping timeout: 248 seconds]
gurki has quit [Server closed connection]
gurki has joined #riscv
guerby has joined #riscv
Tenkawa has quit [Quit: Was I really ever here?]
guerby has quit [Ping timeout: 246 seconds]
MaxGanzII_ has quit [Remote host closed the connection]
MaxGanzII_ has joined #riscv
loki_val has quit []
<deathmist>
anyone else seeing broken ethernet on next-20230831 with VisionFive 2? I'm seeing starfive-dwmac logs of e.g. "16030000.ethernet: unsupported interface -22" & "probe of 16030000.ethernet failed with error -22" which are similarly also repeated for 16040000.ethernet as well, rebooting a few times didn't change anything
<deathmist>
I've been thinking of hopping off from -next now that most important patches are in mainline pending the v6.6-rc1 tag so perhaps this was my signal ;)
<deathmist>
ok so my last known working build was next-20230828 and the only change between that and next-20230831 was a014c35556b9045ece8426df2b38eb3c5e1c1aa0 which I already linked, there's a good chance it'll just work if I revert that for now
MaxGanzII_ has quit [Remote host closed the connection]
<Esmil>
it's 6.5 + jh7110 patches that's going into 6.6 + pcie and some more from the mailing list
<deathmist>
that likely will boot and work since it won't have that super recent change, I'll try the revert first
<deathmist>
but thanks, I've been interested in something like this and was probably going to do the same but on top of latest mainline so I wouldn't have to do so much work of hunting down the patches etc