MaxGanzII has quit [Remote host closed the connection]
MaxGanzII has joined #riscv
jcarr has joined #riscv
stolen has joined #riscv
ntwk has quit [Read error: Connection reset by peer]
notgull has joined #riscv
mlw has quit [Ping timeout: 264 seconds]
mlw has joined #riscv
jcarr has quit [Ping timeout: 245 seconds]
mahk has quit [Ping timeout: 260 seconds]
psydroid2 has joined #riscv
<drewfustini>
Esmil: or anyone else, do you know if the PLDA PCIe controller for the JH7110 can distribute interrupts to all the cores?
andyc has quit [Quit: Connection closed for inactivity]
<sorear>
it's not using a PLIC for interrupt distribution?
MaxGanzII has quit [Remote host closed the connection]
jcarr has joined #riscv
MaxGanzII has joined #riscv
<Tenkawa>
this shows up in my dmesg
<Tenkawa>
[ 0.000000] plic: interrupt-controller@c000000: mapped 136 interrupts with 4 handlers for 9 contexts.
josuah has joined #riscv
josuah has quit [Client Quit]
MaxGanzII has quit [Remote host closed the connection]
josuah has joined #riscv
MaxGanzII has joined #riscv
<Esmil>
drewfustini: hmm.. you're right after letting my vf2 idle a bit and cat'ing /proc/interrupts it seems like all the interrupts are taken on cpu0
<Esmil>
did you try setting the irq affinities?
A1ice has quit [Ping timeout: 240 seconds]
<drewfustini>
Thanks, let me check
<Esmil>
cat /proc/irq/{47,48,49,50,51,52}/smp_affinity returns f f 1 2 4 8, so it looks like they should be already spread out. yet they're all still taken on cpu0. i guess the driver still has room for improvement
<Esmil>
(those are the nvmeXq interrupts)
<Esmil>
*nvme0qX rather
Andre_Z has quit [Quit: Leaving.]
<Tenkawa>
interesting
<Tenkawa>
Esmil: I get f only
heat_ has joined #riscv
<Tenkawa>
ok I found a few diff.. checking what procs they are
<Esmil>
i happened to have booted this board on archlinux. i wonder if the archlinux udev rules has some logic to spread the irqs out
heat has quit [Read error: Connection reset by peer]
<Tenkawa>
I found 4 irqs with different affinity
<Tenkawa>
out of 60
___nick___ has joined #riscv
___nick___ has quit [Client Quit]
___nick___ has joined #riscv
jcarr has quit [Remote host closed the connection]
jcarr has joined #riscv
<sorear>
what's the problem that you're hoping to solve by sending interrupts to all the cores?
stolen has quit [Quit: Connection closed for inactivity]
BootLayer has quit [Quit: Leaving]
<Tenkawa>
sorear: agreed.. I haven't seen any interrupt conflicts (as of yet)
somlo has quit [Remote host closed the connection]
ntwk has joined #riscv
<drewfustini>
Not a problem per se. I was just curious if it was capable of spreading it out.
<drewfustini>
Thanks for the feedback Tenkawa and Esmil. I'm thinking about trying the latest PLDA patches from linux-riscv to see if there is a difference
stolen has joined #riscv
notgull has quit [Ping timeout: 255 seconds]
somlo has joined #riscv
shamoe has quit [Quit: Connection closed for inactivity]
dionysos is now known as dilfridge
davidlt has quit [Ping timeout: 256 seconds]
hightower2 has joined #riscv
<hightower2>
Hey for Beagleboard Ahead, can I somehow boot the Linux install from USB and install like a normal machine, or I really must flash the image onto internal storage?
<Tenkawa>
hightower2: it has a microsd on it right?
<hightower2>
it does, yes (in addition to internal storage)
<Tenkawa>
You should be able to chainload a usb root with a microsd boot
<Tenkawa>
I do it with my XFS root filesystems now
<Tenkawa>
on other sbcs
<hightower2>
or if it would automatically prefer microsd boot, I could put the installer onto that?
<hightower2>
(directly)
<hightower2>
lemme try
<hightower2>
nah still boots internal first. I just need a way to boot anything other than internal storage
<drewfustini>
hightower2: the trouble with booting the ahead from SD is due to thead's u-boot spl. I believe what would work is using the U-Boot SPL -> OpenSBI -> U-Boot from the eMMC and then have it boot from the SD card.
<drewfustini>
the trouble with booting the ahead from _directly_ SD is due to thead's u-boot spl.
<Tenkawa>
drewfustini: yeah even in the arm world we recompile u-boot quite often to change the boot order
<Tenkawa>
(and sometimes just to add options/settings)
<hightower2>
yeah I realized that this isn't a normal thing where I'd just boot... and even then there's image only for yocto and ubuntu
<hightower2>
bottom line, not a gadget I'm gonna waste my time on
<hightower2>
thanks for comments/input
mwette has quit [Quit: ERC 5.5.0.29.1 (IRC client for GNU Emacs 29.1)]
<palmer>
is anyone else seeing some spinlock magic BUGs? Mine started last night, backtrace shows sbi_ipi_init()
mlw has quit [Ping timeout: 255 seconds]
<Tenkawa>
palmer: which unit?
<khem>
boot order could be a u-boot env thing. why recompile ?
<Tenkawa>
khem: that's my point... we had a vendor that built their u-boot to boot onboard emmc - microsd
<Tenkawa>
a bit of a problem when the emmc crashes/etc
<Tenkawa>
(with no access to reset it or the spi)
<Tenkawa>
I never said this was a "good scenario"
___nick___ has quit [Ping timeout: 245 seconds]
jfsimon1981_c has joined #riscv
jfsimon1981_c has quit [Remote host closed the connection]
<khem>
palmer: I should add your tree to my CI for vf2, we can spin full images and run yocto ptests
<drewfustini>
Esmil: I see similar to what you did in /proc/irq. I have ubuntu 23.10 on vf2 with nvme. smp_affinity is 0xf for all the PCIe interrupts other than nvme0q{1-4}, yet, all the interrupts are serviced by cpu0. https://gist.github.com/pdp7/2c29d5fc052bbea22707311970510430
foton_x has joined #riscv
foton has quit [Ping timeout: 260 seconds]
<Esmil>
drewfustini: the kernel i had booted earlier was my jh7110 branch with Minda's latest pci patches
<drewfustini>
Ah okay, so it seems it is lacking that functionality to spread the interrupts across cpus
<palmer>
khem: ya, if you've got something for CI that'd be awesome
heat_ is now known as heat
notgull has quit [Ping timeout: 240 seconds]
MaxGanzII has quit [Ping timeout: 240 seconds]
jjido has quit [Quit: My laptop has gone to sleep. ZZZzzz…]
thodg has quit [Read error: Connection reset by peer]