<Esmil>
smaeul: drewfustini: conchuod: No, the JH7100 doesn't work without the DMA_GLOBAL_POOL unfortunately, so chances are that's the same for the rzfive :/
<Esmil>
*without CONFIG_DMA_GLOBAL_POOL=y i should say. the pool is still there in the device tree
hightower2 has quit [Read error: Connection reset by peer]
<drewfustini>
palmer: Esmil: interesting point by icenowy
<drewfustini>
Well, considering RZ/Five enables some spec-non-conformant local memory
<drewfustini>
(which bypasses MMU) that makes even running generic user space
<drewfustini>
doesn't look like a big deal.
<drewfustini>
binaries not so viable (PIE ones may still run, but those built to be
<drewfustini>
on the default fixed location of binutils will conflict with the MMU-
<drewfustini>
bypassing local memory), not supporting it in a multi-platform kernel
<drewfustini>
Anyways, I'll post a patch that removes the select of DMA_GLOBAL_POOL and the discussion can move there instead of the TH1520 MMC driver series
GenTooMan has quit [Ping timeout: 258 seconds]
<Esmil>
yeah, unfortunately that also means the JH7100 can't be enabled without NONPORTABLE and that and the unmatched are still the two most reliable machines i have :/
<drewfustini>
For the distro perspective, do you think RZ/Five could have been supported by a single userspace image? I am uncertain about the issues there.
GenTooMan has joined #riscv
<Esmil>
i heard someone talk about a really slow workaround by letting opensbi handle faults, but without that it doesn't seem like it
<sorear>
does anyone actually have an rz/five to test? the soc documentation is confusing and I'm not confident in any of the interpretations that are floating around
<sorear>
contradictory on the extent to which the TCM visibility and address can be configured at runtime
<drewfustini>
I was looking last night to see if I could get one but I only saw the official eval board which was about $1,000
<drewfustini>
which is beyond my curiosity budget :)
<geertu>
sorear: I have an RZ/Five, and got the latest firmware from Prabhakar earlier today.
jacklsw has quit [Quit: Back to the real world]
<geertu>
Until now, my testing was limited (boot modern kernel until unable to mount root, or boot old kernel the board came with)
<geertu>
drewfustini: 560.42€ < $1000, AFAIK ;-)
<geertu>
OK, still above curiosity budget
<drewfustini>
oh, I didn't see that
<drewfustini>
yeah :)
<geertu>
search RZ/Five at digikey
prabhakarlad has joined #riscv
<geertu>
OMG, Digikey has a picture with the CAN-transceiver mounted ;-)
<drewfustini>
geertu: what do you think is a good solution for the DMA_GLOBAL_POOL issue? One hand, I can add a global pool to my th1520 device tree and it works okay with the defconfig. On the other hand, I'm not sure if that is the right decision?
<geertu>
drewfustini: I guess the kernel should be fixed to handle both cases
<geertu>
Cfr. what Robin says
<drewfustini>
I'm less certain about how that could be done, and ideally I would like to decouple it from my mmc series. I would love to have booting from eMMC on the th1520 in 6.7.
<geertu>
Sure
<drewfustini>
Palmer suggested I send a patch to remove the DMA_GLOBAL_POOL select.and that could start the thread of discussion
<drewfustini>
alright, then that serves the purpose of bringing the discussion out of the mmc series and hopefully making sure everyone notices the discussion
<geertu>
Depending on non-portable would be fine for now, I guess. And could be final, if it was just for RZ/Five, as it cannot run a standard distro anyway :-(
<geertu>
But if you bring JH7100 and Starlight (and what else?) into the picture...
<drewfustini>
Esmil: It just effect JH7100, right? JH7110 is okay?
<drewfustini>
geertu: " it cannot run a standard distro anyway" thanks for clarifying, i was uncertain about that
<Esmil>
drewfustini: yes, for most peripherals, but i'm told some of the graphics may be non-coherent
<geertu>
What about ERRATA_THEAD_CMO?
<drewfustini>
do you mean as to whether it should be selected?
<drewfustini>
ARCH_THEAD selects ERRATA_THEAD_CMO. defconfig has CONFIG_ARCH_THEAD=y. Are you thinking that should change?
jacklsw has joined #riscv
<Esmil>
geertu: the ubuntu kernels already enable ERRATA_THEAD_CMO and co-exists with sifive, microchip and JH7110 support just ine
<Esmil>
(from the lunar 6.2 kernel onwards that is)
hightower2 has joined #riscv
Kedleston has quit [Ping timeout: 260 seconds]
Kedleston has joined #riscv
mwette has joined #riscv
crabbedhaloablut has quit [Read error: Connection reset by peer]
crabbedhaloablut has joined #riscv
Andre_Z has joined #riscv
GenTooMan has quit [Read error: No route to host]
cybernaut has joined #riscv
cybernaut has quit [Read error: Connection reset by peer]
GenTooMan has joined #riscv
ntwk has joined #riscv
Andre_Z has quit [Quit: Leaving.]
Kedleston has quit [Ping timeout: 240 seconds]
Kedleston_ has joined #riscv
Kedleston_ is now known as Kedleston
jacklsw has quit [Ping timeout: 255 seconds]
Andre_Z has joined #riscv
Andre_Z has quit [Remote host closed the connection]