02:36
zsun has joined #fedora-riscv
02:49
pbsds has joined #fedora-riscv
03:09
davidlt has joined #fedora-riscv
03:17
fuwei has joined #fedora-riscv
03:52
zsun has quit [Quit: Leaving.]
04:20
davidlt has quit [Ping timeout: 255 seconds]
04:30
fuwei has quit [Quit: Konversation terminated!]
05:14
masami has joined #fedora-riscv
05:15
masami has quit [Client Quit]
06:12
davidlt has joined #fedora-riscv
06:18
<
davidlt >
efi zboot kernel works with edk2 firmware
06:18
<
davidlt >
(which was fully expected)
09:06
<
davidlt >
yeah, so once we get edk2-stable bumped in Fedora I can continue with Pungi
09:06
<
davidlt >
I haven't looked yet, but annaconda might be the next thing I need to review
10:13
fuwei has joined #fedora-riscv
10:54
underscore has joined #fedora-riscv
11:16
underscore has quit [Quit: Quit]
11:44
<
davidlt >
I am starting a new GCC 13 build with a couple more backports
11:55
<
rwmjones >
hey davidlt ... on the subject of "cpuid"
11:55
<
rwmjones >
am I right in thinking that the only two mechanisms we have today are
11:56
<
rwmjones >
MISA CSR (which only has a very simplistic "alphabetic" list of extensions)
11:56
<
rwmjones >
riscv,isa string in device tree, which seems to be fairly free-form
11:56
<
rwmjones >
nothing obvious in SBI, except it offers a way to read misa
11:56
<
rwmjones >
are there any extension proposals I'm missing?
11:57
<
rwmjones >
like it'd be nice to be able to get all the extensions (including S* and Z*) and their version numbers
11:57
<
rwmjones >
that's a minimum really
11:57
<
davidlt >
The only thing that we should care about is hwprobe syscall (only starting v6.4 kernel)
11:57
<
davidlt >
MISA CSR basically shows nothing interesting anymore
11:58
<
davidlt >
and riscv,isa is kinda the same
11:58
<
davidlt >
glibc ifunc decisions are made using hwprobe info
11:59
<
rwmjones >
ok hwprobe (arch/riscv/include/uapi/asm/hwprobe.h) seems limited as well
11:59
<
davidlt >
there are also discussions about what cpuinfo should show
11:59
* rwmjones
pulls latest linux
12:00
<
davidlt >
there is also hwcap, which has way more stuff than MISA
12:00
<
davidlt >
but it will run out of bits sooner than later :)
12:00
<
davidlt >
hwprobe is kinda small, but it has stuff that kinda exists in SoCs.
12:01
<
rwmjones >
I'm just looking at how hwprobe is implemented
12:01
<
davidlt >
Today your biggest new extensions in SoC land is Zba and Zbb :) Not much.
12:01
<
davidlt >
It's key-value thing IIRC. You can check glibc patches.
12:02
<
davidlt >
Also T-HEAD has fast misaligned access (i.e. handled in Hw side), while U74 doesn't.
12:02
<
rwmjones >
hwprobe is basically using SBI which is using misa afaict
12:03
<
davidlt >
That information most likely will come from different sources
12:03
<
rwmjones >
& using DT riscv,isa too
12:05
<
rwmjones >
looks like ventana added ACPI support for the isa string
12:05
<
davidlt >
yeah, but it also could come from ACPI tables
12:06
<
davidlt >
Ventana is now working on ACPI bits IIRC (the whole stack).
12:08
<
davidlt >
IIUC ECRs by UEFI/ACPI Forum have been approved
12:10
<
davidlt >
rwmjones, qemu also got (or patches are available) for max cpu variant.
12:10
<
davidlt >
You could toy with that + hwprobe if you want.
12:18
<
davidlt >
RISCV Dev Partners program is something new to me.
12:20
<
rwmjones >
excellent! ... and... intel..?!?
12:20
<
davidlt >
well, it's kinda an old thing :)
12:20
<
davidlt >
Intel has been pushing riscv patch in some projects for some time now
12:21
<
davidlt >
Intel is one of founding RISE members IIRC
12:24
<
rwmjones >
wouldn't that just work anyway?
12:24
<
davidlt >
I think I saw U-Boot patch this morning
12:24
<
rwmjones >
ok I guess "x86 has identity-mapped requirements for the MSI doorbell and these (APIC/IOAPIC) live right below 4GiB." is the problem
12:25
<
davidlt >
rwmjones, it seems it requires a bit of work.
12:25
<
davidlt >
QEMU, U-Boot and EDK2 side.
12:25
<
rwmjones >
let me send that page to alex
12:25
<
davidlt >
I haven't seen this for some time.
12:26
<
davidlt >
"Vector extension discovery using HWPROBE" is also on the list
12:27
<
davidlt >
I really want this 'SE_01_011 - QEMU RVA22U64/S64 profile support'
12:27
<
davidlt >
Seems to be 8.2 QEMU as a target.
12:29
<
davidlt >
o, "StandaloneMmPkg" for EDK2 is also on the list
12:30
<
davidlt >
There are also 24 H1 priorities too.
12:31
<
davidlt >
that incl. RVA23 extension support
12:31
<
davidlt >
(on the kernel side)
12:32
<
davidlt >
tons of ACPI related work (which means no proper ACPI until maybe late 2024?)
12:32
<
davidlt >
I bet that will happen around new year (that's typical IIRC)
12:33
<
davidlt >
"Debug console support" would be nice to have, I already noticed patches for it
12:34
zsun has joined #fedora-riscv
12:52
fuwei has quit [Ping timeout: 245 seconds]
12:52
fuwei has joined #fedora-riscv
13:13
kwizart has left #fedora-riscv [Leaving]
13:57
iooi has quit [Quit: iooi]
13:58
iooi has joined #fedora-riscv
14:03
iooi has quit [Ping timeout: 255 seconds]
14:25
fuwei has quit [Quit: Konversation terminated!]
15:56
zsun has quit [Quit: Leaving.]
17:03
jednorozec has joined #fedora-riscv
17:31
<
dgilmore >
I broke NVMe on the VF2
17:33
<
dgilmore >
I updated u-boot to upstream. and it works in u-boot with the patch about to be merged added. however the debian image I have on the sdcard until I get Fedora on it throws timeout errors
17:33
<
dgilmore >
I guess their u-boot was doing something that upstream is not
18:18
jednorozec has joined #fedora-riscv
18:30
davidlt has quit [Ping timeout: 255 seconds]
18:50
jednorozec has joined #fedora-riscv
21:51
rwmjones has quit [Ping timeout: 248 seconds]
21:51
rwmjones_ has joined #fedora-riscv