<xypron>
When running EDK II under KVM a crash (EXCEPT_RISCV_ILLEGAL_INST) occurs when hitting OpenSSL's 'fcvt.d.w fa5,a5' instruction. OpenSSL is used in the EDK II TLS implementation. When running with the QEMU TCG accelerator the crash does not occur.
<xypron>
In the profile spec I find: The RVA22S64 mandatory unprivileged extensions include all the mandatory unprivileged extensions in RVA22U64. So shouldn't floating point be supported under KVM in VS mode?
craigo has joined #riscv
damian101 has quit [Ping timeout: 252 seconds]
damian101_ has joined #riscv
damian101 has joined #riscv
damian101_ has quit [Ping timeout: 252 seconds]
___nick___ has joined #riscv
<drmpeg>
xypron: I've found a weird regression in u-boot 2024.01 for HiFive Unmatched.
damian101 has quit [Ping timeout: 252 seconds]
test924 has quit [Ping timeout: 252 seconds]
<xypron>
drmpeg: Do you mean upstream? Or Ubuntu?
<drmpeg>
Upstream.
<xypron>
The problem with EDK II is reproducible with Linux next-20240913.
<xypron>
drmpeg: Do you have some more detail?
<drmpeg>
Yes, it takes about 10 seconds for SPL to run.
<xypron>
Is the problem reproducible with origin/master and origin/next?
<drmpeg>
I've bisected it, and have the solution. The weird part is I don't see this on VF2.
<drmpeg>
It's a bit over my head as to what's going on.
<drmpeg>
But if I revert that patch, it works as before.
<drmpeg>
Do you still have an Unmatched?
<xypron>
drmpeg: Yes I have an Unmatched board. - What was the issue that you saw? We don't carry the patch in Ubuntu's U-Boot and it boots fine for me. I would not know why calling clk_enable() twice should lead to problems.
dlan has quit [Ping timeout: 252 seconds]
pabs3 has quit [Ping timeout: 252 seconds]
<drmpeg>
There's just a long 10 second pause from when it prints "Trying to boot from MMC1" to when it gets to u-boot.
dlan has joined #riscv
<drmpeg>
I'm seeing this on the Ubuntu version of u-boot also.
coldfeet has quit [Remote host closed the connection]