[U-Boot] [PATCH 12/12] riscv: Add QEMU virt board support

[Auer, Lukas]

Hi Bin,

On Thu, 2018-09-06 at 11:14 +0800, Bin Meng wrote:
> Hi Lukas,
> 
> On Wed, Sep 5, 2018 at 5:35 PM Auer, Lukas
> <lukas.auer at aisec.fraunhofer.de> wrote:
> > 
> > On Wed, 2018-09-05 at 09:28 +0800, Rick Chen wrote:
> > >  > > From: Auer, Lukas [mailto:lukas.auer at aisec.fraunhofer.de]
> > >  > > Sent: Wednesday, September 05, 2018 5:53 AM
> > >  > > To: bmeng.cn at gmail.com
> > >  > > Cc: Rick Jian-Zhi Chen(陳建志); u-boot at lists.denx.de
> > >  > > Subject: Re: [U-Boot] [PATCH 12/12] riscv: Add QEMU virt
> > > board
> > > support
> > >  > >
> > >  > > On Tue, 2018-09-04 at 17:31 +0800, Bin Meng wrote:
> > >  > > > Hi Lukas,
> > >  > > >
> > >  > > > On Tue, Sep 4, 2018 at 5:39 AM Auer, Lukas
> > >  > > > <lukas.auer at aisec.fraunhofer.de> wrote:
> > >  > > > >
> > >  > > > > On Thu, 2018-08-30 at 00:54 -0700, Bin Meng wrote:
> > >  > > > > > This adds QEMU RISC-V 'virt' board target support,
> > > with
> > > the hope
> > >  > > > > > of helping people easily test U-Boot on RISC-V.
> > >  > > > > >
> > >  > > > > > The QEMU virt machine models a generic RISC-V virtual
> > > machine
> > >  > > > > > with support for the VirtIO standard networking and
> > > block
> > >  > > > > > storage devices.
> > >  > > > > > It has CLINT, PLIC, 16550A UART devices in addition to
> > > VirtIO
> > >  > > > > > and it also uses device-tree to pass configuration
> > > information
> > >  > > > > > to guest software. It implements RISC-V privileged
> > > architecture
> > >  > > > > > spec v1.10.
> > >  > > > > >
> > >  > > > > > Both 32-bit and 64-bit builds are supported. Support
> > > is
> > > pretty
> > >  > > > > > much preliminary, only booting to U-Boot shell with
> > > the
> > > UART
> > >  > > > > > driver on a single core. Booting Linux is not
> > > supported
> > > yet.
> > >  > > > > >
> > >  > > > >
> > >  > > > > For your information and to avoid duplicate work, I am
> > > working on
> > >  > > > > a patch set that improves RISC-V support in u-boot. I am
> > > currently
> > >  > > > > able to boot Linux on a multi-core setup in QEMU, but
> > > they
> > > are not
> > >  > > > > quite ready to submit yet.
> > >  > > > >
> > >  > > >
> > >  > > > This is great! My next step is to work on virtio driver
> > > support in
> > >  > > > U-Boot as qemu-riscv virt machine has these devices but we
> > > don't
> > >  > > > have corresponding drivers in U-Boot. Then I will try to
> > > boot
> > > Linux
> > >  > > > after that. Good to hear you already boot Linux with qemu-
> > > riscv!
> > >  > > > Have you already supported virtio drivers in your port? If
> > > yes, I
> > >  > > > will just hold on and wait for your patch series :-)
> > >  > > >
> > >  > >
> > >  > > Hi Bin,
> > >  > >
> > >  > > Support for the virtio devices would be great! I don't
> > > support
> > > them in
> > >  > > my port, I can only boot a kernel image from RAM.
> > >  > > I only have a driver for the clint0 (core local interrupt
> > > controller),
> > >  > > which I need for software interrupts to other cores and as a
> > > timer.
> > >  > > Software interrupts also work over the supervisor binary
> > > interface
> > >  > > (SBI), which allows u-boot to run in supervisor mode with
> > > bbl
> > > running
> > >  > > in machine mode to handle the SBI calls.
> > >  > >
> > > 
> > > Hi Bin and Auer
> > > 
> > > I have already boot bbl run in S-mode and riscv-linux in M-mode
> > > via
> > > u-boot from SD card or FLASH.
> > > It mean after booting riscv-linux, u-boot will be dead. And no
> > > matter
> > > about kernel.
> > > Please refer to doc/README.ae350
> > > 
> > > May I figure out more clearly what are you going to do ?
> > > What are you going to do is let u-boot run in S-mode and boot bbl
> > > and
> > > riscv-linux in M-mode, right ?
> > > It mean after booting bbl and riscv-linux, u-boot will still
> > > alive
> > > and
> > > handle SBI calls and somethings in S-mode.
> > > 
> > > Or u-boot is going to replace the role of bbl ?
> > > 
> > > Rick
> > > 
> > 
> > Hi Rick,
> > 
> > Not exactly, my current boot flow is as follows.
> > 
> > 1. u-boot SPL starts in machine mode and jumps to bbl
> > 2. bbl starts u-boot proper in supervisor mode
> > 3. u-boot boots the kernel
> > 
> 
> If this is QEMU virt target, there is no need to boot from SPL then
> U-Boot proper.
> 

That's true. My goal was to load bbl without having to embed either  
u-boot or Linux in it. That boot flow is similar to armv8, where ARM
trusted firmware can be loaded by SPL.
With an SBI implementation in u-boot this is not required of course. I
think I'll simplify my boot flow for now and remove SPL. Either u-boot
or Linux must then be embedded into bbl until we have our own SBI
implementation.

Thanks,
Lukas

> > bbl is still active once Linux has booted and is used there for its
> > SBI
> > implementation. Hope this helps.
> 
> Yes, I am not quite convinced why Linux kernel was designed this way.
> This is something like x86's SMM or EFI runtime services...
> 
> Regards,
> Bin