Re: [PATCH v4 1/2] arch/riscv: initial RISC-V support to build/run minimal Xen

[Oleksii <oleksii.kurochko@xxxxxxxxx>]

On Thu, 2023-01-05 at 15:48 +0000, Andrew Cooper wrote:
> On 05/01/2023 1:40 pm, Jan Beulich wrote:
> > On 05.01.2023 13:01, Oleksii Kurochko wrote:
> > > To run in debug mode should be done the following instructions:
> > >  $ qemu-system-riscv64 -M virt -smp 1 -nographic -m 2g \
> > >         -kernel xen/xen -s -S
> > >  # In separate terminal:
> > >  $ riscv64-buildroot-linux-gnu-gdb
> > >  $ target remote :1234
> > >  $ add-symbol-file <xen_src>/xen/xen-syms 0x80200000
> > >  $ hb *0x80200000
> > >  $ c # it should stop at instruction j 0x80200000 <start>
> > This suggests to me that Xen is meant to run at VA 0x80200000,
> > whereas ...
> > 
> > > --- a/xen/arch/riscv/include/asm/config.h
> > > +++ b/xen/arch/riscv/include/asm/config.h
> > > @@ -1,6 +1,9 @@
> > >  #ifndef __RISCV_CONFIG_H__
> > >  #define __RISCV_CONFIG_H__
> > >  
> > > +#include <xen/const.h>
> > > +#include <xen/page-size.h>
> > > +
> > >  #if defined(CONFIG_RISCV_64)
> > >  # define LONG_BYTEORDER 3
> > >  # define ELFSIZE 64
> > > @@ -28,7 +31,7 @@
> > >  
> > >  /* Linkage for RISCV */
> > >  #ifdef __ASSEMBLY__
> > > -#define ALIGN .align 2
> > > +#define ALIGN .align 4
> > >  
> > >  #define ENTRY(name)                                \
> > >    .globl name;                                     \
> > > @@ -36,6 +39,10 @@
> > >    name:
> > >  #endif
> > >  
> > > +#define XEN_VIRT_START  _AT(UL, 0x00200000)
> > ... here you specify a much lower address (and to be honest even
> > 0x80200000
> > looks pretty low to me for 64-bit, and perhaps even for 32-bit).
> > Could you
> > clarify what the plans here are? Maybe this is merely a temporary
> > thing,
> > but not called out as such?
> 
> It's stale from v1 which had:
> 
> #define XEN_VIRT_START  0x80200000
> 
> 
> But honestly, I don't think the qemu details in the commit message
> are
> useful.  This series is just about making "make build" work.
> 
> The next series (being worked on, but not posted yet) is only a few
> patches and gets a full Gitlab CI smoke test, at which point the
> smoke
> test shell script is the reference for how to invoke qemu.
> 
> 
> I'm happy to R-by this series and drop that part of the commit
> message
> on commit.
> 
I'm happy with that. Thanks.
> ~Andrew