qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] RISC-V: virt: This is a "sifive,test1" test finisher


From: Palmer Dabbelt
Subject: Re: [PATCH] RISC-V: virt: This is a "sifive,test1" test finisher
Date: Wed, 20 Nov 2019 18:40:24 -0800 (PST)

On Sun, 10 Nov 2019 13:10:33 PST (-0800), address@hidden wrote:
On Fri, Nov 08, 2019 at 10:13:16AM -0800, Palmer Dabbelt wrote:
On Fri, 08 Nov 2019 10:04:47 PST (-0800), Peter Maydell wrote:
> On Fri, 8 Nov 2019 at 17:15, Alistair Francis <address@hidden> wrote:
> > > > On Fri, Nov 8, 2019 at 9:05 AM Palmer Dabbelt <address@hidden> wrote:
> > >
> > > The test finisher implements the reset command, which means it's a
> > > "sifive,test1" device.  This is a backwards compatible change, so it's
> > > also a "sifive,test0" device.  I copied the odd idiom for adding a
> > > two-string compatible field from the ARM virt board.
> > >
> > > Fixes: 9a2551ed6f ("riscv: sifive_test: Add reset functionality")
> > > Signed-off-by: Palmer Dabbelt <address@hidden>
> > > Signed-off-by: Palmer Dabbelt <address@hidden>
> > > ---
> > >  hw/riscv/virt.c | 5 ++++-
> > >  1 file changed, 4 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/hw/riscv/virt.c b/hw/riscv/virt.c
> > > index 23f340df19..74f2dce81c 100644
> > > --- a/hw/riscv/virt.c
> > > +++ b/hw/riscv/virt.c
> > > @@ -359,7 +359,10 @@ static void create_fdt(RISCVVirtState *s, const 
struct MemmapEntry *memmap,
> > >      nodename = g_strdup_printf("/test@%lx",
> > >          (long)memmap[VIRT_TEST].base);
> > >      qemu_fdt_add_subnode(fdt, nodename);
> > > -    qemu_fdt_setprop_string(fdt, nodename, "compatible", "sifive,test0");
> > > +    {
> > > +        const char compat[] = "sifive,test1\0sifive,test0";
> > > > Does this really work? Why not use qemu_fdt_setprop_cells()? > > > > Alistair > > > > > + qemu_fdt_setprop(fdt, nodename, "compatible", compat, sizeof(compat));
> > > +    }
> > qemu_fdt_setprop_cells() is for "set this property to
> contain this list of 32-bit integers" (and it does a byteswap
> of each 32-bit value from host to BE). That's not what
> you want for a string (or a string list, which is what
> we have here).
> > Cc'ing David Gibson who's our device tree expert to see if there's
> a nicer way to write this. Oddly, given that it's used in the
> ubiquitous 'compatible' prop, the dtc Documentation/manual.txt
> doesn't say anything about properties being able to be
> 'string lists', only 'strings', '32 bit numbers', 'lists of
> 32-bit numbers' and 'byte sequences'. You have to dig through
> the header file comments to deduce that a string list is
> represented by a string with embedded NULs separating
> each list item.

I copied this from hw/arm/virt.c, but messed up.  There they use

       const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
       qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
                        compat, sizeof(compat));

I'm not sure what you're saying is messed up.  AFAICT, this matches
the code you have above, and both should be correct.

Sorry, I must have been hallucinating.  For some reason I though I wrote
qemu_fdt_setprop_string(... compat).

I'd like to take this for 4.2 if possible, but I don't think I have a
reviewed-by (I just got my email set up on my Google computer, so it might be
messy for a bit).  I'm happy to submit the cleaner valist version after 4.2, as
per Peter's suggestion.

Alistair: are you OK with this?

I'll send a v2, but I'd be happy to add some sort of setprop_stringlist
function.  Maybe we just indicate the length with two '\0's?  IIRC that's
how other similar-looking data structures are encoded.


--
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson



reply via email to

[Prev in Thread] Current Thread [Next in Thread]