Re: [PATCH v3 3/6] target/riscv/tcg: add user flag for profile support

[Andrew Jones]

On Mon, Oct 23, 2023 at 02:00:00PM -0300, Daniel Henrique Barboza wrote:
> 
> 
> On 10/23/23 05:16, Andrew Jones wrote:
> > On Fri, Oct 20, 2023 at 07:39:48PM -0300, Daniel Henrique Barboza wrote:
> > > The TCG emulation implements all the extensions described in the
> > > RVA22U64 profile, both mandatory and optional. The mandatory extensions
> > > will be enabled via the profile flag. We'll leave the optional
> > > extensions to be enabled by hand.
> > > 
> > > Given that this is the first profile we're implementing in TCG we'll
> > > need some ground work first:
> > > 
> > > - all profiles declared in riscv_profiles[] will be exposed to users.
> > > TCG is the main accelerator we're considering when adding profile
> > > support in QEMU, so for now it's safe to assume that all profiles in
> > > riscv_profiles[] will be relevant to TCG;
> > > 
> > > - we'll not support user profile settings for vendor CPUs. The flags
> > > will still be exposed but users won't be able to change them. The idea
> > > is that vendor CPUs in the future can enable profiles internally in
> > > their cpu_init() functions, showing to the external world that the CPU
> > > supports a certain profile. But users won't be able to enable/disable
> > > it;
> > > 
> > > - Setting a profile to 'true' means 'enable all mandatory extensions of
> > > this profile, setting it to 'false' means disabling all its mandatory
> > > extensions. Disabling a profile is discouraged for regular use and will
> > > issue an user warning. User choices for individual extensions will take
> > > precedence, i.e. enabling a profile will not enable extensions that the
> > > user set to 'false', and vice-versa. This will make us independent of
> > > left-to-right ordering in the QEMU command line, i.e. the following QEMU
> > > command lines:
> > > 
> > > -cpu rv64,zicbom=false,rva22u64=true,Zifencei=false
> > > -cpu rv64,zicbom=false,Zifencei=false,rva22u64=true
> > > -cpu rv64,rva22u64=true,zicbom=false,Zifencei=false
> > > 
> > > They mean the same thing: "enable all mandatory extensions of the
> > > rva22u64 profile while keeping zicbom and Zifencei disabled".
> > 
> > Hmm, I'm not sure I agree with special-casing profiles like this. I think
> > the left-to-right processing should be consistent for all. I'm also not
> > sure we should always warn when disabling a profile. For example, if a
> > user does
> > 
> >   -cpu rv64,rva22u64=true,rva22u64=false
> > 
> > then they'll get a warning, even though all they're doing is restoring the
> > cpu model. While that looks like an odd thing to do, a script may be
> > adding the rva22u64=true and the rva22u64=false is the user input which
> > undoes what the script did.
> 
> QEMU options do not work with a "the user enabled then disabled the same 
> option,
> thus it'll count as nothing happened" logic. The last instance of the option 
> will
> overwrite all previous instances. In the example you mentioned above the user 
> would
> disable all mandatory extensions of rva22u64 in the CPU, doesn't matter if the
> same profile was enabled beforehand.

Yup, I'm aware, but I keep thinking that we'll only be using profiles with
a base cpu type. If you start with nothing (a base) and then add a profile
and take the same one away, you shouldn't be taking away anything else. I
agree that if you use a profile on some cpu type that already enabled a
bunch of stuff itself, then disabling a profile would potentially remove
some of those too, but mixing cpu types that have their own extensions and
profiles seems like a great way to confuse oneself as to what extensions
will be present.  IOW, we should be adding a base cpu type at the same
time we're adding these profiles.

> 
> Sure, the can put code in place to make this happen, but then this would make
> profiles act different than regular extensions. "-cpu rv64,zicbom=true -cpu 
> rv64,zicbom=false"
> will disable zicbom, it will not preserve the original 'zicbom' rv64 default. 
> If
> we're going to keep left-to-right ordering consistent, this behavior should 
> also
> be consistent as well.

It will be consistent if we always override whatever was on the left with
what's on the right, which means with

 -cpu rv64,rva22u64=true -cpu rv64,zicbom=false

zicbom will be disabled, but with

 -cpu rv64,zicbom=false -cpu rv64,rva22u64=true

it will be enabled. The same goes if the properties are given to the same
-cpu parameter.

> 
> 
> As for warnings, I agree that we'll throw warnings even when nothing of 
> notice happened.
> For example:
> 
> -cpu rv64,rva22u64=false -cpu rv64,rva22u64=true
> 
> This will throw a warning even though the user ended up enabling the extension
> in the end.
> 
> 
> We can fix it by postponing warnings to realize().
> 
> 
> > 
> > As far as warnings go, it'd be nice to warn when mandatory profile
> > extensions are disabled from an enabled profile. Doing that might be
> > useful for debug, but users which do it without being aware they're
> > "breaking" the profile may learn from that warning. Note, the warning
> > should only come when the profile is actually enabled and when the
> > extension would actually be disabled, i.e.
> > 
> >   -cpu rv64,rva22u64=true,c=off
> > 
> > should warn
> > 
> >   -cpu rv64,c=off,rva22u64=true
> > 
> > should not warn (rva22u64 overrides c=off since it's to the right)
> > 
> >   -cpu rv64,rva22u64=true,rva22u64=false,c=off
> > 
> > should not warn (rva22u64 is not enabled)
> 
> Ack for all the above.
> 
> > 
> > And,
> > 
> >   -cpu rv64,rva22u64=true,rva24u64=false
> > 
> > should warn for each extension which is mandatory in both profiles.
> 
> The way I'm imagining this happening is to cycle through all profiles during 
> realize(),
> see which ones are enabled, and then warn if the user disabled their mandatory
> extensions. In this example we would warn for all rva22 mandatory extensions
> that were disabled because we disabled rva24, but we won't emit any warnings 
> for
> rva24 mandatory extensions given that the profile is marked as disabled.

Yup, sounds good.

Thanks,
drew