qemu-devel
[Top][All Lists]
Advanced

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

Re: [Qemu-devel] [PATCH 20/24] machine: drop MachineState::cpu_model


From: Eduardo Habkost
Subject: Re: [Qemu-devel] [PATCH 20/24] machine: drop MachineState::cpu_model
Date: Thu, 18 Jan 2018 17:18:09 -0200
User-agent: Mutt/1.9.1 (2017-09-22)

On Thu, Jan 18, 2018 at 11:10:35AM +0100, Igor Mammedov wrote:
> On Wed, 17 Jan 2018 23:48:46 -0200
> Eduardo Habkost <address@hidden> wrote:
> 
> > On Wed, Jan 17, 2018 at 04:43:32PM +0100, Igor Mammedov wrote:
> > > The last user of it was machine type 'none', which used field
> > > to create CPU id user requested it on CLI with -cpu option.
> > > 
> > > We could compare pointers of MachineState::cpu_type and
> > > MachineClass::default_cpu_type to check for the same condition,
> > > and drop cpu_model concept completly from machine/boards code
> > > So that no one would try to reuse obsolete field and only
> > > place to deal with cpu model would be vl.c and
> > > foo_cpu_class_by_name() callbacks.
> > > 
> > > Signed-off-by: Igor Mammedov <address@hidden>
> > > ---
> > > CC: Eduardo Habkost <address@hidden>
> > > CC: Marcel Apfelbaum <address@hidden>
> > > CC: Paolo Bonzini <address@hidden>
> > > ---
> > >  include/hw/boards.h    |  1 -
> > >  hw/core/null-machine.c | 10 +++++++---
> > >  vl.c                   |  8 +++++++-
> > >  3 files changed, 14 insertions(+), 5 deletions(-)
> > > 
> > > diff --git a/include/hw/boards.h b/include/hw/boards.h
> > > index 156b16f..decd0ec 100644
> > > --- a/include/hw/boards.h
> > > +++ b/include/hw/boards.h
> > > @@ -246,7 +246,6 @@ struct MachineState {
> > >      char *kernel_filename;
> > >      char *kernel_cmdline;
> > >      char *initrd_filename;
> > > -    const char *cpu_model;
> > >      const char *cpu_type;
> > >      AccelState *accelerator;
> > >      CPUArchIdList *possible_cpus;
> > > diff --git a/hw/core/null-machine.c b/hw/core/null-machine.c
> > > index 864832d..c2e466c 100644
> > > --- a/hw/core/null-machine.c
> > > +++ b/hw/core/null-machine.c
> > > @@ -23,10 +23,13 @@
> > >  static void machine_none_init(MachineState *mch)
> > >  {
> > >      CPUState *cpu = NULL;
> > > +    MachineClass *mc = MACHINE_GET_CLASS(mch);
> > >  
> > > -    /* Initialize CPU (if a model has been specified) */
> > > -    if (mch->cpu_model) {
> > > -        cpu = cpu_init(mch->cpu_model);
> > > +    /* Initialize CPU if cpu_type pointer is user provided
> > > +     * (i.e. != to pointer tot static default cpu type string)
> > > +     */
> > > +    if (mch->cpu_type != mc->default_cpu_type) {
> > > +        cpu = cpu_create(mch->cpu_type);  
> > 
> > This is a big assumption about the code that sets mch->cpu_type.
> > A simple g_strdup(machine_class->default_cpu_type) would break
> > this silently (as it won't trigger the assert() below).
> Yes, it's a bit of a hack to figure out is user has requested
> cpu type explicitly. But so far there isn't need to do
>    g_strdup(machine_class->default_cpu_type)
> when copying default in vl.c and guarding against it
> looks like overkill currently.
> 
> Cleaner way would be to make cpu_type property, add new property
> API to set/check flags and use that here, there are other places
> that would benefit from such API as well.
> But it looks beyond scope of this series, so I used simple
> hackish way to make it work with current code.
> 
> I can amend comment here:
>     /* Initialize CPU if cpu_type pointer is user provided
>      * (i.e. != to pointer to static default cpu type string)
>      * MachineClass::default_cpu_type must be assigned to
>      * MachineState::cpu_type directly for this to work.
>      * TODO:
>      *   - make cpu_type a property
>      *   - add API to add/check user_set flag to property
>      *   - use new API to check if property was user set
>      */
> 
> > 
> > 
> > >          if (!cpu) {
> > >              error_report("Unable to initialize CPU");
> > >              exit(1);
> > > @@ -54,6 +57,7 @@ static void machine_none_machine_init(MachineClass *mc)
> > >      mc->init = machine_none_init;
> > >      mc->max_cpus = 1;
> > >      mc->default_ram_size = 0;
> > > +    mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;  
> > 
> > Why do you need this?  Isn't it simpler to just leave
> > default_cpu_type=NULL here?
> vl.c "-cpu" parsing depends on default_cpu_type being set
>  ...
>  if (machine_class->default_cpu_type) {
>    ...
>  }

Right, this makes sense now.

It looks like default_cpu_type is being overloaded for two
different roles: 1) specifying the default CPU type; 2) finding
the arch-specific class to be used to parse -cpu.

In the case of null-machine, these two roles conflict with each
other.  I believe we can find other solutions instead of this
hack that involves lying on MachineClass::default_cpu_type (and
then having to work around the lie on machine_none_init()).

I see multiple options: adding a new MachineClass field for that
(e.g.  resolving_cpu_type, which defaults to default_cpu_type if
NULL); moving the CPU parsing code to arch_init.c (so it could
use CPU_RESOLVING_TYPE or something similar); adding a optional
MachineClass::parse_cpu_model hook.  We could even try to get rid
of CPUClass::parse_features completely


> when we get rid of requirement for proxy type in
>  cpu_parse_cpu_model()
> we can drop this ugliness in null-machine.

I'd prefer to not introduce this ugliness in the first place.

> 
> I'm doing it dirty way to prevent cpu_model resurgence
> in boards code as it happened with nios2, even though
> I've tried to monitor list for such patches.

To be honest, I don't think the harm in having new code using
MachineState::cpu_model is so big to justify this hack.


>  
> > >  }
> > >  
> > >  DEFINE_MACHINE("none", machine_none_machine_init)
> > > diff --git a/vl.c b/vl.c
> > > index 2586f25..8aa0131 100644
> > > --- a/vl.c
> > > +++ b/vl.c
> > > @@ -4609,7 +4609,6 @@ int main(int argc, char **argv, char **envp)
> > >      current_machine->maxram_size = maxram_size;
> > >      current_machine->ram_slots = ram_slots;
> > >      current_machine->boot_order = boot_order;
> > > -    current_machine->cpu_model = cpu_model;
> > >  
> > >      parse_numa_opts(current_machine);
> > >  
> > > @@ -4619,6 +4618,13 @@ int main(int argc, char **argv, char **envp)
> > >          if (cpu_model) {
> > >              current_machine->cpu_type =
> > >                  cpu_parse_cpu_model(machine_class->default_cpu_type, 
> > > cpu_model);
> > > +
> > > +            /* machine 'none' depends on default cpu type pointer not 
> > > being
> > > +             * equal to resolved type name pointer to fugure out if type 
> > > was
> > > +             * user provided, make sure that if it becomes not true in 
> > > future
> > > +             * it won't beark silently */
> > > +            g_assert(
> > > +                current_machine->cpu_type != 
> > > machine_class->default_cpu_type);
> > >          }
> > >      }
> > >  
> > > -- 
> > > 2.7.4
> > >   
> > 
> 

-- 
Eduardo



reply via email to

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