qemu-devel
[Top][All Lists]
Advanced

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

Re: [PATCH v10 11/12] hw/pci: Convert rom_bar into OnOffAuto


From: Michael S. Tsirkin
Subject: Re: [PATCH v10 11/12] hw/pci: Convert rom_bar into OnOffAuto
Date: Wed, 3 Jul 2024 09:35:56 -0400

On Wed, Jul 03, 2024 at 01:00:21PM +0200, BALATON Zoltan wrote:
> On Wed, 3 Jul 2024, Michael S. Tsirkin wrote:
> > On Wed, Jul 03, 2024 at 04:15:23AM +0200, BALATON Zoltan wrote:
> > > On Tue, 2 Jul 2024, Michael S. Tsirkin wrote:
> > > > On Thu, Jun 27, 2024 at 03:08:00PM +0900, Akihiko Odaki wrote:
> > > > > rom_bar is tristate but was defined as uint32_t so convert it into
> > > > > OnOffAuto.
> > > > > 
> > > > > Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
> > > > 
> > > > Commit log should explain why this is an improvement,
> > > > not just what's done.
> > > > 
> > > > 
> > > > > diff --git a/docs/igd-assign.txt b/docs/igd-assign.txt
> > > > > index e17bb50789ad..35c6c8e28493 100644
> > > > > --- a/docs/igd-assign.txt
> > > > > +++ b/docs/igd-assign.txt
> > > > > @@ -35,7 +35,7 @@ IGD has two different modes for assignment using 
> > > > > vfio-pci:
> > > > >        ISA/LPC bridge device (vfio-pci-igd-lpc-bridge) on the root 
> > > > > bus at
> > > > >        PCI address 1f.0.
> > > > >      * The IGD device must have a VGA ROM, either provided via the 
> > > > > romfile
> > > > > -      option or loaded automatically through vfio (standard).  
> > > > > rombar=0
> > > > > +      option or loaded automatically through vfio (standard).  
> > > > > rombar=off
> > > > >        will disable legacy mode support.
> > > > >      * Hotplug of the IGD device is not supported.
> > > > >      * The IGD device must be a SandyBridge or newer model device.
> > > > 
> > > > ...
> > > > 
> > > > > diff --git a/hw/vfio/pci-quirks.c b/hw/vfio/pci-quirks.c
> > > > > index 39dae72497e0..0e920ed0691a 100644
> > > > > --- a/hw/vfio/pci-quirks.c
> > > > > +++ b/hw/vfio/pci-quirks.c
> > > > > @@ -33,7 +33,7 @@
> > > > >   * execution as noticed with the BCM 57810 card for lack of a
> > > > >   * more better way to handle such issues.
> > > > >   * The  user can still override by specifying a romfile or
> > > > > - * rombar=1.
> > > > > + * rombar=on.
> > > > >   * Please see https://bugs.launchpad.net/qemu/+bug/1284874
> > > > >   * for an analysis of the 57810 card hang. When adding
> > > > >   * a new vendor id/device id combination below, please also add
> > > > 
> > > > 
> > > > So we are apparently breaking a bunch of users who followed
> > > > documentation to the dot. Why is this a good idea?
> > > 
> > > On/off is clearer than 1/0. But isn't 1/0 a synonym for on/off so previous
> > > command lines would still work?
> > > 
> > > Regards,
> > > BALATON Zoltan
> > 
> > I see nothing in code that would make it so:
> > 
> > 
> > const QEnumLookup OnOffAuto_lookup = {
> >    .array = (const char *const[]) {
> >        [ON_OFF_AUTO_AUTO] = "auto",
> >        [ON_OFF_AUTO_ON] = "on",
> >        [ON_OFF_AUTO_OFF] = "off",
> >    },
> >    .size = ON_OFF_AUTO__MAX
> > };
> > 
> > I also tried with an existing property:
> > 
> > $ ./qemu-system-x86_64 -device intel-hda,msi=0
> > qemu-system-x86_64: -device intel-hda,msi=0: Parameter 'msi' does not 
> > accept value '0'
> 
> Then it was probably bit properties that also accept 0/1, on/off,
> true/false.

I mean, the code is open, why do you keep guessing?


No, these reuse the bool parsing logic:

static void prop_get_bit(Object *obj, Visitor *v, const char *name,
                         void *opaque, Error **errp)
{
    Property *prop = opaque;
    uint32_t *p = object_field_prop_ptr(obj, prop);
    bool value = (*p & qdev_get_prop_mask(prop)) != 0;

    visit_type_bool(v, name, &value, errp);
}

and that never accepted 0 or 1:


bool qapi_bool_parse(const char *name, const char *value, bool *obj, Error 
**errp)
{       
    if (g_str_equal(value, "on") ||
        g_str_equal(value, "yes") ||
        g_str_equal(value, "true") ||
        g_str_equal(value, "y")) {
        *obj = true; 
        return true;
    }
    if (g_str_equal(value, "off") ||
        g_str_equal(value, "no") ||
        g_str_equal(value, "false") ||
        g_str_equal(value, "n")) {
        *obj = false;
        return true;
    }
    
    error_setg(errp, QERR_INVALID_PARAMETER_VALUE, name,
               "'on' or 'off'");
    return false;
}



> Maybe similar aliases could be added to on/off/auto?

Could be, but even then switching to that would mean that user sets 1
but query returns "on".  Might or might not surprise some users.

Adding true/false yes/no y/n aliases to on/off/auto might make sense
though, for consistency. Donnu if QAPI guys will agree, though,
and not directly related to this patchset.

One other idea is to add a generic way to detect that a property is set
by user. This requirement comes up, once in a while.




> In any case when I first saw rombar I thought it would set the BAR of the
> ROM so wondered why it's 1 and not 5 or 6 or an offset. So on/off is clearer
> in this case.
> 
> Regards,
> BALATON Zoltan


I agree here, but it's been here for a long time.

-- 
MST




reply via email to

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