Re: [Xen-devel] [PATCH v2] iommu / p2m: add a page_order parameter to iommu_map/unmap_page()

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> Sent: 30 October 2018 17:05
> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> Cc: Julien Grall <julien.grall@xxxxxxx>; Andrew Cooper
> <Andrew.Cooper3@xxxxxxxxxx>; George Dunlap <George.Dunlap@xxxxxxxxxx>; Ian
> Jackson <Ian.Jackson@xxxxxxxxxx>; Wei Liu <wei.liu2@xxxxxxxxxx>; Jun
> Nakajima <jun.nakajima@xxxxxxxxx>; Kevin Tian <kevin.tian@xxxxxxxxx>;
> Stefano Stabellini <sstabellini@xxxxxxxxxx>; xen-devel <xen-
> devel@xxxxxxxxxxxxxxxxxxxx>; Konrad Rzeszutek Wilk
> <konrad.wilk@xxxxxxxxxx>; Tim (Xen.org) <tim@xxxxxxx>
> Subject: RE: [PATCH v2] iommu / p2m: add a page_order parameter to
> iommu_map/unmap_page()
> 
> >>> On 30.10.18 at 17:56, <Paul.Durrant@xxxxxxxxxx> wrote:
> >> From: Jan Beulich [mailto:JBeulich@xxxxxxxx]
> >> Sent: 30 October 2018 16:08
> >>
> >> >>> On 29.10.18 at 14:29, <paul.durrant@xxxxxxxxxx> wrote:
> >> > --- a/xen/common/grant_table.c
> >> > +++ b/xen/common/grant_table.c
> >> > @@ -1142,12 +1142,14 @@ map_grant_ref(
> >> >          {
> >> >              if ( !(kind & MAPKIND_WRITE) )
> >> >                  err = iommu_map_page(ld, _dfn(mfn_x(mfn)), mfn,
> >> > +                                     PAGE_ORDER_4K,
> >> >                                       IOMMUF_readable |
> >> IOMMUF_writable);
> >> >          }
> >> >          else if ( act_pin && !old_pin )
> >> >          {
> >> >              if ( !kind )
> >> >                  err = iommu_map_page(ld, _dfn(mfn_x(mfn)), mfn,
> >> > +                                     PAGE_ORDER_4K,
> >> >                                       IOMMUF_readable);
> >> >          }
> >> >          if ( err )
> >> > @@ -1396,10 +1398,11 @@ unmap_common(
> >> >
> >> >          kind = mapkind(lgt, rd, op->mfn);
> >> >          if ( !kind )
> >> > -            err = iommu_unmap_page(ld, _dfn(mfn_x(op->mfn)));
> >> > +            err = iommu_unmap_page(ld, _dfn(mfn_x(op->mfn)),
> >> > +                                   PAGE_ORDER_4K);
> >> >          else if ( !(kind & MAPKIND_WRITE) )
> >> >              err = iommu_map_page(ld, _dfn(mfn_x(op->mfn)), op->mfn,
> >> > -                                 IOMMUF_readable);
> >> > +                                 PAGE_ORDER_4K, IOMMUF_readable);
> >> >
> >> >          double_gt_unlock(lgt, rgt);
> >>
> >> I am, btw, uncertain that using PAGE_ORDER_4K is correct here:
> >> Other than in the IOMMU code, grant table code isn't tied to a
> >> particular architecture, and hence ought to work fine on a port
> >> to an architecture with 8k, 16k, or 32k pages.
> >
> > Would you suggest I add an arch specific #define for a grant table page
> > order and then use that?
> 
> No, I'd prefer if you used liter 0 zero here.
> 

Ok.

> >> > --- a/xen/drivers/passthrough/iommu.c
> >> > +++ b/xen/drivers/passthrough/iommu.c
> >> > @@ -305,47 +305,76 @@ void iommu_domain_destroy(struct domain *d)
> >> >  }
> >> >
> >> >  int iommu_map_page(struct domain *d, dfn_t dfn, mfn_t mfn,
> >> > -                   unsigned int flags)
> >> > +                   unsigned int page_order, unsigned int flags)
> >> >  {
> >> >      const struct domain_iommu *hd = dom_iommu(d);
> >> > -    int rc;
> >> > +    unsigned long i;
> >> >
> >> >      if ( !iommu_enabled || !hd->platform_ops )
> >> >          return 0;
> >> >
> >> > -    rc = hd->platform_ops->map_page(d, dfn, mfn, flags);
> >> > -    if ( unlikely(rc) )
> >> > +    ASSERT(!(dfn_x(dfn) & ((1ul << page_order) - 1)));
> >> > +    ASSERT(!(mfn_x(mfn) & ((1ul << page_order) - 1)));
> >> > +
> >> > +    for ( i = 0; i < (1ul << page_order); i++ )
> >> >      {
> >> > +        int ignored, err = hd->platform_ops->map_page(d,
> dfn_add(dfn,
> >> i),
> >> > +                                                      mfn_add(mfn,
> i),
> >> > +                                                      flags);
> >> > +
> >> > +        if ( likely(!err) )
> >> > +            continue;
> >> > +
> >> >          if ( !d->is_shutting_down && printk_ratelimit() )
> >> >              printk(XENLOG_ERR
> >> >                     "d%d: IOMMU mapping dfn %"PRI_dfn" to mfn
> %"PRI_mfn"
> >> failed: %d\n",
> >> > -                   d->domain_id, dfn_x(dfn), mfn_x(mfn), rc);
> >> > +                   d->domain_id, dfn_x(dfn_add(dfn, i)),
> >> > +                   mfn_x(mfn_add(mfn, i)), err);
> >> > +
> >> > +        while (i--)
> >> > +            /* assign to something to avoid compiler warning */
> >> > +            ignored = hd->platform_ops->unmap_page(d, dfn_add(dfn,
> i));
> >>
> >> Hmm, as said on v1 - please use the original mode (while-if-continue)
> >> here. This lets you get away without a local variable that's never
> >> read, and which hence future compiler versions may legitimately warn
> >> about.
> >>
> >
> > Ok, I clearly don't understand what you mean by 'while-if-continue'
> then.
> > Above I have for-if-continue, which is what I thought you wanted. What
> code
> > structure are you actually looking for?
> 
> The one your patch removes elsewhere:
> 
> -                        while ( i-- )
> -                            /* If statement to satisfy __must_check. */
> -                            if ( iommu_unmap_page(p2m->domain,
> -                                                  dfn_add(dfn, i)) )
> -                                continue;

Oh, ok. Thanks. I'll change it.

  Paul

> 
> Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel