Re: [Qemu-devel] i386: pc: align gpa<->hpa on 1GB boundary (v3)

[Igor Mammedov <imammedo@redhat.com>]

On Wed, 06 Nov 2013 13:07:26 +0100
Paolo Bonzini <pbonzini@redhat.com> wrote:

> Il 06/11/2013 12:59, Igor Mammedov ha scritto:
> > On Tue, 5 Nov 2013 23:55:43 -0200
> > Marcelo Tosatti <mtosatti@redhat.com> wrote:
> > 
> >>
> >>
> >> v2: condition enablement of new mapping to new machine types (Paolo)
> >> v3: fix changelog
> >>
> >> -----
> >>
> >>
> >> Align guest physical address and host physical address
> >> beyond guest 4GB on a 1GB boundary.
> >>
> >> Otherwise 1GB TLBs cannot be cached for the range.
> >>
> >> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> >>
> >> diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> >> index 0c313fe..534e067 100644
> >> --- a/hw/i386/pc.c
> >> +++ b/hw/i386/pc.c
> >> @@ -1116,7 +1116,7 @@ FWCfgState *pc_memory_init(MemoryRegion *system_memory,
> > [...]
> >> +        /*
> >> +         *
> >> +         * If 1GB hugepages are used to back guest RAM, map guest address
> >> +         * space in the range [ramsize,ramsize+holesize] to the ram block
> >> +         * range [holestart, 4GB]
> >> +         *
> >> +         *                      0      h     4G     [ramsize,ramsize+holesize]
> >> +         *
> >> +         * guest-addr-space     [      ]     [      ][xxx]
> >> +         *                                /----------/
> >> +         * contiguous-ram-block [      ][xxx][     ]
> >> +         *
> >> +         * So that memory beyond 4GB is aligned on a 1GB boundary,
> >> +         * at the host physical address space.
> >> +         *
> >> +         */
> >> +        if (guest_info->gb_align) {
> > 'gb_align' is one shot usage, it would be better to just add it as an argument
> > to pc_memory_init(). That would allow to avoid extending PcGuestInfo needlessly,
> > since gb_align isn't reused.
> 
> No, Marcelo's way is better.  pc_memory_init already has too many
> arguments, moving them to PcGuestInfo (which ultimately might become
> properties of the /machine QOM object) is the right thing to do.
In general I agree. But unless there is plans to reuse gb_align in future,
it doesn't really belong to PcGuestInfo (this change however looks like
cannibalizing structure for argument passing only).

> 
> Paolo
> 
> >> +            unsigned long holesize = 0x100000000ULL - below_4g_mem_size;
> >> +
> >> +            memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
> >> +                                    0x100000000ULL,
> >> +                                    above_4g_mem_size - holesize);
> >> +            memory_region_add_subregion(system_memory, 0x100000000ULL,
> >>                                      ram_above_4g);
> > [...]
> >> diff --git a/include/hw/i386/pc.h b/include/hw/i386/pc.h
> >> index 6083839..00afe4a 100644
> >> --- a/include/hw/i386/pc.h
> >> +++ b/include/hw/i386/pc.h
> >> @@ -20,6 +20,7 @@ typedef struct PcPciInfo {
> >>  struct PcGuestInfo {
> >>      bool has_pci_info;
> >>      bool isapc_ram_fw;
> >> +    bool gb_align;
> >>      FWCfgState *fw_cfg;
> >>  };
> > it doesn't apply anymore.
> > 
> 
>