Re: [Qemu-devel] i386: pc: align gpa<->hpa on 1GB boundary (v4)

[Marcelo Tosatti]

On Thu, Nov 07, 2013 at 04:24:59PM +0100, Igor Mammedov wrote:
> On Wed, 6 Nov 2013 19:31:19 -0200
> Marcelo Tosatti <address@hidden> wrote:
> 
> > 
> > v2: condition enablement of new mapping to new machine types (Paolo)
> > v3: fix changelog
> > v4: rebase
> > 
> > -----
> > 
> > 
> > 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 <address@hidden>
> > 
> > 
> > diff --git a/hw/i386/pc.c b/hw/i386/pc.c
> > index 12c436e..d29196d 100644
> > --- a/hw/i386/pc.c
> > +++ b/hw/i386/pc.c
> > @@ -1156,7 +1156,7 @@ FWCfgState *pc_memory_init(MemoryRegion 
> > *system_memory,
> >  {
> >      int linux_boot, i;
> >      MemoryRegion *ram, *option_rom_mr;
> > -    MemoryRegion *ram_below_4g, *ram_above_4g;
> > +    MemoryRegion *ram_below_4g, *ram_above_4g, *ram_above_4g_piecetwo;
> >      FWCfgState *fw_cfg;
> >  
> >      linux_boot = (kernel_filename != NULL);
> > @@ -1177,10 +1177,45 @@ FWCfgState *pc_memory_init(MemoryRegion 
> > *system_memory,
> >      e820_add_entry(0, below_4g_mem_size, E820_RAM);
> >      if (above_4g_mem_size > 0) {
> >          ram_above_4g = g_malloc(sizeof(*ram_above_4g));
> > -        memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
> > -                                 below_4g_mem_size, above_4g_mem_size);
> > -        memory_region_add_subregion(system_memory, 0x100000000ULL,
> > +        /*
> > +         *
> > +         * 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.
> > +         *
> > +         */
> I did some corner cases testing and there is alias overlapping in case 
>   -m 4096 -mem-path /var/lib/hugetlbfs/global/pagesize-1GB
> 
>   0000000100000000-000000011fffffff (prio 0, RW): alias ram-above-4g-piecetwo 
> @pc.ram 00000000e0000000-00000000ffffffff
>   0000000100000000-0000000100000000 (prio 0, RW): alias ram-above-4g @pc.ram 
> 0000000100000000-0000000100000000
> 
> perhaps zero sized ram-above-4g shouldn't be created at all?

Right.

> in addition ram-above-4g-piecetwo starts at half page offset 00000000e0000000 
> but guest sees it 4Gb offset,
> wouldn't it cause the same issue patch tries to solve but only for 
> ram-above-4g-piecetwo tail sync host/guest offsets
> are not 1Gb aligned?

Piece 1 is aligned. Piece 2 maps from tail of RAM (gpa) to start of hole
(ramblock).

> there is more misalignment with:
>  -m 4097 -mem-path /var/lib/hugetlbfs/global/pagesize-1GB
> 
>   0000000100000000-00000001000fffff (prio 0, RW): alias ram-above-4g @pc.ram 
> 0000000100000000-00000001000fffff

Piece 1 is aligned.

>   0000000100100000-00000001200fffff (prio 0, RW): alias ram-above-4g-piecetwo 
> @pc.ram 00000000e0000000-00000000ffffffff

Piece 2 is not. Should allocate one extra MB of RAM to align that. I'll
resend, thanks.

> where ram-above-4g-piecetwo is aligned with 1Gb+1Mb GPA offset, in addition 
> to 500Mb offset of HPA.
> which would cause the same issue as above prehaps?