Re: [Xen-devel] Can't see more than 3.5GB of RAM / UEFI / no e820 memory map detected

[Jonathan Tripathy <jonnyt@xxxxxxxxxxx>]

On 28/08/2012 22:50, Jonathan Tripathy wrote:
> On 28/08/2012 22:41, Keir Fraser wrote:
> > On 28/08/2012 22:13, "Jonathan Tripathy" <jonnyt@xxxxxxxxxxx> wrote:
> >
> > > Keir Fraser wrote:
> > > > On 28/08/2012 20:36, "Jonathan Tripathy" <jonnyt@xxxxxxxxxxx> wrote:
> > > >
> > > > > > > Thanks for the clarification.
> > > > > > >
> > > > > > > So from a security/reliability standpoint, nothing will be 
> > > > > > > affected by
> > > > > > > flipping the if block?
> > > > > > It should simply make it more likely that Xen sees all your RAM. ;)
> > > > > >
> > > > > >     -- Keir
> > > > > Hi Everyone,
> > > > >
> > > > > I reversed the if block in setup.c and now my server can see the full
> > > > > 32GB of RAM. I haven't submitted a patch yet as we have run into 
> > > > > another
> > > > > (possibly unrelated to xen) issue with this server build that we are
> > > > > working on. Once we complete our full testing, a patch will be 
> > > > > submitted :)
> > > > In this case, I will re-make the patch myself and check it in. 
> > > > Since it is a
> > > > trivial one.
> > > >
> > > >     Thanks,
> > > >    Keir
> > > Thanks Keir
> > Now done. xen-unstable:25786
> >
> >   -- Keir
> Hi Keir,
>
> Thanks for doing that. However, the ordering of the if block in my 
> code that I used was slightly different from your commit. Perhaps this 
> doesn't make too much of a difference?
> Thanks
Ooops, typo'd the code (I had to recreate the changes as I deleted my 
source tree). Here is the version I used:
if ( mbi->flags & MBI_MEMMAP )
    {
        memmap_type = "Multiboot-e820";
        while ( (bytes < mbi->mmap_length) && (e820_raw_nr < E820MAX) )
        {
            memory_map_t *map = __va(mbi->mmap_addr + bytes);

            /*
             * This is a gross workaround for a BIOS bug. Some 
bootloaders do
             * not write e820 map entries into pre-zeroed memory. This is
             * okay if the BIOS fills in all fields of the map entry, but
             * some broken BIOSes do not bother to write the high word of
             * the length field if the length is smaller than 4GB. We
             * detect and fix this by flagging sections below 4GB that
             * appear to be larger than 4GB in size.
             */
            if ( (map->base_addr_high == 0) && (map->length_high != 0) )
            {
                if ( !e820_warn )
                {
                    printk("WARNING: Buggy e820 map detected and fixed "
                           "(truncated length fields).\n");
                    e820_warn = 1;
                }
                map->length_high = 0;
            }

            e820_raw[e820_raw_nr].addr =
                ((u64)map->base_addr_high << 32) | 
(u64)map->base_addr_low;
            e820_raw[e820_raw_nr].size =
                ((u64)map->length_high << 32) | (u64)map->length_low;
            e820_raw[e820_raw_nr].type = map->type;
            e820_raw_nr++;

            bytes += map->size + 4;
        }
    }
    else if ( mbi->flags & MBI_MEMLIMITS )
    {
        memmap_type = "Multiboot-e801";
        e820_raw[0].addr = 0;
        e820_raw[0].size = mbi->mem_lower << 10;
        e820_raw[0].type = E820_RAM;
        e820_raw[1].addr = 0x100000;
        e820_raw[1].size = mbi->mem_upper << 10;
        e820_raw[1].type = E820_RAM;
        e820_raw_nr = 2;
    }
    else if ( e820_raw_nr != 0 )
    {
        memmap_type = "Xen-e820";
    }
    else if ( bootsym(lowmem_kb) )
    {
        memmap_type = "Xen-e801";
        e820_raw[0].addr = 0;
        e820_raw[0].size = bootsym(lowmem_kb) << 10;
        e820_raw[0].type = E820_RAM;
        e820_raw[1].addr = 0x100000;
        e820_raw[1].size = bootsym(highmem_kb) << 10;
        e820_raw[1].type = E820_RAM;
        e820_raw_nr = 2;
    }
    else
    {
        EARLY_FAIL("Bootloader provided no memory information.\n");
    }


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel