Re: [Minios-devel] [PATCH 03/40] arm64: add the boot code

[Julien Grall <julien.grall@xxxxxxxxxx>]

Hi Shijie,

On 08/11/17 06:17, Huang Shijie wrote:
> On Tue, Nov 07, 2017 at 01:39:44PM +0000, Julien Grall wrote:
> > Hi Shijie,
> >
> > More question about the boot code...
> >
> > On 03/11/17 03:11, Huang Shijie wrote:
> > > +ENTRY(_start)
> > > +    /* Save the FDT pointer */
> > > +    mov     x20, x0
> > > +
> > > +    /* Calculate where we are */
> > > +    bl      _calc_offset
> > > +
> > > +    PRINT("- Mini-OS booting -\n")
> > > +
> > > +    PRINT("- Setup CPU -\n")
> > > +    /* Setup CPU for turning on the MMU. */
> > > +    bl      _setup_cpu
> > > +
> > > +    PRINT("- Setup booting pagetable -\n")
> > > +    /* Setup the initial page table. */
> > > +    bl      _setup_initial_pgtable
> > > +
> > > +    /* Setup the identity mapping */
> > > +    bl      _setup_idmap_pgtable
> > > +
> > > +    /* Load TTBRx */
> > > +    msr     ttbr1_el1, x4
> > > +    msr     ttbr0_el1, x5
> > First at all, this is not entirely obvious where x4 and x5 are been set.
> > Proper documentation would be needed for that.
> Okay, I will add more document for it.
>
> > Now, IIUC the code TTBR0_EL1 will contain the identity mappings. I was
> > expecting to see it removed as soon as MMU is enabled. But this does not
> Why should we remove it after the MMU is enabled?
>
> Is there any problem if we keep it?
Well, after the MMU is enabled the OS should never address the identity 
mapping. If there are a programming error, that region may be accessed 
by errror. So instead of getting a fault you would corrupt memory. The 
latter is really hard to debug.
Cheers,

--
Julien Grall

_______________________________________________
Minios-devel mailing list
Minios-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/cgi-bin/mailman/listinfo/minios-devel