Mini-OS: new_pt_frame fails

Mini-OS: new_pt_frame fails

[Julian Stecklina]

Hello,

as some might have read in my previous postings, I have a Mini-OS dom0
that constructs a Mini-OS domU on x86 xen-3.0.4-testing. As the memory
layout my Dom0 builds does not guarantee, that pages the DomU Mini-OS 
tries to
use as page tables are mapped to existing memory, I changed new_pt_frame
in arch/x86/mm.c in the following way:

The pt_pfn is mapped R/W, then it is zeroed by a memset(). After that it
is mapped with the correct protection flags and pinned. The problem is
that the first mapping fails:

Mapping memory range 0xc0400000 - 0xc1000000
MINI_OS(file=mm.c, line=68) Allocating new L1 pt frame for pt_pfn=36,
mfn=ca8, prev_l_mfn=cad, offset=301
(XEN) .../xen-3.0.4-testing.hg/xen/include/asm/mm.h:184:d1 Error pfn 0:
rd=ffbe8100, od=ffbf4100, caf=80000002, taf=e8000002
(XEN) mm.c:456:d1 Could not get page ref for pfn 0
(XEN) mm.c:2225:d1 Could not get page for normal update
New PT could not be mapped R/W.

Now I am puzzled what these hypervisor errors mean, and what I could 
have done wrong? The code in question is the following:

     mmu_updates[0].ptr = ((pgentry_t)tab[l2_table_offset(pt_page)] &
PAGE_MASK) +
                          sizeof(pgentry_t) * l1_table_offset(pt_page);
     mmu_updates[0].val = ((pgentry_t)pfn_to_mfn(*pt_pfn) << PAGE_SHIFT) 
| _PAGE_RW | _PAGE_PRESENT;
     if(HYPERVISOR_mmu_update(mmu_updates, 1, NULL, DOMID_SELF) < 0)
     {
          printk("New PT could not be mapped R/W.\n");
          do_exit();
     }


Any help is appreciated.

Regards,
Julian

Re: Mini-OS: new_pt_frame fails

[Keir Fraser]

On 8/3/07 14:32, "Julian Stecklina" <der_julian@web.de> wrote:

> Mapping memory range 0xc0400000 - 0xc1000000
> MINI_OS(file=mm.c, line=68) Allocating new L1 pt frame for pt_pfn=36,
> mfn=ca8, prev_l_mfn=cad, offset=301
> (XEN) .../xen-3.0.4-testing.hg/xen/include/asm/mm.h:184:d1 Error pfn 0:
> rd=ffbe8100, od=ffbf4100, caf=80000002, taf=e8000002
> (XEN) mm.c:456:d1 Could not get page ref for pfn 0
> (XEN) mm.c:2225:d1 Could not get page for normal update
> New PT could not be mapped R/W.

'pfn 0' (actually it's mfn 0) looks suspicious. Like you passed the wrong
ptr value to the mmu_update.

 -- Keir

Re: Mini-OS: new_pt_frame fails

[Jacob Gorm Hansen]

On Thu, 2007-03-08 at 15:32 +0100, Julian Stecklina wrote:
> Hello,
> 
> as some might have read in my previous postings, I have a Mini-OS dom0
> that constructs a Mini-OS domU on x86 xen-3.0.4-testing. As the memory
> layout my Dom0 builds does not guarantee, that pages the DomU Mini-OS 
> tries to
> use as page tables are mapped to existing memory, I changed new_pt_frame
> in arch/x86/mm.c in the following way:

Cool, on a related note I have recently written at blkfront for mini-os,
but not had time to cook up a proper patch for submission. This means
that my 'cstrap' bootloader now works from disk, as well as over HTTP.
If anyone would like to play with my minios blkfront please let me know.

The blockfront can also be found in my xen-3.0.3-based hg tree at
http://www.distlab.dk/hg/index.cgi/xen-evilman.hg

Regards,
Jacob

Re: Mini-OS: new_pt_frame fails

[Grzegorz Milos]

I'm interested in blkfront. I'll port it to unstable and clean up if 
necessary.

Thanks
Gregor

Jacob Gorm Hansen wrote:
> On Thu, 2007-03-08 at 15:32 +0100, Julian Stecklina wrote:
>> Hello,
>>
>> as some might have read in my previous postings, I have a Mini-OS dom0
>> that constructs a Mini-OS domU on x86 xen-3.0.4-testing. As the memory
>> layout my Dom0 builds does not guarantee, that pages the DomU Mini-OS 
>> tries to
>> use as page tables are mapped to existing memory, I changed new_pt_frame
>> in arch/x86/mm.c in the following way:
> 
> Cool, on a related note I have recently written at blkfront for mini-os,
> but not had time to cook up a proper patch for submission. This means
> that my 'cstrap' bootloader now works from disk, as well as over HTTP.
> If anyone would like to play with my minios blkfront please let me know.
> 
> The blockfront can also be found in my xen-3.0.3-based hg tree at
> http://www.distlab.dk/hg/index.cgi/xen-evilman.hg
> 
> Regards,
> Jacob
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel