[XenARM] Moving Guest RAM address to other location than 0x80000000

[XenARM] Moving Guest RAM address to other location than 0x80000000

[Eric Trudeau]

[-- Attachment #1.1: Type: text/plain, Size: 380 bytes --]

Can I move the Guest RAM address to something other than 0x80000000?
I would like to move the Guest RAM to 0 if possible.

I get a crash while creating the domain when I just change the following in tools/libxc/xc_dom_armzimageloader.c:

#define GUEST_RAM_BASE 0x00000000 /* was 0x80000000 */

Thanks,

Eric Trudeau
200 Brickstone Square Suite 401
Andover, MA 01810


[-- Attachment #1.2: Type: text/html, Size: 2414 bytes --]

[-- Attachment #2: Type: text/plain, Size: 126 bytes --]

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

Re: [XenARM] Moving Guest RAM address to other location than 0x80000000

[Julien Grall]

On 10/09/2013 04:57 PM, Eric Trudeau wrote:

Hi,

> Can I move the Guest RAM address to something other than 0x80000000?
> I would like to move the Guest RAM to 0 if possible.
> 
>  
> 
> I get a crash while creating the domain when I just change the following
> in tools/libxc/xc_dom_armzimageloader.c:


> #define GUEST_RAM_BASE 0x00000000 /* was 0x80000000 */

Did you modify the guest DTS? If not, you need to fix the "reg" property
in memory node.

Cheers,

-- 
Julien Grall

Re: [XenARM] Moving Guest RAM address to other location than 0x80000000

[Eric Trudeau]

> -----Original Message-----
> From: Julien Grall [mailto:julien.grall@linaro.org]
> Sent: Wednesday, October 09, 2013 12:20 PM
> To: Eric Trudeau
> Cc: xen-devel (xen-devel@lists.xen.org)
> Subject: Re: [Xen-devel] [XenARM] Moving Guest RAM address to other location
> than 0x80000000
> 
> On 10/09/2013 04:57 PM, Eric Trudeau wrote:
> 
> Hi,
> 
> > Can I move the Guest RAM address to something other than 0x80000000?
> > I would like to move the Guest RAM to 0 if possible.
> >
> >
> >
> > I get a crash while creating the domain when I just change the following
> > in tools/libxc/xc_dom_armzimageloader.c:
> 
> 
> > #define GUEST_RAM_BASE 0x00000000 /* was 0x80000000 */
> 
> Did you modify the guest DTS? If not, you need to fix the "reg" property
> in memory node.
> 
Yes, I have a memory region of 0 in the DT.

The crash is very early.  Could there be another code dependency on RAM starting at 0x80000000 (e.g. guest mapping)?
BTW:   xl destroy dom1 hangs when trying to destroy this guest.

-------------------------------------------------------------------------------
dom0 # xl create -c dom1.cfg
Parsing config from dom1.cfg
(XEN) physdev.c:57: dom1: gic_route_irq_to_guest mapped in IRQ 108
(XEN) memory_map:add: dom1 gfn=0 mfn=0 nr=80000
(XEN) memory_map:add: dom1 gfn=a0000 mfn=a0000 nr=20000
(XEN) memory_map:add: dom1 gfn=f0000 mfn=f0000 nr=1000
device vif1.0 entered promiscuous mode
(XEN) Hypervisor Trap. HSR=0x80000006 EC=0x20 IL=0 Syndrome=6
Daemon running with PID 681(XEN) CPU3: Unexpected Trap: Hypervisor

(XEN) ----[ Xen-4.4-unstable  arm32  debug=y  Tainted:    C ]----
(XEN) CPU:    3
(XEN) PC:     ff7f4014
(XEN) CPSR:   400001d7 MODE:32-bit Guest ABT
(XEN)      R0: 00000000 R1: 41007a2f R2: 00000067 R3: 00000000
(XEN)      R4: 41007a18 R5: 00000000 R6: 00000000 R7: 00000000
(XEN)      R8: 00000000 R9: 00000000 R10:00000000 R11:00000000 R12:00000000
(XEN) USR: SP: 00000000 LR: 00000000
(XEN) SVC: SP: 00000058 LR: 40008008 SPSR:00000000
(XEN) ABT: SP: 00000000 LR: 00000014 SPSR:400001f3
(XEN) UND: SP: 00000000 LR: 00000000 SPSR:00000000
(XEN) IRQ: SP: 00000000 LR: 00000000 SPSR:00000000
(XEN) FIQ: SP: 00000000 LR: c2c2c2c2 SPSR:00000000
(XEN) FIQ: R8: 00000000 R9: 00000000 R10:00000000 R11:00000000 R12:00000000
(XEN) 
(XEN)      SCTLR: 00c50078
(XEN)        TCR: 00000000
(XEN)      TTBR0: 0000000000000000
(XEN)      TTBR1: 0000000000000000
(XEN)       IFAR: 00000000, IFSR: 00000000
(XEN)       DFAR: 00000067, DFSR: 00000001
(XEN) 
(XEN)   VTCR_EL2: 80002558
(XEN)  VTTBR_EL2: 000200009fe6e000
(XEN) 
(XEN)  SCTLR_EL2: 30cd187f
(XEN)    HCR_EL2: 0000000000282835
(XEN)  TTBR0_EL2: 0000000096039000
(XEN) 
(XEN)    ESR_EL2: 80000006
(XEN)  HPFAR_EL2: 0000000000ff7f40
(XEN)      HDFAR: c8800f00
(XEN)      HIFAR: ff7f4014
(XEN) 
(XEN) No stack trace for 32-bit guest kernel-mode
xenbr0: port 1(eth0) entered forwarding state
dom0 # 
dom0 # 
dom0 # xl list
Name                                        ID   Mem VCPUs      State   Time(s)
Domain-0                                     0   122     4     r-----       5.0
dom1                                         1    96     1     r-----      16.9
dom0 # xl destroy dom1


> Cheers,
> 
> --
> Julien Grall

Re: [XenARM] Moving Guest RAM address to other location than 0x80000000

[Julien Grall]

On 10/09/2013 05:34 PM, Eric Trudeau wrote:
>> -----Original Message-----
>> From: Julien Grall [mailto:julien.grall@linaro.org]
>> Sent: Wednesday, October 09, 2013 12:20 PM
>> To: Eric Trudeau
>> Cc: xen-devel (xen-devel@lists.xen.org)
>> Subject: Re: [Xen-devel] [XenARM] Moving Guest RAM address to other location
>> than 0x80000000
>>
>> On 10/09/2013 04:57 PM, Eric Trudeau wrote:
>>
>> Hi,
>>
>>> Can I move the Guest RAM address to something other than 0x80000000?
>>> I would like to move the Guest RAM to 0 if possible.
>>>
>>>
>>>
>>> I get a crash while creating the domain when I just change the following
>>> in tools/libxc/xc_dom_armzimageloader.c:
>>
>>
>>> #define GUEST_RAM_BASE 0x00000000 /* was 0x80000000 */
>>
>> Did you modify the guest DTS? If not, you need to fix the "reg" property
>> in memory node.
>>
> Yes, I have a memory region of 0 in the DT.
> 
> The crash is very early.  Could there be another code dependency on RAM starting at 0x80000000 (e.g. guest mapping)?

I don't see other dependencies in the code. Did you check that the jump
address is correct (dom->parms.virt_entry)?

> BTW:   xl destroy dom1 hangs when trying to destroy this guest.

That's normal, one of the CPU has received an non-handled hypervisor
trap that is result to an infinite loop in hyp mode for this CPU.
After few minutes you should see strange warning from the dom0 kernel.

I think we should correctly handle EC=0x20 (ie prefetch abort).

> 
> -------------------------------------------------------------------------------
> dom0 # xl create -c dom1.cfg

Can you paste the log of the following command?
xl -vvv create -c dom1.cfg

> Parsing config from dom1.cfg
> (XEN) physdev.c:57: dom1: gic_route_irq_to_guest mapped in IRQ 108
> (XEN) memory_map:add: dom1 gfn=0 mfn=0 nr=80000
> (XEN) memory_map:add: dom1 gfn=a0000 mfn=a0000 nr=20000
> (XEN) memory_map:add: dom1 gfn=f0000 mfn=f0000 nr=1000
> device vif1.0 entered promiscuous mode
> (XEN) Hypervisor Trap. HSR=0x80000006 EC=0x20 IL=0 Syndrome=6

It seems Xen received a prefetch abort, which means that the kernel is
trying to fetch an instruction in a non mapped address.

> Daemon running with PID 681(XEN) CPU3: Unexpected Trap: Hypervisor
> 
> (XEN) ----[ Xen-4.4-unstable  arm32  debug=y  Tainted:    C ]----
> (XEN) CPU:    3
> (XEN) PC:     ff7f4014

The PC address looks weird. Silly question, Did you try to boot the
guest with the current GUEST_RAM_BASE (ie 0x80000000)?

-- 
Julien Grall