headers
Kamal R. Prasad | 1 Jan 2005 09:40
Picon
Favicon

Which kernel to use for Triton chipset

Hello,

 I have cross-compiled netbsd with a 
#./build.sh -m evbarm release

The process built many kernels -probably one for each
cpu type and an integrator-install kernel. The board I
have is an Intel Xscale board with a Triton chipset.
Can someone tell me which kernel to load on it?
The message from bootup is pasted below:-
--------------------------------------
RedBoot(tm) bootstrap and debug environment [ROM]
Non-certified release, version 2.3 - built 13:38:55,
Jun 11 2004

Platform: TRITON ECO --STK1--  (XScale) developed and
produced by Ka-Ro electroy
Copyright (C) 2000, 2001, 2002, Red Hat, Inc.

RAM: 0xa0000000-0xa1000000, 0xa0013f68-0xa0edd000
available
FLASH: 0x00000000 - 0x01000000, 128 blocks of
0x00020000 bytes each.
--------------------------------------

thanks
-kamal
p.s:- not subscribed to either mailing lists
(Continue reading)

Permalink | Reply | 0 comments |
headers
Kamal R. Prasad | 1 Jan 2005 20:06
Picon
Favicon

Re:-Which kernel to use for Triton chipset

Hello,

The answer as I found out is Lubbock. The chip used is
PXA255. Im not yet out of the woods with the
bootstrap.  but can someone tell me how to print
diagnostic messages from within  assembly code
(machdep_lubbock.S)?

thanks
-kamal

=====================================
Subject: Which kernel to use for Triton chipset
To: None <tech-kern <at> netbsd.org>
From: Kamal R. Prasad <kamalpr <at> yahoo.com>
List: tech-kern
Date: 01/01/2005 00:40:41

Hello,

 I have cross-compiled netbsd with a 
#./build.sh -m evbarm release

The process built many kernels -probably one for each
cpu type and an integrator-install kernel. The board I
have is an Intel Xscale board with a Triton chipset.
Can someone tell me which kernel to load on it?
The message from bootup is pasted below:-
--------------------------------------
RedBoot(tm) bootstrap and debug environment [ROM]
(Continue reading)

Permalink | Reply | 0 comments |
headers
Kamal R. Prasad | 2 Jan 2005 06:47
Picon
Favicon

pmap initlialization

Hello,

 I have managed to bootstrap netbsd on my Intel pxa255
board -with a machdep glue. But there is a problem
with pmap initialization. If someone can give me a
clue, it might help. Pasted below is the output on
bootup.

thanks
-kamal
---------------------------------------
NetBSD/evbarm (triton) booting ...
initarm: Configuring system ...
physmemory: 4096 pages at 0xa0000000 -> 0xa0ffffff
Allocating page tables
freestart = 0xa0009000, free_pages = 503 (0x000001f7)
IRQ stack: p0xa01f1000 v0xc01f1000
ABT stack: p0xa01f0000 v0xc01f0000
UND stack: p0xa01ef000 v0xc01ef000
SVC stack: p0xa01ed000 v0xc01ed000
Creating L1 page table at 0xa01fc000
Mapping kernel
pmap_map_chunk: pa=0xa0200000 va=0xc0200000
size=0x10d000 resid=0x10d000 prot=01
SPPPPPPPPPPPPP
pmap_map_chunk: pa=0xa030d000 va=0xc030d000
size=0x3d000 resid=0x3d000 prot=0x31
PPPLLLPPPPPPPPPP
Constructing L2 page tables
pmap_map_chunk: pa=0xa01f1000 va=0xc01f1000
(Continue reading)

Permalink | Reply | 0 comments |
headers
Kamal R. Prasad | 4 Jan 2005 09:21
Picon
Favicon

Re:-pmap initlialization

Hello,

 I found out that as soon as tlb is invalidated, the
system goes for a toss. Does the configuration below
look OK? The kernel is loaded at 0xa0200000 and the L1
page table is at 0xa01fc000.
Other than an incorrect base addr for the page table,
what else could cause the system to go for a toss? Why
wouldn't the system generate an exception in such an
event? Appreciate any answers.

thanks
-kamal

Subject: pmap initlialization
To: None <tech-kern <at> netbsd.org, port-arm <at> netbsd.org>
From: Kamal R. Prasad <kamalpr <at> yahoo.com>
List: tech-kern
Date: 01/01/2005 21:47:59

Hello,

 I have managed to bootstrap netbsd on my Intel pxa255
board -with a machdep glue. But there is a problem
with pmap initialization. If someone can give me a
clue, it might help. Pasted below is the output on
bootup.

thanks
-kamal
(Continue reading)

Permalink | Reply | 544 comments |
headers
Kamal R. Prasad | 4 Jan 2005 09:21
Picon
Favicon

Re:-pmap initlialization

Hello,

 I found out that as soon as tlb is invalidated, the
system goes for a toss. Does the configuration below
look OK? The kernel is loaded at 0xa0200000 and the L1
page table is at 0xa01fc000.
Other than an incorrect base addr for the page table,
what else could cause the system to go for a toss? Why
wouldn't the system generate an exception in such an
event? Appreciate any answers.

thanks
-kamal

Subject: pmap initlialization
To: None <tech-kern <at> netbsd.org, port-arm <at> netbsd.org>
From: Kamal R. Prasad <kamalpr <at> yahoo.com>
List: tech-kern
Date: 01/01/2005 21:47:59

Hello,

 I have managed to bootstrap netbsd on my Intel pxa255
board -with a machdep glue. But there is a problem
with pmap initialization. If someone can give me a
clue, it might help. Pasted below is the output on
bootup.

thanks
-kamal
(Continue reading)

Permalink | Reply | 0 comments |
headers
Kamal R. Prasad | 4 Jan 2005 09:18
Picon
Favicon

Re:-pmap initlialization

Hello,

 I found out that as soon as tlb is invalidated, the
system goes for a toss. Does the configuration below
look OK? The kernel is loaded at 0xa0200000 and the L1
page table is at 0xa01fc000.
Other than an incorrect base addr for the page table,
what else could cause the system to go for a toss? Why
wouldn't the system generate an exception in such an
event? Appreciate any answers.

thanks
-kamal

Subject: pmap initlialization
To: None <tech-kern <at> netbsd.org, port-arm <at> netbsd.org>
From: Kamal R. Prasad <kamalpr <at> yahoo.com>
List: tech-kern
Date: 01/01/2005 21:47:59

Hello,

 I have managed to bootstrap netbsd on my Intel pxa255
board -with a machdep glue. But there is a problem
with pmap initialization. If someone can give me a
clue, it might help. Pasted below is the output on
bootup.

thanks
-kamal
(Continue reading)

Permalink | Reply | 0 comments |
headers
Kamal R. Prasad | 4 Jan 2005 09:19
Picon
Favicon

Re:-pmap initlialization

Hello,

 I found out that as soon as tlb is invalidated, the
system goes for a toss. Does the configuration below
look OK? The kernel is loaded at 0xa0200000 and the L1
page table is at 0xa01fc000.
Other than an incorrect base addr for the page table,
what else could cause the system to go for a toss? Why
wouldn't the system generate an exception in such an
event? Appreciate any answers.

thanks
-kamal

Subject: pmap initlialization
To: None <tech-kern <at> netbsd.org, port-arm <at> netbsd.org>
From: Kamal R. Prasad <kamalpr <at> yahoo.com>
List: tech-kern
Date: 01/01/2005 21:47:59

Hello,

 I have managed to bootstrap netbsd on my Intel pxa255
board -with a machdep glue. But there is a problem
with pmap initialization. If someone can give me a
clue, it might help. Pasted below is the output on
bootup.

thanks
-kamal
(Continue reading)

Permalink | Reply | 0 comments |
headers
Toru Nishimura | 4 Jan 2005 15:06

Re:-pmap initlialization

Kamal R. Prasad asked;

> I found out that as soon as tlb is invalidated, the
> system goes for a toss. Does the configuration below
> look OK? The kernel is loaded at 0xa0200000 and the L1
> page table is at 0xa01fc000.
> Other than an incorrect base addr for the page table,
> what else could cause the system to go for a toss?

- xxx_machdep.c of any evbarm allocates some pmap structure
for kernel address below 2MB range.  PLS take a close look at
#define alloc_pages in the middle of initarm().

- The following debug message implicates your data segment
base is va=0xc030d000.  PLS check objdump --headers result to
see it makes sense.

> pmap_map_chunk: pa=0xa0200000 va=0xc0200000
> size=0x10d000 resid=0x10d000 prot=01
> SPPPPPPPPPPPPP
> pmap_map_chunk: pa=0xa030d000 va=0xc030d000
> size=0x3d000 resid=0x3d000 prot=0x31
> PPPLLLPPPPPPPPPP

Toru Nishimura/ALKYL Technology
Permalink | Reply | 0 comments |
headers
Kamal R. Prasad | 4 Jan 2005 15:44
Picon
Favicon

Re:-pmap initlialization

First of all, let me apologize for multiple posts -it
happened inadverently.
The hang happens in
sys/arch/arm/arm/cpufunc_asm_xscale.S in
xscale_setttb() which sets the MMU TLB base address
register. The address is 0xa01fc000. Is this the right
address? Does the datasheet mention a phys addr for
the TLB base addr?

--- Toru Nishimura <locore64 <at> alkyltechnology.com>
wrote:

> Kamal R. Prasad asked;
> 
> > I found out that as soon as tlb is invalidated,
> the
> > system goes for a toss. Does the configuration
> below
> > look OK? The kernel is loaded at 0xa0200000 and
> the L1
> > page table is at 0xa01fc000.
> > Other than an incorrect base addr for the page
> table,
> > what else could cause the system to go for a toss?
> 
> - xxx_machdep.c of any evbarm allocates some pmap
> structure
> for kernel address below 2MB range.  PLS take a
> close look at
> #define alloc_pages in the middle of initarm().
(Continue reading)

Permalink | Reply | 0 comments |
headers
Richard Earnshaw | 4 Jan 2005 18:06
Picon

Re: KVA pitfall in initarm()

On Sun, 2004-12-26 at 11:53, Toru Nishimura wrote:
> Many ARM ports arrange KVA for text and data segments as two
> adjuscent address ranges in the following way;
> 
>         {
>                 extern char etext[], _end[];
>                 size_t textsize = (uintptr_t) etext - KERNEL_TEXT_BASE;
>                 size_t totalsize = (uintptr_t) _end - KERNEL_TEXT_BASE;
>                 u_int logical;
> 
>                 textsize = (textsize + PGOFSET) & ~PGOFSET;
>                 totalsize = (totalsize + PGOFSET) & ~PGOFSET;
>                 
>                 logical = boofoowoo;    /* offset of kernel in RAM */
>                 logical += pmap_map_chunk(l1pagetable, KERNEL_BASE + logical,
>                     physical_start + logical, textsize,
>                     VM_PROT_READ|VM_PROT_WRITE, PTE_CACHE);
>                 logical += pmap_map_chunk(l1pagetable, KERNEL_BASE + logical,
>                     physical_start + logical, totalsize - textsize,
>                     VM_PROT_READ|VM_PROT_WRITE, PTE_CACHE);
>         }
> 
> This is a potetial pitfall for the case when there is a hole (segment alignment)
> between text segment and data segment.  Data segment would result in
> be mapped incorrect KVA (it badly bited me, indeed).  The right way is to map
> data segment at its own offset, not the end of text.  

There certainly is an assumption that the pages of the kernel text and
data have a 1:1 correspondence between the image loaded at boot and the
pages after mapping.  That doesn't mean there can be no alignment of the
(Continue reading)

Permalink | Reply | 0 comments |

Gmane