Skip to contentSkip to conversation

DBCR0 reset problem in MPC8548CDS

U-Boot Werner Nedel 4 participants 7 messages Apr 9, 2009 – Apr 17, 2009

lore
Share & tools
lore
#0 lore raw

Hi,

I'm trying to update my u-boot version (1.2) to the last one (2009.03). My board is very similar to MPC8548CDS. Everything was working fine till I tried to use the reset command. It hangs the processor in cpu_init_early_f, when it realocates ccsbar. I'd always used the hard reset in DBCR0 register instead HRESET_REQ that ins't wired in my board.

I saw that the value of MSR register came different before reset when I compare with the old u-boot (0x8000 -> 0x29200). There are new writes in this register in this new version, could any of them causing my problem?

The major difference beetwen my configuration and default resides in u-boot.lds. I changed ADDR(.text) by the fisical init address of my boot flash, that is different of my TEXT_BASE, but bootpg and resetvec still in the same address, so I think that it couldn't be the problem.

u-boot.lds:

Quoted 1 line

.bootpg 0xfff80000 + 0x7f000 :

< .bootpg ADDR(.text) + 0x7f000 :
{
cpu/mpc85xx/start.o (.bootpg)
} :text = 0xffff

Quoted 1 line

.resetvec 0xfff80000 + 0x7fffc :

< .resetvec ADDR(.text) + 0x7fffc :

  {
    *(.resetvec)

} :text = 0xffff

Quoted 1 line

. = 0xfff80000 + 0x80000;

< . = ADDR(.text) + 0x80000;

Thanks in advance,
Werner Nedel.

#1 lore raw

Actually, it hangs changing gd pointer, before set_tlb() call. In my older u-boot version, it was made in cpu_init_f function.

    /* Pointer is writable since we allocated a register for it */
    gd = (gd_t *) (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_GBL_DATA_OFFSET);
Quoted 36 lines · Werner Nedel

Hi,

I'm trying to update my u-boot version (1.2) to the last one (2009.03). My board is very similar to MPC8548CDS. Everything was working fine till I tried to use the reset command. It hangs the processor in cpu_init_early_f, when it realocates ccsbar. I'd always used the hard reset in DBCR0 register instead HRESET_REQ that ins't wired in my board.

I saw that the value of MSR register came different before reset when I compare with the old u-boot (0x8000 -> 0x29200). There are new writes in this register in this new version, could any of them causing my problem?

The major difference beetwen my configuration and default resides in u-boot.lds. I changed ADDR(.text) by the fisical init address of my boot flash, that is different of my TEXT_BASE, but bootpg and resetvec still in the same address, so I think that it couldn't be the problem.

u-boot.lds:

Quoted 1 line

.bootpg 0xfff80000 + 0x7f000 :

< .bootpg ADDR(.text) + 0x7f000 :
{
cpu/mpc85xx/start.o (.bootpg)
} :text = 0xffff

Quoted 1 line

.resetvec 0xfff80000 + 0x7fffc :

< .resetvec ADDR(.text) + 0x7fffc :

  {
    *(.resetvec)

} :text = 0xffff

Quoted 1 line

. = 0xfff80000 + 0x80000;

< . = ADDR(.text) + 0x80000;

Thanks in advance,
Werner Nedel.

#2 lore raw
Quoted 5 lines · Werner Nedel

Actually, it hangs changing gd pointer, before set_tlb() call. In my older u-boot version, it was made in cpu_init_f function.

? ?/* Pointer is writable since we allocated a register for it */ ? ?gd = (gd_t *) (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_GBL_DATA_OFFSET);

That usually means that your LAWBARs are misprogrammed.

Signature · 2 lines
Timur Tabi
Linux kernel developer at Freescale
#3 lore raw
Quoted 8 lines · Timur Tabi
Quoted 5 lines · Werner Nedel

Actually, it hangs changing gd pointer, before set_tlb() call. In my older u-boot version, it was made in cpu_init_f function.

   /* Pointer is writable since we allocated a register for it */
   gd = (gd_t *) (CONFIG_SYS_INIT_RAM_ADDR + CONFIG_SYS_GBL_DATA_OFFSET);

That usually means that your LAWBARs are misprogrammed.

The above line just loads a constant into a register, so the source of the problem is likely elsewhere (possibly the code itself can't be fetched).

-Scott

#4 lore raw
Quoted 2 lines · Scott Wood

The above line just loads a constant into a register, so the source of the problem is likely elsewhere (possibly the code itself can't be fetched).

I don't remember exactly why, but I once had the same symptom that Werner is reporting, and the cause was a misprogrammed LAWBAR.

Signature · 2 lines
Timur Tabi
Linux kernel developer at Freescale
#5 lore raw
In reply to Werner Nedel
Quoted 11 lines · Werner Nedel

I'm trying to update my u-boot version (1.2) to the last one (2009.03). My
board is very similar to MPC8548CDS. Everything was working fine till I
tried to use the reset command. It hangs the processor in cpu_init_early_f,
when it realocates ccsbar. I'd always used the hard reset in DBCR0 register
instead HRESET_REQ that ins't wired in my board.

Quoted 6 lines

I saw that the value of MSR register came different before reset when I
compare with the old u-boot (0x8000 -> 0x29200). There are new writes in
this register in this new version, could any of them causing my problem?

It would probably be useful to have any obvious differences in the physical address space from the MPC8548CDS config that exist.

- k

#6 lore raw

Sure. My complete phys addresses mapping:

DDR_SDRAM_BASE:     0x00000000 - 0x80000000 (2Gb)
NAND:            0x80000000 - 0xc0000000 (1Gb)
PCI1:            0xc0000000 - 0xd0000000 (256Mb)
FPGA:            0xd0000000 - 0xd0100000 (1Mb)
CCSRBAR (= PHYS):    0xe0000000 - 0xe4000000 (64Mb)
PCI_IO:            0xe3000000 - 0xe3100000 (1Mb)
INIT_RAM:        0xe4010000
CCSRBAR_DEFAULT:    0xff700000
BOOT_BLOCK=FLASH_BASE:    0xfff80000 - 0xffffffff (512Kb)

(TEXT_BASE=FLASH_BASE + 0x10000)

RIO, CADMUS, PCIE and PCI2 are undefined.

My law.c:
- SET_LAW(CONFIG_SYS_LBC_SDRAM_BASE, LAW_SIZE_256M, LAW_TRGT_IF_LBC),
+ SET_LAW(0x00000000, LAW_SIZE_512M, LAW_TRGT_IF_DDR),

and NAND, FPGA windows (and PCI was resized to 256Mb).

My tlb.c:
SET_TLB_ENTRY(1, CONFIG_SYS_BOOT_BLOCK, CONFIG_SYS_BOOT_BLOCK,

              MAS3_SX|MAS3_SW|MAS3_SR, MAS2_I|MAS2_G,

- 0, 0, BOOKE_PAGESZ_16M, 1),
+ 0, 0, BOOKE_PAGESZ_1M, 1),

SDRAM and CADMUS was changed by FPGA and NAND tlbs (and PCI was resized to 256Mb).

(nor)
#define CONFIG_SYS_BR0_PRELIM 0xfff81001
#define CONFIG_SYS_OR0_PRELIM 0xfff80e65

(fpga)
#define CONFIG_SYS_BR1_PRELIM 0xd0000801
#define CONFIG_SYS_OR1_PRELIM 0xfff06633

(nand)
#define CONFIG_SYS_BR2_PRELIM 0x80000801
#define CONFIG_SYS_OR2_PRELIM 0xc0008ef7

Something must be very wrong. Recently I tried to change my 512kb NOR by a 1Mb NOR and some accesses to fff00000 to fff80000 addresses range hangs my processor (fff80000 to ffffffff works fine), but it can be a problem in my flash driver.

Thanks for your help,

Werner Nedel.

Quoted 22 lines · Kumar Gala
Quoted 8 lines · Werner Nedel

I'm trying to update my u-boot version (1.2) to the last one (2009.03). My board is very similar to MPC8548CDS. Everything was working fine till I tried to use the reset command. It hangs the processor in cpu_init_early_f,
when it realocates ccsbar. I'd always used the hard reset in DBCR0 register
instead HRESET_REQ that ins't wired in my board.

Quoted 4 lines

I saw that the value of MSR register came different before reset when I compare with the old u-boot (0x8000 -> 0x29200). There are new writes in this register in this new version, could any of them causing my problem?

It would probably be useful to have any obvious differences in the physical address space from the MPC8548CDS config that exist.

- k