[U-Boot] Cannot access memory at address 0xd8013fa8 when using gdb/BDI3000 to debug u-boot
I'm trying to get u-boot version 1.3.4 working a custom MPC8548 based board (version 1.1.4 currently works fine on this board so the hardware is known to be good). I'm encountering the following problem during the early stages of the u-boot initialization and any insights as to what the problem may be would be greatly appreciated:
I have a BDI3000 and i'm using gdb for debug sessions. I started with the MPC8548CDS.h as my template for the build configuration file for this board.
1) in the start.S file the L1 DCache is configured to be used for initial RAM, this occurs in the code at the 'switch_as:' label. I set the CFG_INIT_RAM_ADDR to the value 0xd8010000 and this value is used to configure the L1CFG0 register. 2) execution soon jumps to the '_start_cont:' label and the stack is set up in this initial RAM area 3) I set a breakpoint at '_start_cont' and tried to examine the initial RAM memory with the gdb 'x' command: (gdb) x/16x 0xd8010000 0xd8010000: Cannot access memory at address 0xd8010000
So at this point should gdb be able to examine this memory area?
I compared the code in start.S for 1.3.4 with the code from 1.1.4 and although not identical the instructions to setup the L1 cache and use it for initial RAM end up using the same values and doing the same operations...
I can continue to execute code and soon we jump into the c routines where the global data pointer is initialized, attempts to display this gd_t structure give the same error message (no surprise there), so I'm guessing this structure does not get initialized properly and function using it read bogus values...
but I can continue stepping thru the initialzation functions, and for example when the console is initialized a long string of garbage shows up in the console terminal window, the baudrate comes from the gd_t struct so its not what I configured...
so i think this is close to working but something seems broken with using the L1 Dcache for initial ram...
thanks in advance for any ideas and suggestion and please let me know if there is additional info that might be useful for you to understand the problem...
davis mcpherson
Dear davis mcpherson,
In message 4D3B0525.4080107@gmail.com you wrote:
I'm trying to get u-boot version 1.3.4 working a custom MPC8548 based board (version 1.1.4 currently works fine on this board so the hardware is known to be good). I'm encountering the following problem during the
Sorry, but I don't even continue reading. v1.3.4 is 2.5 years old, i. e. hopelessly obsolete. Any efforts on this old stuff are only a waste of time.
Do yourself (and us) a favour and use current code instead (at least v2010.12, or bettter current top of tree from the git repository).
Best regards,
Wolfgang Denk
Ok, I'm updated my u-boot source code to the v2010.12 release and i'm still seeing the same behaviour as with the earlier version...
in summary:
early in the boot process, in the start.S file, the L1 DCache is configured for use as some initial RAM, the stack is located there, and the global data structure, gd_t, is located there... when I debug my u-boot build i can set breakpoints, e.g. in cpu_init_f, and the breakpoint is hit and I can single step thru the code, but when the return from the function is executed i get the 'Cannot access memory at address..." error, but apparently the call does return, because I can single step on to the next function that is call, board_init_f...
Additionally, if I set a breakpoint in board_init_f, after the statement that initializes the gd_t pointer variable 'gd', and then try to print the structure or just examine the memory locations i get the same "Cannot access memory at address..." error...
so i'm assuming gdb working with a bdi3000 should be able to access this L1Dcache when configured to be used as this initial RAM area for u-boot, is this assumption correct?
the code in start.S is coded to be common for all mpc85xx boards (its in the arch/powerpc/cpu/mpc85xx directory) with some config options from the board specific <board>_config.h file. My board is custom but modeled after the MPC8548CDS and at least at this point of execution I think components common to any MPC8548 board are being set up... I left the #define that specifies the address for this initial RAM unchanged, 0xe4010000 ( i had changed it to 0xd8010000 when i was using 1.3.4, but could see a reason that was required...)
could my bdi3000 config file affect this in any way? i've attached it for reference...
my thanks for your time and any ideas or suggestions are greatly appreciated...
davis mcpherson
On 01/22/2011 12:39 PM, Wolfgang Denk wrote:
Dear davis mcpherson,
In message4D3B0525.4080107@gmail.com you wrote:
I'm trying to get u-boot version 1.3.4 working a custom MPC8548 based board (version 1.1.4 currently works fine on this board so the hardware is known to be good). I'm encountering the following problem during the
Sorry, but I don't even continue reading. v1.3.4 is 2.5 years old, i. e. hopelessly obsolete. Any efforts on this old stuff are only a waste of time.
Do yourself (and us) a favour and use current code instead (at least v2010.12, or bettter current top of tree from the git repository).
Best regards,
Wolfgang Denk
participants (2)
-
davis mcpherson -
Wolfgang Denk