debug application in sdram (microblaze system)

  • Thread starter Frank van Eijkelenburg
  • Start date
F

Frank van Eijkelenburg

Guest
Hi,

Is it possible to debug an application which is in sdram by use of xmdstub?
I have a small bootloader program which programs a final application into
sdram (by use of xmodem). Now I want to debug this application, is that
possible or should I use the opb mdm device?! If it's possible, I guess I
have to build a bootloader with the xmdstub and make a connection. But how
to get my application into the sdram? Can I use the "dow" command at the xmd
command line or an option inside gdb to download the application into sdram
(I'm asking this, because in the normal way the user has to give some input
via the serial port, but that's the same port for debugging thus I can't
give any input). And how to step through the code when the application is in
the sdram?!

TIA,
Frank
 
Frank,

Yes, you can do exactly what you have described. When you use the "dow"
command from the XMD command prompt (after an mbconnect), XMD will use
the available memory interface core to place the .elf at whatever
location has been specified in your linker script. Therefore, you want
to "dow" the final application, not the bootloader. You are essentially
letting MDM/XMD act as your bootloader in this case.

Once the .elf is downloaded, you can then "con" to run or "stp" (step)
from the command line and see what instruction is causing the problem
you described in your previous post. If the system hangs on one of your
first steps then you probably have a problem with your hardware (likely
logical or timing issue with your memory interface and/or board). If
this is the case, try using the "mwr" and "mrd" commands from XMD to see
if you can get data in and out of your memory correctly.

You can also try to "dow" your .elf file, do an object dump (mb-objdump)
of the .elf, then use "mrd" on a large block of the instruction memory,
and do a comparison between the dumped .elf and what's in the memory.
If there are any mismatches then there's likely a problem with the
memory interface hardware/timing.

If it's many instructions into the run, and that instruction memory
location looks correct, then the software is likely at fault. Use that
dumped .elf file to figure out which instruction is causing the problem
and debug accordingly.

Hopefully you are able to resolve the problem using this information.
If not then please contact our support hotline for more help.


Best regards,
Ryan Laity
Xilinx Applications

Frank van Eijkelenburg wrote:
Hi,

Is it possible to debug an application which is in sdram by use of xmdstub?
I have a small bootloader program which programs a final application into
sdram (by use of xmodem). Now I want to debug this application, is that
possible or should I use the opb mdm device?! If it's possible, I guess I
have to build a bootloader with the xmdstub and make a connection. But how
to get my application into the sdram? Can I use the "dow" command at the xmd
command line or an option inside gdb to download the application into sdram
(I'm asking this, because in the normal way the user has to give some input
via the serial port, but that's the same port for debugging thus I can't
give any input). And how to step through the code when the application is in
the sdram?!

TIA,
Frank
Click to expand...
 
It looks like there are some troubles with debugging an application in sdram
while caching is enabled (both instruction cache and data cache are
enabled). If I remove the cache, I am able to debug my application from
sdram. Could anyone confirm this (I saw a few postings at the xilinx
microblaze forum, but I couldn't find an confirmation from xilinx about this
behaviour)?!

Frank


"Ryan Laity" <ryan_dot_laity@x-i-l-i-n-x_pleasenospam_dot_com> wrote in
message news:c0dod1$5kk1@cliff.xsj.xilinx.com...
Frank,

Yes, you can do exactly what you have described. When you use the "dow"
command from the XMD command prompt (after an mbconnect), XMD will use
the available memory interface core to place the .elf at whatever
location has been specified in your linker script. Therefore, you want
to "dow" the final application, not the bootloader. You are essentially
letting MDM/XMD act as your bootloader in this case.

Once the .elf is downloaded, you can then "con" to run or "stp" (step)
from the command line and see what instruction is causing the problem
you described in your previous post. If the system hangs on one of your
first steps then you probably have a problem with your hardware (likely
logical or timing issue with your memory interface and/or board). If
this is the case, try using the "mwr" and "mrd" commands from XMD to see
if you can get data in and out of your memory correctly.

You can also try to "dow" your .elf file, do an object dump (mb-objdump)
of the .elf, then use "mrd" on a large block of the instruction memory,
and do a comparison between the dumped .elf and what's in the memory.
If there are any mismatches then there's likely a problem with the
memory interface hardware/timing.

If it's many instructions into the run, and that instruction memory
location looks correct, then the software is likely at fault. Use that
dumped .elf file to figure out which instruction is causing the problem
and debug accordingly.

Hopefully you are able to resolve the problem using this information.
If not then please contact our support hotline for more help.


Best regards,
Ryan Laity
Xilinx Applications

Frank van Eijkelenburg wrote:
Hi,

Is it possible to debug an application which is in sdram by use of
xmdstub?
I have a small bootloader program which programs a final application
into
sdram (by use of xmodem). Now I want to debug this application, is that
possible or should I use the opb mdm device?! If it's possible, I guess
I
have to build a bootloader with the xmdstub and make a connection. But
how
to get my application into the sdram? Can I use the "dow" command at the
xmd
command line or an option inside gdb to download the application into
sdram
(I'm asking this, because in the normal way the user has to give some
input
via the serial port, but that's the same port for debugging thus I can't
give any input). And how to step through the code when the application
is in
the sdram?!

TIA,
Frank
Click to expand...
 
Frank,

I was not aware that you were using cache with your SDRAM in this
system. To debug with cache support, you must use the MDM.

Ryan Laity
Xilinx



Frank van Eijkelenburg wrote:
It looks like there are some troubles with debugging an application in sdram
while caching is enabled (both instruction cache and data cache are
enabled). If I remove the cache, I am able to debug my application from
sdram. Could anyone confirm this (I saw a few postings at the xilinx
microblaze forum, but I couldn't find an confirmation from xilinx about this
behaviour)?!

Frank


"Ryan Laity" <ryan_dot_laity@x-i-l-i-n-x_pleasenospam_dot_com> wrote in
message news:c0dod1$5kk1@cliff.xsj.xilinx.com...

Frank,

Yes, you can do exactly what you have described. When you use the "dow"
command from the XMD command prompt (after an mbconnect), XMD will use
the available memory interface core to place the .elf at whatever
location has been specified in your linker script. Therefore, you want
to "dow" the final application, not the bootloader. You are essentially
letting MDM/XMD act as your bootloader in this case.

Once the .elf is downloaded, you can then "con" to run or "stp" (step)
from the command line and see what instruction is causing the problem
you described in your previous post. If the system hangs on one of your
first steps then you probably have a problem with your hardware (likely
logical or timing issue with your memory interface and/or board). If
this is the case, try using the "mwr" and "mrd" commands from XMD to see
if you can get data in and out of your memory correctly.

You can also try to "dow" your .elf file, do an object dump (mb-objdump)
of the .elf, then use "mrd" on a large block of the instruction memory,
and do a comparison between the dumped .elf and what's in the memory.
If there are any mismatches then there's likely a problem with the
memory interface hardware/timing.

If it's many instructions into the run, and that instruction memory
location looks correct, then the software is likely at fault. Use that
dumped .elf file to figure out which instruction is causing the problem
and debug accordingly.

Hopefully you are able to resolve the problem using this information.
If not then please contact our support hotline for more help.


Best regards,
Ryan Laity
Xilinx Applications

Frank van Eijkelenburg wrote:

Hi,

Is it possible to debug an application which is in sdram by use of

xmdstub?

I have a small bootloader program which programs a final application

into

sdram (by use of xmodem). Now I want to debug this application, is that
possible or should I use the opb mdm device?! If it's possible, I guess

I

have to build a bootloader with the xmdstub and make a connection. But

how

to get my application into the sdram? Can I use the "dow" command at the

xmd

command line or an option inside gdb to download the application into

sdram

(I'm asking this, because in the normal way the user has to give some

input

via the serial port, but that's the same port for debugging thus I can't
give any input). And how to step through the code when the application

is in

the sdram?!

TIA,
Frank
Click to expand...
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top