• State Not Answered
  • Locked Locked
  • Replies 3 replies
  • Subscribers 21 subscribers
  • Views 1304 views
  • Users 0 members are here
2025 survey

We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.

Options
Related
How was your experience today?
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

Why Doesn’t FVP Initialize Unused Memory to Zero (e.g., 0xcfdfdfdfdfdfdfcf)?

Qingyu Ma

I have a question about FVP (Fixed Virtual Platform) and its memory initialization behavior. Specifically, I noticed that unused memory in FVP is not set to 0, but instead contains non-zero values like 0xcfdfdfdfdfdfdfcf.

For example, when reading a non-used memory area in FVP, the values appear as non-zero by default. This behavior can lead to issues, such as:

  • If the OS running on FVP does not explicitly clear the .bss section, global variables in the OS could end up initialized with these non-zero default values.
  • This, in turn, may cause unexpected behavior or bugs during execution.

Given that FVP operates in a virtualized memory environment, I’m wondering why it doesn’t simply initialize all memory to zero. Is there a specific design rationale or technical limitation for this behavior?

  • +1 Arm Employee Badge

    I'm not sure which specific FVP you are using, but enter the command:

    <fvp_executable> --list-params > params.txt


    Then search for "fill" therein. You should find lines similar to:

    bp.dram.fill1=3755990991 # (int , init-time) default = '0xdfdfdfcf' : Fill pattern 1, initialise memory at start of simulation with alternating fill1, fill2 pattern
bp.dram.fill2=3487555551 # (int , init-time) default = '0xcfdfdfdf' : Fill pattern 2, initialise memory at start of simulation with alternating fill1, fill2 pattern


    You can then launch the FVP with command options (-C) to change these values, or specify them in an options file (-f), for example:
    -C bp.dram.fill1=0

    See also https://developer.arm.com/documentation/100966/1127/Getting-Started-with-Fixed-Virtual-Platforms/Configuring-the-model

  • 0 in reply to Ronan Synnott

    Thanks you very much, I successfully clear the memory of the OS's bss section.

  • 0 in reply to Ronan Synnott

    I have tested these configurations one by one and confirmed that the memory region where my kernel put is seems to be controlled by bp.s_dram. However, I’m still unsure about why this is the case.

    Could it be related to the security settings of the memory regions? For example, I understand that "P security memory regions block access from both Secure and Non-secure domains." And my kernel is put in `P` memory region, Could it be that regions blocking Non-secure access are controlled by bp.s_dram? Actually I didn't find any description of what there fill params mean.

    UPDATE:

    I set bp.secrue_memory=0, and the three DRAM regions is still controlled by bp.s_dram.fill1 & fill2, so that may not associate with security of the memory region.