In ARM Cortex-A7 platform which includes L1 and L2 level caches，I start cache setup flow as follows:
1. Enable SMP bit and disable MMU.
2. Disable I cache in L1, and invalidate it , then enable it.
3. Invalidate D cache in L1 and unified cache of L2.
4. Enable MMU.
5. Bl back to my main routine.
But after enabling MMU, pc hangs before pop and bl, and I don't know why.
By tracing the source , I found that signal "nxt_ls_valid_wr" has asserted for one clk, and causing stall_wr asserted and kept high, then the signal "en_pc_ret" keeps low, so that pc did't change anymore.
What are the mainly possible causes that would leading to such situation?
From past experience most likely causes are X propagation (I'm assuming you running a simulation based on you viewing signals) or your translation tables are incorrect.
If you are simulating RTL I suspect you have a support contract with ARM. Which means you could submit the question direct to ARM's support team:
https://silver.arm.com/support/
Thank you for replying,
I will check my translation table and see if that's the problem.