window_ret_fault on SPARC

[Baurjan Ismagulov]

Hi,

my kernel gets into window_ret_fault while returning from trap to
user mode.

In the trap return code, SRMMU sets "Fault Address Valid" during
"save; LOAD_WINDOW(sp); restore". After that, SFSR=00000127 and
SFAR=efffed18. Loaded %fp value is efffece0, and [efffece0+38]
contains efffed50 -- at least, that is what I see in kgdb using "p/x
*(struct reg_window *)0xefffece0". I also tried to see that using
printk from window_ret_fault, but the kernel hung solidly (L1-A didn't
work).

I followed all vm_next links in current->mm->mmap; efffd000-effff000
area seems to be there. However, I was unable to track it via
current->pgd.

So:

1. Why do I get multiple (SFSR.OV=1) faults in srmmu_rett_stackchk?

2. How can I see whether page directories of the current process are
   set up correctly?

I would be very grateful if someone could help me to solve this problem. I'm using 2.2.19 on sun4m.

Thanks in advance,
Baurjan.