[PATCH 5.10.y/5.15.y 0/4] ARM: fix might_sleep() WARNING for mmap_write_lock() around show_pte()

[Xie Yuanbin]

On linux-5.10.y and linux-5.15.y, commit 07e4d5380f2a844ab7a1 ("ARM:
9485/1: mm: acquire mmap write lock around show_pte() for user faults")
introduces a might_sleep() WARNING when using mmap write lock. This is
due to the lack of upstream commit 59e4f3b45b96a24fc9b7 ("ARM: ensure
interrupts are enabled in __do_user_fault()") and
commit e79ca91165d4fd18549c ("ARM: 9484/1: enable interrupts when
unhandled user faults are triggered") .

This series backports all these upstream patches into the linux-5.10.y:
1. commit 7733bc7d299d682f2723 ("ARM: fix hash_name() fault")
2. commit fd2dee1c6e2256f726ba ("ARM: fix branch predictor hardening")
3. commit 59e4f3b45b96a24fc9b7 ("ARM: ensure interrupts are enabled in
   __do_user_fault()")
4. commit e79ca91165d4fd18549c ("ARM: 9484/1: enable interrupts when
   unhandled user faults are triggered")

In the upstream version, the following patches are a patch set:
a. commit dea20281ac8822661576 ("ARM: group is_permission_fault() with
   is_translation_fault()")
b. commit 40b466db1dffb41f0529 ("ARM: allow __do_kernel_fault() to report
   execution of memory faults")
c. commit 7733bc7d299d682f2723 ("ARM: fix hash_name() fault"):
   which is 1.
d. commit fd2dee1c6e2256f726ba ("ARM: fix branch predictor hardening"):
   which is 2.
The fixed issues by this patch set also exist in linux-5.10 && linux-5.15.
a. and b. do not need to be backported, and after c. and d.
is backported, 3. and 4. can be safely patched.