Devices that make up DPU, i.e. graphics card, request their interrupts
from this "virtual" interrupt chip. The interrupt chip builds upon a GIC
SPI interrupt that raises high when any of the interrupts in the DPU's
irq status register are triggered. From the kernel's perspective this is
a chained irq chip, so requesting a flow handler for the GIC SPI and
then calling generic IRQ handling code from that irq handler is not
completely proper. It's better to convert this to a chained irq so that
the GIC SPI irq doesn't appear in /proc/interrupts, can't have CPU
affinity changed, and won't be accounted for with irq stats. Doing this
also silences a recursive lockdep warning because we can specify a
different lock class for the chained interrupts, silencing a warning
that is easy to see with 'threadirqs' on the kernel commandline.

 WARNING: inconsistent lock state
 4.19.10 #76 Tainted: G        W
 --------------------------------
 inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
 irq/40-dpu_mdss/203 [HC0[0]:SC0[2]:HE1:SE0] takes:
 0000000053ea9021 (&irq_desc_lock_class){?.-.}, at: handle_level_irq+0x34/0x26c
 {IN-HARDIRQ-W} state was registered at:
   lock_acquire+0x244/0x360
   _raw_spin_lock+0x64/0xa0
   handle_fasteoi_irq+0x54/0x2ec
   generic_handle_irq+0x44/0x5c
   __handle_domain_irq+0x9c/0x11c
   gic_handle_irq+0x208/0x260
   el1_irq+0xb4/0x130
   arch_cpu_idle+0x178/0x3cc
   default_idle_call+0x3c/0x54
   do_idle+0x1a8/0x3dc
   cpu_startup_entry+0x24/0x28
   rest_init+0x240/0x270
   start_kernel+0x5a8/0x6bc
 irq event stamp: 18
 hardirqs last  enabled at (17): [<ffffff9042385e80>] _raw_spin_unlock_irq+0x40/0xc0
 hardirqs last disabled at (16): [<ffffff904237a1f4>] __schedule+0x20c/0x1bbc
 softirqs last  enabled at (0): [<ffffff9040f318d0>] copy_process+0xb50/0x3964
 softirqs last disabled at (18): [<ffffff9041036364>] local_bh_disable+0x8/0x20

 other info that might help us debug this:
  Possible unsafe locking scenario:

        CPU0
        ----
   lock(&irq_desc_lock_class);
   <Interrupt>
     lock(&irq_desc_lock_class);

  *** DEADLOCK ***

 no locks held by irq/40-dpu_mdss/203.

 stack backtrace:
 CPU: 0 PID: 203 Comm: irq/40-dpu_mdss Tainted: G        W         4.19.10 #76
 Call trace:
  dump_backtrace+0x0/0x2f8
  show_stack+0x20/0x2c
  __dump_stack+0x20/0x28
  dump_stack+0xcc/0x10c
  mark_lock+0xbe0/0xe24
  __lock_acquire+0x4cc/0x2708
  lock_acquire+0x244/0x360
  _raw_spin_lock+0x64/0xa0
  handle_level_irq+0x34/0x26c
  generic_handle_irq+0x44/0x5c
  dpu_mdss_irq+0x64/0xec
  irq_forced_thread_fn+0x58/0x9c
  irq_thread+0x120/0x1dc
  kthread+0x248/0x260
  ret_from_fork+0x10/0x18
 ------------[ cut here ]------------
 irq 169 handler irq_default_primary_handler+0x0/0x18 enabled interrupts

Cc: Sean Paul <seanpaul@chromium.org>
Cc: Jordan Crouse <jcrouse@codeaurora.org>
Cc: Jayant Shekhar <jshekhar@codeaurora.org>
Cc: Rajesh Yadav <ryadav@codeaurora.org>
Cc: Jeykumar Sankaran <jsanka@codeaurora.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Sean Paul <seanpaul@chromium.org>