ee4ad3b25d
Add patch 0017 for kernel-rt to fix severe performance regression
in RT kernel's block multi-queue layer.
In RT kernel (PREEMPT_RT), commit 6bda857bcbb86 ("block: fix ordering
between checking QUEUE_FLAG_QUIESCED request adding") introduced
spinlock_t queue_lock usage in blk_mq_run_hw_queue(). While this works
correctly in standard kernel, it causes catastrophic serialization in
RT kernel where spinlock_t converts to sleeping rt_mutex.
The issue occurs when multiple MSI-X interrupt vectors contend on
queue_lock in blk_mq_run_hw_queue(), causing IRQ threads to serialize
and enter D-state.
The patch converts quiesce_depth to atomic_t and uses it directly for
quiesce state checking, eliminating QUEUE_FLAG_QUIESCED entirely. This
removes the need for any locking in the hot path.
Tested on MegaRAID 12GSAS with 128 MSI-X vectors (120 hw queues):
- Before: 153 MB/s, IRQ threads in D-state
- After: 640 MB/s, no IRQ threads blocked
Verification:
- Build kernel and out of tree modules success for rt and std.
- Build iso success for rt and std.
- Install success onto a All-in-One lab with std kernel, and can
switch to rt kernel.
- Boot up successfully in the lab.
- The sanity testing was run and the test results PASS.
Story: 2011644
Task: 53580
Closes-Bug: 2139132
Change-Id: Ida9e378127637dc4abb0737846405b740c9913b8
Signed-off-by: Ionut Nechita <ionut.nechita@windriver.com>