RHEL9: performance regression on ext4 writes over md/RAID5

Issue

• High 'await' is observed at the md, volume group, and ext4 layers.
• The perf top indicates that the kernel is spending a significant amount of time in the ext4_mb_good_group function.

# perf top
Samples: 87K of event 'cycles:P', 4000 Hz, Event count (approx.): 45245285879 lost: 0/0 drop: 0/0
Overhead  Shared Object                                    Symbol
  20.85%  [kernel]                                         [k] ext4_mb_good_group
  16.72%  [kernel]                                         [k] ext4_get_group_info
   9.98%  [kernel]                                         [k] ext4_mb_choose_next_group_cr1.constprop.0
   6.16%  [kernel]                                         [k] rwsem_spin_on_owner
   4.43%  [kernel]                                         [k] __rcu_read_unlock