Re: [PATCH] scsi: ufs: core: move some irq handling back to hardirq (with time limit)

["André Draszik" <andre.draszik@linaro.org>]

On Thu, 2025-07-24 at 10:54 +0100, André Draszik wrote:
> fio results on Pixel 6:
>   read / 1 job     original    after    this commit
>     min IOPS        4,653.60   2,704.40    3,902.80
>     max IOPS        6,151.80   4,847.60    6,103.40
>     avg IOPS        5,488.82   4,226.61    5,314.89
>     cpu % usr           1.85       1.72        1.97
>     cpu % sys          32.46      28.88       33.29
>     bw MB/s            21.46      16.50       20.76
> 
>   read / 8 jobs    original    after    this commit
>     min IOPS       18,207.80  11,323.00   17,911.80
>     max IOPS       25,535.80  14,477.40   24,373.60
>     avg IOPS       22,529.93  13,325.59   21,868.85
>     cpu % usr           1.70       1.41        1.67
>     cpu % sys          27.89      21.85       27.23
>     bw MB/s            88.10      52.10       84.48
> 
>   write / 1 job    original    after    this commit
>     min IOPS        6,524.20   3,136.00    5,988.40
>     max IOPS        7,303.60   5,144.40    7,232.40
>     avg IOPS        7,169.80   4,608.29    7,014.66
>     cpu % usr           2.29       2.34        2.23
>     cpu % sys          41.91      39.34       42.48
>     bw MB/s            28.02      18.00       27.42
> 
>   write / 8 jobs   original    after    this commit
>     min IOPS       12,685.40  13,783.00   12,622.40
>     max IOPS       30,814.20  22,122.00   29,636.00
>     avg IOPS       21,539.04  18,552.63   21,134.65
>     cpu % usr           2.08       1.61        2.07
>     cpu % sys          30.86      23.88       30.64
>     bw MB/s            84.18      72.54       82.62

Given the severe performance drop introduced by the culprit
commit, it might make sense to instead just revert it for
6.16 now, while this patch here can mature and be properly
reviewed. At least then 6.16 will not have any performance
regression of such a scale.

Cheers,
Andre'