Re: [PATCH v2] mm/page-writeback: Raise wb_thresh to prevent write blocking with strictlimit

[Kemeng Shi]

on 11/21/2024 4:05 PM, Jim Zhao wrote:
>> on 11/19/2024 8:29 PM, Jim Zhao wrote:
>>> Thanks, Jan, I just sent patch v2, could you please review it ?
>>>
>>> And I found the debug info in the bdi stats.
>>> The BdiDirtyThresh value may be greater than DirtyThresh, and after applying this patch, the value of BdiDirtyThresh could become even larger.
>>>
>>> without patch:
>>> ---
>>> root@ubuntu:/sys/kernel/debug/bdi/8:0# cat stats
>>> BdiWriteback:                0 kB
>>> BdiReclaimable:             96 kB
>>> BdiDirtyThresh:        1346824 kB
>>> DirtyThresh:            673412 kB
>>> BackgroundThresh:       336292 kB
>>> BdiDirtied:              19872 kB
>>> BdiWritten:              19776 kB
>>> BdiWriteBandwidth:           0 kBps
>>> b_dirty:                     0
>>> b_io:                        0
>>> b_more_io:                   0
>>> b_dirty_time:                0
>>> bdi_list:                    1
>>> state:                       1
>>>
>>> with patch:
>>> ---
>>> root@ubuntu:/sys/kernel/debug/bdi/8:0# cat stats
>>> BdiWriteback:               96 kB
>>> BdiReclaimable:            192 kB
>>> BdiDirtyThresh:        3090736 kB
>>> DirtyThresh:            650716 kB
>>> BackgroundThresh:       324960 kB
>>> BdiDirtied:             472512 kB
>>> BdiWritten:             470592 kB
>>> BdiWriteBandwidth:      106268 kBps
>>> b_dirty:                     2
>>> b_io:                        0
>>> b_more_io:                   0
>>> b_dirty_time:                0
>>> bdi_list:                    1
>>> state:                       1
>>>
>>>
>>> @kemeng, is this a normal behavior or an issue ?
>> Hello, this is not a normal behavior, could you aslo send the content in
>> wb_stats and configuired bdi_min_ratio.
>> I think the improper use of bdi_min_ratio may cause the issue.
> 
> the min_ratio is 0
> ---
> root@ubuntu:/sys/class/bdi/8:0# cat min_bytes
> 0
> root@ubuntu:/sys/class/bdi/8:0# cat min_ratio
> 0
> root@ubuntu:/sys/class/bdi/8:0# cat min_ratio_fine
> 0
> 
> wb_stats:
> ---
> 
> root@ubuntu:/sys/kernel/debug/bdi/8:0# cat stats
> BdiWriteback:                0 kB
> BdiReclaimable:            480 kB
> BdiDirtyThresh:        1664700 kB
> DirtyThresh:            554900 kB
> BackgroundThresh:       277108 kB
> BdiDirtied:              82752 kB
> BdiWritten:              82752 kB
> BdiWriteBandwidth:      205116 kBps
> b_dirty:                     6
> b_io:                        0
> b_more_io:                   0
> b_dirty_time:                0
> bdi_list:                    1
> state:                       1
> root@ubuntu:/sys/kernel/debug/bdi/8:0# cat wb_stats
...
> 
> WbCgIno:                  416
> WbWriteback:                0 kB
> WbReclaimable:            288 kB
> WbDirtyThresh:         554836 kB
> WbDirtied:              47616 kB
> WbWritten:              47424 kB
> WbWriteBandwidth:         168 kBps
> b_dirty:                    1
> b_io:                       0
> b_more_io:                  0
> b_dirty_time:               0
> state:                      5
> 
...
> WbCgIno:                 3186
> WbWriteback:                0 kB
> WbReclaimable:             96 kB
> WbDirtyThresh:         554788 kB
> WbDirtied:               1056 kB
> WbWritten:               1152 kB
> WbWriteBandwidth:         152 kBps
> b_dirty:                    1
> b_io:                       0
> b_more_io:                  0
> b_dirty_time:               0
> state:                      5
...
> WbCgIno:                   72
> WbWriteback:                0 kB
> WbReclaimable:              0 kB
> WbDirtyThresh:         554836 kB
> WbDirtied:                 96 kB
> WbWritten:                192 kB
> WbWriteBandwidth:           4 kBps
> b_dirty:                    1
> b_io:                       0
> b_more_io:                  0
> b_dirty_time:               0
> state:                      5
Hi Jim,
Sorry for late reply.
The dirty thresh of these three groups is as high as dirty thresh
of whole bdi which is unusual. In __wb_calc_thresh, we calculate
dirty thresh of group by (numerator / denominator) * (thresh of
whole bdi) roughly, so (numerator / denominator) of these three
groups is 1. However, the sum of (numerator / denominator) of
all groups is suppose to be 1.
In fprop_fraction_percpu, we know numerator and denominator are
retrieved from percpu_counter, I think it's because of percpu
counter errors make this happen. Do we cat wb_stats and stats
when writeback load is low. If so, it's likely caused by percpu
counter errors .


> ubuntu24.04 desktop + kernel 6.12.0
> default cgroups, not configured manually.
> 
> ---
> Thanks
> Jim Zhao
>