Fwd: ubi: fastmap: Fix a series of wear leveling problems

Fwd: ubi: fastmap: Fix a series of wear leveling problems

[Bagas Sanjaya]

Hi,

I notice a bug report with proposed fixes on Bugzilla [1]. Quoting from it
(only the first problem that is quoted):

> Problem 1: large erase counter for single fastmap data PEB
> 
> Config:
> x86_64 qemu
> flash: nandsim
> CONFIG_MTD_UBI_WL_THRESHOLD=128
> CONFIG_MTD_UBI_FASTMAP=y
> ubi->beb_rsvd_pebs=0
> 
> Running fsstress on ubifs for 3h(fastmap data PEB has large erase counter than others):
> =========================================================
> from              to     count      min      avg      max
> ---------------------------------------------------------
> 0        ..        9:        0        0        0        0
> 10       ..       99:      532       84       92       99
> 100      ..      999:    15787      100      147      229
> 1000     ..     9999:       64     4699     4765     4826
> 10000    ..    99999:        0        0        0        0
> 100000   ..      inf:        1   272935   272935   272935
> ---------------------------------------------------------
> Total               :    16384       84      180   272935
> PEB 8031(ec=272935) is always taken for fastmap data.
> 
> After fix, running fsstress on ubifs for 12h(no pool reservation), no individual peb has big erase counter:
> =========================================================
> from              to     count      min      avg      max
> ---------------------------------------------------------
> 0        ..        9:        0        0        0        0
> 10       ..       99:        0        0        0        0
> 100      ..      999:    16320      609      642      705
> 1000     ..     9999:        0        0        0        0
> 10000    ..    99999:       64    18176    18234    18303
> 100000   ..      inf:        0        0        0        0
> ---------------------------------------------------------
> Total               :    16384      609      710    18303

See Bugzilla for the full thread (with other problems mentioned) and
attached patch series that fixes them.

Zhihao: I asked you on BZ to send your patches to linux-mtd list,
but you didn't respond there. Would you like to send them for
review?

Thanks.

[1]: https://bugzilla.kernel.org/show_bug.cgi?id=217787

-- 
An old man doll... just what I always wanted! - Clara

Re: Fwd: ubi: fastmap: Fix a series of wear leveling problems

[Zhihao Cheng]

在 2023/8/16 17:56, Bagas Sanjaya 写道:
Hi,
> Hi,
> 
> I notice a bug report with proposed fixes on Bugzilla [1]. Quoting from it
> (only the first problem that is quoted):
> 
>> Problem 1: large erase counter for single fastmap data PEB
>>
>> Config:
>> x86_64 qemu
>> flash: nandsim
>> CONFIG_MTD_UBI_WL_THRESHOLD=128
>> CONFIG_MTD_UBI_FASTMAP=y
>> ubi->beb_rsvd_pebs=0
>>
>> Running fsstress on ubifs for 3h(fastmap data PEB has large erase counter than others):
>> =========================================================
>> from              to     count      min      avg      max
>> ---------------------------------------------------------
>> 0        ..        9:        0        0        0        0
>> 10       ..       99:      532       84       92       99
>> 100      ..      999:    15787      100      147      229
>> 1000     ..     9999:       64     4699     4765     4826
>> 10000    ..    99999:        0        0        0        0
>> 100000   ..      inf:        1   272935   272935   272935
>> ---------------------------------------------------------
>> Total               :    16384       84      180   272935
>> PEB 8031(ec=272935) is always taken for fastmap data.
>>
>> After fix, running fsstress on ubifs for 12h(no pool reservation), no individual peb has big erase counter:
>> =========================================================
>> from              to     count      min      avg      max
>> ---------------------------------------------------------
>> 0        ..        9:        0        0        0        0
>> 10       ..       99:        0        0        0        0
>> 100      ..      999:    16320      609      642      705
>> 1000     ..     9999:        0        0        0        0
>> 10000    ..    99999:       64    18176    18234    18303
>> 100000   ..      inf:        0        0        0        0
>> ---------------------------------------------------------
>> Total               :    16384      609      710    18303
> 
> See Bugzilla for the full thread (with other problems mentioned) and
> attached patch series that fixes them.
> 
> Zhihao: I asked you on BZ to send your patches to linux-mtd list,
> but you didn't respond there. Would you like to send them for
> review?
> 

Sorry for the delayed response. Yes, I have sent the fix patches in 
https://patchwork.ozlabs.org/project/linux-mtd/list/?series=368534.
Any suggestions are welcomed.


> Thanks.
> 
> [1]: https://bugzilla.kernel.org/show_bug.cgi?id=217787
>