Jan Kara wrote:
> On Wed 30-04-08 22:13:35, Edward Shishkin wrote:
>   
>> Jan Kara wrote:
>>
>>     
>>>  Hi,
>>>
>>>  I was just looking into reiserfs remount code and it seems it calls
>>> finish_unfinished() whenever filesystem is remounted and isn't in 
>>> read-only
>>> mode. Isn't this unnecessary? I thought finish_unfinished() is needed only
>>> when we remount from read-only to read-write state for the first time...
>>>  
>>>       
>> Perhaps, you are right (if we don't miss some points). But we need to keep 
>> a track
>> of first successful remounts to rw state, and it requires a special flag in 
>> in-memory
>> superblock (I don't see another way).
>>
>>     
>>>  Well, I wouldn't mind the performance impact to remount that much
>>>
>>>       
>> yeah,  definitely benchmarks lack "1000 remounts"statistics ;)
>>
>>     
>>> but it
>>> would make my life with journaled quota a bit easier ;). Thanks for an
>>> answer in advance.
>>>       
>   OK, so would you accept the patch below?
>
>   

I think it should work properly, however Linus doesn't like mysterious
values.
The attached patch might look better: it does the same things, but uses
status flags.

Thanks,
Edward.