[f2fs-dev] f2fs: cp_mutex starvation issue

[f2fs-dev] f2fs: cp_mutex starvation issue

[stummala]

Hi Chao,

Can we consider to move the sbi->cp_mutex as well to use rw_sem similar 
to this patch of yours -
https://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs.git/commit/fs/f2fs?id=fb24fea75ca5ceef59f753494b2efd453606e08a

We are observing issues where fsync() of a high priority task getting 
starved when there is heavy IO from several low priority tasks.
I would like to know if this was considered before and if any issues in 
doing so? Please share your inputs.

Thanks,
Sahitya.


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Re: [f2fs-dev] f2fs: cp_mutex starvation issue

[Chao Yu]

Hi Sahitya,

On 2020/11/18 17:47, stummala@codeaurora.org wrote:
> Hi Chao,
> 
> Can we consider to move the sbi->cp_mutex as well to use rw_sem similar
> to this patch of yours -
> https://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs.git/commit/fs/f2fs?id=fb24fea75ca5ceef59f753494b2efd453606e08a
> 
> We are observing issues where fsync() of a high priority task getting
> starved when there is heavy IO from several low priority tasks.
> I would like to know if this was considered before and if any issues in
> doing so? Please share your inputs.

I missed to check cp_mutex lock, because I haven't been reported with
starving issue on this lock, anyway, I agree to do the same change on
cp_mutex lock, and I don't think there is any problem after change.

Could you please make one patch to fix this?

Thanks,

> 
> Thanks,
> Sahitya.
> .
> 


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel

Re: [f2fs-dev] f2fs: cp_mutex starvation issue

[stummala]

Hi Chao,

On 2020-11-18 15:37, Chao Yu wrote:
> Hi Sahitya,
> 
> On 2020/11/18 17:47, stummala@codeaurora.org wrote:
>> Hi Chao,
>> 
>> Can we consider to move the sbi->cp_mutex as well to use rw_sem 
>> similar
>> to this patch of yours -
>> https://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs.git/commit/fs/f2fs?id=fb24fea75ca5ceef59f753494b2efd453606e08a
>> 
>> We are observing issues where fsync() of a high priority task getting
>> starved when there is heavy IO from several low priority tasks.
>> I would like to know if this was considered before and if any issues 
>> in
>> doing so? Please share your inputs.
> 
> I missed to check cp_mutex lock, because I haven't been reported with
> starving issue on this lock, anyway, I agree to do the same change on
> cp_mutex lock, and I don't think there is any problem after change.
> 
> Could you please make one patch to fix this?

Thanks for the comments.
Sure, I will put up a patch for this.

Thanks,

> Thanks,
> 
>> 
>> Thanks,
>> Sahitya.
>> .
>> 


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel