Re: [PATCH 00/13] bcache: fixes and update for 4.14

[Jens Axboe <axboe@kernel.dk>]

On 09/07/2017 12:51 PM, Eddie Chapman wrote:
> On 06/09/17 18:38, Coly Li wrote:
>> On 2017/9/6 下午11:46, Jens Axboe wrote:
>>> On 09/06/2017 09:41 AM, Coly Li wrote:
>>>> On 2017/9/6 下午10:20, Jens Axboe wrote:
>>>>> On Wed, Sep 06 2017, Coly Li wrote:
>>>>>> Hi Jens,
>>>>>>
>>>>>> Here are 12 patchs for bcache fixes and updates, most of them were posted
>>>>>> by Eric Wheeler in 4.13 merge window, but delayed due to lack of code
>>>>>> review.
>>>>>
>>>>> Next time, please send this _before_ the merge window opens. Not a huge
>>>>> problem for this series, since it has been posted numerous times in the
>>>>> past and already has good review coverage, but it really should have
>>>>> been in linux-next for a week or two before heading upstream.
>>>>>
>>>> Hi Jens,
>>>>
>>>> Copied, send patches _before_ merge window opens.
>>>>
>>>> But could you please to give me a hint, how can I find when a specific
>>>> merge window will open ? I search LKML, and see people send pull
>>>> requests for 4.14 merge window, but I don't see any announcement for
>>>> 4.14 merge window time slot.
>>>
>>> The merge window opens when Linus releases the previous kernel. So you
>>> have to try and keep track of when he expects to do that. That really
>>> isn't that difficult - Linus always cuts a release on Sundays. We always
>>> have 7 or 8 rc releases, so a good time to check is his mail on -rc6 or
>>> -rc7 release, that usually gives a good indication of when he expects to
>>> release the final.
>>>
>>> To keep things simple, if you always have things ready when -rc7 is
>>> released, then you are at least one week out from the merge window,
>>> possibly two if we end up doing an -rc8. That means you don't have to
>>> track anything, you know the exact date of when -rc7 is released since
>>> Linus's schedule is usually reliable.
>>>
>>
>> Hi Jens,
>>
>> Copied, I will follow the above rule in next cycle.
>>
>> And I just post the last patch
>> (0013-bcache-initialize-stripe_sectors_dirty-correctly-for.patch) to you
>> for 4.14 cycle.
>>
>> This patch is an important fix for bcache writeback rate calculation, it
>> is depended by another patch I sent to you
>> (0006-bcache-correct-cache_dirty_target-in-__update_writeb.patch),
>> please have it in 4.14.
>>
>> Thanks in advance.
> 
> Hello Jens,
> 
> FWIW I'd like to support Coly by reporting that patches 0001, 0002, 0006 
> and 0013 (the last one mentioned above) have been tested by me on 2 
> production servers with a good deal of bcache activity for the last 6 
> weeks without any issues having arisen.
> 
> But this may not be much use to you to know, as these are running 
> vanilla kernel.org kernel 4.4.77.  Still, these 4 patches exactly as 
> Coly has sent them to you apply without any changes to the code, so I 
> guess it vouches for the code at least somewhat.

That's all good and fine, and it's better than no testing at all. But it
doesn't change the fact that anyone using bcache on -next would have
been using the changes for a while before release, instead of just one
or two people. Additionally, your base is drastically different than
mainline, since you're running a kernel that's a year and a half old.

Next time I'm not adding anything post merge window open, unless it's
addressing a problem that was added in the merge window.

-- 
Jens Axboe