Re: squashfs performance regression and readahea

Re: squashfs performance regression and readahea

[Phillip Lougher]

On 10/05/2022 03:35, Matthew Wilcox wrote:
> On Tue, May 10, 2022 at 02:11:41AM +0100, Phillip Lougher wrote:
>> On 09/05/2022 14:21, Matthew Wilcox wrote:
>>> On Mon, May 09, 2022 at 08:43:45PM +0800, Xiongwei Song wrote:
>>>> Hi Hsin-Yi and Matthew,
>>>>
>>>> With the patch from the attachment on linux 5.10, ran the command as I
>>>> mentioned earlier,
>>>> got the results below:
>>>> 1:40.65 (1m + 40.65s)
>>>> 1:10.12
>>>> 1:11.10
>>>> 1:11.47
>>>> 1:11.59
>>>> 1:11.94
>>>> 1:11.86
>>>> 1:12.04
>>>> 1:12.21
>>>> 1:12.06
>>>>
>>>> The performance has improved obviously, but compared to linux 4.18, the
>>>> performance is not so good.
>>>>
>>>> Moreover, I wanted to test on linux 5.18. But I think I should revert
>>>> 9eec1d897139 ("squashfs: provide backing_dev_info in order to disable
>>>> read-ahead"),
>>>> right?  Otherwise, the patch doesn't work?
>>>
>>> I've never seen patch 9eec1d897139 before.  If you're going to point
>>> out bugs in my code, at least have the decency to cc me on it.  It
>>> should never have gone in, and should be reverted so the problem can
>>> be fixed properly.
>>
>> You are not in charge of what patches goes into Squashfs, that is my
>> perogative as maintainer of Squashfs.
> 
> I think you mean 'prerogative'.  And, no, your filesystem is not your
> little fiefdom, it's part of a collaborative effort.
> 

This isn't a spelling contest, and if that's the best you can do you
have already failed.

Be carefull here also, I have been maintainer of Squashfs for 20 years,
and was kernel maintainer for both Ubuntu and Redhat for 10 years, and
so I am experienced member of the community.

You reply is bordering on offensive and arrogant, especially considering
it is unwarranted.  I did not set out to offend you, and I don't
appreciate it.

About 8 years ago I decided to refrain from active involvement in the
kernel community, because I decided the level of discourse was
disgusting, and I had enough of it.

I poped up now to defend my approval of the Huawei patch.  I am *quite*
happy not to have any more involvement until necessary.

So having said what I want to say, I will leave it at that. You have
just proved why I have minimised my involvement.

No doubt you'll throw your toys out the pram, but, I'm no
longer listening so don't bother.


>> That patch (by Huawei) fixes the performance regression in Squashfs
>> by disabling readahead, and it is good workaround until something
>> better.
> 
> You *didn't even report the problem to me*.  How can it be fixed if I'm
> not aware of it?
> 

There was a email discussion last year, which I responded to, and got
ignored.  I will find it out tomorrow, perhaps.  But I will probably
not bother, because life is too short.

Cheers

Phillip

Re: squashfs performance regression and readahea

[Phillip Lougher]

On 10/05/2022 04:20, Phillip Lougher wrote:
> On 10/05/2022 03:35, Matthew Wilcox wrote:
>> On Tue, May 10, 2022 at 02:11:41AM +0100, Phillip Lougher wrote:
>>> On 09/05/2022 14:21, Matthew Wilcox wrote:
>>>> On Mon, May 09, 2022 at 08:43:45PM +0800, Xiongwei Song wrote:
>>>>> Hi Hsin-Yi and Matthew,
>>>>>
>>>>> With the patch from the attachment on linux 5.10, ran the command as I
>>>>> mentioned earlier,
>>>>> got the results below:
>>>>> 1:40.65 (1m + 40.65s)
>>>>> 1:10.12
>>>>> 1:11.10
>>>>> 1:11.47
>>>>> 1:11.59
>>>>> 1:11.94
>>>>> 1:11.86
>>>>> 1:12.04
>>>>> 1:12.21
>>>>> 1:12.06
>>>>>
>>>>> The performance has improved obviously, but compared to linux 4.18, 
>>>>> the
>>>>> performance is not so good.
>>>>>
>>>>> Moreover, I wanted to test on linux 5.18. But I think I should revert
>>>>> 9eec1d897139 ("squashfs: provide backing_dev_info in order to disable
>>>>> read-ahead"),
>>>>> right?  Otherwise, the patch doesn't work?
>>>>
>>>> I've never seen patch 9eec1d897139 before.  If you're going to point
>>>> out bugs in my code, at least have the decency to cc me on it.  It
>>>> should never have gone in, and should be reverted so the problem can
>>>> be fixed properly.
>>>
>>> You are not in charge of what patches goes into Squashfs, that is my
>>> perogative as maintainer of Squashfs.
>>
>> I think you mean 'prerogative'.  And, no, your filesystem is not your
>> little fiefdom, it's part of a collaborative effort.
>>
> 
> This isn't a spelling contest, and if that's the best you can do you
> have already failed.
> 
> Be carefull here also, I have been maintainer of Squashfs for 20 years,
> and was kernel maintainer for both Ubuntu and Redhat for 10 years, and
> so I am experienced member of the community.
> 
> You reply is bordering on offensive and arrogant, especially considering
> it is unwarranted.  I did not set out to offend you, and I don't
> appreciate it.
> 
> About 8 years ago I decided to refrain from active involvement in the
> kernel community, because I decided the level of discourse was
> disgusting, and I had enough of it.
> 
> I poped up now to defend my approval of the Huawei patch.  I am *quite*
> happy not to have any more involvement until necessary.
> 
> So having said what I want to say, I will leave it at that. You have
> just proved why I have minimised my involvement.
> 
> No doubt you'll throw your toys out the pram, but, I'm no
> longer listening so don't bother.
> 
> 
>>> That patch (by Huawei) fixes the performance regression in Squashfs
>>> by disabling readahead, and it is good workaround until something
>>> better.
>>
>> You *didn't even report the problem to me*.  How can it be fixed if I'm
>> not aware of it?

Despite having been insulted, I have done your homework for you.

This is where the problem was raised last year, with you directly
emailed.

https://lore.kernel.org/all/CAJMQK-g9G6KQmH-V=BRGX0swZji9Wxe_2c7ht-MMAapdFy2pXw@mail.gmail.com/T/

>>
> 
> There was a email discussion last year, which I responded to, and got
> ignored.  I will find it out tomorrow, perhaps.  But I will probably
> not bother, because life is too short.
> 

Afterwards you started a thread on "Readahead for compressed data",
which I responded to.

https://lore.kernel.org/all/YXHK5HrQpJu9oy8w@casper.infradead.org/T/


> Cheers
> 
> Phillip