[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <140065e3-0368-0b5d-8a0d-afe49b741ad2@kernel.dk>
Date: Mon, 24 Jul 2023 09:50:24 -0600
From: Jens Axboe <axboe@...nel.dk>
To: Greg KH <gregkh@...uxfoundation.org>,
Phil Elwell <phil@...pberrypi.com>
Cc: andres@...razel.de, asml.silence@...il.com, david@...morbit.com,
hch@....de, io-uring@...r.kernel.org,
LKML <linux-kernel@...r.kernel.org>, linux-xfs@...r.kernel.org,
stable <stable@...r.kernel.org>
Subject: Re: [PATCH] io_uring: Use io_schedule* in cqring wait
On 7/24/23 9:48?AM, Greg KH wrote:
> On Mon, Jul 24, 2023 at 04:35:43PM +0100, Phil Elwell wrote:
>> Hi Andres,
>>
>> With this commit applied to the 6.1 and later kernels (others not
>> tested) the iowait time ("wa" field in top) in an ARM64 build running
>> on a 4 core CPU (a Raspberry Pi 4 B) increases to 25%, as if one core
>> is permanently blocked on I/O. The change can be observed after
>> installing mariadb-server (no configuration or use is required). After
>> reverting just this commit, "wa" drops to zero again.
>
> This has been discussed already:
> https://lore.kernel.org/r/12251678.O9o76ZdvQC@natalenko.name
>
> It's not a bug, mariadb does have pending I/O, so the report is correct,
> but the CPU isn't blocked at all.
Indeed - only thing I can think of is perhaps mariadb is having a
separate thread waiting on the ring in perpetuity, regardless of whether
or not it currently has IO.
But yes, this is very much ado about nothing...
--
Jens Axboe
Powered by blists - more mailing lists