Received-SPF: pass (google.com: domain of linux-ext4+bounces-665-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
From: bugzilla-daemon@kernel.org
To: linux-ext4@vger.kernel.org
Subject: [Bug 217965] ext4(?) regression since 6.5.0 on sata hdd
Date: Thu, 04 Jan 2024 05:38:36 +0000
Message-ID: <bug-217965-13602-oNiK3hL8DX@https.bugzilla.kernel.org/>
In-Reply-To: <bug-217965-13602@https.bugzilla.kernel.org/>
References: <bug-217965-13602@https.bugzilla.kernel.org/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Auto-Submitted: auto-generated
Precedence: bulk
MIME-Version: 1.0

https://bugzilla.kernel.org/show_bug.cgi?id=3D217965

--- Comment #61 from Ojaswin Mujoo (ojaswin.mujoo@ibm.com) ---
Hi Matthew, thanks for confirming. So as pointed out in comment 9 [1], for =
the
above steps to disable CR_BEST_AVAIL_LEN code did not fix the issue. My
suspicion is that this issue can occur either in CR_GOAL_LEN_FAST or
CR_BEST_AVAIL_LEN lookup, depending on the block groups being searched.
Probably for you, it was occurring during CR_BEST_AVAIL_LEN lookup and hence
disabling that code fixed it.=20

Further, as Carlos pointed out above, they are able to see this in all 6.*
kernels which means this is happening before CR_BEST_AVAIL_LEN was introduc=
ed
however it seems to be much easier to trigger in 6.5+ kernels. Now, assuming
the above theory is correct, then in cases where this is triggered from
CR_GOAL_LEN_FAST, it should ideally happen easily in pre 6.5 kernels as wel=
l,
but it doesn't, which makes me think that there might be some other related
changes in 6.5 that might be making it easier to trigger.

I'll try to play around a bit more with this. Also, as for higher CPU usage,
how high are we talking about? So CR_BEST_AVAIL_LEN does add some extra cyc=
les
at the cost of generally faster allocation in fragmented filesystems, howev=
er
since you have disabled it we shouldn't ideally be seeing it. Also, does the
CPU util consistently drop when you commented out that code?

[1] https://bugzilla.kernel.org/show_bug.cgi?id=3D217965#c9

--=20
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.=