Received-SPF: pass (google.com: domain of linux-kernel+bounces-45039-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
Mime-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset=UTF-8
Date: Tue, 30 Jan 2024 19:27:44 +0200
Message-Id: <CYS7UOQQ3L7P.2V7HDL1U2VOD4@suppilovahvero>
Cc: <keyrings@vger.kernel.org>, <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v3] keys: update key quotas in key_put()
From: "Jarkko Sakkinen" <jarkko@kernel.org>
To: "Luis Henriques" <lhenriques@suse.de>, "David Howells"
 <dhowells@redhat.com>, "Eric Biggers" <ebiggers@kernel.org>
References: <20240130101344.28936-1-lhenriques@suse.de>
In-Reply-To: <20240130101344.28936-1-lhenriques@suse.de>

On Tue Jan 30, 2024 at 12:13 PM EET, Luis Henriques wrote:
> Delaying key quotas update when key's refcount reaches 0 in key_put() has
> been causing some issues in fscrypt testing, specifically in fstest
> generic/581.  This commit fixes this test flakiness by dealing with the
> quotas immediately, and leaving all the other clean-ups to the key garbag=
e
> collector.
>
> This is done by moving the updates to the qnkeys and qnbytes fields in
> struct key_user from key_gc_unused_keys() into key_put().  Unfortunately,
> this also means that we need to switch to the irq-version of the spinlock
> that protects these fields and use spin_lock_{irqsave,irqrestore} in all =
the
> code that touches these fields.
>
> Signed-off-by: Luis Henriques <lhenriques@suse.de>

OK this is great. I mean in this commit it is pretty essentiual to
document that there is an ownership change. Such changes have by
far the biggest impact to kernel semantics, and thus very useful
to mark such commits for e.g. bisection.

Reviewed-by: Jarkko Sakkinen <jarkko@kernel.org>

BR, Jarkko