[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAHk-=wgzJf_UTORN+j-Bn-faWBgM=oNgj9feUu-d7F6BqK5hdw@mail.gmail.com>
Date: Thu, 24 Dec 2020 13:57:59 -0800
From: Linus Torvalds <torvalds@...ux-foundation.org>
To: Thomas Gleixner <tglx@...utronix.de>,
Eran Ben Elisha <eranbe@...dia.com>,
Saeed Mahameed <saeedm@...dia.com>
Cc: Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
"the arch/x86 maintainers" <x86@...nel.org>,
linux-rdma <linux-rdma@...r.kernel.org>
Subject: Re: [GIT pull] irq/core for v5.11-rc1
On Tue, Dec 22, 2020 at 3:58 PM Thomas Gleixner <tglx@...utronix.de> wrote:
>
> A treewide cleanup of interrupt descriptor (ab)use with all sorts of racy
> accesses, inefficient and disfunctional code. The goal is to remove the
> export of irq_to_desc() to prevent these things from creeping up again.
This exposed that the mellanox mlx5 driver had grown another use of
'irq_desc' in commit 145e5637d941 ("net/mlx5e: Add TX PTP port object
support").
Except that use was write-only, with the only acces being this assignment:
c->irq_desc = irq_to_desc(irq);
so I removed the 'irq_desc' field and that assignment in my merge.
I didn't even replace it with an affinity mask thing like commit
6e745db4ddd0 ("net/mlx5: Replace irq_to_desc() abuse") did for the
previous case, because the irq_desc thing was literally pointless and
unused. I suspect it was some cut-and-paste programming.
Just writing this note so that everybody knows what's going on.
Linus
Powered by blists - more mailing lists