Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <9db9d9ae751a50c9fd1da1ee3bd4c564546ce1c5.camel@sipsolutions.net>
Subject: Re: systemd-rfkill regression on 5.11 and later kernels
From:   Johannes Berg <johannes@sipsolutions.net>
To:     Takashi Iwai <tiwai@suse.de>
Cc:     Emmanuel Grumbach <emmanuel.grumbach@intel.com>,
        linux-wireless@vger.kernel.org, linux-kernel@vger.kernel.org
Date:   Thu, 18 Mar 2021 12:16:54 +0100
In-Reply-To: <s5hczvwiumf.wl-tiwai@suse.de>
References: <s5ha6r0kgt5.wl-tiwai@suse.de>
         <54859a03b8789a2800596067e06c8adb49a107f5.camel@sipsolutions.net>
         <s5ho8fgixl9.wl-tiwai@suse.de>
         <c196f9cb7ba2487fb5aceceedf860cc24c6843f2.camel@sipsolutions.net>
         <s5hczvwiumf.wl-tiwai@suse.de>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.38.4 (3.38.4-1.fc33) 
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Thu, 2021-03-18 at 12:11 +0100, Takashi Iwai wrote:
> > That said, we can "fix" this like this, and hope we'll not get this
> > again? And if we do get it again ... well, we keep renaming the structs
> > and add "struct rfkill_event_v3" next time?
> 
> Yeah, that's a dilemma.  An oft-seen trick is to add more bytes for
> the future use, e.g. extend to 16 bytes and fill 0 for the remaining.

Yeah, I guess I could stick a reserved[15] there, it's small enough.

> In the sound driver, we introduced an ioctl to inform from user-space
> which API protocol it can speak, and the kernel falls back to the old
> API/ABI if it's a lower version or it's not told at all.  But I'm not
> sure whether such an implementation is optimal for rfkill.

I thought about it, but it ... doesn't really help.

Somebody's going to do

	ioctl(..., sizeof(ev)) == sizeof(ev)

and break on older kernels, or == my_fixed_size, or ... something. It's
not really going to address the issue entirely.

And it's more complexity.

johannes