Re: [PATCH 1/2] tracing: Simplify & fix saved_tgids logic

From: Steven Rostedt <rostedt@goodmis.org>
To: Paul Burton <paulburton@google.com>
Cc: Joel Fernandes <joelaf@google.com>,
	linux-kernel@vger.kernel.org, Ingo Molnar <mingo@redhat.com>,
	stable@vger.kernel.org
Subject: Re: [PATCH 1/2] tracing: Simplify & fix saved_tgids logic
Date: Thu, 1 Jul 2021 14:07:54 -0400	[thread overview]
Message-ID: <20210701140754.5847a50f@oasis.local.home> (raw)
In-Reply-To: <YN38D3dg0fLzL0Ia@google.com>

On Thu, 1 Jul 2021 10:31:59 -0700
Paul Burton <paulburton@google.com> wrote:

> I was tempted to just add the redundant checks anyway (pick your battles
> and all) but for show() in particular it wound up making things seem
> non-sensical to me ("display the value describing this non-NULL pointer
> into tgid_map only if tgid_map is not NULL?").

I agree with your assessment, and will actually take your first patch,
as I don't think the comment is that helpful, not to mention, we don't
use '//' comments in the kernel, so that would have to be changed.

But for cases like this, I usually have something like:

	if (WARN_ON_ONCE(!tgid_map))
		return -1;

Because the logic is what makes tgid_map not being NULL, but as
experience has taught me, the logic can sometimes be mistaken, at least
as time goes by. And things that are protected by logic, deserve a
WARN*() when it doesn't go as planned.

We can always add that later, if needed.

-- Steve