Received-SPF: neutral (google.com: 212.227.17.24 is neither permitted nor denied by best guess record for domain of arnd@arndb.de) client-ip=212.227.17.24;
From: Arnd Bergmann <arnd@arndb.de>
To: y2038@lists.linaro.org
Cc: Deepa Dinamani <deepa.kernel@gmail.com>, outreachy-kernel@googlegroups.com, Y2038@lists.linaro.org
Subject: Re: [Y2038] [PATCH] staging: ft1000: replace timeval with ktime_t
Date: Sat, 17 Oct 2015 00:14:57 +0200
Message-ID: <30046181.iodvR1OlJa@wuerfel>
User-Agent: KMail/4.11.5 (Linux/3.16.0-10-generic; KDE/4.11.5; x86_64; ; )
In-Reply-To: <1444869780-26658-1-git-send-email-deepa.kernel@gmail.com>
References: <1444869780-26658-1-git-send-email-deepa.kernel@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"

On Wednesday 14 October 2015 17:43:00 Deepa Dinamani wrote:
> Replace use of struct timeval with ktime_t to calculate connection
> duration. This avoids y2038 issues. Use ktime_get monotonic clock
> instead of gettimeofday realtime clock to guard against negative
> durations. Fix pr_debug to use matching %ul instead of converting
> value to int, so that debug does not provide misleading information.
> Fix connection time conversion to use interface type unsigned long
> instead of assuming it maps to u32.
> 
> Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>

The change looks correct, and your changelog also looks reasonable.

However, I notice that the code only ever looks at the second
portion of the time stamp. Therefore it would be nicer to use
ktime_get_seconds() to avoid the expensive ktime_divns(), and to
slightly simplify the code further.

You can use either time64_t or a shorter type like 'u32' or 'long'
to store monotonic seconds, so pick one of them and explain why
you have that one.

	Arnd