[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <50343373035f4f2694aac4df86dae618@AcuMS.aculab.com>
Date: Tue, 16 Nov 2021 13:06:49 +0000
From: David Laight <David.Laight@...LAB.COM>
To: 'Thomas Gleixner' <tglx@...utronix.de>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
CC: "David S. Miller" <davem@...emloft.net>,
Voon Weifeng <weifeng.voon@...el.com>,
Wong Vee Khee <vee.khee.wong@...el.com>,
"Giuseppe Cavallaro" <peppe.cavallaro@...com>,
Alexandre Torgue <alexandre.torgue@...s.st.com>,
Jose Abreu <joabreu@...opsys.com>,
"Benedikt Spranger" <b.spranger@...utronix.de>
Subject: RE: [PATCH] net: stmmac: Fix signed/unsigned wreckage
From: Thomas Gleixner
> Sent: 15 November 2021 15:21
>
> The recent addition of timestamp correction to compensate the CDC error
> introduced a subtle signed/unsigned bug in stmmac_get_tx_hwtstamp() while
> it managed for some obscure reason to avoid that in stmmac_get_rx_hwtstamp().
>
> The issue is:
>
> s64 adjust = 0;
> u64 ns;
>
> adjust += -(2 * (NSEC_PER_SEC / priv->plat->clk_ptp_rate));
> ns += adjust;
>
> works by chance on 64bit, but falls apart on 32bit because the compiler
> knows that adjust fits into 32bit and then treats the addition as a u64 +
> u32 resulting in an off by ~2 seconds failure.
The problem is earlier.
NSEC_PER_SEC and clk_ptp_rate are almost certainly 32bit and unsigned.
So you have:
adjust = (s64)((u64)adjust + (u32)-(2 * (NSEC_PER_SEC/...)));
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
Powered by blists - more mailing lists