Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 03 Mar 2018 12:14:42 -0700
From:      Ian Lepore <ian@freebsd.org>
To:        Hans Petter Selasky <hselasky@FreeBSD.org>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   Re: svn commit: r330352 - in head/sys/compat/linuxkpi/common: include/linux src
Message-ID:  <1520104482.23690.15.camel@freebsd.org>
In-Reply-To: <201803031854.w23IsHxm086835@repo.freebsd.org>
References:  <201803031854.w23IsHxm086835@repo.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 2018-03-03 at 18:54 +0000, Hans Petter Selasky wrote:
> Author: hselasky
> Date: Sat Mar  3 18:54:16 2018
> New Revision: 330352
> URL: https://svnweb.freebsd.org/changeset/base/330352
> 
> Log:
>   Implement msleep_interruptible() in the LinuxKPI. While at it use pause_sbt()
>   instead of pause() in the msleep() function to avoid rounding errors when
>   converting delay values forth and back. Add a guard for a delay value
>   of zero milliseconds which is undefined.
>   
>   MFC after:	1 week
>   Requested by:	Johannes Lundberg <johalun0@gmail.com>
>   Sponsored by:	Mellanox Technologies
> 
> Modified:
>   head/sys/compat/linuxkpi/common/include/linux/delay.h
>   head/sys/compat/linuxkpi/common/src/linux_schedule.c
> 
> Modified: head/sys/compat/linuxkpi/common/include/linux/delay.h
> ==============================================================================
> --- head/sys/compat/linuxkpi/common/include/linux/delay.h	Sat Mar  3 18:52:19 2018	(r330351)
> +++ head/sys/compat/linuxkpi/common/include/linux/delay.h	Sat Mar  3 18:54:16 2018	(r330352)
> @@ -36,14 +36,20 @@
>  #include 
>  
>  static inline void
> -linux_msleep(int ms)
> +linux_msleep(unsigned int ms)
>  {
> -	pause("lnxsleep", msecs_to_jiffies(ms));
> +	/* guard against invalid values */
> +	if (ms == 0)
> +		ms = 1;
> +	pause_sbt("lnxsleep", SBT_1MS * ms, 0, C_HARDCLOCK);
>  }
>  

To really avoid roundoff errors, use mstosbt(ms) instead of SBT_1MS*ms.

The rounding errors when multiplying or dividing by SBT_1MS is not
severe like it is with SBT_1NS, but the inline functions still do the
arithmetic more correctly to avoid roundoff error.

-- Ian



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1520104482.23690.15.camel>