Date: Sun, 23 Dec 2012 12:33:39 +0200 From: Andriy Gapon <avg@FreeBSD.org> To: =?ISO-8859-1?Q?Bernhard_Fr=F6hlich?= <decke@bluelife.at> Cc: freebsd-emulation@FreeBSD.org, vbox@FreeBSD.org Subject: Re: incorrect usage of callout_reset in vbox 4.2.4 ? Message-ID: <50D6DE03.50707@FreeBSD.org> In-Reply-To: <CAE-m3X2QzsGvw6ekGdssL1MMBQucRy9ve99M4P1BhJ4=tQBh0w@mail.gmail.com> References: <50C9D369.6040204@FreeBSD.org> <CAE-m3X2QzsGvw6ekGdssL1MMBQucRy9ve99M4P1BhJ4=tQBh0w@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
on 20/12/2012 13:14 Bernhard Fröhlich said the following: > On Thu, Dec 13, 2012 at 2:08 PM, Andriy Gapon <avg@freebsd.org> wrote: >> >> It looks like in timer-r0drv-freebsd.c the code tries to pass absolute time as a >> timeout parameter to callout_reset while that function actually expects relative >> time (period). >> >> I am testing the following patch, but I am sure that the code can be made more >> efficient. >> >> --- timer-r0drv-freebsd.c.orig 2012-12-12 20:13:27.623202784 +0200 >> +++ timer-r0drv-freebsd.c 2012-12-12 20:19:43.368202795 +0200 >> @@ -172,15 +172,16 @@ >> /* >> * Calc when it should start firing. >> */ >> - u64First += RTTimeNanoTS(); >> + const uint64_t u64Now = RTTimeNanoTS(); >> + u64First += u64Now; >> >> pTimer->fSuspended = false; >> pTimer->iTick = 0; >> pTimer->u64StartTS = u64First; >> pTimer->u64NextTS = u64First; >> >> - tv.tv_sec = u64First / 1000000000; >> - tv.tv_usec = (u64First % 1000000000) / 1000; >> + tv.tv_sec = (u64First - u64Now) / 1000000000; >> + tv.tv_usec = ((u64First - u64Now) % 1000000000) / 1000; >> callout_reset(&pTimer->Callout, tvtohz(&tv), rtTimerFreeBSDCallback, pTimer); >> >> return VINF_SUCCESS; >> @@ -247,8 +248,8 @@ >> if (pTimer->u64NextTS < u64NanoTS) >> pTimer->u64NextTS = u64NanoTS + RTTimerGetSystemGranularity() / 2; >> >> - tv.tv_sec = pTimer->u64NextTS / 1000000000; >> - tv.tv_usec = (pTimer->u64NextTS % 1000000000) / 1000; >> + tv.tv_sec = (pTimer->u64NextTS - u64NanoTS) / 1000000000; >> + tv.tv_usec = ((pTimer->u64NextTS - u64NanoTS) % 1000000000) / 1000; >> callout_reset(&pTimer->Callout, tvtohz(&tv), rtTimerFreeBSDCallback, pTimer); >> } > > What is your results from that tests? Is the patch correct so should we include > it into the port? I believe that the patch is correct. I haven't seen any regressions nor any visible improvements from it. -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?50D6DE03.50707>