[Cluster-devel] [GFS2 PATCH] GFS2: Use average srttb value in congestion calculations
Steven Whitehouse
swhiteho at redhat.com
Wed Apr 15 14:43:17 UTC 2015
Hi,
On 15/04/15 15:39, Bob Peterson wrote:
> Hi,
>
> This patch changes function gfs2_rgrp_congested so that it uses an
> average srttb (smoothed round trip time for blocking rgrp glocks)
> rather than the CPU-specific value. If we use the CPU-specific value
> it can incorrectly report no contention when there really is contention
> due to the glock processing occurring on a different CPU.
>
> Signed-off-by: Bob Peterson <rpeterso at redhat.com>
Looks like a good catch... Acked-by: Steven Whitehouse <swhiteho at redhat.com>
Steve.
> ---
> diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
> index 6af2396..f39eedc 100644
> --- a/fs/gfs2/rgrp.c
> +++ b/fs/gfs2/rgrp.c
> @@ -1850,14 +1850,19 @@ static bool gfs2_rgrp_congested(const struct gfs2_rgrpd *rgd, int loops)
> const struct gfs2_sbd *sdp = gl->gl_sbd;
> struct gfs2_lkstats *st;
> s64 r_dcount, l_dcount;
> - s64 r_srttb, l_srttb;
> + s64 l_srttb, a_srttb = 0;
> s64 srttb_diff;
> s64 sqr_diff;
> s64 var;
> + int cpu;
>
> preempt_disable();
> + for_each_present_cpu(cpu) {
> + st = &per_cpu_ptr(sdp->sd_lkstats, cpu)->lkstats[LM_TYPE_RGRP];
> + a_srttb += st->stats[GFS2_LKS_SRTTB];
> + }
> st = &this_cpu_ptr(sdp->sd_lkstats)->lkstats[LM_TYPE_RGRP];
> - r_srttb = st->stats[GFS2_LKS_SRTTB];
> + a_srttb /= num_present_cpus();
> r_dcount = st->stats[GFS2_LKS_DCOUNT];
> var = st->stats[GFS2_LKS_SRTTVARB] +
> gl->gl_stats.stats[GFS2_LKS_SRTTVARB];
> @@ -1866,10 +1871,10 @@ static bool gfs2_rgrp_congested(const struct gfs2_rgrpd *rgd, int loops)
> l_srttb = gl->gl_stats.stats[GFS2_LKS_SRTTB];
> l_dcount = gl->gl_stats.stats[GFS2_LKS_DCOUNT];
>
> - if ((l_dcount < 1) || (r_dcount < 1) || (r_srttb == 0))
> + if ((l_dcount < 1) || (r_dcount < 1) || (a_srttb == 0))
> return false;
>
> - srttb_diff = r_srttb - l_srttb;
> + srttb_diff = a_srttb - l_srttb;
> sqr_diff = srttb_diff * srttb_diff;
>
> var *= 2;
>
More information about the Cluster-devel
mailing list