Messages in this thread | | | Date | Thu, 22 Aug 2019 11:09:48 +0800 | From | Ming Lei <> | Subject | Re: [PATCH V6 2/2] genirq/affinity: Spread vectors on node according to nr_cpu ratio |
| |
On Mon, Aug 19, 2019 at 04:02:21PM +0200, Thomas Gleixner wrote: > On Mon, 19 Aug 2019, Ming Lei wrote: > > On Mon, Aug 19, 2019 at 03:13:58PM +0200, Thomas Gleixner wrote: > > > On Mon, 19 Aug 2019, Ming Lei wrote: > > > > > > > Cc: Jon Derrick <jonathan.derrick@intel.com> > > > > Cc: Jens Axboe <axboe@kernel.dk> > > > > Reported-by: Jon Derrick <jonathan.derrick@intel.com> > > > > Reviewed-by: Jon Derrick <jonathan.derrick@intel.com> > > > > Reviewed-by: Keith Busch <kbusch@kernel.org> > > > > > > This version is sufficiently different from the previous one, so I do not > > > consider the reviewed-by tags still being valid and meaningful. Don't > > > include them unless you just do cosmetic changes. > > > > Fine. > > > > However, the V6 only change isn't big, just for addressing the un-initialized > > warning, and the change is only done on function of irq_build_affinity_masks(). > > They are not trivial either: > > affinity.c | 28 +++++++++++++--------------- > 1 file changed, 13 insertions(+), 15 deletions(-) > > --- a/kernel/irq/affinity.c > +++ b/kernel/irq/affinity.c > @@ -339,7 +339,7 @@ static int irq_build_affinity_masks(unsi > unsigned int firstvec, > struct irq_affinity_desc *masks) > { > - unsigned int curvec = startvec, nr_present, nr_others; > + unsigned int curvec = startvec, nr_present = 0, nr_others = 0; > cpumask_var_t *node_to_cpumask; > cpumask_var_t nmsk, npresmsk; > int ret = -ENOMEM; > @@ -354,19 +354,17 @@ static int irq_build_affinity_masks(unsi > if (!node_to_cpumask) > goto fail_npresmsk; > > - ret = 0; > /* Stabilize the cpumasks */ > get_online_cpus(); > build_node_to_cpumask(node_to_cpumask); > > /* Spread on present CPUs starting from affd->pre_vectors */ > - nr_present = __irq_build_affinity_masks(curvec, numvecs, > - firstvec, node_to_cpumask, > - cpu_present_mask, nmsk, masks); > - if (nr_present < 0) { > - ret = nr_present; > + ret = __irq_build_affinity_masks(curvec, numvecs, firstvec, > + node_to_cpumask, cpu_present_mask, > + nmsk, masks); > + if (ret < 0) > goto fail_build_affinity; > - } > + nr_present = ret; > > /* > * Spread on non present CPUs starting from the next vector to be > @@ -379,16 +377,16 @@ static int irq_build_affinity_masks(unsi > else > curvec = firstvec + nr_present; > cpumask_andnot(npresmsk, cpu_possible_mask, cpu_present_mask); > - nr_others = __irq_build_affinity_masks(curvec, numvecs, > - firstvec, node_to_cpumask, > - npresmsk, nmsk, masks); > - if (nr_others < 0) > - ret = nr_others; > + ret = __irq_build_affinity_masks(curvec, numvecs, firstvec, > + node_to_cpumask, npresmsk, nmsk, > + masks); > + if (ret >= 0) > + nr_others = ret; > > fail_build_affinity: > put_online_cpus(); > > - if (min(nr_present, nr_others) >= 0) > + if (ret >= 0) > WARN_ON(nr_present + nr_others < numvecs); > > free_node_to_cpumask(node_to_cpumask); > @@ -398,7 +396,7 @@ static int irq_build_affinity_masks(unsi > > fail_nmsk: > free_cpumask_var(nmsk); > - return ret; > + return ret < 0 ? ret : 0; > } > > static void default_calc_sets(struct irq_affinity *affd, unsigned int affvecs)
Hi Keith & Jon,
Could you review the above V6 extra change so that we can move on?
BTW, in-balanced numa nodes can be made easily via passing 'possible_cpus=N'.
Thanks, Ming
| |