Message ID | 20130704030506.GA2396@iris.ozlabs.ibm.com (mailing list archive) |
---|---|
State | Superseded |
Headers | show |
On 07/04/2013 08:35 AM, Paul Mackerras wrote: > Some systems have an ibm,chip-id property in the cpu nodes in the > device tree. On these systems, we now use that to compute the > cpu_core_mask (i.e. the set of core siblings) rather than looking > at cache properties. > Paul, I wanted to test this patch but not able to apply this patch on top of Linux tree. Looks like I'm missing traverse_core_siblings() related patch. I searched in ppc mailing list and couldn't figure out. Can you point me to related patch? -Vasant > Signed-off-by: Paul Mackerras <paulus@samba.org> > --- > diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c > index b72d8c9..3b7a118 100644 > --- a/arch/powerpc/kernel/smp.c > +++ b/arch/powerpc/kernel/smp.c > @@ -587,6 +587,32 @@ int cpu_first_thread_of_core(int core) > } > EXPORT_SYMBOL_GPL(cpu_first_thread_of_core); > > +static void traverse_siblings_chip_id(int cpu, int add, int chipid) > +{ > + const struct cpumask *mask; > + struct device_node *np; > + int i, plen; > + const int *prop; > + > + mask = add ? cpu_online_mask : cpu_present_mask; > + for_each_cpu(i, mask) { > + np = of_get_cpu_node(i, NULL); > + if (!np) > + continue; > + prop = of_get_property(np, "ibm,chip-id", &plen); > + if (prop && plen == sizeof(int) && *prop == chipid) { > + if (add) { > + cpumask_set_cpu(cpu, cpu_core_mask(i)); > + cpumask_set_cpu(i, cpu_core_mask(cpu)); > + } else { > + cpumask_clear_cpu(cpu, cpu_core_mask(i)); > + cpumask_clear_cpu(i, cpu_core_mask(cpu)); > + } > + } > + of_node_put(np); > + } > +} > + > /* Must be called when no change can occur to cpu_present_mask, > * i.e. during cpu online or offline. > */ > @@ -611,14 +637,29 @@ static struct device_node *cpu_to_l2cache(int cpu) > > static void traverse_core_siblings(int cpu, int add) > { > - struct device_node *l2_cache; > + struct device_node *l2_cache, *np; > const struct cpumask *mask; > - int i; > + int i, chip, plen; > + const int *prop; > + > + /* First see if we have ibm,chip-id properties in cpu nodes */ > + np = of_get_cpu_node(cpu, NULL); > + if (np) { > + chip = -1; > + prop = of_get_property(np, "ibm,chip-id", &plen); > + if (prop && plen == sizeof(int)) > + chip = *(int *)prop; > + of_node_put(np); > + if (chip >= 0) { > + traverse_siblings_chip_id(cpu, add, chip); > + return; > + } > + } > > l2_cache = cpu_to_l2cache(cpu); > mask = add ? cpu_online_mask : cpu_present_mask; > for_each_cpu(i, mask) { > - struct device_node *np = cpu_to_l2cache(i); > + np = cpu_to_l2cache(i); > if (!np) > continue; > if (np == l2_cache) { > _______________________________________________ > Linuxppc-dev mailing list > Linuxppc-dev@lists.ozlabs.org > https://lists.ozlabs.org/listinfo/linuxppc-dev >
On Fri, Aug 09, 2013 at 12:52:12PM +0530, Vasant Hegde wrote: > Paul, > > I wanted to test this patch but not able to apply this patch on top > of Linux tree. Looks like I'm missing traverse_core_siblings() > related patch. I searched in ppc mailing list and couldn't figure > out. Oops, my fault, there were actually 2 patches in that series but somehow I only posted the second one. I'll repost both patches. Paul.
diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c index b72d8c9..3b7a118 100644 --- a/arch/powerpc/kernel/smp.c +++ b/arch/powerpc/kernel/smp.c @@ -587,6 +587,32 @@ int cpu_first_thread_of_core(int core) } EXPORT_SYMBOL_GPL(cpu_first_thread_of_core); +static void traverse_siblings_chip_id(int cpu, int add, int chipid) +{ + const struct cpumask *mask; + struct device_node *np; + int i, plen; + const int *prop; + + mask = add ? cpu_online_mask : cpu_present_mask; + for_each_cpu(i, mask) { + np = of_get_cpu_node(i, NULL); + if (!np) + continue; + prop = of_get_property(np, "ibm,chip-id", &plen); + if (prop && plen == sizeof(int) && *prop == chipid) { + if (add) { + cpumask_set_cpu(cpu, cpu_core_mask(i)); + cpumask_set_cpu(i, cpu_core_mask(cpu)); + } else { + cpumask_clear_cpu(cpu, cpu_core_mask(i)); + cpumask_clear_cpu(i, cpu_core_mask(cpu)); + } + } + of_node_put(np); + } +} + /* Must be called when no change can occur to cpu_present_mask, * i.e. during cpu online or offline. */ @@ -611,14 +637,29 @@ static struct device_node *cpu_to_l2cache(int cpu) static void traverse_core_siblings(int cpu, int add) { - struct device_node *l2_cache; + struct device_node *l2_cache, *np; const struct cpumask *mask; - int i; + int i, chip, plen; + const int *prop; + + /* First see if we have ibm,chip-id properties in cpu nodes */ + np = of_get_cpu_node(cpu, NULL); + if (np) { + chip = -1; + prop = of_get_property(np, "ibm,chip-id", &plen); + if (prop && plen == sizeof(int)) + chip = *(int *)prop; + of_node_put(np); + if (chip >= 0) { + traverse_siblings_chip_id(cpu, add, chip); + return; + } + } l2_cache = cpu_to_l2cache(cpu); mask = add ? cpu_online_mask : cpu_present_mask; for_each_cpu(i, mask) { - struct device_node *np = cpu_to_l2cache(i); + np = cpu_to_l2cache(i); if (!np) continue; if (np == l2_cache) {
Some systems have an ibm,chip-id property in the cpu nodes in the device tree. On these systems, we now use that to compute the cpu_core_mask (i.e. the set of core siblings) rather than looking at cache properties. Signed-off-by: Paul Mackerras <paulus@samba.org> ---