Hi Greg,
did you have a chance to have a look at my replies below?
It would be really helpful to understand from you how to move forward
with this set.
Best Regards,
- Juri
On 13/02/17 15:09, Juri Lelli wrote:
> Hi Greg,
>
> On 10/02/17 15:28, Greg KH wrote:
> > On Thu, Feb 09, 2017 at 09:25:22AM +0000, Juri Lelli wrote:
> > > arm and arm64 share lot of code relative to parsing CPU capacity
> > > information from DT, using that information for appropriate scaling and
> > > exposing a sysfs interface for chaging such values at runtime.
> > >
> > > Factorize such code in a common place (driver/base/arch_topology.c) in
> > > preparation for further additions.
> > >
> > > Suggested-by: Will Deacon <[email protected]>
> > > Suggested-by: Mark Rutland <[email protected]>
> > > Suggested-by: Catalin Marinas <[email protected]>
> > > Cc: Russell King <[email protected]>
> > > Cc: Catalin Marinas <[email protected]>
> > > Cc: Will Deacon <[email protected]>
> > > Cc: Greg Kroah-Hartman <[email protected]>
> > > Signed-off-by: Juri Lelli <[email protected]>
> > > ---
> > >
> > > Changes from v1:
> > > - keep the original GPLv2 header
> > > ---
> > > arch/arm/Kconfig | 1 +
> > > arch/arm/kernel/topology.c | 213 ++------------------------------------
> > > arch/arm64/Kconfig | 1 +
> > > arch/arm64/kernel/topology.c | 219 +--------------------------------------
> > > drivers/base/Kconfig | 8 ++
> > > drivers/base/Makefile | 1 +
> > > drivers/base/arch_topology.c | 237 +++++++++++++++++++++++++++++++++++++++++++
> > > 7 files changed, 257 insertions(+), 423 deletions(-)
> > > create mode 100644 drivers/base/arch_topology.c
> >
> > Ah, so you want _me_ to maintain this, ok, I better review it...
> >
>
> This has been suggested as a possible way to stop replicating code between arm
> and arm64 (and possibly other archs in the future). Are you in principle OK
> with it?
>
> Thanks a lot for your comments, please find my answers below.
>
> > > --- a/drivers/base/Kconfig
> > > +++ b/drivers/base/Kconfig
> > > @@ -339,4 +339,12 @@ config CMA_ALIGNMENT
> > >
> > > endif
> > >
> > > +config GENERIC_ARCH_TOPOLOGY
> > > + bool
> > > + help
> > > + Enable support for architectures common topology code: e.g., parsing
> > > + CPU capacity information from DT, usage of such information for
> > > + appropriate scaling, sysfs interface for changing capacity values at
> > > + runtime.
> >
> > Mix of spaces and tabs :(
> >
>
> Argh. :(
>
> > > +
> > > endmenu
> > > diff --git a/drivers/base/Makefile b/drivers/base/Makefile
> > > index f2816f6ff76a..397e5c344e6a 100644
> > > --- a/drivers/base/Makefile
> > > +++ b/drivers/base/Makefile
> > > @@ -23,6 +23,7 @@ obj-$(CONFIG_SOC_BUS) += soc.o
> > > obj-$(CONFIG_PINCTRL) += pinctrl.o
> > > obj-$(CONFIG_DEV_COREDUMP) += devcoredump.o
> > > obj-$(CONFIG_GENERIC_MSI_IRQ_DOMAIN) += platform-msi.o
> > > +obj-$(CONFIG_GENERIC_ARCH_TOPOLOGY) += arch_topology.o
> > >
> > > obj-y += test/
> > >
> > > diff --git a/drivers/base/arch_topology.c b/drivers/base/arch_topology.c
> > > new file mode 100644
> > > index 000000000000..c1dd430adad2
> > > --- /dev/null
> > > +++ b/drivers/base/arch_topology.c
> > > @@ -0,0 +1,237 @@
> > > +/*
> > > + * driver/base/arch_topology.c - Arch specific cpu topology information
> >
> > No need to keep the filename in the file, you know what it is called :)
> >
>
> OK, removed.
>
> > > + *
> > > + * Copyright (C) 2016, ARM Ltd.
> > > + * Written by: Juri Lelli, ARM Ltd.
> > > + *
> > > + * This file is subject to the terms and conditions of the GNU General Public
> > > + * License. See the file "COPYING" in the main directory of this archive
> > > + * for more details.
> >
> > So, v2 only? Please be specific. Even better yet, use a SPDX header if
> > you want to, those are always nice.
> >
>
> Yes, v2 only.
>
> * for more details.
> + *
> + * Released under the GPLv2 only.
> + * SPDX-License-Identifier: GPL-2.0
>
> Would do, right?
>
> > > + */
> > > +
> > > +#include <linux/acpi.h>
> > > +#include <linux/cpu.h>
> > > +#include <linux/cpufreq.h>
> > > +#include <linux/device.h>
> > > +#include <linux/of.h>
> > > +#include <linux/slab.h>
> > > +#include <linux/string.h>
> > > +#include <linux/topology.h>
> > > +
> > > +static DEFINE_MUTEX(cpu_scale_mutex);
> > > +static DEFINE_PER_CPU(unsigned long, cpu_scale) = SCHED_CAPACITY_SCALE;
> > > +
> > > +unsigned long arch_scale_cpu_capacity(struct sched_domain *sd, int cpu)
> >
> > Why do you have sd here? You never use it:
> >
> > > +{
> > > + return per_cpu(cpu_scale, cpu);
> >
> > See? What am I missing?
> >
>
> This is how this function is defined in kernel/sched/sched.h:
>
> #ifndef arch_scale_cpu_capacity
> static __always_inline
> unsigned long arch_scale_cpu_capacity(struct sched_domain *sd, int cpu)
> {
> if (sd && (sd->flags & SD_SHARE_CPUCAPACITY) && (sd->span_weight > 1))
> return sd->smt_gain / sd->span_weight;
>
> return SCHED_CAPACITY_SCALE;
> }
> #endif
>
> and in this case the sd argument is used: there is a call site in fair.c
> that passes a non NULL sd, updated_cpu_capacity().
>
> A following set of patches will re-define the function so that the
> drivers one gets used by the kernel (only arm and arm64 will currently
> want this), with something like this in arch code
>
> #define arch_scale_cpu_capacity atd_scale_cpu_capacity
>
> Please note that last patch of this set renames this function atd_scale_
> cpu_capacity, to (hopefully) make this approach more clear.
>
> Does it make more sense to you?
>
> > > +}
> > > +
> > > +void set_capacity_scale(unsigned int cpu, unsigned long capacity)
> > > +{
> > > + per_cpu(cpu_scale, cpu) = capacity;
> > > +}
> > > +
> > > +static ssize_t cpu_capacity_show(struct device *dev,
> > > + struct device_attribute *attr,
> > > + char *buf)
> > > +{
> > > + struct cpu *cpu = container_of(dev, struct cpu, dev);
> > > +
> > > + return sprintf(buf, "%lu\n",
> > > + arch_scale_cpu_capacity(NULL, cpu->dev.id));
> > > +}
> > > +
> > > +static ssize_t cpu_capacity_store(struct device *dev,
> > > + struct device_attribute *attr,
> > > + const char *buf,
> > > + size_t count)
> > > +{
> > > + struct cpu *cpu = container_of(dev, struct cpu, dev);
> > > + int this_cpu = cpu->dev.id, i;
> >
> > new line for:
> > int i;
> > please.
> >
>
> Sure.
>
> > > + unsigned long new_capacity;
> > > + ssize_t ret;
> > > +
> > > + if (count) {
> >
> > if (!count)
> > return 0;
> >
> > then you can get on with the rest of the logic. Don't indent if you
> > don't have to.
> >
>
> Right.
>
> > > + ret = kstrtoul(buf, 0, &new_capacity);
> > > + if (ret)
> > > + return ret;
> > > + if (new_capacity > SCHED_CAPACITY_SCALE)
> > > + return -EINVAL;
> > > +
> > > + mutex_lock(&cpu_scale_mutex);
> > > + for_each_cpu(i, &cpu_topology[this_cpu].core_sibling)
> > > + set_capacity_scale(i, new_capacity);
> > > + mutex_unlock(&cpu_scale_mutex);
> > > + }
> > > +
> > > + return count;
> > > +}
> >
> > No documentation for these sysfs file? Not good :(
> >
>
> Patch 2/9 introduces some documentation. There is already more in
> Documentation/devicetree/bindings/arm/cpu-capacity.txt.
>
> Do you think I should improve further?
>
> > > +
> > > +static DEVICE_ATTR_RW(cpu_capacity);
> > > +
> > > +static int register_cpu_capacity_sysctl(void)
> > > +{
> > > + int i;
> > > + struct device *cpu;
> > > +
> > > + for_each_possible_cpu(i) {
> > > + cpu = get_cpu_device(i);
> > > + if (!cpu) {
> > > + pr_err("%s: too early to get CPU%d device!\n",
> > > + __func__, i);
> >
> > What is this going to help with?
> >
>
> Not much I guess, I can remove it.
>
> > > + continue;
> > > + }
> > > + device_create_file(cpu, &dev_attr_cpu_capacity);
> >
> > You realize you just raced userspace, right? Why do it this way and not
> > register the files when the CPU device is created/removed?
> >
>
> Humm, my intention for doing it this way is that I'd like to make all
> the code dealing with cpu_capacity confined in a single place (this
> file), without the need to modify other files.
>
> > > + }
> > > +
> > > + return 0;
> > > +}
> > > +subsys_initcall(register_cpu_capacity_sysctl);
>
> AFAIU, for both arm and arm64 CPU device is registered with a
> subsys_initcall(topology_init), so I'm doing the same. Other archs seem to do
> similar things. Could you explain a little more why this is a problem?
>
> > > +
> > > +u32 capacity_scale;
> > > +u32 *raw_capacity;
> > > +bool cap_parsing_failed;
> >
> > globals? really? That's bold :(
> >
>
> Yeah, ugly. However, patch 7/9 is making cap_parsing_failed static. The other
> two can be made static already, I should have done that in the first place. :(
>
> BTW, with this set I'm trying to incrementally fix things (after moving code in
> the new place), does it look reasonable to you or would you prefer to squash
> intermediate steps?
>
> > > +
> > > +void normalize_cpu_capacity(void)
> >
> > naming is hard, but try to put a good, descriptive, prefix on everything
> > you are exporting in the same file, the same prefix.
> >
> > cpu_capacity_normalize()?
> > cpu_capacity_register_sysctl()?
> >
> > and so on.
> >
> > > +{
> > > + u64 capacity;
> > > + int cpu;
> > > +
> > > + if (!raw_capacity || cap_parsing_failed)
> > > + return;
> > > +
> > > + pr_debug("cpu_capacity: capacity_scale=%u\n", capacity_scale);
> > > + mutex_lock(&cpu_scale_mutex);
> > > + for_each_possible_cpu(cpu) {
> > > + pr_debug("cpu_capacity: cpu=%d raw_capacity=%u\n",
> > > + cpu, raw_capacity[cpu]);
> > > + capacity = (raw_capacity[cpu] << SCHED_CAPACITY_SHIFT)
> > > + / capacity_scale;
> > > + set_capacity_scale(cpu, capacity);
> > > + pr_debug("cpu_capacity: CPU%d cpu_capacity=%lu\n",
> > > + cpu, arch_scale_cpu_capacity(NULL, cpu));
> > > + }
> > > + mutex_unlock(&cpu_scale_mutex);
> > > +}
> > > +
> > > +int __init parse_cpu_capacity(struct device_node *cpu_node, int cpu)
> >
> > cpu_capacity_parse()?
> >
>
> OK, I'll try to fix the naming as you suggest. Thanks!
>
> Best,
>
> - Juri