[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CH2PR13MB3368C0B3E991478BDA7665578C3F0@CH2PR13MB3368.namprd13.prod.outlook.com>
Date: Tue, 7 Jan 2020 03:55:49 +0000
From: Yash Shah <yash.shah@...ive.com>
To: Anup Patel <anup@...infault.org>
CC: Rob Herring <robh+dt@...nel.org>,
Mark Rutland <mark.rutland@....com>,
"Paul Walmsley ( Sifive)" <paul.walmsley@...ive.com>,
Palmer Dabbelt <palmer@...belt.com>,
Albert Ou <aou@...s.berkeley.edu>,
Bin Meng <bmeng.cn@...il.com>,
Green Wan <green.wan@...ive.com>,
Allison Randal <allison@...utok.net>,
Alexios Zavras <alexios.zavras@...el.com>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
Thomas Gleixner <tglx@...utronix.de>,
"bp@...e.de" <bp@...e.de>,
"devicetree@...r.kernel.org" <devicetree@...r.kernel.org>,
linux-riscv <linux-riscv@...ts.infradead.org>,
"linux-kernel@...r.kernel.org List" <linux-kernel@...r.kernel.org>,
Sachin Ghadi <sachin.ghadi@...ive.com>
Subject: RE: [PATCH v2 2/2] riscv: cacheinfo: Add support to determine no. of
L2 cache way enabled
> -----Original Message-----
> From: Anup Patel <anup@...infault.org>
> Sent: 06 January 2020 14:40
> To: Yash Shah <yash.shah@...ive.com>
> Cc: Rob Herring <robh+dt@...nel.org>; Mark Rutland
> <mark.rutland@....com>; Paul Walmsley ( Sifive)
> <paul.walmsley@...ive.com>; Palmer Dabbelt <palmer@...belt.com>;
> Albert Ou <aou@...s.berkeley.edu>; Bin Meng <bmeng.cn@...il.com>;
> Green Wan <green.wan@...ive.com>; Allison Randal <allison@...utok.net>;
> Alexios Zavras <alexios.zavras@...el.com>; Greg Kroah-Hartman
> <gregkh@...uxfoundation.org>; Thomas Gleixner <tglx@...utronix.de>;
> bp@...e.de; devicetree@...r.kernel.org; linux-riscv <linux-
> riscv@...ts.infradead.org>; linux-kernel@...r.kernel.org List <linux-
> kernel@...r.kernel.org>; Sachin Ghadi <sachin.ghadi@...ive.com>
> Subject: Re: [PATCH v2 2/2] riscv: cacheinfo: Add support to determine no. of
> L2 cache way enabled
>
> On Fri, Jan 3, 2020 at 9:44 AM Yash Shah <yash.shah@...ive.com> wrote:
> >
> > In order to determine the number of L2 cache ways enabled at runtime,
> > implement a private attribute using cache_get_priv_group() in
> > cacheinfo framework. Reading this attribute
> ("number_of_ways_enabled")
> > will return the number of enabled L2 cache ways at runtime.
> >
> > Signed-off-by: Yash Shah <yash.shah@...ive.com>
> > ---
> > arch/riscv/include/asm/sifive_l2_cache.h | 2 ++
> > arch/riscv/kernel/cacheinfo.c | 31
> +++++++++++++++++++++++++++++++
> > drivers/soc/sifive/sifive_l2_cache.c | 5 +++++
> > 3 files changed, 38 insertions(+)
> >
> > diff --git a/arch/riscv/include/asm/sifive_l2_cache.h
> > b/arch/riscv/include/asm/sifive_l2_cache.h
> > index 04f6748..217a42f 100644
> > --- a/arch/riscv/include/asm/sifive_l2_cache.h
> > +++ b/arch/riscv/include/asm/sifive_l2_cache.h
> > @@ -10,6 +10,8 @@
> > extern int register_sifive_l2_error_notifier(struct notifier_block
> > *nb); extern int unregister_sifive_l2_error_notifier(struct
> > notifier_block *nb);
> >
> > +int sifive_l2_largest_wayenabled(void);
> > +
> > #define SIFIVE_L2_ERR_TYPE_CE 0
> > #define SIFIVE_L2_ERR_TYPE_UE 1
> >
> > diff --git a/arch/riscv/kernel/cacheinfo.c
> > b/arch/riscv/kernel/cacheinfo.c index 4c90c07..29bdb21 100644
> > --- a/arch/riscv/kernel/cacheinfo.c
> > +++ b/arch/riscv/kernel/cacheinfo.c
> > @@ -7,6 +7,7 @@
> > #include <linux/cpu.h>
> > #include <linux/of.h>
> > #include <linux/of_device.h>
> > +#include <asm/sifive_l2_cache.h>
> >
> > static void ci_leaf_init(struct cacheinfo *this_leaf,
> > struct device_node *node, @@ -16,6 +17,36 @@
> > static void ci_leaf_init(struct cacheinfo *this_leaf,
> > this_leaf->type = type;
> > }
> >
> > +#ifdef CONFIG_SIFIVE_L2
> > +static ssize_t number_of_ways_enabled_show(struct device *dev,
> > + struct device_attribute *attr,
> > + char *buf) {
> > + return sprintf(buf, "%u\n", sifive_l2_largest_wayenabled()); }
> > +
> > +static DEVICE_ATTR_RO(number_of_ways_enabled);
> > +
> > +static struct attribute *priv_attrs[] = {
> > + &dev_attr_number_of_ways_enabled.attr,
> > + NULL,
> > +};
> > +
> > +static const struct attribute_group priv_attr_group = {
> > + .attrs = priv_attrs,
> > +};
> > +
> > +const struct attribute_group *
> > +cache_get_priv_group(struct cacheinfo *this_leaf) {
> > + /* We want to use private group for L2 cache only */
> > + if (this_leaf->level == 2)
> > + return &priv_attr_group;
> > + else
> > + return NULL;
> > +}
> > +#endif /* CONFIG_SIFIVE_L2 */
> > +
>
> Instead of this, I would suggest to implement a generic ops structure.
>
> In arch/riscv/include/asm/cacheinfo.h:
>
> struct riscv_caceinfo_ops {
> const struct attribute_group * (*get_priv_group)(struct cacheinfo
> *this_leaf); };
>
> void riscv_set_cacheinfo_ops(struct riscv_caceinfo_ops *ops);
>
> In arch/riscv/riscv/kernel/cacheinfo.h
>
> static struct riscv_caceinfo_ops *rv_cache_ops;
>
> void riscv_set_cacheinfo_ops(struct riscv_caceinfo_ops *ops) {
> rv_cache_ops = ops;
> }
> EXPORT_SYMBOL_GPL(riscv_set_cacheinfo_ops);
>
> const struct attribute_group *
> cache_get_priv_group(struct cacheinfo *this_leaf) {
> if (rv_cache_ops && rv_cache_ops->get_priv_group)
> return rv_cache_ops->get_priv_group(this_leaf)
> return NULL;
> }
>
> Above will be a separate patch. In future, we can add more ops for SOC
> specific cacheinfo.
>
> Using riscv_set_cacheinfo_ops() you can have another patch to implement
> SiFive L2 info entirely in drivers/soc/sifive/sifive_l2_cache.c
>
Yes I agree, the above approach is better. Will work on this approach and send new patches.
> Also, I would strongly recommend moving
> arch/riscv/include/asm/sifive_l2_cache.h
> TO
> include/soc/sifive/sifive_l2_cache.h
Sure, will send a separate patch for this.
>
> > static int __init_cache_level(unsigned int cpu) {
> > struct cpu_cacheinfo *this_cpu_ci = get_cpu_cacheinfo(cpu);
> > diff --git a/drivers/soc/sifive/sifive_l2_cache.c
> > b/drivers/soc/sifive/sifive_l2_cache.c
> > index a9ffff3..f1a5f2c 100644
> > --- a/drivers/soc/sifive/sifive_l2_cache.c
> > +++ b/drivers/soc/sifive/sifive_l2_cache.c
> > @@ -107,6 +107,11 @@ int unregister_sifive_l2_error_notifier(struct
> > notifier_block *nb) }
> > EXPORT_SYMBOL_GPL(unregister_sifive_l2_error_notifier);
> >
> > +int sifive_l2_largest_wayenabled(void)
> > +{
> > + return readl(l2_base + SIFIVE_L2_WAYENABLE); }
> > +
> > static irqreturn_t l2_int_handler(int irq, void *device) {
> > unsigned int add_h, add_l;
> > --
> > 2.7.4
> >
>
> Regards,
> Anup
Thanks for your comments
- Yash
Powered by blists - more mailing lists