Date:   Mon, 06 Feb 2023 12:48:53 +0000
Message-ID: <86v8kfyn62.wl-maz@kernel.org>
From:   Marc Zyngier <maz@kernel.org>
To:     Mason Huo <mason.huo@starfivetech.com>
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Palmer Dabbelt <palmer@dabbelt.com>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        <linux-kernel@vger.kernel.org>, <linux-riscv@lists.infradead.org>,
        Ley Foon Tan <leyfoon.tan@starfivetech.com>,
        Sia Jee Heng <jeeheng.sia@starfivetech.com>
Subject: Re: [PATCH v1] irqchip/irq-sifive-plic: Add syscore callbacks for hibernation
In-Reply-To: <c851138c-148a-bc5f-219b-1573d7e7e318@starfivetech.com>
References: <20230113094216.116036-1-mason.huo@starfivetech.com>
        <864js01j26.wl-maz@kernel.org>
        <c851138c-148a-bc5f-219b-1573d7e7e318@starfivetech.com>
User-Agent: Wanderlust/2.15.9 (Almost Unreal) SEMI-EPG/1.14.7 (Harue)
 FLIM-LB/1.14.9 (=?UTF-8?B?R29qxY0=?=) APEL-LB/10.8 EasyPG/1.0.0 Emacs/28.2
 (aarch64-unknown-linux-gnu) MULE/6.0 (HANACHIRUSATO)
MIME-Version: 1.0 (generated by SEMI-EPG 1.14.7 - "Harue")
Content-Type: text/plain; charset=US-ASCII
Precedence: bulk

On Mon, 06 Feb 2023 06:13:11 +0000,
Mason Huo <mason.huo@starfivetech.com> wrote:
> 
> 
> 
> On 2023/2/5 18:51, Marc Zyngier wrote:
> > On Fri, 13 Jan 2023 09:42:16 +0000,
> > Mason Huo <mason.huo@starfivetech.com> wrote:
> >> 
> >> The priority and enable registers of plic will be reset
> >> during hibernation power cycle in poweroff mode,
> >> add the syscore callbacks to save/restore those registers.
> >> 
> >> Signed-off-by: Mason Huo <mason.huo@starfivetech.com>
> >> Reviewed-by: Ley Foon Tan <leyfoon.tan@starfivetech.com>
> >> Reviewed-by: Sia Jee Heng <jeeheng.sia@starfivetech.com>
> >> ---
> >>  drivers/irqchip/irq-sifive-plic.c | 93 ++++++++++++++++++++++++++++++-
> >>  1 file changed, 91 insertions(+), 2 deletions(-)
> >> 
> >> diff --git a/drivers/irqchip/irq-sifive-plic.c b/drivers/irqchip/irq-sifive-plic.c
> >> index ff47bd0dec45..80306de45d2b 100644
> >> --- a/drivers/irqchip/irq-sifive-plic.c
> >> +++ b/drivers/irqchip/irq-sifive-plic.c
> >> @@ -17,6 +17,7 @@
> >>  #include <linux/of_irq.h>
> >>  #include <linux/platform_device.h>
> >>  #include <linux/spinlock.h>
> >> +#include <linux/syscore_ops.h>
> >>  #include <asm/smp.h>
> >>  
> >>  /*
> >> @@ -67,6 +68,8 @@ struct plic_priv {
> >>  	struct irq_domain *irqdomain;
> >>  	void __iomem *regs;
> >>  	unsigned long plic_quirks;
> >> +	unsigned int nr_irqs;
> >> +	u32 *priority_reg;
> >>  };
> >>  
> >>  struct plic_handler {
> >> @@ -79,10 +82,13 @@ struct plic_handler {
> >>  	raw_spinlock_t		enable_lock;
> >>  	void __iomem		*enable_base;
> >>  	struct plic_priv	*priv;
> >> +	/* To record interrupts that are enabled before suspend. */
> >> +	u32 enable_reg[MAX_DEVICES / 32];
> > 
> > What does MAX_DEVICES represent here? How is it related to the number
> > of interrupts you're trying to save? It seems to be related to the
> > number of CPUs, so it hardly makes any sense so far.
> > 
> The comment of this macro describes that "The largest number supported
> by devices marked as 'sifive,plic-1.0.0', is 1024, of which
> device 0 is defined as non-existent by the RISC-V Privileged Spec."
> As far as I understand, the *device* here means HW IRQ source,
> and the HW IRQ 0 is non-existent.

So why is it sized to that maximum value? The binding gives you the
*real* value that the HW implements.

> 
> >>  };
> >>  static int plic_parent_irq __ro_after_init;
> >>  static bool plic_cpuhp_setup_done __ro_after_init;
> >>  static DEFINE_PER_CPU(struct plic_handler, plic_handlers);
> >> +static struct plic_priv *priv_data;
> >>  
> >>  static int plic_irq_set_type(struct irq_data *d, unsigned int type);
> >>  
> >> @@ -229,6 +235,78 @@ static int plic_irq_set_type(struct irq_data *d, unsigned int type)
> >>  	return IRQ_SET_MASK_OK;
> >>  }
> >>  
> >> +static void plic_irq_resume(void)
> >> +{
> >> +	unsigned int i, cpu;
> >> +	u32 __iomem *reg;
> >> +
> >> +	for (i = 0; i < priv_data->nr_irqs; i++)
> >> +		writel(priv_data->priority_reg[i],
> >> +				priv_data->regs + PRIORITY_BASE + i * PRIORITY_PER_ID);
> > 
> > From what I can tell, this driver uses exactly 2 priorities: 0 and 1.
> > And yet you use a full 32bit to encode those. Does it seem like a good
> > idea?
> > 
> Yes, currently this driver uses oly 2 priorities.
> But, according to the sifive spec, the priority register is a 32bit register,
> and it supports 7 levels of priority.

And? This is a Linux driver, not an implementation validation
tool. What is the point of saving/restoring stuff that is *never*
used? :-(

	M.

-- 
Without deviation from the norm, progress is not possible.