[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20220506183051.wimo7p4nuqfnl2aj@pali>
Date: Fri, 6 May 2022 20:30:51 +0200
From: Pali Rohár <pali@...nel.org>
To: Marc Zyngier <maz@...nel.org>
Cc: Thomas Gleixner <tglx@...utronix.de>,
Rob Herring <robh+dt@...nel.org>,
Bjorn Helgaas <bhelgaas@...gle.com>,
Andrew Lunn <andrew@...n.ch>,
Gregory Clement <gregory.clement@...tlin.com>,
Sebastian Hesselbarth <sebastian.hesselbarth@...il.com>,
Thomas Petazzoni <thomas.petazzoni@...tlin.com>,
Lorenzo Pieralisi <lorenzo.pieralisi@....com>,
Krzysztof Wilczyński <kw@...ux.com>,
Marek Behún <kabel@...nel.org>,
linux-kernel@...r.kernel.org, devicetree@...r.kernel.org,
linux-pci@...r.kernel.org, linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH 2/6] irqchip/armada-370-xp: Implement SoC Error interrupts
On Friday 06 May 2022 19:19:46 Marc Zyngier wrote:
> On Fri, 06 May 2022 14:40:25 +0100,
> Pali Rohár <pali@...nel.org> wrote:
> >
> > MPIC IRQ 4 is used as SoC Error Summary interrupt and provides access to
> > another hierarchy of SoC Error interrupts. Implement a new IRQ chip and
> > domain for accessing this IRQ hierarchy.
> >
> > Signed-off-by: Pali Rohár <pali@...nel.org>
> > ---
> > drivers/irqchip/irq-armada-370-xp.c | 213 +++++++++++++++++++++++++++-
> > 1 file changed, 210 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/irqchip/irq-armada-370-xp.c b/drivers/irqchip/irq-armada-370-xp.c
> > index ebd76ea1c69b..71578b65f5c8 100644
> > --- a/drivers/irqchip/irq-armada-370-xp.c
> > +++ b/drivers/irqchip/irq-armada-370-xp.c
> > @@ -117,6 +117,8 @@
> > /* Registers relative to main_int_base */
> > #define ARMADA_370_XP_INT_CONTROL (0x00)
> > #define ARMADA_370_XP_SW_TRIG_INT_OFFS (0x04)
> > +#define ARMADA_370_XP_INT_SOC_ERR_0_CAUSE_OFFS (0x20)
> > +#define ARMADA_370_XP_INT_SOC_ERR_1_CAUSE_OFFS (0x24)
> > #define ARMADA_370_XP_INT_SET_ENABLE_OFFS (0x30)
> > #define ARMADA_370_XP_INT_CLEAR_ENABLE_OFFS (0x34)
> > #define ARMADA_370_XP_INT_SOURCE_CTL(irq) (0x100 + irq*4)
> > @@ -130,6 +132,8 @@
> > #define ARMADA_370_XP_CPU_INTACK_OFFS (0x44)
> > #define ARMADA_370_XP_INT_SET_MASK_OFFS (0x48)
> > #define ARMADA_370_XP_INT_CLEAR_MASK_OFFS (0x4C)
> > +#define ARMADA_370_XP_INT_SOC_ERR_0_MASK_OFF (0x50)
> > +#define ARMADA_370_XP_INT_SOC_ERR_1_MASK_OFF (0x54)
> > #define ARMADA_370_XP_INT_FABRIC_MASK_OFFS (0x54)
> > #define ARMADA_370_XP_INT_CAUSE_PERF(cpu) (1 << cpu)
> >
> > @@ -146,6 +150,8 @@
> > static void __iomem *per_cpu_int_base;
> > static void __iomem *main_int_base;
> > static struct irq_domain *armada_370_xp_mpic_domain;
> > +static struct irq_domain *armada_370_xp_soc_err_domain;
> > +static unsigned int soc_err_irq_num_regs;
> > static u32 doorbell_mask_reg;
> > static int parent_irq;
> > #ifdef CONFIG_PCI_MSI
> > @@ -156,6 +162,8 @@ static DEFINE_MUTEX(msi_used_lock);
> > static phys_addr_t msi_doorbell_addr;
> > #endif
> >
> > +static void armada_370_xp_soc_err_irq_unmask(struct irq_data *d);
> > +
> > static inline bool is_percpu_irq(irq_hw_number_t irq)
> > {
> > if (irq <= ARMADA_370_XP_MAX_PER_CPU_IRQS)
> > @@ -509,6 +517,27 @@ static void armada_xp_mpic_reenable_percpu(void)
> > armada_370_xp_irq_unmask(data);
> > }
> >
> > + /* Re-enable per-CPU SoC Error interrupts that were enabled before suspend */
> > + for (irq = 0; irq < soc_err_irq_num_regs * 32; irq++) {
> > + struct irq_data *data;
> > + int virq;
> > +
> > + virq = irq_linear_revmap(armada_370_xp_soc_err_domain, irq);
> > + if (virq == 0)
> > + continue;
> > +
> > + data = irq_get_irq_data(virq);
> > +
> > + if (!irq_percpu_is_enabled(virq))
> > + continue;
> > +
> > + armada_370_xp_soc_err_irq_unmask(data);
> > + }
>
> So you do this loop and all these lookups, both here and in the resume
> function (duplicated code!) just to be able to call the unmask
> function? This would be better served by two straight writes of the
> mask register, which you'd conveniently save on suspend.
>
> Yes, you have only duplicated the existing logic. But surely there is
> something better to do.
Yes, I just used existing logic.
I'm not rewriting driver or doing big refactor of it, as this is not in
the scope of the PCIe AER interrupt support.
> > +
> > + /* Unmask summary SoC Error Interrupt */
> > + if (soc_err_irq_num_regs > 0)
> > + writel(4, per_cpu_int_base + ARMADA_370_XP_INT_CLEAR_MASK_OFFS);
> > +
> > ipi_resume();
> > }
> >
> > @@ -546,8 +575,8 @@ static struct irq_chip armada_370_xp_irq_chip = {
> > static int armada_370_xp_mpic_irq_map(struct irq_domain *h,
> > unsigned int virq, irq_hw_number_t hw)
> > {
> > - /* IRQs 0 and 1 cannot be mapped, they are handled internally */
> > - if (hw <= 1)
> > + /* IRQs 0, 1 and 4 cannot be mapped, they are handled internally */
> > + if (hw <= 1 || hw == 4)
> > return -EINVAL;
> >
> > armada_370_xp_irq_mask(irq_get_irq_data(virq));
> > @@ -577,6 +606,99 @@ static const struct irq_domain_ops armada_370_xp_mpic_irq_ops = {
> > .xlate = irq_domain_xlate_onecell,
> > };
> >
> > +static DEFINE_RAW_SPINLOCK(armada_370_xp_soc_err_lock);
> > +
> > +static void armada_370_xp_soc_err_irq_mask(struct irq_data *d)
> > +{
> > + irq_hw_number_t hwirq = irqd_to_hwirq(d);
> > + u32 reg, mask;
> > +
> > + reg = hwirq >= 32 ? ARMADA_370_XP_INT_SOC_ERR_1_MASK_OFF
> > + : ARMADA_370_XP_INT_SOC_ERR_0_MASK_OFF;
> > +
> > + raw_spin_lock(&armada_370_xp_soc_err_lock);
> > + mask = readl(per_cpu_int_base + reg);
> > + mask &= ~BIT(hwirq % 32);
> > + writel(mask, per_cpu_int_base + reg);
> > + raw_spin_unlock(&armada_370_xp_soc_err_lock);
> > +}
> > +
> > +static void armada_370_xp_soc_err_irq_unmask(struct irq_data *d)
> > +{
> > + irq_hw_number_t hwirq = irqd_to_hwirq(d);
> > + u32 reg, mask;
> > +
> > + reg = hwirq >= 32 ? ARMADA_370_XP_INT_SOC_ERR_1_MASK_OFF
> > + : ARMADA_370_XP_INT_SOC_ERR_0_MASK_OFF;
> > +
> > + raw_spin_lock(&armada_370_xp_soc_err_lock);
> > + mask = readl(per_cpu_int_base + reg);
> > + mask |= BIT(hwirq % 32);
> > + writel(mask, per_cpu_int_base + reg);
> > + raw_spin_unlock(&armada_370_xp_soc_err_lock);
> > +}
> > +
> > +static int armada_370_xp_soc_err_irq_mask_on_cpu(void *par)
> > +{
> > + struct irq_data *d = par;
> > + armada_370_xp_soc_err_irq_mask(d);
> > + return 0;
> > +}
> > +
> > +static int armada_370_xp_soc_err_irq_unmask_on_cpu(void *par)
> > +{
> > + struct irq_data *d = par;
> > + armada_370_xp_soc_err_irq_unmask(d);
> > + return 0;
> > +}
> > +
> > +static int armada_xp_soc_err_irq_set_affinity(struct irq_data *d,
> > + const struct cpumask *mask,
> > + bool force)
> > +{
> > + unsigned int cpu;
> > +
> > + cpus_read_lock();
> > +
> > + /* First disable IRQ on all cores */
> > + for_each_online_cpu(cpu)
> > + smp_call_on_cpu(cpu, armada_370_xp_soc_err_irq_mask_on_cpu, d, true);
> > +
> > + /* Select a single core from the affinity mask which is online */
> > + cpu = cpumask_any_and(mask, cpu_online_mask);
> > + smp_call_on_cpu(cpu, armada_370_xp_soc_err_irq_unmask_on_cpu, d, true);
> > +
> > + cpus_read_unlock();
> > +
> > + irq_data_update_effective_affinity(d, cpumask_of(cpu));
> > +
> > + return IRQ_SET_MASK_OK;
> > +}
>
> Aren't these per-CPU interrupts anyway? What does it mean to set their
> affinity? /me rolls eyes...
Yes, they are per-CPU interrupts. But to mask or unmask particular
interrupt for specific CPU is possible only from that CPU. CPU 0 just
cannot move interrupt from CPU 0 to CPU 1. CPU 0 can only mask that
interrupt and CPU 1 has to unmask it.
> > +
> > +static struct irq_chip armada_370_xp_soc_err_irq_chip = {
> > + .name = "MPIC SOC",
> > + .irq_mask = armada_370_xp_soc_err_irq_mask,
> > + .irq_unmask = armada_370_xp_soc_err_irq_unmask,
> > + .irq_set_affinity = armada_xp_soc_err_irq_set_affinity,
> > +};
> > +
> > +static int armada_370_xp_soc_err_irq_map(struct irq_domain *h,
> > + unsigned int virq, irq_hw_number_t hw)
> > +{
> > + armada_370_xp_soc_err_irq_mask(irq_get_irq_data(virq));
> > + irq_set_status_flags(virq, IRQ_LEVEL);
> > + irq_set_percpu_devid(virq);
> > + irq_set_chip_and_handler(virq, &armada_370_xp_soc_err_irq_chip,
> > + handle_percpu_devid_irq);
> > + irq_set_probe(virq);
> > + return 0;
> > +}
> > +
> > +static const struct irq_domain_ops armada_370_xp_soc_err_irq_ops = {
> > + .map = armada_370_xp_soc_err_irq_map,
> > + .xlate = irq_domain_xlate_onecell,
> > +};
> > +
> > #ifdef CONFIG_PCI_MSI
> > static void armada_370_xp_handle_msi_irq(struct pt_regs *regs, bool is_chained)
> > {
> > @@ -605,6 +727,32 @@ static void armada_370_xp_handle_msi_irq(struct pt_regs *regs, bool is_chained)
> > static void armada_370_xp_handle_msi_irq(struct pt_regs *r, bool b) {}
> > #endif
> >
> > +static void armada_370_xp_handle_soc_err_irq(void)
> > +{
> > + unsigned long status, bit;
> > + u32 mask, cause;
> > +
> > + if (soc_err_irq_num_regs < 1)
> > + return;
> > +
> > + mask = readl(per_cpu_int_base + ARMADA_370_XP_INT_SOC_ERR_0_MASK_OFF);
> > + cause = readl(main_int_base + ARMADA_370_XP_INT_SOC_ERR_0_CAUSE_OFFS);
> > + status = cause & mask;
> > +
> > + for_each_set_bit(bit, &status, 32)
> > + generic_handle_domain_irq(armada_370_xp_soc_err_domain, bit);
> > +
> > + if (soc_err_irq_num_regs < 2)
> > + return;
> > +
> > + mask = readl(per_cpu_int_base + ARMADA_370_XP_INT_SOC_ERR_1_MASK_OFF);
> > + cause = readl(main_int_base + ARMADA_370_XP_INT_SOC_ERR_1_CAUSE_OFFS);
> > + status = cause & mask;
> > +
> > + for_each_set_bit(bit, &status, 32)
> > + generic_handle_domain_irq(armada_370_xp_soc_err_domain, bit + 32);
> > +}
> > +
> > static void armada_370_xp_mpic_handle_cascade_irq(struct irq_desc *desc)
> > {
> > struct irq_chip *chip = irq_desc_get_chip(desc);
> > @@ -630,6 +778,11 @@ static void armada_370_xp_mpic_handle_cascade_irq(struct irq_desc *desc)
> > continue;
> > }
> >
> > + if (irqn == 4) {
> > + armada_370_xp_handle_soc_err_irq();
> > + continue;
> > + }
> > +
> > generic_handle_domain_irq(armada_370_xp_mpic_domain, irqn);
> > }
> >
> > @@ -649,7 +802,7 @@ armada_370_xp_handle_irq(struct pt_regs *regs)
> > if (irqnr > 1022)
> > break;
> >
> > - if (irqnr > 1) {
> > + if (irqnr > 1 && irqnr != 4) {
> > generic_handle_domain_irq(armada_370_xp_mpic_domain,
> > irqnr);
> > continue;
> > @@ -659,6 +812,10 @@ armada_370_xp_handle_irq(struct pt_regs *regs)
> > if (irqnr == 1)
> > armada_370_xp_handle_msi_irq(regs, false);
> >
> > + /* SoC Error handling */
> > + if (irqnr == 4)
> > + armada_370_xp_handle_soc_err_irq();
> > +
> > #ifdef CONFIG_SMP
> > /* IPI Handling */
> > if (irqnr == 0) {
> > @@ -722,6 +879,26 @@ static void armada_370_xp_mpic_resume(void)
> > }
> > }
> >
> > + /* Re-enable per-CPU SoC Error interrupts */
> > + for (irq = 0; irq < soc_err_irq_num_regs * 32; irq++) {
> > + struct irq_data *data;
> > + int virq;
> > +
> > + virq = irq_linear_revmap(armada_370_xp_soc_err_domain, irq);
> > + if (virq == 0)
> > + continue;
> > +
> > + data = irq_get_irq_data(virq);
> > +
> > + /*
> > + * Re-enable on the current CPU,
> > + * armada_xp_mpic_reenable_percpu() will take
> > + * care of secondary CPUs when they come up.
> > + */
> > + if (irq_percpu_is_enabled(virq))
> > + armada_370_xp_soc_err_irq_unmask(data);
> > + }
>
> As I said above, this is duplicated code that should be replaced with
> a simple write to the corresponding MMIO registers.
>
> > +
> > /* Reconfigure doorbells for IPIs and MSIs */
> > writel(doorbell_mask_reg,
> > per_cpu_int_base + ARMADA_370_XP_IN_DRBEL_MSK_OFFS);
> > @@ -730,6 +907,10 @@ static void armada_370_xp_mpic_resume(void)
> > if (doorbell_mask_reg & PCI_MSI_DOORBELL_MASK)
> > writel(1, per_cpu_int_base + ARMADA_370_XP_INT_CLEAR_MASK_OFFS);
> >
> > + /* Unmask summary SoC Error Interrupt */
> > + if (soc_err_irq_num_regs > 0)
> > + writel(4, per_cpu_int_base + ARMADA_370_XP_INT_CLEAR_MASK_OFFS);
>
> Magic value?
>
> Also, writing to this register tends to indicate that the whole thing
> should really be a chained irqchip... Maybe that's overkill in this
> instance, but the whole thing is rather oddly architected.
I used exactly same code style like it is used for MSI doorbell
interrupts.
Yes, it looks like chained irqchip, but it is overkill.
> > +
> > ipi_resume();
> > }
> >
> > @@ -742,6 +923,7 @@ static int __init armada_370_xp_mpic_of_init(struct device_node *node,
> > struct device_node *parent)
> > {
> > struct resource main_int_res, per_cpu_int_res;
> > + struct device_node *soc_err_node;
> > int nr_irqs, i;
> > u32 control;
> >
> > @@ -775,12 +957,37 @@ static int __init armada_370_xp_mpic_of_init(struct device_node *node,
> > BUG_ON(!armada_370_xp_mpic_domain);
> > irq_domain_update_bus_token(armada_370_xp_mpic_domain, DOMAIN_BUS_WIRED);
> >
> > + soc_err_node = of_get_next_child(node, NULL);
> > + if (!soc_err_node) {
> > + pr_warn("Missing SoC Error Interrupt Controller node\n");
> > + pr_warn("Extended interrupts are not supported\n");
> > + } else {
> > + pr_info("Registering MPIC SoC Error Interrupt Controller\n");
> > + /*
> > + * Armada 370 and XP have only 32 SoC Error IRQs in one register
> > + * and other Armada platforms have 64 IRQs in two registers.
> > + */
> > + soc_err_irq_num_regs =
> > + of_machine_is_compatible("marvell,armada-370-xp") ? 1 : 2;
>
> Don't you have an actual compatible string for the interrupt
> controller? It seems odd to rely on the SoC name.
Compatible string is same for all those 32-bit Armada SoCs. So it cannot
be used to distinguish between XP and 385. That is why I used
of_machine_is_compatible.
> > + armada_370_xp_soc_err_domain =
> > + irq_domain_add_hierarchy(armada_370_xp_mpic_domain, 0,
> > + soc_err_irq_num_regs * 32,
> > + soc_err_node,
> > + &armada_370_xp_soc_err_irq_ops,
> > + NULL);
> > + BUG_ON(!armada_370_xp_soc_err_domain);
> > + }
> > +
> > /* Setup for the boot CPU */
> > armada_xp_mpic_perf_init();
> > armada_xp_mpic_smp_cpu_init();
> >
> > armada_370_xp_msi_init(node, main_int_res.start);
> >
> > + /* Unmask summary SoC Error Interrupt */
> > + if (soc_err_irq_num_regs > 0)
> > + writel(4, per_cpu_int_base + ARMADA_370_XP_INT_CLEAR_MASK_OFFS);
> > +
>
> Magic value, duplicated this time?
>
> > parent_irq = irq_of_parse_and_map(node, 0);
> > if (parent_irq <= 0) {
> > irq_set_default_host(armada_370_xp_mpic_domain);
> > --
> > 2.20.1
> >
> >
>
> M.
>
> --
> Without deviation from the norm, progress is not possible.
Powered by blists - more mailing lists