From: Marc Zyngier <marc.zyngier@arm.com>
To: Anup Patel <anup@brainfault.org>
Cc: "kvmarm@lists.cs.columbia.edu" <kvmarm@lists.cs.columbia.edu>,
linux-arm-kernel <linux-arm-kernel@lists.infradead.org>,
"linux-kernel@vger.kernel.org List"
<linux-kernel@vger.kernel.org>,
Catalin Marinas <Catalin.Marinas@arm.com>,
Thomas Gleixner <tglx@linutronix.de>,
Will Deacon <Will.Deacon@arm.com>,
Christoffer Dall <christoffer.dall@linaro.org>,
Eric Auger <eric.auger@linaro.org>,
Ian Campbell <ian.campbell@citrix.com>,
Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Subject: Re: [RFC PATCH 3/9] irqchip: GIC: Convert to EOImode == 1
Date: Wed, 25 Jun 2014 15:24:02 +0100 [thread overview]
Message-ID: <53AADB82.3040808@arm.com> (raw)
In-Reply-To: <CAAhSdy11HQtiOD6XLqQcZ+zf_Z9Mx0nPrJ0ECNNuRkZt=CfdpA@mail.gmail.com>
Hi Anup,
On 25/06/14 14:56, Anup Patel wrote:
> Hi Marc,
>
> On Wed, Jun 25, 2014 at 2:58 PM, Marc Zyngier <marc.zyngier@arm.com> wrote:
>> So far, GICv2 has been used in with EOImode == 0. The effect of this
>> mode is to perform the priority drop and the deactivation of the
>> interrupt at the same time.
>>
>> While this works perfectly for Linux (we only have a single priority),
>> it causes issues when an interrupt is forwarded to a guest, and when
>> we want the guest to perform the EOI itself.
>>
>> For this case, the GIC architecture provides EOImode == 1, where:
>> - A write to the EOI register drops the priority of the interrupt and leaves
>> it active. Other interrupts at the same priority level can now be taken,
>> but the active interrupt cannot be taken again
>> - A write to the DIR marks the interrupt as inactive, meaning it can
>> now be taken again.
>>
>> We only enable this feature when booted in HYP mode. Also, as most device
>> trees are broken (they report the CPU interface size to be 4kB, while
>> the GICv2 CPU interface size is 8kB), output a warning if we're booted
>> in HYP mode, and disable the feature.
>>
>> Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
>> ---
>> drivers/irqchip/irq-gic.c | 61 +++++++++++++++++++++++++++++++++++++----
>> include/linux/irqchip/arm-gic.h | 4 +++
>> 2 files changed, 59 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/irqchip/irq-gic.c b/drivers/irqchip/irq-gic.c
>> index 508b815..9295bf2 100644
>> --- a/drivers/irqchip/irq-gic.c
>> +++ b/drivers/irqchip/irq-gic.c
>> @@ -45,6 +45,7 @@
>> #include <asm/irq.h>
>> #include <asm/exception.h>
>> #include <asm/smp_plat.h>
>> +#include <asm/virt.h>
>>
>> #include "irq-gic-common.h"
>> #include "irqchip.h"
>> @@ -94,6 +95,10 @@ struct irq_chip gic_arch_extn = {
>> .irq_set_wake = NULL,
>> };
>>
>> +static struct irq_chip *gic_chip;
>> +
>> +static bool supports_deactivate = false;
>> +
>> #ifndef MAX_GIC_NR
>> #define MAX_GIC_NR 1
>> #endif
>> @@ -185,6 +190,12 @@ static void gic_eoi_irq(struct irq_data *d)
>> writel_relaxed(gic_irq(d), gic_cpu_base(d) + GIC_CPU_EOI);
>> }
>>
>> +static void gic_eoi_dir_irq(struct irq_data *d)
>> +{
>> + gic_eoi_irq(d);
>> + writel_relaxed(gic_irq(d), gic_cpu_base(d) + GIC_CPU_DEACTIVATE);
>> +}
>> +
>> static int gic_set_type(struct irq_data *d, unsigned int type)
>> {
>> void __iomem *base = gic_dist_base(d);
>> @@ -277,6 +288,7 @@ static void __exception_irq_entry gic_handle_irq(struct pt_regs *regs)
>> }
>> if (irqnr < 16) {
>> writel_relaxed(irqstat, cpu_base + GIC_CPU_EOI);
>> + writel_relaxed(irqstat, cpu_base + GIC_CPU_DEACTIVATE);
>> #ifdef CONFIG_SMP
>> handle_IPI(irqnr, regs);
>> #endif
>> @@ -313,7 +325,7 @@ static void gic_handle_cascade_irq(unsigned int irq, struct irq_desc *desc)
>> chained_irq_exit(chip, desc);
>> }
>>
>> -static struct irq_chip gic_chip = {
>> +static struct irq_chip gicv1_chip = {
>> .name = "GIC",
>> .irq_mask = gic_mask_irq,
>> .irq_unmask = gic_unmask_irq,
>> @@ -326,6 +338,19 @@ static struct irq_chip gic_chip = {
>> .irq_set_wake = gic_set_wake,
>> };
>>
>> +static struct irq_chip gicv2_chip = {
>> + .name = "GIC",
>> + .irq_mask = gic_mask_irq,
>> + .irq_unmask = gic_unmask_irq,
>> + .irq_eoi = gic_eoi_dir_irq,
>> + .irq_set_type = gic_set_type,
>> + .irq_retrigger = gic_retrigger,
>> +#ifdef CONFIG_SMP
>> + .irq_set_affinity = gic_set_affinity,
>> +#endif
>> + .irq_set_wake = gic_set_wake,
>> +};
>> +
>> void __init gic_cascade_irq(unsigned int gic_nr, unsigned int irq)
>> {
>> if (gic_nr >= MAX_GIC_NR)
>> @@ -377,6 +402,16 @@ static void __init gic_dist_init(struct gic_chip_data *gic)
>> writel_relaxed(1, base + GIC_DIST_CTRL);
>> }
>>
>> +static void gic_cpu_if_up(void)
>> +{
>> + void __iomem *cpu_base = gic_data_cpu_base(&gic_data[0]);
>> + u32 val = 0;
>> +
>> + if (supports_deactivate)
>> + val = GIC_CPU_CTRL_EOImodeNS;
>> + writel_relaxed(val | 1, cpu_base + GIC_CPU_CTRL);
>> +}
>> +
>> static void gic_cpu_init(struct gic_chip_data *gic)
>> {
>> void __iomem *dist_base = gic_data_dist_base(gic);
>> @@ -402,7 +437,7 @@ static void gic_cpu_init(struct gic_chip_data *gic)
>> gic_cpu_config(dist_base, NULL);
>>
>> writel_relaxed(0xf0, base + GIC_CPU_PRIMASK);
>> - writel_relaxed(1, base + GIC_CPU_CTRL);
>> + gic_cpu_if_up();
>> }
>>
>> void gic_cpu_if_down(void)
>> @@ -543,7 +578,7 @@ static void gic_cpu_restore(unsigned int gic_nr)
>> writel_relaxed(0xa0a0a0a0, dist_base + GIC_DIST_PRI + i * 4);
>>
>> writel_relaxed(0xf0, cpu_base + GIC_CPU_PRIMASK);
>> - writel_relaxed(1, cpu_base + GIC_CPU_CTRL);
>> + gic_cpu_if_up();
>> }
>>
>> static int gic_notifier(struct notifier_block *self, unsigned long cmd, void *v)
>> @@ -770,11 +805,11 @@ static int gic_irq_domain_map(struct irq_domain *d, unsigned int irq,
>> {
>> if (hw < 32) {
>> irq_set_percpu_devid(irq);
>> - irq_set_chip_and_handler(irq, &gic_chip,
>> + irq_set_chip_and_handler(irq, gic_chip,
>> handle_percpu_devid_irq);
>> set_irq_flags(irq, IRQF_VALID | IRQF_NOAUTOEN);
>> } else {
>> - irq_set_chip_and_handler(irq, &gic_chip,
>> + irq_set_chip_and_handler(irq, gic_chip,
>> handle_fasteoi_irq);
>> set_irq_flags(irq, IRQF_VALID | IRQF_PROBE);
>>
>> @@ -951,6 +986,11 @@ void __init gic_init_bases(unsigned int gic_nr, int irq_start,
>>
>> gic_irqs -= hwirq_base; /* calculate # of irqs to allocate */
>>
>> + if (!supports_deactivate)
>> + gic_chip = &gicv1_chip;
>> + else
>> + gic_chip = &gicv2_chip;
>> +
>> if (of_property_read_u32(node, "arm,routable-irqs",
>> &nr_routable_irqs)) {
>> irq_base = irq_alloc_descs(irq_start, 16, gic_irqs,
>> @@ -980,7 +1020,7 @@ void __init gic_init_bases(unsigned int gic_nr, int irq_start,
>> set_handle_irq(gic_handle_irq);
>> }
>>
>> - gic_chip.flags |= gic_arch_extn.flags;
>> + gic_chip->flags |= gic_arch_extn.flags;
>> gic_dist_init(gic);
>> gic_cpu_init(gic);
>> gic_pm_init(gic);
>> @@ -994,6 +1034,7 @@ gic_of_init(struct device_node *node, struct device_node *parent)
>> {
>> void __iomem *cpu_base;
>> void __iomem *dist_base;
>> + struct resource cu_res;
>> u32 percpu_offset;
>> int irq;
>>
>> @@ -1006,6 +1047,14 @@ gic_of_init(struct device_node *node, struct device_node *parent)
>> cpu_base = of_iomap(node, 1);
>> WARN(!cpu_base, "unable to map gic cpu registers\n");
>>
>> + of_address_to_resource(node, 1, &cpu_res);
>> + if (is_hyp_mode_available()) {
>
> Interrupt deactivation feature is not dependent on whether hyp-mode
> is available or not.
Indeed. But for the Linux use-case, this is a pointless feature, unless
you want to use it for things like threaded interrupts (but that's not
what this patch series is about).
> We can use GICC_DIR even if CPU does not have virtualization
> extension. Also, we can use GICV_DIR when running as Guest or VM.
I know that. But what's the point? The joy of doing two MMIO accesses
instead of one? It is likely to be a net performance loss if you're not
actively using it.
> I think "supports_deactivate" should depend on
> the GIC compatible string.
Probably, but again, what is the point of using it if there is no benefit?
>> + if (resource_size(&cpu_res) >= SZ_8K)
>> + supports_deactivate = true;
>> + else
>> + pr_warn("GIC: CPU interface size is %x, DT is probably wrong\n", (int)resource_size(&cpu_res));
>
> This will not work on APM X-Gene because, for X-Gene first CPU page
> is at 0x78020000 and second CPU page is at 0x78030000.
Then, as far as I'm concerned, it is not compliant with the architecture
(the GICv2 spec says clearly that GICC_DIR is at offset 0x1000, not
0x10000).
We can work around it, but that's not the purpose of this series.
> Ian had send-out a patch long time back to extend GIC dt-bindings for
> addressing this issue.
> (http://www.spinics.net/lists/arm-kernel/msg283767.html)
Wow. That's really horrible. Blimey! Ian, you owe me a few beers, just
so I can forget this... ;-)
M.
--
Jazz is not dead. It just smells funny...
next prev parent reply other threads:[~2014-06-25 14:24 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-25 9:28 [RFC PATCH 0/9] ARM: Forwarding physical interrupts to a guest VM Marc Zyngier
2014-06-25 9:28 ` [RFC PATCH 1/9] genirq: Add IRQD_IRQ_FORWARDED flag and accessors Marc Zyngier
2014-06-25 9:28 ` [RFC PATCH 2/9] genirq: Allow the state of a forwarded irq to be save/restored Marc Zyngier
2014-06-27 13:10 ` Will Deacon
2014-07-07 8:40 ` Marc Zyngier
2014-06-25 9:28 ` [RFC PATCH 3/9] irqchip: GIC: Convert to EOImode == 1 Marc Zyngier
2014-06-25 12:50 ` Rob Herring
2014-06-25 13:03 ` Marc Zyngier
2014-06-25 13:18 ` Rob Herring
2014-06-25 13:56 ` Anup Patel
2014-06-25 14:03 ` Ian Campbell
2014-06-25 14:31 ` Marc Zyngier
2014-06-25 14:08 ` Rob Herring
2014-06-25 14:24 ` Marc Zyngier [this message]
2014-06-25 14:27 ` Ian Campbell
2014-06-25 20:14 ` Joel Schopp
2014-06-30 19:09 ` Stefano Stabellini
2014-07-01 8:24 ` Marc Zyngier
2014-07-01 16:34 ` Stefano Stabellini
2014-07-01 16:42 ` Marc Zyngier
2014-06-25 14:06 ` Peter Maydell
2014-06-25 14:46 ` Marc Zyngier
2014-08-06 11:30 ` Christoffer Dall
2014-07-25 12:42 ` Eric Auger
2014-06-25 9:28 ` [RFC PATCH 4/9] irqchip: GIC: add support for forwarded interrupts Marc Zyngier
2014-06-27 13:17 ` Will Deacon
2014-07-07 10:43 ` Marc Zyngier
2014-08-06 11:30 ` Christoffer Dall
2014-06-25 9:28 ` [RFC PATCH 5/9] irqchip: GICv3: Convert to EOImode == 1 Marc Zyngier
2014-06-25 9:28 ` [RFC PATCH 6/9] irqchip: GICv3: add support for forwarded interrupts Marc Zyngier
2014-06-25 9:28 ` [RFC PATCH 7/9] KVM: arm: vgic: allow dynamic mapping of physical/virtual interrupts Marc Zyngier
2014-08-03 9:48 ` Eric Auger
2014-08-04 13:13 ` Marc Zyngier
2014-08-07 15:47 ` Eric Auger
2014-08-11 8:01 ` Christoffer Dall
2014-08-11 13:22 ` Eric Auger
2014-06-25 9:28 ` [RFC PATCH 8/9] arm: KVM: timer: move the timer switch into the non-preemptible section Marc Zyngier
2014-06-25 9:28 ` [RFC PATCH 9/9] KVM: arm: timer: make the interrupt state part of the timer state Marc Zyngier
2014-06-25 14:52 ` [RFC PATCH 0/9] ARM: Forwarding physical interrupts to a guest VM Eric Auger
2014-06-26 9:31 ` Marc Zyngier
2014-06-26 12:58 ` Eric Auger
2014-06-26 14:12 ` Marc Zyngier
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=53AADB82.3040808@arm.com \
--to=marc.zyngier@arm.com \
--cc=Catalin.Marinas@arm.com \
--cc=Will.Deacon@arm.com \
--cc=anup@brainfault.org \
--cc=christoffer.dall@linaro.org \
--cc=eric.auger@linaro.org \
--cc=ian.campbell@citrix.com \
--cc=kvmarm@lists.cs.columbia.edu \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=stefano.stabellini@eu.citrix.com \
--cc=tglx@linutronix.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).