| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Date: Tue, 4 Jan 2022 19:23:41 +0000
From: Sunil Muthuswamy <sunilmut@...rosoft.com>
To: "Michael Kelley (LINUX)" <mikelley@...rosoft.com>,
Sunil Muthuswamy <sunilmut@...ux.microsoft.com>,
KY Srinivasan <kys@...rosoft.com>,
Haiyang Zhang <haiyangz@...rosoft.com>,
Stephen Hemminger <sthemmin@...rosoft.com>,
"wei.liu@...nel.org" <wei.liu@...nel.org>,
"maz@...nel.org" <maz@...nel.org>,
Dexuan Cui <decui@...rosoft.com>,
"tglx@...utronix.de" <tglx@...utronix.de>,
"mingo@...hat.com" <mingo@...hat.com>,
"bp@...en8.de" <bp@...en8.de>, "hpa@...or.com" <hpa@...or.com>,
"lorenzo.pieralisi@....com" <lorenzo.pieralisi@....com>,
"robh@...nel.org" <robh@...nel.org>, "kw@...ux.com" <kw@...ux.com>,
"bhelgaas@...gle.com" <bhelgaas@...gle.com>,
"arnd@...db.de" <arnd@...db.de>
CC: "x86@...nel.org" <x86@...nel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"linux-hyperv@...r.kernel.org" <linux-hyperv@...r.kernel.org>,
"linux-pci@...r.kernel.org" <linux-pci@...r.kernel.org>,
"linux-arch@...r.kernel.org" <linux-arch@...r.kernel.org>
Subject: RE: [PATCH v7 2/2] PCI: hv: Add arm64 Hyper-V vPCI support
On Mon, 27 Dec 2021 17:38:07 +0000,
"Michael Kelley (LINUX)" <mikelley@...rosoft.com> wrote:
>
> From: Sunil Muthuswamy <sunilmut@...ux.microsoft.com> Sent: Friday,
> December 17, 2021 10:52 AM
> >
> > Add arm64 Hyper-V vPCI support by implementing the arch specific
> > interfaces. Introduce an IRQ domain and chip specific to Hyper-v vPCI that
> > is based on SPIs. The IRQ domain parents itself to the arch GIC IRQ domain
> > for basic vector management.
> >
> > Signed-off-by: Sunil Muthuswamy <sunilmut@...rosoft.com>
> > ---
> > In v2, v3, v4, v5, v6 & v7:
> > Changes are described in the cover letter.
> >
> > arch/arm64/include/asm/hyperv-tlfs.h | 9 +
> > drivers/pci/Kconfig | 2 +-
> > drivers/pci/controller/Kconfig | 2 +-
> > drivers/pci/controller/pci-hyperv.c | 241
> ++++++++++++++++++++++++++-
> > 4 files changed, 251 insertions(+), 3 deletions(-)
> >
> > diff --git a/arch/arm64/include/asm/hyperv-tlfs.h
> b/arch/arm64/include/asm/hyperv-tlfs.h
> > index 4d964a7f02ee..bc6c7ac934a1 100644
> > --- a/arch/arm64/include/asm/hyperv-tlfs.h
> > +++ b/arch/arm64/include/asm/hyperv-tlfs.h
> > @@ -64,6 +64,15 @@
> > #define HV_REGISTER_STIMER0_CONFIG 0x000B0000
> > #define HV_REGISTER_STIMER0_COUNT 0x000B0001
> >
> > +union hv_msi_entry {
> > + u64 as_uint64[2];
> > + struct {
> > + u64 address;
> > + u32 data;
> > + u32 reserved;
> > + } __packed;
> > +};
> > +
> > #include <asm-generic/hyperv-tlfs.h>
> >
> > #endif
> > diff --git a/drivers/pci/Kconfig b/drivers/pci/Kconfig
> > index 43e615aa12ff..d98fafdd0f99 100644
> > --- a/drivers/pci/Kconfig
> > +++ b/drivers/pci/Kconfig
> > @@ -184,7 +184,7 @@ config PCI_LABEL
> >
> > config PCI_HYPERV
> > tristate "Hyper-V PCI Frontend"
> > - depends on X86_64 && HYPERV && PCI_MSI &&
> PCI_MSI_IRQ_DOMAIN && SYSFS
> > + depends on ((X86 && X86_64) || ARM64) && HYPERV && PCI_MSI
> && PCI_MSI_IRQ_DOMAIN && SYSFS
> > select PCI_HYPERV_INTERFACE
> > help
> > The PCI device frontend driver allows the kernel to import arbitrary
> > diff --git a/drivers/pci/controller/Kconfig b/drivers/pci/controller/Kconfig
> > index 93b141110537..2536abcc045a 100644
> > --- a/drivers/pci/controller/Kconfig
> > +++ b/drivers/pci/controller/Kconfig
> > @@ -281,7 +281,7 @@ config PCIE_BRCMSTB
> >
> > config PCI_HYPERV_INTERFACE
> > tristate "Hyper-V PCI Interface"
> > - depends on X86 && HYPERV && PCI_MSI && PCI_MSI_IRQ_DOMAIN
> && X86_64
> > + depends on ((X86 && X86_64) || ARM64) && HYPERV && PCI_MSI
> && PCI_MSI_IRQ_DOMAIN
> > help
> > The Hyper-V PCI Interface is a helper driver allows other drivers to
> > have a common interface with the Hyper-V PCI frontend driver.
> > diff --git a/drivers/pci/controller/pci-hyperv.c b/drivers/pci/controller/pci-
> hyperv.c
> > index ead7d6cb6bf1..02ba2e7e2618 100644
> > --- a/drivers/pci/controller/pci-hyperv.c
> > +++ b/drivers/pci/controller/pci-hyperv.c
> > @@ -47,6 +47,8 @@
> > #include <linux/msi.h>
> > #include <linux/hyperv.h>
> > #include <linux/refcount.h>
> > +#include <linux/irqdomain.h>
> > +#include <linux/acpi.h>
> > #include <asm/mshyperv.h>
> >
> > /*
> > @@ -614,7 +616,236 @@ static int hv_msi_prepare(struct irq_domain
> *domain, struct device *dev,
> > {
> > return pci_msi_prepare(domain, dev, nvec, info);
> > }
> > -#endif /* CONFIG_X86 */
> > +#elif defined(CONFIG_ARM64)
> > +/*
> > + * SPI vectors to use for vPCI; arch SPIs range is [32, 1019], but leaving a bit
> > + * of room at the start to allow for SPIs to be specified through ACPI and
> > + * starting with a power of two to satisfy power of 2 multi-MSI
> requirement.
> > + */
> > +#define HV_PCI_MSI_SPI_START 64
> > +#define HV_PCI_MSI_SPI_NR (1020 - HV_PCI_MSI_SPI_START)
> > +#define DELIVERY_MODE 0
> > +#define FLOW_HANDLER NULL
> > +#define FLOW_NAME NULL
> > +#define hv_msi_prepare NULL
> > +
> > +struct hv_pci_chip_data {
> > + DECLARE_BITMAP(spi_map, HV_PCI_MSI_SPI_NR);
> > + struct mutex map_lock;
> > +};
> > +
> > +/* Hyper-V vPCI MSI GIC IRQ domain */
> > +static struct irq_domain *hv_msi_gic_irq_domain;
> > +
> > +/* Hyper-V PCI MSI IRQ chip */
> > +static struct irq_chip hv_arm64_msi_irq_chip = {
> > + .name = "MSI",
> > + .irq_set_affinity = irq_chip_set_affinity_parent,
> > + .irq_eoi = irq_chip_eoi_parent,
> > + .irq_mask = irq_chip_mask_parent,
> > + .irq_unmask = irq_chip_unmask_parent
> > +};
> > +
> > +static unsigned int hv_msi_get_int_vector(struct irq_data *irqd)
> > +{
> > + return irqd->parent_data->hwirq;
> > +}
> > +
> > +static void hv_set_msi_entry_from_desc(union hv_msi_entry
> *msi_entry,
> > + struct msi_desc *msi_desc)
> > +{
> > + msi_entry->address = ((u64)msi_desc->msg.address_hi << 32) |
> > + msi_desc->msg.address_lo;
> > + msi_entry->data = msi_desc->msg.data;
> > +}
> > +
> > +/*
> > + * @nr_bm_irqs: Indicates the number of IRQs that were
> allocated from
> > + * the bitmap.
> > + * @nr_dom_irqs: Indicates the number of IRQs that were allocated
> from
> > + * the parent domain.
> > + */
> > +static void hv_pci_vec_irq_free(struct irq_domain *domain,
> > + unsigned int virq,
> > + unsigned int nr_bm_irqs,
> > + unsigned int nr_dom_irqs)
> > +{
> > + struct hv_pci_chip_data *chip_data = domain->host_data;
> > + struct irq_data *d = irq_domain_get_irq_data(domain, virq);
>
> FWIW, irq_domain_get_irq_data() can return NULL. Maybe that's an
> error in the "should never happen" category. Throughout kernel code,
> some callers check for a NULL result, but a lot do not.
>
Marc's response to this comment makes sense. Will keep this as is, going
by that.
> > + int first = d->hwirq - HV_PCI_MSI_SPI_START;
> > + int i;
> > +
> > + mutex_lock(&chip_data->map_lock);
> > + bitmap_release_region(chip_data->spi_map,
> > + first,
> > + get_count_order(nr_bm_irqs));
> > + mutex_unlock(&chip_data->map_lock);
> > + for (i = 0; i < nr_dom_irqs; i++) {
> > + if (i)
> > + d = irq_domain_get_irq_data(domain, virq + i);
>
> Same here.
>
Same response as above.
> > + irq_domain_reset_irq_data(d);
> > + }
> > +
> > + irq_domain_free_irqs_parent(domain, virq, nr_dom_irqs);
> > +}
> > +
> > +static void hv_pci_vec_irq_domain_free(struct irq_domain *domain,
> > + unsigned int virq,
> > + unsigned int nr_irqs)
> > +{
> > + hv_pci_vec_irq_free(domain, virq, nr_irqs, nr_irqs);
> > +}
> > +
> > +static int hv_pci_vec_alloc_device_irq(struct irq_domain *domain,
> > + unsigned int nr_irqs,
> > + irq_hw_number_t *hwirq)
> > +{
> > + struct hv_pci_chip_data *chip_data = domain->host_data;
> > + unsigned int index;
> > +
> > + /* Find and allocate region from the SPI bitmap */
> > + mutex_lock(&chip_data->map_lock);
> > + index = bitmap_find_free_region(chip_data->spi_map,
> > + HV_PCI_MSI_SPI_NR,
> > + get_count_order(nr_irqs));
> > + mutex_unlock(&chip_data->map_lock);
> > + if (index < 0)
> > + return -ENOSPC;
> > +
> > + *hwirq = index + HV_PCI_MSI_SPI_START;
> > +
> > + return 0;
> > +}
> > +
> > +static int hv_pci_vec_irq_gic_domain_alloc(struct irq_domain *domain,
> > + unsigned int virq,
> > + irq_hw_number_t hwirq)
> > +{
> > + struct irq_fwspec fwspec;
> > + struct irq_data *d;
> > + int ret;
> > +
> > + fwspec.fwnode = domain->parent->fwnode;
> > + fwspec.param_count = 2;
> > + fwspec.param[0] = hwirq;
> > + fwspec.param[1] = IRQ_TYPE_EDGE_RISING;
> > +
> > + ret = irq_domain_alloc_irqs_parent(domain, virq, 1, &fwspec);
> > + if (ret)
> > + return ret;
> > +
> > + /*
> > + * Since the interrupt specifier is not coming from ACPI or DT, the
> > + * trigger type will need to be set explicitly. Otherwise, it will be
> > + * set to whatever is in the GIC configuration.
> > + */
> > + d = irq_domain_get_irq_data(domain->parent, virq);
>
> And here.
>
Same response as above.
> > +
> > + return d->chip->irq_set_type(d, IRQ_TYPE_EDGE_RISING);
> > +}
> > +
> > +static int hv_pci_vec_irq_domain_alloc(struct irq_domain *domain,
> > + unsigned int virq, unsigned int nr_irqs,
> > + void *args)
> > +{
> > + irq_hw_number_t hwirq;
> > + unsigned int i;
> > + int ret;
> > +
> > + ret = hv_pci_vec_alloc_device_irq(domain, nr_irqs, &hwirq);
> > + if (ret)
> > + return ret;
> > +
> > + for (i = 0; i < nr_irqs; i++) {
> > + ret = hv_pci_vec_irq_gic_domain_alloc(domain, virq + i,
> > + hwirq + i);
> > + if (ret)
> > + goto free_irq;
> > +
> > + ret = irq_domain_set_hwirq_and_chip(domain, virq + i,
> > + hwirq + i,
> > + &hv_arm64_msi_irq_chip,
> > + domain->host_data);
> > + if (ret)
> > + goto free_irq;
>
> This error path doesn't clean up correctly. While parent IRQs allocated
> in previous iterations of the loop is cleaned up, the parent IRQ
> allocated in the current iteration is not.
>
'irq_domain_set_hwirq_and_chip' really shouldn't fail. If you still feel
that we should address this, I can.
> > +
> > + pr_debug("pID:%d vID:%u\n", (int)(hwirq + i), virq + i);
> > + }
> > +
> > + return 0;
> > +
> > +free_irq:
> > + hv_pci_vec_irq_free(domain, virq, nr_irqs, i);
> > +
> > + return ret;
> > +}
> > +
> > +/*
> > + * Pick the first online cpu as the irq affinity that can be temporarily used
> > + * for composing MSI from the hypervisor. GIC will eventually set the right
> > + * affinity for the irq and the 'unmask' will retarget the interrupt to that
> > + * cpu.
> > + */
> > +static int hv_pci_vec_irq_domain_activate(struct irq_domain *domain,
> > + struct irq_data *irqd, bool reserve)
> > +{
> > + int cpu = cpumask_first(cpu_online_mask);
>
> Using the cpu_online_mask may work correctly, though its usage here
> raises a red flag for me. The problem is that the CPU selected can go offline
> immediately after the above line of code, as the cpus_read_lock() does
> not appear to be held anywhere in the call stack. If the CPU is only
> a temporary placeholder and will never actually be targeted with the
> interrupt, then maybe the online/offline state doesn't matter. But if
> that's the case, I'd suggest using the cpu_present_mask instead of the
> cpu_online_mask. Hyper-V doesn't hot-add CPUs, so cpu_present_mask
> should be stable even if the cpus_read_lock() isn't held, and the code
> doesn't incorrectly imply that it's important for the CPU to be online.
>
> Michael
>
Thanks. I can change this to 'cpu_present_mask' in the next iteration.
- Sunil
Powered by blists - more mailing lists