| Message ID | 1461863358-12496-1-git-send-email-keith.busch@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Delegated to: | Bjorn Helgaas |
| Headers | show |
Hi Keith, On Thu, Apr 28, 2016 at 11:09:18AM -0600, Keith Busch wrote: > This adds driver support for root and downstream ports that implement > the PCI-Express Downstream Port Containment extended capability. DPC is > an optional capability to contain uncorrectable errors below a port. > > When a DPC event is triggered, the h/w disables downstream links, so > the DPC driver schedules removal for all devices below this port. This > may happen concurrently with a PCI-e hotplug driver if enabled. When all > downstream devices are removed and the link state transitions to disabled, > the DPC driver clears the DPC status and interrupt bits so the link may > retrain for a newly connected device. > > The pcie device naming is updated to accomodate the additional service > driver. From Lukas Wunner <lukas@wunner.de>: > > The names of port service devices previously used one nibble to encode > the port type and another nibble to encode the service type. Since this > commit introduces a fifth service type, it changes device names to use > one *byte* to encode the service type. E.g. a hotplug port service on a > downstream bridge was previously called pcie24 and is now called pcie204. Can you include a pointer to the DPC spec here? > Signed-off-by: Keith Busch <keith.busch@intel.com> > Cc: Lukas Wunner <lukas@wunner.de> > --- > v2 -> v3: > > Device naming fixes and changelog from Lukas. > > Using tristate for driver instead of boolean, and update Kconfig > description. > > drivers/pci/pcie/Kconfig | 1 + > drivers/pci/pcie/Makefile | 1 + > drivers/pci/pcie/dpc/Kconfig | 18 ++++ > drivers/pci/pcie/dpc/Makefile | 5 ++ > drivers/pci/pcie/dpc/dpcdrv.c | 182 ++++++++++++++++++++++++++++++++++++++++ Seems overkill to add a new directory for a single file. Can you just put it in drivers/pci/pcie and fold the Makefile and Kconfig changes into the existing files? The module is named "dpcdrv" -- that will show up in lsmod and oops output, and I'm not sure that's as helpful as it could be. The "drv" part doesn't really say anything useful, and there's no hint about it being PCI-related. Maybe "pcie-dpc"? > drivers/pci/pcie/portdrv.h | 4 +- > drivers/pci/pcie/portdrv_acpi.c | 2 +- > drivers/pci/pcie/portdrv_core.c | 6 +- > include/linux/pcieport_if.h | 2 + > include/uapi/linux/pci_regs.h | 27 +++++- > 10 files changed, 242 insertions(+), 6 deletions(-) > create mode 100644 drivers/pci/pcie/dpc/Kconfig > create mode 100644 drivers/pci/pcie/dpc/Makefile > create mode 100644 drivers/pci/pcie/dpc/dpcdrv.c > > diff --git a/drivers/pci/pcie/Kconfig b/drivers/pci/pcie/Kconfig > index e294713..bd9fca7 100644 > --- a/drivers/pci/pcie/Kconfig > +++ b/drivers/pci/pcie/Kconfig > @@ -23,6 +23,7 @@ config HOTPLUG_PCI_PCIE > When in doubt, say N. > > source "drivers/pci/pcie/aer/Kconfig" > +source "drivers/pci/pcie/dpc/Kconfig" > > # > # PCI Express ASPM > diff --git a/drivers/pci/pcie/Makefile b/drivers/pci/pcie/Makefile > index 00c62df..18604a6 100644 > --- a/drivers/pci/pcie/Makefile > +++ b/drivers/pci/pcie/Makefile > @@ -12,5 +12,6 @@ obj-$(CONFIG_PCIEPORTBUS) += pcieportdrv.o > > # Build PCI Express AER if needed > obj-$(CONFIG_PCIEAER) += aer/ > +obj-$(CONFIG_PCIEDPC) += dpc/ > > obj-$(CONFIG_PCIE_PME) += pme.o > diff --git a/drivers/pci/pcie/dpc/Kconfig b/drivers/pci/pcie/dpc/Kconfig > new file mode 100644 > index 0000000..384202f > --- /dev/null > +++ b/drivers/pci/pcie/dpc/Kconfig > @@ -0,0 +1,18 @@ > +# > +# PCI Express Device DPC Configuration > +# > + > +config PCIEDPC > + tristate "PCI-e Downstream Port Containment support" > + depends on PCIEPORTBUS > + default n > + help > + This enables PCI-Express Downstream Port Containment (DPC) s/PCI-e/PCIe/ s/PCI-Express/PCI Express/ (you have both variants below) > + driver support. DPC events from Root and Downstream ports > + will be handled by the DPC driver. If your system doesn't > + have this capability or you do not want to use this feature, > + it is safe to answer N. > + > + To compile this driver as a module, choose M here: the module > + will be called dpcdrv. > + > diff --git a/drivers/pci/pcie/dpc/Makefile b/drivers/pci/pcie/dpc/Makefile > new file mode 100644 > index 0000000..60273ec > --- /dev/null > +++ b/drivers/pci/pcie/dpc/Makefile > @@ -0,0 +1,5 @@ > +# > +# Makefile for PCI-Express Downstream Port Containment Driver > +# > + > +obj-$(CONFIG_PCIEDPC) += dpcdrv.o > diff --git a/drivers/pci/pcie/dpc/dpcdrv.c b/drivers/pci/pcie/dpc/dpcdrv.c > new file mode 100644 > index 0000000..8cba6ef > --- /dev/null > +++ b/drivers/pci/pcie/dpc/dpcdrv.c > @@ -0,0 +1,182 @@ > +/* > + * PCI-Express Downstream Port Containment services driver > + * Copyright (C) 2016 Intel Corp. > + * > + * This file is subject to the terms and conditions of the GNU General Public > + * License. See the file "COPYING" in the main directory of this archive > + * for more details. > + */ > + > +#include <linux/delay.h> > +#include <linux/interrupt.h> > +#include <linux/module.h> > +#include <linux/pci.h> > +#include <linux/pcieport_if.h> > + > +struct event_info { > + struct pcie_device *dev; > + struct work_struct work; > +}; > + > +static void dpc_wait_link_inactive(struct pci_dev *pdev) > +{ > + unsigned long timeout = jiffies + HZ; > + u16 lnk_status; > + > + pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status); > + while (lnk_status & PCI_EXP_LNKSTA_DLLLA && > + !time_after(jiffies, timeout)) { > + msleep(10); > + pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status); > + } > + if (lnk_status & PCI_EXP_LNKSTA_DLLLA) > + dev_warn(&pdev->dev, "Link state not disabled for DPC"); > +} > + > +static void interrupt_event_handler(struct work_struct *work) > +{ > + int pos; > + > + struct event_info *info = container_of(work, struct event_info, work); > + struct pci_dev *dev, *temp, *pdev = info->dev->port; > + struct pci_bus *parent = pdev->subordinate; > + > + pci_lock_rescan_remove(); > + list_for_each_entry_safe_reverse(dev, temp, &parent->devices, > + bus_list) { > + pci_dev_get(dev); > + pci_stop_and_remove_bus_device(dev); > + pci_dev_put(dev); > + } > + pci_unlock_rescan_remove(); > + > + dpc_wait_link_inactive(pdev); > + > + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); > + pci_write_config_word(pdev, pos + PCI_EXP_DPC_STATUS, > + PCI_EXP_DPC_STATUS_TRIGGER | PCI_EXP_DPC_STATUS_INTERRUPT); > + > + kfree(info); > +} > + > +static void dpc_queue_event(struct pcie_device *dev) > +{ > + struct event_info *info; > + > + info = kmalloc(sizeof(*info), GFP_ATOMIC); Use kzalloc(). This should be a low-frequency path, and it's a hassle to verify that we initialize everything, and it's easy to add a field and forget to add the initializer. > + if (!info) { > + dev_warn(&dev->device, "dropped containment event\n"); > + return; > + } > + > + INIT_WORK(&info->work, interrupt_event_handler); > + info->dev = dev; > + > + schedule_work(&info->work); > +} > + > +static irqreturn_t dpc_irq(int irq, void *context) > +{ > + int pos; > + u16 status, source; > + > + struct pcie_device *dev = (struct pcie_device *)context; > + struct pci_dev *pdev = dev->port; > + > + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); I think you should cache the capability offset at probe-time instead of searching for it every time you get an IRQ. I know this is low-frequency also, so it's not a performance issue, but it just feels clunky. > + pci_read_config_word(pdev, pos + PCI_EXP_DPC_STATUS, &status); > + pci_read_config_word(pdev, pos + PCI_EXP_DPC_SOURCE_ID, &source); > + > + if (!status) > + return IRQ_NONE; > + > + dev_warn(&dev->device, "dpc status:%04x source:%04x\n", status, source); Could use more informative text here. This will show up in dmesg, followed by devices mysteriously (to the user) disappearing. > + > + if (status & PCI_EXP_DPC_STATUS_TRIGGER) > + dpc_queue_event(dev); > + > + return IRQ_HANDLED; > +} > + > +#define FLAG(x, y) (((x) & (y)) ? '+' : '-') > + > +static void dpc_enable_port(struct pcie_device *dev) Not sure splitting dpc_enable_port() and dpc_disable_port() out into their own functions is really worthwhile. They're only called from one place, and the callers are trivial anyway. > +{ > + struct pci_dev *pdev = dev->port; > + int pos; > + u16 ctl, cap; > + > + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); > + pci_read_config_word(pdev, pos + PCI_EXP_DPC_CAP, &cap); > + pci_read_config_word(pdev, pos + PCI_EXP_DPC_CTL, &ctl); > + > + ctl |= PCI_EXP_DPC_CTL_EN_NONFATAL | PCI_EXP_DPC_CTL_INT_EN; > + pci_write_config_word(pdev, pos + PCI_EXP_DPC_CTL, ctl); > + > + dev_info(&dev->device, > + "DPC Int Msg #%d, RPExt%c PoisonedTLP%c SwTrigger%c RP PIO Log %d, DL_ActiveErr%c\n", Maybe a more informative message here too? "DPC error containment" or something? > + cap & 0xf, FLAG(cap, PCI_EXP_DPC_CAP_RP_EXT), > + FLAG(cap, PCI_EXP_DPC_CAP_POISONED_TLP), > + FLAG(cap, PCI_EXP_DPC_CAP_SW_TRIGGER), (cap >> 8) & 0xf, > + FLAG(cap, PCI_EXP_DPC_CAP_DL_ACTIVE)); > +} > + > +static void dpc_disable_port(struct pcie_device *dev) > +{ > + struct pci_dev *pdev = dev->port; > + int pos; > + u16 ctl; > + > + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); > + pci_read_config_word(pdev, pos + PCI_EXP_DPC_CTL, &ctl); > + > + ctl |= ~(PCI_EXP_DPC_CTL_EN_NONFATAL | PCI_EXP_DPC_CTL_INT_EN); > + pci_write_config_word(pdev, pos + PCI_EXP_DPC_CTL, ctl); > +} > + > +static int dpc_probe(struct pcie_device *dev) > +{ > + int status; > + > + status = request_irq(dev->irq, dpc_irq, IRQF_SHARED, "dpcdrv", dev); > + if (status) { > + dev_warn(&dev->device, "request IRQ failed\n"); You have the IRQ, so why not include it in the warning message? > + return status; > + } > + dpc_enable_port(dev); > + > + return status; > +} > + > +static void dpc_remove(struct pcie_device *dev) > +{ > + dpc_disable_port(dev); > + free_irq(dev->irq, dev); > +} > + > +static struct pcie_port_service_driver dpcdriver = { > + .name = "pciedpc", > + .port_type = PCI_EXP_TYPE_ROOT_PORT | PCI_EXP_TYPE_DOWNSTREAM, > + .service = PCIE_PORT_SERVICE_DPC, > + .probe = dpc_probe, > + .remove = dpc_remove, > +}; > + > +static int __init dpc_service_init(void) > +{ > + /* XXX: Add kernel parameters to control PCIe DPC module */ > + return pcie_port_service_register(&dpcdriver); > +} > + > +static void __exit dpc_service_exit(void) > +{ > + pcie_port_service_unregister(&dpcdriver); > +} > + > +MODULE_DESCRIPTION("PCI Express Downstream Port Containment driver"); > +MODULE_AUTHOR("Keith Busch <keith.busch@intel.com>"); > +MODULE_LICENSE("GPL"); > +MODULE_VERSION("0.1"); > + > +module_init(dpc_service_init); > +module_exit(dpc_service_exit); > diff --git a/drivers/pci/pcie/portdrv.h b/drivers/pci/pcie/portdrv.h > index d525548..7d82f6d 100644 > --- a/drivers/pci/pcie/portdrv.h > +++ b/drivers/pci/pcie/portdrv.h > @@ -11,14 +11,14 @@ > > #include <linux/compiler.h> > > -#define PCIE_PORT_DEVICE_MAXSERVICES 4 > +#define PCIE_PORT_DEVICE_MAXSERVICES 5 > /* > * According to the PCI Express Base Specification 2.0, the indices of > * the MSI-X table entries used by port services must not exceed 31 > */ > #define PCIE_PORT_MAX_MSIX_ENTRIES 32 > > -#define get_descriptor_id(type, service) (((type - 4) << 4) | service) > +#define get_descriptor_id(type, service) (((type - 4) << 8) | service) > > extern struct bus_type pcie_port_bus_type; > int pcie_port_device_register(struct pci_dev *dev); > diff --git a/drivers/pci/pcie/portdrv_acpi.c b/drivers/pci/pcie/portdrv_acpi.c > index b4d2894..44296eb 100644 > --- a/drivers/pci/pcie/portdrv_acpi.c > +++ b/drivers/pci/pcie/portdrv_acpi.c > @@ -51,7 +51,7 @@ int pcie_port_acpi_setup(struct pci_dev *port, int *srv_mask) > > flags = root->osc_control_set; > > - *srv_mask = PCIE_PORT_SERVICE_VC; > + *srv_mask = PCIE_PORT_SERVICE_VC | PCIE_PORT_SERVICE_DPC; > if (flags & OSC_PCI_EXPRESS_NATIVE_HP_CONTROL) > *srv_mask |= PCIE_PORT_SERVICE_HP; > if (flags & OSC_PCI_EXPRESS_PME_CONTROL) > diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c > index 88122dc..2ab0f42 100644 > --- a/drivers/pci/pcie/portdrv_core.c > +++ b/drivers/pci/pcie/portdrv_core.c > @@ -262,7 +262,7 @@ static int get_port_device_capability(struct pci_dev *dev) > return 0; > > cap_mask = PCIE_PORT_SERVICE_PME | PCIE_PORT_SERVICE_HP > - | PCIE_PORT_SERVICE_VC; > + | PCIE_PORT_SERVICE_VC | PCIE_PORT_SERVICE_DPC; > if (pci_aer_available()) > cap_mask |= PCIE_PORT_SERVICE_AER; > > @@ -311,6 +311,8 @@ static int get_port_device_capability(struct pci_dev *dev) > */ > pcie_pme_interrupt_enable(dev, false); > } > + if (pci_find_ext_capability(dev, PCI_EXT_CAP_ID_DPC)) > + services |= PCIE_PORT_SERVICE_DPC; > > return services; > } > @@ -338,7 +340,7 @@ static int pcie_device_init(struct pci_dev *pdev, int service, int irq) > device = &pcie->device; > device->bus = &pcie_port_bus_type; > device->release = release_pcie_device; /* callback to free pcie dev */ > - dev_set_name(device, "%s:pcie%02x", > + dev_set_name(device, "%s:pcie%03x", > pci_name(pdev), > get_descriptor_id(pci_pcie_type(pdev), service)); > device->parent = &pdev->dev; > diff --git a/include/linux/pcieport_if.h b/include/linux/pcieport_if.h > index 4f1089f..afcd130 100644 > --- a/include/linux/pcieport_if.h > +++ b/include/linux/pcieport_if.h > @@ -21,6 +21,8 @@ > #define PCIE_PORT_SERVICE_HP (1 << PCIE_PORT_SERVICE_HP_SHIFT) > #define PCIE_PORT_SERVICE_VC_SHIFT 3 /* Virtual Channel */ > #define PCIE_PORT_SERVICE_VC (1 << PCIE_PORT_SERVICE_VC_SHIFT) > +#define PCIE_PORT_SERVICE_DPC_SHIFT 4 /* Downstream Port Containment */ > +#define PCIE_PORT_SERVICE_DPC (1 << PCIE_PORT_SERVICE_DPC_SHIFT) > > struct pcie_device { > int irq; /* Service IRQ/MSI/MSI-X Vector */ > diff --git a/include/uapi/linux/pci_regs.h b/include/uapi/linux/pci_regs.h > index 1becea8..c008294 100644 > --- a/include/uapi/linux/pci_regs.h > +++ b/include/uapi/linux/pci_regs.h > @@ -670,7 +670,8 @@ > #define PCI_EXT_CAP_ID_SECPCI 0x19 /* Secondary PCIe Capability */ > #define PCI_EXT_CAP_ID_PMUX 0x1A /* Protocol Multiplexing */ > #define PCI_EXT_CAP_ID_PASID 0x1B /* Process Address Space ID */ > -#define PCI_EXT_CAP_ID_MAX PCI_EXT_CAP_ID_PASID > +#define PCI_EXT_CAP_ID_DPC 0x1D /* Downstream Port Containment */ > +#define PCI_EXT_CAP_ID_MAX PCI_EXT_CAP_ID_DPC > > #define PCI_EXT_CAP_DSN_SIZEOF 12 > #define PCI_EXT_CAP_MCAST_ENDPOINT_SIZEOF 40 > @@ -946,4 +947,28 @@ > #define PCI_TPH_CAP_ST_SHIFT 16 /* st table shift */ > #define PCI_TPH_BASE_SIZEOF 12 /* size with no st table */ > > +/* Downstream Port Containment */ > +#define PCI_EXP_DPC_CAP 4 /* DPC Capability */ > +#define PCI_EXP_DPC_CAP_RP_EXT 0x20 /* Root Port Extensions for DPC */ > +#define PCI_EXP_DPC_CAP_POISONED_TLP 0x40 /* Poisoned TLP Egress Blocking Supported */ > +#define PCI_EXP_DPC_CAP_SW_TRIGGER 0x80 /* Software Triggering Supported */ > +#define PCI_EXP_DPC_CAP_DL_ACTIVE 0x1000 /* ERR_COR signal on DL_Active supported */ > + > +#define PCI_EXP_DPC_CTL 6 /* DPC control */ > +#define PCI_EXP_DPC_CTL_DISABLE 0x00 /* Disable trigger */ Only define the bits you're using. Then it's a little easier to figure out whether something is supported by Linux. > +#define PCI_EXP_DPC_CTL_EN_FATAL 0x01 /* Enable trigger on ERR_FATAL message */ > +#define PCI_EXP_DPC_CTL_EN_NONFATAL 0x02 /* Enable trigger on ERR_NONFATAL message */ > +#define PCI_EXP_DPC_CTL_UR 0x04 /* Unsupported Request Completion Status */ > +#define PCI_EXP_DPC_CTL_INT_EN 0x08 /* DPC Interrupt Enable */ > +#define PCI_EXP_DPC_CTL_ERR_COR_EN 0x10 /* Enable ERRO_COR Message on DPC triggered */ > +#define PCI_EXP_DPC_CTL_POSIONED_TLP 0x20 /* Enabled Poisoned TLP Egress Blocking */ And it helps avoid typos like this one (POSIONED :)) > +#define PCI_EXP_DPC_CTL_SW_T_EN 0x40 /* Trigger DPC Status */ > +#define PCI_EXP_DPC_CTL_DL_A_ERR_COR 0x80 /* Signal ER_COR on DL_Active */ > + > +#define PCI_EXP_DPC_STATUS 8 /* DPC Status */ > +#define PCI_EXP_DPC_STATUS_TRIGGER 0x01 /* Trigger Status */ > +#define PCI_EXP_DPC_STATUS_INTERRUPT 0x08 /* Interrupt Status */ > + > +#define PCI_EXP_DPC_SOURCE_ID 0x0A > + > #endif /* LINUX_PCI_REGS_H */ > -- > 2.7.2 > > -- > To unsubscribe from this list: send the line "unsubscribe linux-pci" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/pci/pcie/Kconfig b/drivers/pci/pcie/Kconfig index e294713..bd9fca7 100644 --- a/drivers/pci/pcie/Kconfig +++ b/drivers/pci/pcie/Kconfig @@ -23,6 +23,7 @@ config HOTPLUG_PCI_PCIE When in doubt, say N. source "drivers/pci/pcie/aer/Kconfig" +source "drivers/pci/pcie/dpc/Kconfig" # # PCI Express ASPM diff --git a/drivers/pci/pcie/Makefile b/drivers/pci/pcie/Makefile index 00c62df..18604a6 100644 --- a/drivers/pci/pcie/Makefile +++ b/drivers/pci/pcie/Makefile @@ -12,5 +12,6 @@ obj-$(CONFIG_PCIEPORTBUS) += pcieportdrv.o # Build PCI Express AER if needed obj-$(CONFIG_PCIEAER) += aer/ +obj-$(CONFIG_PCIEDPC) += dpc/ obj-$(CONFIG_PCIE_PME) += pme.o diff --git a/drivers/pci/pcie/dpc/Kconfig b/drivers/pci/pcie/dpc/Kconfig new file mode 100644 index 0000000..384202f --- /dev/null +++ b/drivers/pci/pcie/dpc/Kconfig @@ -0,0 +1,18 @@ +# +# PCI Express Device DPC Configuration +# + +config PCIEDPC + tristate "PCI-e Downstream Port Containment support" + depends on PCIEPORTBUS + default n + help + This enables PCI-Express Downstream Port Containment (DPC) + driver support. DPC events from Root and Downstream ports + will be handled by the DPC driver. If your system doesn't + have this capability or you do not want to use this feature, + it is safe to answer N. + + To compile this driver as a module, choose M here: the module + will be called dpcdrv. + diff --git a/drivers/pci/pcie/dpc/Makefile b/drivers/pci/pcie/dpc/Makefile new file mode 100644 index 0000000..60273ec --- /dev/null +++ b/drivers/pci/pcie/dpc/Makefile @@ -0,0 +1,5 @@ +# +# Makefile for PCI-Express Downstream Port Containment Driver +# + +obj-$(CONFIG_PCIEDPC) += dpcdrv.o diff --git a/drivers/pci/pcie/dpc/dpcdrv.c b/drivers/pci/pcie/dpc/dpcdrv.c new file mode 100644 index 0000000..8cba6ef --- /dev/null +++ b/drivers/pci/pcie/dpc/dpcdrv.c @@ -0,0 +1,182 @@ +/* + * PCI-Express Downstream Port Containment services driver + * Copyright (C) 2016 Intel Corp. + * + * This file is subject to the terms and conditions of the GNU General Public + * License. See the file "COPYING" in the main directory of this archive + * for more details. + */ + +#include <linux/delay.h> +#include <linux/interrupt.h> +#include <linux/module.h> +#include <linux/pci.h> +#include <linux/pcieport_if.h> + +struct event_info { + struct pcie_device *dev; + struct work_struct work; +}; + +static void dpc_wait_link_inactive(struct pci_dev *pdev) +{ + unsigned long timeout = jiffies + HZ; + u16 lnk_status; + + pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status); + while (lnk_status & PCI_EXP_LNKSTA_DLLLA && + !time_after(jiffies, timeout)) { + msleep(10); + pcie_capability_read_word(pdev, PCI_EXP_LNKSTA, &lnk_status); + } + if (lnk_status & PCI_EXP_LNKSTA_DLLLA) + dev_warn(&pdev->dev, "Link state not disabled for DPC"); +} + +static void interrupt_event_handler(struct work_struct *work) +{ + int pos; + + struct event_info *info = container_of(work, struct event_info, work); + struct pci_dev *dev, *temp, *pdev = info->dev->port; + struct pci_bus *parent = pdev->subordinate; + + pci_lock_rescan_remove(); + list_for_each_entry_safe_reverse(dev, temp, &parent->devices, + bus_list) { + pci_dev_get(dev); + pci_stop_and_remove_bus_device(dev); + pci_dev_put(dev); + } + pci_unlock_rescan_remove(); + + dpc_wait_link_inactive(pdev); + + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); + pci_write_config_word(pdev, pos + PCI_EXP_DPC_STATUS, + PCI_EXP_DPC_STATUS_TRIGGER | PCI_EXP_DPC_STATUS_INTERRUPT); + + kfree(info); +} + +static void dpc_queue_event(struct pcie_device *dev) +{ + struct event_info *info; + + info = kmalloc(sizeof(*info), GFP_ATOMIC); + if (!info) { + dev_warn(&dev->device, "dropped containment event\n"); + return; + } + + INIT_WORK(&info->work, interrupt_event_handler); + info->dev = dev; + + schedule_work(&info->work); +} + +static irqreturn_t dpc_irq(int irq, void *context) +{ + int pos; + u16 status, source; + + struct pcie_device *dev = (struct pcie_device *)context; + struct pci_dev *pdev = dev->port; + + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); + pci_read_config_word(pdev, pos + PCI_EXP_DPC_STATUS, &status); + pci_read_config_word(pdev, pos + PCI_EXP_DPC_SOURCE_ID, &source); + + if (!status) + return IRQ_NONE; + + dev_warn(&dev->device, "dpc status:%04x source:%04x\n", status, source); + + if (status & PCI_EXP_DPC_STATUS_TRIGGER) + dpc_queue_event(dev); + + return IRQ_HANDLED; +} + +#define FLAG(x, y) (((x) & (y)) ? '+' : '-') + +static void dpc_enable_port(struct pcie_device *dev) +{ + struct pci_dev *pdev = dev->port; + int pos; + u16 ctl, cap; + + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); + pci_read_config_word(pdev, pos + PCI_EXP_DPC_CAP, &cap); + pci_read_config_word(pdev, pos + PCI_EXP_DPC_CTL, &ctl); + + ctl |= PCI_EXP_DPC_CTL_EN_NONFATAL | PCI_EXP_DPC_CTL_INT_EN; + pci_write_config_word(pdev, pos + PCI_EXP_DPC_CTL, ctl); + + dev_info(&dev->device, + "DPC Int Msg #%d, RPExt%c PoisonedTLP%c SwTrigger%c RP PIO Log %d, DL_ActiveErr%c\n", + cap & 0xf, FLAG(cap, PCI_EXP_DPC_CAP_RP_EXT), + FLAG(cap, PCI_EXP_DPC_CAP_POISONED_TLP), + FLAG(cap, PCI_EXP_DPC_CAP_SW_TRIGGER), (cap >> 8) & 0xf, + FLAG(cap, PCI_EXP_DPC_CAP_DL_ACTIVE)); +} + +static void dpc_disable_port(struct pcie_device *dev) +{ + struct pci_dev *pdev = dev->port; + int pos; + u16 ctl; + + pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_DPC); + pci_read_config_word(pdev, pos + PCI_EXP_DPC_CTL, &ctl); + + ctl |= ~(PCI_EXP_DPC_CTL_EN_NONFATAL | PCI_EXP_DPC_CTL_INT_EN); + pci_write_config_word(pdev, pos + PCI_EXP_DPC_CTL, ctl); +} + +static int dpc_probe(struct pcie_device *dev) +{ + int status; + + status = request_irq(dev->irq, dpc_irq, IRQF_SHARED, "dpcdrv", dev); + if (status) { + dev_warn(&dev->device, "request IRQ failed\n"); + return status; + } + dpc_enable_port(dev); + + return status; +} + +static void dpc_remove(struct pcie_device *dev) +{ + dpc_disable_port(dev); + free_irq(dev->irq, dev); +} + +static struct pcie_port_service_driver dpcdriver = { + .name = "pciedpc", + .port_type = PCI_EXP_TYPE_ROOT_PORT | PCI_EXP_TYPE_DOWNSTREAM, + .service = PCIE_PORT_SERVICE_DPC, + .probe = dpc_probe, + .remove = dpc_remove, +}; + +static int __init dpc_service_init(void) +{ + /* XXX: Add kernel parameters to control PCIe DPC module */ + return pcie_port_service_register(&dpcdriver); +} + +static void __exit dpc_service_exit(void) +{ + pcie_port_service_unregister(&dpcdriver); +} + +MODULE_DESCRIPTION("PCI Express Downstream Port Containment driver"); +MODULE_AUTHOR("Keith Busch <keith.busch@intel.com>"); +MODULE_LICENSE("GPL"); +MODULE_VERSION("0.1"); + +module_init(dpc_service_init); +module_exit(dpc_service_exit); diff --git a/drivers/pci/pcie/portdrv.h b/drivers/pci/pcie/portdrv.h index d525548..7d82f6d 100644 --- a/drivers/pci/pcie/portdrv.h +++ b/drivers/pci/pcie/portdrv.h @@ -11,14 +11,14 @@ #include <linux/compiler.h> -#define PCIE_PORT_DEVICE_MAXSERVICES 4 +#define PCIE_PORT_DEVICE_MAXSERVICES 5 /* * According to the PCI Express Base Specification 2.0, the indices of * the MSI-X table entries used by port services must not exceed 31 */ #define PCIE_PORT_MAX_MSIX_ENTRIES 32 -#define get_descriptor_id(type, service) (((type - 4) << 4) | service) +#define get_descriptor_id(type, service) (((type - 4) << 8) | service) extern struct bus_type pcie_port_bus_type; int pcie_port_device_register(struct pci_dev *dev); diff --git a/drivers/pci/pcie/portdrv_acpi.c b/drivers/pci/pcie/portdrv_acpi.c index b4d2894..44296eb 100644 --- a/drivers/pci/pcie/portdrv_acpi.c +++ b/drivers/pci/pcie/portdrv_acpi.c @@ -51,7 +51,7 @@ int pcie_port_acpi_setup(struct pci_dev *port, int *srv_mask) flags = root->osc_control_set; - *srv_mask = PCIE_PORT_SERVICE_VC; + *srv_mask = PCIE_PORT_SERVICE_VC | PCIE_PORT_SERVICE_DPC; if (flags & OSC_PCI_EXPRESS_NATIVE_HP_CONTROL) *srv_mask |= PCIE_PORT_SERVICE_HP; if (flags & OSC_PCI_EXPRESS_PME_CONTROL) diff --git a/drivers/pci/pcie/portdrv_core.c b/drivers/pci/pcie/portdrv_core.c index 88122dc..2ab0f42 100644 --- a/drivers/pci/pcie/portdrv_core.c +++ b/drivers/pci/pcie/portdrv_core.c @@ -262,7 +262,7 @@ static int get_port_device_capability(struct pci_dev *dev) return 0; cap_mask = PCIE_PORT_SERVICE_PME | PCIE_PORT_SERVICE_HP - | PCIE_PORT_SERVICE_VC; + | PCIE_PORT_SERVICE_VC | PCIE_PORT_SERVICE_DPC; if (pci_aer_available()) cap_mask |= PCIE_PORT_SERVICE_AER; @@ -311,6 +311,8 @@ static int get_port_device_capability(struct pci_dev *dev) */ pcie_pme_interrupt_enable(dev, false); } + if (pci_find_ext_capability(dev, PCI_EXT_CAP_ID_DPC)) + services |= PCIE_PORT_SERVICE_DPC; return services; } @@ -338,7 +340,7 @@ static int pcie_device_init(struct pci_dev *pdev, int service, int irq) device = &pcie->device; device->bus = &pcie_port_bus_type; device->release = release_pcie_device; /* callback to free pcie dev */ - dev_set_name(device, "%s:pcie%02x", + dev_set_name(device, "%s:pcie%03x", pci_name(pdev), get_descriptor_id(pci_pcie_type(pdev), service)); device->parent = &pdev->dev; diff --git a/include/linux/pcieport_if.h b/include/linux/pcieport_if.h index 4f1089f..afcd130 100644 --- a/include/linux/pcieport_if.h +++ b/include/linux/pcieport_if.h @@ -21,6 +21,8 @@ #define PCIE_PORT_SERVICE_HP (1 << PCIE_PORT_SERVICE_HP_SHIFT) #define PCIE_PORT_SERVICE_VC_SHIFT 3 /* Virtual Channel */ #define PCIE_PORT_SERVICE_VC (1 << PCIE_PORT_SERVICE_VC_SHIFT) +#define PCIE_PORT_SERVICE_DPC_SHIFT 4 /* Downstream Port Containment */ +#define PCIE_PORT_SERVICE_DPC (1 << PCIE_PORT_SERVICE_DPC_SHIFT) struct pcie_device { int irq; /* Service IRQ/MSI/MSI-X Vector */ diff --git a/include/uapi/linux/pci_regs.h b/include/uapi/linux/pci_regs.h index 1becea8..c008294 100644 --- a/include/uapi/linux/pci_regs.h +++ b/include/uapi/linux/pci_regs.h @@ -670,7 +670,8 @@ #define PCI_EXT_CAP_ID_SECPCI 0x19 /* Secondary PCIe Capability */ #define PCI_EXT_CAP_ID_PMUX 0x1A /* Protocol Multiplexing */ #define PCI_EXT_CAP_ID_PASID 0x1B /* Process Address Space ID */ -#define PCI_EXT_CAP_ID_MAX PCI_EXT_CAP_ID_PASID +#define PCI_EXT_CAP_ID_DPC 0x1D /* Downstream Port Containment */ +#define PCI_EXT_CAP_ID_MAX PCI_EXT_CAP_ID_DPC #define PCI_EXT_CAP_DSN_SIZEOF 12 #define PCI_EXT_CAP_MCAST_ENDPOINT_SIZEOF 40 @@ -946,4 +947,28 @@ #define PCI_TPH_CAP_ST_SHIFT 16 /* st table shift */ #define PCI_TPH_BASE_SIZEOF 12 /* size with no st table */ +/* Downstream Port Containment */ +#define PCI_EXP_DPC_CAP 4 /* DPC Capability */ +#define PCI_EXP_DPC_CAP_RP_EXT 0x20 /* Root Port Extensions for DPC */ +#define PCI_EXP_DPC_CAP_POISONED_TLP 0x40 /* Poisoned TLP Egress Blocking Supported */ +#define PCI_EXP_DPC_CAP_SW_TRIGGER 0x80 /* Software Triggering Supported */ +#define PCI_EXP_DPC_CAP_DL_ACTIVE 0x1000 /* ERR_COR signal on DL_Active supported */ + +#define PCI_EXP_DPC_CTL 6 /* DPC control */ +#define PCI_EXP_DPC_CTL_DISABLE 0x00 /* Disable trigger */ +#define PCI_EXP_DPC_CTL_EN_FATAL 0x01 /* Enable trigger on ERR_FATAL message */ +#define PCI_EXP_DPC_CTL_EN_NONFATAL 0x02 /* Enable trigger on ERR_NONFATAL message */ +#define PCI_EXP_DPC_CTL_UR 0x04 /* Unsupported Request Completion Status */ +#define PCI_EXP_DPC_CTL_INT_EN 0x08 /* DPC Interrupt Enable */ +#define PCI_EXP_DPC_CTL_ERR_COR_EN 0x10 /* Enable ERRO_COR Message on DPC triggered */ +#define PCI_EXP_DPC_CTL_POSIONED_TLP 0x20 /* Enabled Poisoned TLP Egress Blocking */ +#define PCI_EXP_DPC_CTL_SW_T_EN 0x40 /* Trigger DPC Status */ +#define PCI_EXP_DPC_CTL_DL_A_ERR_COR 0x80 /* Signal ER_COR on DL_Active */ + +#define PCI_EXP_DPC_STATUS 8 /* DPC Status */ +#define PCI_EXP_DPC_STATUS_TRIGGER 0x01 /* Trigger Status */ +#define PCI_EXP_DPC_STATUS_INTERRUPT 0x08 /* Interrupt Status */ + +#define PCI_EXP_DPC_SOURCE_ID 0x0A + #endif /* LINUX_PCI_REGS_H */
This adds driver support for root and downstream ports that implement the PCI-Express Downstream Port Containment extended capability. DPC is an optional capability to contain uncorrectable errors below a port. When a DPC event is triggered, the h/w disables downstream links, so the DPC driver schedules removal for all devices below this port. This may happen concurrently with a PCI-e hotplug driver if enabled. When all downstream devices are removed and the link state transitions to disabled, the DPC driver clears the DPC status and interrupt bits so the link may retrain for a newly connected device. The pcie device naming is updated to accomodate the additional service driver. From Lukas Wunner <lukas@wunner.de>: The names of port service devices previously used one nibble to encode the port type and another nibble to encode the service type. Since this commit introduces a fifth service type, it changes device names to use one *byte* to encode the service type. E.g. a hotplug port service on a downstream bridge was previously called pcie24 and is now called pcie204. Signed-off-by: Keith Busch <keith.busch@intel.com> Cc: Lukas Wunner <lukas@wunner.de> --- v2 -> v3: Device naming fixes and changelog from Lukas. Using tristate for driver instead of boolean, and update Kconfig description. drivers/pci/pcie/Kconfig | 1 + drivers/pci/pcie/Makefile | 1 + drivers/pci/pcie/dpc/Kconfig | 18 ++++ drivers/pci/pcie/dpc/Makefile | 5 ++ drivers/pci/pcie/dpc/dpcdrv.c | 182 ++++++++++++++++++++++++++++++++++++++++ drivers/pci/pcie/portdrv.h | 4 +- drivers/pci/pcie/portdrv_acpi.c | 2 +- drivers/pci/pcie/portdrv_core.c | 6 +- include/linux/pcieport_if.h | 2 + include/uapi/linux/pci_regs.h | 27 +++++- 10 files changed, 242 insertions(+), 6 deletions(-) create mode 100644 drivers/pci/pcie/dpc/Kconfig create mode 100644 drivers/pci/pcie/dpc/Makefile create mode 100644 drivers/pci/pcie/dpc/dpcdrv.c