linux_dsm_epyc7002/drivers/pci/hotplug/pciehp.h

201 lines
8.0 KiB
C
Raw Normal View History

/* SPDX-License-Identifier: GPL-2.0+ */
/*
* PCI Express Hot Plug Controller Driver
*
* Copyright (C) 1995,2001 Compaq Computer Corporation
* Copyright (C) 2001 Greg Kroah-Hartman (greg@kroah.com)
* Copyright (C) 2001 IBM Corp.
* Copyright (C) 2003-2004 Intel Corporation
*
* All rights reserved.
*
* Send feedback to <greg@kroah.com>, <kristen.c.accardi@intel.com>
*
*/
#ifndef _PCIEHP_H
#define _PCIEHP_H
#include <linux/types.h>
#include <linux/pci.h>
#include <linux/pci_hotplug.h>
#include <linux/delay.h>
#include <linux/mutex.h>
PCI: pciehp: Avoid slot access during reset The ->reset_slot callback introduced by commits: 2e35afaefe64 ("PCI: pciehp: Add reset_slot() method") and 06a8d89af551 ("PCI: pciehp: Disable link notification across slot reset") disables notification of Presence Detect Changed and Data Link Layer State Changed events for the duration of a secondary bus reset. However a bus reset not only triggers these events, but may also clear the Presence Detect State bit in the Slot Status register and the Data Link Layer Link Active bit in the Link Status register momentarily. According to Sinan Kaya: "I know for a fact that bus reset clears the Data Link Layer Active bit as soon as link goes down. It gets set again following link up. Presence detect depends on the HW implementation. QDT root ports don't change presence detect for instance since nobody actually removed the card. If an implementation supports in-band presence detect, the answer is yes. As soon as the link goes down, presence detect bit will get cleared until recovery." https://lkml.kernel.org/r/42e72f83-3b24-f7ef-e5bc-290fae99259a@codeaurora.org In-band presence detect is also covered in Table 4-15 in PCIe r4.0, sec 4.2.6. pciehp should therefore ensure that any parts of the driver that access those bits do not run concurrently to a bus reset. The only precaution the commits took to that effect was to halt interrupt polling. They made no effort to drain the slot workqueue, cancel an outstanding Attention Button work, or block slot enable/disable requests via sysfs and in the ->probe hook. Now that pciehp is converted to enable/disable the slot exclusively from the IRQ thread, the only places accessing the two above-mentioned bits are the IRQ thread and the ->probe hook. Add locking to serialize them with a bus reset. This obviates the need to halt interrupt polling. Do not add locking to the ->get_adapter_status sysfs callback to afford users unfettered access to that bit. Use an rw_semaphore in lieu of a regular mutex to allow parallel execution of the non-reset code paths accessing the critical bits, i.e. the IRQ thread and the ->probe hook. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Rajat Jain <rajatja@google.com> Cc: Alex Williamson <alex.williamson@redhat.com> Cc: Sinan Kaya <okaya@kernel.org>
2018-07-28 12:18:00 +07:00
#include <linux/rwsem.h>
#include <linux/workqueue.h>
#include "../pcie/portdrv.h"
extern bool pciehp_poll_mode;
extern int pciehp_poll_time;
/*
* Set CONFIG_DYNAMIC_DEBUG=y and boot with 'dyndbg="file pciehp* +p"' to
* enable debug messages.
*/
#define ctrl_dbg(ctrl, format, arg...) \
pci_dbg(ctrl->pcie->port, format, ## arg)
#define ctrl_err(ctrl, format, arg...) \
pci_err(ctrl->pcie->port, format, ## arg)
#define ctrl_info(ctrl, format, arg...) \
pci_info(ctrl->pcie->port, format, ## arg)
#define ctrl_warn(ctrl, format, arg...) \
pci_warn(ctrl->pcie->port, format, ## arg)
#define SLOT_NAME_SIZE 10
/**
* struct controller - PCIe hotplug controller
* @pcie: pointer to the controller's PCIe port service device
* @slot_cap: cached copy of the Slot Capabilities register
* @slot_ctrl: cached copy of the Slot Control register
* @ctrl_lock: serializes writes to the Slot Control register
* @cmd_started: jiffies when the Slot Control register was last written;
* the next write is allowed 1 second later, absent a Command Completed
* interrupt (PCIe r4.0, sec 6.7.3.2)
* @cmd_busy: flag set on Slot Control register write, cleared by IRQ handler
* on reception of a Command Completed event
* @queue: wait queue to wake up on reception of a Command Completed event,
* used for synchronous writes to the Slot Control register
* @pending_events: used by the IRQ handler to save events retrieved from the
* Slot Status register for later consumption by the IRQ thread
* @notification_enabled: whether the IRQ was requested successfully
* @power_fault_detected: whether a power fault was detected by the hardware
* that has not yet been cleared by the user
* @poll_thread: thread to poll for slot events if no IRQ is available,
* enabled with pciehp_poll_mode module parameter
* @state: current state machine position
* @state_lock: protects reads and writes of @state;
* protects scheduling, execution and cancellation of @button_work
* @button_work: work item to turn the slot on or off after 5 seconds
* in response to an Attention Button press
PCI: hotplug: Embed hotplug_slot When the PCI hotplug core and its first user, cpqphp, were introduced in February 2002 with historic commit a8a2069f432c, cpqphp allocated a slot struct for its internal use plus a hotplug_slot struct to be registered with the hotplug core and linked the two with pointers: https://git.kernel.org/tglx/history/c/a8a2069f432c Nowadays, the predominant pattern in the tree is to embed ("subclass") such structures in one another and cast to the containing struct with container_of(). But it wasn't until July 2002 that container_of() was introduced with historic commit ec4f214232cf: https://git.kernel.org/tglx/history/c/ec4f214232cf pnv_php, introduced in 2016, did the right thing and embedded struct hotplug_slot in its internal struct pnv_php_slot, but all other drivers cargo-culted cpqphp's design and linked separate structs with pointers. Embedding structs is preferrable to linking them with pointers because it requires fewer allocations, thereby reducing overhead and simplifying error paths. Casting an embedded struct to the containing struct becomes a cheap subtraction rather than a dereference. And having fewer pointers reduces the risk of them pointing nowhere either accidentally or due to an attack. Convert all drivers to embed struct hotplug_slot in their internal slot struct. The "private" pointer in struct hotplug_slot thereby becomes unused, so drop it. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com> # drivers/pci/hotplug/rpa* Acked-by: Sebastian Ott <sebott@linux.ibm.com> # drivers/pci/hotplug/s390* Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com> # drivers/platform/x86 Cc: Len Brown <lenb@kernel.org> Cc: Scott Murray <scott@spiteful.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Oliver OHalloran <oliveroh@au1.ibm.com> Cc: Gavin Shan <gwshan@linux.vnet.ibm.com> Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com> Cc: Corentin Chary <corentin.chary@gmail.com> Cc: Darren Hart <dvhart@infradead.org>
2018-09-08 14:59:01 +07:00
* @hotplug_slot: structure registered with the PCI hotplug core
* @reset_lock: prevents access to the Data Link Layer Link Active bit in the
* Link Status register and to the Presence Detect State bit in the Slot
* Status register during a slot reset which may cause them to flap
* @ist_running: flag to keep user request waiting while IRQ thread is running
PCI: pciehp: Enable/disable exclusively from IRQ thread Besides the IRQ thread, there are several other places in the driver which enable or disable the slot: - pciehp_probe() enables the slot if it's occupied and the pciehp_force module parameter is used. - pciehp_resume() enables or disables the slot after system sleep. - pciehp_queue_pushbutton_work() enables or disables the slot after the 5 second delay following an Attention Button press. - pciehp_sysfs_enable_slot() and pciehp_sysfs_disable_slot() enable or disable the slot on sysfs write. This requires locking and complicates pciehp's state machine. A simplification can be achieved by enabling and disabling the slot exclusively from the IRQ thread. Amend the functions listed above to request slot enable/disablement from the IRQ thread by either synthesizing a Presence Detect Changed event or, in the case of a disable user request (via sysfs or an Attention Button press), submitting a newly introduced force disable request. The latter is needed because the slot shall be forced off despite being occupied. For this force disable request, avoid colliding with Slot Status register bits by using a bit number greater than 16. For synchronous execution of requests (on sysfs write), wait for the request to finish and retrieve the result. There can only ever be one sysfs write in flight due to the locking in kernfs_fop_write(), hence there is no risk of returning the result of a different sysfs request to user space. The POWERON_STATE and POWEROFF_STATE is now no longer entered by the above-listed functions, but solely by the IRQ thread when it begins a power transition. Afterwards, it moves to STATIC_STATE. The same applies to canceling the Attention Button work, it likewise becomes an IRQ thread only operation. An immediate consequence is that the POWERON_STATE and POWEROFF_STATE is never observed by the IRQ thread itself, only by functions called in a different context, such as pciehp_sysfs_enable_slot(). So remove handling of these states from pciehp_handle_button_press() and pciehp_handle_link_change() which are exclusively called from the IRQ thread. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2018-07-20 05:27:46 +07:00
* @request_result: result of last user request submitted to the IRQ thread
* @requester: wait queue to wake up on completion of user request,
* used for synchronous slot enable/disable request via sysfs
*
* PCIe hotplug has a 1:1 relationship between controller and slot, hence
* unlike other drivers, the two aren't represented by separate structures.
*/
struct controller {
struct pcie_device *pcie;
u32 slot_cap; /* capabilities and quirks */
u16 slot_ctrl; /* control register access */
struct mutex ctrl_lock;
unsigned long cmd_started;
unsigned int cmd_busy:1;
wait_queue_head_t queue;
atomic_t pending_events; /* event handling */
unsigned int notification_enabled:1;
unsigned int power_fault_detected;
struct task_struct *poll_thread;
u8 state; /* state machine */
struct mutex state_lock;
struct delayed_work button_work;
PCI: hotplug: Embed hotplug_slot When the PCI hotplug core and its first user, cpqphp, were introduced in February 2002 with historic commit a8a2069f432c, cpqphp allocated a slot struct for its internal use plus a hotplug_slot struct to be registered with the hotplug core and linked the two with pointers: https://git.kernel.org/tglx/history/c/a8a2069f432c Nowadays, the predominant pattern in the tree is to embed ("subclass") such structures in one another and cast to the containing struct with container_of(). But it wasn't until July 2002 that container_of() was introduced with historic commit ec4f214232cf: https://git.kernel.org/tglx/history/c/ec4f214232cf pnv_php, introduced in 2016, did the right thing and embedded struct hotplug_slot in its internal struct pnv_php_slot, but all other drivers cargo-culted cpqphp's design and linked separate structs with pointers. Embedding structs is preferrable to linking them with pointers because it requires fewer allocations, thereby reducing overhead and simplifying error paths. Casting an embedded struct to the containing struct becomes a cheap subtraction rather than a dereference. And having fewer pointers reduces the risk of them pointing nowhere either accidentally or due to an attack. Convert all drivers to embed struct hotplug_slot in their internal slot struct. The "private" pointer in struct hotplug_slot thereby becomes unused, so drop it. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com> # drivers/pci/hotplug/rpa* Acked-by: Sebastian Ott <sebott@linux.ibm.com> # drivers/pci/hotplug/s390* Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com> # drivers/platform/x86 Cc: Len Brown <lenb@kernel.org> Cc: Scott Murray <scott@spiteful.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Oliver OHalloran <oliveroh@au1.ibm.com> Cc: Gavin Shan <gwshan@linux.vnet.ibm.com> Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com> Cc: Corentin Chary <corentin.chary@gmail.com> Cc: Darren Hart <dvhart@infradead.org>
2018-09-08 14:59:01 +07:00
struct hotplug_slot hotplug_slot; /* hotplug core interface */
struct rw_semaphore reset_lock;
unsigned int ist_running;
PCI: pciehp: Enable/disable exclusively from IRQ thread Besides the IRQ thread, there are several other places in the driver which enable or disable the slot: - pciehp_probe() enables the slot if it's occupied and the pciehp_force module parameter is used. - pciehp_resume() enables or disables the slot after system sleep. - pciehp_queue_pushbutton_work() enables or disables the slot after the 5 second delay following an Attention Button press. - pciehp_sysfs_enable_slot() and pciehp_sysfs_disable_slot() enable or disable the slot on sysfs write. This requires locking and complicates pciehp's state machine. A simplification can be achieved by enabling and disabling the slot exclusively from the IRQ thread. Amend the functions listed above to request slot enable/disablement from the IRQ thread by either synthesizing a Presence Detect Changed event or, in the case of a disable user request (via sysfs or an Attention Button press), submitting a newly introduced force disable request. The latter is needed because the slot shall be forced off despite being occupied. For this force disable request, avoid colliding with Slot Status register bits by using a bit number greater than 16. For synchronous execution of requests (on sysfs write), wait for the request to finish and retrieve the result. There can only ever be one sysfs write in flight due to the locking in kernfs_fop_write(), hence there is no risk of returning the result of a different sysfs request to user space. The POWERON_STATE and POWEROFF_STATE is now no longer entered by the above-listed functions, but solely by the IRQ thread when it begins a power transition. Afterwards, it moves to STATIC_STATE. The same applies to canceling the Attention Button work, it likewise becomes an IRQ thread only operation. An immediate consequence is that the POWERON_STATE and POWEROFF_STATE is never observed by the IRQ thread itself, only by functions called in a different context, such as pciehp_sysfs_enable_slot(). So remove handling of these states from pciehp_handle_button_press() and pciehp_handle_link_change() which are exclusively called from the IRQ thread. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2018-07-20 05:27:46 +07:00
int request_result;
wait_queue_head_t requester;
};
/**
* DOC: Slot state
*
* @OFF_STATE: slot is powered off, no subordinate devices are enumerated
* @BLINKINGON_STATE: slot will be powered on after the 5 second delay,
* Power Indicator is blinking
* @BLINKINGOFF_STATE: slot will be powered off after the 5 second delay,
* Power Indicator is blinking
* @POWERON_STATE: slot is currently powering on
* @POWEROFF_STATE: slot is currently powering off
* @ON_STATE: slot is powered on, subordinate devices have been enumerated
*/
#define OFF_STATE 0
#define BLINKINGON_STATE 1
#define BLINKINGOFF_STATE 2
#define POWERON_STATE 3
#define POWEROFF_STATE 4
#define ON_STATE 5
PCI: pciehp: Enable/disable exclusively from IRQ thread Besides the IRQ thread, there are several other places in the driver which enable or disable the slot: - pciehp_probe() enables the slot if it's occupied and the pciehp_force module parameter is used. - pciehp_resume() enables or disables the slot after system sleep. - pciehp_queue_pushbutton_work() enables or disables the slot after the 5 second delay following an Attention Button press. - pciehp_sysfs_enable_slot() and pciehp_sysfs_disable_slot() enable or disable the slot on sysfs write. This requires locking and complicates pciehp's state machine. A simplification can be achieved by enabling and disabling the slot exclusively from the IRQ thread. Amend the functions listed above to request slot enable/disablement from the IRQ thread by either synthesizing a Presence Detect Changed event or, in the case of a disable user request (via sysfs or an Attention Button press), submitting a newly introduced force disable request. The latter is needed because the slot shall be forced off despite being occupied. For this force disable request, avoid colliding with Slot Status register bits by using a bit number greater than 16. For synchronous execution of requests (on sysfs write), wait for the request to finish and retrieve the result. There can only ever be one sysfs write in flight due to the locking in kernfs_fop_write(), hence there is no risk of returning the result of a different sysfs request to user space. The POWERON_STATE and POWEROFF_STATE is now no longer entered by the above-listed functions, but solely by the IRQ thread when it begins a power transition. Afterwards, it moves to STATIC_STATE. The same applies to canceling the Attention Button work, it likewise becomes an IRQ thread only operation. An immediate consequence is that the POWERON_STATE and POWEROFF_STATE is never observed by the IRQ thread itself, only by functions called in a different context, such as pciehp_sysfs_enable_slot(). So remove handling of these states from pciehp_handle_button_press() and pciehp_handle_link_change() which are exclusively called from the IRQ thread. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2018-07-20 05:27:46 +07:00
/**
* DOC: Flags to request an action from the IRQ thread
*
* These are stored together with events read from the Slot Status register,
* hence must be greater than its 16-bit width.
*
* %DISABLE_SLOT: Disable the slot in response to a user request via sysfs or
* an Attention Button press after the 5 second delay
PCI: pciehp: Support interrupts sent from D3hot If a hotplug port is able to send an interrupt, one would naively assume that it is accessible at that moment. After all, if it wouldn't be accessible, i.e. if its parent is in D3hot and the link to the hotplug port is thus down, how should an interrupt come through? It turns out that assumption is wrong at least for Thunderbolt: Even though its parents are in D3hot, a Thunderbolt hotplug port is able to signal interrupts. Because the port's config space is inaccessible and resuming the parents may sleep, the hard IRQ handler has to defer runtime resuming the parents and reading the Slot Status register to the IRQ thread. If the hotplug port uses a level-triggered INTx interrupt, it needs to be masked until the IRQ thread has cleared the signaled events. For simplicity, this commit also masks edge-triggered MSI/MSI-X interrupts. Note that if the interrupt is shared (which can only happen for INTx), other devices are starved from receiving interrupts until the IRQ thread is scheduled, has runtime resumed the hotplug port's parents and has read and cleared the Slot Status register. That delay is dominated by the 10 ms D3hot->D0 transition time of each parent port. The worst case is a Thunderbolt downstream port at the end of a daisy chain: There may be up to six Thunderbolt controllers in-between it and the root port, each comprising an upstream and downstream port, plus its own upstream port. That's 13 x 10 = 130 ms. Possible mitigations are polling the interrupt while it's disabled or reducing the d3_delay of Thunderbolt ports if possible. Open code masking of the interrupt instead of requesting it with the IRQF_ONESHOT flag to minimize the period during which it is masked. (IRQF_ONESHOT unmasks the IRQ only after the IRQ thread has finished.) PCIe r4.0 sec 6.7.3.4 states that "If wake generation is required by the associated form factor specification, a hotplug capable Downstream Port must support generation of a wakeup event (using the PME mechanism) on hotplug events that occur when the system is in a sleep state or the Port is in device state D1, D2, or D3Hot." This would seem to imply that PME needs to be enabled on the hotplug port when it is runtime suspended. pci_enable_wake() currently doesn't enable PME on bridges, it may be necessary to add an exemption for hotplug bridges there. On "Light Ridge" Thunderbolt controllers, the PME_Status bit is not set when an interrupt occurs while the hotplug port is in D3hot, even if PME is enabled. (I've tested this on a Mac and we hardcode the OSC_PCI_EXPRESS_PME_CONTROL bit to 0 on Macs in negotiate_os_control(), modifying it to 1 didn't change the behavior.) (Side note: Section 6.7.3.4 also states that "PME and Hot-Plug Event interrupts (when both are implemented) always share the same MSI or MSI-X vector". That would only seem to apply to Root Ports, however the section never mentions Root Ports, only Downstream Ports. This is explained in the definition of "Downstream Port" in the "Terms and Acronyms" section of the PCIe Base Spec: "The Ports on a Switch that are not the Upstream Port are Downstream Ports. All Ports on a Root Complex are Downstream Ports.") Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Ashok Raj <ashok.raj@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Yinghai Lu <yinghai@kernel.org>
2018-07-28 12:18:00 +07:00
* %RERUN_ISR: Used by the IRQ handler to inform the IRQ thread that the
* hotplug port was inaccessible when the interrupt occurred, requiring
* that the IRQ handler is rerun by the IRQ thread after it has made the
* hotplug port accessible by runtime resuming its parents to D0
PCI: pciehp: Enable/disable exclusively from IRQ thread Besides the IRQ thread, there are several other places in the driver which enable or disable the slot: - pciehp_probe() enables the slot if it's occupied and the pciehp_force module parameter is used. - pciehp_resume() enables or disables the slot after system sleep. - pciehp_queue_pushbutton_work() enables or disables the slot after the 5 second delay following an Attention Button press. - pciehp_sysfs_enable_slot() and pciehp_sysfs_disable_slot() enable or disable the slot on sysfs write. This requires locking and complicates pciehp's state machine. A simplification can be achieved by enabling and disabling the slot exclusively from the IRQ thread. Amend the functions listed above to request slot enable/disablement from the IRQ thread by either synthesizing a Presence Detect Changed event or, in the case of a disable user request (via sysfs or an Attention Button press), submitting a newly introduced force disable request. The latter is needed because the slot shall be forced off despite being occupied. For this force disable request, avoid colliding with Slot Status register bits by using a bit number greater than 16. For synchronous execution of requests (on sysfs write), wait for the request to finish and retrieve the result. There can only ever be one sysfs write in flight due to the locking in kernfs_fop_write(), hence there is no risk of returning the result of a different sysfs request to user space. The POWERON_STATE and POWEROFF_STATE is now no longer entered by the above-listed functions, but solely by the IRQ thread when it begins a power transition. Afterwards, it moves to STATIC_STATE. The same applies to canceling the Attention Button work, it likewise becomes an IRQ thread only operation. An immediate consequence is that the POWERON_STATE and POWEROFF_STATE is never observed by the IRQ thread itself, only by functions called in a different context, such as pciehp_sysfs_enable_slot(). So remove handling of these states from pciehp_handle_button_press() and pciehp_handle_link_change() which are exclusively called from the IRQ thread. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2018-07-20 05:27:46 +07:00
*/
#define DISABLE_SLOT (1 << 16)
PCI: pciehp: Support interrupts sent from D3hot If a hotplug port is able to send an interrupt, one would naively assume that it is accessible at that moment. After all, if it wouldn't be accessible, i.e. if its parent is in D3hot and the link to the hotplug port is thus down, how should an interrupt come through? It turns out that assumption is wrong at least for Thunderbolt: Even though its parents are in D3hot, a Thunderbolt hotplug port is able to signal interrupts. Because the port's config space is inaccessible and resuming the parents may sleep, the hard IRQ handler has to defer runtime resuming the parents and reading the Slot Status register to the IRQ thread. If the hotplug port uses a level-triggered INTx interrupt, it needs to be masked until the IRQ thread has cleared the signaled events. For simplicity, this commit also masks edge-triggered MSI/MSI-X interrupts. Note that if the interrupt is shared (which can only happen for INTx), other devices are starved from receiving interrupts until the IRQ thread is scheduled, has runtime resumed the hotplug port's parents and has read and cleared the Slot Status register. That delay is dominated by the 10 ms D3hot->D0 transition time of each parent port. The worst case is a Thunderbolt downstream port at the end of a daisy chain: There may be up to six Thunderbolt controllers in-between it and the root port, each comprising an upstream and downstream port, plus its own upstream port. That's 13 x 10 = 130 ms. Possible mitigations are polling the interrupt while it's disabled or reducing the d3_delay of Thunderbolt ports if possible. Open code masking of the interrupt instead of requesting it with the IRQF_ONESHOT flag to minimize the period during which it is masked. (IRQF_ONESHOT unmasks the IRQ only after the IRQ thread has finished.) PCIe r4.0 sec 6.7.3.4 states that "If wake generation is required by the associated form factor specification, a hotplug capable Downstream Port must support generation of a wakeup event (using the PME mechanism) on hotplug events that occur when the system is in a sleep state or the Port is in device state D1, D2, or D3Hot." This would seem to imply that PME needs to be enabled on the hotplug port when it is runtime suspended. pci_enable_wake() currently doesn't enable PME on bridges, it may be necessary to add an exemption for hotplug bridges there. On "Light Ridge" Thunderbolt controllers, the PME_Status bit is not set when an interrupt occurs while the hotplug port is in D3hot, even if PME is enabled. (I've tested this on a Mac and we hardcode the OSC_PCI_EXPRESS_PME_CONTROL bit to 0 on Macs in negotiate_os_control(), modifying it to 1 didn't change the behavior.) (Side note: Section 6.7.3.4 also states that "PME and Hot-Plug Event interrupts (when both are implemented) always share the same MSI or MSI-X vector". That would only seem to apply to Root Ports, however the section never mentions Root Ports, only Downstream Ports. This is explained in the definition of "Downstream Port" in the "Terms and Acronyms" section of the PCIe Base Spec: "The Ports on a Switch that are not the Upstream Port are Downstream Ports. All Ports on a Root Complex are Downstream Ports.") Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Ashok Raj <ashok.raj@intel.com> Cc: Keith Busch <keith.busch@intel.com> Cc: Yinghai Lu <yinghai@kernel.org>
2018-07-28 12:18:00 +07:00
#define RERUN_ISR (1 << 17)
PCI: pciehp: Enable/disable exclusively from IRQ thread Besides the IRQ thread, there are several other places in the driver which enable or disable the slot: - pciehp_probe() enables the slot if it's occupied and the pciehp_force module parameter is used. - pciehp_resume() enables or disables the slot after system sleep. - pciehp_queue_pushbutton_work() enables or disables the slot after the 5 second delay following an Attention Button press. - pciehp_sysfs_enable_slot() and pciehp_sysfs_disable_slot() enable or disable the slot on sysfs write. This requires locking and complicates pciehp's state machine. A simplification can be achieved by enabling and disabling the slot exclusively from the IRQ thread. Amend the functions listed above to request slot enable/disablement from the IRQ thread by either synthesizing a Presence Detect Changed event or, in the case of a disable user request (via sysfs or an Attention Button press), submitting a newly introduced force disable request. The latter is needed because the slot shall be forced off despite being occupied. For this force disable request, avoid colliding with Slot Status register bits by using a bit number greater than 16. For synchronous execution of requests (on sysfs write), wait for the request to finish and retrieve the result. There can only ever be one sysfs write in flight due to the locking in kernfs_fop_write(), hence there is no risk of returning the result of a different sysfs request to user space. The POWERON_STATE and POWEROFF_STATE is now no longer entered by the above-listed functions, but solely by the IRQ thread when it begins a power transition. Afterwards, it moves to STATIC_STATE. The same applies to canceling the Attention Button work, it likewise becomes an IRQ thread only operation. An immediate consequence is that the POWERON_STATE and POWEROFF_STATE is never observed by the IRQ thread itself, only by functions called in a different context, such as pciehp_sysfs_enable_slot(). So remove handling of these states from pciehp_handle_button_press() and pciehp_handle_link_change() which are exclusively called from the IRQ thread. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2018-07-20 05:27:46 +07:00
#define ATTN_BUTTN(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_ABP)
#define POWER_CTRL(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_PCP)
#define MRL_SENS(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_MRLSP)
#define ATTN_LED(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_AIP)
#define PWR_LED(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_PIP)
#define HP_SUPR_RM(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_HPS)
#define EMI(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_EIP)
#define NO_CMD_CMPL(ctrl) ((ctrl)->slot_cap & PCI_EXP_SLTCAP_NCCS)
#define PSN(ctrl) (((ctrl)->slot_cap & PCI_EXP_SLTCAP_PSN) >> 19)
PCI: pciehp: Enable/disable exclusively from IRQ thread Besides the IRQ thread, there are several other places in the driver which enable or disable the slot: - pciehp_probe() enables the slot if it's occupied and the pciehp_force module parameter is used. - pciehp_resume() enables or disables the slot after system sleep. - pciehp_queue_pushbutton_work() enables or disables the slot after the 5 second delay following an Attention Button press. - pciehp_sysfs_enable_slot() and pciehp_sysfs_disable_slot() enable or disable the slot on sysfs write. This requires locking and complicates pciehp's state machine. A simplification can be achieved by enabling and disabling the slot exclusively from the IRQ thread. Amend the functions listed above to request slot enable/disablement from the IRQ thread by either synthesizing a Presence Detect Changed event or, in the case of a disable user request (via sysfs or an Attention Button press), submitting a newly introduced force disable request. The latter is needed because the slot shall be forced off despite being occupied. For this force disable request, avoid colliding with Slot Status register bits by using a bit number greater than 16. For synchronous execution of requests (on sysfs write), wait for the request to finish and retrieve the result. There can only ever be one sysfs write in flight due to the locking in kernfs_fop_write(), hence there is no risk of returning the result of a different sysfs request to user space. The POWERON_STATE and POWEROFF_STATE is now no longer entered by the above-listed functions, but solely by the IRQ thread when it begins a power transition. Afterwards, it moves to STATIC_STATE. The same applies to canceling the Attention Button work, it likewise becomes an IRQ thread only operation. An immediate consequence is that the POWERON_STATE and POWEROFF_STATE is never observed by the IRQ thread itself, only by functions called in a different context, such as pciehp_sysfs_enable_slot(). So remove handling of these states from pciehp_handle_button_press() and pciehp_handle_link_change() which are exclusively called from the IRQ thread. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2018-07-20 05:27:46 +07:00
void pciehp_request(struct controller *ctrl, int action);
void pciehp_handle_button_press(struct controller *ctrl);
void pciehp_handle_disable_request(struct controller *ctrl);
void pciehp_handle_presence_or_link_change(struct controller *ctrl, u32 events);
int pciehp_configure_device(struct controller *ctrl);
void pciehp_unconfigure_device(struct controller *ctrl, bool presence);
void pciehp_queue_pushbutton_work(struct work_struct *work);
struct controller *pcie_init(struct pcie_device *dev);
int pcie_init_notification(struct controller *ctrl);
void pcie_shutdown_notification(struct controller *ctrl);
PCI: pciehp: Clear spurious events earlier on resume Thunderbolt hotplug ports that were occupied before system sleep resume with their downstream link in "off" state. Only after the Thunderbolt controller has reestablished the PCIe tunnels does the link go up. As a result, a spurious Presence Detect Changed and/or Data Link Layer State Changed event occurs. The events are not immediately acted upon because tunnel reestablishment happens in the ->resume_noirq phase, when interrupts are still disabled. Also, notification of events may initially be disabled in the Slot Control register when coming out of system sleep and is reenabled in the ->resume_noirq phase through: pci_pm_resume_noirq() pci_pm_default_resume_early() pci_restore_state() pci_restore_pcie_state() It is not guaranteed that the events are acted upon at all: PCIe r4.0, sec 6.7.3.4 says that "a port may optionally send an MSI when there are hot-plug events that occur while interrupt generation is disabled, and interrupt generation is subsequently enabled." Note the "optionally". If an MSI is sent, pciehp will gratuitously turn the slot off and back on once the ->resume_early phase has commenced. If an MSI is not sent, the extant, unacknowledged events in the Slot Status register will prevent future notification of presence or link changes. Commit 13c65840feab ("PCI: pciehp: Clear Presence Detect and Data Link Layer Status Changed on resume") fixed the latter by clearing the events in the ->resume phase. Move this to the ->resume_noirq phase to also fix the gratuitous disable/enablement of the slot. The commit further restored the Slot Control register in the ->resume phase, but that's dispensable because as shown above it's already been done in the ->resume_noirq phase. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com>
2018-07-20 05:27:53 +07:00
void pcie_clear_hotplug_events(struct controller *ctrl);
void pcie_enable_interrupt(struct controller *ctrl);
void pcie_disable_interrupt(struct controller *ctrl);
int pciehp_power_on_slot(struct controller *ctrl);
void pciehp_power_off_slot(struct controller *ctrl);
void pciehp_get_power_status(struct controller *ctrl, u8 *status);
#define INDICATOR_NOOP -1 /* Leave indicator unchanged */
void pciehp_set_indicators(struct controller *ctrl, int pwr, int attn);
void pciehp_get_latch_status(struct controller *ctrl, u8 *status);
int pciehp_query_power_fault(struct controller *ctrl);
PCI: pciehp: Prevent deadlock on disconnect This addresses deadlocks in these common cases in hierarchies containing two switches: - All involved ports are runtime suspended and they are unplugged. This can happen easily if the drivers involved automatically enable runtime PM (xHCI for example does that). - System is suspended (e.g., closing the lid on a laptop) with a dock + something else connected, and the dock is unplugged while suspended. These cases lead to the following deadlock: INFO: task irq/126-pciehp:198 blocked for more than 120 seconds. irq/126-pciehp D 0 198 2 0x80000000 Call Trace: schedule+0x2c/0x80 schedule_timeout+0x246/0x350 wait_for_completion+0xb7/0x140 kthread_stop+0x49/0x110 free_irq+0x32/0x70 pcie_shutdown_notification+0x2f/0x50 pciehp_remove+0x27/0x50 pcie_port_remove_service+0x36/0x50 device_release_driver+0x12/0x20 bus_remove_device+0xec/0x160 device_del+0x13b/0x350 device_unregister+0x1a/0x60 remove_iter+0x1e/0x30 device_for_each_child+0x56/0x90 pcie_port_device_remove+0x22/0x40 pcie_portdrv_remove+0x20/0x60 pci_device_remove+0x3e/0xc0 device_release_driver_internal+0x18c/0x250 device_release_driver+0x12/0x20 pci_stop_bus_device+0x6f/0x90 pci_stop_bus_device+0x31/0x90 pci_stop_and_remove_bus_device+0x12/0x20 pciehp_unconfigure_device+0x88/0x140 pciehp_disable_slot+0x6a/0x110 pciehp_handle_presence_or_link_change+0x263/0x400 pciehp_ist+0x1c9/0x1d0 irq_thread_fn+0x24/0x60 irq_thread+0xeb/0x190 kthread+0x120/0x140 INFO: task irq/190-pciehp:2288 blocked for more than 120 seconds. irq/190-pciehp D 0 2288 2 0x80000000 Call Trace: __schedule+0x2a2/0x880 schedule+0x2c/0x80 schedule_preempt_disabled+0xe/0x10 mutex_lock+0x2c/0x30 pci_lock_rescan_remove+0x15/0x20 pciehp_unconfigure_device+0x4d/0x140 pciehp_disable_slot+0x6a/0x110 pciehp_handle_presence_or_link_change+0x263/0x400 pciehp_ist+0x1c9/0x1d0 irq_thread_fn+0x24/0x60 irq_thread+0xeb/0x190 kthread+0x120/0x140 What happens here is that the whole hierarchy is runtime resumed and the parent PCIe downstream port, which got the hot-remove event, starts removing devices below it, taking pci_lock_rescan_remove() lock. When the child PCIe port is runtime resumed it calls pciehp_check_presence() which ends up calling pciehp_card_present() and pciehp_check_link_active(). Both of these use pcie_capability_read_word(), which notices that the underlying device is already gone and returns PCIBIOS_DEVICE_NOT_FOUND with the capability value set to 0. When pciehp gets this value it thinks that its child device is also hot-removed and schedules its IRQ thread to handle the event. The deadlock happens when the child's IRQ thread runs and tries to acquire pci_lock_rescan_remove() which is already taken by the parent and the parent waits for the child's IRQ thread to finish. Prevent this from happening by checking the return value of pcie_capability_read_word() and if it is PCIBIOS_DEVICE_NOT_FOUND stop performing any hot-removal activities. [bhelgaas: add common scenarios to commit log] Link: https://lore.kernel.org/r/20191029170022.57528-2-mika.westerberg@linux.intel.com Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com> Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2019-10-30 00:00:22 +07:00
int pciehp_card_present(struct controller *ctrl);
int pciehp_card_present_or_link_active(struct controller *ctrl);
int pciehp_check_link_status(struct controller *ctrl);
PCI: pciehp: Prevent deadlock on disconnect This addresses deadlocks in these common cases in hierarchies containing two switches: - All involved ports are runtime suspended and they are unplugged. This can happen easily if the drivers involved automatically enable runtime PM (xHCI for example does that). - System is suspended (e.g., closing the lid on a laptop) with a dock + something else connected, and the dock is unplugged while suspended. These cases lead to the following deadlock: INFO: task irq/126-pciehp:198 blocked for more than 120 seconds. irq/126-pciehp D 0 198 2 0x80000000 Call Trace: schedule+0x2c/0x80 schedule_timeout+0x246/0x350 wait_for_completion+0xb7/0x140 kthread_stop+0x49/0x110 free_irq+0x32/0x70 pcie_shutdown_notification+0x2f/0x50 pciehp_remove+0x27/0x50 pcie_port_remove_service+0x36/0x50 device_release_driver+0x12/0x20 bus_remove_device+0xec/0x160 device_del+0x13b/0x350 device_unregister+0x1a/0x60 remove_iter+0x1e/0x30 device_for_each_child+0x56/0x90 pcie_port_device_remove+0x22/0x40 pcie_portdrv_remove+0x20/0x60 pci_device_remove+0x3e/0xc0 device_release_driver_internal+0x18c/0x250 device_release_driver+0x12/0x20 pci_stop_bus_device+0x6f/0x90 pci_stop_bus_device+0x31/0x90 pci_stop_and_remove_bus_device+0x12/0x20 pciehp_unconfigure_device+0x88/0x140 pciehp_disable_slot+0x6a/0x110 pciehp_handle_presence_or_link_change+0x263/0x400 pciehp_ist+0x1c9/0x1d0 irq_thread_fn+0x24/0x60 irq_thread+0xeb/0x190 kthread+0x120/0x140 INFO: task irq/190-pciehp:2288 blocked for more than 120 seconds. irq/190-pciehp D 0 2288 2 0x80000000 Call Trace: __schedule+0x2a2/0x880 schedule+0x2c/0x80 schedule_preempt_disabled+0xe/0x10 mutex_lock+0x2c/0x30 pci_lock_rescan_remove+0x15/0x20 pciehp_unconfigure_device+0x4d/0x140 pciehp_disable_slot+0x6a/0x110 pciehp_handle_presence_or_link_change+0x263/0x400 pciehp_ist+0x1c9/0x1d0 irq_thread_fn+0x24/0x60 irq_thread+0xeb/0x190 kthread+0x120/0x140 What happens here is that the whole hierarchy is runtime resumed and the parent PCIe downstream port, which got the hot-remove event, starts removing devices below it, taking pci_lock_rescan_remove() lock. When the child PCIe port is runtime resumed it calls pciehp_check_presence() which ends up calling pciehp_card_present() and pciehp_check_link_active(). Both of these use pcie_capability_read_word(), which notices that the underlying device is already gone and returns PCIBIOS_DEVICE_NOT_FOUND with the capability value set to 0. When pciehp gets this value it thinks that its child device is also hot-removed and schedules its IRQ thread to handle the event. The deadlock happens when the child's IRQ thread runs and tries to acquire pci_lock_rescan_remove() which is already taken by the parent and the parent waits for the child's IRQ thread to finish. Prevent this from happening by checking the return value of pcie_capability_read_word() and if it is PCIBIOS_DEVICE_NOT_FOUND stop performing any hot-removal activities. [bhelgaas: add common scenarios to commit log] Link: https://lore.kernel.org/r/20191029170022.57528-2-mika.westerberg@linux.intel.com Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com> Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
2019-10-30 00:00:22 +07:00
int pciehp_check_link_active(struct controller *ctrl);
void pciehp_release_ctrl(struct controller *ctrl);
int pciehp_sysfs_enable_slot(struct hotplug_slot *hotplug_slot);
int pciehp_sysfs_disable_slot(struct hotplug_slot *hotplug_slot);
int pciehp_reset_slot(struct hotplug_slot *hotplug_slot, int probe);
int pciehp_get_attention_status(struct hotplug_slot *hotplug_slot, u8 *status);
int pciehp_set_raw_indicator_status(struct hotplug_slot *h_slot, u8 status);
int pciehp_get_raw_indicator_status(struct hotplug_slot *h_slot, u8 *status);
static inline const char *slot_name(struct controller *ctrl)
{
PCI: hotplug: Embed hotplug_slot When the PCI hotplug core and its first user, cpqphp, were introduced in February 2002 with historic commit a8a2069f432c, cpqphp allocated a slot struct for its internal use plus a hotplug_slot struct to be registered with the hotplug core and linked the two with pointers: https://git.kernel.org/tglx/history/c/a8a2069f432c Nowadays, the predominant pattern in the tree is to embed ("subclass") such structures in one another and cast to the containing struct with container_of(). But it wasn't until July 2002 that container_of() was introduced with historic commit ec4f214232cf: https://git.kernel.org/tglx/history/c/ec4f214232cf pnv_php, introduced in 2016, did the right thing and embedded struct hotplug_slot in its internal struct pnv_php_slot, but all other drivers cargo-culted cpqphp's design and linked separate structs with pointers. Embedding structs is preferrable to linking them with pointers because it requires fewer allocations, thereby reducing overhead and simplifying error paths. Casting an embedded struct to the containing struct becomes a cheap subtraction rather than a dereference. And having fewer pointers reduces the risk of them pointing nowhere either accidentally or due to an attack. Convert all drivers to embed struct hotplug_slot in their internal slot struct. The "private" pointer in struct hotplug_slot thereby becomes unused, so drop it. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Acked-by: Tyrel Datwyler <tyreld@linux.vnet.ibm.com> # drivers/pci/hotplug/rpa* Acked-by: Sebastian Ott <sebott@linux.ibm.com> # drivers/pci/hotplug/s390* Acked-by: Andy Shevchenko <andy.shevchenko@gmail.com> # drivers/platform/x86 Cc: Len Brown <lenb@kernel.org> Cc: Scott Murray <scott@spiteful.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Oliver OHalloran <oliveroh@au1.ibm.com> Cc: Gavin Shan <gwshan@linux.vnet.ibm.com> Cc: Gerald Schaefer <gerald.schaefer@de.ibm.com> Cc: Corentin Chary <corentin.chary@gmail.com> Cc: Darren Hart <dvhart@infradead.org>
2018-09-08 14:59:01 +07:00
return hotplug_slot_name(&ctrl->hotplug_slot);
}
static inline struct controller *to_ctrl(struct hotplug_slot *hotplug_slot)
{
return container_of(hotplug_slot, struct controller, hotplug_slot);
}
#endif /* _PCIEHP_H */