OpenCloudOS-Kernel/drivers/pci/hotplug/pciehp_ctrl.c

437 lines
10 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>
*
*/
#include <linux/kernel.h>
#include <linux/types.h>
PCI: pciehp: Resume to D0 on enable/disable pciehp's IRQ thread ensures accessibility of the port by runtime resuming its parent to D0. However when the slot is enabled/disabled, the port itself needs to be in D0 because its secondary bus is accessed in: pciehp_check_link_status(), pciehp_configure_device() (both called from board_added()) and pciehp_unconfigure_device() (called from remove_board()). Thus, acquire a runtime PM ref on enable/disablement of the slot. Yinghai Lu additionally discovered that some SkyLake servers feature a Power Controller for their PCIe hotplug ports (PCIe r3.1, sec 6.7.1.8) which requires the port to be in D0 when invoking pciehp_power_on_slot() (likewise called from board_added()). If slot power is turned on while in D3hot, link training later fails: https://lkml.kernel.org/r/20170205073454.GA253@wunner.de The spec is silent about such a requirement, but it seems prudent to assume that any hotplug port with a Power Controller may need this. The present commit holds a runtime PM ref whenever slot power is turned on and off, but it doesn't keep the port in D0 as long as slot power is on. If vendors determine that's necessary, they need to amend pciehp to acquire a runtime PM ref in pciehp_power_on_slot() and release one in pciehp_power_off_slot(). Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> 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>
2018-07-20 06:27:56 +08:00
#include <linux/pm_runtime.h>
#include <linux/pci.h>
#include "pciehp.h"
/* The following routines constitute the bulk of the
hotplug controller logic
*/
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
#define SAFE_REMOVAL true
#define SURPRISE_REMOVAL false
static void set_slot_off(struct controller *ctrl, struct slot *pslot)
{
/* turn off slot, turn on Amber LED, turn off Green LED if supported*/
if (POWER_CTRL(ctrl)) {
pciehp_power_off_slot(pslot);
/*
* After turning power off, we must wait for at least 1 second
* before taking any action that relies on power having been
* removed from the slot/adapter.
*/
msleep(1000);
}
pciehp_green_led_off(pslot);
pciehp_set_attention_status(pslot, 1);
}
/**
* board_added - Called after a board has been added to the system.
* @p_slot: &slot where board is added
*
* Turns power on for the board.
* Configures board.
*/
static int board_added(struct slot *p_slot)
{
int retval = 0;
struct controller *ctrl = p_slot->ctrl;
struct pci_bus *parent = ctrl->pcie->port->subordinate;
if (POWER_CTRL(ctrl)) {
/* Power on slot */
retval = pciehp_power_on_slot(p_slot);
if (retval)
return retval;
}
pciehp_green_led_blink(p_slot);
/* Check link training status */
retval = pciehp_check_link_status(ctrl);
if (retval) {
ctrl_err(ctrl, "Failed to check link status\n");
goto err_exit;
}
/* Check for a power fault */
if (ctrl->power_fault_detected || pciehp_query_power_fault(p_slot)) {
ctrl_err(ctrl, "Slot(%s): Power fault\n", slot_name(p_slot));
retval = -EIO;
goto err_exit;
}
retval = pciehp_configure_device(p_slot);
if (retval) {
if (retval != -EEXIST) {
ctrl_err(ctrl, "Cannot add device at %04x:%02x:00\n",
pci_domain_nr(parent), parent->number);
goto err_exit;
}
}
pciehp_green_led_on(p_slot);
pciehp_set_attention_status(p_slot, 0);
return 0;
err_exit:
set_slot_off(ctrl, p_slot);
return retval;
}
/**
* remove_board - Turns off slot and LEDs
* @p_slot: slot where board is being removed
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
* @safe_removal: whether the board is safely removed (versus surprise removed)
*/
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
static void remove_board(struct slot *p_slot, bool safe_removal)
{
struct controller *ctrl = p_slot->ctrl;
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
pciehp_unconfigure_device(p_slot, safe_removal);
if (POWER_CTRL(ctrl)) {
pciehp_power_off_slot(p_slot);
/*
* After turning power off, we must wait for at least 1 second
* before taking any action that relies on power having been
* removed from the slot/adapter.
*/
msleep(1000);
}
/* turn off Green LED */
pciehp_green_led_off(p_slot);
}
static int pciehp_enable_slot(struct slot *slot);
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
static int pciehp_disable_slot(struct slot *slot, bool safe_removal);
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 06:27:46 +08:00
void pciehp_request(struct controller *ctrl, int action)
{
atomic_or(action, &ctrl->pending_events);
if (!pciehp_poll_mode)
irq_wake_thread(ctrl->pcie->irq, ctrl);
}
void pciehp_queue_pushbutton_work(struct work_struct *work)
{
struct slot *p_slot = container_of(work, struct slot, work.work);
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 06:27:46 +08:00
struct controller *ctrl = p_slot->ctrl;
mutex_lock(&p_slot->lock);
switch (p_slot->state) {
case BLINKINGOFF_STATE:
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 06:27:46 +08:00
pciehp_request(ctrl, DISABLE_SLOT);
break;
case BLINKINGON_STATE:
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 06:27:46 +08:00
pciehp_request(ctrl, PCI_EXP_SLTSTA_PDC);
break;
default:
break;
}
mutex_unlock(&p_slot->lock);
}
void pciehp_handle_button_press(struct slot *p_slot)
{
struct controller *ctrl = p_slot->ctrl;
mutex_lock(&p_slot->lock);
switch (p_slot->state) {
case OFF_STATE:
case ON_STATE:
if (p_slot->state == ON_STATE) {
p_slot->state = BLINKINGOFF_STATE;
ctrl_info(ctrl, "Slot(%s): Powering off due to button press\n",
slot_name(p_slot));
} else {
p_slot->state = BLINKINGON_STATE;
ctrl_info(ctrl, "Slot(%s) Powering on due to button press\n",
slot_name(p_slot));
}
/* blink green LED and turn off amber */
pciehp_green_led_blink(p_slot);
pciehp_set_attention_status(p_slot, 0);
schedule_delayed_work(&p_slot->work, 5 * HZ);
break;
case BLINKINGOFF_STATE:
case BLINKINGON_STATE:
/*
* Cancel if we are still blinking; this means that we
* press the attention again before the 5 sec. limit
* expires to cancel hot-add or hot-remove
*/
ctrl_info(ctrl, "Slot(%s): Button cancel\n", slot_name(p_slot));
cancel_delayed_work(&p_slot->work);
if (p_slot->state == BLINKINGOFF_STATE) {
p_slot->state = ON_STATE;
pciehp_green_led_on(p_slot);
} else {
p_slot->state = OFF_STATE;
pciehp_green_led_off(p_slot);
}
pciehp_set_attention_status(p_slot, 0);
ctrl_info(ctrl, "Slot(%s): Action canceled due to button press\n",
slot_name(p_slot));
break;
default:
ctrl_err(ctrl, "Slot(%s): Ignoring invalid state %#x\n",
slot_name(p_slot), p_slot->state);
break;
}
mutex_unlock(&p_slot->lock);
}
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 06:27:46 +08:00
void pciehp_handle_disable_request(struct slot *slot)
{
struct controller *ctrl = slot->ctrl;
mutex_lock(&slot->lock);
switch (slot->state) {
case BLINKINGON_STATE:
case BLINKINGOFF_STATE:
cancel_delayed_work(&slot->work);
break;
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 06:27:46 +08:00
}
slot->state = POWEROFF_STATE;
mutex_unlock(&slot->lock);
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
ctrl->request_result = pciehp_disable_slot(slot, SAFE_REMOVAL);
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 06:27:46 +08:00
}
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
void pciehp_handle_presence_or_link_change(struct slot *slot, u32 events)
{
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
struct controller *ctrl = slot->ctrl;
bool link_active;
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
u8 present;
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
/*
* If the slot is on and presence or link has changed, turn it off.
* Even if it's occupied again, we cannot assume the card is the same.
*/
mutex_lock(&slot->lock);
switch (slot->state) {
case BLINKINGOFF_STATE:
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
cancel_delayed_work(&slot->work);
/* fall through */
case ON_STATE:
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
slot->state = POWEROFF_STATE;
mutex_unlock(&slot->lock);
if (events & PCI_EXP_SLTSTA_DLLSC)
ctrl_info(ctrl, "Slot(%s): Link Down\n",
slot_name(slot));
if (events & PCI_EXP_SLTSTA_PDC)
ctrl_info(ctrl, "Slot(%s): Card not present\n",
slot_name(slot));
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
pciehp_disable_slot(slot, SURPRISE_REMOVAL);
break;
default:
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
mutex_unlock(&slot->lock);
break;
}
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
/* Turn the slot on if it's occupied or link is up */
mutex_lock(&slot->lock);
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
pciehp_get_adapter_status(slot, &present);
link_active = pciehp_check_link_active(ctrl);
if (!present && !link_active) {
mutex_unlock(&slot->lock);
return;
}
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 06:27:46 +08:00
switch (slot->state) {
case BLINKINGON_STATE:
cancel_delayed_work(&slot->work);
/* fall through */
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
case OFF_STATE:
slot->state = POWERON_STATE;
mutex_unlock(&slot->lock);
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
if (present)
ctrl_info(ctrl, "Slot(%s): Card present\n",
slot_name(slot));
if (link_active)
ctrl_info(ctrl, "Slot(%s): Link Up\n",
slot_name(slot));
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 06:27:46 +08:00
ctrl->request_result = pciehp_enable_slot(slot);
PCI: pciehp: Become resilient to missed events A hotplug port's Slot Status register does not count how often each type of event occurred, it only records the fact *that* an event has occurred. Previously pciehp queued a work item for each event. But if it missed an event, e.g. removal of a card in-between two back-to-back insertions, it queued up the wrong work item or no work item at all. Commit fad214b0aa72 ("PCI: pciehp: Process all hotplug events before looking for new ones") sought to improve the situation by shrinking the window during which events may be missed. But Stefan Roese reports unbalanced Card present and Link Up events, suggesting that we're still missing events if they occur very rapidly. Bjorn Helgaas responds that he considers pciehp's event handling "baroque" and calls for its simplification and rationalization: https://lkml.kernel.org/r/20180202192045.GA53759@bhelgaas-glaptop.roam.corp.google.com It gets worse once a hotplug port is runtime suspended: The port can signal an interrupt while it and its parents are in D3hot, i.e. while it is inaccessible. By the time we've runtime resumed all parents to D0 and read the port's Slot Status register, we may have missed an arbitrary number of events. Event handling therefore needs to be reworked to become resilient to missed events. Assume that a Presence Detect Changed event has occurred. Consider the following truth table: - Slot is in OFF_STATE and is currently empty. => Do nothing. (The event is trailing a Link Down or we've missed an insertion and subsequent removal.) - Slot is in OFF_STATE and is currently occupied. => Turn the slot on. - Slot is in ON_STATE and is currently empty. => Turn the slot off. - Slot is in ON_STATE and is currently occupied. => Turn the slot off, (Be cautious and assume the card in then back on. the slot isn't the same as before.) This leads to the following simple algorithm: 1 If the slot is in ON_STATE, turn it off unconditionally. 2 If the slot is currently occupied, turn it on. Because those actions are now carried out synchronously, rather than by scheduled work items, pciehp reacts to the *current* situation and missed events no longer matter. Data Link Layer State Changed events can be handled identically to Presence Detect Changed events. Note that in the above truth table, a Link Up trailing a Card present event didn't have to be accounted for: It is filtered out by pciehp_check_link_status(). As for Attention Button Pressed events, PCIe r4.0, sec 6.7.1.5 says: "Once the Power Indicator begins blinking, a 5-second abort interval exists during which a second depression of the Attention Button cancels the operation." In other words, the user can only expect the system to react to a button press after it starts blinking. Missed button presses that occur in-between are irrelevant. Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Stefan Roese <sr@denx.de> Cc: Mayurkumar Patel <mayurkumar.patel@intel.com> Cc: Mika Westerberg <mika.westerberg@linux.intel.com> Cc: Kenji Kaneshige <kaneshige.kenji@jp.fujitsu.com>
2018-07-20 06:27:49 +08:00
break;
default:
mutex_unlock(&slot->lock);
break;
}
}
static int __pciehp_enable_slot(struct slot *p_slot)
{
u8 getstatus = 0;
struct controller *ctrl = p_slot->ctrl;
pciehp_get_adapter_status(p_slot, &getstatus);
if (!getstatus) {
ctrl_info(ctrl, "Slot(%s): No adapter\n", slot_name(p_slot));
return -ENODEV;
}
if (MRL_SENS(p_slot->ctrl)) {
pciehp_get_latch_status(p_slot, &getstatus);
if (getstatus) {
ctrl_info(ctrl, "Slot(%s): Latch open\n",
slot_name(p_slot));
return -ENODEV;
}
}
if (POWER_CTRL(p_slot->ctrl)) {
pciehp_get_power_status(p_slot, &getstatus);
if (getstatus) {
ctrl_info(ctrl, "Slot(%s): Already enabled\n",
slot_name(p_slot));
return 0;
}
}
return board_added(p_slot);
}
static int pciehp_enable_slot(struct slot *slot)
{
struct controller *ctrl = slot->ctrl;
int ret;
PCI: pciehp: Resume to D0 on enable/disable pciehp's IRQ thread ensures accessibility of the port by runtime resuming its parent to D0. However when the slot is enabled/disabled, the port itself needs to be in D0 because its secondary bus is accessed in: pciehp_check_link_status(), pciehp_configure_device() (both called from board_added()) and pciehp_unconfigure_device() (called from remove_board()). Thus, acquire a runtime PM ref on enable/disablement of the slot. Yinghai Lu additionally discovered that some SkyLake servers feature a Power Controller for their PCIe hotplug ports (PCIe r3.1, sec 6.7.1.8) which requires the port to be in D0 when invoking pciehp_power_on_slot() (likewise called from board_added()). If slot power is turned on while in D3hot, link training later fails: https://lkml.kernel.org/r/20170205073454.GA253@wunner.de The spec is silent about such a requirement, but it seems prudent to assume that any hotplug port with a Power Controller may need this. The present commit holds a runtime PM ref whenever slot power is turned on and off, but it doesn't keep the port in D0 as long as slot power is on. If vendors determine that's necessary, they need to amend pciehp to acquire a runtime PM ref in pciehp_power_on_slot() and release one in pciehp_power_off_slot(). Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> 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>
2018-07-20 06:27:56 +08:00
pm_runtime_get_sync(&ctrl->pcie->port->dev);
ret = __pciehp_enable_slot(slot);
if (ret && ATTN_BUTTN(ctrl))
pciehp_green_led_off(slot); /* may be blinking */
PCI: pciehp: Resume to D0 on enable/disable pciehp's IRQ thread ensures accessibility of the port by runtime resuming its parent to D0. However when the slot is enabled/disabled, the port itself needs to be in D0 because its secondary bus is accessed in: pciehp_check_link_status(), pciehp_configure_device() (both called from board_added()) and pciehp_unconfigure_device() (called from remove_board()). Thus, acquire a runtime PM ref on enable/disablement of the slot. Yinghai Lu additionally discovered that some SkyLake servers feature a Power Controller for their PCIe hotplug ports (PCIe r3.1, sec 6.7.1.8) which requires the port to be in D0 when invoking pciehp_power_on_slot() (likewise called from board_added()). If slot power is turned on while in D3hot, link training later fails: https://lkml.kernel.org/r/20170205073454.GA253@wunner.de The spec is silent about such a requirement, but it seems prudent to assume that any hotplug port with a Power Controller may need this. The present commit holds a runtime PM ref whenever slot power is turned on and off, but it doesn't keep the port in D0 as long as slot power is on. If vendors determine that's necessary, they need to amend pciehp to acquire a runtime PM ref in pciehp_power_on_slot() and release one in pciehp_power_off_slot(). Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> 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>
2018-07-20 06:27:56 +08:00
pm_runtime_put(&ctrl->pcie->port->dev);
mutex_lock(&slot->lock);
slot->state = ret ? OFF_STATE : ON_STATE;
mutex_unlock(&slot->lock);
return ret;
}
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
static int __pciehp_disable_slot(struct slot *p_slot, bool safe_removal)
{
u8 getstatus = 0;
struct controller *ctrl = p_slot->ctrl;
if (POWER_CTRL(p_slot->ctrl)) {
pciehp_get_power_status(p_slot, &getstatus);
if (!getstatus) {
ctrl_info(ctrl, "Slot(%s): Already disabled\n",
slot_name(p_slot));
return -EINVAL;
}
}
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
remove_board(p_slot, safe_removal);
return 0;
}
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
static int pciehp_disable_slot(struct slot *slot, bool safe_removal)
{
PCI: pciehp: Resume to D0 on enable/disable pciehp's IRQ thread ensures accessibility of the port by runtime resuming its parent to D0. However when the slot is enabled/disabled, the port itself needs to be in D0 because its secondary bus is accessed in: pciehp_check_link_status(), pciehp_configure_device() (both called from board_added()) and pciehp_unconfigure_device() (called from remove_board()). Thus, acquire a runtime PM ref on enable/disablement of the slot. Yinghai Lu additionally discovered that some SkyLake servers feature a Power Controller for their PCIe hotplug ports (PCIe r3.1, sec 6.7.1.8) which requires the port to be in D0 when invoking pciehp_power_on_slot() (likewise called from board_added()). If slot power is turned on while in D3hot, link training later fails: https://lkml.kernel.org/r/20170205073454.GA253@wunner.de The spec is silent about such a requirement, but it seems prudent to assume that any hotplug port with a Power Controller may need this. The present commit holds a runtime PM ref whenever slot power is turned on and off, but it doesn't keep the port in D0 as long as slot power is on. If vendors determine that's necessary, they need to amend pciehp to acquire a runtime PM ref in pciehp_power_on_slot() and release one in pciehp_power_off_slot(). Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> 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>
2018-07-20 06:27:56 +08:00
struct controller *ctrl = slot->ctrl;
int ret;
PCI: pciehp: Resume to D0 on enable/disable pciehp's IRQ thread ensures accessibility of the port by runtime resuming its parent to D0. However when the slot is enabled/disabled, the port itself needs to be in D0 because its secondary bus is accessed in: pciehp_check_link_status(), pciehp_configure_device() (both called from board_added()) and pciehp_unconfigure_device() (called from remove_board()). Thus, acquire a runtime PM ref on enable/disablement of the slot. Yinghai Lu additionally discovered that some SkyLake servers feature a Power Controller for their PCIe hotplug ports (PCIe r3.1, sec 6.7.1.8) which requires the port to be in D0 when invoking pciehp_power_on_slot() (likewise called from board_added()). If slot power is turned on while in D3hot, link training later fails: https://lkml.kernel.org/r/20170205073454.GA253@wunner.de The spec is silent about such a requirement, but it seems prudent to assume that any hotplug port with a Power Controller may need this. The present commit holds a runtime PM ref whenever slot power is turned on and off, but it doesn't keep the port in D0 as long as slot power is on. If vendors determine that's necessary, they need to amend pciehp to acquire a runtime PM ref in pciehp_power_on_slot() and release one in pciehp_power_off_slot(). Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> 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>
2018-07-20 06:27:56 +08:00
pm_runtime_get_sync(&ctrl->pcie->port->dev);
PCI: pciehp: Differentiate between surprise and safe removal When removing PCI devices below a hotplug bridge, pciehp marks them as disconnected if the card is no longer present in the slot or it quiesces them if the card is still present (by disabling INTx interrupts, bus mastering and SERR# reporting). To detect whether the card is still present, pciehp checks the Presence Detect State bit in the Slot Status register. The problem with this approach is that even if the card is present, the link to it may be down, and it that case it would be better to mark the devices as disconnected instead of trying to quiesce them. Moreover, if the card in the slot was quickly replaced by another one, the Presence Detect State bit would be set, yet trying to quiesce the new card's devices would be wrong and the correct thing to do is to mark the previous card's devices as disconnected. Instead of looking at the Presence Detect State bit, it is better to differentiate whether the card was surprise removed versus safely removed (via sysfs or an Attention Button press). On surprise removal, the devices should be marked as disconnected, whereas on safe removal it is correct to quiesce the devices. The knowledge whether a surprise removal or a safe removal is at hand does exist further up in the call stack: A surprise removal is initiated by pciehp_handle_presence_or_link_change(), a safe removal by pciehp_handle_disable_request(). Pass that information down to pciehp_unconfigure_device() and use it in lieu of the Presence Detect State bit. While there, add kernel-doc to pciehp_unconfigure_device() and pciehp_configure_device(). Tested-by: Alexandru Gagniuc <mr.nuke.me@gmail.com> Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Cc: Keith Busch <keith.busch@intel.com>
2018-07-31 13:50:37 +08:00
ret = __pciehp_disable_slot(slot, safe_removal);
PCI: pciehp: Resume to D0 on enable/disable pciehp's IRQ thread ensures accessibility of the port by runtime resuming its parent to D0. However when the slot is enabled/disabled, the port itself needs to be in D0 because its secondary bus is accessed in: pciehp_check_link_status(), pciehp_configure_device() (both called from board_added()) and pciehp_unconfigure_device() (called from remove_board()). Thus, acquire a runtime PM ref on enable/disablement of the slot. Yinghai Lu additionally discovered that some SkyLake servers feature a Power Controller for their PCIe hotplug ports (PCIe r3.1, sec 6.7.1.8) which requires the port to be in D0 when invoking pciehp_power_on_slot() (likewise called from board_added()). If slot power is turned on while in D3hot, link training later fails: https://lkml.kernel.org/r/20170205073454.GA253@wunner.de The spec is silent about such a requirement, but it seems prudent to assume that any hotplug port with a Power Controller may need this. The present commit holds a runtime PM ref whenever slot power is turned on and off, but it doesn't keep the port in D0 as long as slot power is on. If vendors determine that's necessary, they need to amend pciehp to acquire a runtime PM ref in pciehp_power_on_slot() and release one in pciehp_power_off_slot(). Signed-off-by: Lukas Wunner <lukas@wunner.de> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> 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>
2018-07-20 06:27:56 +08:00
pm_runtime_put(&ctrl->pcie->port->dev);
mutex_lock(&slot->lock);
slot->state = OFF_STATE;
mutex_unlock(&slot->lock);
return ret;
}
int pciehp_sysfs_enable_slot(struct slot *p_slot)
{
struct controller *ctrl = p_slot->ctrl;
mutex_lock(&p_slot->lock);
switch (p_slot->state) {
case BLINKINGON_STATE:
case OFF_STATE:
mutex_unlock(&p_slot->lock);
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 06:27:46 +08:00
/*
* The IRQ thread becomes a no-op if the user pulls out the
* card before the thread wakes up, so initialize to -ENODEV.
*/
ctrl->request_result = -ENODEV;
pciehp_request(ctrl, PCI_EXP_SLTSTA_PDC);
wait_event(ctrl->requester,
!atomic_read(&ctrl->pending_events));
return ctrl->request_result;
case POWERON_STATE:
ctrl_info(ctrl, "Slot(%s): Already in powering on state\n",
slot_name(p_slot));
break;
case BLINKINGOFF_STATE:
case ON_STATE:
case POWEROFF_STATE:
ctrl_info(ctrl, "Slot(%s): Already enabled\n",
slot_name(p_slot));
break;
default:
ctrl_err(ctrl, "Slot(%s): Invalid state %#x\n",
slot_name(p_slot), p_slot->state);
break;
}
mutex_unlock(&p_slot->lock);
return -ENODEV;
}
int pciehp_sysfs_disable_slot(struct slot *p_slot)
{
struct controller *ctrl = p_slot->ctrl;
mutex_lock(&p_slot->lock);
switch (p_slot->state) {
case BLINKINGOFF_STATE:
case ON_STATE:
mutex_unlock(&p_slot->lock);
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 06:27:46 +08:00
pciehp_request(ctrl, DISABLE_SLOT);
wait_event(ctrl->requester,
!atomic_read(&ctrl->pending_events));
return ctrl->request_result;
case POWEROFF_STATE:
ctrl_info(ctrl, "Slot(%s): Already in powering off state\n",
slot_name(p_slot));
break;
case BLINKINGON_STATE:
case OFF_STATE:
case POWERON_STATE:
ctrl_info(ctrl, "Slot(%s): Already disabled\n",
slot_name(p_slot));
break;
default:
ctrl_err(ctrl, "Slot(%s): Invalid state %#x\n",
slot_name(p_slot), p_slot->state);
break;
}
mutex_unlock(&p_slot->lock);
return -ENODEV;
}