Re: [RFC PATCH 2/2] PCI: pciehp: Add support for delayed power-on

[Guenter Roeck]

On Wed, Oct 21, 2015 at 03:23:23PM -0500, Bjorn Helgaas wrote:
> Hi Guenter,
> 
> On Mon, Oct 12, 2015 at 12:10:13PM -0700, Guenter Roeck wrote:
> > Some oddball devices may experience a PCIe link flap after power-on.
> > This may result in the following sequence of events.
> > 
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: Card present on Slot(0)
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: slot(0): Link Up event
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: Link Up event ignored on slot(0): already powering on
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: slot(0): Link Down event
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: Link Down event queued on slot(0): currently getting powered on
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: slot(0): Link Up event
> > fpc0 kernel: pciehp 0000:02:08.0:pcie24: Link Up event queued on slot(0): currently getting powered off
> 
> I'm really interested in how this happens.  I'm not happy with the
> pciehp state machine, and I wonder if it is causing or obscuring this
> problem.
> 
Obscuring, maybe, but not causing it. The hardware I am dealing with
is a bit special.

> For example, pcie_isr() reads PCI_EXP_SLTSTA to find out what
> happened.  Then it queues events (attention button, presence detect
> changed, etc.)  Along the way, we re-read PCI_EXP_SLTSTA.  That seems
> wrong to me -- I think we should capture the state *once*, save it,
> and act on it.  If we re-read it, we'll see transitions that might
> confuse us.
> 
... especially if PCI_EXP_SLTSTA isn't stable.

> Similarly, pcie_isr() reads and acts on PCI_EXP_LNKSTA (via
> pciehp_check_link_active()).  I think we should capture and save that
> at the same time we capture PCI_EXP_SLTSTA, before we queue up work
> events that are going to change things.
> 
I tried that. It didn't help to get rid of the ignored Link Up event,
because the events are reported with separate interrupts.

> And I'm not sure it's really necessary for pcie_isr() to queue up
> *separate* events for attention button, presence detect, power fault,
> and link state events.  I suspect that could throw in extraneous
> events that confuse things.  For instance, I think it's possible for
> pcie_isr() to see a single interrupt with PCI_EXP_SLTSTA showing both
> a presence detect change (card now present) and a link state change
> (link now up).  It will queue two events and we'll probably see a
> "Link Up event ignored" message.  I think it would be better if
> pcie_isr() captured the register values we need, bundled them up,
> and queued a single work item to deal with them.
> 
Agreed. Problem, however, is that - again, in my case - the events are
reported with separate interrupts.

Also, the problem isn't really the ignored link up event. The problem is the
sequence of link up - link down - link up, which makes things really messy.

What happens is along the line of

- link up is queued
- (at least sometimes) link up event starts to be processed (ultimately
  causing all devices to be instantiated)
- link down is queued (unavoidable if the link up event is already being
  processed)
- link up even is queued

I _might_ be able to "catch" the link up event before it is being handled,
at least in some cases. I had previously thought about it, but was unable
to come up with a solution.

Here is something that might work, though: In pciehp_power_thread(),
if ENABLE_REQ is handled, ignore it if the slot state is POWEROFF_STATE.
Would that be acceptable if it works ? It would not solve the problem for good,
but at least it might limit its impact.

> > This causes the driver for affected devices to be instantiated, removed,
> > and re-instantiated. This can result in problems, for example if the device
> > in question provides gpio pins or interrupts which are used by other
> > drivers. For example, the removal can result in errors such as
> > the following.
> > 
> > fpc0 kernel: remove_proc_entry: removing non-empty directory 'irq/148',
> > 	leaking at least 'pic0'
> > fpc0 kernel: ------------[ cut here ]------------
> > fpc0 kernel: WARNING: at /home/p2020/linux-freescale/fs/proc/generic.c:575
> 
> Something's definitely wrong here, but shouldn't we be able to add and
> remove a device as many times as we want, as quickly as we want,
> without problems?  Maybe this particular issue is a driver problem or
> a core problem with the proc file maintenance?  I wonder if this is
> something we could reproduce without pciehp at all, just by inserting
> and removing a module?
> 
This may be special for my situation. The pcie driver instantiates a large
number of gpio pins, some of which are used to provide interrupts and status
pins for various other devices. Those devices are still instantiated when
the PCIe interface goes down. Those devices are also not connected to the
PCIe bus, or at least not to the same PCIe bus, so they are still active.

I would need some means to remove all those devices in a clean way, but I have
no idea how I could do that. The infrastructure provides EPROBE_DEFER to handle
the case where the gpio pins are not yet available, and our code makes heavy
use of that. However, there isn't really a means to handle the opposite
situation, where a driver providing gpio pins used by other drivers
"disappears".

Granted, this has a lot to do with the hardware I am dealing with. There is
nothing I can do about that, though, except maybe influence the next generation
of that hardware.

> > Add support for per-port power-on delay to avoid this situation. This can
> > be enabled globally with the pciehp_poweron_delay module parameter, or
> > per port (using a quirks function) with a new poweron_delay flag in
> > struct pci_dev.
> > 
> > With this patch, the link flap still occurs, but because of the delayed
> > insertion the driver is not immediately instantiated, and the above error
> > is no longer seen.
> 
> We might have to do something like this eventually, but I'd really
> like to see if we can simplify the pciehp state machine a little
> before we add more stuff to it.
> 
Fair enough.

Thanks,
Guenter

> > Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> > ---
> > We started seeing this problem after the recent rework of link status
> > change handling. I think the up/down/up sequence was previously just
> > ignored or folded into a single event (and sometimes resulted in a stuck
> > state machine).
> > I would like to see this patch upstream, but I am not sure if the problem
> > is seen anywhere but on the hardware I am dealing with, and I can
> > understand if others don't want to pollute the pcie hotplug code with
> > such workarounds. Also, maybe someone has a better idea on how to handle
> > the situation, so I marked the patch as RFC.
> > 
> >  drivers/pci/hotplug/pciehp.h      |  5 ++++-
> >  drivers/pci/hotplug/pciehp_core.c |  3 +++
> >  drivers/pci/hotplug/pciehp_ctrl.c | 44 +++++++++++++++++++++++++++------------
> >  drivers/pci/hotplug/pciehp_hpc.c  |  2 ++
> >  include/linux/pci.h               |  1 +
> >  5 files changed, 41 insertions(+), 14 deletions(-)
> > 
> > diff --git a/drivers/pci/hotplug/pciehp.h b/drivers/pci/hotplug/pciehp.h
> > index 62d6fe6c3714..5047bde7d51d 100644
> > --- a/drivers/pci/hotplug/pciehp.h
> > +++ b/drivers/pci/hotplug/pciehp.h
> > @@ -40,6 +40,7 @@
> >  
> >  #define MY_NAME	"pciehp"
> >  
> > +extern bool pciehp_poweron_delay;
> >  extern bool pciehp_poll_mode;
> >  extern int pciehp_poll_time;
> >  extern bool pciehp_debug;
> > @@ -98,6 +99,7 @@ struct controller {
> >  	unsigned int cmd_busy:1;
> >  	unsigned int link_active_reporting:1;
> >  	unsigned int notification_enabled:1;
> > +	unsigned int poweron_delay:1;	/* Delay poweron for this slot */
> >  	unsigned int power_fault_detected;
> >  };
> >  
> > @@ -112,7 +114,8 @@ struct controller {
> >  #define BLINKINGON_STATE		1
> >  #define BLINKINGOFF_STATE		2
> >  #define POWERON_STATE			3
> > -#define POWEROFF_STATE			4
> > +#define DELAYED_POWERON_STATE		4
> > +#define POWEROFF_STATE			5
> >  
> >  #define ATTN_BUTTN(ctrl)	((ctrl)->slot_cap & PCI_EXP_SLTCAP_ABP)
> >  #define POWER_CTRL(ctrl)	((ctrl)->slot_cap & PCI_EXP_SLTCAP_PCP)
> > diff --git a/drivers/pci/hotplug/pciehp_core.c b/drivers/pci/hotplug/pciehp_core.c
> > index 612b21a14df5..cc69fd10d884 100644
> > --- a/drivers/pci/hotplug/pciehp_core.c
> > +++ b/drivers/pci/hotplug/pciehp_core.c
> > @@ -38,6 +38,7 @@
> >  #include <linux/time.h>
> >  
> >  /* Global variables */
> > +bool pciehp_poweron_delay;
> >  bool pciehp_debug;
> >  bool pciehp_poll_mode;
> >  int pciehp_poll_time;
> > @@ -51,10 +52,12 @@ MODULE_AUTHOR(DRIVER_AUTHOR);
> >  MODULE_DESCRIPTION(DRIVER_DESC);
> >  MODULE_LICENSE("GPL");
> >  
> > +module_param(pciehp_poweron_delay, bool, 0644);
> >  module_param(pciehp_debug, bool, 0644);
> >  module_param(pciehp_poll_mode, bool, 0644);
> >  module_param(pciehp_poll_time, int, 0644);
> >  module_param(pciehp_force, bool, 0644);
> > +MODULE_PARM_DESC(pciehp_poweron_delay, "Delay port power-on");
> >  MODULE_PARM_DESC(pciehp_debug, "Debugging mode enabled or not");
> >  MODULE_PARM_DESC(pciehp_poll_mode, "Using polling mechanism for hot-plug events or not");
> >  MODULE_PARM_DESC(pciehp_poll_time, "Polling mechanism frequency, in seconds");
> > diff --git a/drivers/pci/hotplug/pciehp_ctrl.c b/drivers/pci/hotplug/pciehp_ctrl.c
> > index ef96a1d51fac..fd829c2b7418 100644
> > --- a/drivers/pci/hotplug/pciehp_ctrl.c
> > +++ b/drivers/pci/hotplug/pciehp_ctrl.c
> > @@ -204,11 +204,19 @@ static void pciehp_power_thread(struct work_struct *work)
> >  	kfree(info);
> >  }
> >  
> > -void pciehp_queue_power_work(struct slot *p_slot, int req)
> > +void pciehp_queue_power_work(struct slot *p_slot, int req, bool delay)
> >  {
> >  	struct power_work_info *info;
> >  
> > -	p_slot->state = req == ENABLE_REQ ? POWERON_STATE : POWEROFF_STATE;
> > +	if (req == ENABLE_REQ) {
> > +		p_slot->state = delay ? DELAYED_POWERON_STATE : POWERON_STATE;
> > +		if (delay) {
> > +			mod_delayed_work(p_slot->wq, &p_slot->work, HZ);
> > +			return;
> > +		}
> > +	} else {
> > +		p_slot->state = POWEROFF_STATE;
> > +	}
> >  
> >  	info = kmalloc(sizeof(*info), GFP_KERNEL);
> >  	if (!info) {
> > @@ -229,10 +237,11 @@ void pciehp_queue_pushbutton_work(struct work_struct *work)
> >  	mutex_lock(&p_slot->lock);
> >  	switch (p_slot->state) {
> >  	case BLINKINGOFF_STATE:
> > -		pciehp_queue_power_work(p_slot, DISABLE_REQ);
> > +		pciehp_queue_power_work(p_slot, DISABLE_REQ, false);
> >  		break;
> >  	case BLINKINGON_STATE:
> > -		pciehp_queue_power_work(p_slot, ENABLE_REQ);
> > +	case DELAYED_POWERON_STATE:
> > +		pciehp_queue_power_work(p_slot, ENABLE_REQ, false);
> >  		break;
> >  	default:
> >  		break;
> > @@ -263,7 +272,7 @@ static void handle_button_press_event(struct slot *p_slot)
> >  		/* blink green LED and turn off amber */
> >  		pciehp_green_led_blink(p_slot);
> >  		pciehp_set_attention_status(p_slot, 0);
> > -		queue_delayed_work(p_slot->wq, &p_slot->work, 5*HZ);
> > +		mod_delayed_work(p_slot->wq, &p_slot->work, 5 * HZ);
> >  		break;
> >  	case BLINKINGOFF_STATE:
> >  	case BLINKINGON_STATE:
> > @@ -285,6 +294,7 @@ static void handle_button_press_event(struct slot *p_slot)
> >  		break;
> >  	case POWEROFF_STATE:
> >  	case POWERON_STATE:
> > +	case DELAYED_POWERON_STATE:
> >  		/*
> >  		 * Ignore if the slot is on power-on or power-off state;
> >  		 * this means that the previous attention button action
> > @@ -305,12 +315,12 @@ static void handle_surprise_event(struct slot *p_slot)
> >  {
> >  	u8 getstatus;
> >  
> > +	if (p_slot->state == DELAYED_POWERON_STATE)
> > +		cancel_delayed_work(&p_slot->work);
> > +
> >  	pciehp_get_adapter_status(p_slot, &getstatus);
> > -	if (!getstatus) {
> > -		pciehp_queue_power_work(p_slot, DISABLE_REQ);
> > -	} else {
> > -		pciehp_queue_power_work(p_slot, ENABLE_REQ);
> > -	}
> > +	pciehp_queue_power_work(p_slot, getstatus ? ENABLE_REQ : DISABLE_REQ,
> > +				p_slot->ctrl->poweron_delay);
> >  }
> >  
> >  /*
> > @@ -327,8 +337,13 @@ static void handle_link_event(struct slot *p_slot, u32 event)
> >  		/* Fall through */
> >  	case STATIC_STATE:
> >  		pciehp_queue_power_work(p_slot, event == INT_LINK_UP ?
> > -					ENABLE_REQ : DISABLE_REQ);
> > +					ENABLE_REQ : DISABLE_REQ,
> > +					ctrl->poweron_delay);
> >  		break;
> > +	case DELAYED_POWERON_STATE:
> > +		if (event != INT_LINK_UP)
> > +			cancel_delayed_work(&p_slot->work);
> > +		/* Fall through */
> >  	case POWERON_STATE:
> >  		if (event == INT_LINK_UP) {
> >  			ctrl_info(ctrl,
> > @@ -338,7 +353,7 @@ static void handle_link_event(struct slot *p_slot, u32 event)
> >  			ctrl_info(ctrl,
> >  				  "Link Down event queued on slot(%s): currently getting powered on\n",
> >  				  slot_name(p_slot));
> > -			pciehp_queue_power_work(p_slot, DISABLE_REQ);
> > +			pciehp_queue_power_work(p_slot, DISABLE_REQ, false);
> >  		}
> >  		break;
> >  	case POWEROFF_STATE:
> > @@ -346,7 +361,8 @@ static void handle_link_event(struct slot *p_slot, u32 event)
> >  			ctrl_info(ctrl,
> >  				  "Link Up event queued on slot(%s): currently getting powered off\n",
> >  				  slot_name(p_slot));
> > -			pciehp_queue_power_work(p_slot, ENABLE_REQ);
> > +			pciehp_queue_power_work(p_slot, ENABLE_REQ,
> > +						ctrl->poweron_delay);
> >  		} else {
> >  			ctrl_info(ctrl,
> >  				  "Link Down event ignored on slot(%s): already powering off\n",
> > @@ -482,6 +498,7 @@ int pciehp_sysfs_enable_slot(struct slot *p_slot)
> >  		p_slot->state = STATIC_STATE;
> >  		break;
> >  	case POWERON_STATE:
> > +	case DELAYED_POWERON_STATE:
> >  		ctrl_info(ctrl, "Slot %s is already in powering on state\n",
> >  			  slot_name(p_slot));
> >  		break;
> > @@ -522,6 +539,7 @@ int pciehp_sysfs_disable_slot(struct slot *p_slot)
> >  		break;
> >  	case BLINKINGON_STATE:
> >  	case POWERON_STATE:
> > +	case DELAYED_POWERON_STATE:
> >  		ctrl_info(ctrl, "Already disabled on slot %s\n",
> >  			  slot_name(p_slot));
> >  		break;
> > diff --git a/drivers/pci/hotplug/pciehp_hpc.c b/drivers/pci/hotplug/pciehp_hpc.c
> > index 5c24e938042f..7c7a598eec9f 100644
> > --- a/drivers/pci/hotplug/pciehp_hpc.c
> > +++ b/drivers/pci/hotplug/pciehp_hpc.c
> > @@ -809,6 +809,8 @@ struct controller *pcie_init(struct pcie_device *dev)
> >  	pcie_capability_read_dword(pdev, PCI_EXP_LNKCAP, &link_cap);
> >  	if (link_cap & PCI_EXP_LNKCAP_DLLLARC)
> >  		ctrl->link_active_reporting = 1;
> > +	if (pciehp_poweron_delay || dev->port->poweron_delay)
> > +		ctrl->poweron_delay = 1;
> >  
> >  	/* Clear all remaining event bits in Slot Status register */
> >  	pcie_capability_write_word(pdev, PCI_EXP_SLTSTA,
> > diff --git a/include/linux/pci.h b/include/linux/pci.h
> > index e90eb22de628..9cb0fe3037b9 100644
> > --- a/include/linux/pci.h
> > +++ b/include/linux/pci.h
> > @@ -359,6 +359,7 @@ struct pci_dev {
> >  	unsigned int	io_window_1k:1;	/* Intel P2P bridge 1K I/O windows */
> >  	unsigned int	irq_managed:1;
> >  	unsigned int	has_secondary_link:1;
> > +	unsigned int    poweron_delay:1;	/* Port needs poweron delay */
> >  	pci_dev_flags_t dev_flags;
> >  	atomic_t	enable_cnt;	/* pci_enable_device has been called */
> >  
> > -- 
> > 2.1.4
> > 
> > --
> > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> > the body of a message to majordomo@xxxxxxxxxxxxxxx
> > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > Please read the FAQ at  http://www.tux.org/lkml/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/