Skip to contentSkip to conversation

PCI: pciehp: Do not turn off slot if presence comes up after link

PATCH In review Stuart Hayes 3 participants 10 messages Oct 1, 2019 – Oct 3, 2019 PCI

lore
Share & tools
lore
#0 lore raw

cover PCI: pciehp: Do not turn off slot if presence comes up after link

In older PCIe specs, PDS (presence detect) would come up when the "in-band" presence detect pin connected, and would be up before DLLLA (link active).

In PCIe 4.0 (as an ECN) and in PCIe 5.0, there is a new bit to show if in-band presence detection can be disabled for the slot, and another bit that disables it--and a recommendation that it should be disabled if it can be. In addition, certain OEMs disable in-band presence detection without implementing these bits.

This means it is possible to get a "card present" interrupt after the link is up and the driver is loaded. This causes an erroneous removal of the device driver, followed by an immediate re-probing.

This patch set defines these new bits, uses them to disable in-band presence detection if it can be, waits for PDS to go up if in-band presence detection is disabled, and adds a DMI table that will let us know if we should assume in-band presence is disabled on a system.

This patch set is based on a patch set [1] submitted many months ago by Alexandru Gagniuc, who is no longer working on it.

[1] https://patchwork.kernel.org/cover/10909167/
[v3,0/4] PCI: pciehp: Do not turn off slot if presence comes up after link

Stuart Hayes (3):
PCI: pciehp: Add support for disabling in-band presence
PCI: pciehp: Wait for PDS when in-band presence is disabled
PCI: pciehp: Add dmi table for systems with in-band presence disabled

 drivers/pci/hotplug/pciehp.h     |  1 +
 drivers/pci/hotplug/pciehp_hpc.c | 45 +++++++++++++++++++++++++++++++-
 include/uapi/linux/pci_regs.h    |  2 ++
 3 files changed, 47 insertions(+), 1 deletion(-)
Signature · 1 line
2.18.1
#1 lore raw

PATCH 1/3 PCI: pciehp: Add support for disabling in-band presence

The presence detect state (PDS) is normally a logical or of in-band and out-of-band presence. As of PCIe 4.0, there is the option to disable in-band presence so that the PDS bit always reflects the state of the out-of-band presence.

The recommendation of the PCIe spec is to disable in-band presence whenever supported.

---
 drivers/pci/hotplug/pciehp.h     | 1 +
 drivers/pci/hotplug/pciehp_hpc.c | 9 ++++++++-
 include/uapi/linux/pci_regs.h    | 2 ++
 3 files changed, 11 insertions(+), 1 deletion(-)
drivers/pci/hotplug/pciehp.h +1 -0
index 654c972b8ea0..27e4cd6529b0 100644
@@ -83,6 +83,7 @@ struct controller {
 	struct pcie_device *pcie;
 
 	u32 slot_cap;				/* capabilities and quirks */
+	unsigned int inband_presence_disabled:1;
 
 	u16 slot_ctrl;				/* control register access */
 	struct mutex ctrl_lock;
drivers/pci/hotplug/pciehp_hpc.c +8 -1
index 1a522c1c4177..dc109d521f30 100644
@@ -811,7 +811,7 @@ static inline void dbg_ctrl(struct controller *ctrl)
 struct controller *pcie_init(struct pcie_device *dev)
 {
 	struct controller *ctrl;
-	u32 slot_cap, link_cap;
+	u32 slot_cap, slot_cap2, link_cap;
 	u8 poweron;
 	struct pci_dev *pdev = dev->port;
 	struct pci_bus *subordinate = pdev->subordinate;
@@ -869,6 +869,13 @@ struct controller *pcie_init(struct pcie_device *dev)
 		FLAG(link_cap, PCI_EXP_LNKCAP_DLLLARC),
 		pdev->broken_cmd_compl ? " (with Cmd Compl erratum)" : "");
 
+	pcie_capability_read_dword(pdev, PCI_EXP_SLTCAP2, &slot_cap2);
+	if (slot_cap2 & PCI_EXP_SLTCAP2_IBPD) {
+		pcie_write_cmd_nowait(ctrl, PCI_EXP_SLTCTL_IBPD_DISABLE,
+				      PCI_EXP_SLTCTL_IBPD_DISABLE);
+		ctrl->inband_presence_disabled = 1;
+	}
+
 	/*
 	 * If empty slot's power status is on, turn power off.  The IRQ isn't
 	 * requested yet, so avoid triggering a notification with this command.
include/uapi/linux/pci_regs.h +2 -0
index 29d6e93fd15e..ea1cf9546e4d 100644
@@ -604,6 +604,7 @@
 #define  PCI_EXP_SLTCTL_PWR_OFF        0x0400 /* Power Off */
 #define  PCI_EXP_SLTCTL_EIC	0x0800	/* Electromechanical Interlock Control */
 #define  PCI_EXP_SLTCTL_DLLSCE	0x1000	/* Data Link Layer State Changed Enable */
+#define  PCI_EXP_SLTCTL_IBPD_DISABLE	0x4000 /* In-band PD disable */
 #define PCI_EXP_SLTSTA		26	/* Slot Status */
 #define  PCI_EXP_SLTSTA_ABP	0x0001	/* Attention Button Pressed */
 #define  PCI_EXP_SLTSTA_PFD	0x0002	/* Power Fault Detected */
@@ -676,6 +677,7 @@
 #define PCI_EXP_LNKSTA2		50	/* Link Status 2 */
 #define PCI_CAP_EXP_ENDPOINT_SIZEOF_V2	52	/* v2 endpoints with link end here */
 #define PCI_EXP_SLTCAP2		52	/* Slot Capabilities 2 */
+#define  PCI_EXP_SLTCAP2_IBPD	0x0001	/* In-band PD Disable Supported */
 #define PCI_EXP_SLTCTL2		56	/* Slot Control 2 */
 #define PCI_EXP_SLTSTA2		58	/* Slot Status 2 */
 
Signature · 1 line
2.18.1
#2 lore raw

PATCH 2/3 PCI: pciehp: Wait for PDS if in-band presence is disabled

In reply to Stuart Hayes

When inband presence is disabled, PDS may come up at any time, or not at all. PDS being low may indicate that the card is still mating, and we could expect contact bounce to bring down the link as well.

It is reasonable to assume that most cards will mate in a hotplug slot in about a second. Thus, when we know PDS only reflects out-of-band presence, it's worthwhile to wait the extra second or so to make sure the card is properly mated before loading the driver, and to prevent the hotplug code from disabling a device if the presence detect change goes active after the device is enabled.

---
 drivers/pci/hotplug/pciehp_hpc.c | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
drivers/pci/hotplug/pciehp_hpc.c +22 -0
index dc109d521f30..1282641c6458 100644
@@ -242,6 +242,25 @@ static bool pci_bus_check_dev(struct pci_bus *bus, int devfn)
 	return found;
 }
 
+static void pcie_wait_for_presence(struct pci_dev *pdev)
+{
+	int timeout = 1250;
+	bool pds;
+	u16 slot_status;
+
+	while (true) {
+		pcie_capability_read_word(pdev, PCI_EXP_SLTSTA, &slot_status);
+		pds = !!(slot_status & PCI_EXP_SLTSTA_PDS);
+		if (pds || timeout <= 0)
+			break;
+		msleep(10);
+		timeout -= 10;
+	}
+
+	if (!pds)
+		pci_info(pdev, "Presence Detect state not set in 1250 msec\n");
+}
+
 int pciehp_check_link_status(struct controller *ctrl)
 {
 	struct pci_dev *pdev = ctrl_dev(ctrl);
@@ -251,6 +270,9 @@ int pciehp_check_link_status(struct controller *ctrl)
 	if (!pcie_wait_for_link(pdev, true))
 		return -1;
 
+	if (ctrl->inband_presence_disabled)
+		pcie_wait_for_presence(pdev);
+
 	found = pci_bus_check_dev(ctrl->pcie->port->subordinate,
 					PCI_DEVFN(0, 0));
 
Signature · 1 line
2.18.1
#3 lore raw

PATCH 3/3 PCI: pciehp: Add dmi table for in-band presence disabled

In reply to Stuart Hayes

Some systems have in-band presence detection disabled for hot-plug PCI slots, but do not report this in the slot capabilities 2 (SLTCAP2) register. On these systems, presence detect can become active well after the link is reported to be active, which can cause the slots to be disabled after a device is connected.

Add a dmi table to flag these systems as having in-band presence disabled.

---
 drivers/pci/hotplug/pciehp_hpc.c | 14 ++++++++++++++
 1 file changed, 14 insertions(+)
drivers/pci/hotplug/pciehp_hpc.c +14 -0
index 1282641c6458..1dd01e752587 100644
@@ -14,6 +14,7 @@
 
 #define dev_fmt(fmt) "pciehp: " fmt
 
+#include <linux/dmi.h>
 #include <linux/kernel.h>
 #include <linux/types.h>
 #include <linux/jiffies.h>
@@ -26,6 +27,16 @@
 #include "../pci.h"
 #include "pciehp.h"
 
+static const struct dmi_system_id inband_presence_disabled_dmi_table[] = {
+	{
+		.ident = "Dell System",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
+		},
+	},
+	{}
+};
+
 static inline struct pci_dev *ctrl_dev(struct controller *ctrl)
 {
 	return ctrl->pcie->port;
@@ -898,6 +909,9 @@ struct controller *pcie_init(struct pcie_device *dev)
 		ctrl->inband_presence_disabled = 1;
 	}
 
+	if (dmi_first_match(inband_presence_disabled_dmi_table))
+		ctrl->inband_presence_disabled = 1;
+
 	/*
 	 * If empty slot's power status is on, turn power off.  The IRQ isn't
 	 * requested yet, so avoid triggering a notification with this command.
Signature · 1 line
2.18.1
#4 lore raw

Re: [PATCH 3/3] PCI: pciehp: Add dmi table for in-band presence disabled

Quoted 29 lines · Stuart Hayes

Some systems have in-band presence detection disabled for hot-plug PCI slots, but do not report this in the slot capabilities 2 (SLTCAP2) register. On these systems, presence detect can become active well after the link is reported to be active, which can cause the slots to be disabled after a device is connected.

Add a dmi table to flag these systems as having in-band presence disabled.

---
  drivers/pci/hotplug/pciehp_hpc.c | 14 ++++++++++++++
  1 file changed, 14 insertions(+)
drivers/pci/hotplug/pciehp_hpc.c +1 -0
index 1282641c6458..1dd01e752587 100644
@@ -14,6 +14,7 @@
 
  #define dev_fmt(fmt) "pciehp: " fmt
 
+#include <linux/dmi.h>
  #include <linux/kernel.h>
  #include <linux/types.h>
  #include <linux/jiffies.h>
@@ -26,6 +27,16 @@
  #include "../pci.h"
  #include "pciehp.h"
 

drivers/pci/hotplug/pciehp_hpc.c line 39

Quoted 10 lines
drivers/pci/hotplug/pciehp_hpc.c +10 -0
+static const struct dmi_system_id inband_presence_disabled_dmi_table[] = {
+	{
+		.ident = "Dell System",
+		.matches = {
+			DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
+		},
+	},
+	{}
+};
+

I'm not sure that all Dell systems that were ever made or will be made have in-band presence disabled on all their hotplug slots.

This was a problem with the NVMe hot-swap bays on 14G servers. I can't guarantee that any other slot or machine will need this workaround. The best way I found to implement this is to check for the PCI-ID of the switches behind the port.

Alex

#5 lore raw

Re: [PATCH 3/3] PCI: pciehp: Add dmi table for in-band presence disabled

Quoted 47 lines · Alex G.
Quoted 29 lines · Stuart Hayes

Some systems have in-band presence detection disabled for hot-plug PCI slots, but do not report this in the slot capabilities 2 (SLTCAP2) register. On these systems, presence detect can become active well after the link is reported to be active, which can cause the slots to be disabled after a device is connected.

Add a dmi table to flag these systems as having in-band presence disabled.

---
  drivers/pci/hotplug/pciehp_hpc.c | 14 ++++++++++++++
  1 file changed, 14 insertions(+)
drivers/pci/hotplug/pciehp_hpc.c +1 -0
index 1282641c6458..1dd01e752587 100644
@@ -14,6 +14,7 @@
 
  #define dev_fmt(fmt) "pciehp: " fmt
 
+#include <linux/dmi.h>
  #include <linux/kernel.h>
  #include <linux/types.h>
  #include <linux/jiffies.h>
@@ -26,6 +27,16 @@
  #include "../pci.h"
  #include "pciehp.h"
 

drivers/pci/hotplug/pciehp_hpc.c line 39

Quoted 10 lines
drivers/pci/hotplug/pciehp_hpc.c +10 -0
+static const struct dmi_system_id inband_presence_disabled_dmi_table[] = {
+     {
+             .ident = "Dell System",
+             .matches = {
+                     DMI_MATCH(DMI_SYS_VENDOR, "Dell Inc"),
+             },
+     },
+     {}
+};
+

I'm not sure that all Dell systems that were ever made or will be made have in-band presence disabled on all their hotplug slots.

Quoted 6 lines

This was a problem with the NVMe hot-swap bays on 14G servers. I can't guarantee that any other slot or machine will need this workaround. The best way I found to implement this is to check for the PCI-ID of the switches behind the port.

Alex

That is definitely true, not all Dell systems actually disable in-band presence detect and need this. However, we have a number of systems that need this, and we don't have the PCI IDs for all of these yet, so we decided it was preferable to just make all Dell systems wait for presence detect to go active. Since all of our systems support presence detect (either in-band or out-of-band), it shouldn't have any negative effects--on systems that support in-band presence, it will already be active and it won't spend any extra time waiting for it. If someone plugs in a device that has hot-plug slots with no support for presence detect at all (even though Dell doesn't support any), it should still work--it'll just take an extra second for them to come up.

#6 lore raw
In reply to Stuart Hayes
Quoted 5 lines · Stuart Hayes

This patch set is based on a patch set [1] submitted many months ago by Alexandru Gagniuc, who is no longer working on it.

[1] https://patchwork.kernel.org/cover/10909167/
[v3,0/4] PCI: pciehp: Do not turn off slot if presence comes up after link

If I'm not mistaken, these two are identical to Alex' patches, right?

PCI: pciehp: Add support for disabling in-band presence
PCI: pciehp: Wait for PDS when in-band presence is disabled

I'm not sure if it's appropriate to change the author and omit Alex' Signed-off-by.

Otherwise I have no objections against this series.

Thanks,

Lukas

#7 lore raw
Quoted 13 lines · Lukas Wunner
Quoted 5 lines · Stuart Hayes

This patch set is based on a patch set [1] submitted many months ago by Alexandru Gagniuc, who is no longer working on it.

[1] https://patchwork.kernel.org/cover/10909167/
[v3,0/4] PCI: pciehp: Do not turn off slot if presence comes up after link

If I'm not mistaken, these two are identical to Alex' patches, right?

PCI: pciehp: Add support for disabling in-band presence
PCI: pciehp: Wait for PDS when in-band presence is disabled

Quoted 8 lines

I'm not sure if it's appropriate to change the author and omit Alex' Signed-off-by.

Otherwise I have no objections against this series.

Thanks,

Lukas

Thanks! The first patch is identical to the one Alex submitted, and the second is nearly so... they both basically his work. I wasn't sure what proper etiquette was--I was thinking the signed-off-by was taking responsibility that the patch was ok (functional, not copyrighted by someone else, etc) rather than giving credit, but he definitely deserves credit for them. I'm happy to add a signed-off-by for Alex on the first two and resubmit if he doesn't object.

#8 lore raw
In reply to Lukas Wunner
Quoted 7 lines · Lukas Wunner
Quoted 5 lines · Stuart Hayes

This patch set is based on a patch set [1] submitted many months ago by Alexandru Gagniuc, who is no longer working on it.

[1] https://patchwork.kernel.org/cover/10909167/
[v3,0/4] PCI: pciehp: Do not turn off slot if presence comes up after link

Quoted 7 lines

If I'm not mistaken, these two are identical to Alex' patches, right?

PCI: pciehp: Add support for disabling in-band presence
PCI: pciehp: Wait for PDS when in-band presence is disabled

I'm not sure if it's appropriate to change the author and omit Alex' Signed-off-by.

Legally Dell owns the patches. I have no objection on my end.

Alex

Quoted 6 lines

Otherwise I have no objections against this series.

Thanks,

Lukas

#9 lore raw
Quoted 9 lines · Alex G.
Quoted 6 lines · Lukas Wunner
Quoted 2 lines · Stuart Hayes

This patch set is based on a patch set [1] submitted many months ago by Alexandru Gagniuc, who is no longer working on it.

I'm not sure if it's appropriate to change the author and omit Alex' Signed-off-by.

Legally Dell owns the patches. I have no objection on my end.

From a kernel community POV, I don't think it matters (in this case) who legally owns the copyright to the contributed code. It's just that we go to great lengths to provide proper attribution even for small contributions (e.g. Tested-by).

The benefit to the community is that we know who to cc if that portion of the code is changed again and someone knowledgable should take a look.

The benefit to contributors is that they can change jobs and their past contributions are still visible in the git history and attributed to their names. By contrast, if you've worked on closed source code and changed jobs, that work isn't visible to future employers or even yourself, and it may happen that someone else takes credit for your past work without you even knowing about it or being able to stop that. (I've seen it before.)

In this case, there should be a S-o-b line for Alex preceding that for Stuart, and the author of the commit should be Alex unless a significant portion of the patch was changed.

Thanks,

Lukas