PCI: MSI: Deal with devices lying about their masking capability

PATCH Applied Marc Zyngier 5 participants 10 messages Nov 4, 2021 – Nov 19, 2021

Subsystems (2)

Stable branch PCI

lore

Share & tools

lore

Marc Zyngier

author

2021-11-04 18:01 UTC #0 lore raw

cover PCI: MSI: Deal with devices lying about their masking capability

Rui reported[1] that his Nvidia ION system stopped working with 5.15, with the AHCI device failing to get any MSI. A rapid investigation revealed that although the device doesn't advertise MSI masking, it actually needs it. Quality hardware indeed.

Anyway, the couple of patches below are an attempt at dealing with the issue in a more or less generic way.

[1] https://lore.kernel.org/r/CALjTZvbzYfBuLB+H=fj2J+9=DxjQ2Uqcy0if_PvmJ-nU-qEgkg@mail.gmail.com

Marc Zyngier (2):
PCI: MSI: Deal with devices lying about their MSI mask capability
PCI: Add MSI masking quirk for Nvidia ION AHCI

 drivers/pci/msi.c    | 3 +++
 drivers/pci/quirks.c | 6 ++++++
 include/linux/pci.h  | 2 ++
 3 files changed, 11 insertions(+)

Signature · 1 line

2.30.2

Marc Zyngier

author

2021-11-04 18:01 UTC #1 lore raw

PATCH 1/2 PCI: MSI: Deal with devices lying about their MSI mask capability

It appears that some devices are lying about their mask capability, pretending that they don't have it, while they actually do. The net result is that now that we don't enable MSIs on such endpoint.

Add a new per-device flag to deal with this. Further patches will make use of it, sadly.

Signed-off-by Marc Zyngier <[email protected]>

---
 drivers/pci/msi.c   | 3 +++
 include/linux/pci.h | 2 ++
 2 files changed, 5 insertions(+)

drivers/pci/msi.c +3 -0

index 0099a00af361..2f9ec7210991 100644
@@ -479,6 +479,9 @@ msi_setup_entry(struct pci_dev *dev, int nvec, struct irq_affinity *affd)
 		goto out;
 
 	pci_read_config_word(dev, dev->msi_cap + PCI_MSI_FLAGS, &control);
+	/* Lies, damned lies, and MSIs */
+	if (dev->dev_flags & PCI_DEV_FLAGS_HAS_MSI_MASKING)
+		control |= PCI_MSI_FLAGS_MASKBIT;
 
 	entry->msi_attrib.is_msix	= 0;
 	entry->msi_attrib.is_64		= !!(control & PCI_MSI_FLAGS_64BIT);

include/linux/pci.h +2 -0

index cd8aa6fce204..152a4d74f87f 100644
@@ -233,6 +233,8 @@ enum pci_dev_flags {
 	PCI_DEV_FLAGS_NO_FLR_RESET = (__force pci_dev_flags_t) (1 << 10),
 	/* Don't use Relaxed Ordering for TLPs directed at this device */
 	PCI_DEV_FLAGS_NO_RELAXED_ORDERING = (__force pci_dev_flags_t) (1 << 11),
+	/* Device does honor MSI masking despite saying otherwise */
+	PCI_DEV_FLAGS_HAS_MSI_MASKING = (__force pci_dev_flags_t) (1 << 12),
 };
 
 enum pci_irq_reroute_variant {

Signature · 1 line

2.30.2

tip-bot2 for Marc Zyngier

merge bot Applied

2021-11-11 08:57 UTC #2 lore raw

[tip: irq/urgent] PCI/MSI: Deal with devices lying about their MSI mask capability

The following commit has been merged into the irq/urgent branch of tip:

Commit-ID:     2226667a145db2e1f314d7f57fd644fe69863ab9
Gitweb:        https://git.kernel.org/tip/2226667a145db2e1f314d7f57fd644fe69863ab9
Author:        Marc Zyngier <[email protected]>
AuthorDate:    Thu, 04 Nov 2021 18:01:29 
Committer:     Thomas Gleixner <[email protected]>
CommitterDate: Thu, 11 Nov 2021 09:50:30 +01:00

PCI/MSI: Deal with devices lying about their MSI mask capability

It appears that some devices are lying about their mask capability, pretending that they don't have it, while they actually do. The net result is that now that we don't enable MSIs on such endpoint.

Add a new per-device flag to deal with this. Further patches will make use of it, sadly.

Signed-off-by Marc Zyngier <[email protected]>

Signed-off-by Thomas Gleixner <[email protected]>

Reviewed-by Thomas Gleixner <[email protected]>

Link https://lore.kernel.org/r/[email protected] open here

Cc Bjorn Helgaas <[email protected]>

---
 drivers/pci/msi.c   | 3 +++
 include/linux/pci.h | 2 ++
 2 files changed, 5 insertions(+)

drivers/pci/msi.c +3 -0

index 6da7910..7043301 100644
@@ -477,6 +477,9 @@ msi_setup_entry(struct pci_dev *dev, int nvec, struct irq_affinity *affd)
 		goto out;
 
 	pci_read_config_word(dev, dev->msi_cap + PCI_MSI_FLAGS, &control);
+	/* Lies, damned lies, and MSIs */
+	if (dev->dev_flags & PCI_DEV_FLAGS_HAS_MSI_MASKING)
+		control |= PCI_MSI_FLAGS_MASKBIT;
 
 	entry->msi_attrib.is_msix	= 0;
 	entry->msi_attrib.is_64		= !!(control & PCI_MSI_FLAGS_64BIT);

include/linux/pci.h +2 -0

index c8afbee..d0dba7f 100644
@@ -233,6 +233,8 @@ enum pci_dev_flags {
 	PCI_DEV_FLAGS_NO_FLR_RESET = (__force pci_dev_flags_t) (1 << 10),
 	/* Don't use Relaxed Ordering for TLPs directed at this device */
 	PCI_DEV_FLAGS_NO_RELAXED_ORDERING = (__force pci_dev_flags_t) (1 << 11),
+	/* Device does honor MSI masking despite saying otherwise */
+	PCI_DEV_FLAGS_HAS_MSI_MASKING = (__force pci_dev_flags_t) (1 << 12),
 };
 
 enum pci_irq_reroute_variant {

Marc Zyngier

author

2021-11-04 18:01 UTC #3 lore raw

PATCH 2/2 PCI: Add MSI masking quirk for Nvidia ION AHCI

In reply to Marc Zyngier

The ION AHCI device pretends that MSI masking isn't a thing, while it actually implements it and needs MSIs to be unmasked to work. Add a quirk to that effect.

Reported-by Rui Salvaterra <[email protected]>

Tested-by Rui Salvaterra <[email protected]>

Signed-off-by Marc Zyngier <[email protected]>

Link https://lore.kernel.org/r/CALjTZvbzYfBuLB+H=fj2J+9=DxjQ2Uqcy0if_PvmJ-nU-qEgkg@mail.gmail.com open here

---
 drivers/pci/quirks.c | 6 ++++++
 1 file changed, 6 insertions(+)

drivers/pci/quirks.c +6 -0

index 4537d1ea14fd..89c7c99cd1bb 100644
@@ -5795,3 +5795,9 @@ static void apex_pci_fixup_class(struct pci_dev *pdev)
 }
 DECLARE_PCI_FIXUP_CLASS_HEADER(0x1ac1, 0x089a,
 			       PCI_CLASS_NOT_DEFINED, 8, apex_pci_fixup_class);
+
+static void nvidia_ion_ahci_fixup(struct pci_dev *pdev)
+{
+	pdev->dev_flags |= PCI_DEV_FLAGS_HAS_MSI_MASKING;
+}
+DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_NVIDIA, 0x0ab8, nvidia_ion_ahci_fixup);

Signature · 1 line

2.30.2

Thomas Gleixner

Reviewed-by on the v1 series

2021-11-05 13:14 UTC #4 lore raw

In reply to Marc Zyngier

Quoted 10 lines · Marc Zyngier

Rui reported[1] that his Nvidia ION system stopped working with 5.15, with the AHCI device failing to get any MSI. A rapid investigation revealed that although the device doesn't advertise MSI masking, it actually needs it. Quality hardware indeed.
Anyway, the couple of patches below are an attempt at dealing with the issue in a more or less generic way.
[1] https://lore.kernel.org/r/CALjTZvbzYfBuLB+H=fj2J+9=DxjQ2Uqcy0if_PvmJ-nU-qEgkg@mail.gmail.com

Quoted 8 lines

Marc Zyngier (2):
PCI: MSI: Deal with devices lying about their MSI mask capability
PCI: Add MSI masking quirk for Nvidia ION AHCI
 drivers/pci/msi.c    | 3 +++
 drivers/pci/quirks.c | 6 ++++++
 include/linux/pci.h  | 2 ++
 3 files changed, 11 insertions(+)

Groan.

Reviewed-by Thomas Gleixner <[email protected]>

Rui Salvaterra

2021-11-16 10:21 UTC #5 lore raw

Hi, Thomas,

Quoted 16 lines · Thomas Gleixner

Quoted 14 lines · Marc Zyngier
Rui reported[1] that his Nvidia ION system stopped working with 5.15, with the AHCI device failing to get any MSI. A rapid investigation revealed that although the device doesn't advertise MSI masking, it actually needs it. Quality hardware indeed.
Anyway, the couple of patches below are an attempt at dealing with the issue in a more or less generic way.
[1] https://lore.kernel.org/r/CALjTZvbzYfBuLB+H=fj2J+9=DxjQ2Uqcy0if_PvmJ-nU-qEgkg@mail.gmail.com
Marc Zyngier (2):
PCI: MSI: Deal with devices lying about their MSI mask capability
PCI: Add MSI masking quirk for Nvidia ION AHCI

Quoted 8 lines

Quoted 4 lines
 drivers/pci/msi.c    | 3 +++
 drivers/pci/quirks.c | 6 ++++++
 include/linux/pci.h  | 2 ++
 3 files changed, 11 insertions(+)
Groan.
Reviewed-by Thomas Gleixner <[email protected]>

Just a reminder, to make sure this doesn't fall through the cracks. It's already in 5.16, but needs to be backported to 5.15. I'm not seeing it in Greg's 5.15 stable queue yet.

Thanks,
Rui

Greg KH

maintainer

2021-11-16 10:39 UTC #6 lore raw

Quoted 24 lines · Rui Salvaterra

Hi, Thomas,
Quoted 21 lines · Thomas Gleixner
Quoted 18 lines · Marc Zyngier
Rui reported[1] that his Nvidia ION system stopped working with 5.15, with the AHCI device failing to get any MSI. A rapid investigation revealed that although the device doesn't advertise MSI masking, it actually needs it. Quality hardware indeed.
Anyway, the couple of patches below are an attempt at dealing with the issue in a more or less generic way.
[1] https://lore.kernel.org/r/CALjTZvbzYfBuLB+H=fj2J+9=DxjQ2Uqcy0if_PvmJ-nU-qEgkg@mail.gmail.com
Marc Zyngier (2):
PCI: MSI: Deal with devices lying about their MSI mask capability
PCI: Add MSI masking quirk for Nvidia ION AHCI
 drivers/pci/msi.c    | 3 +++
 drivers/pci/quirks.c | 6 ++++++
 include/linux/pci.h  | 2 ++
 3 files changed, 11 insertions(+)

Quoted 7 lines

Quoted 3 lines
Groan.
Reviewed-by Thomas Gleixner <[email protected]>
Just a reminder, to make sure this doesn't fall through the cracks. It's already in 5.16, but needs to be backported to 5.15. I'm not seeing it in Greg's 5.15 stable queue yet.

What is the git commit ids of these changes in Linus's tree?

thanks,

greg k-h

Rui Salvaterra

2021-11-16 10:47 UTC #7 lore raw

Hi, Greg,

Quoted 2 lines · Greg KH

What is the git commit ids of these changes in Linus's tree?

2226667a145d ("PCI/MSI: Deal with devices lying about their MSI mask capability")
f21082fb20db ("PCI: Add MSI masking quirk for Nvidia ION AHCI")

Thanks,
Rui

Greg KH

maintainer

2021-11-16 10:56 UTC #8 lore raw

Quoted 9 lines · Rui Salvaterra

Hi, Greg,
Quoted 2 lines · Greg KH
What is the git commit ids of these changes in Linus's tree?
2226667a145d ("PCI/MSI: Deal with devices lying about their MSI mask capability")
f21082fb20db ("PCI: Add MSI masking quirk for Nvidia ION AHCI")

Thanks, I'll queue them up in the next round of releases after the current ones are out.

greg k-h

Greg KH

maintainer

2021-11-19 14:06 UTC #9 lore raw

Quoted 5 lines · Greg KH

Quoted 4 lines · Rui Salvaterra
Hi, Greg,
Quoted 1 line · Greg KH

Quoted 8 lines

Quoted 5 lines
Quoted 1 line
What is the git commit ids of these changes in Linus's tree?
2226667a145d ("PCI/MSI: Deal with devices lying about their MSI mask capability")
f21082fb20db ("PCI: Add MSI masking quirk for Nvidia ION AHCI")
Thanks, I'll queue them up in the next round of releases after the current ones are out.

Now queued up.

greg k-h