Skip to contentSkip to conversation

PCI/AER: Add error message when AER_MAX_MULTI_ERR_DEVICES limit is hit during AER handling

PATCH Applied Akshay Jindal 2 participants 3 messages Jun 19, 2025 – Jun 26, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 PCI/AER: Add error message when AER_MAX_MULTI_ERR_DEVICES limit is hit during AER handling

When a PCIe error is detected, the root port receives the error message and the threaded IRQ handler, aer_isr, traverses the hierarchy downward from the root port. It populates the e_info->dev[] array with the PCIe devices that have recorded error status, so that appropriate error handling and recovery can be performed.

The e_info->dev[] array is limited in size by AER_MAX_MULTI_ERR_DEVICES, which is currently defined as 5. If more than five devices report errors in the same event, the array silently truncates the list, and those extra devices are not included in the recovery flow.

Emit an error message when this limit is reached, fulfilling a TODO comment in drivers/pci/pcie/aer.c.
/* TODO: Should print error message here? */

---

Changes since v1:
- Reworded commit message in imperative mood (per Shuah’s feedback)
- Mentioned and quoted related TODO in the message
- Updated recipient list

Testing:
========
Verified log in dmesg on QEMU.

1. Following command created the required environment. As mentioned below a pcie-root-port and a virtio-net-pci device are used on a Q35 machine model.

./qemu-system-x86_64 \
	-M q35,accel=kvm \
	-m 2G -cpu host -nographic \
	-serial mon:stdio \
	-kernel /home/akshayaj/pci/arch/x86/boot/bzImage \
	-initrd /home/akshayaj/Embedded_System_Using_QEMU/rootfs/rootfs.cpio.gz \
	-append "console=ttyS0 root=/ pci=pcie_scan_all" \
	-device pcie-root-port,id=rp0,chassis=1,slot=1 \
	-device virtio-net-pci,bus=rp0

~ # mylspci -t
-[0000:00]-+-00.0

           +-01.0
           +-02.0
           +-03.0-[01]----00.0
           +-1f.0
           +-1f.2
           \-1f.3

00:03.0--> pcie-root-port

2. Kernel bzImage compiled with following changes:
2.1 CONFIG_PCIEAER=y in config
2.2 AER_MAX_MULTI_ERR_DEVICES set to 0
Since there is no pcie-testdev in QEMU, it is impossible to create a 5-level hierarchy of PCIe devices in QEMU. So we simulate the error scenario by changing the limit to 0.
2.3 Log added at the required place in aer.c.

3. Both correctable and uncorrectable errors were injected on pcie-root-port via HMP command (pcie_aer_inject_error) in QEMU. HMP Command used are as follows:

	3.1 pcie_aer_inject_error -c rp0 0x1
	3.2 pcie_aer_inject_error -c rp0 0x40
	3.3 pcie_aer_inject_error rp0 0x10

Resulting dmesg:
================

[    0.380534] pcieport 0000:00:03.0: AER: enabled with IRQ 24
[   55.729530] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
[  225.484456] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
[  356.976253] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
 drivers/pci/pcie/aer.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/pcie/aer.c +2 -1
index 70ac66188367..3995a1db5699 100644
@@ -1039,7 +1039,8 @@ static int find_device_iter(struct pci_dev *dev, void *data)
 		/* List this device */
 		if (add_error_device(e_info, dev)) {
 			/* We cannot handle more... Stop iteration */
-			/* TODO: Should print error message here? */
+			pci_err(dev, "Exceeded max allowed (%d) addition of PCIe "
+				"devices for AER handling\n", AER_MAX_MULTI_ERR_DEVICES);
 			return 1;
 		}
 
Signature · 1 line
2.43.0
#1 lore raw

Hi,
Is there any feedback on the patch?

Thanks,
Akshay

Quoted 92 lines · Akshay Jindal

When a PCIe error is detected, the root port receives the error message and the threaded IRQ handler, aer_isr, traverses the hierarchy downward from the root port. It populates the e_info->dev[] array with the PCIe devices that have recorded error status, so that appropriate error handling and recovery can be performed.

The e_info->dev[] array is limited in size by AER_MAX_MULTI_ERR_DEVICES, which is currently defined as 5. If more than five devices report errors in the same event, the array silently truncates the list, and those extra devices are not included in the recovery flow.

Emit an error message when this limit is reached, fulfilling a TODO comment in drivers/pci/pcie/aer.c.
/* TODO: Should print error message here? */

---

Changes since v1:
- Reworded commit message in imperative mood (per Shuah’s feedback)
- Mentioned and quoted related TODO in the message
- Updated recipient list

Testing:
========
Verified log in dmesg on QEMU.

1. Following command created the required environment. As mentioned below a pcie-root-port and a virtio-net-pci device are used on a Q35 machine model.

./qemu-system-x86_64 \
        -M q35,accel=kvm \
        -m 2G -cpu host -nographic \
        -serial mon:stdio \
        -kernel /home/akshayaj/pci/arch/x86/boot/bzImage \
        -initrd /home/akshayaj/Embedded_System_Using_QEMU/rootfs/rootfs.cpio.gz \
        -append "console=ttyS0 root=/ pci=pcie_scan_all" \
        -device pcie-root-port,id=rp0,chassis=1,slot=1 \
        -device virtio-net-pci,bus=rp0

~ # mylspci -t
-[0000:00]-+-00.0

           +-01.0
           +-02.0
           +-03.0-[01]----00.0
           +-1f.0
           +-1f.2
           \-1f.3

00:03.0--> pcie-root-port

2. Kernel bzImage compiled with following changes:
2.1 CONFIG_PCIEAER=y in config
2.2 AER_MAX_MULTI_ERR_DEVICES set to 0
Since there is no pcie-testdev in QEMU, it is impossible to create a 5-level hierarchy of PCIe devices in QEMU. So we simulate the error scenario by changing the limit to 0.
2.3 Log added at the required place in aer.c.

3. Both correctable and uncorrectable errors were injected on pcie-root-port via HMP command (pcie_aer_inject_error) in QEMU. HMP Command used are as follows:

        3.1 pcie_aer_inject_error -c rp0 0x1
        3.2 pcie_aer_inject_error -c rp0 0x40
        3.3 pcie_aer_inject_error rp0 0x10

Resulting dmesg:
================

[    0.380534] pcieport 0000:00:03.0: AER: enabled with IRQ 24
[   55.729530] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
[  225.484456] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
[  356.976253] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
 drivers/pci/pcie/aer.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/pcie/aer.c +2 -2
index 70ac66188367..3995a1db5699 100644
@@ -1039,7 +1039,8 @@ static int find_device_iter(struct pci_dev *dev, void *data)
                /* List this device */
                if (add_error_device(e_info, dev)) {
                        /* We cannot handle more... Stop iteration */
-                       /* TODO: Should print error message here? */
+                       pci_err(dev, "Exceeded max allowed (%d) addition of PCIe "
+                               "devices for AER handling\n", AER_MAX_MULTI_ERR_DEVICES);
                        return 1;
                }
 
--

2.43.0

#2 lore raw
In reply to Akshay Jindal
Quoted 11 lines · Akshay Jindal

When a PCIe error is detected, the root port receives the error message and the threaded IRQ handler, aer_isr, traverses the hierarchy downward from the root port. It populates the e_info->dev[] array with the PCIe devices that have recorded error status, so that appropriate error handling and recovery can be performed.

The e_info->dev[] array is limited in size by AER_MAX_MULTI_ERR_DEVICES, which is currently defined as 5. If more than five devices report errors in the same event, the array silently truncates the list, and those extra devices are not included in the recovery flow.

Quoted 5 lines

Emit an error message when this limit is reached, fulfilling a TODO comment in drivers/pci/pcie/aer.c.
/* TODO: Should print error message here? */

Applied to pci/aer for v6.17, thanks!

Quoted 75 lines

---

Changes since v1:
- Reworded commit message in imperative mood (per Shuah’s feedback)
- Mentioned and quoted related TODO in the message
- Updated recipient list

Testing:
========
Verified log in dmesg on QEMU.

1. Following command created the required environment. As mentioned below a pcie-root-port and a virtio-net-pci device are used on a Q35 machine model.

./qemu-system-x86_64 \
	-M q35,accel=kvm \
	-m 2G -cpu host -nographic \
	-serial mon:stdio \
	-kernel /home/akshayaj/pci/arch/x86/boot/bzImage \
	-initrd /home/akshayaj/Embedded_System_Using_QEMU/rootfs/rootfs.cpio.gz \
	-append "console=ttyS0 root=/ pci=pcie_scan_all" \
	-device pcie-root-port,id=rp0,chassis=1,slot=1 \
	-device virtio-net-pci,bus=rp0

~ # mylspci -t
-[0000:00]-+-00.0

           +-01.0
           +-02.0
           +-03.0-[01]----00.0
           +-1f.0
           +-1f.2
           \-1f.3

00:03.0--> pcie-root-port

2. Kernel bzImage compiled with following changes:
2.1 CONFIG_PCIEAER=y in config
2.2 AER_MAX_MULTI_ERR_DEVICES set to 0
Since there is no pcie-testdev in QEMU, it is impossible to create a 5-level hierarchy of PCIe devices in QEMU. So we simulate the error scenario by changing the limit to 0.
2.3 Log added at the required place in aer.c.

3. Both correctable and uncorrectable errors were injected on pcie-root-port via HMP command (pcie_aer_inject_error) in QEMU. HMP Command used are as follows:

	3.1 pcie_aer_inject_error -c rp0 0x1
	3.2 pcie_aer_inject_error -c rp0 0x40
	3.3 pcie_aer_inject_error rp0 0x10

Resulting dmesg:
================

[    0.380534] pcieport 0000:00:03.0: AER: enabled with IRQ 24
[   55.729530] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
[  225.484456] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
[  356.976253] pcieport 0000:00:03.0: AER: Exceeded max allowed (0) addition of PCIe devices for AER handling
 drivers/pci/pcie/aer.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/pci/pcie/aer.c +2 -1
index 70ac66188367..3995a1db5699 100644
@@ -1039,7 +1039,8 @@ static int find_device_iter(struct pci_dev *dev, void *data)
 		/* List this device */
 		if (add_error_device(e_info, dev)) {
 			/* We cannot handle more... Stop iteration */
-			/* TODO: Should print error message here? */
+			pci_err(dev, "Exceeded max allowed (%d) addition of PCIe "
+				"devices for AER handling\n", AER_MAX_MULTI_ERR_DEVICES);
 			return 1;
 		}
 
Signature · 1 line
2.43.0