Re: [PATCH v1] PCI/AER: Handle Multi UnCorrectable/Correctable errors properly

From: Sathyanarayanan Kuppuswamy
Date: Fri Mar 18 2022 - 16:28:29 EST




On 3/17/22 3:59 PM, Bjorn Helgaas wrote:
Thanks for the additional details!

After this patch, I guess aer_irq() still reads 0x2
(PCI_ERR_ROOT_MULTI_COR_RCV), but now it writes 0x2 back which clears
PCI_ERR_ROOT_MULTI_COR_RCV.

In addition, aer_irq() will continue on to read PCI_ERR_ROOT_ERR_SRC,
which probably contains either 0 or junk left over from being captured
when PCI_ERR_ROOT_COR_RCV was set.

And aer_irq() will queue an e_src record with status ==
PCI_ERR_ROOT_MULTI_COR_RCV. But since PCI_ERR_ROOT_COR_RCV is not set
in status, aer_isr_one_error() will do nothing, right?

That might not be*terrible* and is definitely better than not being
able to handle future interrupts. But we basically threw away the
information that multiple errors occurred, and we queued an e_src
record that occupies space without being used for anything.

Yes, you are correct. One other way to minimize this race window is to
clear the Root status register as soon as possible. Maybe we can move
it before source ID read as below.

--- a/drivers/pci/pcie/aer.c
+++ b/drivers/pci/pcie/aer.c
@@ -1204,8 +1204,8 @@ static irqreturn_t aer_irq(int irq, void *context)
if (!(e_src.status & AER_ERR_STATUS_MASK))
return IRQ_NONE;

- pci_read_config_dword(rp, aer + PCI_ERR_ROOT_ERR_SRC, &e_src.id);
pci_write_config_dword(rp, aer + PCI_ERR_ROOT_STATUS, e_src.status);
+ pci_read_config_dword(rp, aer + PCI_ERR_ROOT_ERR_SRC, &e_src.id)

--
Sathyanarayanan Kuppuswamy
Linux Kernel Developer