Hotplug events are critical indicators for analyzing hardware health,
particularly in AI supercomputers where surprise link downs can
significantly impact system performance and reliability. The failure
characterization analysis illustrates the significance of failures
caused by the Infiniband link errors. Meta observes that 2% in a machine
learning cluster and 6% in a vision application cluster of Infiniband
failures co-occur with GPU failures, such as falling off the bus, which
may indicate a correlation with PCIe.[1]
PATCH 1/2: Add a generic RAS tracepoint for hotplug event to help healthy check.
PATCH 2/2: Generate tracepoints for PCIe hotplug event
The output like below:
$ echo 1 > /sys/kernel/debug/tracing/events/hotplug/pci_hp_event/enable
$ cat /sys/kernel/debug/tracing/trace_pipe
<...>-206 [001] ..... 40.373870: pci_hp_event: 0000:00:02.0 slot:10, trans_state:Link Down
<...>-206 [001] ..... 40.374871: pci_hp_event: 0000:00:02.0 slot:10, trans_state:Card not present
[1]https://arxiv.org/abs/2410.21680
Shuai Xue (2):
PCI: hotplug: Add a generic RAS tracepoint for hotplug event
pci: pciehp: Generate tracepoints for hotplug event
drivers/pci/hotplug/pciehp_ctrl.c | 33 ++++++++++++---
drivers/pci/hotplug/trace.h | 68 +++++++++++++++++++++++++++++++
include/uapi/linux/pci.h | 7 ++++
3 files changed, 102 insertions(+), 6 deletions(-)
create mode 100644 drivers/pci/hotplug/trace.h
--
2.39.3
