Fix fibrechannel_linux for ppc64le by hamzy · Pull Request #3769 · prometheus/node_exporter

hamzy · 2026-08-01T15:25:24Z

Seen on a master node:

panic: runtime error: invalid memory address or nil pointer dereference
[signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x106cefa4]
goroutine 81 [running]:
github.com/prometheus/node_exporter/collector.(*fibrechannelCollector).Update(0x7e28c1ae20f0, 0x7e28c1826770)
  /go/src/github.com/prometheus/node_exporter/collector/fibrechannel_linux.go:133 +0x604

sh-5.1# ls -l /host/sys/class/fc_host/
total 0
lrwxrwxrwx. 1 root root 0 Jul 30 20:42 host1 -> ../../devices/vio/30000003/host1/fc_host/host1
lrwxrwxrwx. 1 root root 0 Jul 30 20:42 host2 -> ../../devices/vio/30000005/host2/fc_host/host2
sh-5.1# ls -l /host/sys/class/fc_host/host1/
total 0
-rw-r--r--. 1 root root 65536 Jul 30 20:09 dev_loss_tmo
lrwxrwxrwx. 1 root root     0 Jul 30 20:42 device -> ../../../host1
-r--r--r--. 1 root root 65536 Jul 30 20:09 fabric_name
--w-------. 1 root root 65536 Jul 30 20:42 issue_lip
-r--r--r--. 1 root root 65536 Jul 30 20:42 maxframe_size
-r--r--r--. 1 root root 65536 Jul 30 20:09 node_name
-r--r--r--. 1 root root 65536 Jul 30 20:09 port_id
-r--r--r--. 1 root root 65536 Jul 30 20:42 port_name
-r--r--r--. 1 root root 65536 Jul 30 20:09 port_state
-r--r--r--. 1 root root 65536 Jul 30 20:09 port_type
drwxr-xr-x. 2 root root     0 Jul 30 20:42 power
-r--r--r--. 1 root root 65536 Jul 30 20:09 speed
lrwxrwxrwx. 1 root root     0 Jul 30 20:42 subsystem -> ../../../../../../class/fc_host
-r--r--r--. 1 root root 65536 Jul 30 20:09 supported_classes
-rw-r--r--. 1 root root 65536 Jul 30 20:42 tgtid_bind_type
-rw-r--r--. 1 root root 65536 Jul 30 20:42 uevent

Since many filenames for counters are not guaranteed to exist in VMs, allocate if not present.

prb112 · 2026-08-02T23:38:25Z

-		// Note: `procfs` does not guarantee a safe dereference for these counters.
-		//       A disabled host returns no statistics counters.
-		if host.PortState == nil || *host.PortState == "Unknown" {
+		// Note: These are not guaranteed to exist in the filesystem


This defense addition makes sense to me.

prb112 · 2026-08-02T23:40:29Z

Related to a bug found at https://redhat.atlassian.net/browse/OCPBUGS-100376

simonpasquier · 2026-08-05T08:53:28Z

 # TYPE node_fibrechannel_info gauge
 node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host1",port_id="",port_name="",port_state="Online",port_type="",speed="8 Gbit",supported_classes="",supported_speeds="",symbolic_name=""} 1
 node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host0",port_id="000002",port_name="1000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1
+node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host2",port_id="000003",port_name="4000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1


this info metric isn't correct (most of the label values are empty)

Isn't that the whole point of this PR?

The expected output is

node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host1",port_id="",port_name="",port_state="Online",port_type="",speed="8 Gbit",supported_classes="",supported_speeds="",symbolic_name=""} 1 node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host2",port_id="",port_name="",port_state="Online",port_type="",speed="",supported_classes="",supported_speeds="",symbolic_name=""} 1 node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host0",port_id="000002",port_name="1000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1

instead of

node_fibrechannel_info{dev_loss_tmo="",fabric_name="",fc_host="host1",port_id="",port_name="",port_state="Online",port_type="",speed="8 Gbit",supported_classes="",supported_speeds="",symbolic_name=""} 1 node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host0",port_id="000002",port_name="1000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1 node_fibrechannel_info{dev_loss_tmo="30",fabric_name="0",fc_host="host2",port_id="000003",port_name="4000e0071bce95f2",port_state="Online",port_type="Point-To-Point (direct nport connection)",speed="16 Gbit",supported_classes="Class 3",supported_speeds="4 Gbit, 8 Gbit, 16 Gbit",symbolic_name="Emulex SN1100E2P FV12.4.270.3 DV12.4.0.0. HN:gotest. OS:Linux"} 1

Notice how the host2 line is between host1 and host0 and how most of its labels are empty.

I still don't understand what the problem is. The testcase reproduces the problem without the fix and everything passes with the fix.

Can you just provide a patch for whatever it is you are unhappy with?

The CI isn't happy, I'll try to get you a patch.

I appreciate it, thanks!

Seen on a master node: ``` panic: runtime error: invalid memory address or nil pointer dereference [signal SIGSEGV: segmentation violation code=0x1 addr=0x0 pc=0x106cefa4] goroutine 81 [running]: github.com/prometheus/node_exporter/collector.(*fibrechannelCollector).Update(0x7e28c1ae20f0, 0x7e28c1826770) /go/src/github.com/prometheus/node_exporter/collector/fibrechannel_linux.go:133 +0x604 sh-5.1# ls -l /host/sys/class/fc_host/ total 0 lrwxrwxrwx. 1 root root 0 Jul 30 20:42 host1 -> ../../devices/vio/30000003/host1/fc_host/host1 lrwxrwxrwx. 1 root root 0 Jul 30 20:42 host2 -> ../../devices/vio/30000005/host2/fc_host/host2 sh-5.1# ls -l /host/sys/class/fc_host/host1/ total 0 -rw-r--r--. 1 root root 65536 Jul 30 20:09 dev_loss_tmo lrwxrwxrwx. 1 root root 0 Jul 30 20:42 device -> ../../../host1 -r--r--r--. 1 root root 65536 Jul 30 20:09 fabric_name --w-------. 1 root root 65536 Jul 30 20:42 issue_lip -r--r--r--. 1 root root 65536 Jul 30 20:42 maxframe_size -r--r--r--. 1 root root 65536 Jul 30 20:09 node_name -r--r--r--. 1 root root 65536 Jul 30 20:09 port_id -r--r--r--. 1 root root 65536 Jul 30 20:42 port_name -r--r--r--. 1 root root 65536 Jul 30 20:09 port_state -r--r--r--. 1 root root 65536 Jul 30 20:09 port_type drwxr-xr-x. 2 root root 0 Jul 30 20:42 power -r--r--r--. 1 root root 65536 Jul 30 20:09 speed lrwxrwxrwx. 1 root root 0 Jul 30 20:42 subsystem -> ../../../../../../class/fc_host -r--r--r--. 1 root root 65536 Jul 30 20:09 supported_classes -rw-r--r--. 1 root root 65536 Jul 30 20:42 tgtid_bind_type -rw-r--r--. 1 root root 65536 Jul 30 20:42 uevent ``` Since many filenames for counters are not guaranteed to exist in VMs, allocate if not present. Signed-off-by: Mark Hamzy <hamzy@us.ibm.com>

prb112

/lgtm

not on project directly, just reviewing for colleague @hamzy

hamzy · 2026-08-10T15:07:16Z

@simonpasquier What is the next step in the process?

prb112 reviewed Aug 2, 2026

View reviewed changes

hamzy force-pushed the fix-collector-fibrechannel_linux branch 3 times, most recently from e4442c4 to 597d541 Compare August 4, 2026 00:23

simonpasquier reviewed Aug 4, 2026

View reviewed changes

Comment thread collector/fibrechannel_linux_test.go

hamzy force-pushed the fix-collector-fibrechannel_linux branch from 597d541 to ae10239 Compare August 4, 2026 12:17

simonpasquier reviewed Aug 4, 2026

View reviewed changes

Comment thread collector/fixtures/e2e-64k-page-output.txt

hamzy force-pushed the fix-collector-fibrechannel_linux branch 2 times, most recently from 32870c9 to d00fb6f Compare August 4, 2026 15:47

simonpasquier reviewed Aug 5, 2026

View reviewed changes

hamzy force-pushed the fix-collector-fibrechannel_linux branch 2 times, most recently from f6b055d to d135cd5 Compare August 5, 2026 22:03

hamzy force-pushed the fix-collector-fibrechannel_linux branch from d135cd5 to 6a190da Compare August 5, 2026 23:14

prb112 approved these changes Aug 6, 2026

View reviewed changes

simonpasquier approved these changes Aug 6, 2026

View reviewed changes

simonpasquier requested a review from jan--f August 6, 2026 12:39

Navigation Menu

Uh oh!

Fix fibrechannel_linux for ppc64le - #3769

Fix fibrechannel_linux for ppc64le#3769
hamzy wants to merge 1 commit into
prometheus:masterfrom
hamzy:fix-collector-fibrechannel_linux

hamzy commented Aug 1, 2026

Uh oh!

prb112 Aug 2, 2026 •

edited

Loading

Uh oh!

prb112 commented Aug 2, 2026

Uh oh!

Uh oh!

Uh oh!

simonpasquier Aug 5, 2026

Uh oh!

hamzy Aug 5, 2026

Uh oh!

simonpasquier Aug 5, 2026

Uh oh!

hamzy Aug 5, 2026

Uh oh!

simonpasquier Aug 5, 2026

Uh oh!

hamzy Aug 5, 2026

Uh oh!

prb112 left a comment

Uh oh!

hamzy commented Aug 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Conversation

hamzy commented Aug 1, 2026

Uh oh!

prb112 Aug 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

prb112 commented Aug 2, 2026

Uh oh!

Uh oh!

Uh oh!

simonpasquier Aug 5, 2026

Choose a reason for hiding this comment

Uh oh!

hamzy Aug 5, 2026

Choose a reason for hiding this comment

Uh oh!

simonpasquier Aug 5, 2026

Choose a reason for hiding this comment

Uh oh!

hamzy Aug 5, 2026

Choose a reason for hiding this comment

Uh oh!

simonpasquier Aug 5, 2026

Choose a reason for hiding this comment

Uh oh!

hamzy Aug 5, 2026

Choose a reason for hiding this comment

Uh oh!

prb112 left a comment

Choose a reason for hiding this comment

Uh oh!

hamzy commented Aug 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

prb112 Aug 2, 2026 •

edited

Loading