[infiniband] collect info from all active ports of all IB devices

[pmoravec]

currently just the very first active port is checked

Resolves: #1262

Signed-off-by: Pavel Moravec [email protected]

---

Please place an 'X' inside each '[]' to confirm you adhere to our Contributor Guidelines

• Is the commit message split over multiple lines and hard-wrapped at 72 characters?
• Is the subject and message clear and concise?
• Does the subject start with [plugin_name] if submitting a plugin patch or a [section_name] if part of the core sosreport code?
• Does the commit contain a Signed-off-by: First Lastname [email protected]?

[pmoravec]

Thanks Hongang Li for initial version of the patch.

[bmr-cymru]

I'm not really in favour of this approach if the root cause is a deficiency in the tools that we are calling.

If this is confusing to users of sos, how can it not be confusing for users who are running these tools themselves?

[pmoravec]

I dont know if/how confusing the behaviour is, or if it is worth fallback behaviour when not providing options. See e.g. https://linux.die.net/man/8/infiniband-diags and "Multiple port/Multiple CA support:" paragraph.

It makes sense to collect that info from all active ports - it is up to discussion if sos or infiniband tools shall provide the "get data from all ports" feature.

[bmr-cymru]

s is not closed in the event of an exception in s.readline().

[bmr-cymru]

p is not closed in the event of an exception in p.readline().

[bmr-cymru]

It just seems like a bad design choice in these tools: pretty much all system level tools of this kind that I can think of defaults to printing information for all devices when no option is given:

lsblk, lsscsi, ifconfig, ip, lvs, etc. etc. ...

What I really dislike here is the need to poke around /sys in order to discover the list of parameters to pass to the commands. The structure of sys does occasionally change and we don't have great infrastructure to deal with that at the moment.

If there's no choice but to work around this in sos then we will but it seems sub-optimal for both users and support that there is not a single, simple command that just lists all the available ports.

[bmr-cymru]

(but the exception handling in the current proposed patch does need to be fixed).

[bmr-cymru]

Let me take a look at this tomorrow: I don't think we'll see changes in the upstream IB tools given how longstanding this behaviour is (although I still think that it is a defect: it's crazy to not give users a simple way to list the devices and ports on their machine).