[LU-12262] Improve sbi_flags checking

Related — Sun, 10 Oct 2021 13:43:04 +0000

In ~~LU-11930~~, Oleg described a problem with printing sbi_flags:
'

Nowadays if we do get_param on lustre-....sbi_flags we get back:

error: get_param: reading 'llite.lustre-ffff88009a0ca800.sbi_flags': Invalid argument

and then in the kernel

[  788.311079] LustreError: 10351:0:(lproc_llite.c:967:ll_sbi_flags_seq_show()) lustre: Revise array LL_SBI_FLAGS to match sbi flags please.

The result is of course every test that checks feature flags is now not finding anything it's looking for and skips. in particular al layoutswaps (test 184*) and a bunch of other stuff is now in perma-skip mode.

This was broken in ~~LU-11825~~ https://review.whamcloud.com/33912 it appears by removing a flag from LL_SBI_FLAGS which is something that should NEVER be done.

Also I think we ned to convert that error message into an assertion.

Problems like here we'd catch sooner and if some client has some flags set incorrectly we are probably ok panicking there as well as there's clearly some unexpected memory corruption or whatnot going on?"

This ticket is to capture his suggestions at the end, to improve checking so we can't accidentally remove a flag like this in the future.

Whamcloud Community JIRA

[LU-12262] Improve sbi_flags checking