Re: [LTP] [PATCH] tst_kconfig: Don't consider CONFIG_FAULT_INJECTION slow

[Petr Vorel <pvorel@suse.cz>]

Hi all,

> Hi!
> > I didn't use fault injection extensively, but I read that when verbose mode is
> > enabled we might have a big slowdown, so I wouldn't skip this check.

> > Eventually, we can implement a check that verifies if fault injection is enabled
> > with a high verbosity. That would probably make more sense.

> In this case the fault injection looks like something that you have to
> configure before use so I would assume that anyone who configure this
> can also adjust the LTP multipliers.

> Maybe we can just print TINFO message that the config is on and may
> slown donw the execution instead?

What do you mean by "slow down the execution"?

Anyway, I'd really like to get this solved, not only due 4x longer runtime by
default on Tumbleweed (in case of sigkill test really hangs for 2 mins)

tst_kconfig.c:753: TINFO: CONFIG_FAULT_INJECTION kernel option detected which might slow the execution
tst_test.c:1870: TINFO: Overall timeout per run is 0h 02m 00s

but on other machines due slow worker with setup LTP_TIMEOUT_MUL=2 we get even
4 min (8x longer wasted time)
tst_kconfig.c:753: TINFO: CONFIG_FAULT_INJECTION kernel option detected which might slow the execution
tst_test.c:1870: TINFO: Overall timeout per run is 0h 04m 00s

Nobody complained that he's actively using CONFIG_FAULT_INJECTION and needed
this setup, but we are really hit by this. That's why I took a simple approach
to ignore CONFIG_FAULT_INJECTION.

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp