[DTrace-devel] [PATCH] test: fix a stupid typo

[Nick Alcock]

On 30 May 2023, Eugene Loh via DTrace-devel uttered the following:

> On 5/30/23 11:21, Nick Alcock via DTrace-devel wrote:
>
>> Serves me right for doing this at the last minute.
>
> That comment is not needed for the patch.

It was the absolute minimum my guilt at such a stupid error would allow.

(However... I've now discovered that on more recent platforms the
debugfs is mounted in *both* /sys/kernel/tracing *and*
/sys/kernel/debug/tracing, so my guilt is reduced: clearly I didn't
entirely forget to test it.)

> By the way, the typo in the path also leads to test failures on some platforms.

Well, one test failure, of this test.

>> It will break other programs using USDT probes that aren't currently
>> being probed on the same machine, so it's very temporary: but without
>> either this or proper uprobes cleanup code, running tst.manyprobes.sh a
>> few times will denial-of-service-attack the machine.
>
> Not only the path typo should be fixed but also this heavy-handed
> cleanup.  The D script can include something like 'BEGIN {
> printf("target pid is %d\n", \$target); }'.  The script can then go
> through and search uprobes_events for /proc/$target/map_files/ events
> and remove those probes.  If you like, I can prepare that patch.  Or,
> feel free.  Anyhow, if we litter the system with 2000 uprobes, we
> should clean them up without blowing everything out of the water.

Oh please do do it! That sounds much better than this.

It's all temporary until dtprobed can do it for us anyway (which is
coming, I very much hope this week).

-- 
NULL && (void)