On Fri, Jul 10, 2020 at 10:48 AM Naresh Kamboju
<naresh.kamboju@xxxxxxxxxx> wrote:
>
> I have applied your patch and test started in a loop for a million times
> but the test ran for 35 times. Seems like the test got a timeout after 1 hour.
That just means that my test-case was wrong (in the sense that what it
was testing wasn't what was triggering things).
It might be wrong because I got the stack usage calculations wrong,
but it might be wrong simply because the bug you've triggered with LTP
needs something else to trigger.
> Re-running test with long timeouts 4 hours and will share findings.
That test was more intended to trigger the issue reliably, if it was
just a "stack is special, needs that exact 2MB aligned case to be
problematic".
So re-running my "t.c" the test for long times with that kernel patch
that removes the stack randomization isn't going to matter. Either it
triggers the case, or not.
I don't actually see any case where we play with vm_start the way we
used to (it was basically removed with commit 1be7107fbe18: "mm:
larger stack guard gap, between vmas"). So it was kind of a log shot
anyway.
Linus
