LEROY Christophe [off-list ref] writes:

Michael Ellerman [off-list ref] a =C3=A9crit=C2=A0:

quoted

LEROY Christophe [off-list ref] writes:

quoted

Benjamin Herrenschmidt [off-list ref] a =C3=A9crit=C2=A0:

quoted

When hitting below a VM_GROWSDOWN vma (typically growing the stack),
we check whether it's a valid stack-growing instruction and we
check the distance to GPR1. This is largely open coded with lots
of comments, so move it out to a helper.

Did you have a look at the following patch ? It's been waiting for
application for some weeks now.
https://patchwork.ozlabs.org/patch/771869

I actually merged it last merge window, but found I had no good way to
test it, so I took it out again until I can write a test case for it.

The way I realised it wasn't being tested was by removing all the
store_updates_sp logic entirely and having my system run happily for
several days :}

Which demonstrates how unlikely this is, hence doing that get_user()=20=20
at every fault is waste of time.

Yes I agree.

How do you plan to handle that in parralele to ben's serie ?

Not sure :)

I'll be back from vacation next week and may help finding a way to=20=20
test that. (A test program using alloca() ?)

I was thinking hand-crafted asm, but that might be a pain to get working
for 32 & 64-bit, in which case alloca() might work.

cheers