[PATCH] D18526: [tsan] Disable randomized address space on aarch64 linux.

[Adhemerval Zanella via llvm-commits]

zatrazz added a comment.

In http://reviews.llvm.org/D18526#390617, @rengolin wrote:

> I'm confused. This patch seems eerily similar to the one that broke the buildbots... Maybe you forgot to add ifdef Android when you added vReport?
>
> My worry is simple: this patch broke not *all* AArch64 buildbots, but "some". Meaning, the idea may work on some but not on other configurations. This is very worrying.


It was breaking the buildbot because the warning message was being handled as
the expected TSAN output instead of just an warning. The logic is similar as when
running with unlimited stack (ulimit -s unlimited): tsan will warning and reexec, but
the output won't be printed as

> Furthermore, it's not because we're not testing on Android (via buildbots) today, that we won't do that tomorrow. So, if this fix is *guaranteed* to work on *every* possible Android configuration on the planet, present and future, than I'm ok with the change.

> 

> But I'm very sceptical. The previous change was *guaranteed* to work with any Linux kernel, so we accepted for all AArch64, but that was a false assumption. So forgive me if I don't trust this new assumption as well.

> 

> Now Android will be building with LLVM by default, so it's not an experimental tool any more. We used to accept "experimental" patches before because not many people were fiddling with sanitizers on Android, but this will not be the case *very* soon. In fact, we're already planning our Android buildbots as we speak, so the more false assumptions we put in now (because it happens to work on a particular developer's device), the harder it will be to get those bots green.

> 

> Concretely, what I'm asking is proof that this is true for all Android configurations. Virtual address randomization is a crucial security feature, and I'd be surprised if Google never uses it as we move into 64-bit land, especially as Android takes on more roles in automotive, home automation and small laptops.


I can't say for every kernel out there, but for the ones I have access (which is unfortunately
only old ones without the new patch that disrupt it), the testsuite works. I would expect that
yabinc tests on Android on 39-VMA with and without the kernel patch that triggered this
modification.


http://reviews.llvm.org/D18526