[asan] Remove check for stack size

[Hahnfeld]

Summary

This has been introduced in rL304598 and fails for increased stack sizes.

[alekseyshl]

What's the stack size on CentOS? What was the point of the original check?

[vitalybuka]

Matt also hit this on Unbuntu with gcc as a host compiler.

[alekseyshl]

Well, would be nice to know the size of the stack there too. I'm curious what was the reason to check that stack size is less than 256MB. Default max_uar_stack_size_log is 20, so it's 1MB of stack size and I don't see where else stack_size is used.

[Hahnfeld]

In D34876#796951, @alekseyshl wrote:

What's the stack size on CentOS? What was the point of the original check?

The error is as follows:

"((stack_size)) <= ((0x10000000))" (0x40000000, 0x10000000)

Please note that I have ulimit -s unlimited because I often had problems with Fortran applications that required large stack sizes. With ulimit -s 4096, the problem also goes away.
However, I don't understand this check either, so I'd say to remove it as it causes problems.

[alekseyshl]

In D34876#797665, @Hahnfeld wrote:

In D34876#796951, @alekseyshl wrote:

What's the stack size on CentOS? What was the point of the original check?

The error is as follows:

"((stack_size)) <= ((0x10000000))" (0x40000000, 0x10000000)

Please note that I have ulimit -s unlimited because I often had problems with Fortran applications that required large stack sizes. With ulimit -s 4096, the problem also goes away.
However, I don't understand this check either, so I'd say to remove it as it causes problems.

I'm ok with removing this check, we can always bring it back (with proper comment this time).

Kostya, what exactly this check was for?

[kcc]

This check is a sanity check, please keep it, but feel free to change the constant to e.g. 0x100000000 (16x increase)

[Hahnfeld]

In D34876#806869, @kcc wrote:

This check is a sanity check, please keep it, but feel free to change the constant to e.g. 0x100000000 (16x increase)

Sanity check in what sense? If we can increase it arbitrarily, it's not a check anymore IMHO

[kcc]

In D34876#807658, @Hahnfeld wrote:

In D34876#806869, @kcc wrote:

This check is a sanity check, please keep it, but feel free to change the constant to e.g. 0x100000000 (16x increase)

Sanity check in what sense? If we can increase it arbitrarily, it's not a check anymore IMHO

It's a sanity check that asan has computed the stack size correctly, and that it doesn't think the stack is e.g. 42 Gb

compiler-rt/trunk/lib/asan/asan_thread.cc

 }
 
 // We want to create the FakeStack lazyly on the first use, but not eralier
 // than the stack size is known and the procedure has to be async-signal safe.
 FakeStack *AsanThread::AsyncSignalSafeLazyInitFakeStack() {
   uptr stack_size = this->stack_size();
   if (stack_size == 0)  // stack_size is not yet available, don't use FakeStack.
     return nullptr;
-  CHECK_LE(stack_size, 0x10000000);
   uptr old_val = 0;
   // fake_stack_ has 3 states:
   // 0   -- not initialized
   // 1   -- being initialized
   // ptr -- initialized
   // This CAS checks if the state was 0 and if so changes it to state 1,
   // if that was successful, it initializes the pointer.
   if (atomic_compare_exchange_strong(