Change printf format string to avoid typecast warning

Fix style per Vitaly's comments

In D137666#3960839, @paulkirth wrote:

Hi, I think this change is currently breaking Fuchsia's Clang CI. https://luci-milo.appspot.com/ui/p/fuchsia/builders/toolchain.ci/clang-linux-arm64/b8796124255499289009/overview

Can you take a look, and if it will be hard to fix, revert until it can be addressed?

In D137666#3960839, @paulkirth wrote:

Hi, I think this change is currently breaking Fuchsia's Clang CI. https://luci-milo.appspot.com/ui/p/fuchsia/builders/toolchain.ci/clang-linux-arm64/b8796124255499289009/overview

Can you take a look, and if it will be hard to fix, revert until it can be addressed?

@vitalybuka I pushed this simple change without review, to (hopefully) fix the build. Please provide post-commit review.

Whitespace

In D137666#3920077, @eugenis wrote:

If we are changing the mapping + the ABI, we should be 100% confident the new one covers all ASLR possibilities. Why not just run some binary a lot of times and collect the range of addresses, or even inspect the kernel source for possible executable locations? Also, make it a large binary.

Larger mappings

In D137666#3919867, @eugenis wrote:

Ah, so is the problem with ASLR randomizing the initial executable mappings over a region larger than 64Gb? We do not care about app allocating memory, heap placement is defined by msan. Could you confirm that the new mapping covers all possible locations? Even with ex. a huge executable binary.

That's ok I guess, if we do not care about /39 and /42 on aarch64 linux. I'm also curious how much dynamic shadow mapping would cost (cpu and code size). It's not as straightforward as with ASan, because MSan can not allocate shadow in a single contiguous block, most likely. But it can save us a lot of pain down the road.

In D137666#3919581, @eugenis wrote:

Sorry, I don't follow. What is limited to 64Gb (the sum of all app regions? why does it matter?) and why can't an "invalid" region be mapped?

Remove unnnecessary app mapping

Removed comment about coverage of 39- and 42-bit mappings, since we have deprecated them.
(We still leave in the general comment about 39- and 42-bit segments, for reference.)

Minor housekeeping: same as the sanitizer_common changes for lsan (GetTls() on 32-bit Android; and sig_setmask)

On 32-bit Android, GetTls() won't use __get_tls() (which is undefined)
Defines sig_setmask

Add back the sanitizer_common/sanitizer_linux_libcdep.cc TLS changes from diff 3 - I accidentally omitted them.
All other files are the same as per diff 4.

Addresses previous comments

Matches the changes in D23369, to use TLS_SLOT_TSAN only on Android (use the original variables on non-Android platforms).
Also adds named constants offsetof__msan_param_tls etc.

For non-Android, this keeps the ABI the same (uses the original variables).
For Android, this uses TLS_SLOT_TSAN.

This emits the TLS access code inline:

• using Intrinsic::thread_pointer and TLS_SLOT_TSAN if on Android
• using the regular GlobalVariable::InitialExecTLSModel if on non-Android

to match the changes in D23369.

In D22549#490474, @eugenis wrote:

This is way too intrusive.
How about something like this:

Move the contents of MsanTLSSlotStruct directly into MsanThread so that they can be accessed as, for example, getCurrentThread()->in_interceptor_scope. This might slow down other platforms a little bit, but that's acceptable. Again, no #ifdefs - the same code on all platforms. Change __msan::GetCurrentThread to use TLS_SLOT_TSAN on Android and a usual thread-local var on other platforms.

This addresses all the previous comments
(use_tls is not overridden for Android, GetTls() updated instead to return the bionic TLS slots;
GetCurrentThread() changed to GetCurrentThread()->id, getCurrentTLS() changed to GetCurrentThread();
SIG_SETMASK copied from signal.h;
removed some unnecessary code and moved some code around)

There's now an LsanThread struct containing all the thread-local variables. A pointer to the struct is returned from lsan_thread.cc::getCurrentTLS(), which is the single point where we distinguish between Android (use AndroidGetTls()) and non-Android (use THREADLOCAL).

Reuploaded with context

Reuploaded with context

#define SANITIZER_ANDROID is just a serving suggestion