@hans and @dmajor thanks for reporting this issue (https://bugs.llvm.org/show_bug.cgi?id=47905).
Do the latest changes resolve this in Chromium and sqlite3?

Correctly handle __imp_ symbols in .giats and .gfids sections

Thanks @aeubanks! I don't yet have commit access and my colleague is on leave.

Ping @mehdi_amini - is this what you were asking for the CFGuard pass?

Always initialize loadThunkSym to nullptr.

Thanks for the diagnostics @aeubanks! Does this now resolve the issue with the Chrome build?

Rewrite markSymbolsForRVATable() in terms of getSymbolsFromSections()

[CFGuard] Split into two passes and link into opt

In D65761#2102525, @mehdi_amini wrote:

It seems like this pass was added in lib/Transforms but does not have any unit-tests (lit tests) provided. It isn't even linked into opt. As it is an LLVM IR pass it should be tested as such I believe. Can you provide tests for this?

In D75047#1890976, @hans wrote:

Do you have commits access, or would you like me to commit for you?

In D65761#1811097, @hubert.reinterpretcast wrote:

In D65761#1804302, @hubert.reinterpretcast wrote:

I have confirmed that the case I mentioned fails with rGd157a9b.

@ajpaverd, is a fix forthcoming for the issue I mentioned with this patch?

Thanks for the feedback @aaron.ballman and @dmajor!

• Address comments

• Make guard_nocf an attribute of individual call instructions

In D72167#1803333, @dmajor wrote:

Thanks for doing this!

When I applied this patch and added __declspec(guard(nocf)) to the function that was giving us trouble, I still crashed with a CFG failure in the caller, since the nocf function was inlined. When I additionally marked that function as noinline, then the CFG checks were omitted, as desired. Is this behavior with respect to inlining expected?

You may want to mention "PR44096" in the commit message.

In D72167#1803395, @aaron.ballman wrote:

Can you please add some Sema tests that verify the attribute only appertains to functions, accepts the right number of arguments, etc? Also, some tests showing how this attribute works for things like redelcarations and virtual functions would help.

Make guard_nocf an attribute of individual call instructions

In D65761#1772076, @dmajor wrote:

Well, this patch specifically has code (even a test) that nocf_check prevents CFG, so it looks like the intent was there, it's just that there isn't a way to get that attribute onto functions from cpp in a CFG/non-CET world.

Thanks for finding and fixing this @rnk! I've double-checked that Windows doesn't use AL for varargs (https://docs.microsoft.com/en-us/cpp/build/x64-calling-convention?view=vs-2019#varargs), but having a general mechanism for forwarding that doesn't assume a particular calling convention sounds like a good approach. This looks good from the control flow guard perspective.

Split cfguard_setjmp test into target-specific tests and add missing cfguard_longjmp pass for ARM and AArch64.

Formatting fixes for CFGuard patch.

Final changes and tests suggested by @rnk.

In D65761#1660121, @rnk wrote:

Here is some feedback, I apologize for dragging my feet.

Also, I would really like to get feedback from @pcc. He's OOO currently, though.

Improved Control Flow Guard implementation.

In D65761#1646059, @rnk wrote:

I plan to take a look at this tomorrow, sorry for putting this off for a while. :(