1 .. _segmented_stacks: 2 3 ======================== 4 Segmented Stacks in LLVM 5 ======================== 6 7 .. contents:: 8 :local: 9 10 Introduction 11 ============ 12 13 Segmented stack allows stack space to be allocated incrementally than as a 14 monolithic chunk (of some worst case size) at thread initialization. This is 15 done by allocating stack blocks (henceforth called *stacklets*) and linking them 16 into a doubly linked list. The function prologue is responsible for checking if 17 the current stacklet has enough space for the function to execute; and if not, 18 call into the libgcc runtime to allocate more stack space. When using ``llc``, 19 segmented stacks can be enabled by adding ``-segmented-stacks`` to the command 20 line. 21 22 The runtime functionality is `already there in libgcc 23 <http://gcc.gnu.org/wiki/SplitStacks>`_. 24 25 Implementation Details 26 ====================== 27 28 .. _allocating stacklets: 29 30 Allocating Stacklets 31 -------------------- 32 33 As mentioned above, the function prologue checks if the current stacklet has 34 enough space. The current approach is to use a slot in the TCB to store the 35 current stack limit (minus the amount of space needed to allocate a new block) - 36 this slot's offset is again dictated by ``libgcc``. The generated 37 assembly looks like this on x86-64: 38 39 .. code-block:: nasm 40 41 leaq -8(%rsp), %r10 42 cmpq %fs:112, %r10 43 jg .LBB0_2 44 45 # More stack space needs to be allocated 46 movabsq $8, %r10 # The amount of space needed 47 movabsq $0, %r11 # The total size of arguments passed on stack 48 callq __morestack 49 ret # The reason for this extra return is explained below 50 .LBB0_2: 51 # Usual prologue continues here 52 53 The size of function arguments on the stack needs to be passed to 54 ``__morestack`` (this function is implemented in ``libgcc``) since that number 55 of bytes has to be copied from the previous stacklet to the current one. This is 56 so that SP (and FP) relative addressing of function arguments work as expected. 57 58 The unusual ``ret`` is needed to have the function which made a call to 59 ``__morestack`` return correctly. ``__morestack``, instead of returning, calls 60 into ``.LBB0_2``. This is possible since both, the size of the ``ret`` 61 instruction and the PC of call to ``__morestack`` are known. When the function 62 body returns, control is transferred back to ``__morestack``. ``__morestack`` 63 then de-allocates the new stacklet, restores the correct SP value, and does a 64 second return, which returns control to the correct caller. 65 66 Variable Sized Allocas 67 ---------------------- 68 69 The section on `allocating stacklets`_ automatically assumes that every stack 70 frame will be of fixed size. However, LLVM allows the use of the ``llvm.alloca`` 71 intrinsic to allocate dynamically sized blocks of memory on the stack. When 72 faced with such a variable-sized alloca, code is generated to: 73 74 * Check if the current stacklet has enough space. If yes, just bump the SP, like 75 in the normal case. 76 * If not, generate a call to ``libgcc``, which allocates the memory from the 77 heap. 78 79 The memory allocated from the heap is linked into a list in the current 80 stacklet, and freed along with the same. This prevents a memory leak. 81
