cranelift_codegen/isa/
unwind.rs

1//! Represents information relating to function unwinding.
2
3use crate::machinst::RealReg;
4
5#[cfg(feature = "enable-serde")]
6use serde_derive::{Deserialize, Serialize};
7
8#[cfg(feature = "unwind")]
9pub mod systemv;
10
11#[cfg(feature = "unwind")]
12pub mod winx64;
13
14/// CFA-based unwind information used on SystemV.
15pub type CfaUnwindInfo = systemv::UnwindInfo;
16
17/// Expected unwind info type.
18#[derive(Debug, Clone, Copy, PartialEq, Eq)]
19#[non_exhaustive]
20pub enum UnwindInfoKind {
21    /// No unwind info.
22    None,
23    /// SystemV CIE/FDE unwind info.
24    #[cfg(feature = "unwind")]
25    SystemV,
26    /// Windows X64 Unwind info
27    #[cfg(feature = "unwind")]
28    Windows,
29}
30
31/// Represents unwind information for a single function.
32#[derive(Clone, Debug, PartialEq, Eq)]
33#[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
34#[non_exhaustive]
35pub enum UnwindInfo {
36    /// Windows x64 ABI unwind information.
37    #[cfg(feature = "unwind")]
38    WindowsX64(winx64::UnwindInfo),
39    /// System V ABI unwind information.
40    #[cfg(feature = "unwind")]
41    SystemV(CfaUnwindInfo),
42}
43
44/// Unwind pseudoinstruction used in VCode backends: represents that
45/// at the present location, an action has just been taken.
46///
47/// VCode backends always emit unwind info that is relative to a frame
48/// pointer, because we are planning to allow for dynamic frame allocation,
49/// and because it makes the design quite a lot simpler in general: we don't
50/// have to be precise about SP adjustments throughout the body of the function.
51///
52/// We include only unwind info for prologues at this time. Note that unwind
53/// info for epilogues is only necessary if one expects to unwind while within
54/// the last few instructions of the function (after FP has been restored) or
55/// if one wishes to instruction-step through the epilogue and see a backtrace
56/// at every point. This is not necessary for correct operation otherwise and so
57/// we simplify the world a bit by omitting epilogue information. (Note that
58/// some platforms also don't require or have a way to describe unwind
59/// information for epilogues at all: for example, on Windows, the `UNWIND_INFO`
60/// format only stores information for the function prologue.)
61///
62/// Because we are defining an abstraction over multiple unwind formats (at
63/// least Windows/fastcall and System V) and multiple architectures (at least
64/// x86-64 and aarch64), we have to be a little bit flexible in how we describe
65/// the frame. However, it turns out that a least-common-denominator prologue
66/// works for all of the cases we have to worry about today!
67///
68/// We assume the stack looks something like this:
69///
70///
71/// ```plain
72///                  +----------------------------------------------+
73///                  | stack arg area, etc (according to ABI)       |
74///                  | ...                                          |
75///   SP at call --> +----------------------------------------------+
76///                  | return address (pushed by HW or SW)          |
77///                  +----------------------------------------------+
78///                  | old frame pointer (FP)                       |
79///   FP in this --> +----------------------------------------------+
80///   function       | clobbered callee-save registers              |
81///                  | ...                                          |
82///   start of   --> +----------------------------------------------+
83///   clobbers       | (rest of function's frame, irrelevant here)  |
84///                  | ...                                          |
85///   SP in this --> +----------------------------------------------+
86///   function
87/// ```
88///
89/// We assume that the prologue consists of:
90///
91/// * `PushFrameRegs`: A push operation that adds the old FP to the stack (and
92///    maybe the link register, on architectures that do not push return addresses
93///    in hardware)
94/// * `DefineFrame`: An update that sets FP to SP to establish a new frame
95/// * `SaveReg`: A number of stores or pushes to the stack to save clobbered registers
96///
97/// Each of these steps has a corresponding pseudo-instruction. At each step,
98/// we need some information to determine where the current stack frame is
99/// relative to SP or FP. When the `PushFrameRegs` occurs, we need to know how
100/// much SP was decremented by, so we can allow the unwinder to continue to find
101/// the caller's frame. When we define the new frame, we need to know where FP
102/// is in relation to "SP at call" and also "start of clobbers", because
103/// different unwind formats define one or the other of those as the anchor by
104/// which we define the frame. Finally, when registers are saved, we need to
105/// know which ones, and where.
106///
107/// Different unwind formats work differently; here is a whirlwind tour of how
108/// they define frames to help understanding:
109///
110/// - Windows unwind information defines a frame that must start below the
111///   clobber area, because all clobber-save offsets are non-negative. We set it
112///   at the "start of clobbers" in the figure above. The `UNWIND_INFO` contains
113///   a "frame pointer offset" field; when we define the new frame, the frame is
114///   understood to be the value of FP (`RBP`) *minus* this offset. In other
115///   words, the FP is *at the frame pointer offset* relative to the
116///   start-of-clobber-frame. We use the "FP offset down to clobber area" offset
117///   to generate this info.
118///
119/// - System V unwind information defines a frame in terms of the CFA
120///   (call-frame address), which is equal to the "SP at call" above. SysV
121///   allows negative offsets, so there is no issue defining clobber-save
122///   locations in terms of CFA. The format allows us to define CFA flexibly in
123///   terms of any register plus an offset; we define it in terms of FP plus
124///   the clobber-to-caller-SP offset once FP is established.
125///
126/// Note that certain architectures impose limits on offsets: for example, on
127/// Windows, the base of the clobber area must not be more than 240 bytes below
128/// FP.
129///
130/// Unwind pseudoinstructions are emitted inline by ABI code as it generates
131/// a prologue. Thus, for the usual case, a prologue might look like (using x64
132/// as an example):
133///
134/// ```plain
135/// push rbp
136/// unwind UnwindInst::PushFrameRegs { offset_upward_to_caller_sp: 16 }
137/// mov rbp, rsp
138/// unwind UnwindInst::DefineNewFrame { offset_upward_to_caller_sp: 16,
139///                                     offset_downward_to_clobbers: 16 }
140/// sub rsp, 32
141/// mov [rsp+16], r12
142/// unwind UnwindInst::SaveReg { reg: R12, clobber_offset: 0 }
143/// mov [rsp+24], r13
144/// unwind UnwindInst::SaveReg { reg: R13, clobber_offset: 8 }
145/// ...
146/// ```
147#[derive(Clone, Debug, PartialEq, Eq)]
148#[cfg_attr(feature = "enable-serde", derive(Serialize, Deserialize))]
149pub enum UnwindInst {
150    /// The frame-pointer register for this architecture has just been pushed to
151    /// the stack (and on architectures where return-addresses are not pushed by
152    /// hardware, the link register as well). The FP has not been set to this
153    /// frame yet. The current location of SP is such that
154    /// `offset_upward_to_caller_sp` is the distance to SP-at-callsite (our
155    /// caller's frame).
156    PushFrameRegs {
157        /// The offset from the current SP (after push) to the SP at
158        /// caller's callsite.
159        offset_upward_to_caller_sp: u32,
160    },
161    /// The frame-pointer register for this architecture has just been
162    /// set to the current stack location. We wish to define a new
163    /// frame that is anchored on this new FP value. Offsets are provided
164    /// upward to the caller's stack frame and downward toward the clobber
165    /// area. We expect this pseudo-op to come after `PushFrameRegs`.
166    DefineNewFrame {
167        /// The offset from the current SP and FP value upward to the value of
168        /// SP at the callsite that invoked us.
169        offset_upward_to_caller_sp: u32,
170        /// The offset from the current SP and FP value downward to the start of
171        /// the clobber area.
172        offset_downward_to_clobbers: u32,
173    },
174    /// The stack pointer was adjusted to allocate the stack.
175    StackAlloc {
176        /// Size to allocate.
177        size: u32,
178    },
179    /// The stack slot at the given offset from the clobber-area base has been
180    /// used to save the given register.
181    ///
182    /// Given that `CreateFrame` has occurred first with some
183    /// `offset_downward_to_clobbers`, `SaveReg` with `clobber_offset` indicates
184    /// that the value of `reg` is saved on the stack at address `FP -
185    /// offset_downward_to_clobbers + clobber_offset`.
186    SaveReg {
187        /// The offset from the start of the clobber area to this register's
188        /// stack location.
189        clobber_offset: u32,
190        /// The saved register.
191        reg: RealReg,
192    },
193    /// Defines if the aarch64-specific pointer authentication available for ARM v8.3+ devices
194    /// is enabled for certain pointers or not.
195    Aarch64SetPointerAuth {
196        /// Whether return addresses (hold in LR) contain a pointer-authentication code.
197        return_addresses: bool,
198    },
199}