2020Q4
Date of Issue: 21st December 2020
This document describes the use of the DWARF debug table format in the Application Binary Interface (ABI) for the Arm architecture.
DWARF, DWARF 3.0, use of DWARF format
Please check Application Binary Interface for the Arm® Architecture for the latest release of this document.
Please report defects in this specification to the issue tracker page on GitHub.
This work is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by-sa/4.0/ or send a letter to Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.
Grant of Patent License. Subject to the terms and conditions of this license (both the Public License and this Patent License), each Licensor hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer the Licensed Material, where such license applies only to those patent claims licensable by such Licensor that are necessarily infringed by their contribution(s) alone or by combination of their contribution(s) with the Licensed Material to which such contribution(s) was submitted. If You institute patent litigation against any entity (including a cross-claim or counterclaim in a lawsuit) alleging that the Licensed Material or a contribution incorporated within the Licensed Material constitutes direct or contributory patent infringement, then any licenses granted to You under this license for that Licensed Material shall terminate as of the date such litigation is filed.
As identified more fully in the Licence section, this project is licensed under CC-BY-SA-4.0 along with an additional patent license. The language in the additional patent license is largely identical to that in Apache-2.0 (specifically, Section 3 of Apache-2.0 as reflected at https://www.apache.org/licenses/LICENSE-2.0) with two exceptions.
First, several changes were made related to the defined terms so as to reflect the fact that such defined terms need to align with the terminology in CC-BY-SA-4.0 rather than Apache-2.0 (e.g., changing “Work” to “Licensed Material”).
Second, the defensive termination clause was changed such that the scope of defensive termination applies to “any licenses granted to You” (rather than “any patent licenses granted to You”). This change is intended to help maintain a healthy ecosystem by providing additional protection to the community against patent litigation claims.
Contributions to this project are licensed under an inbound=outbound model such that any such contributions are licensed by the contributor under the same terms as those in the Licence section.
The text of and illustrations in this document are licensed by Arm under a Creative Commons Attribution–Share Alike 4.0 International license ("CC-BY-SA-4.0”), with an additional clause on patents. The Arm trademarks featured here are registered trademarks or trademarks of Arm Limited (or its subsidiaries) in the US and/or elsewhere. All rights reserved. Please visit https://www.arm.com/company/policies/trademarks for more information about Arm’s trademarks.
Copyright (c) 2003-2007, 2012, 2018, 2020, Arm Limited and its affiliates. All rights reserved.
The following support level definitions are used by the Arm ABI specifications:
- Release
- Arm considers this specification to have enough implementations, which have received sufficient testing, to verify that it is correct. The details of these criteria are dependent on the scale and complexity of the change over previous versions: small, simple changes might only require one implementation, but more complex changes require multiple independent implementations, which have been rigorously tested for cross-compatibility. Arm anticipates that future changes to this specification will be limited to typographical corrections, clarifications and compatible extensions.
- Beta
- Arm considers this specification to be complete, but existing implementations do not meet the requirements for confidence in its release quality. Arm may need to make incompatible changes if issues emerge from its implementation.
- Alpha
- The content of this specification is a draft, and Arm considers the likelihood of future incompatible changes to be significant.
All content in this document is at the Release quality level.
Issue | Date | Change |
---|---|---|
1.0 | 30th October 2003 | First public release. |
2.0 | 24th March 2005 | Second public release. |
2.01 | 6th October 2006 | Added register numbers for VFP-v3 d0-d31 (DWARF register names). |
2.02 | 5th May 2006 | Minor corrections now that DWARF 3.0 is a standard; incompatible changes to the values of DW_AT_endianity (Describing other endian data) as a result. |
A | 25th October 2007 | Document renumbered (formerly GENC-003533 v2.02). |
B r2.09 | 30th November 2012 | Common information entries: Clarify CIE descriptions of registers that are unused by intention of the user, for example as a consequence of the chosen procedure call standard. |
2018Q4 | 21st December 2018 | Minor typographical fixes, updated links. |
2020Q4 | 21st December 2020 |
|
This document refers to, or is referred to by, the following documents.
Ref | External reference or URL | Title |
---|---|---|
AADWARF | Source for this document | DWARF for the Arm Architecture |
BSABI32 | ABI for the Arm Architecture (Base Standard) | |
GDWARF | http://dwarfstd.org/Dwarf3Std.php | DWARF 3.0, the generic debug table format. |
The ABI for the Arm Architecture uses the following terms and abbreviations.
- AAPCS
- Procedure Call Standard for the Arm Architecture.
- ABI
Application Binary Interface:
- The specifications to which an executable must conform in order to execute in a specific execution environment. For example, the Linux ABI for the Arm Architecture.
- A particular aspect of the specifications to which independently produced relocatable files must conform in order to be statically linkable and executable. For example, the AAELF32, RTABI32, ...
- AEABI
- (Embedded) ABI for the Arm architecture (this ABI...)
- Arm-based
- ... based on the Arm architecture ...
- core registers
- The general purpose registers visible in the Arm architecture’s programmer’s model, typically r0-r12, SP, LR, PC, and CPSR.
- EABI
- An ABI suited to the needs of embedded, and deeply embedded (sometimes called free standing), applications.
- Q-o-I
- Quality of Implementation – a quality, behavior, functionality, or mechanism not required by this standard, but which might be provided by systems conforming to it. Q-o-I is often used to describe the tool-chain-specific means by which a standard requirement is met.
- VFP
- The Arm architecture’s Floating Point architecture and instruction set. In this ABI, this abbreviation includes all floating point variants regardless of whether or not vector (V) mode is supported.
This specification has been developed with the active support of the following organizations. In alphabetical order: Arm, CodeSourcery, Intel, Metrowerks, Montavista, Nexus Electronics, PalmSource, Symbian, Texas Instruments, and Wind River.
The ABI for the Arm architecture specifies the use of DWARF 3.0-format debugging data. For details of the base standard see GDWARF.
The ABI for the Arm architecture gives additional rules for how DWARF 3.0 should be used, and how it is extended in ways specific to the Arm architecture. The following topics are covered in detail.
- The enumeration of DWARF register-numbers for, use in .debug_frame sections (DWARF register names).
- How the machine state (Arm state versus Thumb state) is encoded in DWARF 3.0 line number tables (DWARF line number information (ISA field)).
- How to describe access to Arm architecture v6 other-endian data (Describing other endian data).
- The definition of Canonical Frame Address (CFA) used by this ABI (Canonical Frame Address).
- The generation and interpretation of debug frame Common Information Entries (Common information entries).
To support stack unwinding by debuggers, producers must always generate .debug_frame sections, even when:
- Not generating other debug tables.
- At high optimization levels.
- Assembling hand-written assembly language, if that code calls code compiled from C or C++.
Ideally, a user of a C/C++ source language debugger would like the illusion of:
- Stepping through the source program sequence point (SP) by sequence point.
- Being able to inspect the program’s state at any sequence point, and seeing there the state predicted by the source language semantics.
For the purpose of debugging illusion, we define an observation point (OP) to be a point at which a debugger may (meaningfully) inspect a program’s state. Most sequence points are also observation points. In addition
- There is an OP just after each function call (at the pc value to which the call will return).
- There is no OP at the SP after evaluation of function arguments but before the function call.
A variable’s scope extends from the point of declaration of the identifier to the end of the smallest enclosing block. A variable need not have a value everywhere in its scope – it may be initialized some way after its declaration.
When a user signals to a producer (by Q-o-I means) that it should favour quality of debugging over quality of generated code, the producer should strive (Q-o-I) to generate DWARF tables and code supporting this illusion. Specifically:
- A statement should describe the code between consecutive OPs.
- At each OP, every in-scope, initialized, source code variable should have a location (need not be in memory), and that location should hold the value predicted by the source language semantics.
It is not necessary to describe OPs in code the producer knows can never be
executed (e.g. in if(0){i++;}
).
GDWARF §2.6.1, Register Name Operators, suggests that the mapping from a DWARF register name to a target register number should be defined by the ABI for the target architecture. DWARF register names are encoded as unsigned LEB128 integers. Numbers 0-127 encode in 1 byte, 128-16383 in 2 bytes.
DWARF register number | Arm core or co-processor registers | Description |
---|---|---|
0–15 | R0–R15 | Arm core integer registers |
16–63 | None | Obsolescent: 16–47 were previously used for both FPA and VFP registers (Note 1) |
64–95 | S0–S31 | Legacy VFP-v2 use: D0–D15 alias S0, S2, … S30 (Note 1, Note 4) |
96–103 | F0–F7 | Obsolescent: FPA registers 0-7 (Note 1) |
104–111 | wCGR0–wCGR7 | Intel wireless MMX general purpose registers 0-7 |
ACC0-ACC7 | XScale accumulator register 0-7 (Note 2) | |
112–127 | wR0–wR15 | Intel wireless MMX data registers 0–15 |
128 | SPSR | Current SPSR register |
129 | SPSR_FIQ | FIQ-mode SPSR |
130 | SPSR_IRQ | IRQ-mode SPSR |
131 | SPSR_ABT | ABT-mode SPSR |
132 | SPSR_UND | UND-mode SPSR |
133 | SPSR_SVC | SVC-mode SPSR |
134–143 | None | Reserved for future allocation |
144–150 | R8_USR–R14_USR | User mode registers |
151–157 | R8_FIQ–R14_FIQ | Banked FIQ registers |
158–159 | R13_IRQ–R14_IRQ | Banked IRQ registers |
160–161 | R13_ABT–R14_ABT | Banked ABT registers |
162–163 | R13_UND–R14_UND | Banked UND registers |
164–165 | R13_SVC–R14_SVC | Banked SVC registers |
166–191 | None | Reserved for future allocation |
192–199 | wC0–wC7 | Intel wireless MMX control register in co-processor 0–7 |
200 | TPIDRURO | PL0 Read-Only Software Thread IR register. |
201–255 | None | Reserved for future allocation |
256-287 | VFP-v3/Neon D0-D31 | VFP-v3/Neon 64-bit register file (Note 4) |
288-319 | None | Reserved to VFP/Neon |
320 | TPIDRURO | PL0 Read-Only Software Thread ID register |
321 | TPIDRURW | PL0 Read/Write Software Thread ID register |
322 | TPIDPR | PL1 Software Thread ID register |
323 | HTPIDPR | Hyp Software Thread ID register |
324-8191 | None | Reserved for future allocation |
8192–16383 | Vendor co-processor | Unspecified vendor co-processor register (Note 3) |
Note
- In ADS toolkits, DWARF names 16–23 were used to represent FPA registers F0–F7 and 16-47 were used to represent VFP registers S0–S31. No application needs to use both numberings simultaneously but it can complicate decoding, so in RVDS new, non-overlapping, numbers 64-95 were allocated to VFP S0-S31. Debuggers that need to support legacy objects may need to handle both mappings.
- Current implementations of the version 1 XScale Architecture specification implement only acc0, though eight such registers (acc0–acc7) are defined architecturally in co-processor 0. The version 2 specification defines the Wireless MMX co-processor in Arm co-processor slots 0 and 1. No system can contain both acc0 and MMX, so these numberings can overlap.
- The vendor co-processor space is not specified by this ABI and should be used when there is unlikely to be a requirement for multiple vendors to support debugging such code. By using numbers in this space vendors can be sure that they will not conflict with future ABI allocations. If a set of co-processor registers is likely to be used directly from a high-level language and to require support of multiple toolkit vendors, then an application should be made to Arm for an allocation of a numbering in the reserved space.
- The VFP-v3 and Neon architectures extend the register file to 32 64-bit registers, posing significant difficulties to extending the ABI v2.0 VFP encodings. There is no simple scheme using 1-byte register numbers that is compatible with the legacies. We have, therefore, introduced a new, simple, more precisely specified scheme using 2-byte register numbers. The new numbering scheme should also be used for VFP-v2.
The CPSR, VFP and FPA control registers are not allocated a numbering above. It is considered unlikely that these will be needed for producing a stack back-trace in a debugger.
Architecturally, VFP-v3 and the Neon SIMD unit share a register file comprising 32 64-bit registers, D0-D31. Registers D0-D31 are described by DWARF register numbers 256-287. Register numbers 288-319 are reserved in case of future register file expansion.
DWARF registers 64-95 are obsolescent (and will become obsolete in the next major revision of the ABI for the Arm Architecture).
In DWARF terms:
Register Dx is described as DW_OP_regx(256+x).
Q registers Q0-Q15 are described by composing two D registers together.
Qx = DW_OP_regx(256+2x) DW_OP_piece(8) DW_OP_regx(256+2x+1) [DW_OP_piece(8)]
(Note that the final DW_OP_piece(8) can be omitted because the whole register is used. It is left in above for expositional clarity).
S registers are described as bit-pieces of a register
S[2x] = DW_OP_regx(256 + (x >> 1)) DW_OP_bit_piece(32, 0)
S[2x+1] = DW_OP_regx(256 + (x >> 1)) DW_OP_bit_piece(32, 32)
Neon Half-word lanes and byte lanes are described in a similar way to S registers.
Producers should use this new numbering scheme for VFP-v2 before the ABI-v2.0 scheme (S0-S31 → 64-95) is declared obsolete. Consumers should accept both numberings for as long as there are legacy binaries.
GDWARF §6.2.5.2 Standard Opcodes, item 12, DW_LNS_set_isa, describes a single unsigned LEB128 operand that denotes the instruction set architecture (ISA) at the location identified by the line number table entry. The value of the operand is determined by the ABI for the architecture (this specification).
Under the Arm architecture there are many instruction set versions and variants, but few instruction set states. Under this ABI, the ISA field corresponding to a particular program address denotes the instruction set state encoded by the CPSR when the pc contains that address.
Name | Value | Meaning |
---|---|---|
DW_ISA_UNKNOWN | 0 | I-set state not available or not recorded. |
DW_ISA_ARM_thumb | 1 | T-bit will be set in the CPSR when pc contains this code address. |
DW_ISA_ARM_arm | 2 | T-bit will be clear in the CPSR when pc contains this code address. |
Other | Reserved to the ABI for the Arm architecture. |
Arm architecture version 6 allows programs to access data stored in the other byte order, either by executing REV* instructions, or by juggling the E bit in the PSR. Consequently, there is a need to describe in DWARF tables data that has been statically declared with a particular byte order.
This ABI mandates no particular way to describe the byte order of data manipulated by a programming language, but one could imagine a simple language extension like the following, or use of #pragma:
extern __big_endian T bx; // bx contains big-endian data
extern __little_endian T lx; // lx contains little-endian data
Usually, all data has the same byte order and this is recorded in the EI_DATA field of the header of the ELF file (as the value ELFDATA2MSB or ELFDATA2LSB).
To describe data that is explicitly declared big-endian or little-endian (by
whatever means), use the DWARF 3.0 attribute DW_AT_endianity
(0x65). It takes a
single LEB128 constant argument value that is one of the following:
DW_END_default (= 0)
DW_END_big (= 1) (Was 0 prior to the DWARF 3.0 standard)
DW_END_little (= 2) (Was 1 prior to the DWARF 3.0 standard)
By default the Arm architecture is little endian, so DW_END_default
should be
interpreted as DW_END_little
.
The DW_AT_endianity
attributes can be attached to type entries as follows.
Attached to a base type (GDWARF, §5.1, Base Type Entries), this attribute gives the byte order of the data described by the base type.
If this order differs from the default byte order recorded in the containing ELF file, a debugger should reverse the order of the bytes it fetches or stores when accessing values of that base type.
Attached to any other type (GDWARF, §5, Type Entries), this attribute indicates that the type was labeled explicitly (in some way) with the given byte order.
When representing such a type across its user interface, a debugger should label the representation in some way that indicates it was declared with an explicit byte order. Some possible labels for big-endian follow.
__big_endian T X; __declspec(big_endian) T X; T X __attribute__("big endian"); #pragma arm_big_endian struct BigT { ... }; #pragma no_arm_big_endian BigT X;
Any such representation by a debugger is entirely quality of implementation.
The term Canonical Frame Address (CFA) is defined in GDWARF, §6.4, Call Frame Information.
This ABI adopts the typical definition of CFA given there.
- The CFA is the value of the stack pointer (r13) at the call site in the previous frame.
The DWARF virtual unwinding model is based, conceptually, on a tabular structure with one column for each target register (GDWARF, §6.4.1, Structure of Call Frame Information). A .debug_frame Common Information Entry (CIE) specifies the initial values (on entry to an associated function) of each register.
The variability of processors conforming to the Arm architecture creates a problem for this model. A producer cannot reliably enumerate all the registers in the target. For example, an integer-only function might be included in one executable file for targets with VFP and another for targets without. In effect, it must be acceptable for a producer not to initialize, in a CIE, registers it does not know about. In turn this generates an obligation on consuming debuggers to default missing initial values.
This generates the following obligations on producers and consumers of CIEs.
Consumers must default the CIE initial value of any target register not mentioned explicitly in the CIE.
Callee-saved registers (and registers intentionally unused by the program, for example as a consequence of the procedure call standard) should be initialized as if by DW_CFA_same_value, other registers as if by DW_CFA_undefined.
A debugger can use built-in knowledge of the procedure call standard or can deduce which registers are callee-saved by scanning all CIEs.
To allow consumers to reliably default the initial values of missing entries by scanning a program’s CIEs, without recourse to built-in knowledge, producers must identify registers not preserved by callees, as follows.
If a function uses any register from a particular hardware register class (e.g. Arm core registers), its associated CIE must initialize all the registers of that class that are not callee-saved to DW_CFA_undefined.
(As an optimization, a producer need not initialize registers it can prove cannot be used by any associated functions and their descendants. Although these are not callee-saved, they are not callee-used either).
If a function uses a callee-saved register R, its associated CIE must initialize R using one of the defined value methods (not DW_CFA_undefined).