This repository was archived by the owner on Dec 20, 2019. It is now read-only.
[arm] fix CFA when frame pointer is used #48
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Consider the following prologue:
```
0xce2d34: 0xe92d40f0 push {r4, r5, r6, r7, lr}
0xce2d38: 0xe28d700c add r7, sp, mono#12
0xce2d3c: 0xe92d0c00 push {r10, r11}
```
Without this patch the following unwind information would be emitted:
```
CFA: [0] def_cfa: arm_sp+0x0
CFA: [4] def_cfa: arm_v4+0x8
CFA: [4] offset: pc at cfa-0x4
CFA: [4] offset: arm_v4 at cfa-0x8
CFA: [4] offset: arm_v3 at cfa-0xc
CFA: [4] offset: arm_v2 at cfa-0x10
CFA: [4] offset: arm_v1 at cfa-0x14
CFA: [8] offset: arm_fp at cfa-0x18
CFA: [c] offset: arm_v7 at cfa-0x1c
CFA: [c] def_cfa_offset: 0x14
```
which is _almost_ correct. The last bit messes up our unwinding machine and also doesn't make sense in general. If _anything_ this should be `def_cfa: arm_sp+0x1c`.
Anyway, since the framepointer is always `+8` to the old stack pointer, we can use that to restore the CFA and can omit whatever DefCFAOffsetCandidates would compute.
lewurm
added a commit
to lewurm/mono
that referenced
this pull request
Jul 8, 2019
Instead of
```
push {r4, r5, r6, r7, r10, r11, lr}
```
LLVM will generate
```
push {r4, r5, r6, r7, lr}
add r7, sp, mono#12
push {r10, r11}
```
Seems like this mono/llvm@a04e9e4 assumes that `-disable-fp-elim` is passed for iOS targets. I wondered why this wouldn't break things for them, but when you run `clang` from Xcode with `-v`, you will discover that its driver passes `-mdisable-fp-elim`.
Together with mono/llvm#48 this fixes crashes with FullAOT+LLVM on iOS 32bit.
Fixes mono#15058
Contributes to mono#9621
vargaz
approved these changes
Jul 8, 2019
monojenkins
pushed a commit
to monojenkins/mono
that referenced
this pull request
Jul 8, 2019
Instead of
```
push {r4, r5, r6, r7, r10, r11, lr}
```
LLVM will generate
```
push {r4, r5, r6, r7, lr}
add r7, sp, mono#12
push {r10, r11}
```
Seems like this mono/llvm@a04e9e4 assumes that `-disable-fp-elim` is passed for iOS targets. I wondered why this wouldn't break things for them, but when you run `clang` from Xcode with `-v`, you will discover that its driver passes `-mdisable-fp-elim`.
Together with mono/llvm#48 this fixes crashes with FullAOT+LLVM on iOS 32bit.
Fixes mono#15058
Contributes to mono#9621
monojenkins
added a commit
to mono/mono
that referenced
this pull request
Jul 9, 2019
[2019-06] [llvm] avoid FP elimination on iOS/armv7
Instead of
```
push {r4, r5, r6, r7, r10, r11, lr}
```
LLVM will generate
```
push {r4, r5, r6, r7, lr}
add r7, sp, #12
push {r9, r10}
```
Seems like this mono/llvm@a04e9e4 assumes that `-disable-fp-elim` is passed for iOS targets. I wondered why this wouldn't break things for them, but when you run `clang` from Xcode with `-v`, you will discover that its driver passes `-mdisable-fp-elim`.
Together with mono/llvm#48 this fixes crashes with FullAOT+LLVM on iOS 32bit.
Fixes #15058
Contributes to #9621
Backport of #15617.
/cc @lewurm
monojenkins
pushed a commit
to mono/mono
that referenced
this pull request
Jul 14, 2019
[llvm] avoid FP elimination on iOS/armv7
Instead of
```
push {r4, r5, r6, r7, r10, r11, lr}
```
LLVM will generate
```
push {r4, r5, r6, r7, lr}
add r7, sp, #12
push {r9, r10}
```
Seems like this mono/llvm@a04e9e4 assumes that `-disable-fp-elim` is passed for iOS targets. I wondered why this wouldn't break things for them, but when you run `clang` from Xcode with `-v`, you will discover that its driver passes `-mdisable-fp-elim`.
Together with mono/llvm#48 this fixes crashes with FullAOT+LLVM on iOS 32bit.
Fixes #15058
Contributes to #9621
vargaz
added a commit
that referenced
this pull request
Oct 1, 2019
[arm] fix CFA when frame pointer is used
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Consider the following prologue:
Without this patch the following unwind information would be emitted:
which is almost correct. The last bit messes up our unwinding machine and also doesn't make sense in general. If anything this should be
def_cfa: arm_sp+0x1c.Anyway, since the framepointer is always
+8to the old stack pointer, we can use that to restore the CFA and can omit whatever DefCFAOffsetCandidates would compute.