mark `start_task` with ".cfi_undefined rip" #23074

vtjnash · 2017-08-01T17:45:01Z

Annotate start_task with asm(".cfi_undefined rip");, just like scylladb/seastar@1fa1323

The text was updated successfully, but these errors were encountered:

maleadt · 2017-08-02T07:26:09Z

So maybe that's why #14727 still had some unannotated frames.

shivakrishna9 · 2017-08-02T08:40:02Z

@vtjnash I will take this up. Need more info on where to add annotations.

yuyichao · 2017-08-02T12:42:45Z

We set the RIP to 0 so it shouldn't point to garbage at least...

vtjnash · 2017-08-03T20:20:13Z

I think it's just the start_task function that needs annotation, and testing that backtrace()[end] != C_NULL (only seems to be an issue on Linux).

Closes #23074 Fixes #35155

Closes JuliaLang#23074 Fixes JuliaLang#35155

Fiber call-chains end at `coro_init`, but unwinders don't stop there, trying to use `coro_init` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_init`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for gcc on x86_64 — also provide it for clang and AARCH64. Needed for tarantool#4002 NO_DOC=bug fix NO_CHANGELOG=bug fix

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for gcc on x86_64 — also provide it for clang. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. Needed for tarantool#4002 NO_DOC=bug fix NO_CHANGELOG=bug fix

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/ sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb =ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_6 4/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e 67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27b f8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbde d0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/ libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba 7dd3857aba8 Needed for tarantool#4002 NO_DOC=bug fix NO_CHANGELOG=bug fix

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/ sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb =ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_6 4/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e 67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27b f8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbde d0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/ libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba 7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. Zero out rbp on x86_64 (also marks end of call-chain for unwinders): this requires removing "-fno-omit-frame-pointer" compiler flag, which is anyway redundant. Also remove redundant "-fno-stack-protector" compiler flag. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/ sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb =ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_6 4/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e 67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27b f8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbde d0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/ libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba 7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. Zero out rbp on x86_64 (this also marks end of call-chain for unwinders): this requires setting "-fomit-frame-pointer" compile flag for compiling coro.c. Remove redundant "-fno-omit-frame-pointer" and "-fno-stack-protector" compile flags. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/ sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb =ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_6 4/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e 67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27b f8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbde d0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/ libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba 7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. Zero out rbp on x86_64 (this also marks end of call-chain for unwinders): this requires setting "-fomit-frame-pointer" compile flag for compiling coro.c. Remove redundant "-fno-omit-frame-pointer" and "-fno-stack-protector" compile flags. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb=ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_64/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27bf8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbded0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/ sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb =ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_6 4/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e 67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27b f8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbde d0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/ libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba 7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it for clang. Zero out rbp on x86_64 (to conform to x86_64 ABI): this requires setting "-fomit-frame-pointer" compile flag for coro.c. Remove redundant "-fno-omit-frame-pointer" and "-fno-stack-protector" compile flags for other Tarantool sources. For some reason unwinders ignore platform ABIs regarding ending of call-chains: instead of trying to follow platform ABIs, explicitly invalidate the topmost (`coro_{init, startup}`) stack frame information for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb=ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_64/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27bf8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbded0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it if ENABLE_BACKTRACE is set during configuration. Zero out rbp on x86_64 (to conform to x86_64 ABI): this requires setting "-fomit-frame-pointer" compile flag for coro.c. Backtrace collection from inactive fiber based on pseudo context-switch relied on the stack frame structure: remove redundant "-fno-omit-frame-pointer" and "-fno-stack-protector" compile flags for other Tarantool sources. For some reason unwinders ignore platform ABIs regarding ending of call-chains: explicitly invalidate the topmost (`coro_{init, startup}`) current frame information (CFI) for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb=ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_64/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27bf8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbded0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba7dd3857aba8 Needed for tarantool#4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

Fiber call-chains end at `coro_{init, startup}`, but unwinders don't stop there, trying to use `coro_{init, startup}` stack frame's return address (which points to some garbage) and, in turn, failing. A similar issue was experienced by seastar and julia (see JuliaLang/julia#23074 and scylladb/scylladb#1909). In order to make unwinding stop at `coro_{init, startup}`'s stack frame we need to annotate it with CFI assembly: previously, annotation was provided only for GCC on x86_64 — also provide it if ENABLE_BACKTRACE is set during configuration. Zero out rbp on x86_64 (to conform to x86_64 ABI): this requires setting "-fomit-frame-pointer" compile flag for coro.c. Backtrace collection from inactive fiber based on pseudo context-switch relied on the stack frame structure: remove redundant "-fno-omit-frame-pointer" and "-fno-stack-protector" compile flags for other Tarantool sources. For some reason unwinders ignore platform ABIs regarding ending of call-chains: explicitly invalidate the topmost (`coro_{init, startup}`) current frame information (CFI) for both x86_64 and AARCH64. References: 1. glibc: * clone: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/x86_64/clone.S;h=31ac12da0cc08a934d514fed1de9eba1cb3e8ec5;hb=ebbb8c9f64c3486603ef4ccee4dd2a5574e41039 * start: https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86_64/start.S;h=9edd17b60cd54ec9eef11c76ab02322dcb5d057a;hb=5b736bc9b55115e67129e77db4de6cf193054cd2 2. seastar: * thread_context::main(): https://github.com/scylladb/seastar/blob/d27bf8b5a14e5b9e9c9df18fd1306489b651aa42/src/core/thread.cc#L278-L293 3. julia: * https://github.com/JuliaLang/julia/blob/2e2b1d2ad50fe12999cbded0b5acd3f0a36ec8c5/src/julia_internal.h#L90-L106 4. android: * https://cs.android.com/android/platform/superproject/+/master:bionic/libc/platform/bionic/macros.h;l=52-60;drc=2528dab7419a63f57fe20027886ba7dd3857aba8 Needed for #4002 NO_DOC=internal bug fix NO_CHANGELOG=internal bug fix NO_TEST=unwind information annotation in inline assembly

vtjnash added the help wanted Indicates that a maintainer wants help on an issue or pull request label Aug 1, 2017

vtjnash added a commit that referenced this issue Jan 26, 2021

fix linux unwind info, per x86_64 ABI

3d24848

Closes #23074 Fixes #35155

vtjnash closed this as completed in 9638ba0 Jan 27, 2021

ElOceanografo pushed a commit to ElOceanografo/julia that referenced this issue May 4, 2021

fix linux unwind info, per x86_64 ABI

b1d86aa

Closes JuliaLang#23074 Fixes JuliaLang#35155

antoine-levitt pushed a commit to antoine-levitt/julia that referenced this issue May 9, 2021

fix linux unwind info, per x86_64 ABI

e88cb57

Closes JuliaLang#23074 Fixes JuliaLang#35155

CuriousGeorgiy mentioned this issue Mar 7, 2022

libcoro fibers unwind information fixes tarantool/tarantool#6889

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

mark `start_task` with ".cfi_undefined rip" #23074

mark `start_task` with ".cfi_undefined rip" #23074

vtjnash commented Aug 1, 2017

maleadt commented Aug 2, 2017

shivakrishna9 commented Aug 2, 2017

yuyichao commented Aug 2, 2017

vtjnash commented Aug 3, 2017

mark start_task with ".cfi_undefined rip" #23074

mark start_task with ".cfi_undefined rip" #23074

Comments

vtjnash commented Aug 1, 2017

maleadt commented Aug 2, 2017

shivakrishna9 commented Aug 2, 2017

yuyichao commented Aug 2, 2017

vtjnash commented Aug 3, 2017

mark `start_task` with ".cfi_undefined rip" #23074

mark `start_task` with ".cfi_undefined rip" #23074