[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added Status|ASSIGNED|RESOLVED Resolution|--- |FIXED --- Comment #16 from Martin Liška --- Fixed on all active branches.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #15 from CVS Commits --- The releases/gcc-8 branch has been updated by Martin Liska : https://gcc.gnu.org/g:5d746191e271949e530d9e5f46cde7e7bf08272f commit r8-10307-g5d746191e271949e530d9e5f46cde7e7bf08272f Author: Martin Liska Date: Tue May 19 16:57:56 2020 +0200 Add missing store in emission of asan_stack_free. gcc/ChangeLog: 2020-05-19 Martin Liska PR sanitizer/94910 * asan.c (asan_emit_stack_protection): Emit also **SavedFlagPtr(FakeStack, class_id) = 0 in order to release a stack frame. (cherry picked from commit 8b6731e674c76cb48a417f2eef74ced92a17f469)
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #14 from CVS Commits --- The releases/gcc-9 branch has been updated by Martin Liska : https://gcc.gnu.org/g:877d8d63228579bd56f94e6c56fbfeb015da08e5 commit r9-8671-g877d8d63228579bd56f94e6c56fbfeb015da08e5 Author: Martin Liska Date: Tue May 19 16:57:56 2020 +0200 Add missing store in emission of asan_stack_free. gcc/ChangeLog: 2020-05-19 Martin Liska PR sanitizer/94910 * asan.c (asan_emit_stack_protection): Emit also **SavedFlagPtr(FakeStack, class_id) = 0 in order to release a stack frame. (cherry picked from commit 8b6731e674c76cb48a417f2eef74ced92a17f469)
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #13 from CVS Commits --- The releases/gcc-10 branch has been updated by Martin Liska : https://gcc.gnu.org/g:036b288ca4cf5d3b1d908ef97c25b7f92153ff8a commit r10-8283-g036b288ca4cf5d3b1d908ef97c25b7f92153ff8a Author: Martin Liska Date: Tue May 19 16:57:56 2020 +0200 Add missing store in emission of asan_stack_free. gcc/ChangeLog: 2020-05-19 Martin Liska PR sanitizer/94910 * asan.c (asan_emit_stack_protection): Emit also **SavedFlagPtr(FakeStack, class_id) = 0 in order to release a stack frame. (cherry picked from commit 8b6731e674c76cb48a417f2eef74ced92a17f469)
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added Known to work||11.0 Known to fail||10.1.0, 8.4.0, 9.3.0 --- Comment #12 from Martin Liška --- Fixed on master so far.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #11 from CVS Commits --- The master branch has been updated by Martin Liska : https://gcc.gnu.org/g:8b6731e674c76cb48a417f2eef74ced92a17f469 commit r11-1145-g8b6731e674c76cb48a417f2eef74ced92a17f469 Author: Martin Liska Date: Tue May 19 16:57:56 2020 +0200 Add missing store in emission of asan_stack_free. gcc/ChangeLog: 2020-05-19 Martin Liska PR sanitizer/94910 * asan.c (asan_emit_stack_protection): Emit also **SavedFlagPtr(FakeStack, class_id) = 0 in order to release a stack frame.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added CC||frantisek at sumsal dot cz --- Comment #10 from Martin Liška --- *** Bug 95275 has been marked as a duplicate of this bug. ***
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #9 from Martin Liška --- (In reply to Rafael Avila de Espindola from comment #8) > I can confirm that the proposed patch fixes the issue for me. > > Thank you so much! I thank you for the bug report. We had the regression since the very beginning!
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #8 from Rafael Avila de Espindola --- I can confirm that the proposed patch fixes the issue for me. Thank you so much!
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added Status|NEW |ASSIGNED Assignee|unassigned at gcc dot gnu.org |marxin at gcc dot gnu.org --- Comment #7 from Martin Liška --- I was able to reproduce that on tramp3d as well. I bet the problem is that the inline expansion does not clear: **SavedFlagPtr(FakeStack) = 0 as LLVM does: https://github.com/llvm/llvm-project/blob/master/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp#L3279-L3292 Let me take a look.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #6 from Jakub Jelinek --- (In reply to Martin Liška from comment #5) > > FakeStack (stack_size_log=20):bail out after 8192 > > FakeStack (stack_size_log=20):bail out after 16384 > > FakeStack (stack_size_log=20):bail out after 4096 > > FakeStack (stack_size_log=20):bail out after 8192 > > ... > > This comes after the 27000 success Fake stack allocations. I guess the important question is what is different with clang that it behaves differently, whether it allocates fake stack for fewer functions, or what else. E.g. try ltrace on it to see how many each one allocates and if it is very different, it would be nice to understand why.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #5 from Martin Liška --- > FakeStack (stack_size_log=20):bail out after 8192 > FakeStack (stack_size_log=20):bail out after 16384 > FakeStack (stack_size_log=20):bail out after 4096 > FakeStack (stack_size_log=20):bail out after 8192 > ... This comes after the 27000 success Fake stack allocations.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added Status|WAITING |NEW --- Comment #4 from Martin Liška --- I was able to reproduce that! The problem is the number of fake stacks that are exhausted. I see 27000x: FakeStack (stack_size_log=20):succ after 0 FakeStack (stack_size_log=20):succ after 0 FakeStack (stack_size_log=20):succ after 0 FakeStack (stack_size_log=20):succ after 0 ... FakeStack (stack_size_log=20):bail out after 8192 FakeStack (stack_size_log=20):bail out after 16384 FakeStack (stack_size_log=20):bail out after 4096 FakeStack (stack_size_log=20):bail out after 8192 ... So it seems that the program uses enormous number of stacks. Or the run-time does not release them in an efficient way.
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 --- Comment #3 from Rafael Avila de Espindola --- Yes, our build bots use podman, so you can reproduce with: $ git clone https://github.com/scylladb/seastar $ cd seastar $ podman run -v $PWD:$PWD:z -w $PWD -it docker.io/scylladb/scylla-toolchain:fedora-31-20200512 $ mkdir build $ cd build $ cmake -DCMAKE_BUILD_TYPE=Debug -G Ninja ../ $ ninja tests/unit/chunked_fifo_test $ time ASAN_OPTIONS=detect_stack_use_after_return=0 ./tests/unit/chunked_fifo_test -t chunked_fifo_big $ time ASAN_OPTIONS=detect_stack_use_after_return=1 ./tests/unit/chunked_fifo_test -t chunked_fifo_big $ cd .. $ dnf install clang $ mkdir build-clang $ cd build-clang $ CC=clang CXX=clang++ cmake -DCMAKE_BUILD_TYPE=Debug -G Ninja ../ -DSeastar_CXX_FLAGS=-Wno-error $ ninja tests/unit/chunked_fifo_test $ time ASAN_OPTIONS=detect_stack_use_after_return=0 ./tests/unit/chunked_fifo_test -t chunked_fifo_big $ time ASAN_OPTIONS=detect_stack_use_after_return=1 ./tests/unit/chunked_fifo_test -t chunked_fifo_big
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added Status|ASSIGNED|WAITING --- Comment #2 from Martin Liška --- I'm sorry but I can't install all pre-requisites. Do you have a Docker/podman I can use to build the package? Or do you have a reduced test-case?
[Bug sanitizer/94910] detect_stack_use_after_return=1 is much slower than clang's
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94910 Martin Liška changed: What|Removed |Added Ever confirmed|0 |1 Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2020-05-04 Assignee|unassigned at gcc dot gnu.org |marxin at gcc dot gnu.org --- Comment #1 from Martin Liška --- Thanks for the report, I'll take a look.
