[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-26 Thread Tom Hughes
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #10 from Tom Hughes --- The readlink support was introduced in https://sourceware.org/git/gitweb.cgi?p=valgrind.git;a=commitdiff;h=423bfef15b2e275314263fea8af11a29d5509110 (later extended to readlinkat) and the NEWS entry in

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-26 Thread Tom Hughes
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #9 from Tom Hughes --- Oddly it seems NEWS.old claims that support was added for both open and readlink in 2.4.0 but I'm not sure that's true, or if it is then I think open has been lost... -- You are receiving this mail because: You are

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-26 Thread Tom Hughes
https://bugs.kde.org/show_bug.cgi?id=399087 Tom Hughes changed: What|Removed |Added CC||t...@compton.nu --- Comment #8 from Tom Hughes

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-26 Thread John Reiser
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #7 from John Reiser --- The root cause is a symlink vulnerability! coregrind fails to do the right thing when the target executes int fd_i_am = open("/proc/self/exe", O_RDONLY); upx uses mmap(fd_i_am,) to replicate portions of the

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-26 Thread Julian Seward
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #6 from Julian Seward --- It would be useful if you could re-run with no instrumentation (--tool=none) and re-post the same logs as before. That has two purposes: first, many fewer generated insns to wade through, and secondly, if the

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-26 Thread Julian Seward
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #5 from Julian Seward --- (In reply to John Reiser from comment #0) It's clear that the thing has gone off of the rails somehow, but I'm not sure I agree with your conclusion here: > EXPECTED RESULT > "add r15, r6, r10" is recognized as a

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-25 Thread John Reiser
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #4 from John Reiser --- Executions were on Fedora 29 beta using valgrind-3.13.0-28.fc29.armv7hl.rpm. The same result was obtained using Fedora 28 (released, prior version of OS) with the same valgrind rpm. -- You are receiving this mail

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-25 Thread John Reiser
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #3 from John Reiser --- Created attachment 115241 --> https://bugs.kde.org/attachment.cgi?id=115241=edit console output with --trace-flags valgrind --trace-flags=1001 --trace-notbelow=108 ./foo 2>&1 | more -- You are receiving this

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-25 Thread John Reiser
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #2 from John Reiser --- Created attachment 115240 --> https://bugs.kde.org/attachment.cgi?id=115240=edit console output of plain memcheck Note parameter --smc-check=all as an attempt to check carefully for non-static code. -- You are

[valgrind] [Bug 399087] memcheck escape from user code into memcheck itself via computed goto

2018-09-25 Thread John Reiser
https://bugs.kde.org/show_bug.cgi?id=399087 --- Comment #1 from John Reiser --- Created attachment 115239 --> https://bugs.kde.org/attachment.cgi?id=115239=edit target program, compressed by upx -- You are receiving this mail because: You are watching all bug changes.