![]() Relocation entries, which will not occur if the return thunk is in theĭiff -git a/arch/x86/kernel/ b/arch/x86/kernel/ĭiff -git a/arch/x86/lib/retpoline.S b/arch/x86/lib/retpoline.S Objtool (add_jump_destinations()) only recognizes return thunk jmps that have This is correct as the jmp is written to the correct return sequence. This jmp to the default return thunk should not happen after alternativesĠxffffffff81d12a8a : jmpq 0xffffffff81f0410b This commit broke patching of the return thunk jmp in the retpolineĠxffffffff81d12a8a : jmpq 0xffffffff81d12d60 To: x86, luto +Cc: linux-kernel, Josh Poimboeuf ` (4 subsequent siblings) 5 siblings, 1 reply 78+ messages in thread 17:10 ` x86/vdso: Run objtool on vdso32-setup David Kaplan 17:10 Ensure default return thunk isn't used at runtime David 17:10 ` David Kaplan X86/retpoline: Ensure default return thunk isn't used at runtimeģ files changed, 12 insertions(+), 4 deletions(-) Return thunk because objtool was not being run on them in some cases. Patch 2 fixes an issue where functions in vdso32-setup were using the default Patch 1 reverts a recent commit which resulted in retpoline sequences not beingĪnnotated as containing returns, which was leaving them using the default return Patch 3 enforces this by modifying _x86_return_thunk to be a ud2 after ![]() Which point the default return thunk (_x86_return_thunk) should never be used. ![]() The necessary return thunk is installed at runtime via apply_returns(), after Several CPU side-channel mitigations require the use of a special return thunk. ` (5 more replies) 0 siblings, 6 replies 78+ messages in thread ![]() text._x86.return_thunk section" David Kaplan Ensure default return thunk isn't used at runtime All of help / color / mirror / Atom feed * Ensure default return thunk isn't used at runtime 17:10 David Kaplan ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |