| Message ID | 1710949294-29287-1-git-send-email-xiaojiangfeng@huawei.com (mailing list archive) |
|---|---|
| State | Handled Elsewhere |
| Headers | show |
| Series | [v3] ARM: unwind: improve unwinders for noreturn case | expand |
On Wed, Mar 20, 2024 at 11:41:34PM +0800, Jiangfeng Xiao wrote: > This is an off-by-one bug which is common in unwinders, > due to the fact that the address on the stack points > to the return address rather than the call address. > > So, for example, when the last instruction of a function > is a function call (e.g., to a noreturn function), it can > cause the unwinder to incorrectly try to unwind from > the function after the callee. > > foo: > ... > bl bar > ... end of function and thus next function ... > > which results in LR pointing into the next function. > > Fixed this by subtracting 1 from frmae->pc in the call frame > like ORC on x86 does. > > Refer to the unwind_next_frame function in the unwind_orc.c This came in while I was still replying to your previous reply, so I'm going to ignore this. Please allow at least 24 hours between postings, and please allow discussion to finish before posting a new version - give your reviewers adequate time to compose a reply bearing in mind that timezones might get in the way, but also making supper (as is the case in this instance) may cause several hour delay in reply. Thanks.
diff --git a/arch/arm/kernel/unwind.c b/arch/arm/kernel/unwind.c index 9d21921..abfa8e9 100644 --- a/arch/arm/kernel/unwind.c +++ b/arch/arm/kernel/unwind.c @@ -514,6 +514,14 @@ int unwind_frame(struct stackframe *frame) frame->sp = ctrl.vrs[SP]; frame->lr = ctrl.vrs[LR]; frame->pc = ctrl.vrs[PC]; + /* + * When the last instruction of a function is a function call + * (e.g., to a noreturn function), it can cause the unwinder + * incorrectly try to unwind from the function after the callee, + * fixed this by subtracting 1 from frame->pc in the call frame + * like ORC on x86 does. + */ + frame->pc = frame->pc - 1; frame->lr_addr = ctrl.lr_addr; return URC_OK;
This is an off-by-one bug which is common in unwinders, due to the fact that the address on the stack points to the return address rather than the call address. So, for example, when the last instruction of a function is a function call (e.g., to a noreturn function), it can cause the unwinder to incorrectly try to unwind from the function after the callee. foo: ... bl bar ... end of function and thus next function ... which results in LR pointing into the next function. Fixed this by subtracting 1 from frmae->pc in the call frame like ORC on x86 does. Refer to the unwind_next_frame function in the unwind_orc.c Suggested-by: Josh Poimboeuf <jpoimboe@kernel.org> Link: https://lkml.kernel.org/lkml/20240305175846.qnyiru7uaa7itqba@treble/ Suggested-by: "Russell King (Oracle)" <linux@armlinux.org.uk> Link: https://lkml.kernel.org/lkml/Zeg8wRYFemMjcCxG@shell.armlinux.org.uk/ Signed-off-by: Jiangfeng Xiao <xiaojiangfeng@huawei.com> --- ChangeLog v1->v2 - stay printk("%s...", loglvl, ...) ChangeLog v2->v3 - Redundant code is deleted to simplify the code --- arch/arm/kernel/unwind.c | 8 ++++++++ 1 file changed, 8 insertions(+)