On Thu, Feb 11, 2021 at 03:09:43PM +0100, Christophe Leroy wrote:
Le 11/02/2021 à 12:49, Segher Boessenkool a écrit :
On Thu, Feb 11, 2021 at 07:41:52AM +0000, Christophe Leroy wrote:
powerpc BUG_ON() is based on using twnei or tdnei instruction,
which obliges gcc to format the condition into a 0 or 1 value
in a register.
Huh? Why is that?
Will it work better if this used __builtin_trap? Or does the kernel only
detect very specific forms of trap instructions?
We already made a try with __builtin_trap() 1,5 year ago, see
https://patchwork.ozlabs.org/project/linuxppc-dev/patch/20510ce03cc9463f1c9e743c1d93b939de501b53.1566219503.git.christophe.leroy@xxxxxx/
The main problems encountered are:
- It is only possible to use it for BUG_ON, not for WARN_ON because GCC
considers it as noreturn. Is there any workaround ?
A trap is noreturn by definition:
-- Built-in Function: void __builtin_trap (void)
This function causes the program to exit abnormally.
- The kernel (With CONFIG_DEBUG_BUGVERBOSE) needs to be able to identify
the source file and line corresponding to the trap. How can that be done
with __builtin_trap() ?
The DWARF debug info should be sufficient. Perhaps you can post-process
some way?
You can create a trap that falls through yourself (by having a trap-on
condition with a condition that is always true, but make the compiler
not see that). This isn't efficient though.
Could you file a feature request (in bugzilla)? It is probably useful
for generic code as well, but we could implement this for powerpc only
if needed.