On 15/04/2025 2:00 pm, Jan Beulich wrote: > On 15.04.2025 14:49, Teddy Astie wrote: >> UBSAN complains when trying memcpy with a NULL pointer even if it's going to >> copy zero bytes (which are the only cases where a NULL pointer is used). > If this really was a problem, I think we'd need to go through and find all > instances. However, ... > >> Fix this by only doing the memcpy if the pointer is non-NULL. >> >> (XEN) >> ================================================================================ >> (XEN) UBSAN: Undefined behaviour in arch/x86/mm/shadow/hvm.c:168:5 >> (XEN) null pointer passed as argument 1, declared with nonnull attribute > ... it can only be the compiler who has added the nonnull attribute; we > use it only in very few (other) places. > > Personally I find it absurd to forbid NULL here when the size is zero. Yet > I agree that the spec can be interpreted this way.
https://www.open-std.org/jtc1/sc22/wg14/www/docs/n3322.pdf This is being proposed for fixing in C2Y, because lots of people think it's absurd. However, until we can raise our -std, I think we're stuck with the current behaviour. GCC-15 introduces the nonnull_if_nonzero attribute specifically for memcpy() etc, but I don't see how we could make use of it in this case. ~Andrew
