On Fri, Apr 14, 2023 at 05:17:42PM +0100, Andrew Cooper wrote: > On 14/04/2023 4:19 pm, Roger Pau Monne wrote: > > From: Josh Poimboeuf <[email protected]> > > > > The paravirt_patch_site struct has 12 bytes of data and 4 bytes of > > padding, for a total of 16 bytes. However, when laying out the structs > > in the .parainstructions section, the vmlinux script only aligns before > > each struct's data, not after. So the last entry doesn't have the > > 4-byte padding, which breaks kpatch_regenerate_special_section()'s > > assumption of a 16-byte struct, resulting in a memcpy past the end of > > the section. > > > > Fixes #747. > > > > Signed-off-by: Josh Poimboeuf <[email protected]> > > > > This is commit: > > > > c2dc3836e862 create-diff-object: handle missing padding at end of special > > section > > > > In kpatch repository. > > > > I've seen the .fixup section get an alignment of 16 but a size of 81, > > which makes the error removed in this patch trigger. Overall I'm not > > sure why the original alignment check was done against the size of the > > section, the alignment applies to the address of the section, not its > > size. > > > > Signed-off-by: Roger Pau Monné <[email protected]> > > Seems like a clean backport, so FWIW > > Acked-by: Andrew Cooper <[email protected]> > > However, surely we want a correction to Xen's linker file too, to stop > putting out a badly aligned section?
AFAICT that alignment comes from the per-function-section object files, so that's before the linker has assembled the xen image. And the address of the section is indeed alignment to the value, so it's all correct. Even then, it's my understanding the alignment in sh_addralign applies to the address of the section, not the size, so I'm confused as to why create-diff-object was expecting section sizes to the aligned. IMO it would make sense to pad the start address so it's aligned to the section requirements, but not the section size. Regardless, it's indeed a clean backport from the change upstream so we should take it. Thanks, Roger.
