On 10/15/25 14:21, David Faust wrote: > The BPF backend expansion of setmem was broken, because it could elect > to use stores of HI, SI or DI modes based on the destitnation alignement > when the value was QI, but fail to duplicate the byte value across to > those larger sizes. This resulted in not all bytes bytes of the > destination actually being set to the desired value.
Oof three different typos in two sentences... Sorry about that. Fixed locally. > > Fix bpf_expand_setmem to ensure the desired byte value is really > duplicated as necessary, whether it is constant or a (sub)reg:QI. > > Tested for bpf-unknown-none target on x86_64-linux-gnu host. > > Sanity checked by compiling kernel bpf-next selftests, and > spot-checking the resulting asm for affected __builtin_memsets > used there. >
