On 26/09/2022 20.47, Song Liu wrote:
Changes v1 => v2:
1. Update arch_prepare_bpf_dispatcher to use a RO image and a RW buffer.
(Alexei) Note: I haven't found an existing test to cover this part, so
this part was tested manually (comparing the generated dispatcher is
the same).
Jeff Layton reported CPA W^X warning linux-next [1]. It turns out to be
W^X issue with bpf trampoline and bpf dispatcher. Fix these by:
1. Use bpf_prog_pack for bpf_dispatcher;
2. Set memory permission properly with bpf trampoline.
Indirectly related to your patchset[0].
- TL;DR calling set_memory_x() have side-effects
We are getting reports that loading BPF-progs (jit stage) cause issues
for RT in the form of triggering work on isolated CPUs. It looks like
BTF JIT stage cause a TLB flush on all CPUs, including isolated CPUs.
The triggering function is set_memory_x() (see call-stack[2]).
We have noticed (and appreciate) you have previously improved the
situation in this patchset[3]:
[3]
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=80123f0ac4a6
Is this patchset also part of improving the situation, or does it
introduce more calls to set_memory_x() ?
[1] https://lore.kernel.org/lkml/c84cc27c1a5031a003039748c3c099732a718aec.camel@xxxxxxxxxx/
[2] Call stack triggering issue:
smp_call_function_many_cond+0x1
smp_call_function+0x39
on_each_cpu+0x2a
cpa_flush+0x11a
change_page_attr_set_clr+0x129
set_memory_x+0x37
bpf_int_jit_compile+0x36f
bpf_prog_select_runtime+0xc6
bpf_prepare_filter+0x523
sk_attach_filter+0x13
sock_setsockopt+0x920
__sys_setsockopt+0x16a
__x64_sys_setsockopt+0x20
do_syscall_64+0x87
entry_SYSCALL_64_after_hwframe+0x65
[0] https://lore.kernel.org/all/20220926184739.3512547-1-song@xxxxxxxxxx/#r
--Jesper
