Re: [PATCH] MIPS/bpf: Enable bpf_probe_read{, str}() on MIPS again

[Thomas Bogendoerfer]

On Wed, Mar 17, 2021 at 11:18:48PM +0100, Daniel Borkmann wrote:
> On 3/17/21 8:15 AM, Tiezhu Yang wrote:
> > After commit 0ebeea8ca8a4 ("bpf: Restrict bpf_probe_read{, str}() only to
> > archs where they work"), bpf_probe_read{, str}() functions were not longer
> > available on MIPS, so there exists some errors when running bpf program:
> > 
> > root@linux:/home/loongson/bcc# python examples/tracing/task_switch.py
> > bpf: Failed to load program: Invalid argument
> > [...]
> > 11: (85) call bpf_probe_read#4
> > unknown func bpf_probe_read#4
> > [...]
> > Exception: Failed to load BPF program count_sched: Invalid argument
> > 
> > So select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE in arch/mips/Kconfig,
> > otherwise the bpf old helper bpf_probe_read() will not be available.
> > 
> > This is similar with the commit d195b1d1d1196 ("powerpc/bpf: Enable
> > bpf_probe_read{, str}() on powerpc again").
> > 
> > Fixes: 0ebeea8ca8a4 ("bpf: Restrict bpf_probe_read{, str}() only to archs where they work")
> > Signed-off-by: Tiezhu Yang <yangtiezhu@xxxxxxxxxxx>
> 
> Thomas, I presume you pick this up via mips tree (with typos fixed)? Or do you
> want us to route the fix via bpf with your ACK? (I'm fine either way.)

I'll take it via mips tree.

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]