Re: [PATCH bpf-next v4 04/11] bpf: Rename BPF_XADD and prepare to encode other atomics in .imm

From: John Fastabend <john.fastabend@gmail.com>
To: Brendan Jackman <jackmanb@google.com>,
	John Fastabend <john.fastabend@gmail.com>
Cc: bpf@vger.kernel.org, Alexei Starovoitov <ast@kernel.org>,
	Yonghong Song <yhs@fb.com>,
	Daniel Borkmann <daniel@iogearbox.net>,
	KP Singh <kpsingh@chromium.org>,
	Florent Revest <revest@chromium.org>,
	linux-kernel@vger.kernel.org, Jann Horn <jannh@google.com>
Subject: Re: [PATCH bpf-next v4 04/11] bpf: Rename BPF_XADD and prepare to encode other atomics in .imm
Date: Tue, 08 Dec 2020 21:40:03 -0800	[thread overview]
Message-ID: <5fd06333d11ff_50ce2082d@john-XPS-13-9370.notmuch> (raw)
In-Reply-To: <X89G2kItO2o60+A6@google.com>

Brendan Jackman wrote:
> Hi John, thanks a lot for the reviews!
> 
> On Mon, Dec 07, 2020 at 01:56:53PM -0800, John Fastabend wrote:
> > Brendan Jackman wrote:
> > > A subsequent patch will add additional atomic operations. These new
> > > operations will use the same opcode field as the existing XADD, with
> > > the immediate discriminating different operations.
> > > 
> > > In preparation, rename the instruction mode BPF_ATOMIC and start
> > > calling the zero immediate BPF_ADD.
> > > 
> > > This is possible (doesn't break existing valid BPF progs) because the
> > > immediate field is currently reserved MBZ and BPF_ADD is zero.
> > > 
> > > All uses are removed from the tree but the BPF_XADD definition is
> > > kept around to avoid breaking builds for people including kernel
> > > headers.
> > > 
> > > Signed-off-by: Brendan Jackman <jackmanb@google.com>
> > > ---

[...]

> > > +	case BPF_STX | BPF_ATOMIC | BPF_W:
> > > +	case BPF_STX | BPF_ATOMIC | BPF_DW:
> > > +		if (insn->imm != BPF_ADD) {
> > > +			pr_err("bpf-jit: not supported: atomic operation %02x ***\n",
> > > +			       insn->imm);
> > > +			return -EINVAL;
> > > +		}
> > 
> > Can we standardize the error across jits and the error return code? It seems
> > odd that we use pr_err, pr_info_once, pr_err_ratelimited and then return
> > ENOTSUPP, EFAULT or EINVAL.
> 
> That would be a noble cause but I don't think it makes sense in this
> patchset: they are already inconsistent, so here I've gone for intra-JIT
> consistency over inter-JIT consistency.
> 
> I think it would be more annoying, for example, if the s390 JIT returned
> -EOPNOTSUPP for a bad atomic but -1 for other unsupported ops, than it
> is already that the s390 JIT returns -1 where the MIPS returns -EINVAL.

ok works for me thanks for the explanation.