Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:3 as permitted sender) client-ip=2620:137:e000::3:3;
Message-ID: <c0c9b60a3c14080f93a8b7b38f0277c6dd0bf7fc.camel@gmail.com>
Subject: Re: [devel-ipsec] [PATCH ipsec-next v3 3/9] libbpf: Add
 BPF_CORE_WRITE_BITFIELD() macro
From:   Eduard Zingerman <eddyz87@gmail.com>
To:     Daniel Xu <dxu@dxuuu.xyz>, ndesaulniers@google.com,
        daniel@iogearbox.net, nathan@kernel.org, ast@kernel.org,
        andrii@kernel.org, steffen.klassert@secunet.com,
        antony.antony@secunet.com, alexei.starovoitov@gmail.com,
        yonghong.song@linux.dev, martin.lau@linux.dev, song@kernel.org,
        john.fastabend@gmail.com, kpsingh@kernel.org, sdf@google.com,
        haoluo@google.com, jolsa@kernel.org, trix@redhat.com,
        bpf@vger.kernel.org, linux-kernel@vger.kernel.org,
        llvm@lists.linux.dev, devel@linux-ipsec.org,
        netdev@vger.kernel.org, Jonathan Lemon <jlemon@aviatrix.com>
Date:   Fri, 01 Dec 2023 23:22:52 +0200
In-Reply-To: <2schji4oladptrev3tswmwkbhspz6mdy5u2v7tvll4du7iylri@2u2zmfdzn6fm>
References: <cover.1701462010.git.dxu@dxuuu.xyz>
         <adea997dff6d07332d294ad9cd233f3b71494a81.1701462010.git.dxu@dxuuu.xyz>
         <2schji4oladptrev3tswmwkbhspz6mdy5u2v7tvll4du7iylri@2u2zmfdzn6fm>
Autocrypt: addr=eddyz87@gmail.com; prefer-encrypt=mutual; keydata=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
 nYzhLWGcczc6J71q1Dje0l5vIPaSFOgwmWD4DA+WvuxM/shH4rtWeodbv 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
 t1iq+gsfnXbPz5AnS598ScZI1oP7OrPSFJkt/z4acEbOQDQs8aUqrd46PV jsdqGvKnXZxzylux29UTNby4jTlz9pNJM+wPrDRmGfchLDUmf6CffaUYCbu4FiId+9+dcTCDvxbABRy1C3OJ8QY7cxfJ+pEZW18fRJ0XCl/fiV/ecAOfB3HsqgTzAn555h0rkFgay0hAvMU/mAW/CFNSIxV397zm749ZNLA0L2dMy1AKuOqH+/B+/ImBfJMDjmdyJQ8WU/OFRuGLdqOd2oZrA1iuPIa+yUYyZkaZfz/emQwpIL1+Q4p1R/OplA4yc301AqruXXUcVDbEB+joHW3hy5FwK5t5OwTKatrSJBkydSF9zdXy98fYzGniRyRA65P0Ix/8J3BYB4edY2/w0Ip/mdYsYQljBY0A==
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.50.1 
MIME-Version: 1.0
Precedence: bulk

On Fri, 2023-12-01 at 13:51 -0700, Daniel Xu wrote:
> On Fri, Dec 01, 2023 at 01:23:14PM -0700, Daniel Xu via Devel wrote:
> > =3D=3D=3D Motivation =3D=3D=3D
> >=20
> > Similar to reading from CO-RE bitfields, we need a CO-RE aware bitfield
> > writing wrapper to make the verifier happy.
> >=20
> > Two alternatives to this approach are:
> >=20
> > 1. Use the upcoming `preserve_static_offset` [0] attribute to disable
> >    CO-RE on specific structs.
> > 2. Use broader byte-sized writes to write to bitfields.
> >=20
> > (1) is a bit hard to use. It requires specific and not-very-obvious
> > annotations to bpftool generated vmlinux.h. It's also not generally
> > available in released LLVM versions yet.
> >=20
> > (2) makes the code quite hard to read and write. And especially if
> > BPF_CORE_READ_BITFIELD() is already being used, it makes more sense to
> > to have an inverse helper for writing.
> >=20
> > =3D=3D=3D Implementation details =3D=3D=3D
> >=20
> > Since the logic is a bit non-obvious, I thought it would be helpful
> > to explain exactly what's going on.
> >=20
> > To start, it helps by explaining what LSHIFT_U64 (lshift) and RSHIFT_U6=
4
> > (rshift) is designed to mean. Consider the core of the
> > BPF_CORE_READ_BITFIELD() algorithm:
> >=20
> >         val <<=3D __CORE_RELO(s, field, LSHIFT_U64);
> >                 val =3D val >> __CORE_RELO(s, field, RSHIFT_U64);
> >=20
> > Basically what happens is we lshift to clear the non-relevant (blank)
> > higher order bits. Then we rshift to bring the relevant bits (bitfield)
> > down to LSB position (while also clearing blank lower order bits). To
> > illustrate:
> >=20
> >         Start:    ........XXX......
> >         Lshift:   XXX......00000000
> >         Rshift:   00000000000000XXX
> >=20
> > where `.` means blank bit, `0` means 0 bit, and `X` means bitfield bit.
> >=20
> > After the two operations, the bitfield is ready to be interpreted as a
> > regular integer.
> >=20
> > Next, we want to build an alternative (but more helpful) mental model
> > on lshift and rshift. That is, to consider:
> >=20
> > * rshift as the total number of blank bits in the u64
> > * lshift as number of blank bits left of the bitfield in the u64
> >=20
> > Take a moment to consider why that is true by consulting the above
> > diagram.
> >=20
> > With this insight, we can how define the following relationship:
> >=20
> >               bitfield
> >                  _
> >                 | |
> >         0.....00XXX0...00
> >         |      |   |    |
> >         |______|   |    |
> >          lshift    |    |
> >                    |____|
> >               (rshift - lshift)
> >=20
> > That is, we know the number of higher order blank bits is just lshift.
> > And the number of lower order blank bits is (rshift - lshift).
> >=20
> > Finally, we can examine the core of the write side algorithm:
> >=20
> >         mask =3D (~0ULL << rshift) >> lshift;   // 1
> >         nval =3D new_val;                       // 2
> >         nval =3D (nval << rpad) & mask;         // 3
> >         val =3D (val & ~mask) | nval;           // 4
> >=20
> > (1): Compute a mask where the set bits are the bitfield bits. The first
> >      left shift zeros out exactly the number of blank bits, leaving a
> >      bitfield sized set of 1s. The subsequent right shift inserts the
> >      correct amount of higher order blank bits.
> > (2): Place the new value into a word sized container, nval.
> > (3): Place nval at the correct bit position and mask out blank bits.
> > (4): Mix the bitfield in with original surrounding blank bits.
> >=20
> > [0]: https://reviews.llvm.org/D133361
> > Co-authored-by: Eduard Zingerman <eddyz87@gmail.com>
> > Signed-off-by: Eduard Zingerman <eddyz87@gmail.com>
>=20
> Just pointing out I inserted Eduard's tags here. Eduard - I hope that's
> OK. Not sure what the usual procedure for this is.

Not that I did a big contribution, you and Andrii figured out a much
better (and correct) expression :) I'm fine with and without this tag,
thank you for working on this.