netdev - Re: [PATCH v3 bpf-next 01/10] treewide: remove struct-pass-by-value from tracepoints arguments

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 22 Mar 2018 13:19:34 -0700
From:   Linus Torvalds <torvalds@...ux-foundation.org>
To:     Alexei Starovoitov <ast@...com>
Cc:     Steven Rostedt <rostedt@...dmis.org>,
        David Miller <davem@...emloft.net>,
        Daniel Borkmann <daniel@...earbox.net>,
        Peter Zijlstra <peterz@...radead.org>,
        Network Development <netdev@...r.kernel.org>,
        kernel-team <kernel-team@...com>,
        Linux API <linux-api@...r.kernel.org>
Subject: Re: [PATCH v3 bpf-next 01/10] treewide: remove struct-pass-by-value
 from tracepoints arguments

On Thu, Mar 22, 2018 at 12:31 PM, Alexei Starovoitov <ast@...com> wrote:
>
> yeah. C doesn't allow casting of 'struct s { u64 var };' into u64
> without massive hacks and aliasing warnings by compiler.

Without massive hacks? Yeah, no. But without warnings? You can do it.

  #define UINTTYPE(size) \
        __typeof__(__builtin_choose_expr(size==1, (u8)1, \
                __builtin_choose_expr(size==2, (u16)2, \
                __builtin_choose_expr(size==4, (u32)3, \
                __builtin_choose_expr(size==8, (u64)4, \
                (void)5)))))

  #define CAST_TO_U64(x) ({ \
        typeof(x) __src = (x); \
        UINTTYPE(sizeof(x)) __dst; \
        memcpy(&__dst, &__src, sizeof(__dst)); \
        (u64)__dst; })

Yeah, I'm not proud of the above, but gcc actually seems to do the
right thing for it. Doing

    struct d {
        unsigned char a,b;
    };

it generates

        movzwl %di, %eax

for the CAST_TO_U64() (the above *looks* like it only casts to 32-bit,
but it is actually a full cast to 64 bits because movzwl will also
clear the top bits of the register).

No warnings.

But is ot "massively hacky"? You be the judge. Probably.

Going the other way is trivial, you just use that same UINTTYPE()
again and just the memcpy in reverse.

NOTE! The above obviously only works for things that are actually
proper nice easy sizes.

If you want to encode a 6-byte thing in a u64, the sanest thing is
likely to use a union, and just accept crap in some bytes of the u64
(and just expect it to be undone by the reversal operation). Honestly,
I'd suggest against it, because of just horrible code generation
and/or data leakage

                Linus