Re: Memory corruption due to word sharing

[James Courtier-Dutton]

On 1 February 2012 15:19, Jan Kara <jack@xxxxxxx> wrote:
> ÂHello,
>
> Âwe've spotted the following mismatch between what kernel folks expect
> from a compiler and what GCC really does, resulting in memory corruption on
> some architectures. Consider the following structure:
> struct x {
> Â Âlong a;
> Â Âunsigned int b1;
> Â Âunsigned int b2:1;
> };
>
> We have two processes P1 and P2 where P1 updates field b1 and P2 updates
> bitfield b2. The code GCC generates for b2 = 1 e.g. on ia64 is:
> Â 0: Â 09 00 21 40 00 21 Â Â Â [MMI] Â Â Â adds r32=8,r32
> Â 6: Â 00 00 00 02 00 e0 Â Â Â Â Â Â Â Â Â nop.m 0x0
> Â c: Â 11 00 00 90 Â Â Â Â Â Â Â Â Â Â Â Â mov r15=1;;
> Â10: Â 0b 70 00 40 18 10 Â Â Â [MMI] Â Â Â ld8 r14=[r32];;
> Â16: Â 00 00 00 02 00 c0 Â Â Â Â Â Â Â Â Â nop.m 0x0
> Â1c: Â f1 70 c0 47 Â Â Â Â Â Â Â Â Â Â Â Â dep r14=r15,r14,32,1;;
> Â20: Â 11 00 38 40 98 11 Â Â Â [MIB] Â Â Â st8 [r32]=r14
> Â26: Â 00 00 00 02 00 80 Â Â Â Â Â Â Â Â Â nop.i 0x0
> Â2c: Â 08 00 84 00 Â Â Â Â Â Â Â Â Â Â Â Â br.ret.sptk.many b0;;
>
> Note that gcc used 64-bit read-modify-write cycle to update b2. Thus if P1
> races with P2, update of b1 can get lost. BTW: I've just checked on x86_64
> and there GCC uses 8-bit bitop to modify the bitfield.
>
> We actually spotted this race in practice in btrfs on structure
> fs/btrfs/ctree.h:struct btrfs_block_rsv where spinlock content got
> corrupted due to update of following bitfield and there seem to be other
> places in kernel where this could happen.
>
> I've raised the issue with our GCC guys and they said to me that: "C does
> not provide such guarantee, nor can you reliably lock different
> structure fields with different locks if they share naturally aligned
> word-size memory regions. ÂThe C++11 memory model would guarantee this,
> but that's not implemented nor do you build the kernel with a C++11
> compiler."
>
> So it seems what C/GCC promises does not quite match with what kernel
> expects. I'm not really an expert in this area so I wanted to report it
> here so that more knowledgeable people can decide how to solve the issue...
>
> Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â ÂHonza
> --
> Jan Kara <jack@xxxxxxx>
> SUSE Labs, CR

What is the recommended work around for this problem?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/