[lkml]   [2006]   [Feb]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: [lock validator] inet6_destroy_sock(): soft-safe -> soft-unsafe lock dependency

    * Herbert Xu <> wrote:

    > On Tue, Jan 31, 2006 at 10:24:32PM +0100, Ingo Molnar wrote:
    > >
    > > [<c04de9e8>] _write_lock+0x8/0x10
    > > [<c0499015>] inet6_destroy_sock+0x25/0x100
    > > [<c04b8672>] tcp_v6_destroy_sock+0x12/0x20
    > > [<c046bbda>] inet_csk_destroy_sock+0x4a/0x150
    > > [<c047625c>] tcp_rcv_state_process+0xd4c/0xdd0
    > > [<c047d8e9>] tcp_v4_do_rcv+0xa9/0x340
    > > [<c047eabb>] tcp_v4_rcv+0x8eb/0x9d0
    > OK this is definitely broken. We should never touch the dst lock in
    > softirq context. Since inet6_destroy_sock may be called from that
    > context due to the asynchronous nature of sockets, we can't take the
    > lock there.
    > In fact this sk_dst_reset is totally redundant since all IPv6 sockets
    > use inet_sock_destruct as their socket destructor which always cleans
    > up the dst anyway. So the solution is to simply remove the call.
    > Signed-off-by: Herbert Xu <>

    cool - i've booted your patch and will have results later today [it's
    looking good so far, after 10 minutes of uptime ;)]

    btw., this codepath took some time to trigger, and i'm not sure why:
    maybe because i dont have any true ipv6 traffic? (In fact i dont have
    CONFIG_IPV6 enabled at all in this kernel config - so this codepath must
    be an effect of ipv4/ipv6 unification?) I guess i should run the ipv6
    testsuite to expose all the important codepaths to the validator? [Would
    you have any suggestions for me how to do that quickly & easily?]

    another thing: could you add the string 'lock validator' somewhere into
    the changelog, so that we can grep for such things later on? One reason
    for that is to strenghten my future arguments for mainline inclusion
    ;-), but there's another argument as well:

    the lock validator finds _all_ hidden deadlocks [no matter how obscure,
    interdependent or unlikely they are, as long as the affected codepath is
    triggered at least once], and thus the resulting bug statistics and
    characteristics will be an excellent (and one-time) opportunity to
    objectively judge the absolute code quality (and defect rate) of the
    Linux kernel, for an important category of hard-to-find bugs. (Maybe the
    results can even be extrapolated to other, similarly hard-to-find bug

    in other words, the lock validator is building a runtime set of formal
    "locking requirements", and is automatically proving (and maintaining
    the proof) that all locking activity within the kernel meets those
    requirements, mathematically. It thus enables us to achieve a hard
    ceiling of 100% correctness that we can rarely achieve in complex code
    as the kernel. We should minimize the changeset entropy introduced and
    preserve this historic information.

    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to
    More majordomo info at
    Please read the FAQ at

     \ /
      Last update: 2006-02-01 12:17    [W:0.025 / U:223.732 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site