[lkml]   [2005]   [Dec]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    SubjectRe: [PATCH -mm 1/5] New system call, unshare
    Al Viro wrote:

    >On Fri, Dec 09, 2005 at 11:55:02AM +0100, Ingo Molnar wrote:
    >>* JANAK DESAI <> wrote:
    >>>[PATCH -mm 1/5] unshare system call: System call handler function
    >>>+ if (unshare_flags & ~(CLONE_NEWNS | CLONE_VM))
    >>>+ goto errout;
    >>just curious, did you consider all the other CLONE_* flags as well, to
    >>see whether it makes sense to add unshare support for them?
    >IMO the right thing to do is
    > * accept *all* flags from the very beginning
    > * check constraints ("CLONE_NEWNS must be accompanied by CLONE_FS")
    >and either -EINVAL if they are not satisfied or silently force them.
    > * for each unimplemented flag check if we corresponding thing
    >is shared; -EINVAL otherwise.
    >Then for each flag we care to implement we should replace such check with
    >actual unsharing - a patch per flag.
    >CLONE_FS and CLONE_FILES are *definitely* worth implementing and are
    >trivial to implement. The only thing we must take care of is doing
    >all replacements under task_lock, without dropping it between updates.

    Ok, thanks. I will restructure code and reorganize patches accordingly
    and post
    updated patches.

    To answer Ingo's question, we did look at other flags when I started.
    I wanted to keep the system call simple enough, with atleast namespace
    so it would get accepted. In the original discussion on fsdevel,
    unsharing of vm
    was mentioned as useful so I added that in addition to namespace unsharing.

    >I would say that CLONE_SIGHAND is also an obvious candidate for adding.
    I did have signal handler unsharing in one of the earlier incarnation of
    the patch,
    however Chris Wright alerted me (on IRC) to a possible problem with posix
    real time signals if we allow unsharing of signal handlers. He pointed
    me to the
    way send_sigqueue is stashing sighand lock for later use and since
    timers are
    flushed on exec and exit, it may lead to an oops. Since my primary
    interest was
    in namespace unsharing, I disallowed unsharing of signal handler. I will
    take a
    look at it more detail and come back with specific issues with sighand


    >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

    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: 2005-12-09 15:18    [W:0.024 / U:6.664 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site