[lkml]   [2012]   [Feb]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
    Subject[RFCv7 PATCH 0/4] Add poll_requested_events() function.
    Hi all,

    This is the seventh version of this patch series (the fifth and sixth where
    never posted and where internal iterations only).

    Al Viro had concerns about silent API changes. I have made an extensive
    analysis of that in my comments in patch 2/4.

    This patch series is rebased to v3.3-rc2. The changes compared to the
    previously posted version are:

    - I have renamed the qproc field to pq_proc to prevent any driver that tries
    to access that directly to fail. No kernel driver does this, BTW.

    - I added a new poll_does_not_wait() inline that returns true if it is known
    that poll() will not wait on return. This removes the last reason for
    looking inside the poll_table struct. include/net/sock.h has been adapted
    to use this new inline (and it is the only place inside the kernel that
    need this).

    I hope that the analysis I made answers any remaining concerns about possible
    silent API changes.

    This patch series is also available here:

    It was suggested to me that creating a new poll system call might be an option
    as well. I've attempted that as well and code implementing that can be found

    However, I think this turned out to be very messy. And because some drivers
    call the poll fop directly or through some framework I could not be certain I
    was not introducing any errors.

    If it is really required to change the API in some way, then I would suggest
    changing this:

    typedef struct poll_table_struct {
    poll_queue_proc pq_proc;
    unsigned long key;
    } poll_table;

    to this:

    struct poll_table {
    poll_queue_proc pq_proc;
    unsigned long key;

    and adapting all users.

    However, I honestly do not think this is necessary at all. But if it is the
    only way to get this in, then I'll do the work. The media/video subsystem really
    needs this functionality. Also note that previous versions of this patch have
    been in linux-next for months now.

    The first version of this patch was posted July 1st, 2011. I really hope that
    it won't take another six months to get a review from a fs developer. As this
    LWN article ( said: 'There has been little
    discussion of the patch; it doesn't seem like there is any real reason for it
    not to go in for 3.1.'

    The earliest this can go in now is 3.4. The only reason it takes so long is
    that it has been almost impossible to get a Ack or comments or even just a
    simple reply from the fs developers. That is really frustrating, I'm sorry
    to say.

    Anyway, comments, reviews, etc. are very welcome.



     \ /
      Last update: 2012-02-02 11:37    [W:0.046 / U:69.976 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site