lkml.org 
[lkml]   [2004]   [May]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    Date
    From
    SubjectRe: Cache queue_congestion_on/off_threshold
    On Fri, May 07 2004, Chen, Kenneth W wrote:
    > >>>> Jens Axboe wrote on Friday, May 07, 2004 2:39 AM
    > > On Thu, May 06 2004, Chen, Kenneth W wrote:
    > > > (3) can we allocate request structure up front in __make_request?
    > > > For I/O that cannot be merged, the elevator code executes twice
    > > > in __make_request.
    > > >
    > >
    > > Actually, with the good working batching we might get away with killing
    > > freereq completely. Have you tested that (if not, could you?)
    >
    > Sorry, I'm clueless on "good working batching". If you could please give
    > me some pointers, I will definitely test it.

    Something like this.

    --- linux-2.6.6/drivers/block/ll_rw_blk.c~ 2004-05-10 16:23:45.684726955 +0200
    +++ linux-2.6.6/drivers/block/ll_rw_blk.c 2004-05-10 16:29:04.333792268 +0200
    @@ -2138,8 +2138,8 @@

    static int __make_request(request_queue_t *q, struct bio *bio)
    {
    - struct request *req, *freereq = NULL;
    int el_ret, rw, nr_sectors, cur_nr_sectors, barrier, ra;
    + struct request *req;
    sector_t sector;

    sector = bio->bi_sector;
    @@ -2161,15 +2161,15 @@

    ra = bio->bi_rw & (1 << BIO_RW_AHEAD);

    -again:
    + if (barrier)
    + goto get_rq_nolock;
    +
    spin_lock_irq(q->queue_lock);

    if (elv_queue_empty(q)) {
    blk_plug_device(q);
    goto get_rq;
    }
    - if (barrier)
    - goto get_rq;

    el_ret = elv_merge(q, &req, bio);
    switch (el_ret) {
    @@ -2230,21 +2230,17 @@
    * a free slot.
    */
    get_rq:
    - if (freereq) {
    - req = freereq;
    - freereq = NULL;
    - } else {
    - spin_unlock_irq(q->queue_lock);
    - if ((freereq = get_request(q, rw, GFP_ATOMIC)) == NULL) {
    - /*
    - * READA bit set
    - */
    - if (ra)
    - goto end_io;
    + spin_unlock_irq(q->queue_lock);
    +get_rq_nolock:
    + req = get_request(q, rw, GFP_ATOMIC);
    + if (!req) {
    + /*
    + * READA bit set
    + */
    + if (ra)
    + goto end_io;

    - freereq = get_request_wait(q, rw);
    - }
    - goto again;
    + req = get_request_wait(q, rw);
    }

    req->flags |= REQ_CMD;
    @@ -2276,10 +2272,9 @@
    req->rq_disk = bio->bi_bdev->bd_disk;
    req->start_time = jiffies;

    + spin_lock_irq(q->queue_lock);
    add_request(q, req);
    out:
    - if (freereq)
    - __blk_put_request(q, freereq);

    if (blk_queue_plugged(q)) {
    int nrq = q->rq.count[READ] + q->rq.count[WRITE] - q->in_flight;
    --
    Jens Axboe

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

    \
     
     \ /
      Last update: 2005-03-22 14:03    [W:4.130 / U:0.000 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site