Re: [PATCH resend] bsg: Add support for submitting requests at tail of queue

[Boaz Harrosh <bharrosh@panasas.com>]

Boaz Harrosh wrote:
> Hi Tomo Jens
> 
> Tomo you never ack-by on this patch. I absolutely needs this for the
> user-mode API of osd-initiator. Which is needed with up-coming exofs
> utilities.
> 
> What do you want to do?
> 
> Thanks 
> 

Hi Jens.

I absolutely need this patch for 2.6.30 merge window. It is totally
un-dangerous since defaults are left unchanged.

I need it because the user-mode utilities for osd and exofs that
correspond to code in 2.6.30 will need this patch to compile,
because they use a flag constant defined in this patch.

I'm not getting any response from Tomo, some of the emails I send bounce,
do you know if he is on vacation or something?

Thank you
Boaz

> ---
> From: Boaz Harrosh <bharrosh@panasas.com>
> 
> Currently inherited from sg.c bsg will submit asynchronous request
> at the head-of-the-queue, (using "at_head" set in the call to
> blk_execute_rq_nowait()). This is bad in situation where the queues
> are full, requests will execute out of order, and can cause
> starvation of the first submitted requests.
> 
> The sg_io_v4->flags member is used and a bit is allocated to denote the
> Q_AT_TAIL. Zero is to queue at_head as before, to be compatible with old
> code at the write/read path. SG_IO code path behavior was changed so to
> be the same as write/read behavior. SG_IO was very rarely used and breaking
> compatibility with it is OK at this stage.
> 
> sg_io_hdr at sg.h also has a flags member and uses 3 bits from the first
> nibble and one bit from the last nibble. Even though none of these bits
> are supported by bsg, The second nibble is allocated for use by bsg. Just
> in case.
> 
> Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
> CC: Douglas Gilbert <dgilbert@interlog.com>
> ---
>  block/bsg.c         |    9 +++++++--
>  include/linux/bsg.h |    8 ++++++++
>  2 files changed, 15 insertions(+), 2 deletions(-)
> 
> diff --git a/block/bsg.c b/block/bsg.c
> index 44a2a0f..206060e 100644
> --- a/block/bsg.c
> +++ b/block/bsg.c
> @@ -350,6 +350,8 @@ static void bsg_rq_end_io(struct request *rq, int uptodate)
>  static void bsg_add_command(struct bsg_device *bd, struct request_queue *q,
>  			    struct bsg_command *bc, struct request *rq)
>  {
> +	int at_head = (0 == (bc->hdr.flags & BSG_FLAG_Q_AT_TAIL));
> +
>  	/*
>  	 * add bc command to busy queue and submit rq for io
>  	 */
> @@ -365,7 +367,7 @@ static void bsg_add_command(struct bsg_device *bd, struct request_queue *q,
>  	dprintk("%s: queueing rq %p, bc %p\n", bd->name, rq, bc);
>  
>  	rq->end_io_data = bc;
> -	blk_execute_rq_nowait(q, NULL, rq, 1, bsg_rq_end_io);
> +	blk_execute_rq_nowait(q, NULL, rq, at_head, bsg_rq_end_io);
>  }
>  
>  static struct bsg_command *bsg_next_done_cmd(struct bsg_device *bd)
> @@ -921,6 +923,7 @@ static long bsg_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
>  		struct request *rq;
>  		struct bio *bio, *bidi_bio = NULL;
>  		struct sg_io_v4 hdr;
> +		int at_head;
>  		u8 sense[SCSI_SENSE_BUFFERSIZE];
>  
>  		if (copy_from_user(&hdr, uarg, sizeof(hdr)))
> @@ -933,7 +936,9 @@ static long bsg_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
>  		bio = rq->bio;
>  		if (rq->next_rq)
>  			bidi_bio = rq->next_rq->bio;
> -		blk_execute_rq(bd->queue, NULL, rq, 0);
> +
> +		at_head = (0 == (hdr.flags & BSG_FLAG_Q_AT_TAIL));
> +		blk_execute_rq(bd->queue, NULL, rq, at_head);
>  		ret = blk_complete_sgv4_hdr_rq(rq, &hdr, bio, bidi_bio);
>  
>  		if (copy_to_user(uarg, &hdr, sizeof(hdr)))
> diff --git a/include/linux/bsg.h b/include/linux/bsg.h
> index cf0303a..3f0c64a 100644
> --- a/include/linux/bsg.h
> +++ b/include/linux/bsg.h
> @@ -7,6 +7,14 @@
>  #define BSG_SUB_PROTOCOL_SCSI_TMF	1
>  #define BSG_SUB_PROTOCOL_SCSI_TRANSPORT	2
>  
> +/*
> + * For flags member below
> + * sg.h sg_io_hdr also has bits defined for it's flags member. However
> + * none of these bits are implemented/used by bsg. The bits below are
> + * allocated to not conflict with sg.h ones anyway.
> + */
> +#define BSG_FLAG_Q_AT_TAIL 0x10 /* default, == 0 at this bit, is Q_AT_HEAD */
> +
>  struct sg_io_v4 {
>  	__s32 guard;		/* [i] 'Q' to differentiate from v3 */
>  	__u32 protocol;		/* [i] 0 -> SCSI , .... */