Re: [PATCH] qla2xxx: Protect access to qpair members with qpair->qp_lock

[Hannes Reinecke]

On 06/22/2017 03:43 PM, Johannes Thumshirn wrote:
> In qla2xx_start_scsi_mq() and qla2xx_dif_start_scsi_mq() we grab the
> qpair->qp_lock but do access members of the qpair before having the lock.
> Re-order the locking sequence to have all read and write access to qpair
> members under the qpair->qp_lock.
> 
> Signed-off-by: Johannes Thumshirn <jthumshirn@xxxxxxx>
> ---
>  drivers/scsi/qla2xxx/qla_iocb.c | 20 +++++++++++++-------
>  1 file changed, 13 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/scsi/qla2xxx/qla_iocb.c b/drivers/scsi/qla2xxx/qla_iocb.c
> index 8404f17f3c6c..425ca1646a9a 100644
> --- a/drivers/scsi/qla2xxx/qla_iocb.c
> +++ b/drivers/scsi/qla2xxx/qla_iocb.c
> @@ -1770,10 +1770,6 @@ qla2xxx_start_scsi_mq(srb_t *sp)
>  	struct qla_hw_data *ha = vha->hw;
>  	struct qla_qpair *qpair = sp->qpair;
>  
> -	/* Setup qpair pointers */
> -	rsp = qpair->rsp;
> -	req = qpair->req;
> -
>  	/* So we know we haven't pci_map'ed anything yet */
>  	tot_dsds = 0;
>  
> @@ -1788,6 +1784,10 @@ qla2xxx_start_scsi_mq(srb_t *sp)
>  	/* Acquire qpair specific lock */
>  	spin_lock_irqsave(&qpair->qp_lock, flags);
>  
> +	/* Setup qpair pointers */
> +	rsp = qpair->rsp;
> +	req = qpair->req;
> +
>  	/* Check for room in outstanding command list. */
>  	handle = req->current_outstanding_cmd;
>  	for (index = 1; index < req->num_outstanding_cmds; index++) {
> @@ -1924,24 +1924,33 @@ qla2xxx_dif_start_scsi_mq(srb_t *sp)
>  
>  #define QDSS_GOT_Q_SPACE	BIT_0
>  
> +	/* Acquire ring specific lock */
> +	spin_lock_irqsave(&qpair->qp_lock, flags);
> +
>  	/* Check for host side state */
>  	if (!qpair->online) {
>  		cmd->result = DID_NO_CONNECT << 16;
> +		spin_unlock_irqrestore(&qpair->qp_lock, flags);
>  		return QLA_INTERFACE_ERROR;
>  	}
>  
>  	if (!qpair->difdix_supported &&
>  		scsi_get_prot_op(cmd) != SCSI_PROT_NORMAL) {
>  		cmd->result = DID_NO_CONNECT << 16;
> +		spin_unlock_irqrestore(&qpair->qp_lock, flags);
>  		return QLA_INTERFACE_ERROR;
>  	}
>  
> +	spin_unlock_irqrestore(&qpair->qp_lock, flags);
> +
>  	/* Only process protection or >16 cdb in this routine */
>  	if (scsi_get_prot_op(cmd) == SCSI_PROT_NORMAL) {
>  		if (cmd->cmd_len <= 16)
>  			return qla2xxx_start_scsi_mq(sp);
>  	}
>  
Not sure if that is required; at worst we're missing a device being
online; difdix support is probably not changing that often to be
warranted being protected with a lock here.

So I'd rather omit spinlocks here.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		   Teamlead Storage & Networking
hare@xxxxxxx			               +49 911 74053 688
SUSE LINUX GmbH, Maxfeldstr. 5, 90409 NÃrnberg
GF: F. ImendÃrffer, J. Smithard, J. Guild, D. Upmanyu, G. Norton
HRB 21284 (AG NÃrnberg)