> -----Original Message----- > From: Martin K. Petersen [mailto:martin.petersen@xxxxxxxxxx] > Sent: Tuesday, August 09, 2016 6:35 AM > To: Mike Christie > Cc: Jitendra Bhivare; linux-scsi@xxxxxxxxxxxxxxx > Subject: Re: [PATCH 02/28] be2iscsi: Replace _bh with _irqsave/irqrestore > > >>>>> "Mike" == Mike Christie <mchristi@xxxxxxxxxx> writes: > > >> In beiscsi_alloc_pdu, _bh versions of spin_lock are being used for > >> protecting SGLs and WRBs. _bh versions are needed as the function > >> gets invoked in process context and BLOCK_IOPOLL softirq. > >> > >> In spin_unlock_bh, after releasing the lock and enabling BH, > >> do_softirq is called which executes till last SOFTIRQ. > >> > >> beiscsi_alloc_pdu is called under session lock. Through block layer, > >> iSCSI stack in some cases send IOs with interrupts disabled. In such > >> paths, > > > Mike> What path is this? Is this with mq enabled or disabled? > > Jitendra? > > -- > Martin K. Petersen Oracle Linux Engineering [JB] Sorry for the delayed response, there was some issue with my mail client. There are paths block layer where IRQs are disabled with request_queue queue_lock. - blk_timeout_work : this triggers NOP-OUT thru' iscsi_eh_cmd_timed_out. - blk_execute_rq_nowait -- To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
