On Tue, 2018-05-15 at 09:00 -0700, Matthew Wilcox wrote:> diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c > index 025dc2d3f3de..cdf671c2af61 100644 > --- a/drivers/scsi/qla2xxx/qla_target.c > +++ b/drivers/scsi/qla2xxx/qla_target.c > @@ -3719,7 +3719,8 @@ void qlt_free_cmd(struct qla_tgt_cmd *cmd) > return; > } > cmd->jiffies_at_free = get_jiffies_64(); > - percpu_ida_free(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag); > + sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag, > + cmd->se_cmd.map_cpu); > } > EXPORT_SYMBOL(qlt_free_cmd);Please introduce functions in the target core for allocating and freeing a tag instead of spreading the knowledge of how to allocate and free tags over all target drivers.> +int iscsit_wait_for_tag(struct se_session *se_sess, int state, int *cpup) > +{ > + int tag = -1; > + DEFINE_WAIT(wait); > + struct sbq_wait_state *ws; > + > + if (state == TASK_RUNNING) > + return tag; > + > + ws = &se_sess->sess_tag_pool.ws[0]; > + for (;;) { > + prepare_to_wait_exclusive(&ws->wait, &wait, state); > + if (signal_pending_state(state, current)) > + break;This looks weird to me. Shouldn't target code ignore signals instead of causing tag allocation to fail if a signal is received?> + schedule(); > + tag = sbitmap_queue_get(&se_sess->sess_tag_pool, cpup); > + } > + > + finish_wait(&ws->wait, &wait); > + return tag; > +}Thanks, Bart.
On Tue, Jun 12, 2018 at 03:22:42PM +0000, Bart Van Assche wrote:> On Tue, 2018-05-15 at 09:00 -0700, Matthew Wilcox wrote: > > diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c > > index 025dc2d3f3de..cdf671c2af61 100644 > > --- a/drivers/scsi/qla2xxx/qla_target.c > > +++ b/drivers/scsi/qla2xxx/qla_target.c > > @@ -3719,7 +3719,8 @@ void qlt_free_cmd(struct qla_tgt_cmd *cmd) > > return; > > } > > cmd->jiffies_at_free = get_jiffies_64(); > > - percpu_ida_free(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag); > > + sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag, > > + cmd->se_cmd.map_cpu); > > } > > EXPORT_SYMBOL(qlt_free_cmd); > > Please introduce functions in the target core for allocating and freeing a tag > instead of spreading the knowledge of how to allocate and free tags over all > target drivers.I can't without doing an unreasonably large amount of work on drivers that I have no way to test. Some of the drivers have the se_cmd already; some of them don't. I'd be happy to introduce a common function for freeing a tag.> > +int iscsit_wait_for_tag(struct se_session *se_sess, int state, int *cpup) > > +{ > > + int tag = -1; > > + DEFINE_WAIT(wait); > > + struct sbq_wait_state *ws; > > + > > + if (state == TASK_RUNNING) > > + return tag; > > + > > + ws = &se_sess->sess_tag_pool.ws[0]; > > + for (;;) { > > + prepare_to_wait_exclusive(&ws->wait, &wait, state); > > + if (signal_pending_state(state, current)) > > + break; > > This looks weird to me. Shouldn't target code ignore signals instead of causing > tag allocation to fail if a signal is received?It's what the current code did: - if (signal_pending_state(state, current)) { - tag = -ERESTARTSYS; - break; - } and the current callers literally indicate that they want signals: drivers/infiniband/ulp/isert/ib_isert.c: cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE); drivers/target/iscsi/iscsi_target.c: cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE); (etc)
On Tue, 2018-06-12 at 09:15 -0700, Matthew Wilcox wrote:> On Tue, Jun 12, 2018 at 03:22:42PM +0000, Bart Van Assche wrote: > > On Tue, 2018-05-15 at 09:00 -0700, Matthew Wilcox wrote: > > > diff --git a/drivers/scsi/qla2xxx/qla_target.c b/drivers/scsi/qla2xxx/qla_target.c > > > index 025dc2d3f3de..cdf671c2af61 100644 > > > --- a/drivers/scsi/qla2xxx/qla_target.c > > > +++ b/drivers/scsi/qla2xxx/qla_target.c > > > @@ -3719,7 +3719,8 @@ void qlt_free_cmd(struct qla_tgt_cmd *cmd) > > > return; > > > } > > > cmd->jiffies_at_free = get_jiffies_64(); > > > - percpu_ida_free(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag); > > > + sbitmap_queue_clear(&sess->se_sess->sess_tag_pool, cmd->se_cmd.map_tag, > > > + cmd->se_cmd.map_cpu); > > > } > > > EXPORT_SYMBOL(qlt_free_cmd); > > > > Please introduce functions in the target core for allocating and freeing a tag > > instead of spreading the knowledge of how to allocate and free tags over all > > target drivers. > > I can't without doing an unreasonably large amount of work on drivers that > I have no way to test. Some of the drivers have the se_cmd already; some > of them don't. I'd be happy to introduce a common function for freeing > a tag.Which target drivers are you referring to? If you are referring to the sbp driver: I think that driver is dead and can be removed from the kernel tree. I even don't know whether that driver ever has had any users other than the developer of that driver.> > This looks weird to me. Shouldn't target code ignore signals instead of causing > > tag allocation to fail if a signal is received? > > It's what the current code did: > > - if (signal_pending_state(state, current)) { > - tag = -ERESTARTSYS; > - break; > - } > > and the current callers literally indicate that they want signals: > > drivers/infiniband/ulp/isert/ib_isert.c: cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE); > drivers/target/iscsi/iscsi_target.c: cmd = iscsit_allocate_cmd(conn, TASK_INTERRUPTIBLE);Right, the iSCSI target driver uses signals to wake up threads (see also the send_sig() calls in the iSCSI target code). Bart.