On 6/30/20 10:07 PM, Bob Liu wrote:
> Commit 3ce4196 (scsi: iscsi: Register sysfs for iscsi workqueue) enables
> 'cpumask' support for iscsi workqueues.
>
> While there is a mistake in that commit, it's unnecessary to set
> max_active = 2 since 'cpumask' can be modified when max_active = 1.
>
> This patch change back max_active to 1 so as to keep the same behaviour as
> before.
>
> Signed-off-by: Bob Liu <bob.liu@xxxxxxxxxx>
> ---
> drivers/scsi/libiscsi.c | 2 +-
> drivers/scsi/scsi_transport_iscsi.c | 2 +-
> 2 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/scsi/libiscsi.c b/drivers/scsi/libiscsi.c
> index e5a64d4..49c8a18 100644
> --- a/drivers/scsi/libiscsi.c
> +++ b/drivers/scsi/libiscsi.c
> @@ -2629,7 +2629,7 @@ struct Scsi_Host *iscsi_host_alloc(struct scsi_host_template *sht,
> "iscsi_q_%d", shost->host_no);
> ihost->workq = alloc_workqueue("%s",
> WQ_SYSFS | __WQ_LEGACY | WQ_MEM_RECLAIM | WQ_UNBOUND,
> - 2, ihost->workq_name);
> + 1, ihost->workq_name);
> if (!ihost->workq)
> goto free_host;
> }
> diff --git a/drivers/scsi/scsi_transport_iscsi.c b/drivers/scsi/scsi_transport_iscsi.c
> index f4cc08e..7ae5024 100644
> --- a/drivers/scsi/scsi_transport_iscsi.c
> +++ b/drivers/scsi/scsi_transport_iscsi.c
> @@ -4760,7 +4760,7 @@ static __init int iscsi_transport_init(void)
>
> iscsi_eh_timer_workq = alloc_workqueue("%s",
> WQ_SYSFS | __WQ_LEGACY | WQ_MEM_RECLAIM | WQ_UNBOUND,
> - 2, "iscsi_eh");
> + 1, "iscsi_eh");
> if (!iscsi_eh_timer_workq) {
> err = -ENOMEM;
> goto release_nls;
>
Reviewed-by: Mike Christie <michael.christie@xxxxxxxxxx>
I think it should get it into 5.8 to fix the bug I mentioned in the other thread.
For 5.9, you'll want to send another patch to update the iscsi_destroy_workq that got added in 5.8.
scsi_transport_iscsi.c:
iscsi_destroy_workq = create_singlethread_workqueue("iscsi_destroy");
