2017-06-30 08:03:27

by Dan Carpenter

Subject: [PATCH 1/2] scsi: lpfc: spin_lock_irq() is not nestable

We're calling spin_lock_irq() multiple times, the problem is that on the
first spin_unlock_irq() then we will re-enable IRQs and we don't want
that.

Fixes: 966bb5b71196 ("scsi: lpfc: Break up IO ctx list into a separate get and put list")
Signed-off-by: Dan Carpenter <[email protected]>

diff --git a/drivers/scsi/lpfc/lpfc_nvmet.c b/drivers/scsi/lpfc/lpfc_nvmet.c
index 7dc061a14f95..afc523209845 100644
--- a/drivers/scsi/lpfc/lpfc_nvmet.c
+++ b/drivers/scsi/lpfc/lpfc_nvmet.c
@@ -866,44 +866,44 @@ lpfc_nvmet_cleanup_io_context(struct lpfc_hba *phba)
unsigned long flags;

spin_lock_irqsave(&phba->sli4_hba.nvmet_ctx_get_lock, flags);
- spin_lock_irq(&phba->sli4_hba.nvmet_ctx_put_lock);
+ spin_lock(&phba->sli4_hba.nvmet_ctx_put_lock);
list_for_each_entry_safe(ctx_buf, next_ctx_buf,
&phba->sli4_hba.lpfc_nvmet_ctx_get_list, list) {
- spin_lock_irq(&phba->sli4_hba.abts_nvme_buf_list_lock);
+ spin_lock(&phba->sli4_hba.abts_nvme_buf_list_lock);
list_del_init(&ctx_buf->list);
- spin_unlock_irq(&phba->sli4_hba.abts_nvme_buf_list_lock);
+ spin_unlock(&phba->sli4_hba.abts_nvme_buf_list_lock);
__lpfc_clear_active_sglq(phba,
ctx_buf->sglq->sli4_lxritag);
ctx_buf->sglq->state = SGL_FREED;
ctx_buf->sglq->ndlp = NULL;

- spin_lock_irq(&phba->sli4_hba.sgl_list_lock);
+ spin_lock(&phba->sli4_hba.sgl_list_lock);
list_add_tail(&ctx_buf->sglq->list,
&phba->sli4_hba.lpfc_nvmet_sgl_list);
- spin_unlock_irq(&phba->sli4_hba.sgl_list_lock);
+ spin_unlock(&phba->sli4_hba.sgl_list_lock);

lpfc_sli_release_iocbq(phba, ctx_buf->iocbq);
kfree(ctx_buf->context);
}
list_for_each_entry_safe(ctx_buf, next_ctx_buf,
&phba->sli4_hba.lpfc_nvmet_ctx_put_list, list) {
- spin_lock_irq(&phba->sli4_hba.abts_nvme_buf_list_lock);
+ spin_lock(&phba->sli4_hba.abts_nvme_buf_list_lock);
list_del_init(&ctx_buf->list);
- spin_unlock_irq(&phba->sli4_hba.abts_nvme_buf_list_lock);
+ spin_unlock(&phba->sli4_hba.abts_nvme_buf_list_lock);
__lpfc_clear_active_sglq(phba,
ctx_buf->sglq->sli4_lxritag);
ctx_buf->sglq->state = SGL_FREED;
ctx_buf->sglq->ndlp = NULL;

- spin_lock_irq(&phba->sli4_hba.sgl_list_lock);
+ spin_lock(&phba->sli4_hba.sgl_list_lock);
list_add_tail(&ctx_buf->sglq->list,
&phba->sli4_hba.lpfc_nvmet_sgl_list);
- spin_unlock_irq(&phba->sli4_hba.sgl_list_lock);
+ spin_unlock(&phba->sli4_hba.sgl_list_lock);

lpfc_sli_release_iocbq(phba, ctx_buf->iocbq);
kfree(ctx_buf->context);
}
- spin_unlock_irq(&phba->sli4_hba.nvmet_ctx_put_lock);
+ spin_unlock(&phba->sli4_hba.nvmet_ctx_put_lock);
spin_unlock_irqrestore(&phba->sli4_hba.nvmet_ctx_get_lock, flags);
}

2017-06-30 14:56:37

by James Smart

Subject: Re: [PATCH 1/2] scsi: lpfc: spin_lock_irq() is not nestable

On 6/30/2017 1:02 AM, Dan Carpenter wrote:
> We're calling spin_lock_irq() multiple times, the problem is that on the
> first spin_unlock_irq() then we will re-enable IRQs and we don't want
> that.
>
> Fixes: 966bb5b71196 ("scsi: lpfc: Break up IO ctx list into a separate get and put list")
> Signed-off-by: Dan Carpenter <[email protected]>
>

looks good.

Signed-off-By: James Smart <[email protected]>

2017-07-01 21:09:57

by Martin K. Petersen

Subject: Re: [PATCH 1/2] scsi: lpfc: spin_lock_irq() is not nestable

Dan,

> We're calling spin_lock_irq() multiple times, the problem is that on
> the first spin_unlock_irq() then we will re-enable IRQs and we don't
> want that.

Applied to 4.13/scsi-queue.

--
Martin K. Petersen Oracle Linux Engineering