Hi Sasha,
please consider taking a small fix for this one (also useful for 4.15):
commit d3b9e8ad425cfd5b9116732e057f1b48e4d3bcb8
Author: Max Gurtovoy <m...@mellanox.com>
Date: Mon Mar 5 20:09:48 2018 +0200
RDMA/core: Reduce poll batch for direct cq polling
Fix warning limit for kernel stack consumption:
drivers/infiniband/core/cq.c: In function 'ib_process_cq_direct':
drivers/infiniband/core/cq.c:78:1: error: the frame size of 1032 bytes
is larger than 1024 bytes [-Werror=frame-larger-than=]
Using smaller ib_wc array on the stack brings us comfortably below that
limit again.
Fixes: 246d8b184c10 ("IB/cq: Don't force IB_POLL_DIRECT poll
context for ib_process_cq_direct")
Reported-by: Arnd Bergmann <a...@arndb.de>
Reviewed-by: Sergey Gorenko <serge...@mellanox.com>
Signed-off-by: Max Gurtovoy <m...@mellanox.com>
Signed-off-by: Leon Romanovsky <leo...@mellanox.com>
Reviewed-by: Bart Van Assche <bart.vanass...@wdc.com>
Acked-by: Arnd Bergmann <a...@arndb.de>
Signed-off-by: Jason Gunthorpe <j...@mellanox.com>
-Max.
On 4/9/2018 3:20 AM, Sasha Levin wrote:
From: Sagi Grimberg <s...@grimberg.me>
[ Upstream commit 246d8b184c100e8eb6b4e8c88f232c2ed2a4e672 ]
polling the completion queue directly does not interfere
with the existing polling logic, hence drop the requirement.
Be aware that running ib_process_cq_direct with non IB_POLL_DIRECT
CQ may trigger concurrent CQ processing.
This can be used for polling mode ULPs.
Cc: Bart Van Assche <bart.vanass...@wdc.com>
Reported-by: Steve Wise <sw...@opengridcomputing.com>
Signed-off-by: Sagi Grimberg <s...@grimberg.me>
[maxg: added wcs array argument to __ib_process_cq]
Signed-off-by: Max Gurtovoy <m...@mellanox.com>
Signed-off-by: Doug Ledford <dledf...@redhat.com>
Signed-off-by: Sasha Levin <alexander.le...@microsoft.com>
---
drivers/infiniband/core/cq.c | 23 +++++++++++++----------
1 file changed, 13 insertions(+), 10 deletions(-)
diff --git a/drivers/infiniband/core/cq.c b/drivers/infiniband/core/cq.c
index f2ae75fa3128..c8c5a5a7f433 100644
--- a/drivers/infiniband/core/cq.c
+++ b/drivers/infiniband/core/cq.c
@@ -25,9 +25,10 @@
#define IB_POLL_FLAGS \
(IB_CQ_NEXT_COMP | IB_CQ_REPORT_MISSED_EVENTS)
-static int __ib_process_cq(struct ib_cq *cq, int budget)
+static int __ib_process_cq(struct ib_cq *cq, int budget, struct ib_wc *poll_wc)
{
int i, n, completed = 0;
+ struct ib_wc *wcs = poll_wc ? : cq->wc;
/*
* budget might be (-1) if the caller does not
@@ -35,9 +36,9 @@ static int __ib_process_cq(struct ib_cq *cq, int budget)
* minimum here.
*/
while ((n = ib_poll_cq(cq, min_t(u32, IB_POLL_BATCH,
- budget - completed), cq->wc)) > 0) {
+ budget - completed), wcs)) > 0) {
for (i = 0; i < n; i++) {
- struct ib_wc *wc = &cq->wc[i];
+ struct ib_wc *wc = &wcs[i];
if (wc->wr_cqe)
wc->wr_cqe->done(cq, wc);
@@ -60,18 +61,20 @@ static int __ib_process_cq(struct ib_cq *cq, int budget)
* @cq: CQ to process
* @budget: number of CQEs to poll for
*
- * This function is used to process all outstanding CQ entries on a
- * %IB_POLL_DIRECT CQ. It does not offload CQ processing to a different
- * context and does not ask for completion interrupts from the HCA.
+ * This function is used to process all outstanding CQ entries.
+ * It does not offload CQ processing to a different context and does
+ * not ask for completion interrupts from the HCA.
+ * Using direct processing on CQ with non IB_POLL_DIRECT type may trigger
+ * concurrent processing.
*
* Note: do not pass -1 as %budget unless it is guaranteed that the number
* of completions that will be processed is small.
*/
int ib_process_cq_direct(struct ib_cq *cq, int budget)
{
- WARN_ON_ONCE(cq->poll_ctx != IB_POLL_DIRECT);
+ struct ib_wc wcs[IB_POLL_BATCH];
- return __ib_process_cq(cq, budget);
+ return __ib_process_cq(cq, budget, wcs);
}
EXPORT_SYMBOL(ib_process_cq_direct);
@@ -85,7 +88,7 @@ static int ib_poll_handler(struct irq_poll *iop, int budget)
struct ib_cq *cq = container_of(iop, struct ib_cq, iop);
int completed;
- completed = __ib_process_cq(cq, budget);
+ completed = __ib_process_cq(cq, budget, NULL);
if (completed < budget) {
irq_poll_complete(&cq->iop);
if (ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
@@ -105,7 +108,7 @@ static void ib_cq_poll_work(struct work_struct *work)
struct ib_cq *cq = container_of(work, struct ib_cq, work);
int completed;
- completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE);
+ completed = __ib_process_cq(cq, IB_POLL_BUDGET_WORKQUEUE, NULL);
if (completed >= IB_POLL_BUDGET_WORKQUEUE ||
ib_req_notify_cq(cq, IB_POLL_FLAGS) > 0)
queue_work(ib_comp_wq, &cq->work);
