From: Henry Orosco <henry.orosco-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
To: dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
e1000-rdma-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org,
Shiraz Saleem
<shiraz.saleem-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>,
Henry Orosco
<henry.orosco-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Subject: [PATCH 4/7] i40iw: Improve CQP timeout logic
Date: Mon, 7 Aug 2017 16:32:00 -0500 [thread overview]
Message-ID: <20170807213203.30588-5-henry.orosco@intel.com> (raw)
In-Reply-To: <20170807213203.30588-1-henry.orosco-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
From: Shiraz Saleem <shiraz.saleem-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
The current timeout logic for Control Queue-Pair (CQP) OPs
does not take into account whether CQP makes progress but
rather blindly waits for a large timeout value, 100000 jiffies
for the completion event. Improve this by setting the timeout
based on whether the CQP is making progress or not. If the CQP
is hung, the timeout will happen sooner, in 5000 jiffies. Each
time the CQP progress is detetcted, the timeout extends by 5000
jiffies.
Signed-off-by: Shiraz Saleem <shiraz.saleem-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Christopher N Bednarz <christopher.n.bednarz-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Henry Orosco <henry.orosco-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/infiniband/hw/i40iw/i40iw_ctrl.c | 11 +++++++++++
drivers/infiniband/hw/i40iw/i40iw_p.h | 14 +++++++++-----
drivers/infiniband/hw/i40iw/i40iw_type.h | 5 +++++
drivers/infiniband/hw/i40iw/i40iw_utils.c | 22 ++++++++++++++--------
4 files changed, 39 insertions(+), 13 deletions(-)
diff --git a/drivers/infiniband/hw/i40iw/i40iw_ctrl.c b/drivers/infiniband/hw/i40iw/i40iw_ctrl.c
index 5769080..b55620d 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_ctrl.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_ctrl.c
@@ -54,6 +54,17 @@ static inline void i40iw_insert_wqe_hdr(u64 *wqe, u64 header)
set_64bit_val(wqe, 24, header);
}
+void i40iw_check_cqp_progress(struct i40iw_cqp_timeout *cqp_timeout, struct i40iw_sc_dev *dev)
+{
+ if (cqp_timeout->compl_cqp_cmds != dev->cqp_cmd_stats[OP_COMPLETED_COMMANDS]) {
+ cqp_timeout->compl_cqp_cmds = dev->cqp_cmd_stats[OP_COMPLETED_COMMANDS];
+ cqp_timeout->count = 0;
+ } else {
+ if (dev->cqp_cmd_stats[OP_REQUESTED_COMMANDS] != cqp_timeout->compl_cqp_cmds)
+ cqp_timeout->count++;
+ }
+}
+
/**
* i40iw_get_cqp_reg_info - get head and tail for cqp using registers
* @cqp: struct for cqp hw
diff --git a/drivers/infiniband/hw/i40iw/i40iw_p.h b/drivers/infiniband/hw/i40iw/i40iw_p.h
index 28a92fe..e217a12 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_p.h
+++ b/drivers/infiniband/hw/i40iw/i40iw_p.h
@@ -35,11 +35,13 @@
#ifndef I40IW_P_H
#define I40IW_P_H
-#define PAUSE_TIMER_VALUE 0xFFFF
-#define REFRESH_THRESHOLD 0x7FFF
-#define HIGH_THRESHOLD 0x800
-#define LOW_THRESHOLD 0x200
-#define ALL_TC2PFC 0xFF
+#define PAUSE_TIMER_VALUE 0xFFFF
+#define REFRESH_THRESHOLD 0x7FFF
+#define HIGH_THRESHOLD 0x800
+#define LOW_THRESHOLD 0x200
+#define ALL_TC2PFC 0xFF
+#define CQP_COMPL_WAIT_TIME 0x3E8
+#define CQP_TIMEOUT_THRESHOLD 5
void i40iw_debug_buf(struct i40iw_sc_dev *dev, enum i40iw_debug_flag mask,
char *desc, u64 *buf, u32 size);
@@ -51,6 +53,8 @@ void i40iw_sc_cqp_post_sq(struct i40iw_sc_cqp *cqp);
u64 *i40iw_sc_cqp_get_next_send_wqe(struct i40iw_sc_cqp *cqp, u64 scratch);
+void i40iw_check_cqp_progress(struct i40iw_cqp_timeout *cqp_timeout, struct i40iw_sc_dev *dev);
+
enum i40iw_status_code i40iw_sc_mr_fast_register(struct i40iw_sc_qp *qp,
struct i40iw_fast_reg_stag_info *info,
bool post_sq);
diff --git a/drivers/infiniband/hw/i40iw/i40iw_type.h b/drivers/infiniband/hw/i40iw/i40iw_type.h
index 959ec81..63118f6 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_type.h
+++ b/drivers/infiniband/hw/i40iw/i40iw_type.h
@@ -1345,4 +1345,9 @@ struct i40iw_virtchnl_work_info {
void *worker_vf_dev;
};
+struct i40iw_cqp_timeout {
+ u64 compl_cqp_cmds;
+ u8 count;
+};
+
#endif
diff --git a/drivers/infiniband/hw/i40iw/i40iw_utils.c b/drivers/infiniband/hw/i40iw/i40iw_utils.c
index e311ec5..62f1f45 100644
--- a/drivers/infiniband/hw/i40iw/i40iw_utils.c
+++ b/drivers/infiniband/hw/i40iw/i40iw_utils.c
@@ -445,23 +445,29 @@ static int i40iw_wait_event(struct i40iw_device *iwdev,
{
struct cqp_commands_info *info = &cqp_request->info;
struct i40iw_cqp *iwcqp = &iwdev->cqp;
+ struct i40iw_cqp_timeout cqp_timeout;
bool cqp_error = false;
int err_code = 0;
- int timeout_ret = 0;
+ memset(&cqp_timeout, 0, sizeof(cqp_timeout));
+ cqp_timeout.compl_cqp_cmds = iwdev->sc_dev.cqp_cmd_stats[OP_COMPLETED_COMMANDS];
+ do {
+ if (wait_event_timeout(cqp_request->waitq,
+ cqp_request->request_done, CQP_COMPL_WAIT_TIME))
+ break;
- timeout_ret = wait_event_timeout(cqp_request->waitq,
- cqp_request->request_done,
- I40IW_EVENT_TIMEOUT);
- if (!timeout_ret) {
- i40iw_pr_err("error cqp command 0x%x timed out ret = %d\n",
- info->cqp_cmd, timeout_ret);
+ i40iw_check_cqp_progress(&cqp_timeout, &iwdev->sc_dev);
+
+ if (cqp_timeout.count < CQP_TIMEOUT_THRESHOLD)
+ continue;
+
+ i40iw_pr_err("error cqp command 0x%x timed out", info->cqp_cmd);
err_code = -ETIME;
if (!iwdev->reset) {
iwdev->reset = true;
i40iw_request_reset(iwdev);
}
goto done;
- }
+ } while (1);
cqp_error = cqp_request->compl_info.error;
if (cqp_error) {
i40iw_pr_err("error cqp command 0x%x completion maj = 0x%x min=0x%x\n",
--
2.8.3
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2017-08-07 21:32 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-08-07 21:31 [pull request][PATCH 0/7] i40iw: Fixes for i40iw Henry Orosco
[not found] ` <20170807213203.30588-1-henry.orosco-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
2017-08-07 21:31 ` [PATCH 1/7] i40iw: Fix parsing of query/commit FPM buffers Henry Orosco
2017-08-07 21:31 ` [PATCH 2/7] i40iw: Ignore first SQ wqe_index if it shows up as a CQE Henry Orosco
2017-08-07 21:31 ` [PATCH 3/7] i40iw: Correct variable names Henry Orosco
2017-08-07 21:32 ` Henry Orosco [this message]
2017-08-07 21:32 ` [PATCH 5/7] i40iw: Fix typecast of tcp_seq_num Henry Orosco
2017-08-07 21:32 ` [PATCH 6/7] i40iw: Use correct alignment for CQ0 memory Henry Orosco
2017-08-07 21:32 ` [PATCH 7/7] i40iw: Fix potential fcn_id_array out of bounds Henry Orosco
2017-08-08 14:30 ` [pull request][PATCH 0/7] i40iw: Fixes for i40iw Henry Orosco
[not found] ` <20170808143022.GA35328-ZmvEvTIhtuUfyugFOqMDN/ooFf0ArEBIu+b9c/7xato@public.gmane.org>
2017-08-08 16:47 ` Doug Ledford
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170807213203.30588-5-henry.orosco@intel.com \
--to=henry.orosco-ral2jqcrhueavxtiumwx3w@public.gmane.org \
--cc=dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
--cc=e1000-rdma-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=shiraz.saleem-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox