[PATCH v3 0/4] rt2x00 random fixes

[PATCH v3 0/4] rt2x00 random fixes

[Stanislaw Gruszka]

This is new version of:
https://lore.kernel.org/linux-wireless/1545318971-28351-1-git-send-email-sgruszka@redhat.com/

v3: 
 - move printing errors routines to cfg80211 (patch 1)
 - handle ETIMEDOUT error and decrees number of error checks (patch 3)

Stanislaw Gruszka (4):
  cfg80211: add ratelimited variants of err and warn
  rt2x00: use ratelimited variants dev_warn/dev_err
  rt2x00: check number of EPROTO errors
  rt2x00: do not print error when queue is full

 drivers/net/wireless/ralink/rt2x00/rt2x00.h      |  5 +++--
 drivers/net/wireless/ralink/rt2x00/rt2x00queue.c |  2 +-
 drivers/net/wireless/ralink/rt2x00/rt2x00usb.c   | 22 +++++++++++++++++++---
 include/net/cfg80211.h                           |  5 +++++
 4 files changed, 28 insertions(+), 6 deletions(-)

-- 
2.7.5

[PATCH v3 1/4] cfg80211: add ratelimited variants of err and warn

[Stanislaw Gruszka]

wiphy_{err,warn}_ratelimited will be used by rt2x00

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 include/net/cfg80211.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/net/cfg80211.h b/include/net/cfg80211.h
index bb307a11ee63..13bfeb712d36 100644
--- a/include/net/cfg80211.h
+++ b/include/net/cfg80211.h
@@ -7183,6 +7183,11 @@ void cfg80211_pmsr_complete(struct wireless_dev *wdev,
 #define wiphy_info(wiphy, format, args...)			\
 	dev_info(&(wiphy)->dev, format, ##args)
 
+#define wiphy_err_ratelimited(wiphy, format, args...)		\
+	dev_err_ratelimited(&(wiphy)->dev, format, ##args)
+#define wiphy_warn_ratelimited(wiphy, format, args...)		\
+	dev_warn_ratelimited(&(wiphy)->dev, format, ##args)
+
 #define wiphy_debug(wiphy, format, args...)			\
 	wiphy_printk(KERN_DEBUG, wiphy, format, ##args)
 
-- 
2.7.5

[PATCH v3 2/4] rt2x00: use ratelimited variants dev_warn/dev_err

[Stanislaw Gruszka]

As reported by Randy we can overwhelm logs on some USB error conditions.
To avoid that use dev_warn_ratelimited() and dev_err_ratelimitd().

Reported-and-tested-by: Randy Oostdyk <linux-kernel@oostdyk.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 drivers/net/wireless/ralink/rt2x00/rt2x00.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/wireless/ralink/rt2x00/rt2x00.h b/drivers/net/wireless/ralink/rt2x00/rt2x00.h
index 4b1744e9fb78..5657495cf626 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2x00.h
+++ b/drivers/net/wireless/ralink/rt2x00/rt2x00.h
@@ -69,10 +69,10 @@
 	printk(KERN_ERR KBUILD_MODNAME ": %s: Error - " fmt,		\
 	       __func__, ##__VA_ARGS__)
 #define rt2x00_err(dev, fmt, ...)					\
-	wiphy_err((dev)->hw->wiphy, "%s: Error - " fmt,			\
+	wiphy_err_ratelimited((dev)->hw->wiphy, "%s: Error - " fmt,	\
 		  __func__, ##__VA_ARGS__)
 #define rt2x00_warn(dev, fmt, ...)					\
-	wiphy_warn((dev)->hw->wiphy, "%s: Warning - " fmt,		\
+	wiphy_warn_ratelimited((dev)->hw->wiphy, "%s: Warning - " fmt,	\
 		   __func__, ##__VA_ARGS__)
 #define rt2x00_info(dev, fmt, ...)					\
 	wiphy_info((dev)->hw->wiphy, "%s: Info - " fmt,			\
-- 
2.7.5

[PATCH v3 3/4] rt2x00: check number of EPROTO errors

[Stanislaw Gruszka]

Some USB host devices/drivers on some conditions can always return
EPROTO error on submitted URBs. That can cause infinity loop in the
rt2x00 driver.

Since we can have single EPROTO errors we can not mark as device as
removed to avoid infinity loop. However we can count consecutive
EPROTO errors and mark device as removed if get lot of it.
I choose number 10 as threshold.

Reported-and-tested-by: Randy Oostdyk <linux-kernel@oostdyk.com>
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 drivers/net/wireless/ralink/rt2x00/rt2x00.h    |  1 +
 drivers/net/wireless/ralink/rt2x00/rt2x00usb.c | 22 +++++++++++++++++++---
 2 files changed, 20 insertions(+), 3 deletions(-)

diff --git a/drivers/net/wireless/ralink/rt2x00/rt2x00.h b/drivers/net/wireless/ralink/rt2x00/rt2x00.h
index 5657495cf626..c1519c98ad41 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2x00.h
+++ b/drivers/net/wireless/ralink/rt2x00/rt2x00.h
@@ -1017,6 +1017,7 @@ struct rt2x00_dev {
 	unsigned int extra_tx_headroom;
 
 	struct usb_anchor *anchor;
+	unsigned int num_proto_errs;
 
 	/* Clock for System On Chip devices. */
 	struct clk *clk;
diff --git a/drivers/net/wireless/ralink/rt2x00/rt2x00usb.c b/drivers/net/wireless/ralink/rt2x00/rt2x00usb.c
index 086aad22743d..9cdd7f2c92b5 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2x00usb.c
+++ b/drivers/net/wireless/ralink/rt2x00/rt2x00usb.c
@@ -31,6 +31,22 @@
 #include "rt2x00.h"
 #include "rt2x00usb.h"
 
+static bool rt2x00usb_check_usb_error(struct rt2x00_dev *rt2x00dev, int status)
+{
+	if (status == -ENODEV || status == -ENOENT)
+		return true;
+
+	if (status == -EPROTO || status == -ETIMEDOUT)
+		rt2x00dev->num_proto_errs++;
+	else
+		rt2x00dev->num_proto_errs = 0;
+
+	if (rt2x00dev->num_proto_errs > 3)
+		return true;
+
+	return false;
+}
+
 /*
  * Interfacing with the HW.
  */
@@ -57,7 +73,7 @@ int rt2x00usb_vendor_request(struct rt2x00_dev *rt2x00dev,
 		if (status >= 0)
 			return 0;
 
-		if (status == -ENODEV || status == -ENOENT) {
+		if (rt2x00usb_check_usb_error(rt2x00dev, status)) {
 			/* Device has disappeared. */
 			clear_bit(DEVICE_STATE_PRESENT, &rt2x00dev->flags);
 			break;
@@ -321,7 +337,7 @@ static bool rt2x00usb_kick_tx_entry(struct queue_entry *entry, void *data)
 
 	status = usb_submit_urb(entry_priv->urb, GFP_ATOMIC);
 	if (status) {
-		if (status == -ENODEV || status == -ENOENT)
+		if (rt2x00usb_check_usb_error(rt2x00dev, status))
 			clear_bit(DEVICE_STATE_PRESENT, &rt2x00dev->flags);
 		set_bit(ENTRY_DATA_IO_FAILED, &entry->flags);
 		rt2x00lib_dmadone(entry);
@@ -410,7 +426,7 @@ static bool rt2x00usb_kick_rx_entry(struct queue_entry *entry, void *data)
 
 	status = usb_submit_urb(entry_priv->urb, GFP_ATOMIC);
 	if (status) {
-		if (status == -ENODEV || status == -ENOENT)
+		if (rt2x00usb_check_usb_error(rt2x00dev, status))
 			clear_bit(DEVICE_STATE_PRESENT, &rt2x00dev->flags);
 		set_bit(ENTRY_DATA_IO_FAILED, &entry->flags);
 		rt2x00lib_dmadone(entry);
-- 
2.7.5

[PATCH v3 4/4] rt2x00: do not print error when queue is full

[Stanislaw Gruszka]

For unknown reasons printk() on some context can cause CPU hung on
embedded MT7620 AP/router MIPS platforms. What can result on wifi
disconnects.

This patch move queue full messages to debug level what is consistent
with other mac80211 drivers which drop packet silently if tx queue is
full. This make MT7620 OpenWRT routers more stable, what was reported
by various users.

Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
---
 drivers/net/wireless/ralink/rt2x00/rt2x00queue.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/wireless/ralink/rt2x00/rt2x00queue.c b/drivers/net/wireless/ralink/rt2x00/rt2x00queue.c
index 92ddc19e7bf7..947fc8964e9a 100644
--- a/drivers/net/wireless/ralink/rt2x00/rt2x00queue.c
+++ b/drivers/net/wireless/ralink/rt2x00/rt2x00queue.c
@@ -671,7 +671,7 @@ int rt2x00queue_write_tx_frame(struct data_queue *queue, struct sk_buff *skb,
 	spin_lock(&queue->tx_lock);
 
 	if (unlikely(rt2x00queue_full(queue))) {
-		rt2x00_err(queue->rt2x00dev, "Dropping frame due to full tx queue %d\n",
+		rt2x00_dbg(queue->rt2x00dev, "Dropping frame due to full tx queue %d\n",
 			   queue->qid);
 		ret = -ENOBUFS;
 		goto out;
-- 
2.7.5

Re: [PATCH v3 4/4] rt2x00: do not print error when queue is full

[Tom Psyborg]

On 12/03/2019, Stanislaw Gruszka <sgruszka@redhat.com> wrote:
> This make MT7620 OpenWRT routers more stable, what was reported
> by various users.

Confirmed.

Re: [PATCH v3 2/4] rt2x00: use ratelimited variants dev_warn/dev_err

[Kalle Valo]

Stanislaw Gruszka <sgruszka@redhat.com> wrote:

> As reported by Randy we can overwhelm logs on some USB error conditions.
> To avoid that use dev_warn_ratelimited() and dev_err_ratelimitd().
> 
> Reported-and-tested-by: Randy Oostdyk <linux-kernel@oostdyk.com>
> Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>

3 patches applied to wireless-drivers-next.git, thanks.

bb3b18c92533 rt2x00: use ratelimited variants dev_warn/dev_err
e383c70474db rt2x00: check number of EPROTO errors
61a4e5ff0d72 rt2x00: do not print error when queue is full

-- 
https://patchwork.kernel.org/patch/10848959/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches