2013-10-17 10:07:57

by Stanislaw Gruszka

[permalink] [raw]
Subject: [PATCH 3.12] rt2800usb: slow down TX status polling

Polling TX statuses too frequently has two negative effects. First is
randomly peek CPU usage, causing overall system functioning delays.
Second bad effect is that device is not able to fill TX statuses in
H/W register on some workloads and we get lot of timeouts like below:

ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
ieee80211 phy4: rt2800usb_txdone: Warning - Got TX status for an empty queue 2, dropping

This not only cause flood of messages in dmesg, but also bad throughput,
since rate scaling algorithm can not work optimally.

In the future, we should probably make polling interval be adjusted
automatically, but for now just increase values, this make mentioned
problems gone.

Resolve:
https://bugzilla.kernel.org/show_bug.cgi?id=62781

Cc: [email protected]
Signed-off-by: Stanislaw Gruszka <[email protected]>
---
drivers/net/wireless/rt2x00/rt2800usb.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/net/wireless/rt2x00/rt2800usb.c b/drivers/net/wireless/rt2x00/rt2800usb.c
index 96677ce5..e095e61 100644
--- a/drivers/net/wireless/rt2x00/rt2800usb.c
+++ b/drivers/net/wireless/rt2x00/rt2800usb.c
@@ -176,8 +176,8 @@ static bool rt2800usb_tx_sta_fifo_read_completed(struct rt2x00_dev *rt2x00dev,
queue_work(rt2x00dev->workqueue, &rt2x00dev->txdone_work);

if (rt2800usb_txstatus_pending(rt2x00dev)) {
- /* Read register after 250 us */
- hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 250000),
+ /* Read register after 1 ms */
+ hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 1000000),
HRTIMER_MODE_REL);
return false;
}
@@ -202,8 +202,8 @@ static void rt2800usb_async_read_tx_status(struct rt2x00_dev *rt2x00dev)
if (test_and_set_bit(TX_STATUS_READING, &rt2x00dev->flags))
return;

- /* Read TX_STA_FIFO register after 500 us */
- hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 500000),
+ /* Read TX_STA_FIFO register after 2 ms */
+ hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 2000000),
HRTIMER_MODE_REL);
}

--
1.8.3.1



2013-10-17 14:39:11

by Larry Finger

[permalink] [raw]
Subject: Re: [PATCH 3.12] rt2800usb: slow down TX status polling

On 10/17/2013 05:04 AM, Stanislaw Gruszka wrote:
> Polling TX statuses too frequently has two negative effects. First is
> randomly peek CPU usage, causing overall system functioning delays.
> Second bad effect is that device is not able to fill TX statuses in
> H/W register on some workloads and we get lot of timeouts like below:
>
> ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
> ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
> ieee80211 phy4: rt2800usb_txdone: Warning - Got TX status for an empty queue 2, dropping
>
> This not only cause flood of messages in dmesg, but also bad throughput,
> since rate scaling algorithm can not work optimally.
>
> In the future, we should probably make polling interval be adjusted
> automatically, but for now just increase values, this make mentioned
> problems gone.
>
> Resolve:
> https://bugzilla.kernel.org/show_bug.cgi?id=62781
>
> Cc: [email protected]
> Signed-off-by: Stanislaw Gruszka <[email protected]>
> ---
> drivers/net/wireless/rt2x00/rt2800usb.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/wireless/rt2x00/rt2800usb.c b/drivers/net/wireless/rt2x00/rt2800usb.c
> index 96677ce5..e095e61 100644
> --- a/drivers/net/wireless/rt2x00/rt2800usb.c
> +++ b/drivers/net/wireless/rt2x00/rt2800usb.c
> @@ -176,8 +176,8 @@ static bool rt2800usb_tx_sta_fifo_read_completed(struct rt2x00_dev *rt2x00dev,
> queue_work(rt2x00dev->workqueue, &rt2x00dev->txdone_work);
>
> if (rt2800usb_txstatus_pending(rt2x00dev)) {
> - /* Read register after 250 us */
> - hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 250000),
> + /* Read register after 1 ms */
> + hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 1000000),
> HRTIMER_MODE_REL);
> return false;
> }
> @@ -202,8 +202,8 @@ static void rt2800usb_async_read_tx_status(struct rt2x00_dev *rt2x00dev)
> if (test_and_set_bit(TX_STATUS_READING, &rt2x00dev->flags))
> return;
>
> - /* Read TX_STA_FIFO register after 500 us */
> - hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 500000),
> + /* Read TX_STA_FIFO register after 2 ms */
> + hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 2000000),
> HRTIMER_MODE_REL);
> }

I suggest getting rid of the magic numbers as long as you are making this
change. A single define could handle the delay time for the two cases.

Larry


2013-10-19 08:33:08

by Gertjan van Wingerde

[permalink] [raw]
Subject: Re: [PATCH 3.12] rt2800usb: slow down TX status polling

On 10/18/13 11:42, Stanislaw Gruszka wrote:
> On Thu, Oct 17, 2013 at 09:39:06AM -0500, Larry Finger wrote:
>> I suggest getting rid of the magic numbers as long as you are making
>> this change. A single define could handle the delay time for the two
>> cases.
>
> Thanks for sugestion Larry, though I do not see clear benefit of
> introduce define since those magic numbers are just time values
> expressed in nano seconds. Anyway patch with define added below.
> John can pick it, if he thinks it is better.
>
> Stanislaw
> ---
> From 813e0bde7340bad7d3401c6aa2a3f8635ec49597 Mon Sep 17 00:00:00 2001
> From: Stanislaw Gruszka <[email protected]>
> Date: Fri, 18 Oct 2013 11:36:54 +0200
> Subject: [PATCH] rt2800usb: slow down TX status polling
>
> Polling TX statuses too frequently has two negative effects. First is
> randomly peek CPU usage, causing overall system functioning delays.
> Second bad effect is that device is not able to fill TX statuses in
> H/W register on some workloads and we get lot of timeouts like below:
>
> ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
> ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
> ieee80211 phy4: rt2800usb_txdone: Warning - Got TX status for an empty queue 2, dropping
>
> This not only cause flood of messages in dmesg, but also bad throughput,
> since rate scaling algorithm can not work optimally.
>
> In the future, we should probably make polling interval be adjusted
> automatically, but for now just increase values, this make mentioned
> problems gone.
>
> Resolve:
> https://bugzilla.kernel.org/show_bug.cgi?id=62781
>
> Cc: [email protected]
> Signed-off-by: Stanislaw Gruszka <[email protected]>

I don't care which version gets picked. In both cases:

Acked-by: Gertjan van Wingerde <[email protected]>

> ---
> drivers/net/wireless/rt2x00/rt2800usb.c | 12 ++++++++----
> 1 file changed, 8 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/wireless/rt2x00/rt2800usb.c b/drivers/net/wireless/rt2x00/rt2800usb.c
> index 96677ce5..997df03 100644
> --- a/drivers/net/wireless/rt2x00/rt2800usb.c
> +++ b/drivers/net/wireless/rt2x00/rt2800usb.c
> @@ -148,6 +148,8 @@ static bool rt2800usb_txstatus_timeout(struct rt2x00_dev *rt2x00dev)
> return false;
> }
>
> +#define TXSTATUS_READ_INTERVAL 1000000
> +
> static bool rt2800usb_tx_sta_fifo_read_completed(struct rt2x00_dev *rt2x00dev,
> int urb_status, u32 tx_status)
> {
> @@ -176,8 +178,9 @@ static bool rt2800usb_tx_sta_fifo_read_completed(struct rt2x00_dev *rt2x00dev,
> queue_work(rt2x00dev->workqueue, &rt2x00dev->txdone_work);
>
> if (rt2800usb_txstatus_pending(rt2x00dev)) {
> - /* Read register after 250 us */
> - hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 250000),
> + /* Read register after 1 ms */
> + hrtimer_start(&rt2x00dev->txstatus_timer,
> + ktime_set(0, TXSTATUS_READ_INTERVAL),
> HRTIMER_MODE_REL);
> return false;
> }
> @@ -202,8 +205,9 @@ static void rt2800usb_async_read_tx_status(struct rt2x00_dev *rt2x00dev)
> if (test_and_set_bit(TX_STATUS_READING, &rt2x00dev->flags))
> return;
>
> - /* Read TX_STA_FIFO register after 500 us */
> - hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 500000),
> + /* Read TX_STA_FIFO register after 2 ms */
> + hrtimer_start(&rt2x00dev->txstatus_timer,
> + ktime_set(0, 2*TXSTATUS_READ_INTERVAL),
> HRTIMER_MODE_REL);
> }
>
>


--
---
Gertjan

2013-10-18 09:46:17

by Stanislaw Gruszka

[permalink] [raw]
Subject: Re: [PATCH 3.12] rt2800usb: slow down TX status polling

On Thu, Oct 17, 2013 at 09:39:06AM -0500, Larry Finger wrote:
> I suggest getting rid of the magic numbers as long as you are making
> this change. A single define could handle the delay time for the two
> cases.

Thanks for sugestion Larry, though I do not see clear benefit of
introduce define since those magic numbers are just time values
expressed in nano seconds. Anyway patch with define added below.
John can pick it, if he thinks it is better.

Stanislaw
---
>From 813e0bde7340bad7d3401c6aa2a3f8635ec49597 Mon Sep 17 00:00:00 2001
From: Stanislaw Gruszka <[email protected]>
Date: Fri, 18 Oct 2013 11:36:54 +0200
Subject: [PATCH] rt2800usb: slow down TX status polling

Polling TX statuses too frequently has two negative effects. First is
randomly peek CPU usage, causing overall system functioning delays.
Second bad effect is that device is not able to fill TX statuses in
H/W register on some workloads and we get lot of timeouts like below:

ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7 in queue 2
ieee80211 phy4: rt2800usb_txdone: Warning - Got TX status for an empty queue 2, dropping

This not only cause flood of messages in dmesg, but also bad throughput,
since rate scaling algorithm can not work optimally.

In the future, we should probably make polling interval be adjusted
automatically, but for now just increase values, this make mentioned
problems gone.

Resolve:
https://bugzilla.kernel.org/show_bug.cgi?id=62781

Cc: [email protected]
Signed-off-by: Stanislaw Gruszka <[email protected]>
---
drivers/net/wireless/rt2x00/rt2800usb.c | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/drivers/net/wireless/rt2x00/rt2800usb.c b/drivers/net/wireless/rt2x00/rt2800usb.c
index 96677ce5..997df03 100644
--- a/drivers/net/wireless/rt2x00/rt2800usb.c
+++ b/drivers/net/wireless/rt2x00/rt2800usb.c
@@ -148,6 +148,8 @@ static bool rt2800usb_txstatus_timeout(struct rt2x00_dev *rt2x00dev)
return false;
}

+#define TXSTATUS_READ_INTERVAL 1000000
+
static bool rt2800usb_tx_sta_fifo_read_completed(struct rt2x00_dev *rt2x00dev,
int urb_status, u32 tx_status)
{
@@ -176,8 +178,9 @@ static bool rt2800usb_tx_sta_fifo_read_completed(struct rt2x00_dev *rt2x00dev,
queue_work(rt2x00dev->workqueue, &rt2x00dev->txdone_work);

if (rt2800usb_txstatus_pending(rt2x00dev)) {
- /* Read register after 250 us */
- hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 250000),
+ /* Read register after 1 ms */
+ hrtimer_start(&rt2x00dev->txstatus_timer,
+ ktime_set(0, TXSTATUS_READ_INTERVAL),
HRTIMER_MODE_REL);
return false;
}
@@ -202,8 +205,9 @@ static void rt2800usb_async_read_tx_status(struct rt2x00_dev *rt2x00dev)
if (test_and_set_bit(TX_STATUS_READING, &rt2x00dev->flags))
return;

- /* Read TX_STA_FIFO register after 500 us */
- hrtimer_start(&rt2x00dev->txstatus_timer, ktime_set(0, 500000),
+ /* Read TX_STA_FIFO register after 2 ms */
+ hrtimer_start(&rt2x00dev->txstatus_timer,
+ ktime_set(0, 2*TXSTATUS_READ_INTERVAL),
HRTIMER_MODE_REL);
}

--
1.8.3.1


2013-11-20 12:53:34

by Marc Dietrich

[permalink] [raw]
Subject: Re: [rt2x00-users] [PATCH 3.12] rt2800usb: slow down TX statuspolling

Am Samstag, 19. Oktober 2013, 10:33:06 schrieb Gertjan van Wingerde:
> On 10/18/13 11:42, Stanislaw Gruszka wrote:
> > On Thu, Oct 17, 2013 at 09:39:06AM -0500, Larry Finger wrote:
> >> I suggest getting rid of the magic numbers as long as you are making
> >> this change. A single define could handle the delay time for the two
> >> cases.
> >
> > Thanks for sugestion Larry, though I do not see clear benefit of
> > introduce define since those magic numbers are just time values
> > expressed in nano seconds. Anyway patch with define added below.
> > John can pick it, if he thinks it is better.
> >
> > Stanislaw
> > ---
> > From 813e0bde7340bad7d3401c6aa2a3f8635ec49597 Mon Sep 17 00:00:00 2001
> > From: Stanislaw Gruszka <[email protected]>
> > Date: Fri, 18 Oct 2013 11:36:54 +0200
> > Subject: [PATCH] rt2800usb: slow down TX status polling
> >
> > Polling TX statuses too frequently has two negative effects. First is
> > randomly peek CPU usage, causing overall system functioning delays.
> > Second bad effect is that device is not able to fill TX statuses in
> > H/W register on some workloads and we get lot of timeouts like below:
> >
> > ieee80211 phy4: rt2800usb_entry_txstatus_timeout: Warning - TX status
> > timeout for entry 7 in queue 2 ieee80211 phy4:
> > rt2800usb_entry_txstatus_timeout: Warning - TX status timeout for entry 7
> > in queue 2 ieee80211 phy4: rt2800usb_txdone: Warning - Got TX status for
> > an empty queue 2, dropping
> >
> > This not only cause flood of messages in dmesg, but also bad throughput,
> > since rate scaling algorithm can not work optimally.
> >
> > In the future, we should probably make polling interval be adjusted
> > automatically, but for now just increase values, this make mentioned
> > problems gone.
> >
> > Resolve:
> > https://bugzilla.kernel.org/show_bug.cgi?id=62781
> >
> > Cc: [email protected]
> > Signed-off-by: Stanislaw Gruszka <[email protected]>
>
> I don't care which version gets picked. In both cases:
>
> Acked-by: Gertjan van Wingerde <[email protected]>

did this ever hit a stable queue?

Marc


2013-11-20 13:14:58

by Stanislaw Gruszka

[permalink] [raw]
Subject: Re: [rt2x00-users] [PATCH 3.12] rt2800usb: slow down TX statuspolling

On Wed, Nov 20, 2013 at 01:53:29PM +0100, Marc Dietrich wrote:
> > > Resolve:
> > > https://bugzilla.kernel.org/show_bug.cgi?id=62781
> > >
> > > Cc: [email protected]
> > > Signed-off-by: Stanislaw Gruszka <[email protected]>
> >
> > I don't care which version gets picked. In both cases:
> >
> > Acked-by: Gertjan van Wingerde <[email protected]>
>
> did this ever hit a stable queue?

Not yet, it lend now in Linus' tree, should be queued to stable soon.

Stanislaw