diff mbox series

ath10k: disable cpuidle during downloading firmware.

Message ID 20191101054035.42101-1-ikjn@chromium.org (mailing list archive)
State New, archived
Headers show
Series ath10k: disable cpuidle during downloading firmware. | expand

Commit Message

Ikjoon Jang Nov. 1, 2019, 5:40 a.m. UTC
Downloading ath10k firmware needs a large number of IOs and
cpuidle's miss predictions make it worse. In the worst case,
resume time can be three times longer than the average on sdio.

This patch disables cpuidle during firmware downloading by
applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().

Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
---
 drivers/net/wireless/ath/ath10k/core.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

Comments

Kalle Valo Nov. 6, 2019, 6:22 p.m. UTC | #1
Ikjoon Jang <ikjn@chromium.org> writes:

> Downloading ath10k firmware needs a large number of IOs and
> cpuidle's miss predictions make it worse. In the worst case,
> resume time can be three times longer than the average on sdio.
>
> This patch disables cpuidle during firmware downloading by
> applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().
>
> Signed-off-by: Ikjoon Jang <ikjn@chromium.org>

On what hardware and firmware versions did you test this? I'll add that
to the commit log.

https://wireless.wiki.kernel.org/en/users/drivers/ath10k/submittingpatches#guidelines
Tom Psyborg Nov. 6, 2019, 6:33 p.m. UTC | #2
On 06/11/2019, Kalle Valo <kvalo@codeaurora.org> wrote:
> Ikjoon Jang <ikjn@chromium.org> writes:
>
>> Downloading ath10k firmware needs a large number of IOs and
>> cpuidle's miss predictions make it worse. In the worst case,
>> resume time can be three times longer than the average on sdio.
>>
>> This patch disables cpuidle during firmware downloading by
>> applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().
>>
>> Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
>
> On what hardware and firmware versions did you test this? I'll add that
> to the commit log.
>
> https://wireless.wiki.kernel.org/en/users/drivers/ath10k/submittingpatches#guidelines
>
> --
> https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
>
> _______________________________________________
> ath10k mailing list
> ath10k@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/ath10k
>

Hi

I've tested this on QCA9880. No issues during firmware download.
Ikjoon Jang Nov. 7, 2019, 4:16 a.m. UTC | #3
On Thu, Nov 7, 2019 at 2:23 AM Kalle Valo <kvalo@codeaurora.org> wrote:
>
> Ikjoon Jang <ikjn@chromium.org> writes:
>
> > Downloading ath10k firmware needs a large number of IOs and
> > cpuidle's miss predictions make it worse. In the worst case,
> > resume time can be three times longer than the average on sdio.
> >
> > This patch disables cpuidle during firmware downloading by
> > applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().
> >
> > Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
>
> On what hardware and firmware versions did you test this? I'll add that
> to the commit log.
>
> https://wireless.wiki.kernel.org/en/users/drivers/ath10k/submittingpatches#guidelines

Thank you for sharing it.
It's QCA6174 hw3.2 SDIO WLAN.RMH.4.4.1-00029
on ARMv8 multi cluster platform.

>
> --
> https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
Kalle Valo Nov. 8, 2019, 8:31 a.m. UTC | #4
Tom Psyborg <pozega.tomislav@gmail.com> writes:

> On 06/11/2019, Kalle Valo <kvalo@codeaurora.org> wrote:
>> Ikjoon Jang <ikjn@chromium.org> writes:
>>
>>> Downloading ath10k firmware needs a large number of IOs and
>>> cpuidle's miss predictions make it worse. In the worst case,
>>> resume time can be three times longer than the average on sdio.
>>>
>>> This patch disables cpuidle during firmware downloading by
>>> applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().
>>>
>>> Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
>>
>> On what hardware and firmware versions did you test this? I'll add that
>> to the commit log.
>
> I've tested this on QCA9880. No issues during firmware download.

Great, thanks for testing.
Kalle Valo Nov. 8, 2019, 8:32 a.m. UTC | #5
Ikjoon Jang <ikjn@chromium.org> writes:

> On Thu, Nov 7, 2019 at 2:23 AM Kalle Valo <kvalo@codeaurora.org> wrote:
>>
>> Ikjoon Jang <ikjn@chromium.org> writes:
>>
>> > Downloading ath10k firmware needs a large number of IOs and
>> > cpuidle's miss predictions make it worse. In the worst case,
>> > resume time can be three times longer than the average on sdio.
>> >
>> > This patch disables cpuidle during firmware downloading by
>> > applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().
>> >
>> > Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
>>
>> On what hardware and firmware versions did you test this? I'll add that
>> to the commit log.
>>
>> https://wireless.wiki.kernel.org/en/users/drivers/ath10k/submittingpatches#guidelines
>
> Thank you for sharing it.
> It's QCA6174 hw3.2 SDIO WLAN.RMH.4.4.1-00029
> on ARMv8 multi cluster platform.

Thanks, I added that to the commit log.
Kalle Valo Nov. 8, 2019, 8:54 a.m. UTC | #6
Ikjoon Jang <ikjn@chromium.org> wrote:

> Downloading ath10k firmware needs a large number of IOs and
> cpuidle's miss predictions make it worse. In the worst case,
> resume time can be three times longer than the average on sdio.
> 
> This patch disables cpuidle during firmware downloading by
> applying PM_QOS_CPU_DMA_LATENCY in ath10k_download_fw().
> 
> Tested-on: QCA9880
> Tested-on: QCA6174 hw3.2 SDIO WLAN.RMH.4.4.1-00029
> 
> Signed-off-by: Ikjoon Jang <ikjn@chromium.org>
> Signed-off-by: Kalle Valo <kvalo@codeaurora.org>

Patch applied to ath-next branch of ath.git, thanks.

3b58d6a599ba ath10k: disable cpuidle during downloading firmware
diff mbox series

Patch

diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
index 36c62d66c19e..4f76ba5d78a9 100644
--- a/drivers/net/wireless/ath/ath10k/core.c
+++ b/drivers/net/wireless/ath/ath10k/core.c
@@ -11,6 +11,7 @@ 
 #include <linux/property.h>
 #include <linux/dmi.h>
 #include <linux/ctype.h>
+#include <linux/pm_qos.h>
 #include <asm/byteorder.h>
 
 #include "core.h"
@@ -1027,6 +1028,7 @@  static int ath10k_download_fw(struct ath10k *ar)
 	u32 address, data_len;
 	const void *data;
 	int ret;
+	struct pm_qos_request latency_qos;
 
 	address = ar->hw_params.patch_load_addr;
 
@@ -1060,8 +1062,14 @@  static int ath10k_download_fw(struct ath10k *ar)
 			    ret);
 	}
 
-	return ath10k_bmi_fast_download(ar, address,
-					data, data_len);
+	memset(&latency_qos, 0, sizeof(latency_qos));
+	pm_qos_add_request(&latency_qos, PM_QOS_CPU_DMA_LATENCY, 0);
+
+	ret = ath10k_bmi_fast_download(ar, address, data, data_len);
+
+	pm_qos_remove_request(&latency_qos);
+
+	return ret;
 }
 
 void ath10k_core_free_board_files(struct ath10k *ar)