[Performance regression] BCM4359/9 on S905X2
Neil Armstrong
neil.armstrong at linaro.org
Thu Apr 6 10:41:55 PDT 2023
On 06/04/2023 19:07, Christian Hewitt wrote:
>> On 6 Apr 2023, at 8:35 pm, Marc Gonzalez <marc.w.gonzalez at free.fr> wrote:
>>
>> On 04/04/2023 23:06, Martin Blumenstingl wrote:
>>
>>> Hi Marc,
>>
>> Hello Martin :)
>>
>>> On Tue, Apr 4, 2023 at 6:09 PM Marc Gonzalez wrote:
>>>
>>>> bus-width = <4>;
>>>> cap-sd-highspeed;
>>>> cap-mmc-highspeed;
>>>> max-frequency = <100000000>;
>>>
>>> I would start by comparing the bus mode. You can get it from
>>> /sys/kernel/debug/mmc2/ios
>>> On the vendor kernel it should be in /sys/kernel/debug/sdio/ios (or
>>> similar, I'm typing this from the top of my head).
>>> It will give you insights on the clock and timing that has been
>>> negotiated between the host and card.
>>>
>>> From this information you can get the maximum bus speed, e.g. from [0]
>>> Please note that any card will add overhead for communication, so bus
>>> speed will not be equal to wifi throughput.
>>
>> For the vendor kernel:
>>
>> /sys/kernel/debug/sdio/sdio:0001/state:0x00000001
>> /sys/kernel/debug/sdio/clock:200000000
>> /sys/kernel/debug/sdio/ios:clock: 200000000 Hz
>> /sys/kernel/debug/sdio/ios:actual clock: 199999997 Hz
>> /sys/kernel/debug/sdio/ios:vdd: 21 (3.3 ~ 3.4 V)
>> /sys/kernel/debug/sdio/ios:bus mode: 2 (push-pull)
>> /sys/kernel/debug/sdio/ios:chip select: 0 (don't care)
>> /sys/kernel/debug/sdio/ios:power mode: 1 (up)
>> /sys/kernel/debug/sdio/ios:bus width: 2 (4 bits)
>> /sys/kernel/debug/sdio/ios:timing spec: 6 (sd uhs SDR104)
>> /sys/kernel/debug/sdio/ios:signal voltage: 1 (1.80 V)
>> /sys/kernel/debug/sdio/ios:driver type: 0 (driver type B)
>>
>>
>> For mainline:
>>
>> /sys/kernel/debug/mmc2/mmc2:0001/state:0x00000001
>> /sys/kernel/debug/mmc2/err_stats:# Command Timeout Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Command CRC Errors Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Data Timeout Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Data CRC Errors Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Auto-Cmd Error Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# ADMA Error Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Tuning Error Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ RED Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ GCE Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ ICCE Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# Request Timedout: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ Request Timedout: 0
>> /sys/kernel/debug/mmc2/err_stats:# ICE Config Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# Controller Timedout errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# Unexpected IRQ errors: 0
>> /sys/kernel/debug/mmc2/err_state:0
>> /sys/kernel/debug/mmc2/clock:100000000
>> /sys/kernel/debug/mmc2/caps2:0x00040000
>> /sys/kernel/debug/mmc2/caps:0x40040105
>> /sys/kernel/debug/mmc2/ios:clock: 100000000 Hz
>> /sys/kernel/debug/mmc2/ios:actual clock: 99999999 Hz
>> /sys/kernel/debug/mmc2/ios:vdd: 21 (3.3 ~ 3.4 V)
>> /sys/kernel/debug/mmc2/ios:bus mode: 2 (push-pull)
>> /sys/kernel/debug/mmc2/ios:chip select: 0 (don't care)
>> /sys/kernel/debug/mmc2/ios:power mode: 2 (on)
>> /sys/kernel/debug/mmc2/ios:bus width: 2 (4 bits)
>> /sys/kernel/debug/mmc2/ios:timing spec: 5 (sd uhs SDR50)
>> /sys/kernel/debug/mmc2/ios:signal voltage: 1 (1.80 V)
>> /sys/kernel/debug/mmc2/ios:driver type: 0 (driver type B)
>>
>>
>> The clock is indeed running twice as fast on the vendor system.
>> And in SDR104 vs SDR50 mode.
>>
>> Adjusting mainline device tree...
>>
>>
>> /sys/kernel/debug/mmc2/mmc2:0001/state:0x00000001
>> /sys/kernel/debug/mmc2/err_stats:# Command Timeout Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Command CRC Errors Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Data Timeout Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Data CRC Errors Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Auto-Cmd Error Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# ADMA Error Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# Tuning Error Occurred: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ RED Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ GCE Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ ICCE Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# Request Timedout: 0
>> /sys/kernel/debug/mmc2/err_stats:# CMDQ Request Timedout: 0
>> /sys/kernel/debug/mmc2/err_stats:# ICE Config Errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# Controller Timedout errors: 0
>> /sys/kernel/debug/mmc2/err_stats:# Unexpected IRQ errors: 0
>> /sys/kernel/debug/mmc2/err_state:0
>> /sys/kernel/debug/mmc2/clock:200000000
>> /sys/kernel/debug/mmc2/caps2:0x00040000
>> /sys/kernel/debug/mmc2/caps:0x40080105
>> /sys/kernel/debug/mmc2/ios:clock: 200000000 Hz
>> /sys/kernel/debug/mmc2/ios:actual clock: 199999997 Hz
>> /sys/kernel/debug/mmc2/ios:vdd: 21 (3.3 ~ 3.4 V)
>> /sys/kernel/debug/mmc2/ios:bus mode: 2 (push-pull)
>> /sys/kernel/debug/mmc2/ios:chip select: 0 (don't care)
>> /sys/kernel/debug/mmc2/ios:power mode: 2 (on)
>> /sys/kernel/debug/mmc2/ios:bus width: 2 (4 bits)
>> /sys/kernel/debug/mmc2/ios:timing spec: 6 (sd uhs SDR104)
>> /sys/kernel/debug/mmc2/ios:signal voltage: 1 (1.80 V)
>> /sys/kernel/debug/mmc2/ios:driver type: 0 (driver type B)
>>
>>
>> # curl -o /dev/null http://192.168.1.254:8095/fixed/1G
>> % Total % Received % Xferd Average Speed Time Time Time Current
>> Dload Upload Total Spent Left Speed
>> 100 1024M 100 1024M 0 0 9747k 0 0:01:47 0:01:47 --:--:-- 9544k
>>
>> With this new setting, WiFi throughput increases 20%
>> (from 60 Mbps to 75 Mbps).
>>
>> (Uggg, I've been using 10^9 for the amount transferred,
>> but it looks to be 2^30 actually. Absolute speeds are
>> actually 7% higher, but doesn't change the conclusion)
>>
>> I also noticed that I reported 11 MB/s (88 Mbps) at the beginning
>> of this thread. This would point to a performance regression caused
>> by my defconfig & device tree changes :(
>>
>>
>>>> vendor DTS has the following child node:
>>>>
>>>> sdio {
>>>> pinname = "sdio";
>>>> ocr_avail = <0x200080>; /**VDD voltage 3.3 ~ 3.4 */
>>>> /* max_req_size = <0x20000>; */ /**128KB*/
>>>> max_req_size = <0x400>;
>>>> card_type = <3>;
>>>> /* 3:sdio device(ie:sdio-wifi),
>>>> * 4:SD combo (IO+mem) card
>>>> */
>>>> dmode = "pio";
>>>> };
>>>>
>>>> Maybe the vendor kernel uses the above information to "boost"
>>>> the performance of the SDIO-based WiFi adapter?
>>>
>>> PIO is also what we support upstream with the
>>> amlogic,dram-access-quirk; (which is enabled for &sd_emmc_a).
>>> This suggests that the pinctrl trick that Neil mentioned is not used here.
>>>
>>> I assume that the wifi driver on the vendor kernel is the brcmdhd (out
>>> of tree) driver, while mainline uses brcmfmac.
>>
>> Correct.
>> hardware/wifi/broadcom/drivers/ap6xxx/bcmdhd.100.10.545.x
>> vs
>> drivers/net/wireless/broadcom/brcm80211/brcmfmac
>>
>>> One idea that comes to my mind is to enable the
>>> amlogic,dram-access-quirk; (and use pio mode on the vendor kernel) for
>>> eMMC or SD card and then compare read/write speeds. If they are
>>> similar-ish then the wifi performance difference is likely caused by
>>> the wifi driver (or in the opposite case: if read/write speeds on
>>> mainline with amlogic,dram-access-quirk; perform worse than the vendor
>>> kernel with pio mode then it's likely that it's a meson-gx-mmc driver
>>> limitation).
>>
>> Thanks for the suggestion.
>>
>> Are you aware whether someone has tried running the bcmdhd driver
>> on mainline?
>
> https://gitlab.baylibre.com/baylibre/amlogic/atv/linux/-/commits/narmstrong/v5.1/aml/integ-5.1-bcmdhd
>
> IIRC not long after this ^ support for BCM4359 was added to brcmfmac so
> using bcmdhd wasn’t pursued. Neil?
BCMDHD is used in the Yukawa Android kernel, e.g in 6.1:
https://android.googlesource.com/kernel/hikey-linaro/+log/refs/heads/android-amlogic-bmeson-6.1/drivers/net/wireless/bcmdhd
and FWs:
https://android.googlesource.com/device/amlogic/yukawa/+/refs/heads/master/binaries/bt-wifi-firmware/
Neil
>
> Christian
>
>> Thanks for the test suggestion.
>>
>>> From my work on the rtw88 SDIO wifi driver I can say that the
>>> meson-gx-mmc driver can push (TX direction) at least 120Mbit/s.
>>> I understand that this is half of what you get with the vendor kernel
>>> - and that this is the wrong direction (you're testing RX while I'm
>>> testing TX).
>>> The point that I want to get across is: I think nobody has the one
>>> answer why wifi performance is lower (personally I'm happy with having
>>> it work at all, performance is second).
>>> So it'll be a process to find the reason, and I think it requires
>>> being creative due to the large amounts of different code (MMC driver,
>>> wifi driver, additional patches...) between mainline and the vendor
>>> kernel.
>>
>> I really appreciate you sharing this insight.
>>
>> Regards.
>>
>
More information about the linux-amlogic
mailing list