Re: iwlwifi: Bluetooth and Wifi

[Emmanuel Grumbach <egrumbach@gmail.com>]

On 02/16/2014 06:25 PM, David John wrote:
> On Sunday 16 February 2014 12:07 PM, Emmanuel Grumbach wrote:
>>> Your patch "iwlwifi: mvm: BT Coex - disable BT when TXing probe request
>>> in scan" fixed wifi dropping on my 7260 with scan time outs but now my
>>> Bluetooth mouse freezes and unfreezes as it gets re-detected every few
>>> minutes (3.14-rc2+):
>>
>> All this patches does is to send the probe request of the scan in High
>> Prio so that it will take precedence over BT traffic in case of
>> collision. Since there shouldn't be that many probe request, I am
>> surprised that it kills your BT connection.
>> Also - are you 100% positive that this patch changed the behavior?
>> Have you tried to revert it and see what happens?
>> Also - what was the behavior before? What do you mean when you say
>> that it fixed your "wifi dropping with scan times out".
>> Can I have a log of this?
>>
> 
> With BT on, I used to get Wifi drops with messages similar to the following:
> 
> iwlwifi 0000    3:00.0: Error sending SCAN_REQUEST_CMD: time out after
> 2000ms.
> iwlwifi 0000    3:00.0: Current CMD queue read_ptr 125 write_ptr 127
> iwlwifi 0000    3:00.0: Scan failed! status 0x1 ret -110

Huh... This should *never* happen. We had such an issue but it was very
rare and should be fixed in 3.14:

commit b9439491055a18ee075614139abadfd74c1b887f

    iwlwifi: pcie: keep the NIC awake when commands are in flight

> 
> The only way to recover was to reload the iwlmvm module. I booted up the
> older stock Fedora 3.12 kernel I was using but couldn't reproduce this
> issue (with debug) even after a day of use, both Bluetooth and Wifi were
> fine. I'll try this again later.
> 
> I've attached debug dmesg for the current BT kill problem from
> 3.14-rc2+. Now I'm not sure if the patch I've referenced is the culprit,
> I'll try and narrow it down to a commit later when I have the time.

Actually, I am pretty sure it isn't - since thanks to the log you enable
we can clearly see that BT begins to complain when we are *not*
scanning. I just picked up one example:

[ 3723.246575] iwlwifi 0000:01:00.0: U iwl_mvm_rx_scan_complete Scan
complete: status=0x1 scanned channels=25

<<== HERE SCAN IS FINISHED ==>>

[ 3768.683762] usb 2-6: USB disconnect, device number 7
[ 3769.179405] usb 2-6: new full-speed USB device number 9 using xhci_hcd
[ 3769.344503] usb 2-6: New USB device found, idVendor=8087, idProduct=07dc
[ 3769.344514] usb 2-6: New USB device strings: Mfr=0, Product=0,
SerialNumber=0
[ 3769.360980] Bluetooth: hci0: read Intel version: 370710018002030d33
[ 3769.360988] Bluetooth: hci0: Intel device is already patched. patch
num: 33
[ 3794.222419] hid-generic 0005:046D:B010.0004: unknown main item tag 0x0
[ 3794.257065] input: Bluetooth Mouse M557 as
/devices/pci0000:00/0000:00:14.0/usb2/2-6/2-6:1.0/bluetooth/hci0/hci0:256/0005:046D:B010.0004/input/input16
[ 3794.257485] hid-generic 0005:046D:B010.0004: input,hidraw1: BLUETOOTH
HID v10.00 Mouse [Bluetooth Mouse M557] on 5c:51:4f:44:52:22

<<== HERE NEW SCAN STARTS ==>>

[ 3838.185707] iwlwifi 0000:01:00.0: U iwl_mvm_scan_request Handling
mac80211 scan request