Re: Firmware Crashed

[Ben Greear <greearb@xxxxxxxxxxxxxxx>]

On 03/23/2020 08:30 PM, Ming Chen wrote:
> > > > > > > No, it will be very difficult for us to upgrade the kernel.
> > > > > > > This module(QCA9986 PCIe) is from the a third party company. They
> > > > > > > are using the QCA driver and it works well. We are trying to
> > > > > > > integrate this module to our own device(PPC-64 + Kernel 4.14.83).
> > > > > > > We don't want to upgrade the kernel, since it will be a huge
> > > > > > > challenge work for
> > > > us.
> > > > > > > Thanks
> > > > > > > Ming Chen
> > > > > > >
> > > > > > > > -----Original Message-----
> > > > > > > > From: Tom Psyborg <pozega.tomislav@xxxxxxxxx>
> > > > > > > > Sent: Saturday, March 21, 2020 11:16 PM
> > > > > > > > To: Ming Chen <Ming.Chen@xxxxxxxxxxxxxx>
> > > > > > > > Cc: linux-wireless <linux-wireless@xxxxxxxxxxxxxxx>
> > > > > > > > Subject: Re: Firmware Crashed
> > > > > > > >
> > > > > > > > can you try to boot it in another device and using some distro
> > > > > > > > that has 4.19 or later kernel without backports?
> > > > > >
> > > > > > My suggestion is to try it in x86_64 PC. Then you can just use live
> > > > > > linux distro with different kernel versions. If it does not work
> > > > > > with any of these you may have ran into a hw bug sample. If it does
> > > > > > work then you don't need to update your PPC64 kernel but find an
> > > > > > architecture/device speific bug that prevents it from loading.
> > > > >
> > > > > [Ming Chen]
> > > > > Thanks for the suggestion. I don't have a PC with PCIe socket on my
> > > > > hand, but
> > > > it's indeed worth to have a try. I will try to find a one.
> > > >
> > > > In case you want to try the ath10k-ct firmware for your chip, you can
> > > > send me the firmware crash output and maybe I can understand why it is
> > crashing.
> > > > http://www.candelatech.com/ath10k.php
> > > >
> > > > Thanks,
> > > > Ben
> > > >
> > > > --
> > > > Ben Greear <greearb@xxxxxxxxxxxxxxx>
> > > > Candela Technologies Inc  http://www.candelatech.com
> > > [Ming Chen]
> > > Thanks for reminding. I just realized that the functionality of board-2.bin in
> > ath10k driver was commented out by someone of my company, and now I
> > restored this functionality in ath10k driver. But, firmware is still crashed, and it
> > seems like it is caused by failing to fetch board data. That is said, the driver
> > failed to find a matched (bus=pci,bmi-chip-id=0,bmi-board-id=11) board data
> > from board-2.bin.
> > > Does that mean the vendor of this wireless module has not committed their
> > board data to upstream and build to board-2.bin file?
> > > The below is the piece of BMI log,
> > > [ 9217.343233] ath10k_pci 0000:01:00.0: boot upload otp to 0x1234 len
> > > 8999 for board id [ 9217.343239] ath10k_pci 0000:01:00.0: bmi fast
> > > download address 0x1234 buffer 0x800000000018a084 length 8999 [
> > > 9217.343242] ath10k_pci 0000:01:00.0: bmi lz stream start address
> > > 0x1234 [ 9217.343828] ath10k_pci 0000:01:00.0: bmi lz data buffer
> > > 0x800000000018a084 length 8996 [ 9217.372561] ath10k_pci 0000:01:00.0:
> > > bmi lz data buffer 0xc00000007584fa50 length 4 [ 9217.372782]
> > > ath10k_pci 0000:01:00.0: bmi lz stream start address 0x0 [
> > > 9217.372831] ath10k_pci 0000:01:00.0: bmi execute address 0x1234 param
> > > 0x10 [ 9218.783383] ath10k_pci 0000:01:00.0: bmi cmd took 353 jiffies
> > > hz 250 ret 0 [ 9218.783396] ath10k_pci 0000:01:00.0: bmi execute result
> > 0x2c00 [ 9218.783401] ath10k_pci 0000:01:00.0: boot get otp board id result
> > 0x00002c00 board_id 11 chip_id 0 ext_bid_support 0 [ 9218.783405]
> > ath10k_pci 0000:01:00.0: SMBIOS bdf variant name not set.
> > > [ 9218.783409] ath10k_pci 0000:01:00.0: DT bdf variant name not set.
> > > [ 9218.783415] ath10k_pci 0000:01:00.0: boot using board name
> > 'bus=pci,bmi-chip-id=0,bmi-board-id=11'
> > > [ 9218.783420] ath10k_pci 0000:01:00.0: boot using board name
> > 'bus=pci,bmi-chip-id=0,bmi-board-id=11'
> > > [ 9218.783916] ath10k_pci 0000:01:00.0: boot fw request
> > > 'ath10k/QCA9984/hw1.0/board-2.bin': 0
> >
> > This is about a wave-2 9984 NIC.  You sure yours is 9886 chipset?
> >
> > Thanks,
> > Ben
> >
> > --
> > Ben Greear <greearb@xxxxxxxxxxxxxxx>
> > Candela Technologies Inc  http://www.candelatech.com
> [Ming Chen] Yeah, I am sure it is a QCA9986 (not 9886), and I can see its device ID is 168c:0046. And the ath10k wireless driver will try to load the firmware from the path of ath10k/QCA9984/hw1.0 by default. So I thought we should use the board binary file same as QCA9984 (https://github.com/kvalo/ath10k-firmware/blob/master/QCA9984/hw1.0/board-2.bin).
> Is there a board binary file for QCA9986 specifically?

Ok, I mis-read your earlier email.

You do need a board file that matches your card.  The NIC vendor should be able to provide this,
but you may have to package it in board2.bin format.  I think the ath10k-firmware git project
has some scripts in it to do the packaging...or google can probably find a link.

Thanks,
Ben

--
Ben Greear <greearb@xxxxxxxxxxxxxxx>
Candela Technologies Inc  http://www.candelatech.com