build of rpi-5.4.y (5.4.58) fails on 32 bit

[ncopa]

Describe the bug
Build rpi-5.4.y (commit a682509) with 32bit config fails.

To reproduce
build the rpi-5.4.y kernel with either bcmrpi_defconfig bcm2709_defconfig or bcm2711_defconfig (not sure exactly which one triggered it)

Expected behaviour
build success

Actual behaviour
build fails:

/home/ncopa/aports/main/linux-rpi/src/linux-5.4/drivers/net/wireless/ath/ath9k/hif_usb.c: In function 'ath9k_hif_usb_reg
_in_cb':                                                    
/home/ncopa/aports/main/linux-rpi/src/linux-5.4/drivers/net/wireless/ath/ath9k/hif_usb.c:735:3: error: 'rx_buf' undeclar
ed (first use in this function); did you mean 'tx_buf'?
  735 |   rx_buf->skb = nskb;                               
      |   ^~~~~~                                            
      |   tx_buf                                            
/home/ncopa/aports/main/linux-rpi/src/linux-5.4/drivers/net/wireless/ath/ath9k/hif_usb.c:735:3: note: each undeclared id
entifier is reported only once for each function it appears in

System
Copy and paste the results of the raspinfo command in to this section. Alternatively, copy and paste a pastebin link, or add answers to the following questions:

• Which model of Raspberry Pi? e.g. Pi3B+, PiZeroW
• Which OS and version (cat /etc/rpi-issue)?
• Which firmware version (vcgencmd version)?
• Which kernel version (uname -a)?

This build happens on an arm server from packet net. (but its irrelevant)

$ uname -a
Linux ncopa-edge-armv7 5.4.34-0-lts #1-Alpine SMP Wed, 22 Apr 2020 19:26:07 UTC armv8l Linux

Logs

/home/ncopa/aports/main/linux-rpi/src/linux-5.4/drivers/net/wireless/ath/ath9k/hif_usb.c: In function 'ath9k_hif_usb_reg
_in_cb':                                                    
/home/ncopa/aports/main/linux-rpi/src/linux-5.4/drivers/net/wireless/ath/ath9k/hif_usb.c:735:3: error: 'rx_buf' undeclar
ed (first use in this function); did you mean 'tx_buf'?
  735 |   rx_buf->skb = nskb;                               
      |   ^~~~~~                                            
      |   tx_buf                                            
/home/ncopa/aports/main/linux-rpi/src/linux-5.4/drivers/net/wireless/ath/ath9k/hif_usb.c:735:3: note: each undeclared id
entifier is reported only once for each function it appears in
  LD [M]  drivers/media/usb/gspca/gspca_finepix.o

Additional context
I believe commit e6eb815 needs to be cherry-picked to rpi-5.4.y.

What happened upstream in linux-5.4.y was:

• commit b5c8896 Fix general protection fault in ath9k_hif_usb_rx_cb
• this apparently caused problems so it got reverted in 5a046d7
• a fix those problems was made so the problematic commit was re-applied in e6eb815.
• The fix for the problems was applied on top in commit c15d59b ath9k: Fix regression with Atheros 9271

I think the rpi branch has both the revert commit (5a046d7) and the fix (c15d59b ath9k: Fix regression with Atheros 9271) for the reverted commit but not the re-applied e6eb815.

To get a clear view of the history:

git log origin/linux-5.4.y -- drivers/net/wireless/ath/ath9k/hif_usb.

[pelwell]

Duplicate of #3786.

[ncopa]

Duplicate of #3786.

The second problem (with lan78xx.c which i have delete now) was a duplicate, but the original problem with drivers/net/wireless/ath/ath9k/hif_usb.c is not mentioned in #3786. They will bump into the ath9k error (this issue) once they solve the lan78xx.c problem.

[pelwell]

They're both trivial problems - there's almost more work updating the issues, so I'd rather only do it once.