ar71xx: Fix 5 GHz Wi-Fi on NBG6716 #655
Conversation
|
Hi @malaakso, It seems that manufacturer has different versions of this device or even different versions of the AC card installed inside. Photos of the AC card from FCC don't show any EEPROM on it (see: I88NBG6716). And actually, I have this device here and it has two calibration data in mtd2 partition which means that your change would break these devices. This issue requires more sophisticated fix (e.g. check if calibration data exists in flash and don't use it if it doesn't). Cheers, |
|
@pepe2k Yeah, that's something I feared could be the case. I think it is enough to check the first few bytes of the calibration data on flash to make sure that it is valid, and go from there. I'll update the patch. I assume the calibration-data-in-flash contains the default MAC address (00:03:7f:12:34:56) in your case? |
Sounds good to me.
Both radios MACs in calibration data seem fake and don't match what's on label but they are different than yours. What's more, they are the same in at least two devices I had access to: 2G: 00:02:6F:12:34:56 Cheers, |
|
@pepe2k Please review. |
| @@ -18,12 +18,19 @@ ath10kcal_extract() { | |||
| local part=$1 | |||
| local offset=$2 | |||
| local count=$3 | |||
| local mtd | |||
| local mtd size1 size2 | |||
There was a problem hiding this comment.
I prefer variable name which is more descriptive, maybe cal_size?
(why just one, see below)
|
|
||
| mtd=$(find_mtd_chardev $part) | ||
| [ -n "$mtd" ] || \ | ||
| ath10kcal_die "no mtd device found for partition $part" | ||
|
|
||
| # Check that the calibration data size in header equals the desired size |
There was a problem hiding this comment.
I'm worried about possible regression if we apply this check globally, for all boards.
Are you able point out any information about header layout and if it's the same for all ath10k chipsets?
There was a problem hiding this comment.
Yes, the header layout is the same for all current ath10k chips, both cal and pre-cal data. The same check is made in Qualcomm driver.
There was a problem hiding this comment.
Can you tell me where I can find it (link or what source file look into)?
There was a problem hiding this comment.
For example https://git.kos.org.cn/openkos/qca-wifi/blob/master/qca-wifi/qca-wifi-10.2.4.58.1/offload/os/linux/tools/athdiag/qc98xx_eeprom.h look for base_eep_header.
There was a problem hiding this comment.
Great, thanks!
I think it's safe to keep that check for all boards.
|
|
||
| mtd=$(find_mtd_chardev $part) | ||
| [ -n "$mtd" ] || \ | ||
| ath10kcal_die "no mtd device found for partition $part" | ||
|
|
||
| # Check that the calibration data size in header equals the desired size | ||
| size1=`dd if=$mtd bs=2 count=1 skip=$(( $offset / 2 )) conv=swab 2>/dev/null | hexdump -ve '1/2 "%.4x"'` |
There was a problem hiding this comment.
Please use var=$(...) instead of var=`...` . And, what about something different (no need to change count to hex):
cal_size=$(dd if=$mtd bs=2 count=1 skip=$(($offset/2)) conv=swab 2>/dev/null | hexdump -ve '1/2 "%d"')
There was a problem hiding this comment.
Heh, I was somehow so completely fixated on getting it to hex that the easier route eluded me. I'll send a new version soon.
| size2=`printf %.4x $count` | ||
|
|
||
| [ $size1 != $size2 ] && \ | ||
| ath10kcal_die "no calibration data found from $part" |
There was a problem hiding this comment.
ath10kcal_die "no calibration data found in $part"
| size1=`dd if=$mtd bs=2 count=1 skip=$(( $offset / 2 )) conv=swab 2>/dev/null | hexdump -ve '1/2 "%.4x"'` | ||
| size2=`printf %.4x $count` | ||
|
|
||
| [ $size1 != $size2 ] && \ |
There was a problem hiding this comment.
One more thing: [ "$count" = "$cal_size" ] || \
Some NBG6716 do not have ath10k calibration data in flash, only in chip OTP. To determine if flash has valid calibration data, the first two bytes telling the length of the calibration data are checked against the requested length. If the lengths match, calibration data is valid and read from flash. See: https://forum.lede-project.org/t/17-01-1-2-3-qca988x-ath10k-5ghz-firmware-crashed-zyxel-nbg6716/1911?u=malaakso Signed-off-by: Matti Laakso <matti.laakso@outlook.com>
|
@pepe2k Updated pull request, please review. |
|
@malaakso sorry for the delay, I have merged your changes into my staging tree, thanks! Cheers, |
NBG6716 has calibration data in the PCI-E card EEPROM, not in flash.
See: https://forum.lede-project.org/t/17-01-1-2-3-qca988x-ath10k-5ghz-firmware-crashed-zyxel-nbg6716/1911?u=malaakso
Signed-off-by: Matti Laakso matti.laakso@outlook.com