build: ar71xx remove packages from archer-c7-v1 #1349
Conversation
removed ath10k because the 5 GHz hardware, QCA9880-AR1A, is unsupported and boot loops kmod-ath10k ath10k-firmware-qca988x 5 GHz radio decommissioned, but router works. Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
|
The real bug still lurks in ath10k though. I remember that it would just refuse to load due to the unsupported chip. Now, it crashes because it shares the same board-id with the C7 v2. |
|
There was a discussion on IRC about this. The consensus was that Qualcomm killed support for it since the hardware was defective in some or several ways. |
|
@neheb (and irc? is there a log of that discussion somewhere?) The Archer C7 V1(.1) has a minipcie slot! So changing out the original QCA9880-AR1A and replacing it with a wle900vx or a unex card, etc. is easily possible on the V1. The C7 V2 did away with the minipcie slot and has the QCA9880 soldered on. The problem now is, that even if the card is replaced by a supported QCA9880-2R4E or QCA9880-BR4A based card, "this" Archer C7 v1 image will still happily crash in the very same way. And this is because of the 11-ath10k-caldata extraction script: The Archer C7 V1 does not have a caldata there, instead the area is filled with 0xff... And this is what's crashing ath10k, since it gets fed that 0xff... through the ath10k/cal-pci-0000:01:00.0.bin file. This can easily tested too: just remove the /etc/hotplug.d/firmware/11-ath10k-caldata file from the C7 V1 image. And ath10k will no longer crash for QCA9880-AR1A cards (instead it will say that it is not supported) and for users that have a QCA9880-2R4E or QCA9880-BR4A card, ath10k will work as expected. |
|
Ah interesting. I didn't know it was crashing because of that. Anyway, the V2 and V3 also have a minipcie slot (I have both, completely identical hardware). I keep IRC logs. I can post if you want. |
Oh Ok, thanks. Yes I'm wrong, the pictures I found for the V2 and V3 confirmed that they had the minipcie slot too.
Yes please. I've posted the ART hexdump on debian's paste for my C7 V1. |
|
BTW, that IRC discussion was specifically in reference to this PR. |
|
In any case, ACK. If someone with an Archer C7v1 replaces the miniPCIe card, they can install the appropriate driver. This PR is for image defaults. |
Ok, I'll try it again. If the card has been the unsupported QCA9880v1 has been replaced with a QCA9880v2 and installed the kmod-ath10k package, they will see the same panic. In fact, you can test this too with your C7v2 (I think most QCA988X will do, However please keep a serial cable ready! Because you could get locked-out). Just run the following on your C7v2:
and reload ath10k_pci and watch it panic. |
|
That is indeed a problem. My thinking is that useless packages should not be included by default. |
That's 100% true.
that one does not work, if the replacement card is a QCA9880v2/QCA9880-2R4E/QCA9880-BR4A though. |
|
My view is that this is not a "useless package" but rather a "harmful
package." In particular, there is discussion on various forums where
people who do not understand failsafe think that they have bricked their
router.
I believe that a router that is working but degraded is better than a
router that will not work. The documentation blurb should say that the
stock 5GHz radio is not supported.
This is a marketing disaster for OpenWRT as it stands. Just drop the
offending driver and document why and how for the more advanced users.
…On Wed, Sep 19, 2018 at 12:20 PM Christian Lamparter < ***@***.***> wrote:
My thinking is that useless packages should not be included by default.
That's 100% true.
If someone with an Archer C7v1 replaces the miniPCIe card, they can
install the appropriate driver. This PR is for image defaults.
that one does not work, if the replacement card is a
QCA9880v2/QCA9880-2R4E/QCA9880-BR4A though.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#1349 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AFXsT0--eJMzKCpUZGwe5cfW4jVY6oErks5ucnzMgaJpZM4WYh71>
.
--
Aubrey McIntosh, Ph.D.
1502 Devon Circle
Austin TX 78723
(512) 791-3801
|
hauke
added
the
release/18.06
|
Can we come to some conclusion? :-) My view is, that we should accept this PR as it is, if we can't come up with some better fix/solution like the one suggested by @greearb on that IRC
For the record, download stats for c7-v1 images: |
|
This PR looks correct. If ath10k does not work, space should not be increased needlessly. |
If I understand it correctly, space isn't issue, the issue is crashing kernel, reboot loop. |
|
right. Both size and crashing are problems. |
|
There have been reports as recently as this month that 18.06.2 does a bootloop. |
Yes you can if you want. To stop the crashes when ath10k_pci is loaded (in this case bootloop) requires to revert this upstream change: But since the QCA9880-AR1A is technically no longer supported by the ath10k driver that patch can't be undone in upstream because of that. What's more, the patch already there survived multiple takedowns, so this would need to be done locally in OpenWrt's ath10k-ct and ath10k and I can't see anyone doing that (or come up with another patch) since Michal is no longer working there. (The null caldata is only a problem if the user replaced the QCA9880. However, It should no longer crash as bad, so at least something got fixed 👍 ) |
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
|
So I'm going to push this fix ynezz@c531735 to master later today, once the updated snapshot images are built and ready for testing, I'll ask on FS#1743 for some Thank you all. |
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt/openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt/openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
|
so much wrong informations provided here that i can't afford time to reply to everbody. i'd provide patch to ath10k list but it won't be included in next final release since drivers won't get updated. a rather simple change to keep the driver building without bootloop is to move chip reset after chip supported check so it doesn't crash with AR1A and works if other card is put in mpcie slot. |
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt/openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
Ping. This is last reminder, I'm removing this from my TODO list. |
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. #1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Reviewed-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Reviewed-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Reviewed-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
ath10k_pci driver crashes once loaded and causes boot loops on this
device as 5GHz radio QCA9880-AR1A shipped with this router is broken.
It's not possible to fix this problem in software, miniPCIe radio has to
be replaced.
We could've probably fixed crashing of the ath10k driver by reverting
following upstream commit:
commit 1a7fecb766c83dace747f42b25bbb544b00a0163
Author: Michal Kazior <michal.kazior@tieto.com>
Date: Sat Jan 24 12:14:48 2015 +0200
ath10k: reset chip before reading chip_id in probe
but it's not worth the effort as it wouldn't make that 5GHz radio usable
anyway. So it seems more convenient to just remove the crashing driver
and provide bootable images, as I believe, that a router that is working
but degraded is better than a router that will not work.
For details please see discussions in PR[1] and in FS#1743[2].
1. openwrt#1349
2. https://bugs.openwrt.org/index.php?do=details&task_id=1743
Reviewed-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
Signed-off-by: Aubrey McIntosh, PhD <aubrey.mcintosh@utexas.edu>
[subject and commit message tweaks]
Signed-off-by: Petr Štetiar <ynezz@true.cz>
removed ath10k because the 5 GHz hardware,
QCA9880-AR1A, is unsupported and boot loops
kmod-ath10k ath10k-firmware-qca988x
5 GHz radio decommissioned, but router works.
Signed-off-by: Aubrey McIntosh, PhD aubrey.mcintosh@utexas.edu