Re: iwlwifi: regression due to: allow different csr flags for different device families

On Thu, 2018-05-31 at 11:44 +0300, Kalle Valo wrote:

Luca Coelho [off-list ref] writes:

quoted

On Sun, 2018-05-27 at 14:08 +0200, Alexander Wetzel wrote:

quoted

quoted

From: Golan Ben Ami <redacted>

Different device families may have different flag values
for passing a message to the fw (i.e. SW_RESET).
In order to keep the code readable, and avoid conditioning
upon the family, store a value for each flag, which indicates
the bit that needs to be enabled.

Signed-off-by: Golan Ben Ami <redacted>
Signed-off-by: Luca Coelho <redacted>

This patch generates a Null pointer exception for my Ultimate-N
6300
card when loading the driver, making the card unusable.
(Bisecting linked it to commit
a8cbb46f831d2c101feccdd0e0daf3627b8c1dca.)

Sine the commit could not be simply reverted I applied this
commit to
my
distribution kernel - basically 4.16.12 - and can confirm it is
indeed
causing the issue.

It seems that I forgot to reply to this, even though we looked into
it
and provided a fix[1].  This fix will hopefully be included in 4.17
final.

Thanks a lot for reporting and bisecting!

[1] https://patchwork.kernel.org/patch/10437773/

Something is not right, commit 0513c083d1f4 is in wireless-drivers-
next
so it will be in 4.18. Are we mixing commits or releases or did I
pull
it to wrong tree?

0513c083d1f4 iwlwifi: add csr configuration for 6300 devices

No, you are right, it was my mistake (I mixed this up with the 16 CPUs
which you picked up for 4.17).  The fix for this regression should go
only to 4.18, since the original patch is not even in 4.17-rc*.

Sorry for the confusion.

So, rephrasing my buggy phrase from the previous email:

"This fix will hopefully be included in 4.18-rc1 and probably earlier
in wireless-testing".

--
Cheers,
Luca.