Send patches to the mailing lists below. Kalle Valo reviews the patches within the next few days and, if they are ok, commits them to ath.git.

• To: ath10k@lists.infradead.org
• Cc: linux-wireless@vger.kernel.org

Preferably use ath.git master branch as the baseline for patches. Other trees can be used as well, but then the chances of conflicts are higher.

It's important that linux-wireless is CCed on the patches, otherwise it won't be in patchwork and the maintainer won't see it. You can follow the state of ath10k patch review from patchwork:

https://patchwork.kernel.org/project/linux-wireless/list/?state=*&q=ath10k

More info about submitting patches:

• linux-wireless submitting patches instructions
• linux-wireless git guide
• Linux kernel submitting patches
• How to write proper git commit messages
• Ingo Molnar's points about commit logs

Guidelines for patches are:

• MUST follow Documentation/SubmittingPatches
• MUST follow ath10k coding style
• MUST be compiler and sparse warning free.
• git send-email SHOULD be used to submit the patch to avoid any formatting issues.
• Patchsets SHOULD contain no more than 12 patches and include a cover letter.
• Commit log MUST answer the question "Why?".
• Commit log MUST not be empty.
• Commit log MUST have Tested-on tags.
• SHOULD use Reported-by: and Tested-by: tags if others have reported the issue.
• SHOULD use Fixes: tag if it fixes a regression caused by known commit.
• SHOULD be mostly checkpatch clean (though not all patchworks warnings make sense), it's RECOMMENDED to test patches with ath10k-check as that will disable the useless warnings.

The terminology is from http://www.ietf.org/rfc/rfc2119.txt

The commit log needs to describe the motivation behind the bug, in other words provide the background information why the patch is needed and why is it needed the way it is.

• How does it change the functionality from user's point of view?
• Does it fix a bug? If it does, please describe the bug (doesn't necessarily need to be long). Also if there's a public bug report add a link to the bug report or email describing the issue.
• If a problem has been found during code review and doesn't fix a known issue, mention that in the commit log this is a theoretical fix.

Ingo Molnar has written a nice description about what maintainers are looking from a commit log, read that to understand more about commit logs.

ath10k supports multiple hardware families:

• QCA988X PCI Wave 1 AP devices

• QCA9884 PCI Wave 2 AP devices

• QCA4019 part of IPQ4019 AP SoC

• QCA6174 PCI mobile devices

• QCA6174 SDIO mobile devices

For each of these families have their own firmware branch with different interfaces and possibly different functionality. So then writing a patch for ath10k it's not enough to think about just one family and instead all families need to be taken into account. Otherwise risk of regressions on other families increases significantly.

To speed up patch acceptance the commit log should explain what families are affected and what families are not, and why. If that is not done that task will be left for the driver mantainer to do and that will slow down the review significantly.

As there are so many different hardware families, to make it clear on what hardware and firmware the patch is tested on please use the Tested-on tag. It's similar as Fixes tag, just informing the testing information. The format is:

Tested-on: <hwname> <hwversion> <bus> <fwversion>

Add this information after the commit log text, but before the s-o-b lines, and adding an empty line between Tested-on and s-o-b tags. There can be (and is very much preferred!) to have multiple Tested-on tags, each tag for every hardware tested.

Few examples:

Tested-on: WCN3990 hw1.0 SNOC WLAN.HL.3.1-01040-QCAHLSWMTPLZ-1
Tested-on: QCA6174 hw3.2 SDIO WLAN.RMH.4.4.1-00029
Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00110-QCARMSWP-1
Tested-on: QCA9888 hw2.0 PCI 10.4-3.10-00047

Tested-on tag should be in every patch, as non-trivial patches should not be submitted without testing. For trivial patches it's ok to skip Tested-on tag but then it should say “Compile tested only”.

The ath10k patch flow is this:

• Patch gets posted to the mailing lists.
• Kalle applies ASAP (usually can take 1-3 business days) the patch to pending and master-pending branches for build testing.
• Kalle waits minimum of two business days for patch being under review (unless the patch is urgent).
• If no comments or warnings, Kalle applies the patch either to ath-current branch (for critical fixes) or to ath-next branch (new features, low priority fixes) and sends a “Thanks, applied” reply.
• Kalle merges ath-next and ath-current to master branch immeadiately after the patch is applied.
• Kalle merges ath-next into wireless-drivers-next tree and ath-current to wireless-drivers tree roughly every 2-3 weeks.
• David Miller merges wireless-drivers-next into net-next tree and wireless-drivers to net tree every two weeks or so.
• Linus Torvalds merges net every week and net-next during the merge window into linux.git .

As a rough estimate it takes 2-4 months for a patch to propagate from ath-next to an official Linux release.

To clarify the meaning with ath-current and ath-next let's take a concrete example: let's say that the latest release from Linux is v4.9-rc2. If a patch is applied to ath-current it will most like be in v4.9-rc4 or v4.9-rc5 (usually it takes a minimum of one week to get to Linus' tree, sometimes more). But if the patch is applied to ath-next the first release it will be in is v4.10-rc1.

See also ath10k sources and branches.