Davidc502- wrt1200ac wrt1900acx wrt3200acm wrt32x builds


#1720

It's on auto now...


#1721

Just in case you're new to OpenWrt, be aware that all "auto" does is default to the lowest channel (36 in the case of a 5GHz radio); it doesn't do anything clever like finding out the least contended channel or anything. You benefit from getting some sort of network scanning tool for your laptop or smartphone and having a look to see what the best channel to run on is (avoiding whatever channels are designated DFS in your region if you're likely to pick up weather or aircraft radar).


#1722

I wouldn't say that...
https://git.openwrt.org/?p=openwrt%2Fopenwrt.git&a=search&h=HEAD&st=commit&s=mwlwifi

It's slow these days but far from dead.


#1723

The mwlwifi drivers are being developed at https://github.com/kaloz/mwlwifi/commits/master, and there have been no significant improvements since mid 2018.


#1724

Are there / do we know if there are any modern wireless chipsets with open source drivers and firmware that are fairly active? I'm starting to see that Marvell might be a crutch moving forward, and wondering if I should base my next hardware purchase on a device that's going to have better open source support at the wireless level. It's a shame because the Linksys WRT AC Series is excellent aside from the clunky wireless implementation and wireless support leaving developers guessing and trying to reverse engineer based on comments from BrainSlayer here https://github.com/kaloz/mwlwifi/issues/350


#1725

Unfortunately, i have no good news for you.
Qualcomm/Atheros is even more buggy (non-opensource as well), Broadcom doesn't reveal their 5G up to now... Maybe Mediatek, but it didn't try it yet.

They all utilize the trick invented by Google ("firmware/certs can be not opensource"), to utilize all profits of it but not reveal their "Intellectual property". Not fair but who cares....

So Marvell looks quite good in compare among others...


#1726

So, this is as good as it's going to get for the WRT series as far as mwlwifi drivers?


#1727

Probably. :frowning_face:

If one thinks about it, we aren't using Wrt because of the Wifi anyway. For the sake of argument, let us say that Wifi is paramount (Most Important), and not the router packages.

  1. Buy a separate Wifi AP and connect it to the Wrt router. Turn OpenWrt Wifi Off.
  2. Stick with the manufacturer's firmware and hence Wifi drivers.

Honestly, I've used the Linksys factory firmware on my 1900ac, 1900acs, and 3200acm, and I have not found it to be much better than the opensource. Yeah, Opensource is buggy on 5Ghz in the DFS ranges, and we don't get MU-MIMO < and option I could care less about because it is a bit snake oil-ish to me just like "BeamForming". However, we do get BeamForming on 5Ghz as it is a part of the AC standard.

Bottom line? It is my opinion many people ruined OpenWrt by using Wifi in ways it wasn't intended to be used (Out of Specifications). The FCC has zero tolerance and shut that down by changing the rules that hurt the OpenWrt project.


#1728

Well, we actually have a new wifi commit.


#1729

I gave up on trying to get wifi stable on my wrt3200acm. Now I'm just using it in bridge mode- seems to be fairly stable for that, I couldn't get it to work reliably with anything else no matter what I did. It's a shame because I paid $200 for the thing.

Now I'm looking to buy an Asus 86u. Heard good things about it. I think I could benefit with a tri-band router but I honestly don't want to spend the money for a decent one and I like the Merlin firmware.


#1730

I'm almost in disbelief! Lol. Nice that this got through.


#1731

Well I guess we'll only see "major" bug fixes from now on.

"Busy on 11ax."


#1732

Sad but true. The Marvell developer yuhhaurlin essentially confirmed today mwlwifi efforts are done except bug fixes. They are working on this pointless 802.11ax (Wi-Fi 6) standard that would have everyone buying new hardware and seeing no tangible improvement unless deployed across large networks.

I probably wouldn't buy a Marvell product again. That been said, open source for drivers tends to be poor across the board, so this is par for the course. Even ath10k has problems, and none support MU-MIMU even though it was promised for mwlwifi. Few companies have open source drivers that reach closed source quality (NVIDIA comes to mind as being close or in some ways better).

OpenWrt still is quite good on my WRT32X, one of the top routers for open source, so I'm not complaining. Wi-Fi works fine for the most part, and mvebu is a solid platform that hopefully will continue to improve.


#1733

I don't really agree on this being pointless. As a family of five with 15 Wi-Fi capable devices, and neighbors that have nearby APs using asinine settings from the ISP (HT 40/80, and multiple APs on the same channel, etc.), throughput gets hit hard and packet loss issues spike during peak hours (6PM-10PM). Of course I know that clients also need to support the AX standard, but in the end that's just the nature of technology standards moving forward. However, I myself am probably gonna skip this next gen and wait for full duplex WiFi (I really hope it goes from just an interest/research group to a full working group).

Agreed. Their hardware is great and as a result the WRTAC series are very capable routers - it's just that their software/drivers for their radios aren't great, but still better than some others. Back in the early days of this device series, the issues and mismanagement was already evident when it took almost a year after Linksys declared "it's open source ready" to have a somewhat functioning stable driver. Look at the owner of the repo as well, Kaloz - I think he's left Linksys for QCA since his email handle now is under the code aurora domain. It was just not a good sign from the start. Finally, it is pretty evident to see how much Marvell cares about open source - they're only paying/assigned one dev to the mwlwifi project. They also gave up even trying to upstream it to the linux kernel when they got lambasted by the linux network kernel devs for code quality issues.

I own a 1900ACS and I do agree that it is still a capable router. It's just mired by some radio/WiFi issues.


#1734

Honestly, at least we actually know what's going on now. Before it was radio silent.

It definitely sucks they've essentially dropped all but the most major of issues and have no plans of rectifying any of the existing problems.

As others have mentioned, at least they're providing some / more level of open-source support for their chipsets. For that much, I need to be grateful for. That being said, the level of support is definitely not their top priority. Marvell as a company is probably wanting to get new hardware out to make money off of.

I'll have my eyes on competing open source chipsets moving forward, but the competitors seem to have quite a bit of work ahead of them to get to a decent level. Unfortunately the way things stand isn't a very high quality standard across the board.


#1735

Maybe it’s Stockholm Syndrome, but I have hopes that there will be learnings from the mistakes of 88w8864/8964 and these carry through to 9064.

I kind of hope that all of the issues “that will be fixed later” get fixed in the new development.

wishful thinking :blush:


#1736

A new build has been uploaded to the server.

Kernel version = 4.14.103
WiFi driver = 10.3.8.0-20181210
Build = r9506

  1. Gentle Reminder - The new URL is https://dc502wrt.org so be sure to update any shortcuts/favorites you may have.

  2. Though the Wifi driver version has not changed, there have been some commits/changes since the last build.

  3. I was thinking about dumping Rosy LuCi, but it was pointed out to me that it is still being improved after all, so I kept is as is for now. This is still subject to change depending on the feedback.

  4. OPKG Configuration has been updated to reflect the new domain/URL. I've tested it and is working properly.


#1737

Hey @davidc502
Both r9506/ and r9287/ are missing the cryptsetup-openssl package (and its dependency libopenssl).
However, I am only able to open my encrypted VeraCrypt disks with this specific package; the regular cryptsetup package just does not work (it literally does nothing when trying to open the disks). For now I've installed the package from r9133/ which still contains it.

Any chance of it coming back?


#1738

Good question... I looked through the build menuconfig, and I don't see it included anymore. There's not even an option to build it. However, I do see libopenssl 1.1 available, but that doesn't help you without cryptsetup-openssl.

Also, I went ahead and checked the OpenWrt daily builds, and I do not see cryptsetup-openssl there either, so that isn't good news.

Since I'm not seeing it in the daily builds, I figure this isn't just related to this 3rd party build. So, I would go ahead and start a thread in this forum "Community Builds, Projects and Packages", and see if someone knows what happened to that particular package. Maybe there is some insight as to what might have happened to it and why.


#1739

Thank you, I'll do that. Is there there any inherent problem in installing the package from 9133? Now or in the future? It seems to work fine.