ipq806x: add support for Linksys EA8500

[ianchi]

Tested and working:

• Flashing from original webui using the factory images (it generates one
  for original firmware 1.1.4.168206 an another for 1.1.4.169978, newer US
  Firmwares are not supported due to signing)
• Sysupgrading from within OpenWrt.
• wifi 2.4GHz and 5GHz.
• wired lan
• usb storage

Pending:

• patch wifi iface mac address

[blogic]

looks ok, will pull this once we have v4.4 support in the tree

[plntyk]

I am a little bit unsure about ath10k stuff (the last + base-files patches)

• ath10k_core has an "skip_otp=y" option - this option was working on my ath10k hardware (Sitecom WLR8100) that seemed to have some issues w. AC wireless/firmware/art partition (skip_otp not required w. the recent mac80211 from 2016-05-12).
• why is the art/caldata sized differently compared to ar71xx w. ath10k ?

your changes:
ath10kcal_extract "art" 4096 12064
ath10kcal_extract "art" 20480 12064

ar71xx: 2116

[ianchi]

Hi @plntyk,
I will look into that option to see if I can get rid off the patch. That would be better.
In the meantime I had to add a modiffied version of the patch to match the new package version for mac80211 commited to Source.

Regarding the difference in size, I don't know. That is the size of the data as it comes preloaded in art partition from Linksys. The size is confirmed in the first two bytes of the caldata.
Archer C2600 which also uses qca99x0 has exactly the same size. Perhaps ar71xx uses a different one, like qca988x and size differs.

[bendavid]

Few comments.

1. I looked at the changes in the updated drivers in 2016-05-12 and I think this should solve the issue with the OTP, since the updated logic should cover what the patch was trying to do (same patch also posted here btw http://patchwork.ozlabs.org/patch/619947/).

2. Yes I confirm that the calibration data is a different size for QCA99x0 vs QCA988x. I think that this is tied to the firmware itself. (So possibly QCA9984 will have a slightly different calibration data size again since it appears to have a different firmware.)

[ianchi]

@bendavid I've tried removing the patch but the updated drivers won't work, I'm getting the same old error:

[   12.132467] ath10k_pci 0000:01:00.0: unable to read from the device
[   12.132492] ath10k_pci 0000:01:00.0: could not execute otp for board id check: -110
[   12.137537] ath10k_pci 0000:01:00.0: failed to get board id from otp: -110
[   12.145309] ath10k_pci 0000:01:00.0: could not probe fw (-110)

In the new drivers, core.c seems similar in this critical part. The problem remains as it continues to check for return code of 'ath10k_core_get_board_id_from_otp(ar)'

The option mentioned by @plntyk is also not working, as it gets checked further down the code, and ath10k_core_get_board_id_from_otp still results in an error.

The modified patch I added to his PR is somehow working. But something with the new driver has changed, and even though wlan is UP and I can connect, I get random warnings traces to the console from compat-wireless-2016-05-12/net/mac80211/rx.c (which weren't there with the previous drivers):

[  476.471514] ------------[ cut here ]------------                                                                                                                                           
[  476.471601] WARNING: CPU: 0 PID: 0 at compat-wireless-2016-05-12/net/mac80211/rx.c:4068 ieee80211_rx_napi+0x8c/0x8b4 [mac80211]()                                                          
[  476.475210] Modules linked in: pppoe ppp_async iptable_nat pppox ppp_generic nf_nat_ipv4 nf_conntrack_ipv6 nf_conntrack_ipv4 ipt_REJECT ipt_MASQUERADE xt_time xt_tcpudp xt_state xt_nat xh
[  476.585611] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G        W      3.18.29 #1                                                                                                              
[  476.585976] [<c021e8ac>] (unwind_backtrace) from [<c021b608>] (show_stack+0x10/0x14)                                                                                                       
[  476.593089] [<c021b608>] (show_stack) from [<c038799c>] (dump_stack+0x8c/0xa0)                                                                                                             
[  476.600899] [<c038799c>] (dump_stack) from [<c0226500>] (warn_slowpath_common+0x6c/0x88)                                                                                                   
[  476.607927] [<c0226500>] (warn_slowpath_common) from [<c02265b8>] (warn_slowpath_null+0x1c/0x24)                                                                                           
[  476.616195] [<c02265b8>] (warn_slowpath_null) from [<bf2bea08>] (ieee80211_rx_napi+0x8c/0x8b4 [mac80211])                                                                                  
[  476.624986] [<bf2bea08>] (ieee80211_rx_napi [mac80211]) from [<bf32d7b8>] (ath10k_htt_t2h_msg_handler+0x928/0x984 [ath10k_core])                                                           
[  476.634435] [<bf32d7b8>] (ath10k_htt_t2h_msg_handler [ath10k_core]) from [<bf32e1cc>] (ath10k_htt_txrx_compl_task+0x9b8/0x1178 [ath10k_core])                                              
[  476.646062] [<bf32e1cc>] (ath10k_htt_txrx_compl_task [ath10k_core]) from [<c022923c>] (tasklet_action+0xb8/0x144)                                                                          
[  476.658624] [<c022923c>] (tasklet_action) from [<c0229420>] (__do_softirq+0xf0/0x220)                                                                                                      
[  476.668868] [<c0229420>] (__do_softirq) from [<c0229800>] (irq_exit+0x98/0xe4)                                                                                                             
[  476.676682] [<c0229800>] (irq_exit) from [<c025ab64>] (__handle_domain_irq+0xac/0xd4)                                                                                                      
[  476.683797] [<c025ab64>] (__handle_domain_irq) from [<c02085f8>] (gic_handle_irq+0x40/0x5c)                                                                                                
[  476.691696] [<c02085f8>] (gic_handle_irq) from [<c0209680>] (__irq_svc+0x40/0x74)                                                                                                          
[  476.699848] Exception stack(0xc0713f48 to 0xc0713f90)                                                                                                                                      
[  476.707486] 3f40:                   ffffffed 00000000 0007af22 c020a220 c0712018 c0712000                                                                                                  
[  476.712525] 3f60: 00000000 00000000 ffffffed 00000000 c071a4b4 c0636660 c071b684 c0713f90                                                                                                  
[  476.720682] 3f80: c02190c8 c02190cc 60000013 ffffffff                                                                                                                                      
[  476.728843] [<c0209680>] (__irq_svc) from [<c02190cc>] (arch_cpu_idle+0x34/0x50)                                                                                                           
[  476.733888] [<c02190cc>] (arch_cpu_idle) from [<c0252960>] (cpu_startup_entry+0x150/0x2c8)                                                                                                 
[  476.741353] [<c0252960>] (cpu_startup_entry) from [<c06e7df8>] (start_kernel+0x510/0x51c)                                                                                                  
[  476.749425] ---[ end trace 3e350a6fd7a3d7f4 ]---              

[ianchi]

Hi @blogic ,
I don't know if this is the right way, but I was also trying to switch to 4.4, so I rebased my branch onto your stagging tree and added a few commits to include the device in that version.

Obviously your 4.4 branch is work in progress, but I wanted to share my results with it.
Basically I'm unable to boot because the NAND device is not recognized:

[    1.521561] qcom-nandc 1ac00000.nand: failure submitting descs for command 255
[    2.032489] qcom-nandc 1ac00000.nand: failure submitting descs for command 144
[    2.542044] qcom-nandc 1ac00000.nand: failure submitting descs for command 144
[    3.052525] qcom-nandc 1ac00000.nand: failure submitting descs for command 144
[    3.562469] qcom-nandc 1ac00000.nand: failure submitting descs for command 144
[    3.562504] nand: No NAND device found

Before trying with your stagging tree, I have also been trying to upgrade to 4.4 myself, and I was having the same problem.
For qcom_mandc (which I assume as a possible culprit) instead of your patch version I was trying to use the version that is already upstream (and to do so needed to pull other patches to reach that commit). It is in my repo, branch ea8500_mtd45 (also rebased onto yours).

Anyway, it compiles ok, but I had the exact same problem. MTD is not recognized that way either.

I'm willing to help with this target if I can. Please let me know if you have any hint to try.

Thnks

[blogic]

Hi,

ok, that is what i found. i hope to have time next week to look into this. i am currently waiting on receiving some test hardware. this is why the pot is pending. thanks for your patience. i think this will take another 2-3 weeks to get resolved.

[bendavid]

@ianchi I see the same ath10k stack traces with the new compat-wireless on the C2600, although I don't see any noticable impact on the functionality of the wireless.

[ianchi]

Perhaps @nbd168 can help us with this warning from the new version of Compat-wireless.
It is affecting both devices using QCA9984.

I copy again the log:

[  476.471514] ------------[ cut here ]------------                                                                                                                                           
[  476.471601] WARNING: CPU: 0 PID: 0 at compat-wireless-2016-05-12/net/mac80211/rx.c:4068 ieee80211_rx_napi+0x8c/0x8b4 [mac80211]()                                                          
[  476.475210] Modules linked in: pppoe ppp_async iptable_nat pppox ppp_generic nf_nat_ipv4 nf_conntrack_ipv6 nf_conntrack_ipv4 ipt_REJECT ipt_MASQUERADE xt_time xt_tcpudp xt_state xt_nat xh
[  476.585611] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G        W      3.18.29 #1                                                                                                              
[  476.585976] [<c021e8ac>] (unwind_backtrace) from [<c021b608>] (show_stack+0x10/0x14)                                                                                                       
[  476.593089] [<c021b608>] (show_stack) from [<c038799c>] (dump_stack+0x8c/0xa0)                                                                                                             
[  476.600899] [<c038799c>] (dump_stack) from [<c0226500>] (warn_slowpath_common+0x6c/0x88)                                                                                                   
[  476.607927] [<c0226500>] (warn_slowpath_common) from [<c02265b8>] (warn_slowpath_null+0x1c/0x24)                                                                                           
[  476.616195] [<c02265b8>] (warn_slowpath_null) from [<bf2bea08>] (ieee80211_rx_napi+0x8c/0x8b4 [mac80211])                                                                                  
[  476.624986] [<bf2bea08>] (ieee80211_rx_napi [mac80211]) from [<bf32d7b8>] (ath10k_htt_t2h_msg_handler+0x928/0x984 [ath10k_core])                                                           
[  476.634435] [<bf32d7b8>] (ath10k_htt_t2h_msg_handler [ath10k_core]) from [<bf32e1cc>] (ath10k_htt_txrx_compl_task+0x9b8/0x1178 [ath10k_core])                                              
[  476.646062] [<bf32e1cc>] (ath10k_htt_txrx_compl_task [ath10k_core]) from [<c022923c>] (tasklet_action+0xb8/0x144)                                                                          
[  476.658624] [<c022923c>] (tasklet_action) from [<c0229420>] (__do_softirq+0xf0/0x220)                                                                                                      
[  476.668868] [<c0229420>] (__do_softirq) from [<c0229800>] (irq_exit+0x98/0xe4)                                                                                                             
[  476.676682] [<c0229800>] (irq_exit) from [<c025ab64>] (__handle_domain_irq+0xac/0xd4)                                                                                                      
[  476.683797] [<c025ab64>] (__handle_domain_irq) from [<c02085f8>] (gic_handle_irq+0x40/0x5c)                                                                                                
[  476.691696] [<c02085f8>] (gic_handle_irq) from [<c0209680>] (__irq_svc+0x40/0x74)                                                                                                          
[  476.699848] Exception stack(0xc0713f48 to 0xc0713f90)                                                                                                                                      
[  476.707486] 3f40:                   ffffffed 00000000 0007af22 c020a220 c0712018 c0712000                                                                                                  
[  476.712525] 3f60: 00000000 00000000 ffffffed 00000000 c071a4b4 c0636660 c071b684 c0713f90                                                                                                  
[  476.720682] 3f80: c02190c8 c02190cc 60000013 ffffffff                                                                                                                                      
[  476.728843] [<c0209680>] (__irq_svc) from [<c02190cc>] (arch_cpu_idle+0x34/0x50)                                                                                                           
[  476.733888] [<c02190cc>] (arch_cpu_idle) from [<c0252960>] (cpu_startup_entry+0x150/0x2c8)                                                                                                 
[  476.741353] [<c0252960>] (cpu_startup_entry) from [<c06e7df8>] (start_kernel+0x510/0x51c)                                                                                                  
[  476.749425] ---[ end trace 3e350a6fd7a3d7f4 ]---              

[bendavid]

(I think you mean QCA9980.)

[nbd168]

This happens when the driver reports packets without setting the band properly in the receive status. It might be an upstream bug, or maybe the calibration data is broken.
It could also simply be a firmware bug and maybe a newer firmware helps.

[bendavid]

So, interesting point: The ath10k stack trace goes away after updating kernel from 3.18 to 4.4. (Maybe there has been some implicit/minor incompatibilities introduced between the newer compat-wireless and the older 3.18 kernel). Anyway, looks like the problem will solve itself when the 4.4 update converges regardless.

[bendavid]

Ok, no, spoke too soon, the stack trace did eventually appear again also in 4.4.

[ianchi]

The QCOM RPM driver was included upstream in v4, but a later commit (regulator: qcom: Rework to single platform device) left the smb208 regulator out.
Added a patch to fix it and update the DT.

[bendavid]

Indeed, this probably explains my speed test results here
https://forum.openwrt.org/viewtopic.php?pid=326104#p326104

[ianchi]

Added commit to fix buttons & wifi led.

[ianchi]

@bendavid I haven't tested wifi speed with this new driver. Will do asap.

My patch enables again smb208 in the driver, but there are still some problems with pcie in DTS.
The qcom,pcie driver expects the following properties that are missing, they should reference a regulator: vdda, vdda_phy, vdda_refclk
So for each of the 3 pcie we get the following error in klog (I assume you get them too) :

[    0.315081] 1b900000.pci supply vdda not found, using dummy regulator
[    0.315178] 1b900000.pci supply vdda_phy not found, using dummy regulator
[    0.315284] 1b900000.pci supply vdda_refclk not found, using dummy regulator         

I've been doing some research but couldn't find the missing mappings.

[bendavid]

Ok I was actually referring to the wired network speed.

[ianchi]

Rebased to current trunk.
@blogic I've fixed all conflicts with your updates to ipq806x, so it should be a clean merge.

I've re-tested in EA8500 and is working ok on Kernel 4.4.

[blogic]

ovrall looks good but i see a lot of small problems. i will try to look at this in detail during the coming days.

[bendavid]

@ianchi, if I cherry-pick your regulator patch on top of the current master and build with kernel 4.4 for the c2600, then the cpu clock speed is set properly, but I get corruption issues over wired lan, with ssh connections for example being terminated with errors like
"Corrupted MAC on input."

This doesn't happen without the patch (but then the cpu frequency scaling doesn't work correctly and I think the cpu is stuck at a default low clockspeed).

[bendavid]

Does this router really only have one ethernet controller connected to the switch? Otherwise why not use separate ethernet controllers for wan and lan as for ap148 et al?

[ianchi]

no it has two. But I've been unable to make them work.
This is the only way I've got wired eth working.
Any ideas to try?

[bendavid]

Well what happens if you use
ucidef_add_switch "switch0" \ "1:lan" "2:lan" "3:lan" "4:lan" "6@eth1" "5:wan" "0@eth0" ;;
as above?

[ianchi]

@bendavid I've been running with the regulator patch for more than a day now, with many SSH sessions and I've never got that error.

I've compared with your C2600 dts and we do have some differences, don´t know if they might be causing the trouble (and my problem with eth1 not coming up):

In phy0: ethernet-phy@0 I have one different qca,ar8327-initvals:
C2600: 0x000e4 0xaa545 /* MAC_POWER_SEL /
EA8500: 0x000e4 0x6a545 / MAC_POWER_SEL */

Besides in gmac1: ethernet@37200000 {
I have some additional properties:
qcom,phy_mdio_addr = <4>;
qcom,poll_required = <1>;
qcom,rgmii_delay = <0>;
qcom,emulation = <0>;

When I find some time I'll try using your settings.

[ianchi]

@blogic please any changes you see are needed, pls just tell me and I'll gladly correct them.

[bendavid]

Ok, yes, for the c2600 I just copied the gmac sections verbatim from the ap148 dts. Latest version is here I guess:
https://git.lede-project.org/?p=source.git;a=blob;f=target/linux/ipq806x/patches-4.4/708-ARM-dts-qcom-add-gmac-nodes-to-ipq806x-platforms.patch;h=691ebb66baef2387e6c794b88e3e0c705661f8ec;hb=HEAD#l37

[ianchi]

Sorry, for the bug.
Fixed the commit, and now it is working again.
Please @heinzek and @ryao can you re test it.

[heinzek]

@ianchi firmware for C2600 booting correctly.

[bendavid]

@heinzek do you have the same problem I had with memory and/or wired ethernet corruption? Should be reproducible just ssh'ing in and running dmesg or logread repeatedly (or anything which produces a non-trivial amount of output on the console)

[heinzek]

Yes. Same problem on C2600:
on linux ssh:

Corrupted MAC on input.
ssh_dispatch_run_fatal: Connection to 192.168.1.1: message authentication code incorrect

On windows putty
Incorrect MAC received on packet

[bendavid]

Yup, actually one thing I didn't try is whether the same occurs over a wireless connection (could narrow down whether the corruption is occurring in memory or specifically in the wired ethernet connection)

[ianchi]

I think we should try to commit this branch so that from there we work on a unified code to fix the pending issues.
If the smb patch is causing trouble in your device, we can leave that out for the meantime while we figure it out, and then re submit it in a new PR.
Alternatively, you can disable the regulator in the C2600's DT.

The bootargs patch still needs a definition, and I believe should be made the same on all three targets using it (that was why I send it as a "general" patch).
If we take the new approach, I can submit the changes to the other platforms, but I don't have the hardware to test them.
If we keep the old patch, we must find another way not to break C2600.
@blogic what do you think?

[blogic]

i just pushed everything from last week prior to the boot mangle patch. please rebase any pending work on the current state and send a new PR. i will close this one now. thanks for the effort and sorry for taking so long to process it.