ar71xx: ag71xx minor fixes that help debugging

[danielfdickinson]

This is a trio of minor fixes/tweaks for ar71xx version of ag71xx that help debugging when working with the driver and/or modifying board definitions for ag71xx boards.

[blogic]

no need for the kernel rev #if. we use >=4.9 rest of the series is fine

[blogic]

manually fixed the comment and merged the series, Thanks !

[hnyman]

@blogic @cshoredaniel

I propose that the commits from this PR get reverted, until they are fixed.
(Using the merged commit hashes below)

• commits 0b9f4e8 and 7a3e133 break all connectivity to some ar71xx devices, at least WNDR3800 (and 3700v1 and 3700v2)
• I already pointed out separately earlier that also the third commit 2655fbe has faulty syntax.

I have problems connecting to my ar71xx WNDR3800 (ar7161 based) after compiling the current master r7713-a2a225517d. The reason seems to be the switch changes made by this PR.

Instead of having both eth0 and eth1, kernel log has an error message "Missing PHY for ag71xx.0" from this PR, and the normal eth1 is assigned to be eth0. No eth1 is defined. That pretty much breaks all connectivity to the router.

kernel log with today's build r7713-a2a225517d

[    3.140448] rtl8366s rtl8366s: using GPIO pins 5 (SDA) and 7 (SCK)
[    3.146773] rtl8366s rtl8366s: RTL8366 ver. 1 chip found
[    3.210298] libphy: rtl8366s: probed
[    3.214482] libphy: Fixed MDIO Bus: probed
[    3.551902] Missing PHY for ag71xx.0
[    3.892725] ag71xx ag71xx.1: connected to PHY at rtl8366s:04 [uid=001cc960, driver=Generic PHY]
[    3.902051] eth0: Atheros AG71xx at 0xba000000, irq 5, mode:RGMII
[    3.909536] NET: Registered protocol family 10

kernel log with 29 July 2018 build r7581-7880a6f7fe:

[    3.140732] rtl8366s rtl8366s: using GPIO pins 5 (SDA) and 7 (SCK)
[    3.146995] rtl8366s rtl8366s: RTL8366 ver. 1 chip found
[    3.209997] libphy: rtl8366s: probed
[    3.214134] libphy: Fixed MDIO Bus: probed
[    3.552250] eth0: Atheros AG71xx at 0xb9000000, irq 4, mode:RGMII
[    3.892480] ag71xx ag71xx.1: connected to PHY at rtl8366s:04 [uid=001cc960, driver=Generic PHY]
[    3.901801] eth1: Atheros AG71xx at 0xba000000, irq 5, mode:RGMII
[    3.909248] NET: Registered protocol family 10

Build r7581-7880a6f7fe from 29 July 2018 worked ok, and that is right before this PR got merged.

Reverting both of these commits fixes things:
0b9f4e8 ar71xx: ag71xx: Add connect message: fixed phy
7a3e133 ar71xx: ag71xx: Prevent kernel oops for board def

Log after reverting matches again the expectations and the router works:

[    3.140193] rtl8366s rtl8366s: using GPIO pins 5 (SDA) and 7 (SCK)
[    3.146516] rtl8366s rtl8366s: RTL8366 ver. 1 chip found
[    3.209889] libphy: rtl8366s: probed
[    3.214063] libphy: Fixed MDIO Bus: probed
[    3.552167] eth0: Atheros AG71xx at 0xb9000000, irq 4, mode:RGMII
[    3.892389] ag71xx ag71xx.1: connected to PHY at rtl8366s:04 [uid=001cc960, driver=Generic PHY]
[    3.901706] eth1: Atheros AG71xx at 0xba000000, irq 5, mode:RGMII
[    3.909137] NET: Registered protocol family 10

If only the "Add connect message: fixed phy" commit is reverted, then a kernel OOPS follows. Log below.

I think that the "Add connect message: fixed phy" tries to give a too informative error message and some information is not found with all devices. That leads to OOPS. The second commit "Prevent kernel oops for board def" tries to mitigate that, but at the same time stops the execution of the function too early.

Log of the OOPS

[    3.134539] Realtek RTL8366S ethernet switch driver version 0.2.2
[    3.140658] rtl8366s rtl8366s: using GPIO pins 5 (SDA) and 7 (SCK)
[    3.146983] rtl8366s rtl8366s: RTL8366 ver. 1 chip found
[    3.210307] libphy: rtl8366s: probed
[    3.214492] libphy: Fixed MDIO Bus: probed
[    3.551943] CPU 0 Unable to handle kernel paging request at virtual address 0000002c, epc == 80286f30, ra == 80285894
[    3.562549] Oops[#1]:
[    3.564824] CPU: 0 PID: 1 Comm: swapper Not tainted 4.9.116 #0
[    3.570638] task: 87c2c000 task.stack: 87c30000
[    3.575155] $ 0   : 00000000 00000000 00000000 000003e8
[    3.580401] $ 4   : 80484cd8 00000000 80450000 0000003f
[    3.585648] $ 8   : 87c31fe0 0000d400 00000000 00000004
[    3.590896] $12   : 0018c25d 19a2c84c 00000001 00000000
[    3.596143] $16   : 87c7cc60 80484dc0 80484dc0 00000200
[    3.601391] $20   : 80484cd8 80490000 87c7cc60 87c7cd2c
[    3.606637] $24   : 8047e834 523b05d3
[    3.611885] $28   : 87c30000 87c31cd0 804f0000 80285894
[    3.617132] Hi    : 00000064
[    3.619999] Lo    : ebf00000
[    3.622882] epc   : 80286f30 ag71xx_phy_connect_fixed+0x80/0xc4
[    3.628783] ra    : 80285894 ag71xx_probe+0x3e4/0x588
[    3.633818] Status: 1000d403 KERNEL EXL IE
[    3.638019] Cause : 00800008 (ExcCode 02)
[    3.642018] BadVA : 0000002c
[    3.644885] PrId  : 00019374 (MIPS 24Kc)
[    3.648789] Modules linked in:
[    3.651842] Process swapper (pid: 1, threadinfo=87c30000, task=87c2c000, tls=00000000)
[    3.659732] Stack : 80484cc8 80484dc0 87c7c800 80484cc8 80484dc0 00000200 80484cd8 87c7c800
[    3.668132]         80484cc8 80285894 19000000 00000001 87d5fb40 00000000 00000000 87c7c800
[    3.676531]         804f0000 80183d10 803c0d04 80484cd8 8049c6e4 80484d0c 80520000 8049c6e4
[    3.684929]         00000000 804a5304 804f0000 80249e8c 87d5e900 87d5e900 87c0e4b0 80183a58
[    3.693329]         80484cd8 80484cd8 80520000 80248384 00000000 80490000 87c5de30 87c31dc0
[    3.701728]         ...
[    3.704180] Call Trace:
[    3.706617] [<80286f30>] ag71xx_phy_connect_fixed+0x80/0xc4
[    3.712179] [<80285894>] ag71xx_probe+0x3e4/0x588
[    3.716874] [<80249e8c>] platform_drv_probe+0x2c/0x70
[    3.721925] [<80248384>] really_probe+0x110/0x278
[    3.726622] [<80248590>] __driver_attach+0xa4/0xdc
[    3.731401] [<80246404>] bus_for_each_dev+0x5c/0xbc
[    3.736267] [<802478b0>] bus_add_driver+0xf4/0x214
[    3.741049] [<80248c40>] driver_register+0xac/0xfc
[    3.745844] [<804ca938>] ag71xx_module_init+0x2c/0x50
[    3.750883] [<80060a5c>] do_one_initcall+0xdc/0x1a0
[    3.755761] [<804a5d0c>] kernel_init_freeable+0x14c/0x20c
[    3.761161] [<803a8198>] kernel_init+0x10/0x104
[    3.765687] [<800668d8>] ret_from_kernel_thread+0x14/0x1c
[    3.771071] Code: ae0300e4  8e230008  ae0300e0 <8c46002c> 14c00002  00000000  8c460008  8c4300f0  8c4700f4
[    3.780861]
[    3.782373] ---[ end trace 246723ef878a10cb ]---
[    3.788017] Kernel panic - not syncing: Fatal exception
[    3.794091] Rebooting in 1 seconds..

[danielfdickinson]

I believe the issue is devices like @hnyman WNDR that have non ag71xxx switch attached. I don't think it's worth trying to make them work; they're just tweaks that were supposed to make life easier. Sorry.