Re: BUG: NETDEV WATCHDOG -> Badness in gianfar driver?

["Matvejchikov Ilya" <matvejchikov@gmail.com>]

Good Day!

> The system was running fine for many days and weeks without any problems.
> However, just a few minutes ago I noticed a single clicking sound of the harddisk
> (like a head recalibration), so I checked the system.
> I couldn't connect to it via ssh anymore, but via a serial console I got
> at least endless messages as shown below...
> After a reboot, everything looks fine again but the kernel log grew up
> to several MBytes. I pasted the hopefully interesting part below.
>
> Any ideas of what could be wrong there? I think there could be a problem
> in the gianfar network driver. Or is there a physical problem with the PHY
> (a Marvell MV88E1111)?
>
> Any recommendations of how to debug that thingy?
>
> Jan  1 03:43:23 ecam kernel: ------------[ cut here ]------------
> Jan  1 03:43:23 ecam kernel: Badness at c003d3d8 [verbose debug info unavailable]
> Jan  1 03:43:23 ecam kernel: Call Trace:
> Jan  1 03:43:23 ecam kernel: [C0355C70] [C0008FE0] show_stack+0x3c/0x194 (unreliable)
> Jan  1 03:43:23 ecam kernel: [C0355CA0] [C0135594] report_bug+0xa4/0xac
> Jan  1 03:43:23 ecam kernel: [C0355CB0] [C0003784] program_check_exception+0x2b8/0x460
> Jan  1 03:43:23 ecam kernel: [C0355CD0] [C0002908] ret_from_except_full+0x0/0x4c
> Jan  1 03:43:23 ecam kernel: [C0355D90] [C017B3CC] marvell_ack_interrupt+0x14/0x38
> Jan  1 03:43:23 ecam kernel: [C0355DB0] [C01764A8] stop_gfar+0x54/0xd0
> Jan  1 03:43:23 ecam kernel: [C0355DD0] [C01773D0] gfar_timeout+0x5c/0x68
> Jan  1 03:43:23 ecam kernel: [C0355DE0] [C020A060] dev_watchdog+0x110/0x118
> Jan  1 03:43:23 ecam kernel: [C0355E00] [C0024228] run_timer_softirq+0x148/0x1a8
> Jan  1 03:43:23 ecam kernel: [C0355E40] [C002006C] __do_softirq+0x78/0xe4
> Jan  1 03:43:23 ecam kernel: [C0355E70] [C0007054] do_softirq+0x54/0x58
> Jan  1 03:43:23 ecam kernel: [C0355E80] [C001FE4C] irq_exit+0x48/0x58
> Jan  1 03:43:23 ecam kernel: [C0355E90] [C0004000] timer_interrupt+0x17c/0x224
> Jan  1 03:43:23 ecam kernel: [C0355ED0] [C0002954] ret_from_except+0x0/0x18
> Jan  1 03:43:23 ecam kernel: [C0355F90] [C0009FB8] cpu_idle+0xc0/0xd0
> Jan  1 03:43:23 ecam kernel: [C0355FB0] [C0001A7C] rest_init+0x28/0x38
> Jan  1 03:43:23 ecam kernel: [C0355FC0] [C03568E4] start_kernel+0x220/0x29c
> Jan  1 03:43:23 ecam kernel: [C0355FF0] [C0000388] skpinv+0x2b8/0x2f4
> Jan  1 03:43:23 ecam kernel: ------------[ cut here ]------------
> Jan  1 03:43:23 ecam kernel: Badness at c003d3d8 [verbose debug info unavailable]
> Jan  1 03:43:23 ecam kernel: Call Trace:
> Jan  1 03:43:23 ecam kernel: [C0355C70] [C0008FE0] show_stack+0x3c/0x194 (unreliable)
> [...repeating forever...]
> ----- 8< ----- cut here

This is because gfar_timeout() calls stop_gfar() that calls
phy_write() that must not be called from interrupt context. See
comments to this function.

Best regards,
Matvejchikov Ilya.