| 14:20 | dbsx | Someone to help with 3.11-rc6. |
| 14:20 | dbsx | PC is at io_serial_out+0x8/0x10 |
| 14:20 | dbsx | LR is at serial8250_clear_fifos.part.14+0x18/0x44 |
| 14:22 | _rmk_ | can you pastebin it please? |
| 14:25 | dbsx | http://pastebin.com/W13jSYrH |
| 14:26 | _rmk_ | thanks |
| 14:27 | _rmk_ | I think its a bug in your userspace... |
| 14:27 | _rmk_ | 244. Loading the saved-state of the serial devices... |
| 14:27 | _rmk_ | 245. /dev/ttyS0 at 0x0000 (irq = 7) is a 16550A |
| 14:28 | _rmk_ | so userspace is telling the kernel that there's a serial port at address 0 |
| 14:29 | _rmk_ | it even tells it that it's a valid 16550A port too |
| 14:29 | _rmk_ | so the kernel then tries to initialize the port there, and oopses |
| 14:30 | _rmk_ | I'd expect that this happens with all kernels |
| 14:33 | dbsx | hmm. Could this be caused by a bad entry in the dove.dtsi? There are only a handful of patches applied (none of them have anything to do with serial/tty) |
| 14:34 | _rmk_ | this is the one created by DT: |
| 14:34 | _rmk_ | 104. f1012000.serial: ttyS0 at MMIO 0xf1012000 (irq = 7) is a 16550A |
| 14:34 | _rmk_ | which is correct |
| 14:34 | _rmk_ | the other is being created by userspace |
| 14:35 | _rmk_ | 245. /dev/ttyS0 at 0x0000 (irq = 7) is a 16550A |
| 14:35 | _rmk_ | isn't a kernel message, but a message from setserial |
| 14:36 | _rmk_ | my guess is that setserial saved the address information incorrectly to a file, and is trying to restore that information |
| 14:37 | _rmk_ | yep, it does |
| 14:37 | _rmk_ | I just installed it on ubuntu, and it saves this: |
| 14:37 | dbsx | ok, I will go have a look what the Debian boot is doing. |
| 14:37 | _rmk_ | /dev/ttyS0 uart 16550A port 0x0000 irq 7 baud_base 10416666 spd_normal skip_test |
| 14:38 | _rmk_ | which is basically wrong |
| 14:38 | _rmk_ | its saying that you have a 16550A at IO port address 0 |
| 14:39 | _rmk_ | note that setserial comes with this warning: |
| 14:39 | _rmk_ | CAUTION: Using an invalid port can lock up your machine! |
| 14:39 | _rmk_ | and that's what you're hitting |
| 14:40 | _rmk_ | just purge setserial from your system |
| 14:41 | dbsx | thank you |
| 14:42 | _rmk_ | just checking the setserial source, it never saves the MMIO address |
| 14:42 | _rmk_ | in fact, it doesn't understand MMIO ports at all |
| 14:56 | dbsx | I am using debian testing. The config port addresses are 0x0000. So yes we have the problem. The previous running kernel was/is 3.6.9 which was OK. Purged setserial. |
| 14:56 | dbsx | I have a boot. Thanks |
| 15:29 | dbsx | Next question - 3.11-rc6 adds the orion waitchdog. |
| 15:29 | dbsx | I added to the dove.dtsi |
| 15:29 | dbsx | wdt@20300 { |
| 15:29 | dbsx | compatible = "marvell,orion-wdt"; |
| 15:29 | dbsx | reg = <0x20300 0x28>; |
| 15:29 | dbsx | clocks = <&gate_clk 7>; |
| 15:29 | dbsx | status = "okay"; |
| 15:29 | dbsx | }; |
| 15:29 | dbsx | I get "orion_wdt f1020300.wdt: Orion Watchdog missing clock" |
| 15:29 | dbsx | So what should the clocks entry be? |
| 15:40 | _rmk_ | let me look it up from the non-dt stuff... |
| 15:41 | _rmk_ | tclk - the same clock as the i2c and spi devices use |
| 15:43 | _rmk_ | so I think it wants to be: clocks = <&core_clk 0>; |
| 15:43 | dbsx | Will try it and let you knwo |
| 15:43 | dbsx | know |
| 15:45 | _rmk | 15:45 * _rmk_ has a nice shiney 32" screen on the cubox now, which finally lets me see the entire desktop, rather than permanently overscanning all the time like the last display |
| 15:46 | _rmk_ | ... and being a 32", its a HD panel (smaller TVs have panel resolutions below 1080 pixels vertically) |
| 15:52 | _rmk_ | the down-side is that it is shiney - the screen is almost mirror like |
| 16:10 | dbsx | I hate shiny. The orion_wdt works, "orion_wdt: Initial timeout 25 sec" |
| 16:10 | dbsx | Shame the initial timeout is not a module parameter |
| 16:12 | dbsx | I also changed the spi to |
| 16:12 | dbsx | compatible = "st,n25q032"; |
| 16:12 | dbsx | which gets rid of a warning |