pmunts
Posts: 5
Joined: Fri Nov 08, 2013 7:19 pm

Re: Next firmware tree is now 3.10

Fri Nov 08, 2013 7:23 pm

The spidev kernel module in 3.10 is not creating the device nodes /dev/spidev0.0 and /dev/spidev0.1, so anything that requires those won't work. The spidev module loads, and appears to be the same size as that for 3.6, but no device nodes are created.

Nige C
Posts: 46
Joined: Thu Jul 19, 2012 12:22 am
Location: Hadleigh, Suffolk

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 2:13 am

Yep did rpi-update late yesterday afternoon.
Checking old against new kernel as pmunts says no spidev node is created.
Retired power engineer with a lot of real life experience. Life is an adventure again :)

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 11:45 am

pmunts wrote:The spidev kernel module in 3.10 is not creating the device nodes /dev/spidev0.0 and /dev/spidev0.1, so anything that requires those won't work. The spidev module loads, and appears to be the same size as that for 3.6, but no device nodes are created.
Could you create a github issue:
https://github.com/raspberrypi/linux/issues?state=open

The problem will get better visibility there.

MrEngman
Posts: 3859
Joined: Fri Feb 03, 2012 2:17 pm
Location: Southampton, UK

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 2:00 pm

I'm having real problems trying to compile 3.10.18+. When trying to compile I am getting a weird error but don't know why or how to fix it. Error is:-

Code: Select all

 LD arch/arm/mach-bcm2708/built-in.o
 CC kernel/fork.o
 CC kernel/exec_domain.o
 CC kernel/panic.o
 CC kernel/printk.o
 CC kernel/cpu.o
 CC kernel/exit.o
 CC kernel/itimer.o
 HZFILE kernel/hz.bc
 BC kernel/timeconst.h
 /bin/sh: 1: bc: not found
 make[1]: *** [kernel/timeconst.h] Error 127
 make: *** [kernel] Error 2
Found this http://lkml.indiana.edu/hypermail/linux ... 03271.html but doesn't help as patch is already applied.

More details here https://github.com/lwfinger/rtl8188eu/i ... t-28124972

Anyone any ideas how to fix it? Really need to be able to solve this so I can get the 8188eu wifi driver compiled.


MrEngman
Simplicity is a prerequisite for reliability. Edsger W. Dijkstra

Please post ALL technical questions on the forum. Please Do Not send private messages.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 2:08 pm

MrEngman wrote: /bin/sh: 1: bc: not found
What do you get from:

Code: Select all

$ which bc
/usr/bin/bc
$ bc --version
bc 1.06.95
does

Code: Select all

sudo apt-get install bc
help?

MrEngman
Posts: 3859
Joined: Fri Feb 03, 2012 2:17 pm
Location: Southampton, UK

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 2:29 pm

dom wrote:
MrEngman wrote: /bin/sh: 1: bc: not found
does

Code: Select all

sudo apt-get install bc
help?
It certainly does, Dom.

Code: Select all

  LD      arch/arm/mach-bcm2708/built-in.o
  CC      kernel/fork.o
  CC      kernel/exec_domain.o
  CC      kernel/panic.o
  CC      kernel/printk.o
  CC      kernel/cpu.o
  CC      kernel/exit.o
  CC      kernel/itimer.o
  HZFILE  kernel/hz.bc
  BC      kernel/timeconst.h
  CC      kernel/time.o
  CC      kernel/softirq.o
  CC      kernel/resource.o
  CC      kernel/sysctl.o
  CC      kernel/sysctl_binary.o
Thanks very, very much. :D


MrEngman
Simplicity is a prerequisite for reliability. Edsger W. Dijkstra

Please post ALL technical questions on the forum. Please Do Not send private messages.

B.Jay
Posts: 28
Joined: Thu Jul 18, 2013 10:57 pm

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 9:15 pm

dom wrote:Can you confirm if you applied the config options here:
https://github.com/raspberrypi/linux/is ... t-27763584
Sorry for the delay, been a bit busy and testing took a bit of time...

Tried the method you linked me to.
Doesn't work reliably (read: long-term) with the Transcend card, but at least it doesn't corrupt right away.

With an dynamic overclock to 900MHz and making sure the RasPi keeps the CPU clock at the max speed: Running a simple prime-finder in a endless loop which also writes all primes in the range of 2-->100M into a text-file, to cause I/O on the SD card, it works out for about a day. After 20<->26 hours I see filesystem errors popping up in the syslog, which are recovered upon a reboot.

It should also be noted that the card gets overly warm over time ... mSD's in a SD adapter stay a lot 'cooler' than the Transcend SD.

In a short summary:

1. The situation got better as it doesn't totally corrupt right away

2. Corruption still happens, but not to the point where the system bombs boot (read: can't load init or libs anymore therefore freezes completely)

I'd say the Transcend cards run on a tight "I/O Clock" threshold, as this card, in several sizes, seems to give most trouble when being used with overclocked Pi's (and even non-overclocked ones). Using a Kingston /SanDisk / (another reputable Brand) would be the better recommendation than trying to work around this particular card in software.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Sat Nov 09, 2013 10:26 pm

B.Jay wrote: With an dynamic overclock to 900MHz and making sure the RasPi keeps the CPU clock at the max speed: Running a simple prime-finder in a endless loop which also writes all primes in the range of 2-->100M into a text-file, to cause I/O on the SD card, it works out for about a day. After 20<->26 hours I see filesystem errors popping up in the syslog, which are recovered upon a reboot.
Thanks for testing.
Can you post your exact config.txt (and cmdline.txt if anything is non-standard)

B.Jay
Posts: 28
Joined: Thu Jul 18, 2013 10:57 pm

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 6:48 am

dom wrote:Can you post your exact config.txt (and cmdline.txt if anything is non-standard)
From the Transcend SD used in the RasPi for testing the proposed 'SD card corruption fix':

config.txt

Code: Select all

# Make the RasPi get all cuddly with the monitor
disable_overscan=1
hdmi_force_hotplug=1
config_hdmi_boost=4

# Enable dynamic clocking
# (who worries about a 'warranty bit' if the magic smoke escapes)
force_turbo=0

# Set the defaults for the other frequencies (to be sure)
# GPU core, h264, v3d and isp share a PLL
core_freq=250
h264_freq=250
v3d_freq=250
isp_freq=250
# arm, sdram and gpu each have their own PLLs and can have unrelated frequencies
arm_freq_min=700
arm_freq=900
gpu_freq=250
sdram_freq=400

# Proposed SD card corruption fix (needs Kernel 3.10.18+ and matching Firmware)
# https://github.com/raspberrypi/linux/issues/415#issuecomment-27763584
init_emmc_clock=250000000
emmc_pll_core=1
cmdline.txt

Code: Select all

dwc_otg.lpm_enable=0 dwc_otg.speed=1 console=ttyAMA0,115200 kgdboc=ttyAMA0,115200 console=tty1 root=/dev/mmcblk0p2 rootfstype=ext4 elevator=deadline rootwait
Only non-standard switch would be "dwc_otg.speed=1" to make the Logitech K400 behave ("repeating key" problem without this option)

gstreeter
Posts: 106
Joined: Sun Sep 02, 2012 11:11 am
Location: UK

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 8:56 am

I've updated a Pi to kernel 3.10.18+ #585 with rpi-update and found that Strongswan 5.1.1 IPSEC VPN no longer functions. This works fine on the current 3.6.11+ #538 kernel.

The following error message appears in the IPsec log file:

"received netlink error: Address family not supported by protocol (97)"

Looks like there something missing out of the kernel 3.10.18+ build that was present in the 3.6.11 version.

User avatar
DougieLawson
Posts: 36340
Joined: Sun Jun 16, 2013 11:19 pm
Location: Basingstoke, UK
Contact: Website Twitter

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 11:45 am

It looks like /lib/modules/*/kernel/net/ipv4/ipip.ko module may be missing from the latest kernel build.
Note: Having anything humorous in your signature is completely banned on this forum. Wear a tin-foil hat and you'll get a ban.

Any DMs sent on Twitter will be answered next month.

This is a doctor free zone.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 12:28 pm

DougieLawson wrote:It looks like /lib/modules/*/kernel/net/ipv4/ipip.ko module may be missing from the latest kernel build.
Yes - will add.

bodshal
Posts: 1
Joined: Mon Aug 19, 2013 12:42 am
Location: Leeds Point, NJ, USA
Contact: Website

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 2:19 pm

pmunts wrote:The spidev kernel module in 3.10 is not creating the device nodes /dev/spidev0.0 and /dev/spidev0.1, so anything that requires those won't work. The spidev module loads, and appears to be the same size as that for 3.6, but no device nodes are created.
Oh good, it's not me going nuts then. Built a new pi yesterday to drive some SPI light segments and whilst it autoloaded spi_bcm2708, it did not load spidev. Manually loading spidev does not result in the devices, boo hoo.

I don't see an issue on github yet so I'll go make one.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 5:33 pm

DougieLawson wrote:It looks like /lib/modules/*/kernel/net/ipv4/ipip.ko module may be missing from the latest kernel build.
Update is out with these added:
NET_IPIP, IP_ADVANCED_ROUTER, IP_MULTIPLE_TABLES, IP_ROUTE_MULTIPATH, NETFILTER_MATCH_IPVS

and a newer rtl8192cu wifi driver:
https://github.com/raspberrypi/linux/issues/420

gstreeter
Posts: 106
Joined: Sun Sep 02, 2012 11:11 am
Location: UK

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 6:20 pm

The update has fixed the IPSec issue, thanks for the quick turn-around.

Gary

B.Jay
Posts: 28
Joined: Thu Jul 18, 2013 10:57 pm

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 9:57 pm

Okay... seems there's still a problem with SD card corruption and Overclocking going on (latest update installed) ...

config.txt snippet:

Code: Select all

force_turbo=1
init_emmc_clock=500000000
emmc_pll_core=1
core_freq=500
arm_freq_min=700
arm_freq=1000
sdram_freq_min=400
sdram_freq=500
over_voltage=6
That's from the RasPi running "USB Booted" (Apache/MySQL/PHP for TT-RSS): overclocked to 1GHz (with heatsink), RAM +100MHz to 500MHz, Core to +250MHz to 500MHz, Force Turbo off, "SD card fix" applied. I actually once took that overclock figure from the RasPi Wiki.

System runs stable @63°C max under stress, but the SD card (2GB Sony Class 4 with just a vfat partition on it to boot the Pi) keeps on getting reported as being "not cleanly unmounted". Running a "fsck.vfat -w -r /dev/mmcblk0p1" returns no errors (weird), but if I unmount the card manually and mount it again, I get the same "not cleanly unmounted" in dmesg (and fsck.vfat still doesn't show any errors).

Side note: Yes, I triggered the "warranty bit", but who cares - I won't try to claim warranty if I fry the thing (especially not at that price).

According to the "corruption fix" everything should be set correctly according to their proposal.

Also, as kind-of off-topic question into dom's direction (as you're obviously working at Broadcom):
Is there any hardware hack to make the SOC boot off of USB? On my Samsung devices (Android phone/tablet) I can make the OMAP4 boot off of USB by hacking a few resistors to change the boot-order of the SOC - is there a way to make the RasPi do the same stunt (boot the firmware off a VFAT Partition on a USB HDD/Stick)? That would solve a whole lot of "corruption problems" right there.

Sorry if that question has already been asked or if this looks like a stupid question.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 10:06 pm

B.Jay wrote:Okay... seems there's still a problem with SD card corruption and Overclocking going on (latest update installed) ...

System runs stable @63°C max under stress, but the SD card (2GB Sony Class 4 with just a vfat partition on it to boot the Pi) keeps on getting reported as being "not cleanly unmounted". Running a "fsck.vfat -w -r /dev/mmcblk0p1" returns no errors (weird), but if I unmount the card manually and mount it again, I get the same "not cleanly unmounted" in dmesg (and fsck.vfat still doesn't show any errors).
That doesn't sound like corruption, more like fsck is not clearing the "not-clean unmount" bit.
If you insert the sdcard into a windows pc and have it run checkdisk, does this error go away?

USB boot (from a mass storage device) is not possible on a Pi.
However the linux fs can be on USB, and the GPU never writes to the sdcard, so that should be just as safe for corruption avoidance.

B.Jay
Posts: 28
Joined: Thu Jul 18, 2013 10:57 pm

Re: Next firmware tree is now 3.10

Sun Nov 10, 2013 10:46 pm

Unmounted and pulled the card out of the Pi...
My Linux desktop doesn't see anything wrong with the VFAT filesystem (also, no "uncleanly unmounted" message - "dry run" to not correct any problem it may see) and Windows (Windows 7 in a VMware VM, USB Card Reader handed off to the VM for exclusive access) also doesn't see a problem (forced full filesystem check). Putting the card back into the Pi results in a "/dev/mmcblk0p1 had not been cleanly unmounted yadda yadda" entry in dmesg (also during the boot output where the Pi is mounting fstab entires).

For the time being I commented out the SD card mount line in fstab and put a wrapper around "rpi-update" ... it reads the firmware string, rpi-update's, pulls the firmware rev into a new string, compares, and in case it changed mounts the SD card to /mnt and copies /boot/* over. That way the SD isn't permanentely mounted but only for a brief "update" when a new firmware landed.

EDIT: Thanks for the reply to "USB boot hack possible"... I already expected that answer ("not possible").
And yes, the USB-Booted Pi is running rootfs off of a USB HDD, and the "/boot" firmware is also saved into a tar.gz in case the card really becomes unreadable.

EDIT 2: Just tried on the other Pi (boot and rootfs on SD card) with the latest update...
force_turbo=1 and ONLY arm_freq to 900 (no other clocks raised, no over_voltage) works fine. The moment I set to the same overclock as the USB-booted Pi (only difference is "force_turbo=0" to not trigger the warranty bit, which is also against the recommendation of the "fix") the filesystems, both the vfat and ext4 partition, do corrupt from the get-go. This Pi is running a Samsung 32GB Class 10 mSD in a SD adapter. Take it with a fine grain of salt, but with the "fix" it seems things work fine up to 900MHz but start to go astray at 1GHz (and possibly more).

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Mon Nov 11, 2013 12:39 pm

B.Jay wrote:Unmounted and pulled the card out of the Pi...
See:
http://forum.stmlabs.com/showthread.php?tid=11524
http://lkml.org/lkml/2013/5/6/342

The bug is in fsck, and is not an indication of corruption occurring. Windows can clear the "uncleanly mounted" message.

Ste123
Posts: 19
Joined: Sat Aug 11, 2012 7:10 am

Re: Next firmware tree is now 3.10

Mon Nov 11, 2013 1:50 pm

I have been having the repeating key problem since upgrading to 3.10 using my wireless keyboard. Was OK before the update.

Is this logged anywhere?

mcgyver83
Posts: 358
Joined: Fri Oct 05, 2012 11:49 am

Re: Next firmware tree is now 3.10

Mon Nov 11, 2013 2:27 pm

So the fix should be only:
force_turbo=1 -> set the warranty bit, who care :-)
init_emmc_clock=500000000 -> (the same value as cpu_freq value)
emmc_pll_core=1

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Mon Nov 11, 2013 7:26 pm

mcgyver83 wrote:So the fix should be only:
force_turbo=1 -> set the warranty bit, who care :-)
init_emmc_clock=500000000 -> (the same value as cpu_freq value)
emmc_pll_core=1
Yes.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Tue Nov 12, 2013 1:42 pm

Latest firmware now enables emmc_pll_core by default (and handles the emmc clock and turbo directly).
This change certainly solves an sdcard corruption case we were able to reproduce.

As usual, testers are welcome to test.

gstreeter
Posts: 106
Joined: Sun Sep 02, 2012 11:11 am
Location: UK

Re: Next firmware tree is now 3.10

Tue Nov 12, 2013 8:28 pm

Issue after updating to 3.10.18+ #590
FSTRIM -v / fails with "FITRIM ioctl failed: Operation not supported"

Worked ok in previous 3.10.18 updates.

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5354
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Next firmware tree is now 3.10

Tue Nov 12, 2013 8:32 pm

gstreeter wrote:Issue after updating to 3.10.18+ #590
FSTRIM -v / fails with "FITRIM ioctl failed: Operation not supported"

Worked ok in previous 3.10.18 updates.
I don't see how that will have changed with last update. Can you let me know the most recent kernel that worked?
You can get older firmware with:

Code: Select all

sudo rpi-update <git hash>
and you can get the hashes from here:
https://github.com/Hexxeh/rpi-firmware/commits/master

Return to “Advanced users”