Running a modern kernel on the PandaBoard
Several years ago TI began offering the PandaBoard, a dual-core ARM7 development board based around the OMAP 4430. The relatively fast OMAP made this board an excellent substrate for development (in my case, hacking on the Glasgow Haskell Compiler). Sadly, the group offering the board has since dissolved, leaving me with two PandaBoards with little software support. Moreover, older kernels lack support for much of what this hardware is capable of. This write-up will be my notes for trying to bring up a modern kernel (3.12) on the PandaBoard. I’ll be building on a Debian Wheezy installation installed in the usual way.
Increasingly, new kernels rely on DeviceTree for their configuration.
First, look up the load address of the existing kernel image,
ben@panda2:/boot/uboot$ mkimage -l uImage
Image Name: 3.7.10-x13
Created: Mon Nov 11 11:24:01 2013
Image Type: ARM Linux Kernel Image (uncompressed)
Data Size: 3413776 Bytes = 3333.77 kB = 3.26 MB
Load Address: 80008000
Entry Point: 80008000Now configure and compile the kernel,
$ scp panda2:/boot/config-3.7.10-x13 .config
$ make CROSS_COMPILE=arm-linux-gnueabi- ARCH=arm menuconfig
$ make CROSS_COMPILE=arm-linux-gnueabi- ARCH=arm LOADADDR=0x80008000 uImage modules
$ make CROSS_COMPILE=arm-linux-gnueabi- ARCH=arm tarbz2-pkg
$ make CROSS_COMPILE=arm-linux-gnueabi- ARCH=arm dtbsMove the new kernel to the target,
$ scp linux-3.12.0-arm.tar.bz2 panda2:
$ scp arch/arm/boot/uImage panda2:
$ scp arch/arm/boot/zImage panda2:
$ scp arch/arm/boot/dts/omap4-panda-a4.dtb panda2:Now moving back to the target. Extract the kernel package,
$ sudo tar -C / -jxf linux-3.12.0-arm.tar.bz2
$ sudo cp zImage /boot/uboot/zImage-new
$ sudo cp omap4-panda-a4.dtb /boot/uboot/dtbs/
$ sudo update-initramfs -c -k 3.12.0
$ sudo cp /boot/initrd.img-3.12.0 /boot/uboot/initrd-new.imgOn Debian 7 the bootloader, u-boot, reads its configuration from /boot/uboot/uEnv.txt. We’ll make our changes to a copy of this instead something goes terribly wrong, as it usually does,
$ sudo -i
$ cd /boot/uboot
$ cp uEnv.txt uEnv-new.txt
$ vim uEnv-new.txtEdit the following,
set
kernel_filetozImage-newset
initrd_filetozinitrd-new.imgensure that the “New Device Tree Boot” command is used. This looks something like,
uenvcmd=run boot_fdt; run device_args; bootz 0x80300000 0x81600000:${initrd_size} 0x815f0000
Now the moment of truth,
$ sudo rebootCatch the bootloader on the serial console before it begins booting the old kernel.
Panda # setenv loadbootenv load mmc ${mmcdev}:${mmcpart} ${loadaddr} uEnv-new.txt
Panda # bootOf course, things don’t boot,
Panda # setenv loadbootenv load mmc ${mmcdev}:${mmcpart} ${loadaddr} uEnv-new.txt
Panda # boot
mmc0 is current device
SD/MMC found on device 0
reading boot.scr
** Unable to read file boot.scr **
reading uEnv-new.txt
1161 bytes read in 5 ms (226.6 KiB/s)
Importing environment from mmc (uEnv.txt)...
Checking if uenvcmd is set ...
Running uenvcmd ...
reading zImage-new
3663184 bytes read in 174 ms (20.1 MiB/s)
reading initrd-new.img
3245123 bytes read in 161 ms (19.2 MiB/s)
reading /dtbs/omap4-panda-a4.dtb
18920 bytes read in 9 ms (2 MiB/s)
Kernel image @ 0x80300000 [ 0x000000 - 0x37e550 ]
## Flattened Device Tree blob at 815f0000
Booting using the fdt blob at 0x815f0000
Loading Ramdisk to bfbe4000, end bfefc443 ... OK
Using Device Tree in place at 815f0000, end 815f79e7
Starting kernel ...And there it hangs. Strange that it doesn’t output any useful error. Confirmed that CONFIG_EARLY_PRINTK is set. UART3 is selected as the low-level debugging output port. Let’s ensure that early printk is really enabled. Add earlyprintk to optargs in uEnv-new.txt and reboot.
At least we have something now,
Starting kernel ...
[ 0.000000] Booting Linux on physical CPU 0x0
[ 0.000000] Initializing cgroup subsys cpuset
[ 0.000000] Initializing cgroup subsys cpu
[ 0.000000] Initializing cgroup subsys cpuacct
[ 0.000000] Linux version 3.12.0 (ben@ben-server) (gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-1ubuntu1) ) #8 SMP Sun Nov 17 19:54:16 EST 2013
[ 0.000000] CPU: ARMv7 Processor [411fc093] revision 3 (ARMv7), cr=10c5387d
[ 0.000000] CPU: PIPT / VIPT nonaliasing data cache, VIPT aliasing instruction cache
[ 0.000000] Machine: Generic OMAP4 (Flattened Device Tree), model: TI OMAP4 PandaBoard
[ 0.000000] bootconsole [earlycon0] enabled
[ 0.000000] Memory policy: ECC disabled, Data cache writealloc
[ 0.000000] OMAP4430 ES2.3
[ 0.000000] PERCPU: Embedded 9 pages/cpu @c138a000 s13376 r8192 d15296 u36864
[ 0.000000] Built 1 zonelists in Zone order, mobility grouping on. Total pages: 260368
[ 0.000000] Kernel command line: console=ttyO2,115200n8 console=tty0 earlyprintk vram=16M omapfb.mode=dvi:1280x720MR-16@60 omapdss.def_disp=dvi root=/dev/mmcblk0p5 ro rootfstype=ext4 rootwait fixrtc
[ 0.000000] PID hash table entries: 4096 (order: 2, 16384 bytes)
[ 0.000000] Dentry cache hash table entries: 131072 (order: 7, 524288 bytes)
[ 0.000000] Inode-cache hash table entries: 65536 (order: 6, 262144 bytes)
[ 0.000000] allocated 2097152 bytes of page_cgroup
[ 0.000000] please try 'cgroup_disable=memory' option if you don't want memory cgroups
[ 0.000000] Memory: 1021024K/1047552K available (6967K kernel code, 608K rwdata, 2864K rodata, 441K init, 828K bss, 26528K reserved, 265732K highmem)
[ 0.000000] Virtual kernel memory layout:
[ 0.000000] vector : 0xffff0000 - 0xffff1000 ( 4 kB)
[ 0.000000] fixmap : 0xfff00000 - 0xfffe0000 ( 896 kB)
[ 0.000000] vmalloc : 0xf0000000 - 0xff000000 ( 240 MB)
[ 0.000000] lowmem : 0xc0000000 - 0xef800000 ( 760 MB)
[ 0.000000] pkmap : 0xbfe00000 - 0xc0000000 ( 2 MB)
[ 0.000000] modules : 0xbf000000 - 0xbfe00000 ( 14 MB)
[ 0.000000] .text : 0xc0008000 - 0xc09a1f0c (9832 kB)
[ 0.000000] .init : 0xc09a2000 - 0xc0a10440 ( 442 kB)
[ 0.000000] .data : 0xc0a12000 - 0xc0aaa0a0 ( 609 kB)
[ 0.000000] .bss : 0xc0aaa0a0 - 0xc0b791b8 ( 829 kB)
[ 0.000000] SLUB: HWalign=64, Order=0-3, MinObjects=0, CPUs=2, Nodes=1
[ 0.000000] Hierarchical RCU implementation.
[ 0.000000] NR_IRQS:16 nr_irqs:16 16
[ 0.000000] OMAP clockevent source: timer1 at 32768 Hz
[ 0.000000] sched_clock: 32 bits at 32kHz, resolution 30517ns, wraps every 131071999ms
[ 0.000000] OMAP clocksource: 32k_counter at 32768 Hz
[ 0.000000] smp_twd: clock not found -2
[ 0.000000] Console: colour dummy device 80x30
[ 0.000000] console [tty0] enabled, bootconsole disabledLet’s try removing console=tty0 from optargs. Why was it there to begin with?
Got much father now,
...
[ 1.734283] NET: Registered protocol family 2
[ 1.739685] TCP established hash table entries: 8192 (order: 4, 65536 bytes)
[ 1.747222] TCP bind hash table entries: 8192 (order: 4, 65536 bytes)
[ 1.754150] TCP: Hash tables configured (established 8192 bind 8192)
[ 1.760864] TCP: reno registered
[ 1.764312] UDP hash table entries: 512 (order: 2, 16384 bytes)
[ 1.770538] UDP-Lite hash table entries: 512 (order: 2, 16384 bytes)
[ 1.777648] NET: Registered protocol family 1
[ 1.782623] RPC: Registered named UNIX socket transport module.
[ 1.788848] RPC: Registered udp transport module.
[ 1.822937] [ffbe4000] *pgd=00000000
[ 1.826690] Internal error: Oops: 5 [#1] SMP ARM
[ 1.831512] Modules linked in:
[ 1.834747] CPU: 1 PID: 1 Comm: swapper/0 Not tainted 3.12.0 #8
[ 1.840911] task: ee86c000 ti: ee878000 task.ti: ee878000
[ 1.846588] PC is at unpack_to_rootfs+0x9c/0x28c
[ 1.851409] LR is at unpack_to_rootfs+0x50/0x28c
[ 1.856262] pc : [<c09a43ac>] lr : [<c09a4360>] psr: 20000113
[ 1.856262] sp : ee879e70 ip : a0000113 fp : 00000000
[ 1.868225] r10: c09f8530 r9 : c09a2508 r8 : c09f8530
[ 1.873687] r7 : ffbe4000 r6 : 00318443 r5 : 00000000 r4 : c0b526ac
[ 1.880493] r3 : 00000000 r2 : 00000000 r1 : 00000000 r0 : eea4a000
[ 1.887268] Flags: nzCv IRQs on FIQs on Mode SVC_32 ISA ARM Segment kernel
[ 1.894866] Control: 10c5387d Table: 8000404a DAC: 00000015
[ 1.900878] Process swapper/0 (pid: 1, stack limit = 0xee878248)
[ 1.907135] Stack: (0xee879e70 to 0xee87a000)
[ 1.911682] 9e60: c09a2508 00000000 00000000 c0685114
[ 1.920196] 9e80: c08895b0 ee879e9c c0aaa14c 00000000 c0a0c3a0 c0aaa100 c09a4a4c c09a2508
[ 1.928710] 9ea0: 00000000 c09a4a9c ee905000 c0915d97 c0a83684 c0a81de0 eea31300 00000000
[ 1.937194] 9ec0: c09df3c0 c0a83684 00000000 c0a0c39c c0aaa100 c09df3c0 c09a2508 00000000
[ 1.945709] 9ee0: 00000000 c03f5c88 00000005 c09f849c c0a0c3a0 c0aaa100 c09a4a4c c09a2508
[ 1.954193] 9f00: 00000000 c0008908 ee8efe00 c069d1a0 c0ad081c c0aaa100 00000000 c0a62b44
[ 1.962707] 9f20: 60000113 00000001 000000fe c13870d2 00000000 c00576b0 ee879f64 c096012c
[ 1.971191] 9f40: c095f14c 000000fe 00000005 00000005 00000001 00000005 00000005 c09f849c
[ 1.979675] 9f60: c0a0c3a0 c0aaa100 000000fe c09a2508 c09f84ac c09a2c34 00000005 00000005
[ 1.988189] 9f80: c09a2508 00000000 00000000 c0683738 00000000 00000000 00000000 00000000
[ 1.996673] 9fa0: 00000000 c0683740 00000000 c000ded8 00000000 00000000 00000000 00000000
[ 2.005187] 9fc0: 00000000 00000000 00000000 00000000 00000000 00000000 00000000 00000000
[ 2.013671] 9fe0: 00000000 00000000 00000000 00000000 00000013 00000000 00000000 00000000
[ 2.022186] [<c09a43ac>] (unpack_to_rootfs+0x9c/0x28c) from [<c09a4a9c>] (populate_rootfs+0x50/0x21c)
[ 2.031768] [<c09a4a9c>] (populate_rootfs+0x50/0x21c) from [<c0008908>] (do_one_initcall+0x94/0x138)
[ 2.041259] [<c0008908>] (do_one_initcall+0x94/0x138) from [<c09a2c34>] (kernel_init_freeable+0x160/0x230)
[ 2.051300] [<c09a2c34>] (kernel_init_freeable+0x160/0x230) from [<c0683740>] (kernel_init+0x8/0xe4)
[ 2.060791] [<c0683740>] (kernel_init+0x8/0xe4) from [<c000ded8>] (ret_from_fork+0x14/0x3c)
[ 2.069458] Code: e1c821f0 e1a0a008 e5881000 ea00004a (e5d73000)
[ 2.075836] ---[ end trace 0935e139c88aa1b3 ]---
[ 2.080780] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
[ 2.080780]
[ 2.090362] CPU0: stopping
[ 2.093231] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G D 3.12.0 #8
[ 2.100433] [<c0014660>] (unwind_backtrace+0x0/0xe0) from [<c0011478>] (show_stack+0x10/0x14)
[ 2.109313] [<c0011478>] (show_stack+0x10/0x14) from [<c068a3e4>] (dump_stack+0x68/0x84)
[ 2.117706] [<c068a3e4>] (dump_stack+0x68/0x84) from [<c0012db8>] (handle_IPI+0xc0/0x124)
[ 2.126220] [<c0012db8>] (handle_IPI+0xc0/0x124) from [<c0008698>] (gic_handle_irq+0x58/0x60)
[ 2.135070] [<c0008698>] (gic_handle_irq+0x58/0x60) from [<c068f400>] (__irq_svc+0x40/0x50)
[ 2.143768] Exception stack(0xc0a13f70 to 0xc0a13fb8)
[ 2.149047] 3f60: ffffffed 00000000 0097d000 00000000
[ 2.157531] 3f80: c0a12000 c0a12000 c0699364 c1387040 8000406a 411fc093 00000000 00000000
[ 2.166046] 3fa0: 00000000 c0a13fb8 c000ea14 c000ea18 60000113 ffffffff
[ 2.172943] [<c068f400>] (__irq_svc+0x40/0x50) from [<c000ea18>] (arch_cpu_idle+0x30/0x38)
[ 2.181549] [<c000ea18>] (arch_cpu_idle+0x30/0x38) from [<c007c7a4>] (cpu_startup_entry+0x134/0x1cc)
[ 2.191040] [<c007c7a4>] (cpu_startup_entry+0x134/0x1cc) from [<c09a2a78>] (start_kernel+0x2d8/0x334)
[ 2.200592] [<c09a2a78>] (start_kernel+0x2d8/0x334) from [<80008074>] (0x80008074)There is evidence to suggest that this is due to someone trampling on the initramfs. Let’s try without an initrd,
uenvcmd=run boot_fdt; run device_args; bootz 0x80300000 0x815f0000in /boot/uboot/uEnv-new.txt.
Things boot fine. It seems likely that the devicetree is trampling on the initrd. Perhaps not, even pushing the initrd back to 0x82800000 doesn’t help. This appears to be a big hint. Interesting,
Panda # setenv loadbootenv load mmc ${mmcdev}:${mmcpart} ${loadaddr} uEnv-new.txt
Panda # setenv initrd_high 30000000
Panda # boot
mmc0 is current device
SD/MMC found on device 0
reading boot.scr
** Unable to read file boot.scr **
reading uEnv-new.txt
1128 bytes read in 5 ms (219.7 KiB/s)
Importing environment from mmc (uEnv.txt)...
Checking if uenvcmd is set ...
Running uenvcmd ...
reading zImage-new
3663184 bytes read in 174 ms (20.1 MiB/s)
reading initrd.img
3182391 bytes read in 152 ms (20 MiB/s)
reading /dtbs/omap4-panda-a4.dtb
18920 bytes read in 9 ms (2 MiB/s)
Kernel image @ 0x80300000 [ 0x000000 - 0x37e550 ]
## Flattened Device Tree blob at 815f0000
Booting using the fdt blob at 0x815f0000
ERROR: Failed to allocate 0x308f37 bytes below 0x30000000.
ramdisk - allocation error
FDT creation failed! hanging...### ERROR ### Please RESET the board ###Let’s try updating u-boot,
$ git clone
$ make CROSS_COMPILE=arm-linux-gnueabi- omap4_panda_config
$ make CROSS_COMPILE=arm-linux-gnueabi-
$ scp u-boot.img panda2.local:Move it into place on target,
$ ssh panda2.local
$ cd /boot/uboot
$ cp u-boot.img u-boot.img.old
$ cp ~/u-boot.img u-boot.img
$ rebootTrying booting again (this time without setting initrd_high),
U-Boot# setenv loadbootenv load mmc ${mmcdev}:${mmcpart} ${loadaddr} uEnv-new.txt
U-Boot# boot
mmc0 is current device
SD/MMC found on device 0
reading boot.scr
** Unable to read file boot.scr **
reading uEnv-new.txt
1128 bytes read in 5 ms (219.7 KiB/s)
Importing environment from mmc0 ...
Running uenvcmd ...
reading zImage-new
3663184 bytes read in 173 ms (20.2 MiB/s)
reading initrd.img
3182391 bytes read in 151 ms (20.1 MiB/s)
reading /dtbs/omap4-panda-a4.dtb
18920 bytes read in 9 ms (2 MiB/s)
Kernel image @ 0x80300000 [ 0x000000 - 0x37e550 ]
Wrong Ramdisk Image Format
Ramdisk image is corrupt or invalid
** Invalid partition 2 **Hmm, but even stock kernel fails with the same image. Adding uboot header with,
$ tools/mkimage -A arm -O linux -T ramdisk -a 0x81650000 -d /media/ben/BOOT/initrd-new.img /media/ben/BOOT/uinitrd-new.imgUnfortunately this allows uboot to boot again but kernel crashes at the same point while unpacking initramfs. Perhaps initrd_high really does need to be overridden.
Setting initrd_high=30000000 brings me back to u-boot failing with an allocation failure. Looking at TRM it seems there is no usable memory below 0x30000000; I suspect the address given in the documentation is simply inapplicable here. Using initrd_high=90000000 allows system to boot.
Sadly, while my PandaBoard A4 works fine, the PandaBoard ES appears to hang shortly before starting init. Reported this to the list. It seems that this is due to cpuidle; this patch appears to fix the issue.
After this things seem quite useable, although there is the occassional warning of the form,
gic_timer_retrigger: lost localtimer interruptAs Tony Lindgren points out there are a few things that are known to be broken at the moment,