Recent git kernels and OSK
I just built a kernel from today's git repo and tried booting it on my OSK. The boot hung after uncompressing the kernel at the "Starting kernel ..." message.
Has anyone successfully booted a recent git kernel on the OSK?
Philip
Quoted 3 lines · Philip Balister
I just built a kernel from today's git repo and tried booting it on my OSK. The boot hung after uncompressing the kernel at the "Starting kernel ..." message.
Even after enabling lowlevel debug, yes?
Quoted 1 line
Has anyone successfully booted a recent git kernel on the OSK?
From OMAP git, no.
From Linus' kernel.org git, yes ... albeit many drivers were missing.
- Dave
Quoted 3 lines · Philip Balister
I just built a kernel from today's git repo and tried booting it on my OSK. The boot hung after uncompressing the kernel at the "Starting kernel ..." message.
Has anyone successfully booted a recent git kernel on the OSK?
Just checked it. Seems that recent git on OSK is really broken. Enabling lowlevel debug last thing I get is kernel command line output [3]. On my last working kernel next output would be clock stuff [2]
My last update some days ago including "Menelaus: Fix voltage values for VMEM" [1] was okay. So anything between this patch and now is the bad one.
Any ideas?
Cheers
Dirk
[2]
Clocks: ARM_SYSST: 0x1000 DPLL_CTL: 0x2813 ARM_CKCTL: 0x150f Clocking rate (xtal/DPLL1/MPU): 12.0/192.0/192.0 MHz
...
[3] Recent git with lowlevel debug enabled on OSK:
5>Linux version 2.6.18-omap1 (dirk@asterix) (gcc version 4.1.0) #2 PREEMPT Tue Oct 3 15:40:42 CEST 2006
CPU: ARM926EJ-S [41069263] revision 3 (ARMv5TEJ), cr=00053177
Machine: TI-OSK
Memory policy: ECC disabled, Data cache writeback
<7>On node 0 totalpages: 8192
<7> DMA zone: 8192 pages, LIFO batch:1
OMAP1611b revision 2 handled as 16xx id: 68058c80eb071005
<6>SRAM: Mapped pa 0x20000000 to va 0xd0000000 size: 0x100000
CPU0: D VIVT write-back cache
CPU0: I cache: 16384 bytes, associativity 4, 32 byte lines, 128 sets
CPU0: D cache: 8192 bytes, associativity 4, 32 byte lines, 64 sets
Built 1 zonelists. Total pages: 8192
<5>Kernel command line: mem=32M console=tty0 console=ttyS0,115200 initrd=0x10400000,8M root=/dev/ram0 rw
Quoted 1 line · Dirk Behme
Philip Balister wrote:
...
Quoted 1 line
Quoted 1 line
Has anyone successfully booted a recent git kernel on the OSK?
...
Quoted 3 lines
Just checked it. Seems that recent git on OSK is really broken. Enabling lowlevel debug last thing I get is kernel command line output [3]. On my last working kernel next output would be clock stuff [2]
Any ideas?
Seems to me that it hangs in arch/arm/mach-omap1/clock.c in omap1_clk_init in
...
/* USB_REQ_EN will be disabled later if necessary (usb_dc_ck) */ reg = omap_readw(SOFT_REQ_REG) & (1 << 4); omap_writew(reg, SOFT_REQ_REG);
if (!cpu_is_omap15xx())
omap_writew(0, SOFT_REQ_REG2);
...
If I remove "omap_writew(reg, SOFT_REQ_REG);" in code above it goes on booting but then crashes later in clk_disable() (as the comment above implies ;) ):
...
Initializing OMAP McBSP system
<6>omap_dsp_init() done
USB: hmc 16, usb0 2 wires
<6>i2c_omap i2c_omap.1: bus 0 rev2.2 at 100 kHz
<2>kernel BUG at arch/arm/plat-omap/clock.c:138!
<1>Unable to handle kernel NULL pointer dereference at virtual address 00000000
...
If I'm correct here (not sure, can anybody reproduce this?) I really wonder why it may hang in (unchanged) write access to SOFT_REQ_REG?
Again, any ideas?
Cheers
Dirk
Quoted 5 lines · Dirk Behme
Quoted 3 lines
Just checked it. Seems that recent git on OSK is really broken. Enabling lowlevel debug last thing I get is kernel command line output [3]. On my last working kernel next output would be clock stuff [2]
Any ideas?
Seems to me that it hangs in
arch/arm/mach-omap1/clock.cinomap1_clk_initin
As already mentioned in my last mail, this seems to be related to broken low level debug.
Doing some more debugging (with low level debug disabled now) shows that at least until "ARM: OMAP: DSPGW: mailbox log cleanup" [1] everything is okay. Then, with all changes including "ARM: OMAP: DSPGW: Peripheral (kfunc) control" [2] it seems to be broken.
So maybe DSPGW changes cause the trouble on OSK. Can anybody with DSPGW knowledge re-check this on OSK?
Many thanks,
Dirk
Quoted 1 line · Philip Balister
Do you mind sending me the sequence you use to prepare the kernel for flashing? I have lost confidence in myself ...
For testing I use uboot TFTP download.
Quoted 1 line
I can't get anything to boot. What is the most recent kernel you have booted on the OSK?
Going back to "ARM: OMAP: DSPGW: mailbox log cleanup" [1] results in a kernel which works for me on OSK. If you have most recent git, try
cg-admin-uncommit -t 4b35036f09f89dcc25b0a770c56e2c71d4937cb4
to go back to this version.
Regards
Dirk
Quoted 1 line · Dirk Behme
Doing some more debugging (with low level debug disabled now) shows that at least until "ARM: OMAP: DSPGW: mailbox log cleanup" [1] everything is okay. Then, with all changes including "ARM: OMAP: DSPGW: Peripheral (kfunc) control" [2] it seems to be broken.
After some additional hours of debugging it's a really simple bug: A strcmp() mismatch which results in omap_dsp->mbox being a NULL pointer.
In mach-omap1/mailbox.c the mailbox name is .name = "DSP" while in plat-omap/dsp/dsp_core.c function dsp_mbox_init() asks for "dsp". omap2 is correct, there .name is "dsp".
Patch follows.
There are still some open questions:
Mailbox related:
* Why does
if (IS_ERR(omap_dsp->mbox)) {
in arch/arm/plat-omap/dsp/dsp_core.c in dsp_mbox_init() doesn't catch NULL pointer (omap_dsp->mbox == NULL)?
- Why does system crash later if omap_dsp->mbox being NULL is caught correctly? E.g. using
if (omap_dsp->mbox == NULL) {
instead of IS_ERR above results in
...
<3>failed to get mailbox handler for DSP. <7>omapdsp: unmapping in ARM MMU, v=0xe0fff000, sz=0x1000 <1>Unable to handle kernel paging request at virtual address b0000ffc
...
LowLevel debug related:
- Why does LowLevel debug only works if omap_writew(reg, SOFT_REQ_REG) is disabled in mach-omap1/clock.c ?
Cheers
Dirk