Release 3.2 kernel issue tracking thread


When I feel there are enough reports that kernel is stable (one from SGX test thread) I'll add it to "upgrade OS" package list, along with all this crypto stuff.
 
Last edited by a moderator:
For info, this is my changes.txt when i recompile my own kernel:

-added udlfb for DisplayLink support
-added lzo and xz support for squashfs

-Removed IP: kernel level autoconfiguration (!).
-Added all bunch in Network packet filtering framework (Netfilter) as modules.
-Added IP set
-Added packet mangling
-Added 802.1d Ethernet Bridging as a module.
-Added 802.1Q VLAN Support as a module.
-Added all bunch in QoS and/or fair queueing as modules.

-Added Plan 9 Resource Sharing Support (9P2000) as a module.

-Added Device mapper support as a module + dm-crypt

-Added Bonding driver support as a module (may be fun with two usb adapters).

-added jfs
-Fully-enabled crypto modules.
 
 
-added lzo and xz support for squashfs
Already had lzo, added xz (SGX pnd is xz compressed, saves over 30MB)
-Removed IP: kernel level autoconfiguration (!).
Hmm ok, agree it's not needed.
-Added all bunch in Network packet filtering framework (Netfilter) as modules.

-Added IP set

-Added packet mangling

-Added 802.1d Ethernet Bridging as a module.

-Added 802.1Q VLAN Support as a module.

-Added all bunch in QoS and/or fair queueing as modules.

-Added Plan 9 Resource Sharing Support (9P2000) as a module.

-Added Bonding driver support as a module (may be fun with two usb adapters).

-added jfs
Don't think these are useful for pandora, nobody ever asked for them.
-Fully-enabled crypto modules.
Don't think we need all of them either, don't want things to sit on tiny NAND in everyone's pandora and be never used.
 
For info, again ;^).

Netfilter and mangling are needed to act as a router/nat.

JFS has the reputation of being the less cpu-intensive filesystem.
 
I would like to know how to extract the current kernel sources from the git pandora-kernel folder i've already downloaded.

The idea is to put it in the ram for a faster distcc compilation.
 
And a technical question. I've patched the kernel for real time. Now, i can't compile it as usual:


CC mm/page_alloc.o
mm/page_alloc.c:242:8: warning: type defaults to ‘int’ in declaration of ‘DEFINE_LOCAL_IRQ_LOCK’ [-Wimplicit-int]
mm/page_alloc.c:242:1: warning: parameter names (without types) in function declaration [enabled by default]
mm/page_alloc.c: In function ‘__free_pages_ok’:
mm/page_alloc.c:782:2: error: implicit declaration of function ‘local_lock_irqsave’ [-Werror=implicit-function-declaration]
mm/page_alloc.c:782:21: error: ‘pa_lock’ undeclared (first use in this function)
mm/page_alloc.c:782:21: note: each undeclared identifier is reported only once for each function it appears in
mm/page_alloc.c:788:2: error: implicit declaration of function ‘local_unlock_irqrestore’ [-Werror=implicit-function-declaration]
mm/page_alloc.c: In function ‘drain_pages’:
mm/page_alloc.c:1253:3: error: implicit declaration of function ‘local_lock_irqsave_on’ [-Werror=implicit-function-declaration]
mm/page_alloc.c:1253:3: error: ‘pa_lock’ undeclared (first use in this function)
mm/page_alloc.c:1262:3: error: implicit declaration of function ‘local_unlock_irqrestore_on’ [-Werror=implicit-function-declaration]
mm/page_alloc.c: In function ‘free_hot_cold_page’:
mm/page_alloc.c:1344:21: error: ‘pa_lock’ undeclared (first use in this function)
mm/page_alloc.c: In function ‘buffered_rmqueue’:
mm/page_alloc.c:1495:22: error: ‘pa_lock’ undeclared (first use in this function)
mm/page_alloc.c:1527:3: error: implicit declaration of function ‘local_spin_lock_irqsave’ [-Werror=implicit-function-declaration]
mm/page_alloc.c: In function ‘__zone_pcp_update’:
mm/page_alloc.c:3936:3: error: ‘pa_lock’ undeclared (first use in this function)
mm/page_alloc.c: In function ‘page_alloc_init’:
mm/page_alloc.c:5303:2: error: implicit declaration of function ‘local_irq_lock_init’ [-Werror=implicit-function-declaration]
mm/page_alloc.c:5303:22: error: ‘pa_lock’ undeclared (first use in this function)
mm/page_alloc.c: At top level:
mm/page_alloc.c:242:8: warning: ‘DEFINE_LOCAL_IRQ_LOCK’ declared ‘static’ but never defined [-Wunused-function]
cc1: some warnings being treated as errors
make[1]: *** [mm/page_alloc.o] Error 1
make: *** [mm] Error 2
root@darkstar:/usr/src/linux#



Any tips ?
 
Last edited by a moderator:
I've built a 3.9x kernel, but it produced a zImage, so i did:

mkimage -A arm -O linux -T kernel -C none -a 0x80300000 -e 0x80300000 -d zImage uImage

But it won't load anything, it get stuck at the Pandora logo :/.

---EDIT:

Corrected the gzip compression, now it reboots endlessly:

mkimage -A arm -O linux -T kernel -C gzip -a 0x80300000 -e 0x80300000 -d zImage uImage
 
Last edited by a moderator:
This works for me (run from kernel tree):


./scripts/mkuboot.sh -A arm -O linux -C none  -T kernel -a 0x80008000 -e 0x80008000 -n 'Linux-3.9.0-rc5' -d arch/arm/boot/zImage uImage

Make sure you have the latest update from git, it has a patch that fixes pandora's LCD.
 
Last edited by a moderator:
I found a new issue. I was trying to port "kexecboot" to the Pandora to create a userfriendly method of booting different kernels and/or the same kernel with different options without manually editing "boot.txt" each time. This "soft" bootloader uses the kexec method of the kernel. The option is included in the 3.2.17 and the old 2.6 kernel. But when using it the screen fades to black and that's it (3.2) or the Pandora freezes (2.6.).
I know it's too little too late, but it has been fixed in kernel included in SuperZaxxon 1.54..
 
Last edited by a moderator:
This works for me (run from kernel tree):


./scripts/mkuboot.sh -A arm -O linux -C none  -T kernel -a 0x80008000 -e 0x80008000 -n 'Linux-3.9.0-rc5' -d arch/arm/boot/zImage uImage

Make sure you have the latest update from git, it has a patch that fixes pandora's LCD.
Ok i will try now. I tried the rc4 because  i had shutdown problems with -rc5 on my main PC.

Just curious, autoboot says 0x80300000, so i used this number. How is it related to the 0x80008000 ?

And for information, i use SOLEIL with 7 PXE'd PCs as arm distcc cluster. Compile time is 66mn.

The bottleneck is the postprocessing on the OP (my two best PC are sufficient in fact), and also a bit of SD access (that's why i wanted to offload the sources directly into the ram).

The distcc-3.1 on Slackware -current now uses the "pump" feature which unload the postprocessing to the cluster and make things up to 10X faster.

Can't wait to try that. But first, i need a working kernel ;^).
 
Do you use a special .config, or the one from the 3.2.39 + make oldconfig will suffice ?

--EDIT:

Only had a black screen.

I realized i used the RC5 archive, not the latest git.

Same question about the .config .
 
Last edited by a moderator:
"make omap2plus_defconfig" should generate config that upstream OMAP guys are usually testing, so is most likely to work.
 
Black screen again :/.

Where do i retrieve the more recent than recent sources ?

Took a look at Linus branch but seems it's not that.

--Edit:

And about the 80008000 ?
 
Last edited by a moderator:
I used Linus tree at around -rc2 time or so and it worked, right now I don't have time to test.

80008000 is what pandora always uses, it has no relation to the other address from u-boot.
 
Last edited by a moderator:
Linus git, make omap2plus_defconfig, the mkimage line = black screen.

--Edit: also tried on my CC, same result.
 
Last edited by a moderator:
Will do a try with the -rc6 soon.

Anyway, in case you rebuild one one day, please upload it somewhere ^^.
 
Same result with -rc6: black screen.

Doing better choices in distcc settings leads to a nice 36 mn kernel + modules compilation.
 
To close this off-topic, Slack -current glibc was reverted to support 3.1 kernels and onwards.
 
Back
Top