Age | Commit message (Collapse) | Author |
|
(kernel 3.17 has been removed from oe-core and meta-initramfs)
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
(kernel 3.10 has been removed from oe-core and meta-initramfs)
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Patch landed upstream in 3.17.3 (commit id
255a8e963ea7d9d856d224a76b8a2fb99f0682fe)
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
This makes the v3.17 kernel work with the h3600 target.
As of kernel v3.17, most drivers for the H3600 have been
upstreamed and we only need two patches to boot it
successfully with full support.
Patch 1 is merged upstream by now (v3.18), patch 2 is
being worked on.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
|
|
The kernel behavior has changed and apparently the init symlink
in the initramfs is ignored (no init seems executed).
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Match the changes done by commit 7384d2831c713ac5999aca83c312154dc15cec56
in oe-core.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Follow upstream changes.
While there, fix a potential issue with do_package_qa
like done in meta-initramfs for the main -kexecboot kernel.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Follow upstream changes.
Add CONFIG_KEXEC=y
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Zaurus is using pxa/sa1100 serial so we only need the
pcmcia serial cs for i.e. bluetooth cf cards.
Avoid device address conflict on pxa (fix pending upstream).
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Unfortunately the default for sa1100/collie is 16M so for collie it is necessary
to specify mem=64M on cmdline.
The very rare 5000D users will have to change that to mem=32M.
This reverts commit f4f05028f5ad72a81d6620340fe1c92ac6486890.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
as sent upstream
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Surely rare today, the SL-5000D has only 32MB ram.
Remove historical setting used for mtd-ram.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Revert to the 3.2 pre-yocto settings.
Removed yaffs, reiserfs, btrfs, most xattr and posix options.
Make NFS modular.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The collie used for development has unfortunately an hw flaw
and the CFI table is missing on one NOR chip.
Further testings on other models are giving normal readings.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The collie used for development has unfortunately an hw flaw
and the CFI table is missing on one NOR chip.
Further testings on other models are giving normal readings.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
commit id=b98c1e1c1f71ab32e755e2e417e9a682865578b8
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Alternative images to be booted by kexecboot
require to have an elf vmlinuz in /boot (kexec
on mips expects elf).
The bootloader however needs raw data (vmlinuz.bin) so
we customize KERNEL_IMAGETYPE in the recipe.
Finally, kexec is added to the main kernel recipe for
easier testing and the watchdog timer is increased in
the configuration for the kexecboot kernel.
At the moment linux-kexecboot detects just fine
partitions on external SD but freezes on kexec.
Tests done from shell reveal that both kexec
binaries (2.06 glibc and 2.02 klibc) load the new kernel
but freeze on kexec -e.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Notes:
- strangely on mips /init seems not allowed to be a symlink
- kernel doesn't boot using kexec_2.0.2 klibc-static but maybe
it is a kernel and not userspace issue. Plan is to update to
kexec-tools_2.0.6 as soon as possible and retest.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The machines using it before are now preferring linux-yocto.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The default for fragment-based recipes is '--alldefconfig'
but using defconfig it is necessary to specify it.
(An hidden bug in the kernel tools has masked that until now...)
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Adjust defconfig to match the gpio-charger patch.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
h1940 is the single one left, needs backports
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The restyling of the .bbappend was seemingly a good simplification but
building for qemux86 fails if the meta-handheld layer is included.
* ERROR: Function failed: Fetcher failure for URL: 'file://defconfig'.
* Unable to fetch URL from any source.
Go back to the previous syntax used for the older kernel versions and
fix the contamination of the variables for foreign machines once for all.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Add more fixups for the chip latencies and enable
again Suspend Erase on write.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
* Includes h3600 updates from Linus Walleij <linus.walleij@linaro.org>
* h1940 is currently excluded since it doesn't build with 3.14.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
|
|
This adds the latest iPAQ microcontroller patches to the yocto-dev
recipe, updates the defconfig, and adds the h3600 variant to the bbappend
file.
(Not making this the default since the following patch updates
linux-yocto to 3.14. - Paul)
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
|
|
Avoid duplication between recipes with the same PV.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
include cfg/8250.scc instead
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Remove the single patches and squash together with next patch
preparing for future upstream submission.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Fix compilation warning
linux/drivers/mtd/chips/cfi_util.c:47:2: warni
ng: ISO C90 forbids mixed declarations and code
[-Wdeclaration-after-statement]
unsigned long mask;
^
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
After debugging we have a patch solving the issue.
It appears these chips need the start address for the buffer write program.
Revert af54b196e7d6ced6d308fc84c2d03a98ed2e5bfd
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The kernel recipe for the Ben Nanonote machine is now relying on the
Qi-Hardware kernel source code instead of the JLime kernel source code
initially.
Rename linux-jlime-ben-nanonote recipe to linux-qi-ben-nanonote to
reflect the change.
Signed-off-by: Apelete Seketeli <apelete@seketeli.net>
|
|
Qi-Hardware community kernel now has an upstream compliant driver to
provide USB support for the Ben Nanonote, along with other features
and fixes.
Update the recipe to build from qi-kernel 3.12 source until the
Qi-Hardware patches land upstream.
Signed-off-by: Apelete Seketeli <apelete@seketeli.net>
|
|
The CFI readings of the number of hardware partitions is now fixed
and correctly showing 2.
Further output is valid only for the default case of 2 asymmetric partitions
but the CFI code in kernel does only care about the number of equal partitions.
So we set the PCR accordingly but with 2 symmetric partitions of 2 planes each
without loosing the advantages of the Dual Work.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
The Sharp CFI table doesn't match 100% with CFI 1.3 about the number
of fields reserved for the Burst/Sync Read modes (unsupported here).
The CFI table reports 2 asymmetric partitions (boot defaults).
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
It is still unclear whether collie is a special case or the setup
of 2x16bit chips on 32bit bus needs some bugfix.
Apparently the issue is only with the answer of READ QUERY (0x98):
the logs show that both chips are answering in all other cases.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
We have now a less intrusive hack to 'fix' QRY answer.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Adding a separate fragment with MD, ethernet and debug options
which can be easily reenabled.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Something bad (timeout ?) is happening and the fs is corrupted
on remount/reboot.
Going back to word write offers a stable base for further debugging.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
DISABLE_ERASE_SUSPEND_ON_WRITE doesn't help with the timeouts
originated by the misdetected chip partitioning.
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|
|
Initial hack intended as workaround for the CFI code which seems
not liking these Dual Work / 4 planes chips
(one Status Register for each partition, 2 partitions on boot).
Signed-off-by: Andrea Adami <andrea.adami@gmail.com>
|