Provides support for the low-level power subsystem handling
backlight control on GPD Pocket UMPC systems. Also enable its
dependency CONFIG_PWM=y (Closes: #895164).
Provides support for the battery charger on GPD Pocket UMPC systems.
Also enable its dependencies CONFIG_EXTCON=m,
CONFIG_EXTCON_INTEL_CHT_WC=m, CONFIG_I2C_CHT_WC=m and
CONFIG_INTEL_SOC_PMIC_CHTWC=y.
Reduce armel image size by:
- Set CONFIG_CRYPTO_MANAGER_DISABLE_TESTS=y
- Change MTD, MTD_CMDLINE_PARTS, RTC_DRV_MV, and SPI_ORION from
built-in to module.
- Disable VT, ZSWAP, RD_BZIP2, and RD_LZMA.
So qnap support is back.
Thanks to Leigh Brown <leigh@solinno.co.uk> for his idea to disable VT.
(cherry picked from commit a4fdfa09ce)
Reduce armel image size by:
- Set CONFIG_CRYPTO_MANAGER_DISABLE_TESTS=y
- Change MTD, MTD_CMDLINE_PARTS, RTC_DRV_MV, and SPI_ORION from
built-in to module.
- Disable VT, ZSWAP, RD_BZIP2, and RD_LZMA.
So qnap support is back.
Thanks to Leigh Brown <leigh@solinno.co.uk> for his idea to disable VT.
This enables CONFIG_MACH_DOVE, and a few device-specific drivers.
Signed-off-by: Josua Mayer <josua.mayer97@gmail.com>
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Extend the size limit for kernel image, from 2097080 to 2729712.
This will break a few qnap devices, but keep other armel devices
running.
Also revert two commits that disabled armel previously:
- [2ed70eb] "Add empty featuresets for armel to help abiupdate script"
- [5f62872] "(Temporarily) disable armel kernel image build"
(cherry picked from commit b0a94d07b4)
Extend the size limit for kernel image, from 2097080 to 2729712.
This will break a few qnap devices, but keep other armel devices
running.
Also revert two commits that disabled armel previously:
- [2ed70eb] "Add empty featuresets for armel to help abiupdate script"
- [5f62872] "(Temporarily) disable armel kernel image build"
Commit 1df9e416e647 "Kbuild: always define endianess in kconfig.h"
fixed several cases where some types (and inline functions) might not
be correctly defined according to the host byte order. It might be
possible to avoid an ABI bump, but it would require a lot of work.
Also, the problem may affect OOT modules that therefore should be
rebuilt with the fix.
Enable SND_SOC_INTEL_SST_TOPLEVEL, SND_SOC_INTEL_HASWELL,
SND_SOC_INTEL_BAYTRAIL, SND_SST_ATOM_HIFI2_PLATFORM,
SND_SOC_INTEL_SKYLAKE as modules. This re-enables the board
drivers that now depend on them (Closes: #892629).
Remove config for SND_SOC_INTEL_BYT_MAX98090_MACH and
SND_SOC_INTEL_BYT_RT5640_MACH, which depend on SND_SST_IPC_ACPI = n
and therefore can't be enabled together with the other drivers.
The abiupdate script bails out currently (since we disabled armel image
builds at least temporarily) with
Retrieve config
Traceback (most recent call last):
File "debian/bin/abiupdate.py", line 224, in <module>
Main(url, url_config, **kw)()
File "debian/bin/abiupdate.py", line 95, in __call__
self.update_arch(config, arch)
File "debian/bin/abiupdate.py", line 149, in update_arch
featuresets = config[('base', arch)]['featuresets']
KeyError: 'featuresets'
Possibly this should be handled more gracefully in abiupdate itself, but
workaround the situation first by adding an empty featuressets and
explaining in the comment why we do not build images.
In the long run armel will disapear completely.
Gbp-Dch: Ignore
The armel/marvell kernel size is growing to large and the compressed
image is over the limit.
Given the armel architecture will most likely not be part of Buster,
disable the image build.
Cf. https://lists.debian.org/debian-kernel/2018/01/msg00278.html
Reset drivers are tiny and are needed often early in boot. Kconfig
selects the drivers automatically with arch specific defaults:
config COMMON_RESET_HI6220
default ARCH_HISI
So remove explicit driver selections from arm 64 config and let
kconfig dependencies handle it for us.
As discussed on d-kernel, this flavour is added as experiment on request
of Microsoft. For now it is only tested on Microsoft Azure.
It will be expanded to cover the other public cloud platforms at well.
This platforms will need additional drivers.
-----BEGIN PGP SIGNATURE-----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=M11J
-----END PGP SIGNATURE-----
Merge tag 'debian/4.14.13-1'
Release linux (4.14.13-1).
- Drop ABI reference
- Drop/refresh patches as necessary
- linux-headers: Drop versioned dependency on linux-kbuild, as there has not
been any version of linux-kbuild-4.15 without objtool
This reverts commit 16d7f29d46. There
seem to be a lot of important fixes going into 4.15 for blk-mq, most
of which aren't tagged for stable, making it look a bit risky as a
default for 4.14. I'll probably re-enable this for 4.15-rc.
Now that blk-mq has a similar default scheduler as the old block
layer, it should be safe to switch over.
Note how to revert this at run-time if necessary.
- Change CONNECTOR from built-in to module, and disable PROC_EVENTS
- Disable INTEGRITY and dependent options
- video: Disable USB_APPLEDISPLAY, BACKLIGHT_CLASS_DEVICE
This still only gets us down to 99.26% of maximum size, so further
changes will be needed soon.
The apparmor package is needed for loading profiles. In theory,
enabling AppArmor without any profiles loaded should do nothing, so
this is not really a dependency of the kernel. However, if a systemd
unit specifies a AppArmor profile and the kernel has AppArmor enabled
then failure to load the profile is fatal.
As the linux-image packages select AppArmor as the default LSM, they
should probably also be responsible for getting the necessary userland
support installed. But since the default can be overridden, use
Recommends rather than Depends.
Commit 8009d506a1dd "ALSA: seq: Enable 'use' locking in all
configurations" which was backported into 4.13.10 will result in an
ABI change for !SMP configurations. Ignore this, as I don't expect
there to be any out-of-tree sequencer drivers.
Upstream bumped the default value from 8 to 16 for 4.10 in commit
90866b3a8011 ("[media] Raise adapter number limit")
. Automatically follow the upstream default value by dropping the explicit
value of 8.
The latest build failed on mips for the 4kc-malta flavour (which is the first):
RELOCS vmlinux
Relocations overflow available space!
Please adjust CONFIG_RELOCATION_TABLE_SIZE to at least 0x0010f000
So do the same thing we did for the octeon flavour.
FRAMEBUFFER_CONSOLE is enabled in almost all configurations. So make
this the top level default, and disable it in the few exceptional
configurations (armel/marvell, mips*/octeon, s390x).
- [alpha] fbdev: Re-enable FRAMEBUFFER_CONSOLE as built-in
- [armel] fbdev: Explicitly disable FRAMEBUFFER_CONSOLE, as it can no longer
be a module
- [arm64] Re-enable MMC_QCOM_DML
- Change RC_CORE back to being a module
- power/supply: Enable BATTERY_BQ27XXX and BATTERY_BQ27XXX_HDQ as modules,
replacing W1_SLAVE_BQ27000
Clean up with kconfigeditor2 and delete obsolete symbols.