| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
|
| |
The goal of upstream Linux is to reduce the stackframe size on 32-bit to
1024 and on 64-bit to 1280, inline with how gcc generally works. While
this hasn't been achieved yet everywhere and in all configurations, the
Alpine status quo of 1024 on 64-bit is something that doesn't have plans
to happen. Given that the intent was to be conservative, we raise this
to 1280, rather than something large like 2048.
Reference: https://lore.kernel.org/lkml/CAK8P3a0CS3QzEKEV5==qj8hUYgW+q2v1f13jA+s0TjQd8kYXFA@mail.gmail.com/
|
|
|
|
| |
ref #10044
|
|
|
|
| |
fixes #9975
|
| |
|
| |
|
|
|
|
|
|
|
| |
Build amazon's ena network driver module on x86, x86_64, and aarch64 (for EC2 a1 instances).
At some point we will probably want to add config-virt.aarch64, too.
The ena driver in recent kernels (4.19.x) is in sync with the one in https://github.com/amzn/amzn-drivers, eliminating the necessity for the 'community/aws-ena-driver' aport.
|
| |
|
| |
|
|
|
|
| |
those are used by packet.net's c1.large.arm.xda machines
|
|
|
|
|
| |
AWS Elastic File System works best with NFS v4.1, Docker's cloudstor:aws plugin requires it.
This is already enabled in config-vanilla.s390x; simply enabling it in the other configs.
|
| |
|
| |
|
|
|
|
|
| |
- build xgene net drivers as modules (aarch64)
- enable cbc_mbim (x86/x86_64) fixes #8855
|
|
|
|
|
|
|
|
|
|
| |
Framebuffer simple seems to be needed to get a VGA console working on
xgene.
setting CONFIG_XGENE_SLIMPRO_MBOX=y fixes dmesg error:
xgene-slimpro-i2c soc:i2cslimpro: i2c mailbox channel request failed
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1697407
|
|
|
|
| |
fixes #8778
|
|
|
|
|
|
|
|
| |
- enable thunder2
- set ARM64_VA_BITS to 48
- increase NR_CPUS to 256
fixes #8717
|
| |
|
|
|
|
|
|
|
| |
The Netronome nic driver was renamed in v4.11, the kernel jump from 4.9
to 4.14 in 9b052dbbcfdc3645493c91f1a80012ba3cdb0d16 correctly drops the
old nic config, but does not replace it with the new one. This commit
adds the driver back.
|
|
|
|
| |
fixes #8401
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit updates to kernel version 4.9.75 and enables
CONFIG_PAGE_TABLE_ISOLATION for x86, x86_64 and aarch64. For all
other architectures, CONFIG_PAGE_TABLE_ISOLATION is disabled.
CONFIG_PAGE_TABLE_ISOLATION mitigates the Meltdown security flaw
almost all Intel CPUs and some ARM CPUs are suspect to [1,2].
(This patch does not solve the Spectre security threat [2], which
affects also non-Intel CPUs [3].)
I believe this commit will cause some discussion, especially the
following points seem worth discussing:
a) CONFIG_PAGE_TABLE_ISOLATION has a performance impact on
syscalls, which can slow down specific applications
significantly. AMD users might benefit from a kernel without
KPTI (unless Meltdown turns out to affect them as well)
b) Is disabling this feature a reasonable choice for CPU
architectures different from x86, x86_64 and aarch64?
[1]: https://meltdownattack.com/#faq-systems-meltdown
[2]: http://kroah.com/log/blog/2018/01/06/meltdown-status/
[3]: https://meltdownattack.com/#faq-systems-spectre
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This commit brings kernel and userspace into agreement about squashfs
compression algorithms: zlib, xz, lzo, and lz4 (except on S/390 and PPC64LE).
Some kernels has LZO, some had LZ4, some had both.
Userspace had LZO, but not LZ4.
XZ and LZO were added to squashfs-tools in commit
151deb4c2fe91078c30c588de3fe9e411849a52f
Most compression options were added to kernels through default configs:
* LZO had been enabled in RPI, RPI2, S/390, PPC64le, and squashfs-tools.
* LZ4 had been enabled in x86, x86_64, aarch64, armhf, rpi, rpi2.
Default linux kernel config does not currently include LZ4 for S/390 or
PPC64le, so I did not change their config. XZ and ZLIB are in everything.
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
ref #5191
ref #6486
|
| |
|
| |
|
| |
|
|
|
|
|
| |
serial_of_platform=y. this improves console=ttyS0 handling
on certain boards.
|
| |
|
|
|