Kernel command line: add support for arbitrary configuration settings #1983
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The id heap has the property that allocated values are aligned to the nearest power-of-2 value equal to or greater than the allocation size. In the current code, it can happen that an allocation of a given size served by an id heap range whose start value is not aligned to the allocation size returns a non-aligned value. This patch fixes the above issue by removing the padding of the start argument in the bitmap allocation function (because the start value is already appropriately aligned by the id heap in the id_alloc_from_range() function). In order to optimize performance of bitmap allocations, bitmap ranges added to an id heap have now a start value always aligned to BITMAP_WORDLEN (even if the corresponding id heap range is not aligned), so that the start argument to bitmap_alloc_within_range() is always aligned to BITMAP_WORDLEN for allocation size values greater than (BITMAP_WORDLEN / 2). The bitmap_start field of struct id_range contains the aligned start value of a range, and is used to apply the appropriate offsets during allocations and deallocations. Since now the id heap works correctly with arbitrarily aligned ranges, we can remove the padding to 2MB of the start value of physical memory ranges, which was causing a noticeable waste of memory space in low-memory machines.
rinor
reviewed
Dec 13, 2023
francescolavra
force-pushed
the
feature/ture/cmdline
branch
from
8c8223d
to
b0d4838
Compare
francescolavra
force-pushed
the
feature/ture/cmdline
branch
from
b0d4838
to
21b4176
Compare
This new parameter is used to determine whether a tagged region allocator should be multithread-safe. This change, beside allowing to remove duplicated code, avoids the unnecessary instantiation of locking heap wrappers in the aarch64 code, where the tagged region allocator uses directly the locked heap and thus is always multithread-safe. Initialization of tagged region allocators is being moved from platform-specific code to kernel_runtime_init(). In addition, the management subsystem now uses locking heaps instead of non-locking heaps, so that it can be used safely at runtime.
The general (non-locked) kernel heap is not safe to use after the call to kernel_runtime_init(); this change replaces this heap with the locked heap in a few places where the code is executed from the scheduler runloop.
This function returns true if the value supplied as argument is of a composite type, i.e. it can be operated on using get(), set() and iterate(). Calls to is_tuple() and is_vector() have been refactored to use is_composite() where appropriate; this removes some duplicated code.
This patch enhances the tuple_notifer implementation so that it can be used with a parent vector (alternatively to a parent tuple) and adds a copy-on-write functionality that allows to operate (i.e. execute get, set or iterate operations) on a composite value (tuple or vector) without modifying the original value or any nested composite values. These new features can be used for example to modify the root configuration tuple (e.g. to add or change debug options, environment variables, or program arguments) without persisting these changes in the root filesystem. The get_root_tuple() kernel function now returns the tuple_notifier that wraps the root tuple (instead of the root tuple itself), so that configuration settings are initially retrieved from the root filesystem and can be changed at runtime without affecting the contents of the boot disk.
This patch adds support for changing arbitrary settings in the root
tuple via the kernel command line (which is retrieved by the kernel
when booting under AWS Firecracker on x86).
The PC platform code has been amended by adding a new
REGION_CMDLINE region type holding the location and size of the
command line, which can be accessed later in the boot process. The
existing parsing of the virtio_mmio command line options has been
moved to kernel/init.c, so that it can be called by
virtio_mmio_enum_devs() instead of platform-specific code. The only
platform-specific code is the code that retrieves (and potentially
updates) the command line, while the parsing implementation is in
kernel/init.c.
With these changes, it is possible for example to override the
network settings when starting an instance from a given image
(without modifying the image itself) by specifying those settings
in the kernel command line ("boot_args" parameter in the
Firecracker configuration file), as in the following example:
"en1.ipaddr=10.3.3.6 en1.netmask=255.255.0.0 en1.gateway=10.3.0.1".
In the above example, "en1" identifies the first network interface;
if multiple interfaces are used (en2, en3, etc.), each of them can
be configured independently.
Example to configure a static IPv6 address on the first network
interface:
"en1.ip6addr=20::A8FC:FF:7600:AA"
Example to add an environment variable (or override its value if
the variable is already present in the image):
"environment.VAR_NAME=VAR_VALUE"
Example to modify the program arguments:
"arguments.0=/bin/my-program arguments.1=--my-option"
Closes #1976
francescolavra
force-pushed
the
feature/ture/cmdline
branch
from
21b4176
to
efd391b
Compare
|
I added 2 commits to fix concurrency issues observed when running CI tests. |
eyberg
approved these changes
Dec 14, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch set adds support for changing arbitrary settings in the root tuple via the kernel command line (which is retrieved by the kernel when booting under AWS Firecracker on x86).
With these changes, it is possible for example to override the network settings when starting an instance from a given image (without modifying the image itself) by specifying those settings in the kernel command line ("boot_args" parameter in the Firecracker configuration file), as in the following example:
"en1.ipaddr=10.3.3.6 en1.netmask=255.255.0.0 en1.gateway=10.3.0.1".
In the above example, "en1" identifies the first network interface; if multiple interfaces are used (en2, en3, etc.), each of them can be configured independently.
Example to configure a static IPv6 address on the first network interface:
"en1.ip6addr=20::A8FC:FF:7600:AA"
Example to add an environment variable (or override its value if the variable is already present in the image):
"environment.VAR_NAME=VAR_VALUE"
Example to modify the program arguments:
"arguments.0=/bin/my-program arguments.1=--my-option"
The first 3 commits are fixes for issues that have been exposed by the last 2 commits; the fix in the first commit allows all available physical memory ranges to be used by the kernel, without requiring 2MB alignment of the range start.
Closes #1976