Revision as of 15:47, 10 August 2022

Comments and Explanations
The page source contains comments providing guidance to fill out each section. They are invisible when viewing this page. To read it, choose the "view source" link.
Copy the source to a new page before making changes! DO NOT EDIT THIS TEMPLATE FOR YOUR CHANGE PROPOSAL.

Guidance
For details on how to fill out this form, see the documentation.

KTLS implementation for GnuTLS

This is a proposed Change for Fedora Linux.
This document represents a proposed Change. As part of the Changes process, proposals are publicly announced in order to receive community feedback. This proposal will only be implemented if approved by the Fedora Engineering Steering Committee.

Summary

Acceleration of GnuTLS with software Kernel TLS (KTLS)

Owner

Name: František Krenželok, Daiki Ueno
Email: fkrenzel@redhat.com, dueno@redhat.com

Current status

Targeted release: Fedora Linux 38
Last updated: 2022-08-10
FESCo issue: <will be assigned by the Wrangler>
Tracker bug: <will be assigned by the Wrangler>
Release notes tracker: <will be assigned by the Wrangler>

Detailed Description

The goal of this change is to provide GnuTLS users with a high throughput data transfer mechanism on encrypted channels, with emphasis on network block devices (NBD).

We accomplish this with KTLS which offloads enc/decryption (TLS record) to the kernel, while GnuTLS handles initial connection (TLS handshake). This approach saves us from frequent context switching as well as data copies in userspace when using send_file() function.

Feedback

Benefit to Fedora

The improvement lies in acceleration of large data transfer trough encrypted channels. The send_file function enables us to send data directly trough socket without entering user space, saving us from 2 context switches and 2 additional buffers. This is especially useful for NBD

One of the benefits might be acceleration of live VM migration, which should mitigate the downtime for various services used by both the users and the developers.

packages that might benefit: nbd nbdkit qemu

Scope

Proposal owners:

Other developers:

Release engineering: #Releng issue number

Policies and guidelines: N/A (not needed for this Change)

Trademark approval: N/A (not needed for this Change)

Alignment with Objectives:

Upgrade/compatibility impact

Although this feature will be enabled by default, users will not notice any change, as in case of failure to initialize KTLS, GnuTLS will default to the currently used mode of operation.

How To Test

To enable this feature user has to:

load TLS kernel module (modprobe tls)
enable ktls with crypto policies

Once proposal accepted
KTLS will be enabled by default and this step will not be needed.

$ cat > /etc/crypto-policies/local.d/gnutls-ktls.config <<EOF
[global]
ktls = true
EOF

$ update-crypto-policies

KTLS will not initialize if app uses custom push/pull callback for GnuTLS.

User Experience

This change should accelerate large data transfers especially that of files. This will affect users that use applications which utilize GnuTLS for encrypting communication channels.

Dependencies

Currently KTLS doesn't support key_update (The keys delivered to the kernel can’t be set more than once per session) so a kernel module patch would be needed for this functionality.

Contingency Plan

Contingency mechanism: (What to do? Who will do it?) N/A (not a System Wide Change)
Contingency deadline: N/A (not a System Wide Change)
Blocks release? N/A (not a System Wide Change), Yes/No

Documentation

API

gnutls_transport_is_ktls_enabled() To check if KTLS was properly initialized on the interfaces:

it has to be invoked no earlier that after a TLS-handshake

 gnutls_transport_ktls_enable_flags_t gnutls_transport_is_ktls_enabled(gnutls_session_t session);

gnutls_record_send_file() To send data directly from a file descriptor in a zero-copy manner if KTLS is enabled; otherwise it will just iteratively read from the file descriptor:

 ssize_t gnutls_record_send_file(gnutls_session_t session, int fd, off_t *offset, size_t count);

N/A (not a System Wide Change)

Release Notes

@@ Line 57: / Line 57: @@
 == Benefit to Fedora ==
-<!-- What is the benefit to the distribution?  Will the software we generate be improved? How will the process of creating Fedora releases be improved?
+The improvement lies in acceleration of large data transfer trough encrypted channels.
+The send_file function enables us to send data directly trough socket without entering user space, saving us from 2 context switches and 2 additional buffers. This is especially useful for NBD
-      Be sure to include the following areas if relevant:
-      If this is a major capability update, what has changed?
-           For example: This change introduces Python 5 that runs without the Global Interpreter Lock and is fully multithreaded.
-      If this is a new functionality, what capabilities does it bring?
-           For example: This change allows package upgrades to be performed automatically and rolled-back at will.
-      Does this improve some specific package or set of packages?
-           For example: This change modifies a package to use a different language stack that reduces install size by removing dependencies.
-      Does this improve specific Spins or Editions?
-           For example: This change modifies the default install of Fedora Workstation to be more in line with the base install of Fedora Server.
-      Does this make the distribution more efficient?
-           For example: This change replaces thousands of individual %post scriptlets in packages with one script that runs at the end.
-      Is this an improvement to maintainer processes?
-           For example: Gating Fedora packages on automatic QA tests will make rawhide more stable and allow changes to be implemented more smoothly.
-      Is this an improvement targeted as specific contributors?
-           For example: Ensuring that a minimal set of tools required for contribution to Fedora are installed by default eases the onboarding of new contributors.
-     When a Change has multiple benefits, it's better to list them all.
-     Consider these Change pages from previous editions as inspiration:
-     https://fedoraproject.org/wiki/Changes/Annobin (low-level and technical, invisible to users)
-     https://fedoraproject.org/wiki/Changes/ParallelInstallableDebuginfo (low-level, but visible to advanced users)
-     https://fedoraproject.org/wiki/Changes/VirtualBox_Guest_Integration (primarily a UX change)
-     https://fedoraproject.org/wiki/Changes/NoMoreAlpha (an improvement to distro processes)
-     https://fedoraproject.org/wiki/Changes/perl5.26 (major upgrade to a popular software stack, visible to users of that stack)
--->
-This change will affect broad spectrum of tools/packages that use GnuTLS.
 One of the benefits might be acceleration of ''live VM migration'', which should mitigate the downtime for various services used by both the users and the developers.
-packages that might benefit: {{package|nbdkit}} {{package|qemu}}
+packages that might benefit: {{package|nbd}} {{package|nbdkit}} {{package|qemu}}
 == Scope ==
@@ Line 110: / Line 85: @@
 == Upgrade/compatibility impact ==
-<!-- What happens to systems that have had a previous versions of Fedora installed and are updated to the version containing this change? Will anything require manual configuration or data migration? Will any existing functionality be no longer supported? -->
+Although this feature will be enabled by default, users will not notice any change, as in case of failure to initialize KTLS, GnuTLS will default to the currently used mode of operation.
-<!-- REQUIRED FOR SYSTEM WIDE CHANGES -->
-This feature is completely optional. Nor developers or users will be required to do anything regarding this change unless they would like to use the feature, in which case they would follow instructions in [[#How_To_Test| How to test]] section below.
 == How To Test ==
-<!-- This does not need to be a full-fledged document. Describe the dimensions of tests that this change implementation is expected to pass when it is done.  If it needs to be tested with different hardware or software configurations, indicate them.  The more specific you can be, the better the community testing can be.
+'''To enable this feature user has to:'''
+# load TLS kernel module (`modprobe tls`)
-Remember that you are writing this how to for interested testers to use to check out your change implementation - documenting what you do for testing is OK, but it's much better to document what *I* can do to test your change.
+# enable ktls with crypto policies
+{{admon/note|Once proposal accepted|KTLS will be enabled by default and this step will not be needed.}}
-A good "how to test" should answer these four questions:
-. What special hardware / data / etc. is needed (if any)?
-. How do I prepare my system to test this change? What packages
-need to be installed, config files edited, etc.?
-. What specific actions do I perform to check that the change is
-working like it's supposed to?
-. What are the expected results of those actions?
--->
-<!-- REQUIRED FOR SYSTEM WIDE CHANGES -->
-To enable this feature, user has to load TLS kernel module (`modprobe tls`) and enable ktls with crypto policies
 <pre>$ cat > /etc/crypto-policies/local.d/gnutls-ktls.config <<EOF
 [global]
@@ Line 140: / Line 98: @@
 </pre>
 <pre>$ update-crypto-policies</pre>
+{{admon/important||KTLS will not initialize if app uses custom push/pull callback for GnuTLS.}}
-===The following API is added===
-To check if KTLS was properly initialized on the interfaces, [https://gnutls.org/manual/gnutls.html#index-gnutls_005ftransport_005fis_005fktls_005fenabled see]: (it has to be invoked no earlier that after a TLS-handshake)
-<pre> gnutls_transport_ktls_enable_flags_t gnutls_transport_is_ktls_enabled(gnutls_session_t session);</pre>
-To send data directly from a file descriptor in a zero-copy manner if KTLS is enabled; otherwise it will just iteratively read from the file descriptor:
-<pre> ssize_t gnutls_record_send_file(gnutls_session_t session, int fd, off_t *offset, size_t count);</pre>
 == User Experience ==
-<!-- If this change proposal is noticeable by users, how will their experiences change as a result?
+This change should accelerate large data transfers especially that of files.
+This will affect users that use applications which utilize GnuTLS for encrypting communication channels.
- This section partially overlaps with the Benefit to Fedora section above. This section should be primarily about the User Experience, written in a way that does not assume deep technical knowledge. More detailed technical description should be left for the Benefit to Fedora section.
- Describe what Users will see or notice, for example:
-  - Packages are compressed more efficiently, making downloads and upgrades faster by 10%.
-  - Kerberos tickets can be renewed automatically. Users will now have to authenticate less and become more productive. Credential management improvements mean a user can start their work day with a single sign on and not have to pause for reauthentication during their entire day.
- - Libreoffice is one of the most commonly installed applications on Fedora and it is now available by default to help users "hit the ground running".
- - Green has been scientifically proven to be the most relaxing color. The move to a default background color of green with green text will result in Fedora users being the most relaxed users of any operating system.
--->
-This change might interest package maintainers i.e. developers that picked GnuTLS as their go to solution when it come to transferring data securely, and wish to accelerate it. The acceleration is most noticeable on large data transfer such that of files.
 == Dependencies ==
-<!-- What other packages (RPMs) depend on this package?  Are there changes outside the developers' control on which completion of this change depends?  In other words, completion of another change owned by someone else and might cause you to not be able to finish on time or that you would need to coordinate?  Other upstream projects like the kernel (if this is not a kernel change)? -->
-<!-- REQUIRED FOR SYSTEM WIDE CHANGES -->
-KTLS will not initialize if app uses custom push/pull callback for GnuTLS.
 Currently KTLS doesn't support key_update (The keys delivered to the kernel can’t be set more than once per session) so a kernel module patch would be needed for this functionality.
 == Contingency Plan ==
 <!-- If you cannot complete your feature by the final development freeze, what is the backup plan?  This might be as simple as "Revert the shipped configuration".  Or it might not (e.g. rebuilding a number of dependent packages).  If you feature is not completed in time we want to assure others that other parts of Fedora will not be in jeopardy.  -->
 * Contingency mechanism: (What to do?  Who will do it?) N/A (not a System Wide Change)  <!-- REQUIRED FOR SYSTEM WIDE CHANGES -->
@@ Line 177: / Line 114: @@
 <!-- Does finishing this feature block the release, or can we ship with the feature in incomplete state? -->
 * Blocks release? N/A (not a System Wide Change), Yes/No <!-- REQUIRED FOR SYSTEM WIDE CHANGES -->
 == Documentation ==
-<!-- Is there upstream documentation on this change, or notes you have written yourself?  Link to that material here so other interested developers can get involved. -->
+=== API ===
-[https://gnutls.org/manual/gnutls.html#index-gnutls_005frecord_005fsend_005ffile gnutls_record_send_file()]
+'''[https://gnutls.org/manual/gnutls.html#index-gnutls_005ftransport_005fis_005fktls_005fenabled gnutls_transport_is_ktls_enabled()]'''
-[https://gnutls.org/manual/gnutls.html#index-gnutls_005ftransport_005fis_005fktls_005fenabled gnutls_is_ktls_enabled()]
+To check if KTLS was properly initialized on the interfaces:
+{{admon/important||it has to be invoked no earlier that after a TLS-handshake}}
+<pre> gnutls_transport_ktls_enable_flags_t gnutls_transport_is_ktls_enabled(gnutls_session_t session);</pre>
+'''[https://gnutls.org/manual/gnutls.html#index-gnutls_005frecord_005fsend_005ffile gnutls_record_send_file()]'''
+To send data directly from a file descriptor in a zero-copy manner if KTLS is enabled; otherwise it will just iteratively read from the file descriptor:
+<pre> ssize_t gnutls_record_send_file(gnutls_session_t session, int fd, off_t *offset, size_t count);</pre>
 <!-- REQUIRED FOR SYSTEM WIDE CHANGES -->

Search

Changes/KTLSSupportForGnuTLS: Difference between revisions