vhost-user-blk has connected hypervisors to software-defined storage since around 2017, but it was mainly seen as virtualization technology. Did you know that vhost-user-blk is not specific to virtual machines? I think it's time to use it more generally as a userspace block I/O interface because it's fast, unprivileged, and avoids exposing kernel attack surfaces.
My LWN.net article about Accessing QEMU storage features without a VM already hinted at this, but now it's time to focus on what vhost-user-blk is and why it's easy to integrate into your applications. libblkio is a simple and familiar block I/O API with vhost-user-blk support. You can connect to existing SPDK-based software-defined storage applications, qemu-storage-daemon, and other vhost-user-blk back-ends.
Come see my FOSDEM '23 talk about vhost-user-blk as a fast userspace block I/O interface live on Saturday Feb 4 2023, 11:15 CET. It will be streamed on the FOSDEM website and recordings will be available later. Slides are available here.
Wednesday, February 1st, 11:00 AM - 12:00 PM ET
"Open to Clients, Business Partners, IBMers, IT Architects, Systems Admins, etc."
We’d like to announce the availability of the QEMU 7.2.0 release. This release contains 1800+ commits from 205 authors.
You can grab the tarball from our download page. The full list of changes are available in the Wiki.
Highlights include:
Thank you to everyone involved!
For a long time, the QEMU project hosted its git repository on their own server and used Launchpad for tracking bugs. The self-hosting of the git repository caused some troubles, so the project switched the main repository to Gitlab in January 2021. That change of course also triggered the question whether the bug tracking could be moved from Launchpad to Gitlab, too. This would provide a better integration of the bug tracking with the git repository, and also has the advantage that more QEMU developers have a Gitlab account than a Launchpad account. But after some discussions it was clear that there was the desire to not simply leave the opened bug tickets at Launchpad behind, so for being able to switch, those tickets needed to be migrated to the Gitlab issue tracker instead.
Fortunately, there are APIs for both, Launchpad and Gitlab, so although I was a complete Python newbie, I was indeed able to build a little script that transfers bug tickets from Launchpad to Gitlab. I recently found the script on my hard disk again, and I thought it might maybe be helpful for other people in the same situation, so here it is:
#!/usr/bin/env python3
import argparse
import os
import re
import sys
import time
import gitlab
import textwrap
from launchpadlib.launchpad import Launchpad
import lazr.restfulclient.errors
parser = argparse.ArgumentParser(description=
"Copy bugs from Launchpad to Gitlab")
parser.add_argument('-l',
'--lp-project-name',
dest='lp_project_name',
help='The Launchpad project name.')
parser.add_argument('-g',
'--gl-project-id',
dest='gl_project_id',
help='The Gitlab project ID.')
parser.add_argument('--verbose', '-v',
help='Enable debug logging.',
action="store_true")
parser.add_argument('--open', '-o',
dest='open_url',
help='Open URLs in browser.',
action="store_true")
parser.add_argument('--anonymous', '-a',
help='Use anonymous login to launchpad (no updates!)',
action="store_true")
parser.add_argument('--search-text', '-s',
dest='search_text',
help='Look up bugs by searching for text.')
parser.add_argument('--reporter', '-r',
dest='reporter',
help='Look up bugs from the given reporter only.')
parser.add_argument('-b',
'--batch-size',
dest='batch_size',
default=20,
type=int,
help='The maximum amount of bug tickets to handle.')
args = parser.parse_args()
def get_launchpad():
cache_dir = os.path.expanduser("~/.launchpadlib/cache/")
if not os.path.exists(cache_dir):
os.makedirs(cache_dir, 0o700)
def no_credential():
print("ERROR: Can't proceed without Launchpad credential.")
sys.exit()
if args.anonymous:
launchpad = Launchpad.login_anonymously(args.lp_project_name +
'-bugs',
'production', cache_dir)
else:
launchpad = Launchpad.login_with(args.lp_project_name + '-bugs',
'production',
cache_dir,
credential_save_failed=no_credential)
return launchpad
def convert_tags(tags):
convtab = {
"cve": "Security",
"disk": "Storage",
"documentation": "Documentation",
"ethernet": "Networking",
"feature-request": "kind::Feature Request",
"linux": "os: Linux",
"macos": "os: macOS",
"security": "Security",
"test": "Tests",
"tests": "Tests",
}
labels = []
for tag in tags:
label = convtab.get(tag)
if label:
labels.append(label)
return labels
def show_bug_task(bug_task):
print('*** %s - %s' % (bug_task.bug.web_link,
str(bug_task.bug.title)[0:44] + "..."))
if args.verbose:
print('### Description: %s' % bug_task.bug.description)
print('### Tags: %s' % bug_task.bug.tags)
print('### Status: %s' % bug_task.status)
print('### Assignee: %s' % bug_task.assignee)
print('### Owner: %s' % bug_task.owner)
for attachment in bug_task.bug.attachments:
print('#### Attachment: %s (%s)'
% (attachment.data_link, attachment.title))
#print(sorted(attachment.lp_attributes))
for message in bug_task.bug.messages:
print('#### Message: %s' % message.content)
def mark_lp_bug_moved(bug_task, new_url):
subject = "Moved bug report"
comment = """
This is an automated cleanup. This bug report has been moved to the
new bug tracker on gitlab.com and thus gets marked as 'expired' now.
Please continue with the discussion here:
%s
""" % new_url
bug_task.status = "Expired"
bug_task.assignee = None
try:
bug_task.lp_save()
bug_task.bug.newMessage(subject=subject, content=comment)
if args.verbose:
printf(" ... expired LP bug report %s" % bug_task.web_link)
except lazr.restfulclient.errors.ServerError as e:
print("ERROR: Timeout while saving LP bug update! (%s)" % e, end='')
except Exception as e:
print("ERROR: Failed to save LP bug update! (%s)" % e, end='')
def preptext(txt):
txtwrapper = textwrap.TextWrapper(replace_whitespace = False,
break_long_words = False,
drop_whitespace = True, width = 74)
outtxt = ""
for line in txt.split("\n"):
outtxt += txtwrapper.fill(line) + "\n"
outtxt = outtxt.replace("-", "-")
outtxt = outtxt.replace("<", "<")
outtxt = outtxt.replace(">", ">")
return outtxt
def transfer_to_gitlab(launchpad, project, bug_task):
bug = bug_task.bug
desc = "This bug has been copied automatically from: " \
+ bug_task.web_link \
+ "<br/>\nReported by '[" + bug.owner.display_name \
+ "](https://launchpad.net/~" + bug.owner.name + ")' "
desc += "on " \
+ bug.date_created.date().isoformat() + " :\n\n" \
+ "<pre>" + preptext(bug.description) + "</pre>\n"
issue = project.issues.create({'title': bug.title, 'description': desc},
retry_transient_errors = True)
for msg in bug.messages:
has_attachment = False
attachtxt = "\n**Attachments:**\n\n"
for attachment in bug_task.bug.attachments:
if attachment.message == msg:
has_attachment = True
attachtxt += "* [" + attachment.title + "](" \
+ attachment.data_link + ")\n"
note = "Comment from '[" + msg.owner.display_name \
+ "](" + msg.owner.web_link + ")' on Launchpad (" \
+ msg.date_created.date().isoformat() + "):\n"
if msg == bug.messages[0] or not msg.content.strip():
if not has_attachment:
continue
else:
note += "\n<pre>" + preptext(msg.content) + "</pre>\n"
if has_attachment:
note += attachtxt
issue.notes.create({'body': note}, retry_transient_errors = True)
time.sleep(0.2) # To avoid "spamming"
labels = convert_tags(bug.tags)
labels.append("Launchpad")
issue.labels = labels
issue.save(retry_transient_errors = True)
print(" ==> %s" % issue.web_url)
if not args.anonymous:
mark_lp_bug_moved(bug_task, issue.web_url)
if args.open_url:
os.system("xdg-open " + issue.web_url)
def main():
print("LP2GL", args)
if not args.lp_project_name:
print("Please specify a Launchpad project name (with -l)")
return
launchpad = get_launchpad()
lp_project = launchpad.projects[args.lp_project_name]
if args.reporter:
bug_tasks = lp_project.searchTasks(
status=["New", "Confirmed", "Triaged"],
bug_reporter="https://api.launchpad.net/1.0/~" + args.reporter,
omit_duplicates=True,
order_by="datecreated")
elif args.search_text:
bug_tasks = lp_project.searchTasks(
status=["New", "Confirmed", "Triaged", "In Progress"],
search_text=args.search_text,
omit_duplicates=True,
order_by="datecreated")
else:
bug_tasks = lp_project.searchTasks(
status=["New", "Confirmed", "Triaged"],
omit_duplicates=True,
order_by="datecreated")
if args.gl_project_id:
try:
priv_token = os.environ['GITLAB_PRIVATE_TOKEN']
except Exception as e:
print("Please set the GITLAB_PRIVATE_TOKEN env variable!")
return
gl = gitlab.Gitlab('https://gitlab.com', private_token=priv_token)
gl.auth()
project = gl.projects.get(args.gl_project_id)
else:
print("Provide a Gitlab project ID to transfer the bugs ('-g')")
batch_size = args.batch_size
for bug_task in bug_tasks:
if batch_size < 1 :
break
owner = bug_task.owner.name
if args.open_url:
os.system("xdg-open " + bug_task.bug.web_link)
show_bug_task(bug_task)
if args.gl_project_id:
time.sleep(2) # To avoid "spamming"
transfer_to_gitlab(launchpad, project, bug_task)
batch_size -= 1
print("All done.")
if __name__ == '__main__':
main()
You need to specify at least a Launchpad project name with the -l
parameter
(for example -l qemu-kvm
), and for simple initial tests it might be good to
use -a
for an anonymous Launchpad login, too (the Launchpad ticket won’t be
updated in that case). Without further parameters, this will just list the
tickets in the Launchpad project that are still opened.
To transfer tickets to a Gitlab issue tracker, you need to specify the
Gitlab project ID with the -g
parameter (which can be found on the main
page of your project on Gitlab) and provide a
Gitlab access token
for the API via the GITLAB_PRIVATE_TOKEN
environment variable.
Anyway, if you want to use the script, I recommend to test it with anonymous
access for Launchpad (i.e. with the -a
parameter) and a dummy project on
Gitlab first (which you can just delete afterwards). This way you can get a
basic understanding and impression of the script first, before you use it for
the final transfer of your bug tickets.
The two previous blog posts about why git forges are von Neumann machines and the Radicle peer-to-peer git forge explored models for git forges. In this final post I want to cover yet another model that draws from the previous ones but has its own unique twist.
I previously showed how applications can be built on centralized git forges using CI/CD functionality for executing code, webhooks for interacting with the outside world, and disjoint branches for storing data.
A more elegant architecture is a peer-to-peer one where instead of many clients and one server there are just peers. Each peer has full access to the data. There is no client/server application code split, instead each peer runs an application for itself.
First, this makes it easier to move the data to new hosting infrastructure or fork a project since all data resides in the git repository. Merge requests, issues, wikis, and even the app settings are all stored in the git repo itself.
Second, this gives more power to the users who can process data however they want without being limited by the server's API. All peers are on equal footing and users don't need permission to alter applications, because they run locally.
Finally, it is easier to develop a local application than a client/server application. Being able to open a file and tweak the code is immediate and less hassle than testing and deploying a server-side application.
Internet peer-to-peer systems typically still require some central point for bootstrapping and this is no exception. A publicly-accessible git repository is still needed so that peers can fetch and push changes. However, in this model the git server does not run application code but "git apps" like merge requests, issue trackers, wikis, etc can still be implemented. Here is how it works...
The git server is not allowed to run application code in our model, so apps like merge requests won't be processing data on the server side. However, the repository does need some primitives to make peer-to-peer git apps possible. These primitives are access control policies for refs and directories/files.
Peers run applications locally and the git server is "dumb" with the sole job of enforcing access control. You can imagine this like a multi-user UNIX machine where users have access to a shared directory. UNIX file permissions determine how processes can access the data. By choosing permissions carefully, multiple users can collaborate in the shared directory in a safe and controlled manner.
This is an anti-application server because no application code runs on the server side. The server is just a git repository that stores data and enforces access control on git push.
Repositories that accept push requests need a pre-receive hook (see githooks(5)) that checks incoming requests against the access control policy. If the request complies with the access control policy then the git push is accepted. Otherwise the git push is rejected and changes are not made to the git repository.
The first type of access control is on git refs. Git refs are the namespace where branches and tags are stored in a git repository. If a regular expression matches the ref and the operation type (create, fast-forward, force, delete) then it is allowed. For example, this policy rule allows any user to push to refs/heads/foo but force pushes and deletion are not allowed:
anyone create,fast-forward ^heads/foo$
The operations available on refs include:
Operation | Description |
---|---|
create-branch | Push a new branch that doesn't exist yet |
create-tag | Push a new tag that doesn't exist yet |
fast-forward | Push a commit that is a descendent of the current commit |
force | Push a commit or tag replacing the previous ref |
delete | Delete a ref |
What's more interesting is that $user_id is expanded to the git push user's identifier so we can write rules to limit access to per-user ref namespaces:
anyone create-branch,fast-forward,force,delete ^heads/$user_id/.*$
This would allow Alice to push her own branches but Alice could not push to Bob's branches.
We have covered how to define access control policies on refs. Access control policies are also needed on branches so that multiple users can modify the same branch in a controlled and safe manner. The syntax is similar but the policy applies to changes made by commits to directories/files (what git calls a tree). The following allows users to create files in a directory but not delete or modify them (somewhat similar to the UNIX restricted deletion or "sticky" bit on world-writable directories):
anyone create-file ^shared-dir/.*$
The operations available on branches include:
Operation | Description |
---|---|
create-directory | Create a new directory |
create-file | Create a new file |
create-symlink | Create a symlink |
modify | Change an existing file or symlink |
delete-file | Delete a file |
... |
$user_id expansion is also available for branch access control. Here the user can create, modify, and delete files in a per-user directory:
anyone create-file,modify,delete-file ^$user_id/.*$
You might be wondering how user identifiers work. Git supports GPG-signed push requests with git push --signed. We can use the GPG key ID as the user identifier, eliminating the need for centralized user accounts. Remember that the GPG key ID is based on the public key. Key pairs are randomly generated and it is improbable that the same key will be generated by two different users. That said, GPG key ID uniqueness has been weak in the past when the default size was 32 bits. Git explicitly enables long 64-bit GPG key IDs but I wonder if collisions could be a problem. Maybe an ID with more bits based on the public key should be used instead, but for now let's assume the GPG key ID is unique.
The downside of this approach is that user IDs are not human-friendly. Git apps can allow the user to assign aliases to avoid displaying raw user IDs. Doing this automatically either requires an external ID issuer like confirming email address ownership, which is tedious for new users, or by storing a registry of usernames in the git repo, which means a first-come-first-server policy for username allocation and possible conflicts when merging from two repositories that don't share history. Due to these challenges I think it makes sense to use raw GPG key IDs at the data storage level and make them prettier at the user interface level.
The GPG key ID approach works well for desktop clients but not for web clients. The web application (even if implemently on the client side) would need access to the private key so it can push to the git repository. Users should not trust remotely hosted web applications with their private keys. Maybe there is a standard Web API that can help but I'm not aware one. More thought is needed here.
The pre-receive git hook checks that signature verification passed and has access to the GPG key ID in the GIT_PUSH_CERT_KEY environment variable. Then the access control policy can be checked.
Access control is the first and most fundamental git app. The access control policies that were described above are stored as files in the apps/access-control branch in the repository. Pushes to that branch are also subject to access control checks. Here is the branch's initial layout:
branches/ - access control policies for branches
owner.conf
groups/ - group definitions (see below)
...
refs/ - access control policies for refs
owner.conf
The default branches/owner.conf access control policy is as follows:
owner create-file,create-directory,modify,delete ^.*$
The default refs/owner.conf access control policy is as follows:
owner create-branch,create-tag,fast-foward,force,delete ^.*$
This gives the owner the ability to push refs and modify branches as they wish. The owner can grant other users access by pushing additional access control policy files or changing exsting files on the apps/access-control branch.
Each access control policy file in refs/ or branches/ is processed in turn. If no access control rule matches the operation then the entire git push is rejected.
Groups can be defined to alias one or more user identifiers. This avoids duplicating access control rules when more than one user should have the same access. There are two automatic groups: owner contains just the user who owns the git repository and anyone is the group of all users.
This completes the description of the access control app. Now let's look at how other functionality is built on top of this.
A merge requests app can be built on top of this model. The refs access control policy is as follows:
# The data branch contains the titles, comments, etc
anyone modify ^apps/merge-reqs/data$
# Each merge request revision is pushed as a tag in a per-user namespace
anyone create-tag ^apps/merge-reqs/$user_id/[0-9]+-v[0-9]+$
The branch access control policy is:
# Merge requests are per-user and numbered
anyone create-directory ^merge-reqs/$user_id/[0-9]+$
# Title string
anyone create-file,modify ^merge-reqs/$user_id/[0-9]+/title$
# Labels (open, needs-review, etc) work like this:
#
# merge-reqs/<user-id>/<merge-req-num>/labels/
# needs-review -> /labels/needs-review
# ...
# labels/
# needs-review/
# <user-id>/
# <merge-req-num> -> /merge-reqs/<user-id>/<merge-req-num>
# ...
# ...
# ...
#
# This directory and symlink layout makes it possible to enumerate labels for a
# given merge request and to enumerate merge requests for a given label.
#
# Both the merge request author and maintainers can add/remove labels to/from a
# merge request.
anyone create-directory ^merge-reqs/[^/]+/[0-9]+/labels$
anyone create-symlink,delete ^merge-reqs/$user_id/[0-9]+/labels/.*$
maintainers create-symlink,delete ^merge-reqs/[^/]+/[0-9]+/labels/.*$
maintainers create-directory ^labels/[^/]+$
anyone create-symlink,delete ^labels/[^/]+/$user_id/[0-9]+$
maintainers create-symlink,delete ^labels/[^/]+/[^/]+/[0-9]+$
# Comments are stored as individual files in per-user directories. Each file
# contains a timestamp and the contents of the comment. The timestamp can be
# used to sort comments chronologically.
anyone create-directory ^merge-reqs/[^/]+/[0-9]+/comments$
anyone create-directory ^merge-reqs/[^/]+/[0-9]+/comments/$user_id$
anyone create-file,modify ^merge-reqs/[^/]+/[0-9]+/comments/$user_id/[0-9]+$
When a user creates a merge request they provide a title, an initial comment, apply labels, and push a v1 tag for review and merging. Other users can comment by adding files into the merge request's per-user comments directory. Labels can be added and removed by changing symlinks in the labels directories.
The user can publish a new revision of the merge request by pushing a v2 tag and adding a comment describing the changes. Once the maintainers are satisfied they merge the final revision tag into the relevant branch (e.g. "main") and relabel the merge request from open/needs-review to closed/merged.
This workflow can be implemented by a tool that performs the necessary git operations so users do not need to understand the git app's internal data layout. Users just need to interact with the tool that displays merge requests, allows commenting, provides searches, etc. A natural way to implement this tool is as a git alias so it integrates alongside git's built-in commands.
One issue with this approach is that it uses the file system as a database. Performance and scalability are likely to be worst than using a database or application-specific file format. However, the reason for this approach is that it allows the access control app to enforce a policy that ensures users cannot modify or delete other user's data without running application-specific code on the server and while keeping everything stored in a git repository.
An example where this approach performs poorly is for full-text search. The application would need to search all title and comment files for a string. There is no index for efficient lookups. However, if applications find that git-grep(1) does not perform well they can maintain their own index and cache files locally.
I hope that this has shown how git apps can be built without application code running on the server.
Now that we have the merge requests app it's time to think how a continuous integration service could interface with it. The goal is to run tests on each revision of a merge request and report failures so the author of the merge request can rectify the situation.
A CI bot watches the repository for changes. In particular, it needs to watch for tags created with the ref name apps/merge-reqs/[^/]+/[0-9]+-v[0-9]+.
When a new tag is found the CI bot checks it out and runs tests. The results of the tests are posted as a comment by creating a file in merge-regs/<user-id>/>merge-req-num>/comments/<ci-bot-user-id>/0 on the apps/merge-reqs/data branch. A ci-pass or ci-fail label can also be applied to the merge request so that the CI status can be easily queried by users and tools.
There are many loose ends. How can non-git users participate on issue trackers and wikis? It might be possible to implement a full peer as a client-side web application using isomorphic-git, a JavaScript git implementation. As mentioned above, the GPG key ID approach is not very browser-friendly because it requires revealing the private key to the web page and since keys are user identifiers using temporary keys does not work well.
The data model does not allow efficient queries. A full copy of the data is necessary in order to query it. That's acceptable for local applications because they can maintain their own indexes and are expected to keep the data for a long period of time. It works less well for short-lived web page sessions like a casual user filing a new bug on the issue tracker.
The git push --signed technique is not the only option. Git also supports signed commits and signed tags. The difference between signed pushes and signed tags/commits is significant. The signed push approach only validates the access control policy when the repository is changed and leaves no audit log for future reference. The signed commit/tag approach keeps the signatures in the git history. Signed commits/tags can be propagated in a peer-to-peer network and each peer can validate the access control policy itself. While signed commits/tags apply the access control policy to each object in the repository, signed pushes apply the access control policy to each change made to the repository. The difference is that it's easy to rebase and include work from different authors with signed pushes. Signed commits/tags require re-signing for rebasing and each commit is validated against its signature, which may be different from the user who is making the push request.
There are a lot of interesting approaches and trade-offs to explore here. This model we've discussed fits closely with how I've seen developers use git in open source projects. It is designed around a "main" repository/server that contributors push their code to. But each clone of the repository has all the data and can be published as a new "main" repository, if necessary.
Although these ideas are unfinished I decided to write them up with the knowledge that I probably won't implement them myself. QEMU is moving to GitLab with a traditional centralized git forge. I don't think this is the right time to develop this idea and try to convince the QEMU community to use it. For projects that have fewer infrastructure requirements it would give their contributors more power than being confined to a centralized git forge.
I hope this was an interesting read for anyone thinking about git forges and building git apps.
After releasing RHEL 8.7 a week before, Red Hat now published RHEL 9.1, see the press release here! It ships, among others:
Note that RHEL9.1 is NOT an EUS (Extended Update Support) release, so it will go out of support with the GA of RHEL 9.2. For details, please see the "Red Hat Enterprise Linux Life Cycle" here.
The new version if IBM Cloud Infrastructure Center is available and has several improvements for KVM on IBM zSystems:
At KVM Forum 2022 Kevin Wolf and Stefano Garzarella gave a talk on qemu-storage-daemon, a way to get QEMU's storage functionality without running a VM. It's great for accessing disk images, basically taking the older qemu-nbd to the next level. The cool thing is this makes QEMU's software-defined storage functionality - block devices with snapshots, incremental backup, image file formats, etc - available to other programs. Backup and forensics tools as well as other types of programs can take advantage of qemu-storage-daemon.
Here is the full article about Accessing QEMU storage features without a VM. Enjoy!
Our solution assurance team published a new paper, providing guidance together with hints and tricks and practical examples to help you configure and use the Red Hat Enterprise Linux High Availability Add-on (Red Hat HA).
You can access the paper here.
This summer I worked on adding Zoned Block Device (ZBD) support to virtio-blk as part of the Outreachy internship program. QEMU hasn’t directly supported ZBDs before so this article explains how they work and why QEMU needed to be extended.
Zoned block devices (ZBDs) are divided into regions called zones that can only be written sequentially. By only allowing sequential writes, SSD write amplification can be reduced by eliminating the need for a Flash Translation Layer, and potentially lead to higher throughput and increased capacity. Providing a new storage software stack, zoned storage concepts are standardized as ZBC (SCSI standard), ZAC (ATA standard), and ZNS (NVMe). Meanwhile, the virtio protocol for block devices(virtio-blk) should also be aware of ZBDs instead of taking them as regular block devices. It should be able to pass such devices through to the guest. An overview of necessary work is as follows:
Once the QEMU and Linux patches have been merged it will be possible to expose a virtio-blk ZBD to the guest like this:
-blockdev node-name=drive0,driver=zoned_host_device,filename=/path/to/zbd,cache.direct=on \
-device virtio-blk-pci,drive=drive0 \
And then we can perform zoned block commands on that device in the guest os.
# blkzone report /dev/vda
start: 0x000000000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x000020000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x000040000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x000060000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x000080000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x0000a0000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x0000c0000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x0000e0000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 0(nw) [type: 1(CONVENTIONAL)]
start: 0x000100000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)]
start: 0x000120000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)]
start: 0x000140000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)]
start: 0x000160000, len 0x020000, cap 0x020000, wptr 0x000000 reset:0 non-seq:0, zcond: 1(em) [type: 2(SEQ_WRITE_REQUIRED)]
Currently, QEMU can support zoned devices by virtio-scsi or PCI device passthrough. It needs to specify the device type it is talking to. Whereas storage controller emulation uses block layer APIs instead of directly accessing disk images. Extending virtio-blk emulation avoids code duplication and simplify the support by hiding the device types under a unified zoned storage interface, simplifying VM deployment for different types of zoned devices. Virtio-blk can also be implemented in hardware. If those devices wish to follow the zoned storage model then the virtio-blk specification needs to natively support zoned storage. With such support, individual NVMe namespaces or anything that is a zoned Linux block device can be exposed to the guest without passing through a full device.
For zoned storage emulation, zoned storage APIs support three zoned models (conventional, host-managed, host-aware) , four zone management commands (Report Zone, Open Zone, Close Zone, Finish Zone), and Append Zone. The QEMU block layer has a BlockDriverState graph that propagates device information inside block layer. File-posix driver is the lowest level within the graph where zoned storage APIs reside.
After receiving the block driver states, Virtio-blk emulation recognizes zoned devices and sends the zoned feature bit to guest. Then the guest can see the zoned device in the host. When the guest executes zoned operations, virtio-blk driver issues corresponding requests that will be captured by viritio-blk device inside QEMU. Afterwards, virtio-blk device sends the requests to file-posix driver which will perform zoned operations using Linux ioctls.
Unlike zone management operations, Linux doesn’t have a user API to issue zone append requests to zoned devices from user space. With the help of write pointer emulation tracking locations of write pointer of each zone, QEMU block layer can perform append writes by modifying regular writes. Write pointer locks guarantee the execution of requests. Upon failure it must not update the write pointer location which is only got updated when the request is successfully finished.
Problems can always be solved with right mind and right tools. A good approach to avoid pitfalls of programs is test-driven. In the beginning, users like qemu-io commands utility can invoke new block layer APIs. Moving towards to guest, existing tools like blktests, zonefs-tools, and fio are introduced for broader testing. Depending on the size of the zoned device, some tests may take long enough time to finish. Besides, tracing is also a good tool for spotting bugs. QEMU tracking tools and blktrace monitors block layer IO, providing detailed information to analysis.
As a student interested in computer science, I am enthusiastic about making real applications and fortunate to find the opportunity in this summer. I have a wonderful experience with QEMU where I get chance to work with experienced engineers and meet peers sharing same interests. It is a good starting point for me to continue my search on storage systems and open source projects.
Public communication, reaching out to people and admitting to failures used to be hard for me. Those feelings had faded away as I put more effort to this project over time. For people may having the same trouble as me, it might be useful to focus on the tasks ahead of you instead of worrying about the consequences of rejections from others.
Finally, I would like to thank Stefan Hajnoczi, Damien Le Moal, Dmitry Fomichev, and Hannes Reinecke for mentoring me - they have guided me through this project with patience and expertise, when I hit obstacles on design or implementations, and introduced a fun and vibrant open source world for me. Also thank QEMU community and Outreachy for organizing this program.
The current status for this project is waiting for virtio specifications extension and Linux driver support patches got accepted. And the up-to-date patch series of zoned device support welcome any new comments.
The next step for zoned storage emulation in QEMU is to enable full zoned emulation through virtio-blk. Adding support on top of a regular file, it allows developers accessing a zoned device environment without real zoned storage hardwares. Furthermore, virtio-scsi may need to add full emulation support to complete the zoned storage picture in QEMU. QEMU NVMe ZNS emulation can also use new block layer APIs to attach real zoned storage if the emulation is used in production in future.
RHEL 8.7 is out, see the blog entry here! It ships, among others:
For a detailed list of Linux on Z-specific changes, see the release notes.
vhost-user-blk is a high-performance storage protocol that connects virtual machines to software-defined storage like SPDK or qemu-storage-daemon. Until now, tool support for vhost-user-blk has been lacking. Accessing vhost-user-blk devices involved running a virtual machine, which requires more setup than one would like.
QEMU 7.2 adds vhost-user-blk support to the qemu-img tool. This is possible thanks to libblkio, a library that other programs besides QEMU can use too.
Check for vhost-user-blk support in your installed qemu-img version like this (if it says 0 then you need to update qemu-img or compile it from source with libblkio enabled):
$ qemu-img --help | grep virtio-blk-vhost-user | wc -l
1
You can copy a raw disk image file into a vhost-user-blk device like this:
$ qemu-img convert \
--target-image-opts \
-n \
test.img \
driver=virtio-blk-vhost-user,path=/tmp/vhost-user-blk.sock,cache.direct=on
The contents of the vhost-user-blk device can be saved as a qcow2 image file like this:
$ qemu-img convert \
--image-opts \
-O qcow2 \
driver=virtio-blk-vhost-user,path=/tmp/vhost-user-blk.sock,cache.direct=on out.qcow2
The size of the virtual disk can be read:
$ qemu-img info \
--image-opts \
driver=virtio-blk-vhost-user,path=/tmp/vhost-user-blk.sock,cache.direct=on
image: json:{"driver": "virtio-blk-vhost-user"}
file format: virtio-blk-vhost-user
virtual size: 4 GiB (4294967296 bytes)
disk size: unavailable
Other qemu-img sub-commands like bench and dd are also available for quickly accessing the vhost-user-blk device without running a virtual machine:
$ qemu-img bench \
--image-opts \
driver=virtio-blk-vhost-user,path=/tmp/vhost-user-blk.sock,cache.direct=on
Sending 75000 read requests, 4096 bytes each, 64 in parallel (starting at offset 0, step size 4096)
Run completed in 1.443 seconds.
Being able to access vhost-user-blk devices from qemu-img makes vhost-user-blk a little easier to work with.
Canonical released a new version of their Ubuntu server offering Ubuntu Server 22.10!
One of the highlights is the addition of a new feature providing Secure Execution attestation.
See the announcement on the mailing list here, and the blog entry at Canonical with Z-specific highlights here.
Starting with RHEL8.6, it is no longer necessary to use the RHEL-AV variant to get updated and new versions of QEMU/libvirt etc. Check this link how to configure dnf/yum to use the RHEL versions instead.
We’d like to announce the availability of the QEMU 7.1.0 release. This release contains 2800+ commits from 238 authors.
You can grab the tarball from our download page. The full list of changes are available in the Wiki.
Highlights include:
Thank you to everyone involved!
Some people already noticed and asked questions. So guess I better write things down in my blog so I don't have to answer the questions over and over again, and I hope to also clarify some things on distro firmware builds.
So, yes, the jenkins autobuilder creating the firmware repository at https://www.kraxel.org/repos/jenkins/ has been shutdown yesterday (Jul 19th 2020). The repository will stay online for the time being, so your establish workflows will not instantly break. But the repository will not get updates any more, so it is wise to start looking for alternatives now.
The obvious primary choice would be to just use the firmware builds provided by your distribution. I'll cover edk2 only, which seems to be the by far most popular use, even thought here are also builds for other firmware projects.
Given I'm quite familier with the RHEL / Fedora world I can give
some advise here. The edk2-ovmf
package comes with
multiple images for the firmware code and the varstore template
which allow for various combinations. The most important ones are:
OVMF_CODE.secboot.fd
and OVMF_VARS.secboot.fd
OVMF_CODE.secboot.fd
and OVMF_VARS.fd
OVMF_CODE.fd
and OVMF_VARS.fd
The classic way to setup this in libvirt looks like this:
<domain type='kvm'>
[ ... ]
<os>
<type arch='x86_64' machine='q35'>hvm</type>
<loader readonly='yes' type='pflash'>/usr/share/OVMF/OVMF_CODE.secboot.fd</loader>
<nvram template='/usr/share/OVMF/OVMF_VARS.fd'/>
</os>
To make this easier the firmware builds come with json files
describing the capabilities and requirements. You can find these
files in /usr/share/qemu/firmware/
. libvirt can use
them to automatically find suitable firmware images, so you don't
have to write the firmware image paths into the domain
configuration. You can simply use this instead:
<domain type='kvm'>
[ ... ]
<os firmware='efi'>
<type arch='x86_64' machine='q35'>hvm</type>
</os>
libvirt also allows to ask for specific firmware features. If you don't want use secure boot for example you can ask for the blank varstore template (no secure boot keys enrolled) this way:
<domain type='kvm'>
[ ... ]
<os firmware='efi'>
<type arch='x86_64' machine='q35'>hvm</type>
<firmware>
<feature name='enrolled-keys' enabled='no' />
</firmware>
</os>
In case you change the configuration of an existing virtual machine
you might (depending on the kind of change) have to run virsh
start --reset-nvram domain
once to to start over with
a fresh copy of the varstore template.
The world has moved forward. UEFI isn't a niche use case any more. Linux distributions all provide good packages theys days. The edk2 project got good CI coverage (years ago it was my autobuilder raising the flag when a commit broke the gcc build). The edk2 project got a regular release process distros can (and do) follow.
All in all the effort to maintain the autobuilder doesn't look justified any more.
Are you using CryptoExpress cards with KVM on IBM zSystems or LinuxONE? Sebastian Mitterle has a very good overview on how to make crypto device passthrough persistent.
http://learningbytesting.mathume.com/2022/07/persistent-crypto-device-passthrough-on.html
A recent thread on the Fedora development list about unified kernel images co-incided with work I’m involved in wrt confidential computing (AMD SEV[-SNP], Intel TDX, etc). In exploring the different options for booting virtual machines in a confidential computing environment, one of the problems that keeps coming up is that of validating the boot measurements of the initrd and kernel command line. The initrd is currently generated on the fly at the time the kernel is installed on a host, while the command line typically contains host specific UUIDs for filesystems or LUKS volumes. Before even dealing with those problems, grub2
‘s support for TPMs causes pain due to its need to measure every single grub.conf
configuration line that is executed into a PCR. Even with the most minimal grub.conf
using autodiscovery based on the boot loader spec, the grub.con
f boot measurements are horribly cumbersome to deal with.
With this in mind, in working on confidential virtualization, we’re exploring options for simplifying the boot process by eliminating any per-host variable measurements. A promising way of achieving this is to make use of sd-boot
instead of grub2
, and using unified kernel images pre-built and signed by the OS vendor. I don’t have enough familiarity with this area of Linux, so I’ve been spending time trying out the different options available to better understand their operation. What follows is a short description of how i took an existing Fedora 36 virtual machine and converted it to sd-boot
with a unified kernel image.
First of all, I’m assuming that the virtual machine has been installed using UEFI (EDK2’s OOVMF build) as the firmware, rather than legacy BIOS (aka SeaBIOS). This is not the default with virt-manager
/virt-install
, but an opt-in is possible at time of provisioning the guest. Similarly it is possible to opt-in to adding a virtual TPM to the guest, for the purpose of receiving boot measurements. Latest upstream code for virt-manager
/virt-install
will always add a vTPM if UEFI is requested.
Assuming UEFI + vTPM are enabled for the guest, the default Fedora / RHEL setup will also result in SecureBoot being enabled in the guest. This is good in general, but the sd-boot
shipped in Fedora is not currently signed. Thus for (current) testing, either disable SecureBoot, or manually sign the sd-boot
binary with a local key and enroll that key with UEFI. SecureBoot isn’t immediately important, so the quickest option is disabling SecureBoot with the following libvirt guest XML config setup:
<os firmware='efi'> <type arch='x86_64' machine='pc-q35-6.2'>hvm</type> <firmware> <feature enabled='no' name='secure-boot'/> </firmware> <loader secure='no'/> <boot dev='hd'/> </os>
The next time the guest is cold-booted, the ‘--reset-nvram
‘ flag needs to be passed to ‘virsh start
‘ to make it throwaway the existing SecureBoot enabled NVRAM and replace it with one disabling SecureBoot.
$ virsh start --reset-nvram fedora36test
Inside the guest, surprisingly, there were only two steps required, installing ‘sd-boot’ to the EFI partition, and building the unified kernel images. Installing ‘sd-boot’ will disable the use of grub, so don’t reboot after this first step, until the kernels are setup:
$ bootctl install Created "/boot/efi/EFI/systemd". Created "/boot/efi/loader". Created "/boot/efi/loader/entries". Created "/boot/efi/EFI/Linux". Copied "/usr/lib/systemd/boot/efi/systemd-bootx64.efi" to "/boot/efi/EFI/systemd/systemd-bootx64.efi". Copied "/usr/lib/systemd/boot/efi/systemd-bootx64.efi" to "/boot/efi/EFI/BOOT/BOOTX64.EFI". Updated /etc/machine-info with KERNEL_INSTALL_LAYOUT=bls Random seed file /boot/efi/loader/random-seed successfully written (512 bytes). Not installing system token, since we are running in a virtualized environment. Created EFI boot entry "Linux Boot Manager".
While the ‘/boot/efi/loader
‘ directory could be populated with config files specifying kernel/initrd/cmdline to boot, the desire is to be able to demonstrate booting with zero host local configuration. So the next step is to build and install the unified kernel image. The Arch Linux wiki has a comprehensive guide, but the easiest option for Fedora appears to be to use dracut
with its ‘--uefi
‘ flag
$ for i in /boot/vmlinuz-*x86_64 do kver=${i#/boot/vmlinuz-} echo "Generating $kver" dracut --uefi --kver $kver --kernel-cmdline "root=UUID=5fd49e99-6297-4880-92ef-bc31aef6d2f0 ro rd.luks.uuid=luks-6806c81d-4169-4e7a-9bbc-c7bf65cabcb2 rhgb quiet" done Generating 5.17.13-300.fc36.x86_64 Generating 5.17.5-300.fc36.x86_64
The observant will notice the ‘–kernel-cmdline’ argument refers to install specific UUIDs for the LUKS volume and root filesystem. This ultimately needs to be eliminated too, which would require configuring the guest disk image to comply with the discoverable partitions spec. That is beyond the scope of my current exercise of merely demonstrating use of sd-boot and unified kernels. It ought to be possible to write a kickstart file to automate creation of a suitable cloud image though.
At this point the VM is rebooted, and watching the graphical console confirms that the grub menu has disappeared and display output goes straight from the UEFI splash screen into Linux. There’s no menu shown by sd-boot
by default, but if desired this can be enabled by editing /boot/efi/loader/loader.conf
to uncomment the line timeout 3
, at which point it will show the kernel version selection at boot.
If following this scheme, bear in mind that nothing is wired up to handle this during kernel updates. The kernel RPM triggers will continue to setup grub.conf
and generate standalone initrds. IOW don’t try this on a VM that you care about. I assume there’s some set of commands I could use to uninstall sd-boot
and switch back to grub
, but I’ve not bothered to figure this out.
Overall this exercise was suprisingly simple and painless. The whole idea of using a drastically simplified boot loader instead of grub, along with pre-built unified kernel images, feels like it has alot of promise, especially in the context of virtual machines where the set of possible boot hardware variants is small and well understood.
A new version of the virtio specification has been released! As it has been three years after the 1.1 release, quite a lot of changes have accumulated. I have attempted to list some of them below; for details, you are invited to check out the spec :)
There are already some changes queued for 1.3; let’s hope it won’t take us three years again before the next release ;)
Several new device types have been added.
Enhancements have been added to some already existing device types.
Some general enhancements include:
The Steam Deck is a handheld gaming computer that runs a Linux-based operating system called SteamOS. The machine comes with SteamOS 3 (code name “holo”), which is in turn based on Arch Linux.
Although there is no SteamOS 3 installer for a generic PC (yet), it is very easy to install on a virtual machine using QEMU. This post explains how to do it.
The goal of this VM is not to play games (you can already install Steam on your computer after all) but to use SteamOS in desktop mode. The Gamescope mode (the console-like interface you normally see when you use the machine) requires additional development to make it work with QEMU and will not work with these instructions.
A SteamOS VM can be useful for debugging, development, and generally playing and tinkering with the OS without risking breaking the Steam Deck.
Running the SteamOS desktop in a virtual machine only requires QEMU and the OVMF UEFI firmware and should work in any relatively recent distribution. In this post I’m using QEMU directly, but you can also use virt-manager or some other tool if you prefer, we’re emulating a standard x86_64 machine here.
SteamOS is a single-user operating system and it uses an A/B partition scheme, which means that there are two sets of partitions and two copies of the operating system. The root filesystem is read-only and system updates happen on the partition set that is not active. This allows for safer updates, among other things.
There is one single /home
partition, shared by both partition sets. It contains the games, user files, and anything that the user wants to install there.
Although the user can trivially become root, make the root filesystem read-write and install or change anything (the pacman
package manager is available), this is not recommended because
A simple way for the user to install additional software that survives OS updates and doesn’t touch the root filesystem is Flatpak. It comes preinstalled with the OS and is integrated with the KDE Discover app.
The first thing that we need is the installer. For that we have to download the Steam Deck recovery image from here: https://store.steampowered.com/steamos/download/?ver=steamdeck&snr=
Once the file has been downloaded, we can uncompress it and we’ll get a raw disk image called steamdeck-recovery-4.img
(the number may vary).
Note that the recovery image is already SteamOS (just not the most up-to-date version). If you simply want to have a quick look you can play a bit with it and skip the installation step. In this case I recommend that you extend the image before using it, for example with ‘truncate -s 64G steamdeck-recovery-4.img
‘ or, better, create a qcow2 overlay file and leave the original raw image unmodified: ‘qemu-img create -f qcow2 -F raw -b steamdeck-recovery-4.img steamdeck-recovery-extended.qcow2 64G
‘
But here we want to perform the actual installation, so we need a destination image. Let’s create one:
$ qemu-img create -f qcow2 steamos.qcow2 64G
Now that we have all files we can start the virtual machine:
$ qemu-system-x86_64 -enable-kvm -smp cores=4 -m 8G \ -device usb-ehci -device usb-tablet \ -device intel-hda -device hda-duplex \ -device VGA,xres=1280,yres=800 \ -drive if=pflash,format=raw,readonly=on,file=/usr/share/ovmf/OVMF.fd \ -drive if=virtio,file=steamdeck-recovery-4.img,driver=raw \ -device nvme,drive=drive0,serial=badbeef \ -drive if=none,id=drive0,file=steamos.qcow2
Note that we’re emulating an NVMe drive for steamos.qcow2
because that’s what the installer script expects. This is not strictly necessary but it makes things a bit easier. If you don’t want to do that you’ll have to edit ~/tools/repair_device.sh
and change DISK
and DISK_SUFFIX
.
Once the system has booted we’ll see a KDE Plasma session with a few tools on the desktop. If we select “Reimage Steam Deck” and click “Proceed” on the confirmation dialog then SteamOS will be installed on the destination drive. This process should not take a long time.
Now, once the operation finishes a new confirmation dialog will ask if we want to reboot the Steam Deck, but here we have to choose “Cancel”. We cannot use the new image yet because it would try to boot into the Gamescope session, which won’t work, so we need to change the default desktop session.
SteamOS comes with a helper script that allows us to enter a chroot after automatically mounting all SteamOS partitions, so let’s open a Konsole and make the Plasma session the default one in both partition sets:
$ sudo steamos-chroot --disk /dev/nvme0n1 --partset A
# steamos-readonly disable
# echo '[Autologin]' > /etc/sddm.conf.d/zz-steamos-autologin.conf
# echo 'Session=plasma.desktop' >> /etc/sddm.conf.d/zz-steamos-autologin.conf
# steamos-readonly enable
# exit
$ sudo steamos-chroot --disk /dev/nvme0n1 --partset B
# steamos-readonly disable
# echo '[Autologin]' > /etc/sddm.conf.d/zz-steamos-autologin.conf
# echo 'Session=plasma.desktop' >> /etc/sddm.conf.d/zz-steamos-autologin.conf
# steamos-readonly enable
# exit
After this we can shut down the virtual machine. Our new SteamOS drive is ready to be used. We can discard the recovery image now if we want.
To boot SteamOS we can use a QEMU line similar to the one used during the installation. This time we’re not emulating an NVMe drive because it’s no longer necessary.
$ cp /usr/share/OVMF/OVMF_VARS.fd . $ qemu-system-x86_64 -enable-kvm -smp cores=4 -m 8G \ -device usb-ehci -device usb-tablet \ -device intel-hda -device hda-duplex \ -device VGA,xres=1280,yres=800 \ -drive if=pflash,format=raw,readonly=on,file=/usr/share/ovmf/OVMF.fd \ -drive if=pflash,format=raw,file=OVMF_VARS.fd \ -drive if=virtio,file=steamos.qcow2 \ -device virtio-net-pci,netdev=net0 \ -netdev user,id=net0,hostfwd=tcp::2222-:22
(the last two lines redirect tcp port 2222 to port 22 of the guest to be able to SSH into the VM. If you don’t want to do that you can omit them)
If everything went fine, you should see KDE Plasma again, this time with a desktop icon to launch Steam and another one to “Return to Gaming Mode” (which we should not use because it won’t work). See the screenshot that opens this post.
Congratulations, you’re running SteamOS now. Here are some things that you probably want to do:
deck
user: run ‘passwd
‘ on a terminalsudo systemctl enable sshd
‘ and/or ‘sudo systemctl start sshd
‘.ssh -p 2222 deck@localhost
‘The Steam Deck recovery image doesn’t install the most recent version of SteamOS, so now we should probably do a software update.
-m 8G
). The OS update might fail if you use less.sudo steamos-select-branch beta
‘ (or main
, if you want the bleeding edge)/etc/os-release
(see the BUILD_ID
variable)steamos-update check
‘steamos-update
‘Note: if the last step fails after reaching 100% with a post-install handler error then go to Connections in the system settings, rename Wired Connection 1 to something else (anything, the name doesn’t matter), click Apply and run steamos-update
again. This works around a bug in the update process. Recent images fix this and this workaround is not necessary with them.
As we did with the recovery image, before rebooting we should ensure that the new update boots into the Plasma session, otherwise it won’t work:
$ sudo steamos-chroot --partset other
# steamos-readonly disable
# echo '[Autologin]' > /etc/sddm.conf.d/zz-steamos-autologin.conf
# echo 'Session=plasma.desktop' >> /etc/sddm.conf.d/zz-steamos-autologin.conf
# steamos-readonly enable
# exit
After this we can restart the system.
If everything went fine we should be running the latest SteamOS release. Enjoy!
SteamOS is under active development. If you find problems or want to request improvements please go to the SteamOS community tracker.
Edit 06 Jul 2022: Small fixes, mention how to install the OS without using NVMe.
Queues and their implementation using shared memory ring buffers are a standard tool for communicating with I/O devices and between CPUs. Although ring buffers are widely used, there is no standard memory layout and it's interesting to compare the differences between designs. When defining libblkio's APIs, I surveyed the ring buffer designs in VIRTIO, NVMe, and io_uring. This article examines some of the differences between the ring buffers and queue semantics in VIRTIO, NVMe, and io_uring.
A ring buffer is a circular array where new elements are written or produced on one side and read or consumed on the other side. Often terms such as head and tail or reader and writer are used to describe the array indices at which the next element is accessed. When the end of the array is reached, one moves back to the start of the array. The empty and full conditions are special states that must be checked to avoid underflow and overflow.
VIRTIO, NVMe, and io_uring all use single producer, single consumer shared memory ring buffers. This allows a CPU and an I/O device or two CPUs to communicate across a region of memory to which both sides have access.
At a minimum a ring buffer element, or descriptor, contains the memory address and size of a data buffer:
Offset | Type | Name |
---|---|---|
0x0 | u64 | buf |
0x8 | u64 | len |
In a storage device the data buffer contains a request structure with information about the I/O request (logical block address, number of sectors, etc). In order to process a request, the device first loads the descriptor and then loads the request structure described by the descriptor. Performing two loads is sub-optimal and it would be faster to fetch the request structure in a single load.
Embedding the data buffer in the descriptor is a technique that reduces the number of loads. The descriptor layout looks like this:
Offset | Type | Name |
---|---|---|
0x0 | u64 | remainder_buf |
0x8 | u64 | remainder_len |
0x10 | ... | request structure |
The descriptor is extended to make room for the data. If the size of the data varies and is sometimes too large for a descriptor, then the remainder is put into an external buffer. The common case will only require a single load but larger variable-sized buffers can still be handled with 2 loads as before.
VIRTIO does not embed data in descriptors due to its layered design. The data buffers are defined by the device type (net, blk, etc) and virtqueue descriptors are one layer below device types. They have no knowledge of the data buffer layout and therefore cannot embed data.
NVMe embeds the request structure into the Submission Queue Entry. The Command Dword 10, 11, 12, 13, 14, and 15 fields contain the request data and their meaning depends on the Opcode (request type). I/O buffers are still external and described by Physical Region Pages (PRPs) or Scatter Gather Lists (SGLs).
io_uring's struct io_uring_sqe embeds the request structure. Only I/O buffer(s) need to be external as their size varies, would be too large for the ring buffer, and typically zero-copy is desired due to the size of the data.
It seems that VIRTIO could learn from NVMe and io_uring. Instead of having small 16-byte descriptors, it could embed part of the data buffer into the descriptor so that devices need to perform fewer loads during request processing. The 12-byte struct virtio_net_hdr and 16-byte struct virtio_blk_req request headers would fit into a new 32-byte descriptor layout. I have not prototyped and benchmarked this optimization, so I don't know how effective it is.
I/O requests often include variable size I/O buffers that require scatter-gather lists similar to POSIX struct iovec arrays. Long arrays don't fit into a descriptor so descriptors have fields that point to an external array of descriptors.
Another technique for scatter-gather lists is to chain descriptors together within the ring buffer instead of relying on memory external to the ring buffer. When descriptor chaining is used, I/O requests that don't fit into a single descriptor can occupy multiple descriptors.
Advantages of chaining are better cache locality when a sequence of descriptors is used and no need to allocate separate per-request external descriptor memory.
A consequence of descriptor chaining is that the maximum queue size, or queue depth, becomes variable. It is not possible to guarantee space for specific number of I/O requests because the available number of descriptors depends on the chain size of requests placed into the ring buffer.
VIRTIO supports descriptor chaining although drivers usually forego it when VIRTIO_F_RING_INDIRECT_DESC is available.
NVMe and io_uring do not support descriptor chaining, instead relying on embedded and external descriptors.
The maximum number of in-flight requests depends on the ring buffer design. Designs where descriptors are occupied from submission until completion prevent descriptor reuse for other requests while the current request is in flight.
An alternative design is where the device processes submitted descriptors and they are considered free again as soon as the device has looked at them. This approach is natural when separate submission and completion queues are used and there is no relationship between the two descriptor rings.
VIRTIO requests occupy descriptors for the duration of their lifetime, at least in the Split Virtqueue format. Therefore the number of in-flight requests is influenced by the descriptor table size.
NVMe has separate Submission Queues and Completion Queues, but its design still limits the number of in-flight requests to the queue size. The Completion Queue Entry's SQ Head Pointer (SQHD) field precludes having more requests in flight than the Submission Queue size because the field would no longer be unique. Additionally, the driver has no way of detecting Submission Queue Head changes, so it only knows there is space for more submissions when completions occur.
io_uring has independent submission (SQ) and completions queues (CQ) with support for more in-flight requests than the ring buffer size. When there are more in-flight requests than CQ capacity, it's possible to overflow the CQ. io_uring has a backlog mechanism for this case, although the intention is for applications to properly size queues to avoid hitting the backlog often.
VIRTIO, NVMe, and io_uring have slightly different takes on queue design. The semantics and performance vary due to these differences. VIRTIO lacks data embedding inside descriptors. io_uring supports more in-flight requests than the queue size. NVMe and io_uring rely on external descriptors with no ability to chain descriptors.
Here is a quickstart for everyone who wants (or needs to) deal with edk2 firmware, with a focus on virtual machine firmware. The article assumes you are using a linux machine with gcc.
To build edk2 you need to have a bunch of tools installed. An
compiler and the make
are required of course, but also
iasl
, nasm
and libuuid
. So
install them first (package names are for centos/fedora).
dnf install -y make gcc binutils iasl nasm libuuid-devel
If you want cross-build arm firmware on a x86 machine you also need cross compilers. While being at also set the environment variables needed to make the build system use the cross compilers:
dnf install -y gcc-aarch64-linux-gnu gcc-arm-linux-gnu
export GCC5_AARCH64_PREFIX="aarch64-linux-gnu-"
export GCC5_ARM_PREFIX="arm-linux-gnu-"
Next clone the tiaocore/edk2 repository and also fetch the git submodules.
git clone https://github.com/tianocore/edk2.git
cd edk2
git submodule update --init
The edksetup
script will prepare the build environment
for you. The script must be sourced because it sets some
environment variables (WORKSPACE
being the most
important one). This must be done only once (as long as you keep
the shell with the configured environment variables open).
source edksetup.sh
Next step is building the BaseTools (also needed only once):
make -C BaseTools
Note: Currently (April 2022) BaseTools are being rewritten in Python, so most likely this step will not be needed any more at some point in the future.
Finally the build (for x64 qemu) can be kicked off:
build -t GCC5 -a X64 -p OvmfPkg/OvmfPkgX64.dsc
The firmware volumes built can be found
in Build/OvmfX64/DEBUG_GCC5/FV
.
Building the aarch64 firmware instead:
build -t GCC5 -a AARCH64 -p ArmVirtPkg/ArmVirtQemu.dsc
The build results land
in Build/ArmVirtQemu-AARCH64/DEBUG_GCC5/FV
.
Qemu expects the aarch64 firmware images being 64M im size. The firmware images can't be used as-is because of that, some padding is needed to create an image which can be used for pflash:
dd of="QEMU_EFI-pflash.raw" if="/dev/zero" bs=1M count=64
dd of="QEMU_EFI-pflash.raw" if="QEMU_EFI.fd" conv=notrunc
dd of="QEMU_VARS-pflash.raw" if="/dev/zero" bs=1M count=64
dd of="QEMU_VARS-pflash.raw" if="QEMU_VARS.fd" conv=notrunc
There are a bunch of compile time options, typically enabled
using -D NAME
or -D NAME=TRUE
. Options
which are enabled by default can be turned off using -D
NAME=FALSE
. Available options are defined in
the *.dsc
files referenced by the build
command. So a feature-complete build looks more like this:
build -t GCC5 -a X64 -p OvmfPkg/OvmfPkgX64.dsc \
-D FD_SIZE_4MB \
-D NETWORK_IP6_ENABLE \
-D NETWORK_HTTP_BOOT_ENABLE \
-D NETWORK_TLS_ENABLE \
-D TPM2_ENABLE
Secure boot support (on x64) requires SMM mode. Well, it builds and works without SMM, but it's not secure then. Without SMM nothing prevents the guest OS writing directly to flash, bypassing the firmware, so protected UEFI variables are not actually protected.
Also suspend (S3) support works with enabled SMM only in case parts of the firmware (PEI specifically, see below for details) run in 32bit mode. So the secure boot variant must be compiled this way:
build -t GCC5 -a IA32 -a X64 -p OvmfPkg/OvmfPkgIa32X64.dsc \
-D FD_SIZE_4MB \
-D SECURE_BOOT_ENABLE \
-D SMM_REQUIRE \
[ ... add network + tpm + other options as needed ... ]
The FD_SIZE_4MB
option creates a larger firmware image,
being 4MB instead of 2MB (default) in size, offering more space for
both code and vars. The RHEL/CentOS builds use that. The Fedora
builds are 2MB in size, for historical reasons.
If you need 32-bit firmware builds for some reason, here is how to do it:
build -t GCC5 -a ARM -p ArmVirtPkg/ArmVirtQemu.dsc
build -t GCC5 -a IA32 -p OvmfPkg/OvmfPkgIa32.dsc
The build results will be in
in Build/ArmVirtQemu-ARM/DEBUG_GCC5/FV
and
Build/OvmfIa32/DEBUG_GCC5/FV
The x86 firmware builds create three different images:
OVMF_VARS.fd
/var/lib/libvirt/qemu/nvram
.
OVMF_CODE.fd
OVMF.fd
CODE
and VARS
. This can be loaded as ROM
using -bios
, with two drawbacks: (a) UEFI variables
are not persistent, and (b) it does not work
for SMM_REQUIRE=TRUE
builds.
qemu handles pflash storage as block devices, so we have to create block devices for the firmware images:
CODE=${WORKSPACE}/Build/OvmfX64/DEBUG_GCC5/FV/OVMF_CODE.fd
VARS=${WORKSPACE}/Build/OvmfX64/DEBUG_GCC5/FV/OVMF_VARS.fd
qemu-system-x86_64 \
-blockdev node-name=code,driver=file,filename=${CODE},read-only=on \
-blockdev node-name=vars,driver=file,filename=${VARS},snapshot=on \
-machine q35,pflash0=code,pflash1=vars \
[ ... ]
Here is the arm version of that (using the padded files created
using dd
, see above):
CODE=${WORKSPACE}/Build/ArmVirtQemu-AARCH64/DEBUG_GCC5/FV/QEMU_EFI-pflash.raw
VARS=${WORKSPACE}/Build/ArmVirtQemu-AARCH64/DEBUG_GCC5/FV/QEMU_VARS-pflash.raw
qemu-system-aarch64 \
-blockdev node-name=code,driver=file,filename=${CODE},read-only=on \
-blockdev node-name=vars,driver=file,filename=${VARS},snapshot=on \
-machine virt,pflash0=code,pflash1=vars \
[ ... ]
The core edk2 repo holds a number of packages, each package has its own toplevel directory. Here are the most interesting ones:
The firmware modules in the edk2 repo often named after the boot
phase they are running in. Most drivers are
named SomeThingDxe
for example.
This week, Red Hat Enterprise Linux 9 has been announced, which will also bring us lots of new stuff for our beloved mainframe.
First, compared with RHEL 8, a lot of generic packages have been updated, of course. For example, RHEL 9 on IBM Z comes with:
And of course all of these have been thoroughly tested during the past months, which is also the reason why RHEL sometimes does not ship the very latest bleeding edge versions of the upstream projects – thorough testing needs some time. But you can be sure that Red Hat also backported lots of selected upstream fixes and improvements e.g. for the kernel to their downstream packages, so this is very up to date and stable software here.
The first big news is: There is no need anymore to install the separate virt:av (“Advanced Virtualization”) module to get the latest and greatest virtualization features on IBM Z. Everything is packaged along with the main RHEL distribution for easier installation now and will be kept up-to-date there, with important new features like virtio-fs enabled by default. And of course, as with the latest releases of RHEL 8, there is also no limit to 4 guests anymore, so you don’t have to worry about the number of supported KVM guests (as long as your hardware can handle them).
The versions that will be shipped with RHEL 9.0 are:
To answer the maybe most important question: Yes, this will also support the brand new IBM z16 mainframe already. Basic support for this new generation has already been added to QEMU 6.1.0 and kernel 5.14, and additional z16 features have been enabled by default in QEMU 6.2.0.
Another great new change is that it is now possible to
configure mediated devices directly with the virtualization CLI tools on IBM Z.
You can now add vfio-ap and vfio-ccw mediated devices to your KVM guests
using virt-install
or virt-xml
. With virt-install
, you can
also create a VM that uses an existing DASD mediated device as its primary disk.
Additionally, many small performance improvements (like the specification exception interpretation feature) and bug fixes have been backported to the RHEL 9 kernel and the userspace tools to give you a great virtualization experience with RHEL 9.
One more thing that is worth mentioning (though it is not specific to IBM Z), which you might have noticed by clicking on the links in the previous paragraphs already, there is another big change in RHEL 9: The development of the upcoming minor RHEL 9 releases (i.e. 9.1, 9.2, etc.) is now done in the public via the CentOS Stream repositories. That means you can not only peak on the work that will be integrated in the next 9.y release, you can now even directly participate in the development of these next release if you like! Isn’t that cool?
Anyway, no matter whether you are planning to participate or just want to use the software, please enjoy the new KVM virtualization stack on the mainframe!
Flatpak is a way to distribute applications on Linux. Its container-style approach allows applications to run across Linux distributions. This means native packages (rpm, deb, etc) are not needed and it's relatively easy to get your app to Linux users with fewer worries about distro compatibility. This makes life a lot easier for developers and is also convenient for users.
I've run popular applications like OBS Studio as flatpaks and even publish my own on Flathub, a popular hosting site for applications. Today I figured out how to debug flatpaks, which requires some extra steps that I'll share below so I don't forget them myself!
If you're building a flatpak of your own application it's handy to use the dir sources type in the manifest to compile your application's source code from a local directory instead of a git tag or tarball URL. This way you can make changes to the source code and test them quickly inside Flatpak.
Put something along these lines in the manifest's modules object where /home/user/my-app is you local directory with your app's source code:
{
"name": "my-app",
"sources": [
{
"type": "dir",
"path": "/home/user/my-app"
}
],
...
}
flatpak-builder(1) automatically creates a separate .Debug extension for your flatpak that contains your application's debuginfo. You'll need the .Debug extension if you want proper backtraces and source level debugging. At the time of writing the Flatpak documentation did not mention how to install the locally-built .Debug extension. Here is how:
$ flatpak-builder --user --force-clean --install build my.org.app.json
$ flatpak install --user --reinstall --assumeyes "$(pwd)/.flatpak-builder/cache" my.org.app.Debug
It might be a good idea to install debuginfo for the system libraries in your SDK too in case it's not already installed:
$ flatpak install org.kde.Sdk.Debug # or your runtime's SDK
There is a flatpak(1) option that launches the application with the SDK instead of the Runtime:
$ flatpak run --user --devel my.org.app
The SDK contains development tools whereas the Runtime just has the files needed to run applications.
It can also be handy to launch a shell so you can control the launch of your app and maybe use gdb or strace:
$ flatpak run --user --devel --command=sh my.org.app
[📦 my.org.app ~]$ gdb /app/bin/my-app
If your application crashes it will dump core like any other process. However, existing ways of inspecting core dumps like coredumpctl(1) are not fully functional because the process ran inside namespaces and debuginfo is located inside flatpaks instead of the usual system-wide /usr/lib/debug location. coredumpctl(1), gdb, etc aren't Flatpak-aware and need extra help.
Use the flatpak-coredumpctl wrapper to launch gdb:
$ flatpak-coredumpctl -m <PID> my.org.app
You can get PID from the list printed by coredumpctl(1).
This article showed how to install locally-built .Debug extensions and inspect core dumps when using Flatpak. I hope that over time these manual steps will become unnecessary as flatpak-builder(1) and coredumpctl(1) are extended to automatically install .Debug extensions and handle Flatpak core dumps. For now it just takes a few extra commands compared to debugging regular applications.
Canonical release a new LTS (Long Term Support) version of its Ubuntu server offering Ubuntu Server 22.04!
It ships
Ubuntu Server 21.10 is out!
It ships
TL;DR
We’d like to announce the availability of the QEMU 7.0.0 release. This release contains 2500+ commits from 225 authors.
You can grab the tarball from our download page. The full list of changes are available in the Wiki.
Highlights include:
Thank you to everyone involved!
Today, IBM announced the new IBM z16, with a planned availability date of May 31.
See here for the press release, and here for the offical homepag. For further details, including a list of supported Linux distributions, see Eberhard's blog here.
And for a more hands-on tour of the new box, check out this video.
The solution assurance team started to publish solution setups, recommendations, and step-by-step guidelines for a broad range of topics, for example:
You can access the materials here, but don't forget to check back periodically: There is more to come!