Our current reference leak checker uses finalizers to verify whether an object
has reached zero references before it is garbage collected. There are multiple
problems with this mechanism, so a rewrite is in order.
With finalizers, there is no way to guarantee that a finalizer will run before
the program exits. When an unreachable object with a finalizer is garbage
collected, its finalizer will be added to a queue and run asynchronously. The
best we can do is run garbage collection upon sandbox exit to make sure that
all finalizers are enqueued.
Furthermore, if there is a chain of finalized objects, e.g. A points to B
points to C, garbage collection needs to run multiple times before all of the
finalizers are enqueued. The first GC run will register the finalizer for A but
not free it. It takes another GC run to free A, at which point B's finalizer
can be registered. As a result, we need to run GC as many times as the length
of the longest such chain to have a somewhat reliable leak checker.
Finally, a cyclical chain of structs pointing to one another will never be
garbage collected if a finalizer is set. This is a well-known issue with Go
finalizers (https://github.com/golang/go/issues/7358). Using leak checking on
filesystem objects that produce cycles will not work and even result in memory
leaks.
The new leak checker stores reference counted objects in a global map when
leak check is enabled and removes them once they are destroyed. At sandbox
exit, any remaining objects in the map are considered as leaked. This provides
a deterministic way of detecting leaks without relying on the complexities of
finalizers and garbage collection.
This approach has several benefits over the former, including:
- Always detects leaks of objects that should be destroyed very close to
sandbox exit. The old checker very rarely detected these leaks, because it
relied on garbage collection to be run in a short window of time.
- Panics if we forgot to enable leak check on a ref-counted object (we will try
to remove it from the map when it is destroyed, but it will never have been
added).
- Can store extra logging information in the map values without adding to the
size of the ref count struct itself. With the size of just an int64, the ref
count object remains compact, meaning frequent operations like IncRef/DecRef
are more cache-efficient.
- Can aggregate leak results in a single report after the sandbox exits.
Instead of having warnings littered in the log, which were
non-deterministically triggered by garbage collection, we can print all
warning messages at once. Note that this could also be a limitation--the
sandbox must exit properly for leaks to be detected.
Some basic benchmarking indicates that this change does not significantly
affect performance when leak checking is enabled, which is understandable
since registering/unregistering is only done once for each filesystem object.
Updates #1486.
PiperOrigin-RevId: 338685972
Currently strace+debug is always enabled as the setting from
the upper layer isn't passed to _syscall_test(). And it will
negatively affect the performance tests. This patch fixes this
issue.
The "debug" argument of _syscall_test() is also made mandatory
to prevent this happening again.
//test/perf:getpid_benchmark_runsc_kvm
-----------------------------------------------------
Benchmark Time CPU Iterations
-----------------------------------------------------
Before:
BM_Getpid 28119 ns 28157 ns 25926
After:
BM_Getpid 947 ns 939 ns 777778
Fixes#4509
Signed-off-by: Tiwei Bie <tiwei.btw@antgroup.com>
Some syscall tests, namely uname_test_* modify the host and domain
name, which modifies the execution environment and can have unintended
consequences on other tests. For example, modifying the hostname
causes some networking tests to fail DNS lookups. Run all syscall
tests in their own uts namespaces to isolate these changes.
PiperOrigin-RevId: 329348127
This is done to ease troubleshooting when tests fail. runsc
logs are not stored when tests passe, so this will only
affect failing tests and should not increase log storage
too badly.
PiperOrigin-RevId: 327717551
The code was deleting logs for all tests when a single test
passed. Change it to delete only the logs relevant to the
test at hand.
Also fixed the benchmark lookup code, which was always generating
a single empty benchmark entry if there were not benchmarks.
PiperOrigin-RevId: 326311477
This commit adds an integration test framework for FUSE support. Please
refer to the test example and test/fuse/README.md for further details.
Fixes#3098
... when it is possible.
The guitar gVisorKernel*Workflow-s runs test with the local execution_method.
In this case, blaze runs test cases locally without sandboxes. This means
that all tests run in the same network namespace. We have a few tests which
use hard-coded network ports and they can fail if one of these port will be
used by someone else or by another test cases.
PiperOrigin-RevId: 323137254
This change gates all FUSE commands (by gating /dev/fuse) behind a runsc
flag. In order to use FUSE commands, use the --fuse flag with the --vfs2
flag. Check if FUSE is enabled by running dmesg in the sandbox.
This change adds a layer of abstraction around the internal Docker APIs,
and eliminates all direct dependencies on Dockerfiles in the infrastructure.
A subsequent change will automated the generation of local images (with
efficient caching). Note that this change drops the use of bazel container
rules, as that experiment does not seem to be viable.
PiperOrigin-RevId: 308095430
The root mount is not shared by default, but all other mounts are shared.
So if we create the /tmp mount, this means that we run tests on a shared mount
even if tests run without the --shared option.
PiperOrigin-RevId: 302130790
The benchmark_filter options accepts regex-s, but
the gtest-filter option accepts shell-like wildcards.
Fixes#2034
Signed-off-by: Andrei Vagin <avagin@gmail.com>
Andrei Vagin
Dean Deng
Tiwei Bie
Fabricio Voznika
Rahat Mahmood
gVisor bot
Craig Chi
Ayush Ranjan
Adin Scannell
Kevin Krakauer
Ridwan Sharif
Nicolas Lacasse
Michael Pratt
Gaurav Singh
Jamie Liu
Andrei Vagin