For the callers of get_dirty_log, they must pass in bitmap large enough
to hold the log. Instead of duplicating that calculation inside the
method and the caller, this change provides a public helper usable by
both to calculate the bitmap's minimum size.
TEST=./build_test
BUG=chromium:800626
Change-Id: I349aa0dd7e78e7f9ea3d35ba0acba8088fd77a4e
Reviewed-on: https://chromium-review.googlesource.com/857909
Commit-Ready: ChromeOS CL Exonerator Bot <chromiumos-cl-exonerator@appspot.gserviceaccount.com>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This is needed to support the plugin API, which might unregister such
eventfds.
BUG=chromium:800626
TEST=./build_test
Change-Id: Ic4e74ffd4b2374868644d1f6598509aba741ce00
Reviewed-on: https://chromium-review.googlesource.com/857908
Commit-Ready: ChromeOS CL Exonerator Bot <chromiumos-cl-exonerator@appspot.gserviceaccount.com>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This will be very useful in tests that need a file as it removes the
need for a temporary file to be created.
Change-Id: I7fb08209174c870279e34ee07aad7a3b05baaad9
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/864625
Reviewed-by: Zach Reizner <zachr@chromium.org>
Allow block to take anything that satisfies the traits it uses. This
will allow for different backends to be used in addition to normal
files. The first new backend will be qcow2 added in subsequent commits.
Change-Id: I571ffc8ad4b5d56f286e134f47a60261d639b160
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/862627
Reviewed-by: Zach Reizner <zachr@chromium.org>
Add a top level fuzz directory. Other fuzz tests will be added here in
subsequent commits.
For now fuzzing must be run manually. Soon there will be a way to
extract the fuzz artifacts and upload them to cluster fuzz.
Change-Id: Iddfb55af78af6f412927b2221f22acb882069d36
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/850851
Reviewed-by: Zach Reizner <zachr@chromium.org>
There were a few places that used this to get the page size inside of an
unsafe block, For convenience, this adds a safe wrapper in sys_util and
replaces all extant usage of sysconf with the wrapper version.
BUG=chromium:800626
TEST=./build_test
Change-Id: Ic65bf72aea90eabd4158fbdcdbe25c3f13ca93ac
Reviewed-on: https://chromium-review.googlesource.com/857907
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This is used to make EventFd passed over unix domain socket, such as by
the plugin API.
TEST=./build_test
BUG=chromium:800626
Change-Id: Ifd6c81c51b31a376d57a007bf413a836f3af870a
Reviewed-on: https://chromium-review.googlesource.com/857906
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
TSYNC isn't particularly useful for the device jails since they start
with just a single thread. But a useful side effect of having minijail
use TSYNC is that instead of the default SECCOMP_RET_KILL_THREAD behavior,
minijail switches to SECCOMP_RET_TRAP and uses the default signal disposition
which dumps core.
Until SECCOMP_RET_KILL_PROCESS is available on all kernel versions with crosvm,
using TSYNC this way allows killing the entire device process instead of just
one thread. This ensures if seccomp kills a worker thread in a device, the
entire device process will die, and the crosvm main process will exit.
BUG=chromium:799523
TEST=add banned syscall to net device worker thread and ensure crosvm exits
Change-Id: Ie9ebfc90c79dcf49283cb2628dc8d4c848e8385b
Reviewed-on: https://chromium-review.googlesource.com/853302
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
The memfd_create flags changed with "add memfd seal support", which
breaks the seccomp filter for the wayland device.
TEST=run a wayland app in crosvm
BUG=chromium:799523
Change-Id: I266e305bc3179a7b9fd105a684de92e944fe1fc6
Reviewed-on: https://chromium-review.googlesource.com/852937
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
The jail for the wayland device used chown to ensure that its jail had
the proper permissions for the wayland socket to be bind mounted into
it. This creates some unnecessary complexity because it requires
careful management of the user and group and crosvm runs as (a non-root
user cannot change the owner of a directory) or that crosvm has the
CAP_CHOWN capability.
Instead of trying to make the permissions fit, just have the jail mount
a small tmpfs over the jail's root directory. This is one of the things
that a process inside a user namespace has the ability to do. Bind
mounting the wayland socket into this tmpfs then just works without any
other issues.
BUG=chromium:799523
TEST=linux vm boots with no errors
Change-Id: Ic2240f430c7fd332a15b4fcd4e52374799eb6c9d
Signed-off-by: Chirantan Ekbote <chirantan@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/851413
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Add support for minijail_mount and minijail_mount_with_data. This will
be used by the jail for the wayland device.
BUG=none
TEST=filesystem is mounted inside the jail
Change-Id: I6ad9933d057e7642a7551a6a316ff65d3b95a9dd
Signed-off-by: Chirantan Ekbote <chirantan@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/851412
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This makes process cleanup difficult because minijail calls
setsid(), and that removes the devices from the main process's
process group.
BUG=chromium:799523
TEST=stop crosvm and ensure there are no zombies hanging around
Change-Id: I14c54cf250bdc7339970c886cdab9ff2f4b8a135
Reviewed-on: https://chromium-review.googlesource.com/852987
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
Use ok_or to remove the match witch was harder to read. No functional
change. Pull the check in to a function so that it only needs to be
right once.
Change-Id: I0e5801d9e7e82994e7f1fbda0d2692a4afd59d99
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/850850
Reviewed-by: Zach Reizner <zachr@chromium.org>
The other functions in mmap check that the offset doesn't overflow. The
kernel_loader fuzzer found that read forgot to check.
Change-Id: Ifc89dbe40345c5923a5cf29c9f29e810e9e1a1e8
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/850542
Reviewed-by: Zach Reizner <zachr@chromium.org>
The syslog subsystem tries to figure out the file descriptor for the
connection to the system logger so that it can ensure that it doesn't
get closed in each device process.
However, the check does not work properly if there was already an open
connection to the system logger. In this case the openlog call does not
do anything and we end up guessing the wrong file descriptor number for
the syslog connection.
Work around this by adding a closelog() call before attempting all of
this cleverness. In the long run this should be fixed properly by just
bind mounting /dev/log into each device process's jail.
BUG=none
TEST=Running crosvm under minijail0 does not cause an InvalidFd error.
Change-Id: Iffd535d62acdf8053817af74b9e97444c746a0cf
Signed-off-by: Chirantan Ekbote <chirantan@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/851271
Reviewed-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
Getting and settings seals is useful to ensure the size of files
underlying memory mappings doesn't shrink, which can trigger a SIGBUS on
access to the truncated pages.
This also bumps the libc version to get MFD_ALLOW_SEALING.
TEST=cargo test
BUG=None
CQ-DEPEND=CL:850535
Change-Id: Ifbe1ec2c47d3d5c51b63472f545acc10d3c8eed2
Reviewed-on: https://chromium-review.googlesource.com/849488
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This is needed to support the plugin process API, which may register
guest memory mapped at an offset from the beginning of a file.
TEST=cargo test
BUG=None
Change-Id: Idf1e9f0287df5510728ab2bcf4dd090f9e81a5bf
Reviewed-on: https://chromium-review.googlesource.com/849495
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
The old sequence used tail cails to foward resources to the next stage
of the process, making it cumbersome to add new resources to earlier
stages. Certain tasks, such as IO port bus creation or irqfd
registration, were delayed to late stages for no good reason, causing
the stage names to have no correspondence with their contents.
The new sequence organizes setup into logical functions that only receive
and return resources that make sense for their operations.
BUG=None
TEST=./build_test; cargo run
Change-Id: Idaf1488da515e2473125333bf997638da60fa992
Reviewed-on: https://chromium-review.googlesource.com/822927
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Zach Reizner <zachr@chromium.org>
The old DeviceManager made changes to VM indirectly using VM control
messages. This change gives the DeviceManager a VM reference to make
changes directly. This simplifies things by removing the need for an
extra loop to pending VM control messages.
BUG=None
TEST=./build_test
Change-Id: I6866015ddf3216a1f03d6ac0431548e3b827b750
Reviewed-on: https://chromium-review.googlesource.com/823263
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This is needed for plugin process style executions because they don't
register guest memory though the GuestMemory interface. Their memory is
more dynamic and ill-suited to the GuestMemory abstraction.
TEST=./build_test
BUG=None
Change-Id: I250b0b874ef6a0e342fd6e659301f389d01a4cda
Reviewed-on: https://chromium-review.googlesource.com/848018
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Re-generated the bindings for x86_64 and arm which now use union types
so a little bit of code in the library also changed, and adds bindings
for aarch64 which are required to run an aarch64 guest.
Also, I manually fixed the zero-length array cases where rust bindgen
doesn't properly align the structs. See rust bindgen bug 684.
BUG=chromium:797868
TEST=run crosvm on x86_64, ensure networking works
TEST=./build_test passes on x86
Change-Id: Iab2193a8f1d4c68bd3346ae683c74bbf16fe45d4
Signed-off-by: Sonny Rao <sonnyrao@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/845519
Reviewed-by: Zach Reizner <zachr@chromium.org>
The balloon device is used to take regions of unused memory from the
guest and allow other host processes to use that memory.
Change-Id: I06c821365a58672d605fc7555beaec599cae1b15
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/759306
Reviewed-by: Stephen Barber <smbarber@chromium.org>
This is to clearly separate the Linux VM setup code from main's argument
parsing logic. This will also make it easier to add other kinds of VM
setup code without littering main or the Linux setup sequence.
BUG=None
TEST=./build_test; cargo run
Change-Id: I8e28720df760357cf00c9711ca5f1f8d6c077844
Reviewed-on: https://chromium-review.googlesource.com/823610
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
/var/empty always exists on cros devices and is commonly used by recent
minijail configs throughout the system. Using it here saves several
variables and removes some error paths.
Leave the wayland root as it needs to be owned by the wayland group.
Change-Id: I261915f4419cadb3f121e9c423c79e467b014700
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/816536
Reviewed-by: Zach Reizner <zachr@chromium.org>
Add functions to let the kernel know a given range of memory isn't
needed currently. This function will result in `madvise(DONTNEED)`.
The ability to signal memory as not needed will be used by the balloon
driver to allow system memory to be redistributed away from the VM.
Change-Id: I4ca56e09010eec33989824f5738db4a4be0ec428
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/759305
Reviewed-by: Zach Reizner <zachr@chromium.org>
Recent Linux kernel's fail to start if the mptable is at the start of
RAM (address 0x00). Avoid putting the mptable there so that crosvm can
boot 4.14+ kernels. The kernel scans the last kilobyte of RAM after the
first, move the mptable there.
Change-Id: Ia00f49e7a4cbd0fb3719c21b757e8fdca65584e8
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/780045
Making the project a workspace allows running all local libs unit tests
from the main build.
Change-Id: Ib531dd10c21f324c7d25af69ad12b95c7bc8f3bc
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/759597
Reviewed-by: Zach Reizner <zachr@chromium.org>
The table grew with the addition of the 16 mpc_intsrc structures.
Correct the `compute_mp_size` function, the end check, and add a unit
test for the not having enough memory.
Change-Id: I1ff268629a47a422f50aefef9d6aa95121b94d59
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/777710
Reviewed-by: Slava Malyugin <slavamn@google.com>
Reviewed-by: Zach Reizner <zachr@chromium.org>
There were remnants of using `minijail_enter`, switch them to
`minijail_fork`.
Change-Id: I5c695cbc3bfd6868f25f46f785a2f96391744d90
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/759598
Reviewed-by: Stephen Barber <smbarber@chromium.org>
On older kernels, the memfd_create syscall isn't available. Skip shm
tests if that is the case.
Change-Id: I39c1f1779f1f02e90df727c6ca18b5bdae52e855
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/768102
Reviewed-by: Stephen Barber <smbarber@chromium.org>
The mptable inherited from kvmtool had some missing pieces. On top of that,
crosvm does not use KVM_SET_GSI_ROUTING. The addresses makes mptable match
the default routing in host kernel and removes "noapic".
TEST=cargo build (--release). tatl boot tested on 4.4.0 and 4.4.9
Change-Id: Ibc55abf245cd9d8fca601da204d5a189321c09c7
Reviewed-on: https://chromium-review.googlesource.com/772820
Commit-Ready: Slava Malyugin <slavamn@google.com>
Tested-by: Slava Malyugin <slavamn@google.com>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
This change is to support a related kernel change but is backwards
compatible with kernels without that change.
BUG=chromium:782474
TEST=None
Change-Id: Ic1224b65ed9685f246002f946cfc6bfa2dbb2856
Reviewed-on: https://chromium-review.googlesource.com/770593
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Suspend/resume can cause syscall restarts and will cause KVM_RUN ioctls
to return with EINTR. Handle these so the VM doesn't shut down.
BUG=none
TEST=vm survives suspend/resume
Change-Id: I1fab624cb8fe0949d341408f0c962c859a034205
Reviewed-on: https://chromium-review.googlesource.com/750054
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
Reviewed-by: Zach Reizner <zachr@chromium.org>
This also updates the `build_test.py` to use the command line option to
run certain certain test modules serially.
TEST=./build_test
BUG=None
Change-Id: I8a498514cb6b89fab01f02d0ef8faf39629f717c
Reviewed-on: https://chromium-review.googlesource.com/748824
Commit-Ready: Zach Reizner <zachr@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
If reading from stdin returns EOF or an error, remove it from the list
of pollables.
BUG=none
TEST=`vm_launcher start` and check that crosvm no longer pegs CPU
Change-Id: I7971058701e6145884de9c52a8dd5b829373637b
Reviewed-on: https://chromium-review.googlesource.com/745961
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Zach Reizner <zachr@chromium.org>
crosvm spawns a lot of processes/threads, and having these all use the same
name as the original process can be confusing. So at least in the instances
where Rust threads are spawned (vs. minijail_fork()), use a thread::Builder
to allow setting the thread name.
BUG=none
TEST=start crosvm, check thread names with top
Change-Id: I6e55ff5fd60f258880bda8e656ab7f9da82c656e
Reviewed-on: https://chromium-review.googlesource.com/742394
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
There is a low limit on the maximum memory slot number imposed by the
kernel. On x86_64, that limit is 509. In order to delay hitting that
limit, we attempt to use the lowest unused slot number. As memory
regions are removed from the VM, the slot for that region is stored in a
heap so that that slot number can quickly be reused next time a memory
region is added.
BUG=None
TEST=finish a game of gnome-mahjong using virtio-wayland
Change-Id: I786c2e2b8ff239c19b3c8a18bd0f6e8f8dc2acbf
Reviewed-on: https://chromium-review.googlesource.com/740102
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Tested-by: Zach Reizner <zachr@chromium.org>
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
Add a wayland-sock argument for specifying the path to the
wayland socket. Also add a wayland-group argument for specifying
the group that has access to that socket.
BUG=none
TEST=crosvm starts with wayland
Change-Id: I3e53eb697951200340613663dc3e4ed28d8ed7f8
Reviewed-on: https://chromium-review.googlesource.com/733732
Commit-Ready: Stephen Barber <smbarber@chromium.org>
Tested-by: Stephen Barber <smbarber@chromium.org>
Reviewed-by: Dylan Reid <dgreid@chromium.org>
Using minijail_fork removes the need to manage user and pid namespace
explicitly in crosvm and removes some parent/child synchonization
requirements too.
Change-Id: I47f9d39527d0a3ccf625600e9bfc2cccc3cb27ca
Signed-off-by: Dylan Reid <dgreid@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/719443
Reviewed-by: Stephen Barber <smbarber@chromium.org>
Since /var/run is a symlink to /run, and we want to avoid going
through the stateful /var whenever possible, use the direct path.
BUG=chromium:699880
TEST=precq passes
Change-Id: I5d95f5358c1fb0cb2ca73f9c4b145d8e36b4a361
Reviewed-on: https://chromium-review.googlesource.com/734225
Commit-Ready: Mike Frysinger <vapier@chromium.org>
Tested-by: Mike Frysinger <vapier@chromium.org>
Reviewed-by: Zach Reizner <zachr@chromium.org>
