Rather than always binding to the vfio-pci driver, use the new
function virPCIDeviceFindBestVFIOVariant() to see if the running
kernel has a VFIO variant driver available that is a better match for
the device, and if one is found, use that instead.
virPCIDeviceFindBestVFIOVariant() function reads the modalias file for
the given device from sysfs, then looks through
/lib/modules/${kernel_release}/modules.alias for the vfio_pci alias
that matches with the least number of wildcard ('*') fields.
The appropriate "VFIO variant" driver for a device will be the PCI
driver implemented by the discovered module - these drivers are
compatible with (and provide the entire API of) the standard vfio-pci
driver, but have additional device-specific APIs that can be useful
for, e.g., saving/restoring state for migration.
If a specific driver is named (using <driver model='blah'/> in the
device XML), that will still be used rather than searching
modules.alias; this makes it possible to force binding of vfio-pci if
there is an issue with the auto-selected variant driver.
Signed-off-by: Laine Stump <laine@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
For statically declared arrays one can use G_N_ELEMENTS() instead
of explicit sizeof(array) / sizeof(item). I've noticed couple of
places where the latter was used.
I am not fixing every occurrence because we have some places
which do not use glib (examples and NSS module).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Tim Wiederhake <twiederh@redhat.com>
In a few places we declare a variable (which is optionally
followed by a code not touching it) then set the variable to a
value and return the variable immediately. It's obvious that the
variable is needless and the value can be returned directly
instead.
This patch was generated using this semantic patch:
@@
type T;
identifier ret;
expression E;
@@
- T ret;
... when != ret
when strict
- ret = E;
- return ret;
+ return E;
After that I fixed couple of formatting issues because coccinelle
formatted some lines differently than our coding style.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Add helper functions to virpci to provide means of checking for a VPD
file presence and for VPD resource retrieval using the PCI VPD parser.
The added test assesses the basic functionality of VPD retrieval while
the full parser is tested by virpcivpdtest.
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Dmitrii Shcherbakov <dmitrii.shcherbakov@canonical.com>
VIR_APPEND_ELEMENT doesn't report any errors now so we can remove
VIR_APPEND_ELEMENT_QUIET and replace all uses by VIR_APPEND_ELEMENT
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
For now it was an alias to VIR_APPEND_ELEMENT. Use virAppendElement
directly until VIR_APPEND_ELEMENT is refactored too and we'll be able to
get rid of VIR_APPEND_ELEMENT_QUIET completely.
Signed-off-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Generated using the following spatch:
@@
expression path;
@@
- virFileMakePath(path)
+ g_mkdir_with_parents(path, 0777)
However, 14 occurrences were not replaced, e.g. in
virHostdevManagerNew(). I don't really understand why.
Fixed by hand afterwards.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
The version of macOS running on Apple Silicon doesn't need to
concern itself with backwards compatibility with 32-bit
applications, and so it could jettison all the symbol aliasing
shenanigans involved.
https://gitlab.com/libvirt/libvirt/-/issues/121
Signed-off-by: Andrea Bolognani <abologna@redhat.com>
Reviewed-by: Roman Bolshakov <r.bolshakov@yadro.com>
Tested-by: Roman Bolshakov <r.bolshakov@yadro.com>
This was found by clang-tidy's
"clang-analyzer-security.insecureAPI.bzero" check.
bzero is marked as deprecated ("LEGACY") in POSIX.1-2001 and
removed in POSIX.1-2008.
Besides its deprecation, bzero can be unsafe to use under certain
circumstances, e.g. when used to zero-out memory containing secrects.
These calls can be optimized away by the compiler, if it concludes no
further access happens to the memory, thus leaving the secrets still
in memory. Hence its classification as "insecureAPI".
Signed-off-by: Tim Wiederhake <twiederh@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Some tests in qemuxml2argvtest need opendir() from virpcimock, others
need opendir() from virfilewrapper.
But as of now, only opendir() from virpcimock has an effect.
real_opendir in virpcimock has a pointer to opendir$INODE64 in
libsystem_kernel.dylib instead of pointing to opendir$INODE64 in
qemuxml2argvtest (from virfilewrapper). And because the second one is
never used, tests that rely on prefixes added by virFileWrapperAddPrefix
fail.
That can be fixed if dlsym(3) is asked explicitly to search symbols in
main executable with RTLD_MAIN_ONLY before going to other dylibs.
Existing RTLD_NEXT handle results into libsystem_kernel.dylib being
searched before main executable.
Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
opendir() mocks need to search for decorated function with $INODE64
suffix, like stat mocks.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
In general, it has little sense to use Linux pci mock on macOS but
virPCIDeviceAddressGetIOMMUGroupNum() is relying on the filesystem
layout mocked by virpcimock. And all tests that rely on correct
execution of virPCIDeviceAddressGetIOMMUGroupNum() fail.
The change fixes qemuhotplugtest, qemumemlocktest and qemuxml2xmltest.
Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
real_close() is not inialized by the first invocation of close(). That
causes an issue when the mock is used before others and a call of
real_close() results in a jump to NULL pointer.
Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Andrea Bolognani <abologna@redhat.com>
Various places reported by cppcheck's invalidPrintfArgType_sint
and invalidPrintfArgType_uint.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
The last_component() method is a GNULIB custom function
that returns a pointer to the base name in the path.
This is similar to g_path_get_basename() but without the
malloc. The extra malloc is no trouble for libvirt's
needs so we can use g_path_get_basename().
Reviewed-by: Fabiano Fidêncio <fidencio@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
The device configs (which are actually the same one config)
come from a NVMe disk of mine.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
ACKed-by: Peter Krempa <pkrempa@redhat.com>
Reviewed-by: Cole Robinson <crobinso@redhat.com>
Glib implementation follows the ISO C99 standard so it's safe to replace
the gnulib implementation.
Signed-off-by: Pavel Hrdina <phrdina@redhat.com>
Reviewed-by: Peter Krempa <pkrempa@redhat.com>
Since we are not passing the @fakerootdir variable to any inline function
anymore, we can make the variable static.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Replace all occurrences of
if (VIR_STRDUP(a, b) < 0)
/* effectively dead code */
with:
a = g_strdup(b);
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Since commit 44e7f02915
util: rewrite auto cleanup macros to use glib's equivalent
VIR_AUTOFREE is just an alias for g_autofree. Use the GLib macros
directly instead of our custom aliases.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Use G_GNUC_UNUSED from GLib instead of ATTRIBUTE_UNUSED.
Signed-off-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Previous patch had to add '/sys/kernel/' prefix in opendir() because
the path, which is being mocked, wasn't being considered due to
an 'if SYSFS_PCI_PREFIX' guarding the call to getrealpath().
In fact, all current getrealpath() callers are guarding it with a
conditional to ensure that the function will never be called with
a non-mocked path. In this case, an extra non-NULL verification is
needed for the 'newpath' string to use the variable - which is
counterintuitive, given that getrealpath() will always write the
'newpath' string in any non-error conditon.
However, simply removing the guard of all getrealpath() instances
causes an abort in init_env(). This happens because tests will
execute access() to non-mocked paths even before the
LIBVIRT_FAKE_ROOT_DIR variable is declared in the test files. We
don't need 'fakerootdir' to be created at this point though.
This patch does the following changes to simplify getrealpath()
usage:
- getrealpath() will now guard the init_env() call by checking if
both fakeroot isn't created and the required path is being mocked.
This ensures that we're not failing inside init_env() because
we're too early and LIBVIRT_FAKE_ROOT_DIR wasn't defined yet;
- remove all conditional guards to call getrealpath() from
access(), virMockStatRedirect(), open(), open_2(), opendir()
and virFileCanonicalizePath(). As a bonus, remove all ternary
conditionals with 'newpath';
- a new 'pathPrefixIsMocked()' helper to aggregate all the prefixes
we're mocking, making it easier to add/remove them. If a prefix
is added inside this function, we can be sure that all functions
are mocking them.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This patch adds hostdev test cases in qemuhotplugtest.c.
Note: the small tweak inside virpcimock.c was needed because
the new tests added a code path in which virHostHasIOMMU()
(virutil.c) started being called, and the mocked '/sys/kernel/'
prefix that is mocked in virpcimock.c wasn't being considered
in the opendir() mock. An alternative to avoid these situations
in virpcimock.c is implemented in the next patch.
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The softlink to physfn is the way to know if the device is
VF or not. So, the patch softlinks 'physfn' to the parent function.
The multifunction PCI devices dont have 'physfn' softlinks.
The patch adds few Virtual functions to the mock environment and
changes the existing VFIO test xmls using the VFs to use the newly
added VFs for their use case.
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.vnet.ibm.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This patch adds mock of the /dev/vfio path, needed for proper
implementation of the support for multifunction/multiple devices
per iommu groups.
To do that, the existing bind and unbind operations were adapted
to operate with the mocked filesystem as well.
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Michal Privoznik <mprivozn@redhat.com>
This enum was introduced to model how RHEL-7 kernel behaves - for
some reason going with the old way (via new_id + bind) fails but
using driver_override succeeds. Well, we don't need to care about
that anymore since we don't create new_id file.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Now that PCI attach/detach happens solely via driver_override
these two files are no longer needed.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Now that nothing supports "pci-stub" driver (aka KVM style of PCI
device assignment) there is no need for virpcimock to create it.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Reviewed-by: Ján Tomko <jtomko@redhat.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Mocking of the __open_2 function was added in
commit 459f071cac
Author: Michal Privoznik <mprivozn@redhat.com>
Date: Thu Aug 15 16:37:17 2019 +0200
virpcimock: Mock __open_2()
This function only exists in glibc, however, and the mocking code runs
on systems not using glibc, such as FreeBSD. Even Linux hosts might be
using a different libc impl, though we don't actively try to support
that.
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
So far, we don't need to create anything under
/sys/kernel/iommu_groups/N/devices directory (which is symlinked
from /sys/bus/pci/devices/DDDD:BB:DD.F/iommu_group directory)
because virhostdevtest still tests the old KVM assignment and
thus has no notion of IOMMU groups. This will change in near
future though. And in order to discover devices belonging to the
same IOMMU group we need to do what kernel does - create symlinks
to devices.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
So far, we are creating devices directly under
/sys/bus/pci/devices/*. There is not much problem with it, but if
we really want to model kernel behaviour we need to create them
under /sys/devices/pciDDDD:BB and then only symlink them from the
old location.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
In upcoming patches we will need only some portions of the PCI
address. To construct that easily, it's better if the PCI address
of a device is stored as four integers rather than one string.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Have just one function to generate path to a PCI driver so that
when we change it in near future there's only few of the places
we need to fix.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Have just one function to generate path to a PCI device so that
when we change it in near future there's only few of the places
we need to fix.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
In near future, we will be creating devices under different
location and just symlink them under devices/. Just like real
kernel does. But for that we need the directories to exist.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
We will need to create more directories and instead of
introducing bunch of new variables to hold their actual
paths, we can have one and reuse it.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
The @fakesysfspcidir is derived from @fakerootdir. We don't need
two global variables that contain nearly the same content,
especially when we construct the actual path anyways.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
It saves us couple of lines.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
When creating a PCI device, the pciDevice structure contains @id
member which holds device address (DDDD.BB:DD.F) and is type of
'char *'. But the structure is initialized from a const char and
in fact we never modify or free the @id.
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Newer kernels (v3.16-rc1~29^2~6^4) have 'driver_override' file
which simplifies way of binding a PCI device to desired driver.
Libvirt has support for this for some time too (v2.3.0-rc1~236),
but not our virpcimock. So far we did not care because our code
is designed to deal with this situation. Except for one.
hypothetical case: binding a device to the vfio-pci driver can be
successful only via driver_override. Any attempt to bind a PCI
device to vfio-pci driver using old method (new_id + unbind +
bind) will fail because of b803b29c1a. While on vanilla kernel
I'm able to use the old method successfully, it's failing on RHEL
kernels (not sure why).
Signed-off-by: Michal Privoznik <mprivozn@redhat.com>
Tested-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
