If a device is part of a PM Domain (e.g. power and/or clock domain), its
power state is managed using Runtime PM. Without Runtime PM, the device
may not be powered up or clocked, causing subtle failures, crashes, or
system lock-ups when the device is accessed by the guest.
Fix this by adding Runtime PM support, powering the device when the VFIO
device is opened by the guest.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Simon Horman <horms+renesas@verge.net.au>
Acked-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
There exists a gap at the end of mdev_device_create() where the device
is visible to userspace, but we're not yet ready to handle removal, as
triggered through the 'remove' attribute. We handle this properly in
mdev_device_remove() with an -EAGAIN return, but we can marginally
reduce this gap by adding this attribute as a final step of our sysfs
setup.
Reviewed-by: Kirti Wankhede <kwankhede@nvidia.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
When we create an mdev device, we check for duplicates against the
parent device and return -EEXIST if found, but the mdev device
namespace is global since we'll link all devices from the bus. We do
catch this later in sysfs_do_create_link_sd() to return -EEXIST, but
with it comes a kernel warning and stack trace for trying to create
duplicate sysfs links, which makes it an undesirable response.
Therefore we should really be looking for duplicates across all mdev
parent devices, or as implemented here, against our mdev device list.
Using mdev_list to prevent duplicates means that we can remove
mdev_parent.lock, but in order not to serialize mdev device creation
and removal globally, we add mdev_device.active which allows UUIDs to
be reserved such that we can drop the mdev_list_lock before the mdev
device is fully in place.
Two behavioral notes; first, mdev_parent.lock had the side-effect of
serializing mdev create and remove ops per parent device. This was
an implementation detail, not an intentional guarantee provided to
the mdev vendor drivers. Vendor drivers can trivially provide this
serialization internally if necessary. Second, review comments note
the new -EAGAIN behavior when the device, and in particular the remove
attribute, becomes visible in sysfs. If a remove is triggered prior
to completion of mdev_device_create() the user will see a -EAGAIN
error. While the errno is different, receiving an error during this
period is not, the previous implementation returned -ENODEV for the
same condition. Furthermore, the consistency to the user is improved
in the case where mdev_device_remove_ops() returns error. Previously
concurrent calls to mdev_device_remove() could see the device
disappear with -ENODEV and return in the case of error. Now a user
would see -EAGAIN while the device is in this transitory state.
Reviewed-by: Kirti Wankhede <kwankhede@nvidia.com>
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
Acked-by: Halil Pasic <pasic@linux.ibm.com>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
- Capitalize the first word of error messages,
- Unwrap statements that fit on a single line,
- Use "VFIO" instead of "vfio" as the error message prefix.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Acked-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
MAP_DMA ioctls might be called from various threads within a process,
for example when using QEMU, the vCPU threads are often generating
these calls and we therefore take a reference to that vCPU task.
However, QEMU also supports vCPU hotplug on some machines and the task
that called MAP_DMA may have exited by the time UNMAP_DMA is called,
resulting in the mm_struct pointer being NULL and thus a failure to
match against the existing mapping.
To resolve this, we instead take a reference to the thread
group_leader, which has the same mm_struct and resource limits, but
is less likely exit, at least in the QEMU case. A difficulty here is
guaranteeing that the capabilities of the group_leader match that of
the calling thread, which we resolve by tracking CAP_IPC_LOCK at the
time of calling rather than at an indeterminate time in the future.
Potentially this also results in better efficiency as this is now
recorded once per MAP_DMA ioctl.
Reported-by: Xu Yandong <xuyandong2@huawei.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Pull aio updates from Al Viro:
"Majority of AIO stuff this cycle. aio-fsync and aio-poll, mostly.
The only thing I'm holding back for a day or so is Adam's aio ioprio -
his last-minute fixup is trivial (missing stub in !CONFIG_BLOCK case),
but let it sit in -next for decency sake..."
* 'work.aio-1' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (46 commits)
aio: sanitize the limit checking in io_submit(2)
aio: fold do_io_submit() into callers
aio: shift copyin of iocb into io_submit_one()
aio_read_events_ring(): make a bit more readable
aio: all callers of aio_{read,write,fsync,poll} treat 0 and -EIOCBQUEUED the same way
aio: take list removal to (some) callers of aio_complete()
aio: add missing break for the IOCB_CMD_FDSYNC case
random: convert to ->poll_mask
timerfd: convert to ->poll_mask
eventfd: switch to ->poll_mask
pipe: convert to ->poll_mask
crypto: af_alg: convert to ->poll_mask
net/rxrpc: convert to ->poll_mask
net/iucv: convert to ->poll_mask
net/phonet: convert to ->poll_mask
net/nfc: convert to ->poll_mask
net/caif: convert to ->poll_mask
net/bluetooth: convert to ->poll_mask
net/sctp: convert to ->poll_mask
net/tipc: convert to ->poll_mask
...
Bisection by Amadeusz Sławiński implicates this commit leading to bad
page state issues after VM shutdown, likely due to unbalanced page
references. The original commit was intended only as a performance
improvement, therefore revert for offline rework.
Link: https://lkml.org/lkml/2018/6/2/97
Fixes: 356e88ebe4 ("vfio/type1: Improve memory pinning process for raw PFN mapping")
Cc: Jason Cai (Xiang Feng) <jason.cai@linux.alibaba.com>
Reported-by: Amadeusz Sławiński <amade@asmblr.net>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
These abstract out calls to the poll method in preparation for changes
in how we poll.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
The ioeventfd here is actually irqfd handling of an ioeventfd such as
supported in KVM. A user is able to pre-program a device write to
occur when the eventfd triggers. This is yet another instance of
eventfd-irqfd triggering between KVM and vfio. The impetus for this
is high frequency writes to pages which are virtualized in QEMU.
Enabling this near-direct write path for selected registers within
the virtualized page can improve performance and reduce overhead.
Specifically this is initially targeted at NVIDIA graphics cards where
the driver issues a write to an MMIO register within a virtualized
region in order to allow the MSI interrupt to re-trigger.
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
The iowriteXX/ioreadXX functions assume little endian hardware and
convert to little endian on a write and from little endian on a read.
We currently do our own explicit conversion to negate this. Instead,
add some endian dependent defines to avoid all byte swaps. There
should be no functional change other than big endian systems aren't
penalized with wasted swaps.
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
When using vfio to pass through a PCIe device (e.g. a GPU card) that
has a huge BAR (e.g. 16GB), a lot of cycles are wasted on memory
pinning because PFNs of PCI BAR are not backed by struct page, and
the corresponding VMA has flag VM_PFNMAP.
With this change, when pinning a region which is a raw PFN mapping,
it can skip unnecessary user memory pinning process, and thus, can
significantly improve VM's boot up time when passing through devices
via VFIO. In my test on a Xeon E5 2.6GHz, the time mapping a 16GB
BAR was reduced from about 0.4s to 1.5us.
Signed-off-by: Jason Cai (Xiang Feng) <jason.cai@linux.alibaba.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
This reverts commit 2170dd0431
The intent of commit 2170dd0431 ("vfio-pci: Mask INTx if a device is
not capabable of enabling it") was to disallow the user from seeing
that the device supports INTx if the platform is incapable of enabling
it. The detection of this case however incorrectly includes devices
which natively do not support INTx, such as SR-IOV VFs, and further
discussions reveal gaps even for the target use case.
Reported-by: Arjun Vynipadath <arjun@chelsio.com>
Fixes: 2170dd0431 ("vfio-pci: Mask INTx if a device is not capabable of enabling it")
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
VFIO IOMMU type1 currently upmaps IOVA pages synchronously, which requires
IOTLB flushing for every unmapping. This results in large IOTLB flushing
overhead when handling pass-through devices has a large number of mapped
IOVAs. This can be avoided by using the new IOTLB flushing interface.
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Joerg Roedel <joro@8bytes.org>
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>
[aw - use LIST_HEAD]
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Filesystem-DAX is incompatible with 'longterm' page pinning. Without
page cache indirection a DAX mapping maps filesystem blocks directly.
This means that the filesystem must not modify a file's block map while
any page in a mapping is pinned. In order to prevent the situation of
userspace holding of filesystem operations indefinitely, disallow
'longterm' Filesystem-DAX mappings.
RDMA has the same conflict and the plan there is to add a 'with lease'
mechanism to allow the kernel to notify userspace that the mapping is
being torn down for block-map maintenance. Perhaps something similar can
be put in place for vfio.
Note that xfs and ext4 still report:
"DAX enabled. Warning: EXPERIMENTAL, use at your own risk"
...at mount time, and resolving the dax-dma-vs-truncate problem is one
of the last hurdles to remove that designation.
Acked-by: Alex Williamson <alex.williamson@redhat.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: kvm@vger.kernel.org
Cc: <stable@vger.kernel.org>
Reported-by: Haozhong Zhang <haozhong.zhang@intel.com>
Tested-by: Haozhong Zhang <haozhong.zhang@intel.com>
Fixes: d475c6346a ("dax,ext2: replace XIP read and write with DAX I/O")
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
This is the mindless scripted replacement of kernel use of POLL*
variables as described by Al, done by this script:
for V in IN OUT PRI ERR RDNORM RDBAND WRNORM WRBAND HUP RDHUP NVAL MSG; do
L=`git grep -l -w POLL$V | grep -v '^t' | grep -v /um/ | grep -v '^sa' | grep -v '/poll.h$'|grep -v '^D'`
for f in $L; do sed -i "-es/^\([^\"]*\)\(\<POLL$V\>\)/\\1E\\2/" $f; done
done
with de-mangling cleanups yet to come.
NOTE! On almost all architectures, the EPOLL* constants have the same
values as the POLL* constants do. But they keyword here is "almost".
For various bad reasons they aren't the same, and epoll() doesn't
actually work quite correctly in some cases due to this on Sparc et al.
The next patch from Al will sort out the final differences, and we
should be all done.
Scripted-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull VFIO updates from Alex Williamson:
- Mask INTx from user if pdev->irq is zero (Alexey Kardashevskiy)
- Capability helper cleanup (Alex Williamson)
- Allow mmaps overlapping MSI-X vector table with region capability
exposing this feature (Alexey Kardashevskiy)
- mdev static cleanups (Xiongwei Song)
* tag 'vfio-v4.16-rc1' of git://github.com/awilliam/linux-vfio:
vfio: mdev: make a couple of functions and structure vfio_mdev_driver static
vfio-pci: Allow mapping MSIX BAR
vfio: Simplify capability helper
vfio-pci: Mask INTx if a device is not capabable of enabling it
The functions vfio_mdev_probe, vfio_mdev_remove and the structure
vfio_mdev_driver are only used in this file, so make them static.
Clean up sparse warnings:
drivers/vfio/mdev/vfio_mdev.c:114:5: warning: no previous prototype
for 'vfio_mdev_probe' [-Wmissing-prototypes]
drivers/vfio/mdev/vfio_mdev.c:121:6: warning: no previous prototype
for 'vfio_mdev_remove' [-Wmissing-prototypes]
Signed-off-by: Xiongwei Song <sxwjean@gmail.com>
Reviewed-by: Quan Xu <quan.xu0@gmail.com>
Reviewed-by: Liu, Yi L <yi.l.liu@intel.com>
Reviewed-by: Kirti Wankhede <kwankhede@nvidia.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
By default VFIO disables mapping of MSIX BAR to the userspace as
the userspace may program it in a way allowing spurious interrupts;
instead the userspace uses the VFIO_DEVICE_SET_IRQS ioctl.
In order to eliminate guessing from the userspace about what is
mmapable, VFIO also advertises a sparse list of regions allowed to mmap.
This works fine as long as the system page size equals to the MSIX
alignment requirement which is 4KB. However with a bigger page size
the existing code prohibits mapping non-MSIX parts of a page with MSIX
structures so these parts have to be emulated via slow reads/writes on
a VFIO device fd. If these emulated bits are accessed often, this has
serious impact on performance.
This allows mmap of the entire BAR containing MSIX vector table.
This removes the sparse capability for PCI devices as it becomes useless.
As the userspace needs to know for sure whether mmapping of the MSIX
vector containing data can succeed, this adds a new capability -
VFIO_REGION_INFO_CAP_MSIX_MAPPABLE - which explicitly tells the userspace
that the entire BAR can be mmapped.
This does not touch the MSIX mangling in the BAR read/write handlers as
we are doing this just to enable direct access to non MSIX registers.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
[aw - fixup whitespace, trim function name]
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
The vfio_info_add_capability() helper requires the caller to pass a
capability ID, which it then uses to fill in header fields, assuming
hard coded versions. This makes for an awkward and rigid interface.
The only thing we want this helper to do is allocate sufficient
space in the caps buffer and chain this capability into the list.
Reduce it to that simple task.
Reviewed-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Reviewed-by: Kirti Wankhede <kwankhede@nvidia.com>
Reviewed-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
At the moment VFIO rightfully assumes that INTx is supported if
the interrupt pin is not set to zero in the device config space.
However if that is not the case (the pin is not zero but pdev->irq is),
vfio_intx_enable() fails.
In order to prevent the userspace from trying to enable INTx when we know
that it cannot work, let's mask the PCI_INTERRUPT_PIN register.
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
