The Tx interrupt is now kept disabled in the common case, only
enabled when the number of free descriptors in the queue falls
below a threshold. Transmitted frames are cleared from the VQ
before subsequent transmit, or in the watchdog timer.
This was a very big performance improvement for an experimental
Netmap bhyve backend.
MFC after: 1 month
Prevent the Xen and VirtIO balloon drivers from marking pages as
wired. This prevents them from increasing the system wired page count,
which can lead to mlock failing because of hitting the limit in
vm.max_wired.
In the Xen case make sure pages are zeroed before giving them back to
the hypervisor, or else we might be leaking data. Also remove the
balloon_{append/retrieve} and link pages directly into the
ballooned_pages queue using the plinks.q field in the page struct.
Sponsored by: Citrix Systems R&D
Reviewed by: kib, bryanv
Approved by: gibbs
dev/virtio/balloon/virtio_balloon.c:
- Don't allocate pages with VM_ALLOC_WIRED.
dev/xen/balloon/balloon.c:
- Don't allocate pages with VM_ALLOC_WIRED.
- Make sure pages are zeroed before giving them back to the
hypervisor.
- Remove the balloon_entry struct and the balloon_{append/retrieve}
functions and use the page plinks.q entry to link the pages
directly into the ballooned_pages queue.
the queue where to enqueue pages that are going to be unwired.
- Add stronger checks to the enqueue/dequeue for the pagequeues when
adding and removing pages to them.
Of course, for unmanaged pages the queue parameter of vm_page_unwire() will
be ignored, just as the active parameter today.
This makes adding new pagequeues quicker.
This change effectively modifies the KPI. __FreeBSD_version will be,
however, bumped just when the full cache of free pages will be
evicted.
Sponsored by: EMC / Isilon storage division
Reviewed by: alc
Tested by: pho
The header structure consists of two 1-byte elements, but it must always
be describable by a single SG entry. Note for consistency, specify the
alignment everywhere, even if the structure has the appropriate natural
alignment since it contains a uint16_t.
Obtained from: DragonFlyBSD
MFC after: 1 week
These defines are applicable to userland too, but virtqueue.h contains
the kernel virtqueue interface, and is therefore not usable in userland.
Note that Linux places these defines in virtio_ring.h, but I don't want
the drivers including this header file to keep the VirtIO ring opaque to
everything but the virtqueue.
MFC after: 1 week
The eventual goal is to share this file with userland, so
remove the macro that is only specific for virtio_pci(4).
Instead, add the VIRTIO_PCI_CONFIG_OFF macro from Linux to
get the config size whether MSIX is enabled or not.
MFC after: 1 week
A similar fix should be applied to vmxnet, ixgbe, igb, i40e.
(some of them previously reported by Michael Tuexen)
Drivers using if_transmit are correct, and so are most of the
other drivers that reassing if_transmit.
Among other things, this bug causes panics when using netmap emulation
on top of generic drivers.
Approved by: bryanv
MFC after: 3 days
Reorganize the previous contexts of the file as it is in Linux. The
eventual goal is to install the header files and share them between
the kernel and bhyve.
MFC after: 1 week
interface, in the r241616 a crutch was provided. It didn't work well, and
finally we decided that it is time to break ABI and simply make if_baudrate
a 64-bit value. Meanwhile, the entire struct if_data was reviewed.
o Remove the if_baudrate_pf crutch.
o Make all fields of struct if_data fixed machine independent size. The
notion of data (packet counters, etc) are by no means MD. And it is a
bug that on amd64 we've got a 64-bit counters, while on i386 32-bit,
which at modern speeds overflow within a second.
This also removes quite a lot of COMPAT_FREEBSD32 code.
o Give 16 bit for the ifi_datalen field. This field was provided to
make future changes to if_data less ABI breaking. Unfortunately the
8 bit size of it had effectively limited sizeof if_data to 256 bytes.
o Give 32 bits to ifi_mtu and ifi_metric.
o Give 64 bits to the rest of fields, since they are counters.
__FreeBSD_version bumped.
Discussed with: emax
Sponsored by: Netflix
Sponsored by: Nginx, Inc.
The sglist segment array has grown to a bit over 512 bytes (on
64-bit system) which is more than ideally should be put on the
stack. Instead allocate an appropriately sized sglist and hang
it off each Rx/Tx queue structure.
Bump the maximum number of Tx segments to 64 to make it unlikely
we'll have defragment an mbuf chain. Our previous count was
rounded up to this value since it is the next power of two, so
effective memory usage should not change.
Also only allocate the maximum number of Tx segments if TSO was
negotiated.
to this event, adding if_var.h to files that do need it. Also, include
all includes that now are included due to implicit pollution via if_var.h
Sponsored by: Netflix
Sponsored by: Nginx, Inc.
This is a significant rewrite of much of the previous driver; lots of
misc. cleanup was also performed, and support for a few other minor
features was also added.
Partial support for the EVENT_IDX feature was added a while ago,
but this commit adds an interface for the device driver to hint
how long (in terms of descriptors) the next interrupt should be
delayed.
The first user of this will be used to reduce VirtIO net's Tx
completion interrupts.
additional information, when the page is guaranteed to not belong to a
paging queue. Usually, this results in a lot of type casts which make
reasoning about the code correctness harder.
Sometimes m->object is used instead of pageq, which could cause real
and confusing bugs if non-NULL m->object is leaked. See r141955 and
r253140 for examples.
Change the pageq member into a union containing explicitly-typed
members. Use them instead of type-punning or abusing m->object in x86
pmaps, uma and vm_page_alloc_contig().
Requested and reviewed by: alc
Sponsored by: The FreeBSD Foundation
r252680:
Fix SIM lock not owned panic
The CAM locking requirements of registering an async
callback has changed so the SIM lock must be held. Remove
code that explicitly dropped the lock around the register.
Also return CAM_SEL_TIMEOUT instead of CAM_TID_INVALID
for bad targets to avoid a lot console spam during bus
scans.
MFC after: 1 month
This commit is primarily a significant cleanup to the interrupt
allocation code that had gotten a bit jumbled from having to
support per-vq MSIX, shared MSIX, MSI, and legacy style interrupts.
Contains projects/virtio commits:
r246064:
virtio_pci: Rewrite allocation of interrupts
r246065:
virtio_pci: Remove spaces before a tab
r246066:
virtio_pci: Dynamically allocate the virtqueue array
r246304:
virtio_pci: Clean up after failed virtqueue alloc attempt
r246305:
virtio_pci: Move no interrupt check into the PCI interrupt handlers
r246308:
virtio_pci: Remove unused variable
MFC after: 1 month
Minor changes to the network driver. A multiqueue driver that is
a significant rewrite will be in merged shortly.
Contains projects/virtio commits:
r246058:
vtnet: Move an mbuf ASSERT to the calling function
r246059:
vtnet: Tweak ASSERT message
MFC after: 1 month
Contains projects/virtio commits:
r245717:
virtio_balloon: Make the softc lock a regular mutex
r245718:
virtio_balloon: Remove two unuseful ASSERTs
r245719:
virtio_balloon: More verbose ASSERT messages
r245720:
virtio_balloon: Simplify lowmem handling in vtballoon_inflate()
r252530:
virtio_balloon: Use just a kthread instead of dedciated kproc
r252568:
virtio_balloon: Need to use kthread_exit() after r252530
MFC after: 1 month
The notable changes of this commit are support for disk resizing
and chases updates to the spec regarding write caching.
Contains projects/virtio commits:
r245713:
virtio_blk: Replace __FUNCTION__ with __func__
r245714:
virtio_blk: Use more consistent mutex name
r245715:
virtio_blk: Print device name too if failed to reinit during dump
r245716:
virtio_blk: Remove an unuseful ASSERT
r245723:
virtio_blk: Record the vendor and device information
r245724:
virtio_blk: Add resize support
r245726:
virtio_blk: More verbose ASSERT messages
r245730:
virtio_blk: Tweak resize announcement message
r246061:
virtio_blk: Do not always read entire config
r246062:
virtio_blk: Use topology to set the stripe size/offset
r246307:
virtio_blk: Correct stripe offset calculation
r246063:
virtio_blk: Add support for write cache enable feature
r246303:
virtio_blk: Expand a comment
r252529:
virtio_blk: Improve write cache handling
r252681:
virtio_blk: Remove unneeded curly braces
MFC after: 1 month
Contains projects/virtio commits:
r245709:
Each VirtIO device was scheduling its own taskqueue(9) to do the
off-level interrupt handling. ithreads(9) is the more nature way
to do this. The primary motivation for this work to better support
network multiqueue.
r245710:
virtio: Change virtqueue intr handlers to return void
r245711:
virtio_blk: Remove interrupt taskqueue
r245721:
vtnet: Remove interrupt taskqueue
r245722:
virtio_scsi: Remove interrupt taskqueue
r245747:
vtnet: Remove taskqueue fields missed in r245721
MFC after: 1 month
QEMU 1.4 made the descriptor requirement stricter - the size of buffer
descriptor must exactly match the number of MAC addresses provided.
PR: kern/178955
MFC after: 5 days
Stop abusing xpt_periph in random plases that really have no periph related
to CCB, for example, bus scanning. NULL value is fine in such cases and it
is correctly logged in debug messages as "noperiph". If at some point we
need some real XPT periphs (alike to pmpX now), quite likely they will be
per-bus, and not a single global instance as xpt_periph now.
When the VirtIO barrier feature is not negotiated, the driver
must enforce the proper ordering for BIO_ORDERED BIOs. All the
in-flight BIOs must complete before starting the BIO, and the
ordered BIO must complete before subsequent BIOs can start.
Also fix a few whitespace nits.
Reported by: neel
Approved by: grehan (mentor)
MFC after: 3 days
every architecture's busdma_machdep.c. It is done by unifying the
bus_dmamap_load_buffer() routines so that they may be called from MI
code. The MD busdma is then given a chance to do any final processing
in the complete() callback.
The cam changes unify the bus_dmamap_load* handling in cam drivers.
The arm and mips implementations are updated to track virtual
addresses for sync(). Previously this was done in a type specific
way. Now it is done in a generic way by recording the list of
virtuals in the map.
Submitted by: jeff (sponsored by EMC/Isilon)
Reviewed by: kan (previous version), scottl,
mjacob (isp(4), no objections for target mode changes)
Discussed with: ian (arm changes)
Tested by: marius (sparc64), mips (jmallet), isci(4) on x86 (jharris),
amd64 (Fabian Keil <freebsd-listen@fabiankeil.de>)
