We currently use the whole top PGD entry for the switcher, but that's
hitting the fixmap in some configurations (mainly, large NR_CPUS).
Introduce a variable, currently set to the constant.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Returning EMFILE (process has too many open files) is incorrect to
indicate a port is already open by another process.  Use EBUSY for that.

This does change what we report to userspace, but I believe userspace
can look at it this way: it gets EBUSY, a new error code, instead of
EMFILE.  It's still an error, and that's not changing.
Reported-by: NMateusz Guzik <mguzik@redhat.com>
Signed-off-by: NAmit Shah <amit.shah@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Add hot cpu notifier to reset the request virtqueue affinity
when doing cpu hotplug.

Cc: linux-scsi@vger.kernel.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

This patch adds queue steering to virtio-scsi.  When a target is sent
multiple requests, we always drive them to the same queue so that FIFO
processing order is kept.  However, if a target was idle, we can choose
a queue arbitrarily.  In this case the queue is chosen according to the
current VCPU, so the driver expects the number of request queues to be
equal to the number of VCPUs.  This makes it easy and fast to select
the queue, and also lets the driver optimize the IRQ affinity for the
virtqueues (each virtqueue's affinity is set to the CPU that "owns"
the queue).

The speedup comes from improving cache locality and giving CPU affinity
to the virtqueues, which is why this scheme was selected.  Assuming that
the thread that is sending requests to the device is I/O-bound, it is
likely to be sleeping at the time the ISR is executed, and thus executing
the ISR on the same processor that sent the requests is cheap.

However, the kernel will not execute the ISR on the "best" processor
unless you explicitly set the affinity.  This is because in practice
you will have many such I/O-bound processes and thus many otherwise
idle processors.  Then the kernel will execute the ISR on a random
processor, rather than the one that is sending requests to the device.

The alternative to per-CPU virtqueues is per-target virtqueues.  To
achieve the same locality, we could dynamically choose the virtqueue's
affinity based on the CPU of the last task that sent a request.  This
is less appealing because we do not set the affinity directly---we only
provide a hint to the irqbalanced running in userspace.  Dynamically
changing the affinity only works if the userspace applies the hint
fast enough.

Cc: linux-scsi@vger.kernel.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: NAsias He <asias@redhat.com>
Tested-by: NVenkatesh Srinivas <venkateshs@google.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Avoid duplicated code in all of the callers.

Cc: linux-scsi@vger.kernel.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

This will be needed soon in order to retrieve the per-target
struct.

Cc: linux-scsi@vger.kernel.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

virtio_scsi_target_state is now empty.  We will find new uses for it in
the next few patches, so this patch does not drop it completely.

And as James suggested, we use entries target_alloc and target_destroy
in the host template to allocate and destroy the virtio_scsi_target_state
of each target, attach this struct to scsi_target->hostdata. Now
we can get at it from the sdev with scsi_target(sdev)->hostdata.
No messing around with fixed size arrays and bulk memory allocation
and no need to pass in the maximum target size as a parameter because
everything should now happen dynamically.

Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: linux-scsi@vger.kernel.org
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Those symbols only used within this file, and should be static.
Signed-off-by: NWei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: NAmit Shah <amit.shah@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Fix to return a negative error code from the error handling
case instead of 0, as returned elsewhere in this function.
Signed-off-by: NWei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Some head files were split or moved to uapi/ without
updating MAINTAINERS.
Signed-off-by: Amos Kong <kongjianjun@gmail.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

virtio_balloon.h exports "u16" and "u64" to userspace. Use "__u16" and
"__u64" instead.
Signed-off-by: NPaul Bolle <pebolle@tiscali.nl>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Check that vringh_config is not NULL before using it.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Check on the correct return value from
vringh_notify_enable_kern(). It returns false if
more packets are available, not true.
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Make the rest of the paths use virtqueue_add_sgs or add_outbuf.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

virtio_add_buf() is going away, replaced with virtio_add_sgs() which
takes multiple terminated scatterlists.

Cc: Eric Van Hensbergen <ericvh@gmail.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

We never add buffers with input and output parts, so use the new accessors.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

We never add buffers with input and output parts, so use the new accessors.

Cc: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

We never add buffers with input and output parts, so use the new accessors.

Cc: Sjur Brendeland <sjur.brandeland@stericsson.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

We never add buffers with input and output parts, so use the new accessors.
Acked-by: NAmit Shah <amit.shah@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

We never add buffers with input and output parts, so use the new accessors.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Reviewed-by: NAsias He <asias@redhat.com>

We never add buffers with input and output parts, so use the new accessors.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

It's a bit cleaner to hand multiple sgs, rather than one big one.

Cc: "Michael S. Tsirkin" <mst@redhat.com>
Tested-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

It's a bit clearer, and add_buf is going away.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Reviewed-by: NAsias He <asias@redhat.com>

Using the new virtqueue_add_sgs function lets us simplify the queueing
path.  In particular, all data protected by the tgt_lock is just gone
(multiqueue will find a new use for the lock).
Signed-off-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Acked-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

It's simply a flag as to whether we have data now, so make it an
explicit function parameter rather than a member of struct
virtblk_req.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Reviewed-by: NAsias He <asias@redhat.com>

(This is a respin of Paolo Bonzini's patch, but it calls
virtqueue_add_sgs() instead of his multi-part API).

This is similar to the previous patch, but a bit more radical
because the bio and req paths now share the buffer construction
code.  Because the req path doesn't use vbr->sg, however, we
need to add a couple of arguments to __virtblk_add_req.

We also need to teach __virtblk_add_req how to build SCSI command
requests.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Reviewed-by: NAsias He <asias@redhat.com>

(This is a respin of Paolo Bonzini's patch, but it calls
virtqueue_add_sgs() instead of his multi-part API).

Move the creation of the request header and response footer to
__virtblk_add_req.  vbr->sg only contains the data scatterlist,
the header/footer are added separately using virtqueue_add_sgs().

With this change, virtio-blk (with use_bio) is not relying anymore on
the virtio functions ignoring the end markers in a scatterlist.
The next patch will do the same for the other path.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Reviewed-by: NAsias He <asias@redhat.com>

Right now, both virtblk_add_req and virtblk_add_req_wait call
virtqueue_add_buf.  To prepare for the next patches, abstract the call
to virtqueue_add_buf into a new function __virtblk_add_req, and include
the waiting logic directly in virtblk_add_req.
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Reviewed-by: NAsias He <asias@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

As expected, the simplified accessors are faster.

for i in `seq 50`; do /usr/bin/time -f 'Wall time:%e' ./vringh_test --indirect --eventidx --parallel --fast-vringh; done 2>&1 | stats --trim-outliers:

Before:
	Using CPUS 0 and 3
	Guest: notified 0, pinged 39062-39063(39063)
	Host: notified 39062-39063(39063), pinged 0
	Wall time:1.760000-2.220000(1.789167)

After:
	Using CPUS 0 and 3
	Guest: notified 0, pinged 39037-39063(39062)
	Host: notified 39037-39063(39062), pinged 0
	Wall time:1.640000-1.810000(1.676875)
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

These are specialized versions of virtqueue_add_buf(), which cover
over 80% of cases and are far clearer.

In particular, the scatterlists passed to these functions don't have
to be clean (ie. we ignore end markers).
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

virtio_scsi can really use this, to avoid the current hack of copying
the whole sg array.  Some other things get slightly neater, too.

This causes a slowdown in virtqueue_add_buf(), which is implemented as
a wrapper.  This is addressed in the next patches.

for i in `seq 50`; do /usr/bin/time -f 'Wall time:%e' ./vringh_test --indirect --eventidx --parallel --fast-vringh; done 2>&1 | stats --trim-outliers:

Before:
	Using CPUS 0 and 3
	Guest: notified 0, pinged 39009-39063(39062)
	Host: notified 39009-39063(39062), pinged 0
	Wall time:1.700000-1.950000(1.723542)

After:
	Using CPUS 0 and 3
	Guest: notified 0, pinged 39062-39063(39063)
	Host: notified 39062-39063(39063), pinged 0
	Wall time:1.760000-2.220000(1.789167)
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Reviewed-by: NWanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: NAsias He <asias@redhat.com>

This is useful in places that recycle the same scatterlist multiple
times, and do not want to incur the cost of sg_init_table every
time in hot paths.
Acked-by: NJens Axboe <axboe@kernel.dk>
Signed-off-by: NPaolo Bonzini <pbonzini@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Add the CAIF Virtio shared memory driver for talking
to a modem.

This CAIF Link layer communicates to the modem over
shared memory. It is implemented as a virtio_driver.
The underlying virtio device is managed by the remoteproc
framework. The Virtio queue is used for transmitting data
to the modem, and the new vringh is used for receiving data.

Genalloc is used for managing the shared memory used for TX
data. The default dma-alloc-coherent allocator can only
allocate whole pages, and this wastes too much shared memory.

Flow control is implemented by stopping the TX-queues if the
virtio queues go full or we run out of memory. Queued are
reopened when queues are below the watermark.

NAPI is used in RX path, and a dedicated tasklet is used
for releasing TX buffers.
Signed-off-by: NErwan Yvin <erwan.yvin@stericsson.com>
Acked-by: NDavid S. Miller <davem@davemloft.net>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (minor fixes)

Add wrappers for the host vrings to support loose
coupling between the virtio device and driver.

A new struct vringh_config_ops with the functions
find_vrhs() and del_vrhs() is added to the virtio_device
struct. This enables virtio drivers to manage virtio
host rings without detailed knowledge of how the
vrings are created and deleted.

The function vringh_notify() is added so vringh clients
can notify the other side that buffers are added to the
used-ring.

Cc: Ohad Ben-Cohen <ohad@wizery.com>
Signed-off-by: NSjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> (constified vringh_config)

This is mainly to test the drivers/vhost/vringh.c code, but it also
uses the drivers/virtio/virtio_ring.c code for the guest side.

Usage for testing the basic implementation:

	./vringh_test
	# Test with indirect descriptors
	./vringh_test --indirect
	# Test with indirect descriptors and event indexex
	./vringh_test --indirect --eventidx

You can run a parallel stress test by adding --parallel to any of the
above options.

eg ./vringh_test --parallel:
	Using CPUS 0 and 3
	Guest: notified 10107974, pinged 107970
	Host: notified 108158, pinged 3172148

./vringh_test --indirect --eventidx --parallel:
	Using CPUS 0 and 3
	Guest: notified 156357, pinged 156251
	Host: notified 156251, pinged 78179

Average of 50 times doing ./vringh_test --indirect --eventidx --parallel:
	2.840000-3.040000(2.927292)user
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Getting use of virtio rings correct is tricky, and a recent patch saw
an implementation of in-kernel rings (as separate from userspace).

This abstracts the business of dealing with the virtio ring layout
from the access (userspace or direct); to do this, we use function
pointers, which gcc inlines correctly.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>
Acked-by: NMichael S. Tsirkin <mst@redhat.com>

This makes them a bit more like the kernel headers, so we can include more
real kernel headers in our tests.

In addition this means that we don't break tools/virtio with the next
patch.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

The host side of ring needs this logic too.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

Signed-off-by: NMichael S. Tsirkin <mst@redhat.com>
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>

We haven't been keeping it in sync, so just remove it.
Signed-off-by: NRusty Russell <rusty@rustcorp.com.au>