This optimizes the frobbing of our Switcher map.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

It's always to same, so no need to put in the PTE every time we're
about to run.  Keep a flag to track whether the pagetable has the
Switcher entries allocated, and when allocating always initialize the
Switcher text PTE.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

We currently use the whole top PGD entry for the switcher, so we
simply share a fixed page of PTEs between all guests (actually, it's
one per Host CPU, to ensure isolation between guests).

Changes to a scheme where every guest has its own mappings.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

We will need this in page_table.c soon.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

We want a separate find_pte() function so we can call it for populating the
switcher PTE entries.

We can also use it in page_writable().
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

This is a bit neater: we can immediately return if a PTE/PGD/PMD entry
is invalid (which also kills the guest).  It means we don't risk using
invalid entries as we reshuffle the code.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

ie. SHARED_SWITCHER_PAGES == 1.  It is well under a page, and it's a
minor simplification: it's nice to have *one* simplification in a
patch series!
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

There is a single page with the Switcher in it, but it's followed by 2
pages per Host CPU.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

We can use switcher_addr directly.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

We currently assume that the Switcher the top pgd; we want to remove
this assumption, so check that vaddr is OK, rather then checking pgd
index.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

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: Rusty 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: Mateusz Guzik <mguzik@redhat.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty 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: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: Asias He <asias@redhat.com>
Signed-off-by: Rusty 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: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: Asias He <asias@redhat.com>
Tested-by: Venkatesh Srinivas <venkateshs@google.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

Avoid duplicated code in all of the callers.

Cc: linux-scsi@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: Asias He <asias@redhat.com>
Signed-off-by: Rusty 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: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: Asias He <asias@redhat.com>
Signed-off-by: Rusty 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: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Reviewed-by: Asias He <asias@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

Those symbols only used within this file, and should be static.
Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Rusty 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: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Acked-by: Sjur Brændeland <sjur.brandeland@stericsson.com>
Signed-off-by: Rusty 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: Rusty Russell <rusty@rustcorp.com.au>

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

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

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

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

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

We never add buffers with input and output parts, so use the new accessors.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reviewed-by: Asias 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: Asias He <asias@redhat.com>
Signed-off-by: Rusty 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: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

It's a bit clearer, and add_buf is going away.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reviewed-by: Asias 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: Wanlong Gao <gaowanlong@cn.fujitsu.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Asias He <asias@redhat.com>
Signed-off-by: Rusty 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: Rusty Russell <rusty@rustcorp.com.au>
Reviewed-by: Asias 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: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reviewed-by: Asias 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: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Reviewed-by: Asias 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: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Asias He <asias@redhat.com>
Signed-off-by: Rusty 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: Rusty 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: Rusty Russell <rusty@rustcorp.com.au>