Now everything is ready to enable on demand file paging, so disable
unconditional populate on mmap for files and provide fault() for file
backed vmas. In fact file vma fault() is not different from anon vma
fault() so provide one general vma::fault() handler for both.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Walk page table and write out only dirty pages during file sync instead
of writing back entire mapping.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Currently map_file_page object can be used only once since its internal
state changes during mapping operation. For demand paging we do not want
to allocate/delete an object each time mapping is needed, so, to make
map_file_page object reusable, this patch resets its sate to a starting
one during finalize().

Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Currently pages are allocated just before they are mapped, but sometimes
we want to map preallocated pages (for shared mapping or mapping page
cache page for instance). This patch moves page acquisition process into
separate class. This will allow us to add subclass that provides pages
from shared pages pool or from a page cache.

Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Makes code easier to understand.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Currently fill_file_page is used to populate entire vma, so starting offset
and mapped file offset are always the same, but if fill_file_page will be used
to populate only part of a vma this will no longer be necessarily true. Keep
track of the lowest file offset populated and use it for read.

Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Value passed to std::vector constructor increase not only vector
capacity() but also its size(). We want only former, not later and it
is achieved by reserve() function.

Also remove prev_off which is a leftover of a debug code.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Generic page mapper code provides offset to mapper's small/huge_page
callbacks, but to make this offset meaningful it should be an offset
from a beginning of a vma that is mapped, but currently it is an offset
from a starting address of current mapping operation which means it
will be always zero it mapping is done page by page. Fix this by passing
starting vma address to page mapper for correct offset calculation.

Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Trying to execute unmapped page is not a punishable offense as long as
vma permissions are correct, so check them.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

It uses low level thread::wait_until() now which calls caller supplied
predicate with preemption disabled. If caller supplied code access not yet
mapped memory it will trigger an assertion on a page fault path.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Gleb Natapov <gleb@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Google Cloud Engine enabling.

Signed-off-by: Avi Kivity <avi@cloudius-systems.com>

Now we are showing names in gdb, it helps to name all of them.

Signed-off-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Replace OSv-specific constructs in tst-epoll.cc by their standard C++
counterparts (i.e., std::thread, std::chrono, std::cout).
This test now also runs (and of course, succeeds) on Linux.

In general, it is important at all our Linux-ABI tests (where we test our
implementation of the Linux/glibc functionality) to be able to run on Linux
as well. Otherwise, it is possible our tests don't actually test the right
thing (we may test for some expected behavior, but the actual behavior on
Linux is different).

I'm doing this in preparation for fixing issue #188 (fix edge-triggered
epoll).

Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

When syscall() was asked to perform an unknown syscall, we printed a
message with debug() and then called abort(). Since, recently, the message
from debug() is not visible on the console.

So pass the message directly to abort(), to do the right thing with it.

Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

The line respective to the mgmt was building it instead of cleaning
it up.

Reviewed-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Raphael S. Carvalho <raphaelsc@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

ide_drive::probe(), all the PCI_CLASS_STORAGE devices are probed
including the virtio-scsi device. This way ide driver will hijack
virtio-scsi handling.

Fix it by only matching PCI_CLASS_STORAGE && PCI_SUB_CLASS_STORAGE_IDE
devices

Signed-off-by: Asias He <asias@cloudius-systems.com>

In GCE, the pci bus looks like:

   $ lspci -x
   00:01.0 ISA bridge: Intel Corporation 82371AB/EB/MB PIIX4 ISA (rev 03)
   00: 86 80 10 71 03 01 00 00 03 00 01 06 00 00 80 00
   10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   30: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

   00:01.3 Bridge: Intel Corporation 82371AB/EB/MB PIIX4 ACPI (rev 03)
   00: 86 80 13 71 03 01 00 00 03 00 80 06 00 00 00 00
   10: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   20: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   30: 00 00 00 00 00 00 00 00 00 00 00 00 09 01 00 00

   00:03.0 SCSI storage controller: Red Hat, Inc Virtio SCSI
   00: f4 1a 04 10 03 01 00 00 00 00 00 01 00 00 00 00
   10: 01 c0 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   20: 00 00 00 00 00 00 00 00 00 00 00 00 f4 1a 08 00
   30: 00 00 00 00 00 00 00 00 00 00 00 00 0b 01 00 00

   00:04.0 Ethernet controller: Red Hat, Inc Virtio network device
   00: f4 1a 00 10 03 01 00 00 00 00 00 02 00 00 00 00
   10: 41 c0 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   20: 00 00 00 00 00 00 00 00 00 00 00 00 f4 1a 01 00
   30: 00 00 00 00 00 00 00 00 00 00 00 00 0b 01 00 00

This makes the code goes to !single_bus branch. However, device 00:00:0
is not present. So, No device is scanned! Obviously, this is not what
we want.

To fix, the following is almost correct except bus 0 might not be there.

   void pci_device_enumeration(void)
   {
          check_bus(0);
   }

To address it, we scan starting from bus 0 and exit the scan when the
first non-empty bus is found. The first non-empty bus should find the
entire hierarchy.

Signed-off-by: Asias He <asias@cloudius-systems.com>

Currently we only support scan one target. We scan target 0 by
default.  On GCE, the default target is 1. Let's scan scan target 1.

Signed-off-by: Asias He <asias@cloudius-systems.com>

We already copied req->resp.cmd.response into response. Use the shorter
one.

Signed-off-by: Asias He <asias@cloudius-systems.com>

GCE does not support REPORT LUNS, fallback to TEST_UNIT_READY to scan
luns.

Signed-off-by: Asias He <asias@cloudius-systems.com>

virtio-scsi device can not handle a request of an infinite size. It
reports max_sectors in the config space. Respect it by using
multiplex_strategy and dev->max_io_size to limit max request size.
multiplex_strategy works by splitting large bios into multiple smaller
bios.

Signed-off-by: Asias He <asias@cloudius-systems.com>

tag is a unique identifier of a scsi cmd. tag is required when more than
one concurrent requests are in-flight. Set tag to the address of bio.
Using bio for tag should be fine, since we can not have two same bio
in-flight at the same time. When one bio is reused, the previous one
should be finished already.

Signed-off-by: Asias He <asias@cloudius-systems.com>

resreved -> reserved

Signed-off-by: Asias He <asias@cloudius-systems.com>

Fix the REPORT LUNS CDB format according to the scsi spec.

Signed-off-by: Asias He <asias@cloudius-systems.com>

bio->bio_private is uesed by low-level device driver, e.g. virtio-scsi.
In multiplex_strategy, when one bio is splitted into multiple smaller
bios, we should copy bio_private into the newly created bios.

Signed-off-by: Asias He <asias@cloudius-systems.com>

No need to reset VIRTIO_PCI_QUEUE_PFN to 0 when reset the device. This
is not required by the virtio spec. Set device status to zero is enough
to reset the device.

Signed-off-by: Asias He <asias@cloudius-systems.com>

msi interrupt is not always implemented. Fallback to gsi interrupt when
msi interrupt is not available.

Signed-off-by: Asias He <asias@cloudius-systems.com>

msi interrupt is not always implemented. Fallback to gsi interrupt when
msi interrupt is not available.

Signed-off-by: Asias He <asias@cloudius-systems.com>

msi interrupt is not always implemented. Fallback to gsi interrupt when
msi interrupt is not available.

Signed-off-by: Asias He <asias@cloudius-systems.com>

Introduce unregister_level_triggered_handler to handle unregister of
shared vector.

Signed-off-by: Asias He <asias@cloudius-systems.com>

The overloaded version of register_interrupt_handler with a gsi
parameter is only used by register_level_triggered_handler. Fold it into
register_level_triggered_handler.

Signed-off-by: Asias He <asias@cloudius-systems.com>

nullptr check is done in rcu_dispose.

Signed-off-by: Asias He <asias@cloudius-systems.com>

Only call rcu_defer if the pointer is not nullptr.

Signed-off-by: Asias He <asias@cloudius-systems.com>

Add set_ack_and_handler helper to setup ack and handler function of gsi
level interrupt.

Signed-off-by: Asias He <asias@cloudius-systems.com>

This patch makes multiple devices which share the same interrupt line
works, e.g. on GCE, it shares virito-net and virtio-scsi interrupts.

Signed-off-by: Asias He <asias@cloudius-systems.com>

pre_eoi return 'true' when the interrupt is for the device, 'false'
otherwise. This is the preparation for interrupt sharing support.

Signed-off-by: Asias He <asias@cloudius-systems.com>

When we commit the changes in osv.git, we usually make a mistake to
also commit the changes in the submodules. To avoid this, this patch
is trying to add "ignore = all" to .gitmodules file.

Signed-off-by: Zhi Yong Wu <zwu.kernel@gmail.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

This patch reduces our BSS size from 9MB to less than 1MB, and reduces the
minimum amount of memory required to run the "rogue" image from 55 MB to
just 47 MB.  Together with the previous patch, this fixes #96.

Our almost-lock-free SMP allocator uses cpu-to-cpu rings to send freed
memory from one CPU to another. So for N cpus we need N^2 of these buffers.
But in an apparent attempt to support CPU hotplugging  the current code
allocated queues for 64 CPUs, instead of the real number of CPUs we have.
So even with a single CPU, we always have 64*64 of these buffers, each a
little over 2K, so totaling a whopping 8MB, wile only a tiny fraction of
this will ever be used

This patch changes the code to only allocate the number of these queues we
really need. It does not support CPU hot-plugging, but a lot of other places
in the code don't support hot-plugging either, so I left a FIXME to remind
us where we need to fix.

Reviewed-by: Glauber Costa <glommer@cloudius-systems.com>
Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

Signed-off-by: Zhi Yong Wu <zwu.kernel@gmail.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>

I recently added (thanks to Gleb!) the "openjdk8" module, which provides
Java 8. However, when one hopes to get a shell with OpenJDK8, using:

	make image=openjdk8,mgmt.shell

The mgmt.shell requires the "java" module, and we end up using both
the "java" and "openjdk8" modules.

This simple patch allows openjdk8's module.py to say:

	provides = ['java']

Which means that any later-loaded module which does require('java')
will *not* load the "java" module because opendjk8 claims it already
provides it.

Signed-off-by: Nadav Har'El <nyh@cloudius-systems.com>
Signed-off-by: Pekka Enberg <penberg@cloudius-systems.com>