hw/intc/armv7m_nvic: Don't hardcode M profile ID registers in NVIC

Instead of hardcoding the values of M profile ID registers in the
NVIC, use the fields in the CPU struct. This will allow us to
give different M profile CPU types different ID register values.

This commit includes the addition of the missing ID_ISAR5,
which exists as RES0 in both v7M and v8M.

(The values of the ID registers might be wrong for the M4 --
this commit leaves the behaviour there unchanged.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180209165810.6668-2-peter.maydell@linaro.org

target/arm: Handle SVE registers when using clear_vec_high

When storing to an AdvSIMD FP register, all of the high
bits of the SVE register are zeroed.  Therefore, call it
more often with is_q as a parameter.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180211205848.4568-6-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Enforce access to ZCR_EL at translation

This also makes sure that we get the correct ordering of
SVE vs FP exceptions.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180211205848.4568-5-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Suppress TB end for FPCR/FPSR

Nothing in either register affects the TB.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180211205848.4568-4-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Enforce FP access to FPCR/FPSR

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180211205848.4568-3-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

target/arm: Remove ARM_CP_64BIT from ZCR_EL registers

Because they are ARM_CP_STATE_AA64, ARM_CP_64BIT is implied.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Message-id: 20180211205848.4568-2-richard.henderson@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/aspeed: simplify using the 'unimplemented device' for aspeed_soc.io

(qemu) info mtree
 address-space: cpu-memory-0
   0000000000000000-ffffffffffffffff (prio 0, i/o): system
     0000000000000000-0000000007ffffff (prio 0, rom): aspeed.boot_rom
-    000000001e600000-000000001e7fffff (prio -1, i/o): aspeed_soc.io
+    000000001e600000-000000001e7fffff (prio -1000, i/o): aspeed_soc.io
     000000001e620000-000000001e6200ff (prio 0, i/o): aspeed.smc.ast2500-fmc
     000000001e630000-000000001e6300ff (prio 0, i/o): aspeed.smc.ast2500-spi1
     000000001e631000-000000001e6310ff (prio 0, i/o): aspeed.smc.ast2500-spi2

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Message-id: 20180209085755.30414-3-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

hw/arm/aspeed: directly map the serial device to the system address space

(qemu) info mtree
 address-space: cpu-memory-0
   0000000000000000-ffffffffffffffff (prio 0, i/o): system
     0000000000000000-0000000007ffffff (prio 0, rom): aspeed.boot_rom
     000000001e600000-000000001e7fffff (prio -1, i/o): aspeed_soc.io
-      000000001e784000-000000001e78401f (prio 0, i/o): serial
     000000001e620000-000000001e6200ff (prio 0, i/o): aspeed.smc.ast2500-fmc
     000000001e630000-000000001e6300ff (prio 0, i/o): aspeed.smc.ast2500-spi1
     [...]
     000000001e720000-000000001e728fff (prio 0, ram): aspeed.sram
     000000001e782000-000000001e782fff (prio 0, i/o): aspeed.timer
+    000000001e784000-000000001e78401f (prio 0, i/o): serial
     000000001e785000-000000001e78501f (prio 0, i/o): aspeed.wdt
     000000001e785020-000000001e78503f (prio 0, i/o): aspeed.wdt

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Reviewed-by: Andrew Jeffery <andrew@aj.id.au>
Message-id: 20180209085755.30414-2-f4bug@amsat.org
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/stefanha/tags/block-pull-request' into staging

Pull request

v2:
 * Dropped Fam's git-publish series because there is still ongoing discussion

# gpg: Signature made Thu 15 Feb 2018 09:42:03 GMT
# gpg:                using RSA key 9CA4ABB381AB73C8
# gpg: Good signature from "Stefan Hajnoczi <stefanha@redhat.com>"
# gpg:                 aka "Stefan Hajnoczi <stefanha@gmail.com>"
# Primary key fingerprint: 8695 A8BF D3F9 7CDA AC35  775A 9CA4 ABB3 81AB 73C8

* remotes/stefanha/tags/block-pull-request:
  misc: fix spelling
  ratelimit: don't align wait time with slices
  vl: pause vcpus before stopping iothreads

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/huth/tags/pull-request-2018-02-14' into staging

Various improvements to the qtest checks:
- Clean-ups by Eric Blake with regards to the global_qtest variable
- Some more test cases for the boot-serial tester
- Re-activation of the m48t59-test

# gpg: Signature made Wed 14 Feb 2018 11:07:44 GMT
# gpg:                using RSA key 2ED9D774FE702DB5
# gpg: Good signature from "Thomas Huth <th.huth@gmx.de>"
# gpg:                 aka "Thomas Huth <thuth@redhat.com>"
# gpg:                 aka "Thomas Huth <huth@tuxfamily.org>"
# gpg:                 aka "Thomas Huth <th.huth@posteo.de>"
# Primary key fingerprint: 27B8 8847 EEE0 2501 18F3  EAB9 2ED9 D774 FE70 2DB5

* remotes/huth/tags/pull-request-2018-02-14:
  tests/m48t59: Use the m48t59 test on ppc, too
  tests/Makefile: Derive check-qtest-ppc64-y from check-qtest-ppc-y
  tests/m48t59: Make the test independent of global_qtest
  tests/m48t59: Fix and re-enable the test for sparc
  tests/boot-serial-test: Add support for the aarch64 virt machine
  tests/boot-serial: Add tests for PowerPC Mac machines
  tests/boot-serial: Enable the boot-serial test on SPARC machines, too
  wdt_ib700-test: Drop dependence on global_qtest
  tests/boot-sector: Drop dependence on global_qtest
  qmp-test: Drop dependence on global_qtest
  libqos: Use explicit QTestState for remaining libqos operations
  libqos: Use explicit QTestState for ahci operations
  libqos: Use explicit QTestState for i2c operations
  libqos: Use explicit QTestState for rtas operations
  libqos: Use explicit QTestState for fw_cfg operations
  libqos: Track QTestState with QPCIBus
  libqtest: Use qemu_strtoul()
  tests: Clean up wait for event

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/vivier/tags/m68k-for-2.12-pull-request' into staging

# gpg: Signature made Wed 14 Feb 2018 10:37:02 GMT
# gpg:                using RSA key F30C38BD3F2FBE3C
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>"
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>"
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>"
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier/tags/m68k-for-2.12-pull-request:
  m68k: implement movep instruction

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

misc: fix spelling

s/pupulate/populate

Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20180208162447.10851-1-marcandre.lureau@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

ratelimit: don't align wait time with slices

It is possible for rate limited writes to keep overshooting a slice's
quota by a tiny amount causing the slice-aligned waiting period to
effectively halve the rate.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Message-id: 20180207071758.6818-1-w.bumiller@proxmox.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

vl: pause vcpus before stopping iothreads

Commit dce8921b2baaf95974af8176406881872067adfa ("iothread: Stop threads
before main() quits") introduced iothread_stop_all() to avoid the
following virtio-scsi assertion failure:

  assert(blk_get_aio_context(d->conf.blk) == s->ctx);

Back then the assertion failed because when bdrv_close_all() made
d->conf.blk NULL, blk_get_aio_context() returned the global AioContext
instead of s->ctx.

The same assertion can still fail today when vcpus submit new I/O
requests after iothread_stop_all() has moved the BDS to the global
AioContext.

This patch hardens the iothread_stop_all() approach by pausing vcpus
before calling iothread_stop_all().

Note that the assertion failure is a race condition.  It is not possible
to reproduce it reliably.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-id: 20180201110708.8080-1-stefanha@redhat.com
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>

tests/m48t59: Use the m48t59 test on ppc, too

The ref405ep machine has a memory-mapped m48t59 device, so
we can run the m48t59 test on this machine, too.

Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/Makefile: Derive check-qtest-ppc64-y from check-qtest-ppc-y

ppc64 is a superset of ppc, so the ppc64 tests should include all
the ppc tests.

Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/m48t59: Make the test independent of global_qtest

Stop using the functions that require global_qtest here and pass
around the QTestState instead (global_qtest should finally get
removed since this causes problems with tests running in parallel).

Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/m48t59: Fix and re-enable the test for sparc

The m48t59 test has been disabled in commit baeddded5fe6fa37d13fb94bf8d
("sparc: disable qtest in make check"), likely due to some timing issues
in the bcd_check_time tests which might fail if it gets interrupted for
too long. It should be OK to re-enable this test if we make sure that we
do not run it on timing-sensitive machines, thus it should be OK if we only
run it in the g_test_slow() mode.

Additionally, there are two other issues:

First, the test can not run so easily on sparc64 anymore, since commit
f3b18f35a23c60edbda6420cd ("sun4u: switch m48t59 NVRAM to MMIO access")
moved the m48t59 device to the ebus instead, and for this you first
have to set up the corresponding PCI device (which is currently not
possible from within the m48t59 test). So we can only re-enable this
test on sparc, but not the sparc64 target.

Second, the fuzzing test is executed before the bcd-check-time test
(due to the naming of the tests), without having the base address set
up properly, so the fuzzing test does not really check anything at all.
Fix it by setting up the base address from the main function already
and by moving the qtest_start() to the tests themselves, so that each
test starts with a clean environment (since after the fuzzing, the clock
is unusable for the bcd-check-time test).

Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/boot-serial-test: Add support for the aarch64 virt machine

This patch adds a small binary kernel to test aarch64 virt machine's
UART.

Signed-off-by: Wei Huang <wei@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
[thuth: Fixed contextual conflicts with the hppa and sdhci patches]
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/boot-serial: Add tests for PowerPC Mac machines

OpenBIOS prints out the CPU type on these machine types, so we can use
this string to test whether the CPU detection is working correctly.

Acked-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/boot-serial: Enable the boot-serial test on SPARC machines, too

OpenBIOS prints out the name of the detected CPU here, so looking for
this string is a nice test to verify that the CPU detection is still
working correctly.

Acked-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

wdt_ib700-test: Drop dependence on global_qtest

As a general rule, we prefer avoiding implicit global state
because it makes code harder to safely copy and paste without
thinking about the global state.  Improve this test to be
explicit about the state.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests/boot-sector: Drop dependence on global_qtest

As a general rule, we prefer avoiding implicit global state
because it makes code harder to safely copy and paste without
thinking about the global state.  Adjust the helper code to
use explicit state instead, and update all callers.

Fix some trailing whitespace while touching the file.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

qmp-test: Drop dependence on global_qtest

As a general rule, we prefer avoiding implicit global state
because it makes code harder to safely copy and paste without
thinking about the global state.  Although qmp-test does not
maintain parallel qtest connections, it was the last test
assigning to global_qtest.  It's just as easy to be explicit
about the state; once all tests have been cleaned up, a later
patch can then get rid of global_qtest and a layer of wrappers
in libqtest.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

libqos: Use explicit QTestState for remaining libqos operations

Drop one more client of global_qtest by teaching all remaining
libqos stragglers to pass in an explicit QTestState.  Change the
setting of global_qtest from being implicit in libqos' call to
qtest_start() to instead be explicit in all clients that are
still relying on global_qtest.

Note that qmp_execute() can be greatly simplified in the process,
and that we also get rid of interpolation of a JSON string into a
temporary variable when qtest_qmp() can do it more reliably.

Signed-off-by: Eric Blake <eblake@redhat.com>
Acked-by: Greg Kurz <groug@kaod.org>
Reviewed-by: John Snow <jsnow@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

libqos: Use explicit QTestState for ahci operations

Drop one more client of global_qtest by teaching all ahci test
functionality to pass in an explicit QTestState.  The state was
already available, so no callers had to be adjusted.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

libqos: Use explicit QTestState for i2c operations

Drop one more client of global_qtest by teaching all i2c test
functionality to pass in an explicit QTestState, adjusting all
callers.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

libqos: Use explicit QTestState for rtas operations

Drop one more client of global_qtest by teaching all rtas test
functionality to pass in an explicit QTestState, adjusting all
callers.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
[thuth: Use nicer indentation in rtas.h]
Signed-off-by: Thomas Huth <thuth@redhat.com>

libqos: Use explicit QTestState for fw_cfg operations

Drop one more client of global_qtest by teaching all fw_cfg test
functionality (invoked through alloc-pc) to pass in an explicit
QTestState, adjusting all callers.  In particular, fw_cfg-test
had to reorder things to create the test state prior to creating
the fw_cfg (and drop a pointless strdup in the meantime), but that
test now no longer depends on global_qtest.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
[thuth: Fixed conflict wrt pc_alloc_init() in vhost-user-test.c]
Signed-off-by: Thomas Huth <thuth@redhat.com>

libqos: Track QTestState with QPCIBus

When initializing a QPCIBus, track which QTestState the bus is
associated with (so that a later patch can then explicitly use
that test state for all communication on the bus, rather than
blindly relying on global_qtest).  Update the initialization
functions to take another parameter, and update all callers to
pass in state (for now, most callers get away with passing the
current global_qtest as the current state, although this required
fixing the order of initialization to ensure qtest_start() is
called before qpci_init*() in rtl8139-test, and provided an
opportunity to pass in the allocator in e1000e-test).

Touch up some allocations to use g_new0() rather than g_malloc()
while in the area, and simplify some code (all implementations
of QOSOps provide a .init_allocator() that never fails).

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: John Snow <jsnow@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
[thuth: Removed hunk from vhost-user-test.c that is not required anymore,
 fixed conflict in qtest_vboot() and adjusted qpci_init_pc() in sdhci-test]
Signed-off-by: Thomas Huth <thuth@redhat.com>

m68k: implement movep instruction

This patch implements movep instruction. It moves data between a data register
and alternate bytes within the address space starting at the location
specified and incrementing by two.

It was designed for the original 68000 and used in firmwares for
interfacing the 8-bit peripherals through the 16-bit data bus.
Without this patch opcode for this instruction is recognized as some bitop.

Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Signed-off-by: Mihail Abakumov <mikhail.abakumov@ispras.ru>
Tested-by: Laurent Vivier <laurent@vivier.eu>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20180206124431.31433.91946.stgit@pasha-VirtualBox>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

libqtest: Use qemu_strtoul()

This will keep checkpatch happy when the next patch does code motion.
Fix the include order to match HACKING when adding the needed header.

Signed-off-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>

tests: Clean up wait for event

We still use hacks like qmp("") to wait for an event, even though we
have qmp_eventwait() since commit 8fe941f, and qmp_eventwait_ref()
since commit 7ffe312.  Both commits neglected to convert all the
existing hacks.  Make up what they missed.

Bonus: gets rid of empty format strings.  A step towards compile-time
format string checking without triggering -Wformat-zero-length.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Signed-off-by: Eric Blake <eblake@redhat.com>
[thuth: dropped the hunks from the usb tests - not needed anymore]
Signed-off-by: Thomas Huth <thuth@redhat.com>

Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block layer patches

# gpg: Signature made Tue 13 Feb 2018 17:03:11 GMT
# gpg:                using RSA key 7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>"
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream: (55 commits)
  iotests: Add l2-cache-entry-size to iotest 137
  iotests: Test downgrading an image using a small L2 slice size
  iotests: Test valid values of l2-cache-entry-size
  qcow2: Allow configuring the L2 slice size
  qcow2: Rename l2_table in count_cow_clusters()
  qcow2: Rename l2_table in count_contiguous_clusters_unallocated()
  qcow2: Rename l2_table in count_contiguous_clusters()
  qcow2: Rename l2_table in qcow2_alloc_compressed_cluster_offset()
  qcow2: Update qcow2_truncate() to support L2 slices
  qcow2: Update expand_zero_clusters_in_l1() to support L2 slices
  qcow2: Prepare expand_zero_clusters_in_l1() for adding L2 slice support
  qcow2: Read refcount before L2 table in expand_zero_clusters_in_l1()
  qcow2: Update qcow2_update_snapshot_refcount() to support L2 slices
  qcow2: Prepare qcow2_update_snapshot_refcount() for adding L2 slice support
  qcow2: Update zero_single_l2() to support L2 slices
  qcow2: Update discard_single_l2() to support L2 slices
  qcow2: Update handle_alloc() to support L2 slices
  qcow2: Update handle_copied() to support L2 slices
  qcow2: Update qcow2_alloc_cluster_link_l2() to support L2 slices
  qcow2: Update qcow2_get_cluster_offset() to support L2 slices
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging

* CAN bus (will be under network maintainner)
* scsi-block opblockers (myself)
* Dirty log bitmap cleanup (myself)
* SDHCI improvements and tests (Philippe)
* HAX support for larger guest sizese (Yu Ning)

# gpg: Signature made Tue 13 Feb 2018 15:37:14 GMT
# gpg:                using RSA key BFFBD25F78C7AE83
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>"
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>"
# Primary key fingerprint: 46F5 9FBD 57D6 12E7 BFD4  E2F7 7E15 100C CD36 69B1
#      Subkey fingerprint: F133 3857 4B66 2389 866C  7682 BFFB D25F 78C7 AE83

* remotes/bonzini/tags/for-upstream: (48 commits)
  travis: use libgcc-4.8-dev (libgcc-6-dev is not available on Ubuntu 14.04)
  memory: unify loops to sync dirty log bitmap
  memory: hide memory_region_sync_dirty_bitmap behind DirtyBitmapSnapshot
  memory: remove memory_region_test_and_clear_dirty
  g364fb: switch to using DirtyBitmapSnapshot
  sdhci: add Spec v4.2 register definitions
  sdhci: add a check_capab_v3() qtest
  sdhci: check Spec v3 capabilities qtest
  hw/arm/xilinx_zynqmp: enable the UHS-I mode
  hw/arm/xilinx_zynqmp: fix the capabilities/spec version to match the datasheet
  hw/arm/fsl-imx6: implement SDHCI Spec. v3
  hw/arm/bcm2835_peripherals: change maximum block size to 1kB
  hw/arm/bcm2835_peripherals: implement SDHCI Spec v3
  sdhci: implement CMD/DAT[] fields in the Present State register
  sdhci: implement UHS-I voltage switch
  sdbus: add trace events
  sdhci: implement the Host Control 2 register (tuning sequence)
  sdhci: rename the hostctl1 register
  sdhci: add support for v3 capabilities
  hw/arm/xilinx_zynq: fix the capabilities register to match the datasheet
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

Merge remote-tracking branch 'remotes/mst/tags/for_upstream' into staging

virtio,vhost,pci,pc: features, fixes and cleanups

- new stats in virtio balloon
- virtio eventfd rework for boot speedup
- vhost memory rework for boot speedup
- fixes and cleanups all over the place

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Tue 13 Feb 2018 16:29:55 GMT
# gpg:                using RSA key 281F0DB8D28D5469
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>"
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>"
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17  0970 C350 3912 AFBE 8E67
#      Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA  8A0D 281F 0DB8 D28D 5469

* remotes/mst/tags/for_upstream: (22 commits)
  virtio-balloon: include statistics of disk/file caches
  acpi-test: update FADT
  lpc: drop pcie host dependency
  tests: acpi: fix FADT not being compared to reference table
  hw/pci-bridge: fix pcie root port's IO hints capability
  libvhost-user: Support across-memory-boundary access
  libvhost-user: Fix resource leak
  virtio-balloon: unref the memory region before continuing
  pci: removed the is_express field since a uniform interface was inserted
  virtio-blk: enable multiple vectors when using multiple I/O queues
  pci/bus: let it has higher migration priority
  pci-bridge/i82801b11: clear bridge registers on platform reset
  vhost: Move log_dirty check
  vhost: Merge and delete unused callbacks
  vhost: Clean out old vhost_set_memory and friends
  vhost: Regenerate region list from changed sections list
  vhost: Merge sections added to temporary list
  vhost: Simplify ring verification checks
  vhost: Build temporary section list and deref after commit
  virtio: improve virtio devices initialization time
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

virtio-balloon: include statistics of disk/file caches

Signed-off-by: Tomáš Golembiovský <tgolembi@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

acpi-test: update FADT

Previous commit ("tests: acpi: fix FADT not being compared to reference table")
started tracking changes to the FADT. Generate the expected FACP files -
apparently these weren't updated since 2013.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

lpc: drop pcie host dependency

Doesn't look like that header is used.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

tests: acpi: fix FADT not being compared to reference table

It turns out that FADT isn't actually tested for changes
against reference table, since it happens to be the 1st
table in RSDT which is currently ignored.
Fix it by making sure that all tables from RSDT are added
to test list.

NOTE: FADT contains guest allocated pointers to FACS/DSDT,
zero them out so that possible FACS/DSDT address change
won't affect test results.

Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

hw/pci-bridge: fix pcie root port's IO hints capability

The gen_pcie_root_port mem-reserve and pref32-reserve properties are
defined as size (so uint64_t), but passed as uint32_t when building
the 'IO hints' vendor specific capability.
Passing 4G (or more) gets truncated and passed as a zero reservation.
Is not a huge issue since the guest firmware will always compare the
hints with the default value and take the maximum.

Fix it by passing the values as uint64_t and failing to init the
gen_pcie_root_port id invalid values are used.

Signed-off-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

libvhost-user: Support across-memory-boundary access

The sg list/indirect descriptor table may be contigious
in GPA but not in HVA address space. But libvhost-user
wasn't aware of that. This would cause out-of-bounds
access. Even a malicious guest could use it to get
information from the vhost-user backend.

Introduce a plen parameter in vu_gpa_to_va() so we can
handle this case, returning the actual mapped length.

Signed-off-by: Yongji Xie <xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Maxime Coquelin <maxime.coquelin@redhat.com>

libvhost-user: Fix resource leak

Free the mmaped memory when we need to mmap new memory
space on vu_set_mem_table_exec() and vu_set_log_base_exec() to
avoid memory leak.

Also close the corresponding fd after mmap() on
vu_set_log_base_exec() to avoid fd leak.

Signed-off-by: Yongji Xie <xieyongji@baidu.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>

virtio-balloon: unref the memory region before continuing

Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
Cc: qemu-stable@nongnu.org
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

Merge remote-tracking branch 'mreitz/tags/pull-block-2018-02-13' into queue-block

Block patches for the block queue

# gpg: Signature made Tue Feb 13 17:00:13 2018 CET
# gpg:                using RSA key F407DB0061D5CF40
# gpg: Good signature from "Max Reitz <mreitz@redhat.com>"
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1  1829 F407 DB00 61D5 CF40

* mreitz/tags/pull-block-2018-02-13: (40 commits)
  iotests: Add l2-cache-entry-size to iotest 137
  iotests: Test downgrading an image using a small L2 slice size
  iotests: Test valid values of l2-cache-entry-size
  qcow2: Allow configuring the L2 slice size
  qcow2: Rename l2_table in count_cow_clusters()
  qcow2: Rename l2_table in count_contiguous_clusters_unallocated()
  qcow2: Rename l2_table in count_contiguous_clusters()
  qcow2: Rename l2_table in qcow2_alloc_compressed_cluster_offset()
  qcow2: Update qcow2_truncate() to support L2 slices
  qcow2: Update expand_zero_clusters_in_l1() to support L2 slices
  qcow2: Prepare expand_zero_clusters_in_l1() for adding L2 slice support
  qcow2: Read refcount before L2 table in expand_zero_clusters_in_l1()
  qcow2: Update qcow2_update_snapshot_refcount() to support L2 slices
  qcow2: Prepare qcow2_update_snapshot_refcount() for adding L2 slice support
  qcow2: Update zero_single_l2() to support L2 slices
  qcow2: Update discard_single_l2() to support L2 slices
  qcow2: Update handle_alloc() to support L2 slices
  qcow2: Update handle_copied() to support L2 slices
  qcow2: Update qcow2_alloc_cluster_link_l2() to support L2 slices
  qcow2: Update qcow2_get_cluster_offset() to support L2 slices
  ...

Signed-off-by: Kevin Wolf <kwolf@redhat.com>

iotests: Add l2-cache-entry-size to iotest 137

This test tries reopening a qcow2 image with valid and invalid
options. This patch adds l2-cache-entry-size to the set.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 3d3b7d2dbfc020deaef60fb58739b0801eb9517c.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Test downgrading an image using a small L2 slice size

expand_zero_clusters_in_l1() is used when downgrading qcow2 images
from v3 to v2 (compat=0.10). This is one of the functions that needed
more changes to support L2 slices, so this patch extends iotest 061 to
test downgrading a qcow2 image using a smaller slice size.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 3e5662dce5e4926c8fabbad4c0b9142b2a506dd4.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

iotests: Test valid values of l2-cache-entry-size

The l2-cache-entry-size setting can only contain values that are
powers of two between 512 and the cluster size.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: bd3547b670b8d0af11480c760991a22bcae5b48c.1517840877.git.berto@igalia.com
[mreitz: Changed non-power-of-two test value from 300 to 4242]
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Allow configuring the L2 slice size

Now that the code is ready to handle L2 slices we can finally add an
option to allow configuring their size.

An L2 slice is the portion of an L2 table that is read by the qcow2
cache. Until now the cache was always reading full L2 tables, and
since the L2 table size is equal to the cluster size this was not very
efficient with large clusters. Here's a more detailed explanation of
why it makes sense to have smaller cache entries in order to load L2
data:

   https://lists.gnu.org/archive/html/qemu-block/2017-09/msg00635.html

This patch introduces a new command-line option to the qcow2 driver
named l2-cache-entry-size (cf. l2-cache-size). The cache entry size
has the same restrictions as the cluster size: it must be a power of
two and it has the same range of allowed values, with the additional
requirement that it must not be larger than the cluster size.

The L2 cache entry size (L2 slice size) remains equal to the cluster
size for now by default, so this feature must be explicitly enabled.
Although my tests show that 4KB slices consistently improve
performance and give the best results, let's wait and make more tests
with different cluster sizes before deciding on an optimal default.

Now that the cache entry size is not necessarily equal to the cluster
size we need to reflect that in the MIN_L2_CACHE_SIZE documentation.
That minimum value is a requirement of the COW algorithm: we need to
read two L2 slices (and not two L2 tables) in order to do COW, see
l2_allocate() for the actual code.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: c73e5611ff4a9ec5d20de68a6c289553a13d2354.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Rename l2_table in count_cow_clusters()

This function doesn't need any changes to support L2 slices, but since
it's now dealing with slices intead of full tables, the l2_table
variable is renamed for clarity.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 6107001fc79e6739242f1de7d191375e4f130aac.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Rename l2_table in count_contiguous_clusters_unallocated()

This function doesn't need any changes to support L2 slices, but since
it's now dealing with slices instead of full tables, the l2_table
variable is renamed for clarity.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 78bcc54bc632574dd0b900a77a00a1b6ffc359e6.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Rename l2_table in count_contiguous_clusters()

This function doesn't need any changes to support L2 slices, but since
it's now dealing with slices intead of full tables, the l2_table
variable is renamed for clarity.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 812b0c3505bb1687e51285dccf1a94f0cecb1f74.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Rename l2_table in qcow2_alloc_compressed_cluster_offset()

This function doesn't need any changes to support L2 slices, but since
it's now dealing with slices instead of full tables, the l2_table
variable is renamed for clarity.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 0c5d4b9bf163aa3b49ec19cc512a50d83563f2ad.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update qcow2_truncate() to support L2 slices

The qcow2_truncate() code is mostly independent from whether
we're using L2 slices or full L2 tables, but in full and
falloc preallocation modes new L2 tables are allocated using
qcow2_alloc_cluster_link_l2().  Therefore the code needs to be
modified to ensure that all nb_clusters that are processed in each
call can be allocated with just one L2 slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 1fd7d272b5e7b66254a090b74cf2bed1cc334c0e.1517840877.git.berto@igalia.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update expand_zero_clusters_in_l1() to support L2 slices

expand_zero_clusters_in_l1() expands zero clusters as a necessary step
to downgrade qcow2 images to a version that doesn't support metadata
zero clusters. This function takes an L1 table (which may or may not
be active) and iterates over all its L2 tables looking for zero
clusters.

Since we'll be loading L2 slices instead of full tables we need to add
an extra loop that iterates over all slices of each L2 table, and we
should also use the slice size when allocating the buffer used when
the L1 table is not active.

This function doesn't need any additional changes so apart from that
this patch simply updates the variable name from l2_table to l2_slice.

Finally, and since we have to touch the bdrv_read() / bdrv_write()
calls anyway, this patch takes the opportunity to replace them with
the byte-based bdrv_pread() / bdrv_pwrite().

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 43590976f730501688096cff103f2923b72b0f32.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Prepare expand_zero_clusters_in_l1() for adding L2 slice support

Adding support for L2 slices to expand_zero_clusters_in_l1() needs
(among other things) an extra loop that iterates over all slices of
each L2 table.

Putting all changes in one patch would make it hard to read because
all semantic changes would be mixed with pure indentation changes.

To make things easier this patch simply creates a new block and
changes the indentation of all lines of code inside it. Thus, all
modifications in this patch are cosmetic. There are no semantic
changes and no variables are renamed yet. The next patch will take
care of that.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: c2ae9f31ed5b6e591477ad4654448badd1c89d73.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Read refcount before L2 table in expand_zero_clusters_in_l1()

At the moment it doesn't really make a difference whether we call
qcow2_get_refcount() before of after reading the L2 table, but if we
want to support L2 slices we'll need to read the refcount first.

This patch simply changes the order of those two operations to prepare
for that. The patch with the actual semantic changes will be easier to
read because of this.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 947a91d934053a2dbfef979aeb9568f57ef57c5d.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update qcow2_update_snapshot_refcount() to support L2 slices

qcow2_update_snapshot_refcount() increases the refcount of all
clusters of a given snapshot. In order to do that it needs to load all
its L2 tables and iterate over their entries. Since we'll be loading
L2 slices instead of full tables we need to add an extra loop that
iterates over all slices of each L2 table.

This function doesn't need any additional changes so apart from that
this patch simply updates the variable name from l2_table to l2_slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 5f4db199b9637f4833b58487135124d70add8cf0.1517840877.git.berto@igalia.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Prepare qcow2_update_snapshot_refcount() for adding L2 slice support

Adding support for L2 slices to qcow2_update_snapshot_refcount() needs
(among other things) an extra loop that iterates over all slices of
each L2 table.

Putting all changes in one patch would make it hard to read because
all semantic changes would be mixed with pure indentation changes.

To make things easier this patch simply creates a new block and
changes the indentation of all lines of code inside it. Thus, all
modifications in this patch are cosmetic. There are no semantic
changes and no variables are renamed yet. The next patch will take
care of that.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 8ffaa5e55bd51121f80e498f4045b64902a94293.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update zero_single_l2() to support L2 slices

zero_single_l2() limits the number of clusters to be zeroed to the
amount that fits inside an L2 table. Since we'll be loading L2 slices
instead of full tables we need to update that limit. The function is
renamed to zero_in_l2_slice() for clarity.

Apart from that, this function doesn't need any additional changes, so
this patch simply updates the variable name from l2_table to l2_slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: ebc16e7e79fa6969d8975ef487d679794de4fbcc.1517840877.git.berto@igalia.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update discard_single_l2() to support L2 slices

discard_single_l2() limits the number of clusters to be discarded
to the amount that fits inside an L2 table. Since we'll be loading
L2 slices instead of full tables we need to update that limit. The
function is renamed to discard_in_l2_slice() for clarity.

Apart from that, this function doesn't need any additional changes, so
this patch simply updates the variable name from l2_table to l2_slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-id: 1cb44a5b68be5334cb01b97a3db3a3c5a43396e5.1517840877.git.berto@igalia.com
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update handle_alloc() to support L2 slices

handle_alloc() loads an L2 table and limits the number of checked
clusters to the amount that fits inside that table. Since we'll be
loading L2 slices instead of full tables we need to update that limit.

Apart from that, this function doesn't need any additional changes, so
this patch simply updates the variable name from l2_table to l2_slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: b243299c7136f7014c5af51665431ddbf5e99afd.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update handle_copied() to support L2 slices

handle_copied() loads an L2 table and limits the number of checked
clusters to the amount that fits inside that table. Since we'll be
loading L2 slices instead of full tables we need to update that limit.

Apart from that, this function doesn't need any additional changes, so
this patch simply updates the variable name from l2_table to l2_slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 541ac001a7d6b86bab2392554bee53c2b312148c.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update qcow2_alloc_cluster_link_l2() to support L2 slices

There's a loop in this function that iterates over the L2 entries in a
table, so now we need to assert that it remains within the limits of
an L2 slice.

Apart from that, this function doesn't need any additional changes, so
this patch simply updates the variable name from l2_table to l2_slice.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: f9846a1c2efc51938e877e2a25852d9ab14797ff.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update qcow2_get_cluster_offset() to support L2 slices

qcow2_get_cluster_offset() checks how many contiguous bytes are
available at a given offset. The returned number of bytes is limited
by the amount that can be addressed without having to load more than
one L2 table.

Since we'll be loading L2 slices instead of full tables this patch
changes the limit accordingly using the size of the L2 slice for the
calculations instead of the full table size.

One consequence of this is that with small L2 slices operations such
as 'qemu-img map' will need to iterate in more steps because each
qcow2_get_cluster_offset() call will potentially return a smaller
number. However the code is already prepared for that so this doesn't
break semantics.

The l2_table variable is also renamed to l2_slice to reflect this, and
offset_to_l2_index() is replaced with offset_to_l2_slice_index().

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 6b602260acb33da56ed6af9611731cb7acd110eb.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update get_cluster_table() to support L2 slices

This patch updates get_cluster_table() to return L2 slices instead of
full L2 tables.

The code itself needs almost no changes, it only needs to call
offset_to_l2_slice_index() instead of offset_to_l2_index(). This patch
also renames all the relevant variables and the documentation.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 64cf064c0021ba315d3f3032da0f95db1b615f33.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Refactor get_cluster_table()

After the previous patch we're now always using l2_load() in
get_cluster_table() regardless of whether a new L2 table has to be
allocated or not.

This patch refactors that part of the code to use one single l2_load()
call.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: ce31758c4a1fadccea7a6ccb93951eb01d95fd4c.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update l2_allocate() to support L2 slices

This patch updates l2_allocate() to support the qcow2 cache returning
L2 slices instead of full L2 tables.

The old code simply gets an L2 table from the cache and initializes it
with zeroes or with the contents of an existing table. With a cache
that returns slices instead of tables the idea remains the same, but
the code must now iterate over all the slices that are contained in an
L2 table.

Since now we're operating with slices the function can no longer
return the newly-allocated table, so it's up to the caller to retrieve
the appropriate L2 slice after calling l2_allocate() (note that with
this patch the caller is still loading full L2 tables, but we'll deal
with that in a separate patch).

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Message-id: 20fc0415bf0e011e29f6487ec86eb06a11f37445.1517840877.git.berto@igalia.com
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Prepare l2_allocate() for adding L2 slice support

Adding support for L2 slices to l2_allocate() needs (among other
things) an extra loop that iterates over all slices of a new L2 table.

Putting all changes in one patch would make it hard to read because
all semantic changes would be mixed with pure indentation changes.

To make things easier this patch simply creates a new block and
changes the indentation of all lines of code inside it. Thus, all
modifications in this patch are cosmetic. There are no semantic
changes and no variables are renamed yet. The next patch will take
care of that.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: d0d7dca8520db304524f52f49d8157595a707a35.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Update l2_load() to support L2 slices

Each entry in the qcow2 L2 cache stores a full L2 table (which uses a
complete cluster in the qcow2 image). A cluster is usually too large
to be used efficiently as the size for a cache entry, so we want to
decouple both values by allowing smaller cache entries. Therefore the
qcow2 L2 cache will no longer return full L2 tables but slices
instead.

This patch updates l2_load() so it can handle L2 slices correctly.
Apart from the offset of the L2 table (which we already had) we also
need the guest offset in order to calculate which one of the slices
we need.

An L2 slice has currently the same size as an L2 table (one cluster),
so for now this function will load exactly the same data as before.

This patch also removes a stale comment about the return value being
a pointer to the L2 table. This function returns an error code since
55c17e9821c474d5fcdebdc82ed2fc096777d611.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: b830aa1fc5b6f8e3cb331d006853fe22facca847.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Add offset_to_l2_slice_index()

Similar to offset_to_l2_index(), this function takes a guest offset
and returns the index in the L2 slice that contains its L2 entry.

An L2 slice has currently the same size as an L2 table (one cluster),
so both functions return the same value for now.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: a1c45c5c5a76146dd1712d8d1e7b409ad539c718.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Add l2_slice_size field to BDRVQcow2State

The BDRVQcow2State structure contains an l2_size field, which stores
the number of 64-bit entries in an L2 table.

For efficiency reasons we want to be able to load slices instead of
full L2 tables, so we need to know how many entries an L2 slice can
hold.

An L2 slice is the portion of an L2 table that is loaded by the qcow2
cache. At the moment that cache can only load complete tables,
therefore an L2 slice has the same size as an L2 table (one cluster)
and l2_size == l2_slice_size.

Later we'll allow smaller slices, but until then we have to use this
new l2_slice_size field to make the rest of the code ready for that.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: adb048595f9fb5dfb110c802a8b3c3be3b937f37.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Add offset_to_l1_index()

Similar to offset_to_l2_index(), this function returns the index in
the L1 table for a given guest offset. This is only used in a couple
of places and it's not a particularly complex calculation, but it
makes the code a bit more readable.

Although in the qcow2_get_cluster_offset() case the old code was
taking advantage of the l1_bits variable, we're going to get rid of
the other uses of l1_bits in a later patch anyway, so it doesn't make
sense to keep it just for this.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: a5f626fed526b7459a0425fad06d823d18df8522.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_is_table_offset()

This function was only using the BlockDriverState parameter to pass it
to qcow2_cache_get_table_addr(). This is no longer necessary so this
parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: eb0ed90affcf302e5a954bafb5931b5215483d3a.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_discard()

This function was only using the BlockDriverState parameter to pass it
to qcow2_cache_get_table_idx() and qcow2_cache_table_release(). This
is no longer necessary so this parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 9724f7e38e763ad3be32627c6b7fe8df9edb1476.1517840877.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_clean_unused()

This function was only using the BlockDriverState parameter to pass it
to qcow2_cache_table_release(). This is no longer necessary so this
parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: b74f17591af52f201de0ea3a3b2dd0a81932334d.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_destroy()

This function was never using the BlockDriverState parameter so it can
be safely removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 49c74fe8b3aead9056e61a85b145ce787d06262b.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_put()

This function was only using the BlockDriverState parameter to pass it
to qcow2_cache_get_table_idx(). This is no longer necessary so this
parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 6f98155489054a457563da77cdad1a66ebb3e896.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_entry_mark_dirty()

This function was only using the BlockDriverState parameter to pass it
to qcow2_cache_get_table_idx(). This is no longer necessary so this
parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 5c40516a91782b083c1428b7b6a41bb9e2679bfb.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_table_release()

This function was only using the BlockDriverState parameter to get the
cache table size (since it was equal to the cluster size). This is no
longer necessary so this parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 7c1b262344375d52544525f85bbbf0548d5ba575.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_get_table_idx()

This function was only using the BlockDriverState parameter to get the
cache table size (since it was equal to the cluster size). This is no
longer necessary so this parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: da3575d47c9a181a2cfd4715e53dd84a2c651017.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Remove BDS parameter from qcow2_cache_get_table_addr()

This function was only using the BlockDriverState parameter to get the
cache table size (since it was equal to the cluster size). This is no
longer necessary so this parameter can be removed.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: e1f943a9e89e1deb876f45de1bb22419ccdb6ad3.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Add table size field to Qcow2Cache

The table size in the qcow2 cache is currently equal to the cluster
size. This doesn't allow us to use the cache memory efficiently,
particularly with large cluster sizes, so we need to be able to have
smaller cache tables that are independent from the cluster size. This
patch adds a new field to Qcow2Cache that we can use instead of the
cluster size.

The current table size is still being initialized to the cluster size,
so there are no semantic changes yet, but this patch will allow us to
prepare the rest of the code and simplify a few function calls.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: 67a1bf9e55f417005c567bead95a018dc34bc687.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

qcow2: Fix documentation of get_cluster_table()

This function has not been returning the offset of the L2 table since
commit 3948d1d4876065160583e79533bf604481063833

Signed-off-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Message-id: b498733b6706a859a03678d74ecbd26aeba129aa.1517840876.git.berto@igalia.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

block: maintain persistent disabled bitmaps

To maintain load/store disabled bitmap there is new approach:

 - deprecate @autoload flag of block-dirty-bitmap-add, make it ignored
 - store enabled bitmaps as "auto" to qcow2
 - store disabled bitmaps without "auto" flag to qcow2
 - on qcow2 open load "auto" bitmaps as enabled and others
   as disabled (except in_use bitmaps)

Also, adjust iotests 165 and 176 appropriately.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-id: 20180202160752.143796-1-vsementsov@virtuozzo.com
Signed-off-by: Max Reitz <mreitz@redhat.com>

sheepdog: Allow fully preallocated truncation

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

sheepdog: Pass old and new size to sd_prealloc()

sd_prealloc() will now preallocate the area [old_size, new_size).  As
before, it rounds to buf_size and may thus overshoot and preallocate
areas that were not requested to be preallocated.  For image creation,
this is no change in behavior.  For truncation, this is in accordance
with the documentation for preallocated truncation.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

sheepdog: Make sd_prealloc() take a BDS

We want to use this function in sd_truncate() later on, so taking a
filename is not exactly ideal.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

gluster: Add preallocated truncation

By using qemu_do_cluster_truncate() in qemu_cluster_truncate(), we now
automatically have preallocated truncation.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

gluster: Query current size in do_truncate()

Instead of expecting the current size to be 0, query it and allocate
only the area [current_size, offset) if preallocation is requested.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

gluster: Pull truncation from qemu_gluster_create

Pull out the truncation code from the qemu_cluster_create() function so
we can later reuse it in qemu_gluster_truncate().

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

gluster: Move glfs_close() to create's clean-up

glfs_close() is a classical clean-up operation, as can be seen by the
fact that it is executed even if the truncation before it failed.
Also, moving it to clean-up makes it more clear that if it fails, we do
not want it to overwrite the current ret value if that signifies an
error already.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

travis: use libgcc-4.8-dev (libgcc-6-dev is not available on Ubuntu 14.04)

Travis image is based on Ubuntu Trusty (14.04), since d83414e1fd1 we get:

  $ sudo -E \
    apt-get -yq --no-install-suggests --no-install-recommends --force-yes \
      install \
        libaio-dev libattr1-dev libbrlapi-dev libcap-ng-dev libgcc-6-dev \
        libgnutls-dev libgtk-3-dev libiscsi-dev liblttng-ust-dev \
        libncurses5-dev libnfs-dev libnss3-dev libpixman-1-dev libpng12-dev \
        librados-dev libsdl1.2-dev libseccomp-dev libspice-protocol-dev \
        libspice-server-dev libssh2-1-dev liburcu-dev libusb-1.0-0-dev \
        libvte-2.90-dev sparse uuid-dev
  Reading package lists...
  Building dependency tree...
  Reading state information...
  E: Unable to locate package libgcc-6-dev

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>

memory: unify loops to sync dirty log bitmap

Now that memory_region_sync_dirty_bitmap is NULL, we can unify its
loop with memory_global_dirty_log_sync's.  The only difference is
that memory_region_sync_dirty_bitmap will no longer call log_sync on
FlatRanges that do have a zero dirty_log_mask, but this is okay because
video memory is always registered with the dirty page logging mechanism.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory: hide memory_region_sync_dirty_bitmap behind DirtyBitmapSnapshot

Simplify the users of memory_region_snapshot_and_clear_dirty, so
that they do not have to call memory_region_sync_dirty_bitmap
explicitly.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

memory: remove memory_region_test_and_clear_dirty

It is unused after g364fb has been converted to use DirtyBitmapSnapshot.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

g364fb: switch to using DirtyBitmapSnapshot

This removes the last user of memory_region_test_and_clear_dirty
outside memory.c.

Tested-by: Hervé Poussineau <hpoussin@reactos.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

sdhci: add Spec v4.2 register definitions

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20180208164818.7961-31-f4bug@amsat.org>

sdhci: add a check_capab_v3() qtest

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20180208164818.7961-30-f4bug@amsat.org>

sdhci: check Spec v3 capabilities qtest

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Acked-by: Alistair Francis <alistair.francis@xilinx.com>
Message-Id: <20180208164818.7961-29-f4bug@amsat.org>