qemu.git
2 days agoMerge remote-tracking branch 'remotes/philmd-gitlab/tags/fw_cfg-20200918' into staging master github/master
Peter Maydell [Fri, 18 Sep 2020 15:34:26 +0000 (16:34 +0100)] 
Merge remote-tracking branch 'remotes/philmd-gitlab/tags/fw_cfg-20200918' into staging

fw_cfg patches

Fixes a bug in the recently introduced fw_cfg_add_from_generator().

CI jobs results:
. https://gitlab.com/philmd/qemu/-/pipelines/191714875
. https://travis-ci.org/github/philmd/qemu/builds/728288784
. https://app.shippable.com/github/philmd/qemu/runs/869/1/console

# gpg: Signature made Fri 18 Sep 2020 16:20:56 BST
# gpg:                using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]
# Primary key fingerprint: FAAB E75E 1291 7221 DCFD  6BB2 E3E3 2C2C DEAD C0DE

* remotes/philmd-gitlab/tags/fw_cfg-20200918:
  hw/nvram/fw_cfg: fix FWCfgDataGeneratorClass::get_data() consumption

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 days agohw/nvram/fw_cfg: fix FWCfgDataGeneratorClass::get_data() consumption
Laszlo Ersek [Wed, 16 Sep 2020 15:15:10 +0000 (17:15 +0200)] 
hw/nvram/fw_cfg: fix FWCfgDataGeneratorClass::get_data() consumption

The documentation on g_byte_array_free()
<https://developer.gnome.org/glib/stable/glib-Byte-Arrays.html#g-byte-array-free>
says:

> Returns
>
> the element data if free_segment is FALSE, otherwise NULL. The element
> data should be freed using g_free().

Because we currently call g_byte_array_free() with free_segment=TRUE, we
end up passing data=NULL to fw_cfg_add_file().

On the plus side, fw_cfg_data_read() and fw_cfg_dma_transfer() both deal
with NULL data gracefully: QEMU does not crash when the guest reads such
an item, the guest just gets a properly sized, but zero-filled blob.

However, the bug breaks UEFI HTTPS boot, as the IANA_TLS_CIPHER array,
generated otherwise correctly by the "tls-cipher-suites" object, is in
effect replaced with a zero blob.

Fix the issue by passing free_segment=FALSE to g_byte_array_free():

- the caller (fw_cfg_add_from_generator()) temporarily assumes ownership
  of the generated byte array,

- then ownership of the byte array is transfered to fw_cfg, as
  fw_cfg_add_file() links (not copies) "data" into fw_cfg.

Cc: "Daniel P. Berrangé" <berrange@redhat.com>
Cc: "Philippe Mathieu-Daudé" <philmd@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Fixes: 3203148917d035b09f71986ac2eaa19a352d6d9d
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200916151510.22767-1-lersek@redhat.com>
Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
2 days agoMerge remote-tracking branch 'remotes/legoater/tags/pull-aspeed-20200918' into staging
Peter Maydell [Fri, 18 Sep 2020 12:36:42 +0000 (13:36 +0100)] 
Merge remote-tracking branch 'remotes/legoater/tags/pull-aspeed-20200918' into staging

Aspeed patches :

* Couple of cleanups
* New machine properties to define the flash models

# gpg: Signature made Fri 18 Sep 2020 08:23:19 BST
# gpg:                using RSA key A0F66548F04895EBFE6B0B6051A343C7CFFBECA1
# gpg: Good signature from "Cédric Le Goater <clg@kaod.org>" [undefined]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: A0F6 6548 F048 95EB FE6B  0B60 51A3 43C7 CFFB ECA1

* remotes/legoater/tags/pull-aspeed-20200918:
  misc: aspeed_scu: Update AST2600 silicon id register
  hw/arm/aspeed: Add machine properties to define the flash models
  hw/arm/aspeed: Map the UART5 device unconditionally

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 days agodocker.py: always use --rm
Paolo Bonzini [Thu, 17 Sep 2020 10:44:41 +0000 (12:44 +0200)] 
docker.py: always use --rm

Avoid that containers pile up.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
2 days agomisc: aspeed_scu: Update AST2600 silicon id register
Joel Stanley [Fri, 18 Sep 2020 07:04:36 +0000 (09:04 +0200)] 
misc: aspeed_scu: Update AST2600 silicon id register

Aspeed have released an updated datasheet (v7) containing the silicon id
for the AST2600 A2. It looks like this:

              SCU004      SCU014
  AST2600-A0  0x05000303  0x05000303
  AST2600-A1  0x05010303  0x05010303
  AST2600-A2  0x05010303  0x05020303
  AST2620-A1  0x05010203  0x05010203
  AST2620-A2  0x05010203  0x05020203

The SCU004 (silicon id 1) value matches SCU014 for A0, but for
subsequent revisions it is hard coded to the A1 value.

Qemu effectively dropped support for the A0 in 7582591ae745 ("aspeed:
Support AST2600A1 silicon revision") as the A0 reset table was removed,
so it makes sense to only support the behaviour of A1 and onwards.

Signed-off-by: Joel Stanley <joel@jms.id.au>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20200916082012.776628-1-joel@jms.id.au>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 days agohw/arm/aspeed: Add machine properties to define the flash models
Cédric Le Goater [Fri, 18 Sep 2020 07:04:36 +0000 (09:04 +0200)] 
hw/arm/aspeed: Add machine properties to define the flash models

Some machines don't have much differences a part from the flash model
being used. Introduce new machine properties to change them from the
command line.

For instance, to start the ast2500-evb machine with a different FMC
chip and a 64M SPI chip, use :

  -M ast2500-evb,fmc-model=mx25l25635e,spi-model=mx66u51235f

Cc: 郁雷 <yulei.sh@bytedance.com>
Reviewed-by: Joel Stanley <joel@jms.id.au>
Tested-by: Lei YU <yulei.sh@bytedance.com>
Message-Id: <20200915054859.2338477-1-clg@kaod.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
2 days agohw/arm/aspeed: Map the UART5 device unconditionally
Philippe Mathieu-Daudé [Fri, 18 Sep 2020 07:04:36 +0000 (09:04 +0200)] 
hw/arm/aspeed: Map the UART5 device unconditionally

The UART5 is present on the machine regardless there is a
character device connected to it. Map it unconditionally.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Cédric Le Goater <clg@kaod.org>
Message-Id: <20200905212415.760452-1-f4bug@amsat.org>
Signed-off-by: Cédric Le Goater <clg@kaod.org>
3 days agoMerge remote-tracking branch 'remotes/kraxel/tags/microvm-20200917-pull-request'...
Peter Maydell [Thu, 17 Sep 2020 19:40:59 +0000 (20:40 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/microvm-20200917-pull-request' into staging

microvm: add acpi support

# gpg: Signature made Thu 17 Sep 2020 14:53:06 BST
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/microvm-20200917-pull-request: (21 commits)
  microvm: enable ramfb
  tests/acpi: update expected data files for microvm
  tests/acpi: add microvm test
  tests/acpi: allow override blkdev
  tests/acpi: allow microvm test data updates.
  microvm: wire up hotplug
  x86: move cpu hotplug from pc to x86
  x86: move acpi_dev from pc/microvm
  x86: constify x86_machine_is_*_enabled
  microvm/acpi: disable virtio-mmio cmdline hack
  microvm/acpi: use seabios with acpi=on
  microvm/acpi: use GSI 16-23 for virtio
  microvm/acpi: add acpi_dsdt_add_virtio() for x86
  microvm/acpi: add minimal acpi support
  microvm: make virtio irq base runtime configurable
  acpi: move acpi_dsdt_add_power_button() to ged
  acpi: ged: add x86 device variant.
  acpi: ged: add control regs
  seabios: add bios-microvm.bin binary
  seabios: add microvm config, update build rules
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 days agoMerge remote-tracking branch 'remotes/stsquad/tags/pull-configure-fixes-160920-1...
Peter Maydell [Thu, 17 Sep 2020 13:50:55 +0000 (14:50 +0100)] 
Merge remote-tracking branch 'remotes/stsquad/tags/pull-configure-fixes-160920-1' into staging

configure tweaks for deprecation

  - iotest fix for readlink -f
  - linux-user, report rather than assert on mmap failure
  - clean-up and re-factor the logic
  - add tilegx-linux-user to deprecated_targets_list
  - add [lm32|unicore32]-softmmu deprecated_targets_list
  - add a gitlab deprecated builds test

# gpg: Signature made Wed 16 Sep 2020 10:11:41 BST
# gpg:                using RSA key 6685AE99E75167BCAFC8DF35FBD0DB095A9E2A44
# gpg: Good signature from "Alex Bennée (Master Work Key) <alex.bennee@linaro.org>" [full]
# Primary key fingerprint: 6685 AE99 E751 67BC AFC8  DF35 FBD0 DB09 5A9E 2A44

* remotes/stsquad/tags/pull-configure-fixes-160920-1:
  configure: add [lm32|unicore32]-softmmu to deprecation logic
  gitlab: create a build-deprecated target
  configure: include tilegx-linux-user in the deprecation logic
  configure: clean-up the target-list-exclude logic
  configure: also skip deprecated targets with target-list-exclude
  configure: move deprecated feature processing to supported_target
  iotests: Drop readlink -f
  linux-user: test, don't assert addr != test in pgb_reserved_va

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 days agoMerge remote-tracking branch 'remotes/huth-gitlab/tags/pull-request-2020-09-16' into...
Peter Maydell [Thu, 17 Sep 2020 12:38:08 +0000 (13:38 +0100)] 
Merge remote-tracking branch 'remotes/huth-gitlab/tags/pull-request-2020-09-16' into staging

* Fix "readlink -f" problem in iotests on macOS (to fix the Cirrus-CI tests)
* Some minor qtest improvements
* Fix the unit tests to work on MSYS2, too
* Enable building and testing on MSYS2 in the Cirrus-CI
* Build FreeBSD with one task again in the Cirrus-CI

# gpg: Signature made Wed 16 Sep 2020 12:24:29 BST
# gpg:                using RSA key 27B88847EEE0250118F3EAB92ED9D774FE702DB5
# gpg:                issuer "thuth@redhat.com"
# gpg: Good signature from "Thomas Huth <th.huth@gmx.de>" [full]
# gpg:                 aka "Thomas Huth <thuth@redhat.com>" [full]
# gpg:                 aka "Thomas Huth <huth@tuxfamily.org>" [full]
# gpg:                 aka "Thomas Huth <th.huth@posteo.de>" [unknown]
# Primary key fingerprint: 27B8 8847 EEE0 2501 18F3  EAB9 2ED9 D774 FE70 2DB5

* remotes/huth-gitlab/tags/pull-request-2020-09-16: (24 commits)
  cirrus: Building freebsd in a single shot
  ci: Enable msys2 ci in cirrus
  tests: Fixes test-qdev-global-props.c
  tests: fix test-util-sockets.c
  tests: Fixes test-io-channel-file by mask only owner file state mask bits
  tests: fixes aio-win32 about aio_remove_fd_handler, get it consistence with aio-posix.c
  tests: Fixes test-io-channel-socket.c tests under msys2/mingw
  vmstate: Fixes test-vmstate.c on msys2/mingw
  meson: remove empty else and duplicated gio deps
  meson: Use -b to ignore CR vs. CR-LF issues on Windows
  osdep: file locking functions are not available on Win32
  tests: test-replication disable /replication/secondary/* on msys2/mingw.
  tests: Fixes test-replication.c on msys2/mingw.
  meson: disable crypto tests are empty under win32
  meson: Disable test-char on msys2/mingw for fixing tests stuck
  rcu: fixes test-logging.c by call drain_call_rcu before rmdir_full
  tests: Convert g_free to g_autofree macro in test-logging.c
  rcu: Implement drain_call_rcu
  qga/commands-win32: Fix problem with redundant protype declaration
  Simplify the .gitignore file
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 days agomicrovm: enable ramfb
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:09 +0000 (14:09 +0200)] 
microvm: enable ramfb

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200915120909.20838-22-kraxel@redhat.com

3 days agotests/acpi: update expected data files for microvm
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:08 +0000 (14:09 +0200)] 
tests/acpi: update expected data files for microvm

Also clear tests/qtest/bios-tables-test-allowed-diff.h

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200915120909.20838-21-kraxel@redhat.com

3 days agotests/acpi: add microvm test
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:07 +0000 (14:09 +0200)] 
tests/acpi: add microvm test

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-20-kraxel@redhat.com

3 days agotests/acpi: allow override blkdev
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:06 +0000 (14:09 +0200)] 
tests/acpi: allow override blkdev

microvm needs virtio-blk instead of ide.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200915120909.20838-19-kraxel@redhat.com

3 days agotests/acpi: allow microvm test data updates.
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:05 +0000 (14:09 +0200)] 
tests/acpi: allow microvm test data updates.

Also add empty test data files.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200915120909.20838-18-kraxel@redhat.com

3 days agomicrovm: wire up hotplug
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:04 +0000 (14:09 +0200)] 
microvm: wire up hotplug

The cpu hotplug code handles the initialization of coldplugged cpus
too, so it is needed even in case cpu hotplug is not supported.

Wire cpu hotplug up for microvm.
Without this we get a broken MADT table.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200915120909.20838-17-kraxel@redhat.com

3 days agox86: move cpu hotplug from pc to x86
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:03 +0000 (14:09 +0200)] 
x86: move cpu hotplug from pc to x86

The cpu hotplug code handles the initialization of coldplugged cpus
too, so it is needed even in case cpu hotplug is not supported.

Move the code from pc to x86, so microvm can use it.

Move both plug and unplug to keep everything in one place, even
though microvm needs plug only.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-16-kraxel@redhat.com

3 days agox86: move acpi_dev from pc/microvm
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:02 +0000 (14:09 +0200)] 
x86: move acpi_dev from pc/microvm

Both pc and microvm machine types have a acpi_dev field.
Move it to the common base type.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-15-kraxel@redhat.com

3 days agox86: constify x86_machine_is_*_enabled
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:01 +0000 (14:09 +0200)] 
x86: constify x86_machine_is_*_enabled

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20200915120909.20838-14-kraxel@redhat.com

3 days agomicrovm/acpi: disable virtio-mmio cmdline hack
Gerd Hoffmann [Tue, 15 Sep 2020 12:09:00 +0000 (14:09 +0200)] 
microvm/acpi: disable virtio-mmio cmdline hack

... in case we are using ACPI.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200915120909.20838-13-kraxel@redhat.com

3 days agomicrovm/acpi: use seabios with acpi=on
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:59 +0000 (14:08 +0200)] 
microvm/acpi: use seabios with acpi=on

With acpi=off continue to use qboot.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200915120909.20838-12-kraxel@redhat.com

3 days agomicrovm/acpi: use GSI 16-23 for virtio
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:58 +0000 (14:08 +0200)] 
microvm/acpi: use GSI 16-23 for virtio

With ACPI enabled and IO-APIC being properly declared in the ACPI tables
we can use interrupt lines 16-23 for virtio and avoid shared interrupts.

With acpi disabled we continue to use lines 5-12.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-11-kraxel@redhat.com

3 days agomicrovm/acpi: add acpi_dsdt_add_virtio() for x86
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:57 +0000 (14:08 +0200)] 
microvm/acpi: add acpi_dsdt_add_virtio() for x86

Makes x86 linux kernel find virtio-mmio devices automatically.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-10-kraxel@redhat.com

3 days agomicrovm/acpi: add minimal acpi support
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:56 +0000 (14:08 +0200)] 
microvm/acpi: add minimal acpi support

$subject says all.  Can be controlled using -M microvm,acpi=on/off.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-9-kraxel@redhat.com

3 days agomicrovm: make virtio irq base runtime configurable
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:55 +0000 (14:08 +0200)] 
microvm: make virtio irq base runtime configurable

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-8-kraxel@redhat.com

3 days agoacpi: move acpi_dsdt_add_power_button() to ged
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:54 +0000 (14:08 +0200)] 
acpi: move acpi_dsdt_add_power_button() to ged

Allow reuse for microvm.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-7-kraxel@redhat.com

3 days agoacpi: ged: add x86 device variant.
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:53 +0000 (14:08 +0200)] 
acpi: ged: add x86 device variant.

Set AcpiDeviceIfClass->madt_cpu,
otherwise identical to TYPE_ACPI_GED.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-6-kraxel@redhat.com

3 days agoacpi: ged: add control regs
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:52 +0000 (14:08 +0200)] 
acpi: ged: add control regs

Add control regs (sleep, reset) for hw-reduced acpi.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Igor Mammedov <imammedo@redhat.com>
Message-id: 20200915120909.20838-5-kraxel@redhat.com

3 days agoseabios: add bios-microvm.bin binary
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:51 +0000 (14:08 +0200)] 
seabios: add bios-microvm.bin binary

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200915120909.20838-4-kraxel@redhat.com

3 days agoseabios: add microvm config, update build rules
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:50 +0000 (14:08 +0200)] 
seabios: add microvm config, update build rules

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Message-id: 20200915120909.20838-3-kraxel@redhat.com

3 days agomicrovm: name qboot binary qboot.rom
Gerd Hoffmann [Tue, 15 Sep 2020 12:08:49 +0000 (14:08 +0200)] 
microvm: name qboot binary qboot.rom

qboot isn't a bios and shouldnt be named that way.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Sergio Lopez <slp@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-id: 20200915120909.20838-2-kraxel@redhat.com

4 days agoMerge remote-tracking branch 'remotes/berrange/tags/block-odirect-pull-request' into...
Peter Maydell [Wed, 16 Sep 2020 15:25:30 +0000 (16:25 +0100)] 
Merge remote-tracking branch 'remotes/berrange/tags/block-odirect-pull-request' into staging

block: improve error reporting for unsupported O_DIRECT

# gpg: Signature made Wed 16 Sep 2020 10:45:48 BST
# gpg:                using RSA key DAF3A6FDB26B62912D0E8E3FBE86EBB415104FDF
# gpg: Good signature from "Daniel P. Berrange <dan@berrange.com>" [full]
# gpg:                 aka "Daniel P. Berrange <berrange@redhat.com>" [full]
# Primary key fingerprint: DAF3 A6FD B26B 6291 2D0E  8E3F BE86 EBB4 1510 4FDF

* remotes/berrange/tags/block-odirect-pull-request:
  block/file: switch to use qemu_open/qemu_create for improved errors
  util: give a specific error message when O_DIRECT doesn't work
  util: introduce qemu_open and qemu_create with error reporting
  util: add Error object for qemu_open_internal error reporting
  util: refactor qemu_open_old to split off variadic args handling
  util: rename qemu_open() to qemu_open_old()
  util: split off a helper for dealing with O_CLOEXEC flag
  monitor: simplify functions for getting a dup'd fdset entry

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 days agoMerge remote-tracking branch 'remotes/cschoenebeck/tags/pull-9p-20200915' into staging
Peter Maydell [Wed, 16 Sep 2020 13:47:50 +0000 (14:47 +0100)] 
Merge remote-tracking branch 'remotes/cschoenebeck/tags/pull-9p-20200915' into staging

The intention of the following two patches is making users aware about
the negative file I/O performance impact when using a very low value
for 9P client parameter 'msize', which especially is the case if no
'msize' parameter was supplied by the user with a 9P Linux client at all.

All it does is logging a performance warning on host side (once) in
that case. By setting 'msize' on client side to any value larger than
8192 the performance warning will disappear.

See https://wiki.qemu.org/Documentation/9psetup#msize for details.

# gpg: Signature made Tue 15 Sep 2020 11:37:32 BST
# gpg:                using RSA key 96D8D110CF7AF8084F88590134C2B58765A47395
# gpg:                issuer "qemu_oss@crudebyte.com"
# gpg: Good signature from "Christian Schoenebeck <qemu_oss@crudebyte.com>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg:          There is no indication that the signature belongs to the owner.
# Primary key fingerprint: ECAB 1A45 4014 1413 BA38  4926 30DB 47C3 A012 D5F4
#      Subkey fingerprint: 96D8 D110 CF7A F808 4F88  5901 34C2 B587 65A4 7395

* remotes/cschoenebeck/tags/pull-9p-20200915:
  9pfs: disable msize warning for synth driver
  9pfs: log warning if msize <= 8192

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
4 days agocirrus: Building freebsd in a single shot
Yonggang Luo [Tue, 15 Sep 2020 12:13:07 +0000 (20:13 +0800)] 
cirrus: Building freebsd in a single shot

This reverts commit 45f7b7b9f38f5c4d1529a37c93dedfc26a231bba
("cirrus.yml: Split FreeBSD job into two parts").

freebsd 1 hour limit not hit anymore

I think we going to a wrong direction, I think there is some tests a stall the test runner,
please look at
https://cirrus-ci.com/task/5110577531977728
When its running properly, the consumed time are little, but when tests running too long,
look at the cpu usage, the cpu usage are nearly zero. doesn't consuming time.

And look at
https://cirrus-ci.com/task/6119341601062912

If the tests running properly, the time consuming are little
We should not hide the error by split them

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Ed Maste <emaste@FreeBSD.org>
Message-Id: <20200915121318.247-16-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agoci: Enable msys2 ci in cirrus
Yonggang Luo [Tue, 15 Sep 2020 12:13:16 +0000 (20:13 +0800)] 
ci: Enable msys2 ci in cirrus

Install msys2 in a proper way refer to:
 https://github.com/cirruslabs/cirrus-ci-docs/issues/699
The https://wiki.qemu.org/Hosts/W32#Native_builds_with_MSYS2 need to be updated.
There is no need of --cross-prefix, open mingw64.exe instead of msys2.exe then
we don't need the --cross-prefix, besides we use environment variable settings:
    MSYS: winsymlinks:nativestrict
    MSYSTEM: MINGW64
    CHERE_INVOKING: 1
to opening mingw64 native shell.

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-25-luoyonggang@gmail.com>
[thuth: Add --target-list-exclude to speed it up a little bit, and
        add capstone package to avoid build failure with internal capstone]
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: Fixes test-qdev-global-props.c
Yonggang Luo [Tue, 15 Sep 2020 17:12:29 +0000 (01:12 +0800)] 
tests: Fixes test-qdev-global-props.c

On win32 the line ending are \r\n, so we skip the \n in function test_dynamic_globalprop

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915171234.236-22-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: fix test-util-sockets.c
Yonggang Luo [Tue, 15 Sep 2020 12:13:12 +0000 (20:13 +0800)] 
tests: fix test-util-sockets.c

Fixes following errors:
Running test test-util-sockets
ERROR test-util-sockets - missing test plan

# Start of name tests
**
ERROR:../tests/test-util-sockets.c:93:test_socket_fd_pass_name_good: assertion failed (fd != -1): (-1 != -1)
Bail out! ERROR:../tests/test-util-sockets.c:93:test_socket_fd_pass_name_good: assertion failed (fd != -1): (-1 != -1)

First should call to qemu_init_main_loop before socket_init,
then on win32 doesn't support for SOCKET_ADDRESS_TYPE_FD socket type

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-21-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: Fixes test-io-channel-file by mask only owner file state mask bits
Yonggang Luo [Tue, 15 Sep 2020 17:12:27 +0000 (01:12 +0800)] 
tests: Fixes test-io-channel-file by mask only owner file state mask bits

This is the error on msys2/mingw
Running test test-io-channel-file
**
ERROR:../tests/test-io-channel-file.c:59:test_io_channel_file_helper: assertion failed (TEST_MASK & ~mask == st.st_mode & 0777): (384 == 438)
ERROR test-io-channel-file - Bail out! ERROR:../tests/test-io-channel-file.c:59:test_io_channel_file_helper: assertion failed (TEST_MASK & ~mask == st.st_mode & 0777): (384 == 438)

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915171234.236-20-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: fixes aio-win32 about aio_remove_fd_handler, get it consistence with aio-posix.c
Yonggang Luo [Tue, 15 Sep 2020 17:12:26 +0000 (01:12 +0800)] 
tests: fixes aio-win32 about aio_remove_fd_handler, get it consistence with aio-posix.c

This is a fixes for
(C:\work\xemu\qemu\build\tests\test-aio-multithread.exe:19100): GLib-CRITICAL **: 23:03:24.965: g_source_remove_poll: assertion '!SOURCE_DESTROYED (source)' failed
ERROR test-aio-multithread - Bail out! GLib-FATAL-CRITICAL: g_source_remove_poll: assertion '!SOURCE_DESTROYED (source)' failed

(C:\work\xemu\qemu\build\tests\test-bdrv-drain.exe:21036): GLib-CRITICAL **: 23:03:29.861: g_source_remove_poll: assertion '!SOURCE_DESTROYED (source)' failed
ERROR test-bdrv-drain - Bail out! GLib-FATAL-CRITICAL: g_source_remove_poll: assertion '!SOURCE_DESTROYED (source)' failed

And the idea comes from https://patchwork.kernel.org/patch/9975239/

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Message-Id: <20200915171234.236-19-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agoblock/file: switch to use qemu_open/qemu_create for improved errors
Daniel P. Berrangé [Wed, 1 Jul 2020 14:22:43 +0000 (15:22 +0100)] 
block/file: switch to use qemu_open/qemu_create for improved errors

Currently at startup if using cache=none on a filesystem lacking
O_DIRECT such as tmpfs, at startup QEMU prints

qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: file system may not support O_DIRECT
qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Could not open '/tmp/foo.img': Invalid argument

while at QMP level the hint is missing, so QEMU reports just

  "error": {
      "class": "GenericError",
      "desc": "Could not open '/tmp/foo.img': Invalid argument"
  }

which is close to useless for the end user trying to figure out what
they did wrong.

With this change at startup QEMU prints

qemu-system-x86_64: -drive file=/tmp/foo.img,cache=none: Unable to open '/tmp/foo.img': filesystem does not support O_DIRECT

while at the QMP level QEMU reports a massively more informative

  "error": {
     "class": "GenericError",
     "desc": "Unable to open '/tmp/foo.img': filesystem does not support O_DIRECT"
  }

Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoutil: give a specific error message when O_DIRECT doesn't work
Daniel P. Berrangé [Tue, 21 Jul 2020 15:17:35 +0000 (16:17 +0100)] 
util: give a specific error message when O_DIRECT doesn't work

A common error scenario is to tell QEMU to use O_DIRECT in combination
with a filesystem that doesn't support it. To aid users to diagnosing
their mistake we want to provide a clear error message when this happens.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoutil: introduce qemu_open and qemu_create with error reporting
Daniel P. Berrangé [Wed, 1 Jul 2020 15:30:35 +0000 (16:30 +0100)] 
util: introduce qemu_open and qemu_create with error reporting

qemu_open_old() works like open(): set errno and return -1 on failure.
It has even more failure modes, though.  Reporting the error clearly
to users is basically impossible for many of them.

Our standard cure for "errno is too coarse" is the Error object.
Introduce two new helper methods:

  int qemu_open(const char *name, int flags, Error **errp);
  int qemu_create(const char *name, int flags, mode_t mode, Error **errp);

Note that with this design we no longer require or even accept the
O_CREAT flag. Avoiding overloading the two distinct operations
means we can avoid variable arguments which would prevent 'errp' from
being the last argument. It also gives us a guarantee that the 'mode' is
given when creating files, avoiding a latent security bug.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoutil: add Error object for qemu_open_internal error reporting
Daniel P. Berrangé [Fri, 21 Aug 2020 16:45:02 +0000 (17:45 +0100)] 
util: add Error object for qemu_open_internal error reporting

Instead of relying on the limited information from errno, we can now
also provide detailed error messages to callers that ask for it.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoutil: refactor qemu_open_old to split off variadic args handling
Daniel P. Berrangé [Wed, 1 Jul 2020 15:30:35 +0000 (16:30 +0100)] 
util: refactor qemu_open_old to split off variadic args handling

This simple refactoring prepares for future patches. The variadic args
handling is split from the main bulk of the open logic.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoutil: rename qemu_open() to qemu_open_old()
Daniel P. Berrangé [Tue, 21 Jul 2020 12:25:21 +0000 (13:25 +0100)] 
util: rename qemu_open() to qemu_open_old()

We want to introduce a new version of qemu_open() that uses an Error
object for reporting problems and make this it the preferred interface.
Rename the existing method to release the namespace for the new impl.

Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoutil: split off a helper for dealing with O_CLOEXEC flag
Daniel P. Berrangé [Wed, 2 Sep 2020 15:58:06 +0000 (16:58 +0100)] 
util: split off a helper for dealing with O_CLOEXEC flag

We're going to have multiple callers to open() from qemu_open()
soon. Readability would thus benefit from having a helper for
dealing with O_CLOEXEC.

Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agomonitor: simplify functions for getting a dup'd fdset entry
Daniel P. Berrangé [Thu, 27 Aug 2020 12:27:00 +0000 (13:27 +0100)] 
monitor: simplify functions for getting a dup'd fdset entry

Currently code has to call monitor_fdset_get_fd, then dup
the return fd, and then add the duplicate FD back into the
fdset. This dance is overly verbose for the caller and
introduces extra failure modes which can be avoided by
folding all the logic into monitor_fdset_dup_fd_add and
removing monitor_fdset_get_fd entirely.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
4 days agoconfigure: add [lm32|unicore32]-softmmu to deprecation logic
Alex Bennée [Tue, 15 Sep 2020 13:43:17 +0000 (14:43 +0100)] 
configure: add [lm32|unicore32]-softmmu to deprecation logic

While we are at it move the few places where they are into the
deprecation build bucket.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200915134317.11110-9-alex.bennee@linaro.org>

4 days agogitlab: create a build-deprecated target
Alex Bennée [Tue, 15 Sep 2020 13:43:16 +0000 (14:43 +0100)] 
gitlab: create a build-deprecated target

These targets might be deprecated but we should keep them building
before the final axe comes down. Lets keep them all in one place and
don't hold up the CI if they do fail. They are either poorly tested or
already flaky anyway.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Acked-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20200915134317.11110-8-alex.bennee@linaro.org>

4 days agoconfigure: include tilegx-linux-user in the deprecation logic
Alex Bennée [Tue, 15 Sep 2020 13:43:15 +0000 (14:43 +0100)] 
configure: include tilegx-linux-user in the deprecation logic

The target is already marked as deprecated in the documentation.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200915134317.11110-7-alex.bennee@linaro.org>

4 days agoconfigure: clean-up the target-list-exclude logic
Alex Bennée [Tue, 15 Sep 2020 13:43:14 +0000 (14:43 +0100)] 
configure: clean-up the target-list-exclude logic

Rather than sed and loop just do a grep.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200915134317.11110-6-alex.bennee@linaro.org>

4 days agoconfigure: also skip deprecated targets with target-list-exclude
Alex Bennée [Tue, 15 Sep 2020 13:43:13 +0000 (14:43 +0100)] 
configure: also skip deprecated targets with target-list-exclude

Now the user has to make an even more deliberate decision to
enable a deprecated target rather than getting it as a side effect of
using --target-exclude-list.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200915134317.11110-5-alex.bennee@linaro.org>

4 days agoconfigure: move deprecated feature processing to supported_target
Alex Bennée [Tue, 15 Sep 2020 13:43:12 +0000 (14:43 +0100)] 
configure: move deprecated feature processing to supported_target

This is the common point at which we validate targets so it makes
sense to add_to deprecated_features here. It will make future target
deprecation easier as we only need to tweak one list.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20200915134317.11110-4-alex.bennee@linaro.org>

4 days agoiotests: Drop readlink -f
Max Reitz [Tue, 15 Sep 2020 13:43:11 +0000 (14:43 +0100)] 
iotests: Drop readlink -f

On macOS, (out of the box) readlink does not have -f.  We do not really
need readlink here, though, it was just a replacement for realpath
(which is not available on our BSD test systems), which we needed to
make the $(dirname) into an absolute path.

Instead of using either, just use "cd; pwd" like is done for
$source_iotests.

       ("iotests: Allow running from different directory")

Fixes: b1cbc33a3971b6bb005d5ac3569feae35a71de0f
Reported-by: Claudio Fontana <cfontana@suse.de>
Reported-by: Thomas Huth <thuth@redhat.com>
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20200914145606.94620-1-mreitz@redhat.com>
Message-Id: <20200915134317.11110-3-alex.bennee@linaro.org>

4 days agolinux-user: test, don't assert addr != test in pgb_reserved_va
Alex Bennée [Tue, 15 Sep 2020 13:43:10 +0000 (14:43 +0100)] 
linux-user: test, don't assert addr != test in pgb_reserved_va

On older kernels which don't implement MAP_FIXED_NOREPLACE the kernel
may still fail to give us the address we asked for despite having
already probed the map for a valid hole. Asserting isn't particularly
useful to the user so let us move the check up and expand the
error_report a little to give them a fighting chance of working around
the problem.

Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Cc: Bug 1895080 <1895080@bugs.launchpad.net>
Ameliorates: ee94743034
Message-Id: <20200915134317.11110-2-alex.bennee@linaro.org>

4 days agotests: Fixes test-io-channel-socket.c tests under msys2/mingw
Yonggang Luo [Tue, 15 Sep 2020 12:13:09 +0000 (20:13 +0800)] 
tests: Fixes test-io-channel-socket.c tests under msys2/mingw

Currently test-io-channel-socket doesn't init with
qemu_init_main_loop
and that's cause the qemu_aio_context not inited,
and the following is the stack when null pointer accessed:

qemu_fd_register (c:\work\xemu\qemu\util\main-loop.c:336)
qemu_try_set_nonblock (c:\work\xemu\qemu\util\oslib-win32.c:224)
qemu_set_nonblock (c:\work\xemu\qemu\util\oslib-win32.c:230)
socket_can_bind_connect (c:\work\xemu\qemu\tests\socket-helpers.c:93)
socket_check_protocol_support (c:\work\xemu\qemu\tests\socket-helpers.c:141)
main (c:\work\xemu\qemu\tests\test-io-channel-socket.c:568)
__tmainCRTStartup (@__tmainCRTStartup:142)
mainCRTStartup (@1400014f6..140001539:3)
BaseThreadInitThunk (@BaseThreadInitThunk:9)
RtlUserThreadStart (@RtlUserThreadStart:12)

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-18-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agovmstate: Fixes test-vmstate.c on msys2/mingw
Yonggang Luo [Thu, 10 Sep 2020 10:30:50 +0000 (18:30 +0800)] 
vmstate: Fixes test-vmstate.c on msys2/mingw

The vmstate are valid on win32, just need generate tmp path properly

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200910103059.987-17-luoyonggang@gmail.com>
[thuth: Make indentation a little bit nicer]
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agomeson: remove empty else and duplicated gio deps
Yonggang Luo [Tue, 15 Sep 2020 12:13:05 +0000 (20:13 +0800)] 
meson: remove empty else and duplicated gio deps

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-14-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agomeson: Use -b to ignore CR vs. CR-LF issues on Windows
Yonggang Luo [Tue, 15 Sep 2020 12:13:02 +0000 (20:13 +0800)] 
meson: Use -b to ignore CR vs. CR-LF issues on Windows

Ideally we would use the '--strip-trailing-cr' option, but not
being POSIX is a portability problem (i.e. BSDs and Solaris
based OSes). Instead use the '-b' option which, although doing
slightly more, produce the expected result on Windows."

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-11-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agoosdep: file locking functions are not available on Win32
Yonggang Luo [Tue, 15 Sep 2020 12:13:01 +0000 (20:13 +0800)] 
osdep: file locking functions are not available on Win32

Do not declare the following locking functions on Win32:
int qemu_lock_fd(int fd, int64_t start, int64_t len, bool exclusive);
int qemu_unlock_fd(int fd, int64_t start, int64_t len);
int qemu_lock_fd_test(int fd, int64_t start, int64_t len, bool exclusive);
bool qemu_has_ofd_lock(void);

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-10-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: test-replication disable /replication/secondary/* on msys2/mingw.
Yonggang Luo [Tue, 15 Sep 2020 12:13:00 +0000 (20:13 +0800)] 
tests: test-replication disable /replication/secondary/* on msys2/mingw.

They caused failure on msys2/mingw, that's because file-win32.c not implement
.bdrv_reopen_prepare/commit/abort yet.

This is the error message:
> $ ./tests/test-replication.exe
> # random seed: R02S3f4d1c01af2b0a046990e0235c481faf
> 1..13
> # Start of replication tests
> # Start of primary tests
> ok 1 /replication/primary/read
> ok 2 /replication/primary/write
> ok 3 /replication/primary/start
> ok 4 /replication/primary/stop
> ok 5 /replication/primary/do_checkpoint
> ok 6 /replication/primary/get_error_all
> # End of primary tests
> # Start of secondary tests
> ok 7 /replication/secondary/read
> ok 8 /replication/secondary/write
> Unexpected error in bdrv_reopen_prepare() at ../block.c:4191:
> Block format 'file' used by node '#block4287' does not support reopening
> files

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-9-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: Fixes test-replication.c on msys2/mingw.
Yonggang Luo [Tue, 15 Sep 2020 12:12:59 +0000 (20:12 +0800)] 
tests: Fixes test-replication.c on msys2/mingw.

On Windows there is no path like /tmp/s_local_disk.XXXXXX
Use g_get_tmp_dir instead of /tmp.

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-8-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agomeson: disable crypto tests are empty under win32
Yonggang Luo [Wed, 9 Sep 2020 09:46:07 +0000 (17:46 +0800)] 
meson: disable crypto tests are empty under win32

Disable following tests on msys2/mingw
      'test-crypto-tlscredsx509': ['crypto-tls-x509-helpers.c', 'pkix_asn1_tab.c',
                                   tasn1, crypto],
      'test-crypto-tlssession': ['crypto-tls-x509-helpers.c', 'pkix_asn1_tab.c', 'crypto-tls-psk-helpers.c',
                                 tasn1, crypto],
      'test-io-channel-tls': ['io-channel-helpers.c', 'crypto-tls-x509-helpers.c', 'pkix_asn1_tab.c',
                              tasn1, io, crypto]}
These tests are failure with:
ERROR test-crypto-tlscredsx509 - missing test plan
ERROR test-crypto-tlssession - missing test plan
ERROR test-io-channel-tls - missing test plan

Because on win32 those test case are all disabled.

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Message-Id: <20200909094617.1582-12-luoyonggang@gmail.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agomeson: Disable test-char on msys2/mingw for fixing tests stuck
Yonggang Luo [Mon, 7 Sep 2020 20:04:30 +0000 (04:04 +0800)] 
meson: Disable test-char on msys2/mingw for fixing tests stuck

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Message-Id: <20200907200432.2418-3-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agorcu: fixes test-logging.c by call drain_call_rcu before rmdir_full
Yonggang Luo [Tue, 15 Sep 2020 12:13:14 +0000 (20:13 +0800)] 
rcu: fixes test-logging.c by call drain_call_rcu before rmdir_full

drain_call_rcu is necessary on win32, because under win32, if you
don't close the file before remove it, the remove would be fail.

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Message-Id: <20200915121318.247-23-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests: Convert g_free to g_autofree macro in test-logging.c
Yonggang Luo [Tue, 15 Sep 2020 12:13:08 +0000 (20:13 +0800)] 
tests: Convert g_free to g_autofree macro in test-logging.c

g_autofree are prefer than g_free when possible.

Signed-off-by: Yonggang Luo <luoyonggang@gmail.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200915121318.247-17-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agorcu: Implement drain_call_rcu
Maxim Levitsky [Tue, 15 Sep 2020 12:12:53 +0000 (20:12 +0800)] 
rcu: Implement drain_call_rcu

This will allow is to preserve the semantics of hmp_device_del,
that the device is deleted immediatly which was changed by previos
patch that delayed this to RCU callback

Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Suggested-by: Stefan Hajnoczi <stefanha@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20200915121318.247-2-luoyonggang@gmail.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agoqga/commands-win32: Fix problem with redundant protype declaration
Thomas Huth [Tue, 15 Sep 2020 10:18:03 +0000 (12:18 +0200)] 
qga/commands-win32: Fix problem with redundant protype declaration

When compiling QEMU with MSYS2 on Windows, there is currently the
following error:

../qga/commands-win32.c:62:24: error: redundant redeclaration of
 'CM_Get_DevNode_PropertyW' [-Werror=redundant-decls]
   62 | CMAPI CONFIGRET WINAPI CM_Get_DevNode_PropertyW(
      |                        ^~~~~~~~~~~~~~~~~~~~~~~~
In file included from ../qga/commands-win32.c:26:
C:/tools/msys64/mingw64/x86_64-w64-mingw32/include/cfgmgr32.h:840:26: note:
 previous declaration of 'CM_Get_DevNode_PropertyW' was here
  840 |   CMAPI CONFIGRET WINAPI CM_Get_DevNode_PropertyW(DEVINST dnDevInst,
   const DEVPROPKEY *PropertyKey, DEVPROPTYPE *PropertyType, PBYTE PropertyBuffer,
   PULONG PropertyBufferSize, ULONG ulFlags);

Seems like this protype is sometimes available in the cfgmgr32.h
header, and sometimes not. Let's silence the compiler warning here
to let the build pass with -Werror, too.

Message-Id: <20200915114757.55635-1-thuth@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agoSimplify the .gitignore file
Thomas Huth [Mon, 7 Sep 2020 17:38:01 +0000 (19:38 +0200)] 
Simplify the .gitignore file

Now that we always do out-of-tree builds (and the in-tree builds are
faked via a "build" directory), we can simplify out .gitignore file
quite a bit.

Message-Id: <20200909080305.258961-1-thuth@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests/socket-helpers: Treat EAI_NONAME as EADDRNOTAVAIL
Thomas Huth [Tue, 8 Sep 2020 11:48:13 +0000 (13:48 +0200)] 
tests/socket-helpers: Treat EAI_NONAME as EADDRNOTAVAIL

The tests/test-char test is currently always failing on my system since
the getaddrinfo() in socket_can_bind_connect() returns EAI_NONAME when
it is called from socket_check_protocol_support() to check for IPv6.
socket_check_protocol_support() then returns -1 and thus the tests are
not run at all - even though IPv4 is working fine.
socket_can_bind_connect() connect should return EADDRNOTAVAIL in this
case instead, so that socket_check_protocol_support() does not fail.

Suggested-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200908121543.222872-1-thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agotests/qtest/qmp-cmd-test: Use inclusive language
Thomas Huth [Sat, 12 Sep 2020 07:43:07 +0000 (09:43 +0200)] 
tests/qtest/qmp-cmd-test: Use inclusive language

We simply want to ignore certain queries here, so let's rather
use the term 'ignore' to express this intention.

Message-Id: <20200914163755.42618-1-thuth@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agofuzz: Change the way we write qtest log to stderr
Alexander Bulekov [Wed, 19 Aug 2020 06:10:56 +0000 (02:10 -0400)] 
fuzz: Change the way we write qtest log to stderr

Telling QTest to log to /dev/fd/2, essentially results in dup(2). This
is fine, if other code isn't logging to stderr. Otherwise, the order of
the logs is mixed due to buffering issues, since two file-descriptors
are used to write to the same file. We can avoid this, since just
specifying "-qtest" sets the log fd to stderr. If we want to disable
qtest logs, we can just add -qtest-log none.

Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Signed-off-by: Alexander Bulekov <alxndr@bu.edu>
Message-Id: <20200819061110.1320568-2-alxndr@bu.edu>
Signed-off-by: Thomas Huth <thuth@redhat.com>
4 days agoiotests: Drop readlink -f
Max Reitz [Mon, 14 Sep 2020 14:56:06 +0000 (16:56 +0200)] 
iotests: Drop readlink -f

On macOS, (out of the box) readlink does not have -f.  We do not really
need readlink here, though, it was just a replacement for realpath
(which is not available on our BSD test systems), which we needed to
make the $(dirname) into an absolute path.

Instead of using either, just use "cd; pwd" like is done for
$source_iotests.

Fixes: b1cbc33a3971b6bb005d5ac3569feae35a71de0f
       ("iotests: Allow running from different directory")
Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Reported-by: Claudio Fontana <cfontana@suse.de>
Reported-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20200914145606.94620-1-mreitz@redhat.com>
Signed-off-by: Thomas Huth <thuth@redhat.com>
5 days agoMerge remote-tracking branch 'remotes/kraxel/tags/vga-20200915-pull-request' into...
Peter Maydell [Tue, 15 Sep 2020 13:25:05 +0000 (14:25 +0100)] 
Merge remote-tracking branch 'remotes/kraxel/tags/vga-20200915-pull-request' into staging

vga: fix qxl modular build.
vga: add support for virtio-gpu modular builds.

# gpg: Signature made Tue 15 Sep 2020 14:08:53 BST
# gpg:                using RSA key 4CB6D8EED3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <kraxel@redhat.com>" [full]
# gpg:                 aka "Gerd Hoffmann <gerd@kraxel.org>" [full]
# gpg:                 aka "Gerd Hoffmann (private) <kraxel@gmail.com>" [full]
# Primary key fingerprint: A032 8CFF B93A 17A7 9901  FE7D 4CB6 D8EE D3E8 7138

* remotes/kraxel/tags/vga-20200915-pull-request:
  virtio-gpu: build modular
  virtio-gpu: make virtio_gpu_ops static
  object_initialize: try module load
  meson: remove duplicate qxl sources
  meson: fix module config
  meson: fix qxl dependencies

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 days agovirtio-gpu: build modular
Gerd Hoffmann [Mon, 14 Sep 2020 13:42:24 +0000 (15:42 +0200)] 
virtio-gpu: build modular

Only build virtio-gpu-device modular (the code which actually depends on
the external virglrenderer library).  virtio-gpu-pci and virtio-vga are
compiled into core qemu still.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200914134224.29769-7-kraxel@redhat.com

5 days agovirtio-gpu: make virtio_gpu_ops static
Gerd Hoffmann [Mon, 14 Sep 2020 13:42:23 +0000 (15:42 +0200)] 
virtio-gpu: make virtio_gpu_ops static

Reference it via ops pointer instead, simliar to the vga one.
Removes hard symbol reference, needed to build virtio-gpu modular.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200914134224.29769-6-kraxel@redhat.com

5 days agoobject_initialize: try module load
Gerd Hoffmann [Mon, 14 Sep 2020 13:42:22 +0000 (15:42 +0200)] 
object_initialize: try module load

Needed to allow virtio-gpu-pci initialize the
virtio-gpu-device child device.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200914134224.29769-5-kraxel@redhat.com

5 days agomeson: remove duplicate qxl sources
Gerd Hoffmann [Mon, 14 Sep 2020 13:42:21 +0000 (15:42 +0200)] 
meson: remove duplicate qxl sources

We should add sources to the softmmu_ss or module_ss but not both.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200914134224.29769-4-kraxel@redhat.com

5 days agomeson: fix module config
Gerd Hoffmann [Mon, 14 Sep 2020 13:42:20 +0000 (15:42 +0200)] 
meson: fix module config

Use all config symbols not only the host ones.
Needed to make sure device configs like CONFIG_QXL
are used for modules too.

Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200914134224.29769-3-kraxel@redhat.com

5 days agomeson: fix qxl dependencies
Gerd Hoffmann [Mon, 14 Sep 2020 13:42:19 +0000 (15:42 +0200)] 
meson: fix qxl dependencies

Add pixman and spice deps to qxl module.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 20200914134224.29769-2-kraxel@redhat.com

5 days agoMerge remote-tracking branch 'remotes/otubo/tags/pull-seccomp-20200915' into staging
Peter Maydell [Tue, 15 Sep 2020 12:05:11 +0000 (13:05 +0100)] 
Merge remote-tracking branch 'remotes/otubo/tags/pull-seccomp-20200915' into staging

seccomp branch queue

# gpg: Signature made Tue 15 Sep 2020 11:58:27 BST
# gpg:                using RSA key D67E1B50937486B40723DBABDF32E7C0F0FFF9A2
# gpg:                issuer "otubo@redhat.com"
# gpg: Good signature from "Eduardo Otubo (Senior Software Engineer) <otubo@redhat.com>" [full]
# Primary key fingerprint: D67E 1B50 9374 86B4 0723  DBAB DF32 E7C0 F0FF F9A2

* remotes/otubo/tags/pull-seccomp-20200915:
  seccomp: fix killing of whole process instead of thread

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 days agoseccomp: fix killing of whole process instead of thread
Daniel P. Berrangé [Fri, 11 Sep 2020 14:18:32 +0000 (15:18 +0100)] 
seccomp: fix killing of whole process instead of thread

Back in 2018 we introduced support for killing the whole QEMU process
instead of just one thread, when a seccomp rule is violated:

  commit bda08a5764d470f101fa38635d30b41179a313e1
  Author: Marc-André Lureau <marcandre.lureau@redhat.com>
  Date:   Wed Aug 22 19:02:48 2018 +0200

    seccomp: prefer SCMP_ACT_KILL_PROCESS if available

Fast forward a year and we introduced a patch to avoid killing the
process for resource control syscalls tickled by Mesa.

  commit 9a1565a03b79d80b236bc7cc2dbce52a2ef3a1b8
  Author: Daniel P. Berrangé <berrange@redhat.com>
  Date:   Wed Mar 13 09:49:03 2019 +0000

    seccomp: don't kill process for resource control syscalls

Unfortunately a logic bug effectively reverted the first commit
mentioned so that we go back to only killing the thread, not the whole
process.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Acked-by: Eduardo Otubo <otubo@redhat.com>
5 days agoMerge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2020-09-15' into staging
Peter Maydell [Tue, 15 Sep 2020 10:48:40 +0000 (11:48 +0100)] 
Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2020-09-15' into staging

Block patches:
- Several qcow2 fixes and refactorings
- Let qemu-img convert try to stay at cluster boundaries
- Stable child names for quorum (with x-blockdev-change)
- Explicitly drop vhdx 4k sector support, as it was never actually
  working
- rbd: Mark @namespace a strong runtime option
- iotests.py improvements
- Drop unused runtime_opts objects
- Skip a test case in 030 when run through make check-block

# gpg: Signature made Tue 15 Sep 2020 11:27:05 BST
# gpg:                using RSA key 91BEB60A30DB3E8857D11829F407DB0061D5CF40
# gpg:                issuer "mreitz@redhat.com"
# gpg: Good signature from "Max Reitz <mreitz@redhat.com>" [full]
# Primary key fingerprint: 91BE B60A 30DB 3E88 57D1  1829 F407 DB00 61D5 CF40

* remotes/maxreitz/tags/pull-block-2020-09-15: (22 commits)
  block/rbd: add 'namespace' to qemu_rbd_strong_runtime_opts[]
  qcow2: Convert qcow2_alloc_cluster_offset() into qcow2_alloc_host_offset()
  qcow2: Make preallocate_co() resize the image to the correct size
  block/qcow: remove runtime opts
  block/rbd: remove runtime_opts
  qcow2: Return the original error code in qcow2_co_pwrite_zeroes()
  qcow2: Make qcow2_free_any_clusters() free only one cluster
  qcow2: Handle QCowL2Meta on error in preallocate_co()
  block/vhdx: Support vhdx image only with 512 bytes logical sector size
  iotests: Skip test_stream_parallel in test 030 when doing "make check"
  qemu-img: Explicit number replaced by a constant
  qcow2: Rewrite the documentation of qcow2_alloc_cluster_offset()
  qcow2: Don't check nb_clusters when removing l2meta from the list
  qcow2: Fix removal of list members from BDRVQcow2State.cluster_allocs
  qcow2: Use macros for the L1, refcount and bitmap table entry sizes
  qemu-img: avoid unaligned read requests during convert
  block/quorum.c: stable children names
  qemu-iotests: Simplify FilePath __init__
  qemu-iotests: Merge FilePaths and FilePath
  qemu-iotests: Support varargs syntax in FilePaths
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
5 days ago9pfs: disable msize warning for synth driver
Christian Schoenebeck [Sun, 6 Sep 2020 16:50:32 +0000 (18:50 +0200)] 
9pfs: disable msize warning for synth driver

Previous patch introduced a performance warning being logged on host
side if client connected with an 'msize' <= 8192. Disable this
performance warning for the synth driver to prevent that warning from
being printed whenever the 9pfs (qtest) test cases are running.

Introduce a new export flag V9FS_NO_PERF_WARN for that purpose, which
might also be used to disable such warnings from the CLI in future.

We could have also prevented the warning by simply raising P9_MAX_SIZE
in virtio-9p-test.c to any value larger than 8192, however in the
context of test cases it makes sense running for edge cases, which
includes the lowest 'msize' value supported by the server which is
4096, hence we want to preserve an msize of 4096 for the test client.

Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <E1kEyDy-0006nN-5A@lizzy.crudebyte.com>
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
5 days ago9pfs: log warning if msize <= 8192
Christian Schoenebeck [Thu, 3 Sep 2020 14:20:21 +0000 (16:20 +0200)] 
9pfs: log warning if msize <= 8192

It is essential to choose a reasonable high value for 'msize' to avoid
severely degraded file I/O performance. This parameter can only be
chosen on client/guest side, and a Linux client defaults to an 'msize'
of only 8192 if the user did not explicitly specify a value for 'msize',
which results in very poor file I/O performance.

Unfortunately many users are not aware that they should specify an
appropriate value for 'msize' to avoid severe performance issues, so
log a performance warning (with a QEMU wiki link explaining this issue
in detail) on host side in that case to make it more clear.

Currently a client cannot automatically pick a reasonable value for
'msize', because a good value for 'msize' depends on the file I/O
potential of the underlying storage on host side, i.e. a feature
invisible to the client, and even then a user would still need to trade
off between performance profit and additional RAM costs, i.e. with
growing 'msize' (RAM occupation), performance still increases, but
performance delta will shrink continuously.

Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Message-Id: <e6fc84845c95816ad5baecb0abd6bfefdcf7ec9f.1599144062.git.qemu_oss@crudebyte.com>
Signed-off-by: Christian Schoenebeck <qemu_oss@crudebyte.com>
5 days agoblock/rbd: add 'namespace' to qemu_rbd_strong_runtime_opts[]
Stefano Garzarella [Mon, 14 Sep 2020 19:05:53 +0000 (21:05 +0200)] 
block/rbd: add 'namespace' to qemu_rbd_strong_runtime_opts[]

Commit 19ae9ae014 ("block/rbd: Add support for ceph namespaces")
introduced namespace support for RBD, but we forgot to add the
new 'namespace' options to qemu_rbd_strong_runtime_opts[].

The 'namespace' is used to identify the image, so it is a strong
option since it can changes the data of a BDS.

Buglink: https://bugzilla.redhat.com/show_bug.cgi?id=1821528
Fixes: 19ae9ae014 ("block/rbd: Add support for ceph namespaces")
Cc: Florian Florensa <fflorensa@online.net>
Signed-off-by: Stefano Garzarella <sgarzare@redhat.com>
Message-Id: <20200914190553.74871-1-sgarzare@redhat.com>
Reviewed-by: Jason Dillaman <dillaman@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Convert qcow2_alloc_cluster_offset() into qcow2_alloc_host_offset()
Alberto Garcia [Fri, 11 Sep 2020 14:09:42 +0000 (16:09 +0200)] 
qcow2: Convert qcow2_alloc_cluster_offset() into qcow2_alloc_host_offset()

qcow2_alloc_cluster_offset() takes an (unaligned) guest offset and
returns the (aligned) offset of the corresponding cluster in the qcow2
image.

In practice none of the callers need to know where the cluster starts
so this patch makes the function calculate and return the final host
offset directly. The function is also renamed accordingly.

See 388e581615 for a similar change to qcow2_get_cluster_offset().

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <9bfef50ec9200d752413be4fc2aeb22a28378817.1599833007.git.berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Make preallocate_co() resize the image to the correct size
Alberto Garcia [Fri, 11 Sep 2020 14:09:41 +0000 (16:09 +0200)] 
qcow2: Make preallocate_co() resize the image to the correct size

This function preallocates metadata structures and then extends the
image to its new size, but that new size calculation is wrong because
it doesn't take into account that the host_offset variable is always
cluster-aligned.

This problem can be reproduced with preallocation=metadata when the
original size is not cluster-aligned but the new size is. In this case
the final image size will be shorter than expected.

   qemu-img create -f qcow2 img.qcow2 31k
   qemu-img resize --preallocation=metadata img.qcow2 128k

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <adeb8b059917b141d5f5b3bd2a016262d3052c79.1599833007.git.berto@igalia.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
[mreitz: Mark compat=0.10 unsupported for iotest 125]
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoblock/qcow: remove runtime opts
John Snow [Thu, 6 Aug 2020 21:13:45 +0000 (17:13 -0400)] 
block/qcow: remove runtime opts

Introduced by d85f4222b468,
These were seemingly never used at all.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20200806211345.2925343-3-jsnow@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoblock/rbd: remove runtime_opts
John Snow [Thu, 6 Aug 2020 21:13:44 +0000 (17:13 -0400)] 
block/rbd: remove runtime_opts

This saw its last use in 4bfb274165ba.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20200806211345.2925343-2-jsnow@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Return the original error code in qcow2_co_pwrite_zeroes()
Alberto Garcia [Wed, 9 Sep 2020 12:37:39 +0000 (14:37 +0200)] 
qcow2: Return the original error code in qcow2_co_pwrite_zeroes()

This function checks the current status of a (sub)cluster in order to
see if an unaligned 'write zeroes' request can be done efficiently by
simply updating the L2 metadata and without having to write actual
zeroes to disk.

If the situation does not allow using the fast path then the function
returns -ENOTSUP and the caller falls back to writing zeroes.

If can happen however that the aforementioned check returns an actual
error code so in this case we should pass it to the caller.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20200909123739.719-1-berto@igalia.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Make qcow2_free_any_clusters() free only one cluster
Alberto Garcia [Tue, 8 Sep 2020 14:08:28 +0000 (16:08 +0200)] 
qcow2: Make qcow2_free_any_clusters() free only one cluster

This function takes an L2 entry and a number of clusters to free.
Although in principle it can free any type of cluster (using the L2
entry to determine its type) in practice the API is broken because
compressed clusters have a variable size and there is no way to free
more than one without having the L2 entry of each one of them.

The good news all callers are passing nb_clusters=1 so we can simply
get rid of that parameter.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <77cea0f4616f921d37e971b3c5b18a2faa24b173.1599573989.git.berto@igalia.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Handle QCowL2Meta on error in preallocate_co()
Alberto Garcia [Tue, 8 Sep 2020 14:08:27 +0000 (16:08 +0200)] 
qcow2: Handle QCowL2Meta on error in preallocate_co()

If qcow2_alloc_cluster_offset() or qcow2_alloc_cluster_link_l2() fail
then this function simply returns the error code, potentially leaking
the QCowL2Meta structure and leaving stale items in s->cluster_allocs.

A second problem is that this function calls qcow2_free_any_clusters()
on failure but passing a host cluster offset instead of an L2 entry.
Luckily for normal uncompressed clusters a raw offset also works like
a valid L2 entry so it works just the same, but we should be using
qcow2_free_clusters() instead.

This patch fixes both problems by using qcow2_handle_l2meta().

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <cd3a6b9abd43f9c0b60be413d760f0cacc67eb66.1599573989.git.berto@igalia.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoblock/vhdx: Support vhdx image only with 512 bytes logical sector size
Swapnil Ingle [Fri, 7 Aug 2020 10:03:23 +0000 (10:03 +0000)] 
block/vhdx: Support vhdx image only with 512 bytes logical sector size

block/vhdx uses qemu block layer where sector size is always 512 bytes.
This may have issues  with 4K logical sector sized vhdx image.

For e.g qemu-img convert on such images fails with following assert:

$qemu-img convert -f vhdx -O raw 4KTest1.vhdx test.raw
qemu-img: util/iov.c:388: qiov_slice: Assertion `offset + len <=
qiov->size' failed.
Aborted

This patch adds an check to return ENOTSUP for vhdx images which
have logical sector size other than 512 bytes.

Signed-off-by: Swapnil Ingle <swapnil.ingle@nutanix.com>
Message-Id: <1596794594-44531-1-git-send-email-swapnil.ingle@nutanix.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoiotests: Skip test_stream_parallel in test 030 when doing "make check"
Thomas Huth [Mon, 7 Sep 2020 11:38:24 +0000 (13:38 +0200)] 
iotests: Skip test_stream_parallel in test 030 when doing "make check"

The test_stream_parallel test still occasionally fails in the CI.
Thus let's disable it during "make check" for now so that it does
not cause trouble during merge tests. We can enable it again once
the problem has been resolved.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20200907113824.134788-1-thuth@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqemu-img: Explicit number replaced by a constant
Yi Li [Wed, 19 Aug 2020 01:36:07 +0000 (09:36 +0800)] 
qemu-img: Explicit number replaced by a constant

Signed-off-by: Yi Li <yili@winhong.com>
Message-Id: <20200819013607.32280-1-yili@winhong.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Rewrite the documentation of qcow2_alloc_cluster_offset()
Alberto Garcia [Thu, 3 Sep 2020 16:37:49 +0000 (18:37 +0200)] 
qcow2: Rewrite the documentation of qcow2_alloc_cluster_offset()

The current text corresponds to an earlier, simpler version of this
function and it does not explain how it works now.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <bb5bd06f07c5a05b0818611de0d06ec5b66c8df3.1599150873.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Don't check nb_clusters when removing l2meta from the list
Alberto Garcia [Thu, 3 Sep 2020 16:37:48 +0000 (18:37 +0200)] 
qcow2: Don't check nb_clusters when removing l2meta from the list

In the past, when a new cluster was allocated the l2meta structure was
a variable in the stack so it was necessary to have a way to tell
whether it had been initialized and contained valid data or not. The
nb_clusters field was used for this purpose. Since commit f50f88b9fe
this is no longer the case, l2meta (nowadays a pointer to a list) is
only allocated when needed and nb_clusters is guaranteed to be > 0 so
this check is unnecessary.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <ab0b67c29c7ba26e598db35f12aa5ab5982539c1.1599150873.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Fix removal of list members from BDRVQcow2State.cluster_allocs
Alberto Garcia [Thu, 3 Sep 2020 16:37:47 +0000 (18:37 +0200)] 
qcow2: Fix removal of list members from BDRVQcow2State.cluster_allocs

When a write request needs to allocate new clusters (or change the L2
bitmap of existing ones) a QCowL2Meta structure is created so the L2
metadata can be later updated and any copy-on-write can be performed
if necessary.

A write request can span a region consisting of an arbitrary
combination of previously unallocated and allocated clusters, and if
the unallocated ones can be put contiguous to the existing ones then
QEMU will do so in order to minimize the number of write operations.

In practice this means that a write request has not just one but a
number of QCowL2Meta structures. All of them are added to the
cluster_allocs list that is stored in BDRVQcow2State and is used to
detect overlapping requests. After the write request finishes all its
associated QCowL2Meta are removed from that list. calculate_l2_meta()
takes care of creating and putting those structures in the list, and
qcow2_handle_l2meta() takes care of removing them.

The problem is that the error path in handle_alloc() also tries to
remove an item in that list, a remnant from the time when this was
handled there (that code would not even be correct anymore because
it only removes one struct and not all the ones from the same write
request).

This can trigger a double removal of the same item from the list,
causing a crash. This is not easy to reproduce in practice because
it requires that do_alloc_cluster_offset() fails after a successful
previous allocation during the same write request, but it can be
reproduced with the included test case.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <3440a1c4d53c4fe48312b478c96accb338cbef7c.1599150873.git.berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
5 days agoqcow2: Use macros for the L1, refcount and bitmap table entry sizes
Alberto Garcia [Fri, 28 Aug 2020 11:08:28 +0000 (13:08 +0200)] 
qcow2: Use macros for the L1, refcount and bitmap table entry sizes

This patch replaces instances of sizeof(uint64_t) in the qcow2 driver
with macros that indicate what those sizes are actually referring to.

Signed-off-by: Alberto Garcia <berto@igalia.com>
Message-Id: <20200828110828.13833-1-berto@igalia.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>