OSDN Git Service

qmiga/qemu.git
3 years agoblockdev: Fix a memleak in drive_backup_prepare()
Pan Nengyuan [Fri, 23 Oct 2020 06:12:17 +0000 (14:12 +0800)]
blockdev: Fix a memleak in drive_backup_prepare()

'local_err' seems forgot to propagate in error path, it'll cause
a memleak. Fix it.

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
Signed-off-by: Chen Qun <kuhn.chenqun@huawei.com>
Message-Id: <20201023061218.2080844-7-kuhn.chenqun@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoblock/file-posix: fix a possible undefined behavior
Pan Nengyuan [Fri, 23 Oct 2020 06:12:18 +0000 (14:12 +0800)]
block/file-posix: fix a possible undefined behavior

local_err is not initialized to NULL, it will cause a assert error as below:
qemu/util/error.c:59: error_setv: Assertion `*errp == NULL' failed.

Fixes: c6447510690
Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
Signed-off-by: Chen Qun <kuhn.chenqun@huawei.com>
Message-Id: <20201023061218.2080844-8-kuhn.chenqun@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoelf2dmp/pdb: Plug memleak in pdb_init_from_file
Pan Nengyuan [Fri, 23 Oct 2020 06:12:15 +0000 (14:12 +0800)]
elf2dmp/pdb: Plug memleak in pdb_init_from_file

Missing g_error_free in pdb_init_from_file() error path. Fix that.

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com>
Reviewed-by: Viktor Prutyanov <viktor.prutyanov@phystech.edu>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
Signed-off-by: Chen Qun <kuhn.chenqun@huawei.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201023061218.2080844-5-kuhn.chenqun@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoelf2dmp/qemu_elf: Plug memleak in QEMU_Elf_init
Pan Nengyuan [Fri, 23 Oct 2020 06:12:14 +0000 (14:12 +0800)]
elf2dmp/qemu_elf: Plug memleak in QEMU_Elf_init

Missing g_error_free in QEMU_Elf_init() error path. Fix that.

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Pan Nengyuan <pannengyuan@huawei.com>
Reviewed-by: Viktor Prutyanov <viktor.prutyanov@phystech.edu>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
Signed-off-by: Chen Qun <kuhn.chenqun@huawei.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201023061218.2080844-4-kuhn.chenqun@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoconfigure: Test if $make actually exists
Roman Bolshakov [Tue, 25 Aug 2020 20:27:55 +0000 (23:27 +0300)]
configure: Test if $make actually exists

configure doesn't detect if $make is installed on the build host.
This is also helpful for hosts where an alias for make is used, i.e.
configure would fail if gmake is not present on macOS.

Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Signed-off-by: Roman Bolshakov <r.bolshakov@yadro.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20200825202755.50626-5-r.bolshakov@yadro.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoads7846: moves from the hw/display folder to the hw/input folder.
Gan Qixin [Sun, 15 Nov 2020 12:35:03 +0000 (20:35 +0800)]
ads7846: moves from the hw/display folder to the hw/input folder.

ads7846 is a touch-screen controller that is an input device rather
than a display device, so move it to the hw/input folder.

Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-Id: <20201115123503.1110665-1-ganqixin@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoCODING_STYLE.rst: Be less strict about 80 character limit
Peter Maydell [Fri, 6 Nov 2020 11:29:40 +0000 (11:29 +0000)]
CODING_STYLE.rst: Be less strict about 80 character limit

Relax the wording about line lengths a little bit; this goes with the
checkpatch changes to warn at 100 characters rather than 80.

(Compare the Linux kernel commit bdc48fa11e46f8; our coding style is
not theirs, but the rationale is good and applies to us too.)

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Message-Id: <20201106112940.31300-1-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agofsdev: open brace '{' following struct go on the same line
zhouyang [Wed, 9 Dec 2020 06:07:35 +0000 (14:07 +0800)]
fsdev: open brace '{' following struct go on the same line

I found some style problems while check the code using checkpatch.pl.
This commit fixs the issue below:
ERROR: open brace '{' following struct go on the same line

Signed-off-by: zhouyang <zhouyang789@huawei.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201209060735.2760943-1-zhouyang789@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agohw/pci-host/pam: Replace magic number by PAM_REGIONS_COUNT definition
Philippe Mathieu-Daudé [Wed, 2 Dec 2020 13:20:38 +0000 (14:20 +0100)]
hw/pci-host/pam: Replace magic number by PAM_REGIONS_COUNT definition

While this change helps triskaidekaphobic developers, it
is a good practice to avoid magic values and using constant
definitions instead.

Introduce the PAM_REGIONS_COUNT and use it. No logical change.

Signed-off-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Reviewed-by: Julia Suvorova <jusual@redhat.com>
Message-Id: <20201202132038.1276404-1-philmd@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agohw/xen: Don't use '#' flag of printf format
Xinhao Zhang [Wed, 4 Nov 2020 13:37:09 +0000 (21:37 +0800)]
hw/xen: Don't use '#' flag of printf format

Fix code style. Don't use '#' flag of printf format ('%#') in
format strings, use '0x' prefix instead

Signed-off-by: Xinhao Zhang <zhangxinhao1@huawei.com>
Signed-off-by: Kai Deng <dengkai1@huawei.com>
Message-Id: <20201104133709.3326630-1-zhangxinhao1@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoMAINTAINERS: update my email address
Michael Roth [Tue, 27 Oct 2020 06:02:53 +0000 (01:02 -0500)]
MAINTAINERS: update my email address

I've recently switched employers and the current email address is out
of date.

Signed-off-by: Michael Roth <michael.roth@amd.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201027060253.353054-1-michael.roth@amd.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoqemu-options.hx: Fix minor issues in icount documentation
Peter Maydell [Sat, 21 Nov 2020 21:35:06 +0000 (21:35 +0000)]
qemu-options.hx: Fix minor issues in icount documentation

The documentation for the icount documentation has some minor issues:
 * in a couple of places it says "sleep=on|off" when in the context of the
   sentence it means specifically "sleep=on"
 * the synopsis line for the documentation has drifted out of sync
   with the synopsis line in the DEF() macro (used for "-help" output)
 * the synopsis line in the DEF() macro is missing a "][" between
   the sleep= part and the rr= part
 * the synopsis line doesn't indicate that rrsnapshot is an optional
   part of the rr=mode,rrfile=filename subgrouping
 * we don't document that sleep=on can't be used with shift=auto
   or align=on
 * the rr option description had some minor grammar and formatting
   errors and was a bit terse
 * in commit f1f4b57e88ff in 2015 the documentation of the sleep=
   suboption got added between the two paragraphs defining general
   behaviour of the icount option. This meant that the second
   paragraph talking about the behaviour of "this option" reads as
   if it's talking about sleep=on, when it's really describing -icount
   as a whole. The paragraph is better moved back up to above the
   sleep= section.
 * the summary text displayed in "-help" output didn't mention
   the record-and-replay part

Fix these errors.

Fixes: https://bugs.launchpad.net/qemu/+bug/1774412
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Message-Id: <20201121213506.15599-1-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agotarget/i386: tracing: format length values as hex
Dov Murik [Tue, 3 Nov 2020 10:07:45 +0000 (10:07 +0000)]
target/i386: tracing: format length values as hex

Three trace events had a literal "0x" prefix but the 'len' values were
formatted in decimal base.  Keep the prefix and format the numbers as
hex, as is usually the case for length of memory regions.

Signed-off-by: Dov Murik <dovmurik@linux.vnet.ibm.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Message-Id: <20201103100745.160409-1-dovmurik@linux.vnet.ibm.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoMerge remote-tracking branch 'remotes/vivier/tags/m68k-for-6.0-pull-request' into...
Peter Maydell [Sat, 12 Dec 2020 18:33:46 +0000 (18:33 +0000)]
Merge remote-tracking branch 'remotes/vivier/tags/m68k-for-6.0-pull-request' into staging

m68k pull request 20201212

Fix for Coverity CID 1421883
Fix some comment spelling errors
Add m68k vmstate

# gpg: Signature made Sat 12 Dec 2020 17:54:28 GMT
# gpg:                using RSA key CD2F75DDC8E3A4DC2E4F5173F30C38BD3F2FBE3C
# gpg:                issuer "laurent@vivier.eu"
# gpg: Good signature from "Laurent Vivier <lvivier@redhat.com>" [full]
# gpg:                 aka "Laurent Vivier <laurent@vivier.eu>" [full]
# gpg:                 aka "Laurent Vivier (Red Hat) <lvivier@redhat.com>" [full]
# Primary key fingerprint: CD2F 75DD C8E3 A4DC 2E4F  5173 F30C 38BD 3F2F BE3C

* remotes/vivier/tags/m68k-for-6.0-pull-request:
  m68k: fix some comment spelling errors
  target/m68k: Add vmstate definition for M68kCPU
  target/m68k: remove useless qregs array
  hw/m68k/q800.c: Make the GLUE chip an actual QOM device
  hw/m68k/q800: Don't connect two qemu_irqs directly to the same input

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agom68k: fix some comment spelling errors
zhaolichang [Fri, 9 Oct 2020 06:44:43 +0000 (14:44 +0800)]
m68k: fix some comment spelling errors

I found that there are many spelling errors in the comments of qemu/target/m68k.
I used spellcheck to check the spelling errors and found some errors in the folder.

Signed-off-by: zhaolichang <zhaolichang@huawei.com>
Reviewed-by: David Edmondson <david.edmondson@oracle.com>
Reviewed-by: Philippe Mathieu-Daude<f4bug@amsat.org>
Reviewed-by: Laurent Vivier<laurent@vivier.eu>
Message-Id: <20201009064449.2336-9-zhaolichang@huawei.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agotarget/m68k: Add vmstate definition for M68kCPU
Laurent Vivier [Thu, 22 Oct 2020 20:30:00 +0000 (22:30 +0200)]
target/m68k: Add vmstate definition for M68kCPU

Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20201022203000.1922749-3-laurent@vivier.eu>

3 years agotarget/m68k: remove useless qregs array
Laurent Vivier [Thu, 22 Oct 2020 20:29:59 +0000 (22:29 +0200)]
target/m68k: remove useless qregs array

They are unused since the target has been converted to TCG.

Fixes: e1f3808e03f7 ("Convert m68k target to TCG.")
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Message-Id: <20201022203000.1922749-2-laurent@vivier.eu>

3 years agohw/m68k/q800.c: Make the GLUE chip an actual QOM device
Peter Maydell [Fri, 6 Nov 2020 23:51:09 +0000 (23:51 +0000)]
hw/m68k/q800.c: Make the GLUE chip an actual QOM device

The handling of the GLUE (General Logic Unit) device is
currently open-coded. Make this into a proper QOM device.

This minor piece of modernisation gets rid of the free
floating qemu_irq array 'pic', which Coverity points out
is technically leaked when we exit the machine init function.
(The replacement glue device is not leaked because it gets
added to the sysbus, so it's accessible via that.)

Fixes: Coverity CID 1421883
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Laurent vivier <laurent@vivier.eu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20201106235109.7066-3-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agohw/m68k/q800: Don't connect two qemu_irqs directly to the same input
Peter Maydell [Fri, 6 Nov 2020 23:51:08 +0000 (23:51 +0000)]
hw/m68k/q800: Don't connect two qemu_irqs directly to the same input

The q800 board code connects both of the IRQ outputs of the ESCC
to the same pic[3] qemu_irq. Connecting two qemu_irqs outputs directly
to the same input is not valid as it produces subtly wrong behaviour
(for instance if both the IRQ lines are high, and then one goes
low, the PIC input will see this as a high-to-low transition
even though the second IRQ line should still be holding it high).

This kind of wiring needs an explicitly created OR gate; add one.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20201106235109.7066-2-peter.maydell@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>
3 years agoMerge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Peter Maydell [Sat, 12 Dec 2020 00:20:45 +0000 (00:20 +0000)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging

Block layer patches:

- Support for FUSE exports
- Fix deadlock in bdrv_co_yield_to_drain()
- Use lock guard macros
- Some preparational patches for 64 bit block layer
- file-posix: Fix request extension to INT64_MAX in raw_do_pwrite_zeroes()

# gpg: Signature made Fri 11 Dec 2020 17:06:19 GMT
# gpg:                using RSA key DC3DEB159A9AF95D3D7456FE7F09B272C88F2FD6
# gpg:                issuer "kwolf@redhat.com"
# gpg: Good signature from "Kevin Wolf <kwolf@redhat.com>" [full]
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74  56FE 7F09 B272 C88F 2FD6

* remotes/kevin/tags/for-upstream: (34 commits)
  block: Fix deadlock in bdrv_co_yield_to_drain()
  block: Fix locking in qmp_block_resize()
  block: Simplify qmp_block_resize() error paths
  block: introduce BDRV_MAX_LENGTH
  block/io: bdrv_check_byte_request(): drop bdrv_is_inserted()
  block/io: bdrv_refresh_limits(): use ERRP_GUARD
  block/file-posix: fix workaround in raw_do_pwrite_zeroes()
  can-host: Fix crash when 'canbus' property is not set
  iotests/221: Discard image before qemu-img map
  file-posix: check the use_lock before setting the file lock
  iotests/308: Add test for FUSE exports
  iotests: Enable fuse for many tests
  iotests: Allow testing FUSE exports
  iotests: Give access to the qemu-storage-daemon
  storage-daemon: Call bdrv_close_all() on exit
  iotests/287: Clean up subshell test image
  iotests: Let _make_test_img guess $TEST_IMG_FILE
  iotests: Restrict some Python tests to file
  iotests/091: Use _cleanup_qemu instad of "wait"
  iotests: Derive image names from $TEST_IMG
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agoMerge remote-tracking branch 'remotes/cohuck/tags/s390x-20201211' into staging
Peter Maydell [Fri, 11 Dec 2020 22:22:50 +0000 (22:22 +0000)]
Merge remote-tracking branch 'remotes/cohuck/tags/s390x-20201211' into staging

First set of 6.0 patches for s390x:
- acceptance test for device detection
- bugfixes

# gpg: Signature made Fri 11 Dec 2020 12:21:45 GMT
# gpg:                using RSA key C3D0D66DC3624FF6A8C018CEDECF6B93C6F02FAF
# gpg:                issuer "cohuck@redhat.com"
# gpg: Good signature from "Cornelia Huck <conny@cornelia-huck.de>" [unknown]
# gpg:                 aka "Cornelia Huck <huckc@linux.vnet.ibm.com>" [full]
# gpg:                 aka "Cornelia Huck <cornelia.huck@de.ibm.com>" [full]
# gpg:                 aka "Cornelia Huck <cohuck@kernel.org>" [unknown]
# gpg:                 aka "Cornelia Huck <cohuck@redhat.com>" [unknown]
# Primary key fingerprint: C3D0 D66D C362 4FF6 A8C0  18CE DECF 6B93 C6F0 2FAF

* remotes/cohuck/tags/s390x-20201211:
  s390x/cpu: Use timer_free() in the finalize function to avoid memleaks
  tests/acceptance: test s390x zpci fid propagation
  tests/acceptance: verify s390x device detection
  tests/acceptance: test virtio-ccw revision handling
  tests/acceptance: add a test for devices on s390x
  hw/watchdog/wdt_diag288: Remove unnecessary includes

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agoMerge remote-tracking branch 'remotes/kraxel/tags/ui-20201211-pull-request' into...
Peter Maydell [Fri, 11 Dec 2020 16:55:52 +0000 (16:55 +0000)]
Merge remote-tracking branch 'remotes/kraxel/tags/ui-20201211-pull-request' into staging

ui/console ui_info tweaks.
ui/vnc: alpha cursor support.
ui/vnc: locking fixes.
ui/sdl: add extra mouse buttons.

# gpg: Signature made Fri 11 Dec 2020 09:12:39 GMT
# 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/ui-20201211-pull-request:
  sdl2: Add extra mouse buttons
  ui/vnc: Add missing lock for send_color_map
  vnc: add alpha cursor support
  vnc: add pseudo encodings
  vnc: drop unused copyrect feature
  vnc: use enum for features
  console: allow con==NULL in dpy_{get, set}_ui_info and dpy_ui_info_supported
  console: drop qemu_console_get_ui_info

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agoblock: Fix deadlock in bdrv_co_yield_to_drain()
Kevin Wolf [Thu, 3 Dec 2020 17:23:11 +0000 (18:23 +0100)]
block: Fix deadlock in bdrv_co_yield_to_drain()

If bdrv_co_yield_to_drain() is called for draining a block node that
runs in a different AioContext, it keeps that AioContext locked while it
yields and schedules a BH in the AioContext to do the actual drain.

As long as executing the BH is the very next thing that the event loop
of the node's AioContext does, this actually happens to work, but when
it tries to execute something else that wants to take the AioContext
lock, it will deadlock. (In the bug report, this other thing is a
virtio-scsi device running virtio_scsi_data_plane_handle_cmd().)

Instead, always drop the AioContext lock across the yield and reacquire
it only when the coroutine is reentered. The BH needs to unconditionally
take the lock for itself now.

This fixes the 'block_resize' QMP command on a block node that runs in
an iothread.

Cc: qemu-stable@nongnu.org
Fixes: eb94b81a94bce112e6b206df846c1551aaf6cab6
Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1903511
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201203172311.68232-4-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock: Fix locking in qmp_block_resize()
Kevin Wolf [Thu, 3 Dec 2020 17:23:10 +0000 (18:23 +0100)]
block: Fix locking in qmp_block_resize()

The drain functions assume that we hold the AioContext lock of the
drained block node. Make sure to actually take the lock.

Cc: qemu-stable@nongnu.org
Fixes: eb94b81a94bce112e6b206df846c1551aaf6cab6
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201203172311.68232-3-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock: Simplify qmp_block_resize() error paths
Kevin Wolf [Thu, 3 Dec 2020 17:23:09 +0000 (18:23 +0100)]
block: Simplify qmp_block_resize() error paths

The only thing that happens after the 'out:' label is blk_unref(blk).
However, blk = NULL in all of the error cases, so instead of jumping to
'out:', we can just return directly.

Cc: qemu-stable@nongnu.org
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201203172311.68232-2-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock: introduce BDRV_MAX_LENGTH
Vladimir Sementsov-Ogievskiy [Thu, 3 Dec 2020 22:27:13 +0000 (01:27 +0300)]
block: introduce BDRV_MAX_LENGTH

We are going to modify block layer to work with 64bit requests. And
first step is moving to int64_t type for both offset and bytes
arguments in all block request related functions.

It's mostly safe (when widening signed or unsigned int to int64_t), but
switching from uint64_t is questionable.

So, let's first establish the set of requests we want to work with.
First signed int64_t should be enough, as off_t is signed anyway. Then,
obviously offset + bytes should not overflow.

And most interesting: (offset + bytes) being aligned up should not
overflow as well. Aligned to what alignment? First thing that comes in
mind is bs->bl.request_alignment, as we align up request to this
alignment. But there is another thing: look at
bdrv_mark_request_serialising(). It aligns request up to some given
alignment. And this parameter may be bdrv_get_cluster_size(), which is
often a lot greater than bs->bl.request_alignment.
Note also, that bdrv_mark_request_serialising() uses signed int64_t for
calculations. So, actually, we already depend on some restrictions.

Happily, bdrv_get_cluster_size() returns int and
bs->bl.request_alignment has 32bit unsigned type, but defined to be a
power of 2 less than INT_MAX. So, we may establish, that INT_MAX is
absolute maximum for any kind of alignment that may occur with the
request.

Note, that bdrv_get_cluster_size() is not documented to return power
of 2, still bdrv_mark_request_serialising() behaves like it is.
Also, backup uses bdi.cluster_size and is not prepared to it not being
power of 2.
So, let's establish that Qemu supports only power-of-2 clusters and
alignments.

So, alignment can't be greater than 2^30.

Finally to be safe with calculations, to not calculate different
maximums for different nodes (depending on cluster size and
request_alignment), let's simply set QEMU_ALIGN_DOWN(INT64_MAX, 2^30)
as absolute maximum bytes length for Qemu. Actually, it's not much less
than INT64_MAX.

OK, then, let's apply it to block/io.

Let's consider all block/io entry points of offset/bytes:

4 bytes/offset interface functions: bdrv_co_preadv_part(),
bdrv_co_pwritev_part(), bdrv_co_copy_range_internal() and
bdrv_co_pdiscard() and we check them all with bdrv_check_request().

We also have one entry point with only offset: bdrv_co_truncate().
Check the offset.

And one public structure: BdrvTrackedRequest. Happily, it has only
three external users:

 file-posix.c: adopted by this patch
 write-threshold.c: only read fields
 test-write-threshold.c: sets obviously small constant values

Better is to make the structure private and add corresponding
interfaces.. Still it's not obvious what kind of interface is needed
for file-posix.c. Let's keep it public but add corresponding
assertions.

After this patch we'll convert functions in block/io.c to int64_t bytes
and offset parameters. We can assume that offset/bytes pair always
satisfy new restrictions, and make
corresponding assertions where needed. If we reach some offset/bytes
point in block/io.c missing bdrv_check_request() it is considered a
bug. As well, if block/io.c modifies a offset/bytes request, expanding
it more then aligning up to request_alignment, it's a bug too.

For all io requests except for discard we keep for now old restriction
of 32bit request length.

iotest 206 output error message changed, as now test disk size is
larger than new limit. Add one more test case with new maximum disk
size to cover too-big-L1 case.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-5-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/io: bdrv_check_byte_request(): drop bdrv_is_inserted()
Vladimir Sementsov-Ogievskiy [Thu, 3 Dec 2020 22:27:12 +0000 (01:27 +0300)]
block/io: bdrv_check_byte_request(): drop bdrv_is_inserted()

Move bdrv_is_inserted() calls into callers.

We are going to make bdrv_check_byte_request() a clean thing.
bdrv_is_inserted() is not about checking the request, it's about
checking the bs. So, it should be separate.

With this patch we probably change error path for some failure
scenarios. But depending on the fact that querying too big request on
empty cdrom (or corrupted qcow2 node with no drv) will result in EIO
and not ENOMEDIUM would be very strange. More over, we are going to
move to 64bit requests, so larger requests will be allowed anyway.

More over, keeping in mind that cdrom is the only driver that has
.bdrv_is_inserted() handler it's strange that we should care so much
about it in generic block layer, intuitively we should just do read and
write, and cdrom driver should return correct errors if it is not
inserted. But it's a work for another series.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-4-vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/io: bdrv_refresh_limits(): use ERRP_GUARD
Vladimir Sementsov-Ogievskiy [Thu, 3 Dec 2020 22:27:11 +0000 (01:27 +0300)]
block/io: bdrv_refresh_limits(): use ERRP_GUARD

This simplifies following commit.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-3-vsementsov@virtuozzo.com>
Reviewed-by: Alberto Garcia <berto@igalia.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/file-posix: fix workaround in raw_do_pwrite_zeroes()
Vladimir Sementsov-Ogievskiy [Thu, 3 Dec 2020 22:27:10 +0000 (01:27 +0300)]
block/file-posix: fix workaround in raw_do_pwrite_zeroes()

We should not set overlap_bytes:

1. Don't worry: it is calculated by bdrv_mark_request_serialising() and
   will be equal to or greater than bytes anyway.

2. If the request was already aligned up to some greater alignment,
   than we may break things: we reduce overlap_bytes, and further
   bdrv_mark_request_serialising() may not help, as it will not restore
   old bigger alignment.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20201203222713.13507-2-vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agocan-host: Fix crash when 'canbus' property is not set
Kevin Wolf [Mon, 30 Nov 2020 10:56:15 +0000 (11:56 +0100)]
can-host: Fix crash when 'canbus' property is not set

Providing the 'if' property, but not 'canbus' segfaults like this:

 #0  0x0000555555b0f14d in can_bus_insert_client (bus=0x0, client=0x555556aa9af0) at ../net/can/can_core.c:88
 #1  0x00005555559c3803 in can_host_connect (ch=0x555556aa9ac0, errp=0x7fffffffd568) at ../net/can/can_host.c:62
 #2  0x00005555559c386a in can_host_complete (uc=0x555556aa9ac0, errp=0x7fffffffd568) at ../net/can/can_host.c:72
 #3  0x0000555555d52de9 in user_creatable_complete (uc=0x555556aa9ac0, errp=0x7fffffffd5c8) at ../qom/object_interfaces.c:23

Add the missing NULL check.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201130105615.21799-5-kwolf@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests/221: Discard image before qemu-img map
Max Reitz [Mon, 7 Dec 2020 15:22:45 +0000 (16:22 +0100)]
iotests/221: Discard image before qemu-img map

See the new comment for why this should be done.

I do not have a reproducer on master, but when using FUSE block exports,
this test breaks depending on the underlying filesystem (for me, it
works on tmpfs, but fails on xfs, because the block allocated by
file-posix has 16 kB there instead of 4 kB).

Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201207152245.66987-1-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agofile-posix: check the use_lock before setting the file lock
Li Feng [Mon, 7 Dec 2020 11:44:06 +0000 (19:44 +0800)]
file-posix: check the use_lock before setting the file lock

The scenario is that when accessing a volume on an NFS filesystem
without supporting the file lock,  Qemu will complain "Failed to lock
byte 100", even when setting the file.locking = off.

We should do file lock related operations only when the file.locking is
enabled, otherwise, the syscall of 'fcntl' will return non-zero.

Signed-off-by: Li Feng <fengli@smartx.com>
Message-Id: <1607341446-85506-1-git-send-email-fengli@smartx.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests/308: Add test for FUSE exports
Max Reitz [Tue, 27 Oct 2020 19:06:00 +0000 (20:06 +0100)]
iotests/308: Add test for FUSE exports

We have good coverage of the normal I/O paths now, but what remains is a
test that tests some more special cases: Exporting an image on itself
(thus turning a formatted image into a raw one), some error cases, and
non-writable and non-growable exports.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-21-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Enable fuse for many tests
Max Reitz [Tue, 27 Oct 2020 19:05:59 +0000 (20:05 +0100)]
iotests: Enable fuse for many tests

Many tests (that do not support generic protocols) can run just fine
with FUSE-exported images, so allow them to.  Note that this is no
attempt at being definitely complete.  There are some tests that might
be modified to run on FUSE, but this patch still skips them.  This patch
only tries to pick the rather low-hanging fruits.

Note that 221 and 250 only pass when .lseek is correctly implemented,
which is only possible with a libfuse that is 3.8 or newer.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-20-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Allow testing FUSE exports
Max Reitz [Tue, 27 Oct 2020 19:05:58 +0000 (20:05 +0100)]
iotests: Allow testing FUSE exports

This pretends FUSE exports are a kind of protocol.  As such, they are
always tested under the format node.  This is probably the best way to
test them, actually, because this will generate more I/O load and more
varied patterns.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-19-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Give access to the qemu-storage-daemon
Max Reitz [Tue, 27 Oct 2020 19:05:57 +0000 (20:05 +0100)]
iotests: Give access to the qemu-storage-daemon

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-18-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agostorage-daemon: Call bdrv_close_all() on exit
Max Reitz [Tue, 27 Oct 2020 19:05:56 +0000 (20:05 +0100)]
storage-daemon: Call bdrv_close_all() on exit

Otherwise, exports and block devices are not properly shut down and
closed, unless the users explicitly issues blockdev-del and
block-export-del commands for each of them.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-17-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests/287: Clean up subshell test image
Max Reitz [Tue, 27 Oct 2020 19:05:55 +0000 (20:05 +0100)]
iotests/287: Clean up subshell test image

287 creates an image in a subshell (thanks to the pipe) to see whether
that is possible with compression_type=zstd.  If _make_test_img were to
modify any global state, this global state would then be lost before we
could cleanup the image.

When using FUSE as the test protocol, this global state is important, so
clean up the image before the state is lost.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-16-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Let _make_test_img guess $TEST_IMG_FILE
Max Reitz [Tue, 27 Oct 2020 19:05:54 +0000 (20:05 +0100)]
iotests: Let _make_test_img guess $TEST_IMG_FILE

When most iotests want to create a test image that is named differently
from the default $TEST_IMG, they do something like this:

    TEST_IMG="$TEST_IMG.base" _make_test_img $options

This works fine with the "file" protocol, but not so much for anything
else: _make_test_img tries to create an image under $TEST_IMG_FILE
first, and only under $TEST_IMG if the former is not set; and on
everything but "file", $TEST_IMG_FILE is set.

There are two ways we can fix this: First, we could make all tests
adjust not only TEST_IMG, but also TEST_IMG_FILE if that is present
(e.g. with something like _set_test_img_suffix $suffix that would affect
not only TEST_IMG but also TEST_IMG_FILE, if necessary).  This is a
pretty clean solution, and this is maybe what we should have done from
the start.

But it would also require changes to most existing bash tests.  So the
alternative is this: Let _make_test_img see whether $TEST_IMG_FILE still
points to the original value.  If so, it is possible that the caller has
adjusted $TEST_IMG but not $TEST_IMG_FILE.  In such a case, we can (for
most protocols) derive the corresponding $TEST_IMG_FILE value from
$TEST_IMG value and thus work around what technically is the caller
misbehaving.

This second solution is less clean, but it is robust against people
keeping their old habit of adjusting TEST_IMG only, and requires much
less changes.  So this patch implements it.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-15-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Restrict some Python tests to file
Max Reitz [Tue, 27 Oct 2020 19:05:53 +0000 (20:05 +0100)]
iotests: Restrict some Python tests to file

Most Python tests are restricted to the file protocol (without
explicitly saying so), but these are the ones that would break
./check -fuse -qcow2.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-14-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests/091: Use _cleanup_qemu instad of "wait"
Max Reitz [Tue, 27 Oct 2020 19:05:52 +0000 (20:05 +0100)]
iotests/091: Use _cleanup_qemu instad of "wait"

If the test environment has some other child processes running (like a
storage daemon that provides a FUSE export), then "wait" will never
finish.  Use wait=yes _cleanup_qemu instead.

(We need to discard the output so there is no change to the reference
output.)

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-13-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Derive image names from $TEST_IMG
Max Reitz [Tue, 27 Oct 2020 19:05:51 +0000 (20:05 +0100)]
iotests: Derive image names from $TEST_IMG

Avoid creating images with custom filenames in $TEST_DIR, because
non-file protocols may want to keep $TEST_IMG (and all other test
images) in some other directory.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-12-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests/046: Avoid renaming images
Max Reitz [Tue, 27 Oct 2020 19:05:50 +0000 (20:05 +0100)]
iotests/046: Avoid renaming images

This generally does not work on non-file protocols.  It is better to
create the image with the final name from the start, and most tests do
this already.  Let 046 follow suit.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-11-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Use convert -n in some cases
Max Reitz [Tue, 27 Oct 2020 19:05:49 +0000 (20:05 +0100)]
iotests: Use convert -n in some cases

qemu-img convert (without -n) can often be replaced by a combination of
_make_test_img + qemu-img convert -n.  Doing so allows converting to
protocols that do not allow direct file creation, such as FUSE exports.
The only problem is that for formats other than qcow2 and qed (qcow1 at
least), this may lead to high disk usage for some reason, so we cannot
do it everywhere.

But we can do it in 028 and 089, so let us do that so they can run on
FUSE exports.  Also, in 028 this allows us to remove a 9-line comment
that used to explain why we cannot safely filter drive-backup's image
creation output.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-10-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Do not pipe _make_test_img
Max Reitz [Tue, 27 Oct 2020 19:05:48 +0000 (20:05 +0100)]
iotests: Do not pipe _make_test_img

Executing _make_test_img as part of a pipe will undo all variable
changes it has done.  As such, this could not work with FUSE (because
we want to remember all of our exports and their qemu instances).

Replace the pipe by a temporary file in 071 and 174 (the two tests that
can run on FUSE).

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-9-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoiotests: Do not needlessly filter _make_test_img
Max Reitz [Tue, 27 Oct 2020 19:05:47 +0000 (20:05 +0100)]
iotests: Do not needlessly filter _make_test_img

In most cases, _make_test_img does not need a _filter_imgfmt on top.  It
does that by itself.

(The exception is when IMGFMT has been overwritten but TEST_IMG has not.
In such cases, we do need a _filter_imgfmt on top to filter the test's
original IMGFMT from TEST_IMG.)

Signed-off-by: Max Reitz <mreitz@redhat.com>
Reviewed-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20201027190600.192171-8-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agofuse: Implement hole detection through lseek
Max Reitz [Tue, 27 Oct 2020 19:05:46 +0000 (20:05 +0100)]
fuse: Implement hole detection through lseek

This is a relatively new feature in libfuse (available since 3.8.0,
which was released in November 2019), so we have to add a dedicated
check whether it is available before making use of it.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-7-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agofuse: (Partially) implement fallocate()
Max Reitz [Tue, 27 Oct 2020 19:05:45 +0000 (20:05 +0100)]
fuse: (Partially) implement fallocate()

This allows allocating areas after the (old) EOF as part of a growing
resize, writing zeroes, and discarding.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-6-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agofuse: Allow growable exports
Max Reitz [Tue, 27 Oct 2020 19:05:44 +0000 (20:05 +0100)]
fuse: Allow growable exports

These will behave more like normal files in that writes beyond the EOF
will automatically grow the export size.

As an optimization, keep the RESIZE permission for growable exports so
we do not have to take it for every post-EOF write.  (This permission is
not released when the export is destroyed, because at that point the
BlockBackend is destroyed altogether anyway.)

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-5-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agofuse: Implement standard FUSE operations
Max Reitz [Tue, 27 Oct 2020 19:05:43 +0000 (20:05 +0100)]
fuse: Implement standard FUSE operations

This makes the export actually useful instead of only producing errors
whenever it is accessed.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-4-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agofuse: Allow exporting BDSs via FUSE
Max Reitz [Tue, 27 Oct 2020 19:05:42 +0000 (20:05 +0100)]
fuse: Allow exporting BDSs via FUSE

block-export-add type=fuse allows mounting block graph nodes via FUSE on
some existing regular file.  That file should then appears like a raw
disk image, and accesses to it result in accesses to the exported BDS.

Right now, we only implement the necessary block export functions to set
it up and shut it down.  We do not implement any access functions, so
accessing the mount point only results in errors.  This will be
addressed by a followup patch.

We keep a hash table of exported mount points, because we want to be
able to detect when users try to use a mount point twice.  This is
because we invoke stat() to check whether the given mount point is a
regular file, but if that file is served by ourselves (because it is
already used as a mount point), then this stat() would have to be served
by ourselves, too, which is impossible to do while we (as the caller)
are waiting for it to settle.  Therefore, keep track of mount point
paths to at least catch the most obvious instances of that problem.

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-3-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agomeson: Detect libfuse
Max Reitz [Tue, 27 Oct 2020 19:05:41 +0000 (20:05 +0100)]
meson: Detect libfuse

Signed-off-by: Max Reitz <mreitz@redhat.com>
Message-Id: <20201027190600.192171-2-mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/iscsi: Use lock guard macros
Gan Qixin [Thu, 3 Dec 2020 07:50:55 +0000 (15:50 +0800)]
block/iscsi: Use lock guard macros

Replace manual lock()/unlock() calls with lock guard macros
(QEMU_LOCK_GUARD/WITH_QEMU_LOCK_GUARD) in block/iscsi.

Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Message-Id: <20201203075055.127773-5-ganqixin@huawei.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/throttle-groups: Use lock guard macros
Gan Qixin [Thu, 3 Dec 2020 07:50:54 +0000 (15:50 +0800)]
block/throttle-groups: Use lock guard macros

Replace manual lock()/unlock() calls with lock guard macros
(QEMU_LOCK_GUARD/WITH_QEMU_LOCK_GUARD) in block/throttle-groups.

Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Message-Id: <20201203075055.127773-4-ganqixin@huawei.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/curl: Use lock guard macros
Gan Qixin [Thu, 3 Dec 2020 07:50:53 +0000 (15:50 +0800)]
block/curl: Use lock guard macros

Replace manual lock()/unlock() calls with lock guard macros
(QEMU_LOCK_GUARD/WITH_QEMU_LOCK_GUARD) in block/curl.

Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201203075055.127773-3-ganqixin@huawei.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoblock/accounting: Use lock guard macros
Gan Qixin [Thu, 3 Dec 2020 07:50:52 +0000 (15:50 +0800)]
block/accounting: Use lock guard macros

Replace manual lock()/unlock() calls with lock guard macros
(QEMU_LOCK_GUARD/WITH_QEMU_LOCK_GUARD) in block/accounting.

Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201203075055.127773-2-ganqixin@huawei.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
3 years agoMerge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20201210' into staging
Peter Maydell [Fri, 11 Dec 2020 15:01:05 +0000 (15:01 +0000)]
Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20201210' into staging

Split CpusAccel for tcg variants

# gpg: Signature made Fri 11 Dec 2020 01:07:33 GMT
# gpg:                using RSA key 7A481E78868B4DB6A85A05C064DF38E8AF7E215F
# gpg:                issuer "richard.henderson@linaro.org"
# gpg: Good signature from "Richard Henderson <richard.henderson@linaro.org>" [full]
# Primary key fingerprint: 7A48 1E78 868B 4DB6 A85A  05C0 64DF 38E8 AF7E 215F

* remotes/rth-gitlab/tags/pull-tcg-20201210:
  accel/tcg: rename tcg-cpus functions to match module name
  accel/tcg: split tcg_start_vcpu_thread
  accel/tcg: split CpusAccel into three TCG variants

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agoMerge remote-tracking branch 'remotes/bonzini-gitlab/tags/for-upstream' into staging
Peter Maydell [Fri, 11 Dec 2020 13:50:35 +0000 (13:50 +0000)]
Merge remote-tracking branch 'remotes/bonzini-gitlab/tags/for-upstream' into staging

* Fix for NULL segments (Bin Meng)
* Support for 32768 CPUs on x86 without IOMMU (David)
* PDEP/PEXT fix and testcase (myself)
* Remove bios_name and ram_size globals (myself)
* qemu_init rationalization (myself)
* Update kernel-doc (myself + upstream patches)
* Propagate MemTxResult across DMA and PCI functions (Philippe)
* Remove master/slave when applicable (Philippe)
* WHPX support for in-kernel irqchip (Sunil)

# gpg: Signature made Thu 10 Dec 2020 17:21:50 GMT
# gpg:                using RSA key F13338574B662389866C7682BFFBD25F78C7AE83
# gpg:                issuer "pbonzini@redhat.com"
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>" [full]
# gpg:                 aka "Paolo Bonzini <pbonzini@redhat.com>" [full]
# 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-gitlab/tags/for-upstream: (113 commits)
  scripts: kernel-doc: remove unnecessary change wrt Linux
  Revert "docs: temporarily disable the kernel-doc extension"
  scripts: kernel-doc: use :c:union when needed
  scripts: kernel-doc: split typedef complex regex
  scripts: kernel-doc: fix typedef parsing
  Revert "kernel-doc: Handle function typedefs that return pointers"
  Revert "kernel-doc: Handle function typedefs without asterisks"
  scripts: kernel-doc: try to use c:function if possible
  scripts: kernel-doc: fix line number handling
  scripts: kernel-doc: allow passing desired Sphinx C domain dialect
  scripts: kernel-doc: don't mangle with parameter list
  scripts: kernel-doc: fix typedef identification
  scripts: kernel-doc: reimplement -nofunction argument
  scripts: kernel-doc: fix troubles with line counts
  scripts: kernel-doc: use a less pedantic markup for funcs on Sphinx 3.x
  scripts: kernel-doc: make it more compatible with Sphinx 3.x
  Revert "kernel-doc: Use c:struct for Sphinx 3.0 and later"
  Revert "scripts/kerneldoc: For Sphinx 3 use c:macro for macros with arguments"
  scripts: kernel-doc: add support for typedef enum
  kernel-doc: add support for ____cacheline_aligned attribute
  ...

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agoMerge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into...
Peter Maydell [Fri, 11 Dec 2020 12:16:33 +0000 (12:16 +0000)]
Merge remote-tracking branch 'remotes/ehabkost/tags/machine-next-pull-request' into staging

Machine queue, 2020-12-10

Some patches that were queued after 5.2 soft freeze.

# gpg: Signature made Thu 10 Dec 2020 22:41:29 GMT
# gpg:                using RSA key 5A322FD5ABC4D3DBACCFD1AA2807936F984DC5A6
# gpg:                issuer "ehabkost@redhat.com"
# gpg: Good signature from "Eduardo Habkost <ehabkost@redhat.com>" [full]
# Primary key fingerprint: 5A32 2FD5 ABC4 D3DB ACCF  D1AA 2807 936F 984D C5A6

* remotes/ehabkost/tags/machine-next-pull-request:
  i386/cpu: Make the Intel PT LIP feature configurable
  sev: add sev-inject-launch-secret
  qom: code hardening - have bound checking while looping with integer value

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
3 years agos390x/cpu: Use timer_free() in the finalize function to avoid memleaks
Gan Qixin [Fri, 4 Dec 2020 08:12:09 +0000 (16:12 +0800)]
s390x/cpu: Use timer_free() in the finalize function to avoid memleaks

When running device-introspect-test, a memory leak occurred in the s390_cpu_initfn
function, this patch use timer_free() in the finalize function to fix it.

ASAN shows memory leak stack:

Direct leak of 3552 byte(s) in 74 object(s) allocated from:
    #0 0xfffeb3d4e1f0 in __interceptor_calloc (/lib64/libasan.so.5+0xee1f0)
    #1 0xfffeb36e6800 in g_malloc0 (/lib64/libglib-2.0.so.0+0x56800)
    #2 0xaaad51a8f9c4 in timer_new_full qemu/include/qemu/timer.h:523
    #3 0xaaad51a8f9c4 in timer_new qemu/include/qemu/timer.h:544
    #4 0xaaad51a8f9c4 in timer_new_ns qemu/include/qemu/timer.h:562
    #5 0xaaad51a8f9c4 in s390_cpu_initfn qemu/target/s390x/cpu.c:304
    #6 0xaaad51e00f58 in object_init_with_type qemu/qom/object.c:371
    #7 0xaaad51e0406c in object_initialize_with_type qemu/qom/object.c:515
    #8 0xaaad51e042e0 in object_new_with_type qemu/qom/object.c:729
    #9 0xaaad51e3ff40 in qmp_device_list_properties qemu/qom/qom-qmp-cmds.c:153
    #10 0xaaad51910518 in qdev_device_help qemu/softmmu/qdev-monitor.c:283
    #11 0xaaad51911918 in qmp_device_add qemu/softmmu/qdev-monitor.c:801
    #12 0xaaad51911e48 in hmp_device_add qemu/softmmu/qdev-monitor.c:916

Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Gan Qixin <ganqixin@huawei.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-Id: <20201204081209.360524-4-ganqixin@huawei.com>
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
3 years agotests/acceptance: test s390x zpci fid propagation
Cornelia Huck [Mon, 30 Nov 2020 18:02:16 +0000 (19:02 +0100)]
tests/acceptance: test s390x zpci fid propagation

Verify that a fid specified on the command line shows up correctly
as the function_id in the guest.

Signed-off-by: Cornelia Huck <cohuck@redhat.com>
[re-formatted overlong lines]
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
Message-Id: <20201130180216.15366-4-cohuck@redhat.com>

3 years agotests/acceptance: verify s390x device detection
Cornelia Huck [Mon, 30 Nov 2020 18:02:15 +0000 (19:02 +0100)]
tests/acceptance: verify s390x device detection

The kernel/initrd combination does not provide the virtio-net
driver; therefore, simply check whether the presented device type
is indeed virtio-net for the two virtio-net-{ccw,pci} devices.

Signed-off-by: Cornelia Huck <cohuck@redhat.com>
[re-formatted overlong lines]
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
Message-Id: <20201130180216.15366-3-cohuck@redhat.com>

3 years agotests/acceptance: test virtio-ccw revision handling
Cornelia Huck [Mon, 30 Nov 2020 18:02:14 +0000 (19:02 +0100)]
tests/acceptance: test virtio-ccw revision handling

The max_revision prop of virtio-ccw devices can be used to force
an older revision for compatibility handling. The easiest way to
check this is to force a device to revision 0, which turns off
virtio-1.

Signed-off-by: Cornelia Huck <cohuck@redhat.com>
[re-formatted overlong lines]
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
Message-Id: <20201130180216.15366-2-cohuck@redhat.com>

3 years agotests/acceptance: add a test for devices on s390x
Cornelia Huck [Thu, 26 Nov 2020 13:01:58 +0000 (14:01 +0100)]
tests/acceptance: add a test for devices on s390x

This adds a very basic test for checking that we present devices
in a way that Linux can consume: boot with both virtio-net-ccw and
virtio-net-pci attached and then verify that Linux is able to see
and detect these devices.

Signed-off-by: Cornelia Huck <cohuck@redhat.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Wainer dos Santos Moschetta <wainersm@redhat.com>
Message-Id: <20201126130158.1471985-1-cohuck@redhat.com>

3 years agohw/watchdog/wdt_diag288: Remove unnecessary includes
Thomas Huth [Wed, 18 Nov 2020 09:03:44 +0000 (10:03 +0100)]
hw/watchdog/wdt_diag288: Remove unnecessary includes

Neither sysbus.h nor module.h are required to compile this file.
diag288 is not a sysbus device, and module.h (for type_init) is
included eventually through qom/object.h.

Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Christian Borntraeger <borntraeger@de.ibm.com>
Message-Id: <20201118090344.243117-1-thuth@redhat.com>
[CH: tweaked description]
Signed-off-by: Cornelia Huck <cohuck@redhat.com>
3 years agosdl2: Add extra mouse buttons
Darrell Walisser [Thu, 10 Dec 2020 14:33:06 +0000 (14:33 +0000)]
sdl2: Add extra mouse buttons

Allows guest to receive mouse buttons 4 and 5, aka "SIDE" button
and "EXTRA" button

Signed-off-by: Darrell Walisser <darrell.walisser@gmail.com>
Message-id: GNUALCfDBqhCUvLwBNJaKqxcPewMtlqCnixk8xTrgI@ubuntu
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
3 years agoui/vnc: Add missing lock for send_color_map
Peng Liang [Mon, 16 Nov 2020 14:13:38 +0000 (22:13 +0800)]
ui/vnc: Add missing lock for send_color_map

vnc_write() should be locked after the RFB protocol is initialized.

Fixes: 0c426e4534b4 ("vnc: Add support for color map")
Cc: qemu-stable@nongnu.org
Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Peng Liang <liangpeng10@huawei.com>
Message-id: 20201116141338.148911-1-liangpeng10@huawei.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
3 years agovnc: add alpha cursor support
Gerd Hoffmann [Tue, 8 Dec 2020 11:57:34 +0000 (12:57 +0100)]
vnc: add alpha cursor support

There is a new vnc extension for cursors with an alpha channel.  Use
it if supported by the vnc client, prefer it over the "rich cursor"
extension which supports only a bitmask for transparency.

This is a visible improvement especially on modern desktops which
actually use the alpha channel when defining cursors.

https://github.com/rfbproto/rfbproto/blob/master/rfbproto.rst#cursor-with-alpha-pseudo-encoding

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Tested-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-7-kraxel@redhat.com

3 years agovnc: add pseudo encodings
Gerd Hoffmann [Tue, 8 Dec 2020 11:57:33 +0000 (12:57 +0100)]
vnc: add pseudo encodings

Add #defines for two new pseudo encodings:
 * cursor with alpha channel.
 * extended desktop resize.

https://github.com/rfbproto/rfbproto/blob/master/rfbproto.rst#pseudo-encodings

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-6-kraxel@redhat.com

3 years agovnc: drop unused copyrect feature
Gerd Hoffmann [Tue, 8 Dec 2020 11:57:32 +0000 (12:57 +0100)]
vnc: drop unused copyrect feature

vnc stopped using the copyrect pseudo encoding in 2017, in commit
50628d3479e4 ("cirrus/vnc: zap bitblit support from console code.")
So we can drop the now unused copyrect feature bit.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-5-kraxel@redhat.com

3 years agovnc: use enum for features
Gerd Hoffmann [Tue, 8 Dec 2020 11:57:31 +0000 (12:57 +0100)]
vnc: use enum for features

Use an enum for the vnc feature bits.  That way they are enumerated
automatically and we don't have to do that manually when adding or
removing features.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Message-id: 20201208115737.18581-4-kraxel@redhat.com

3 years agoconsole: allow con==NULL in dpy_{get, set}_ui_info and dpy_ui_info_supported
Gerd Hoffmann [Tue, 8 Dec 2020 11:57:30 +0000 (12:57 +0100)]
console: allow con==NULL in dpy_{get, set}_ui_info and dpy_ui_info_supported

Use active_console in that case like we do in many other places.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-3-kraxel@redhat.com

3 years agoconsole: drop qemu_console_get_ui_info
Gerd Hoffmann [Tue, 8 Dec 2020 11:57:29 +0000 (12:57 +0100)]
console: drop qemu_console_get_ui_info

Unused and duplicate (there is dpy_get_ui_info).

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-id: 20201208115737.18581-2-kraxel@redhat.com

3 years agoaccel/tcg: rename tcg-cpus functions to match module name
Claudio Fontana [Thu, 15 Oct 2020 14:32:17 +0000 (16:32 +0200)]
accel/tcg: rename tcg-cpus functions to match module name

Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20201015143217.29337-4-cfontana@suse.de>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
3 years agoaccel/tcg: split tcg_start_vcpu_thread
Claudio Fontana [Thu, 15 Oct 2020 14:32:16 +0000 (16:32 +0200)]
accel/tcg: split tcg_start_vcpu_thread

after the initial split into 3 tcg variants, we proceed to also
split tcg_start_vcpu_thread.

We actually split it in 2 this time, since the icount variant
just uses the round robin function.

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Message-Id: <20201015143217.29337-3-cfontana@suse.de>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
3 years agoaccel/tcg: split CpusAccel into three TCG variants
Claudio Fontana [Thu, 15 Oct 2020 14:32:15 +0000 (16:32 +0200)]
accel/tcg: split CpusAccel into three TCG variants

split up the CpusAccel tcg_cpus into three TCG variants:

tcg_cpus_rr (single threaded, round robin cpus)
tcg_cpus_icount (same as rr, but with instruction counting enabled)
tcg_cpus_mttcg (multi-threaded cpus)

Suggested-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Claudio Fontana <cfontana@suse.de>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20201015143217.29337-2-cfontana@suse.de>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
3 years agoi386/cpu: Make the Intel PT LIP feature configurable
Luwei Kang [Wed, 2 Dec 2020 10:10:42 +0000 (18:10 +0800)]
i386/cpu: Make the Intel PT LIP feature configurable

The current implementation will disable the guest Intel PT feature
if the Intel PT LIP feature is supported on the host, but the LIP
feature is comming soon(e.g. SnowRidge and later).

This patch will make the guest LIP feature configurable and Intel
PT feature can be enabled in guest when the guest LIP status same
with the host.

Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Message-Id: <20201202101042.11967-1-luwei.kang@intel.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
3 years agosev: add sev-inject-launch-secret
Tobin Feldman-Fitzthum [Tue, 27 Oct 2020 17:03:03 +0000 (13:03 -0400)]
sev: add sev-inject-launch-secret

AMD SEV allows a guest owner to inject a secret blob
into the memory of a virtual machine. The secret is
encrypted with the SEV Transport Encryption Key and
integrity is guaranteed with the Transport Integrity
Key. Although QEMU facilitates the injection of the
launch secret, it cannot access the secret.

Signed-off-by: Tobin Feldman-Fitzthum <tobin@linux.ibm.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Brijesh Singh <brijesh.singh@amd.com>
Message-Id: <20201027170303.47550-1-tobin@linux.ibm.com>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
3 years agoqom: code hardening - have bound checking while looping with integer value
Ani Sinha [Mon, 21 Sep 2020 09:33:25 +0000 (15:03 +0530)]
qom: code hardening - have bound checking while looping with integer value

Object property insertion code iterates over an integer to get an unused
index that can be used as an unique name for an object property. This loop
increments the integer value indefinitely. Although very unlikely, this can
still cause an integer overflow.
In this change, we fix the above code by checking against INT16_MAX and making
sure that the interger index does not overflow beyond that value. If no
available index is found, the code would cause an assertion failure. This
assertion failure is necessary because the callers of the function do not check
the return value for NULL.

Signed-off-by: Ani Sinha <ani@anisinha.ca>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Daniel P. Berrangé <berrange@redhat.com>
Message-Id: <20200921093325.25617-1-ani@anisinha.ca>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
3 years agoscripts: kernel-doc: remove unnecessary change wrt Linux
Paolo Bonzini [Mon, 30 Nov 2020 12:21:12 +0000 (07:21 -0500)]
scripts: kernel-doc: remove unnecessary change wrt Linux

A comment in kernel-doc mentions QEMU's qatomic_set macro, but since
this code originated in Linux we should just revert it and stay as close
to the kernel's copy of the script as possible.

The change was introduced (more or less unintentionally) in QEMU commit
commit d73415a31547, which did a global search-and-replace of QEMU's
atomic access macros.

Suggested-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoRevert "docs: temporarily disable the kernel-doc extension"
Paolo Bonzini [Mon, 30 Nov 2020 12:17:45 +0000 (07:17 -0500)]
Revert "docs: temporarily disable the kernel-doc extension"

This reverts commit fd68a72875cf318f4310726f842139119c5f45d5.  We're
done with the update of kernel-doc and we can restore kernel-doc's
functionality.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: use :c:union when needed
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:12 +0000 (17:53 +0100)]
scripts: kernel-doc: use :c:union when needed

Sphinx C domain code after 3.2.1 will start complaning if :c:struct
would be used for an union type:

.../Documentation/gpu/drm-kms-helpers:352: ../drivers/video/hdmi.c:851: WARNING: C 'identifier' cross-reference uses wrong tag: reference name is 'union hdmi_infoframe' but found name is 'struct hdmi_infoframe'. Full reference name is 'union hdmi_infoframe'. Full found name is 'struct hdmi_infoframe'.

So, let's address this issue too in advance, in order to
avoid future issues.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/6e4ec3eec914df62389a299797a3880ae4490f35.1603791716.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-30-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: split typedef complex regex
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:11 +0000 (17:53 +0100)]
scripts: kernel-doc: split typedef complex regex

The typedef regex for function prototypes are very complex.
Split them into 3 separate regex and then join them using
qr.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/3a4af999a0d62d4ab9dfae1cdefdfcad93383356.1603792384.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-29-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: fix typedef parsing
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:10 +0000 (17:53 +0100)]
scripts: kernel-doc: fix typedef parsing

The include/linux/genalloc.h file defined this typedef:

typedef unsigned long (*genpool_algo_t)(unsigned long *map,unsigned long size,unsigned long start,unsigned int nr,void *data, struct gen_pool *pool, unsigned long start_addr);

Because it has a type composite of two words (unsigned long),
the parser gets the typedef name wrong:

.. c:macro:: long

   **Typedef**: Allocation callback function type definition

Fix the regex in order to accept composite types when
defining a typedef for a function pointer.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/328e8018041cc44f7a1684e57f8d111230761c4f.1603792384.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-28-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoRevert "kernel-doc: Handle function typedefs that return pointers"
Paolo Bonzini [Tue, 17 Nov 2020 16:53:09 +0000 (17:53 +0100)]
Revert "kernel-doc: Handle function typedefs that return pointers"

This reverts commit 19ab6044be0c55d529e875e3ee16fdd5c3b54d33.
We will replace the commit with the fix from Linux.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-27-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoRevert "kernel-doc: Handle function typedefs without asterisks"
Paolo Bonzini [Tue, 17 Nov 2020 16:53:08 +0000 (17:53 +0100)]
Revert "kernel-doc: Handle function typedefs without asterisks"

This reverts commit 3cd3c5193cde5242e243c25759f85802e267994f.
We will replace the commit with the fix from Linux.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-26-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: try to use c:function if possible
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:07 +0000 (17:53 +0100)]
scripts: kernel-doc: try to use c:function if possible

There are a few namespace clashes by using c:macro everywhere:

basically, when using it, we can't have something like:

.. c:struct:: pwm_capture

.. c:macro:: pwm_capture

So, we need to use, instead:

.. c:function:: int pwm_capture (struct pwm_device * pwm, struct pwm_capture * result, unsigned long timeout)

for the function declaration.

The kernel-doc change was proposed by Jakob Lykke Andersen here:

https://github.com/jakobandersen/linux_docs/commit/6fd2076ec001cca7466857493cd678df4dfe4a65

Although I did a different implementation.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-25-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: fix line number handling
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:06 +0000 (17:53 +0100)]
scripts: kernel-doc: fix line number handling

Address several issues related to pointing to the wrong line
number:

1) ensure that line numbers will always be initialized

   When section is the default (Description), the line number
   is not initializing, producing this:

$ ./scripts/kernel-doc --enable-lineno ./drivers/media/v4l2-core/v4l2-mem2mem.c|less

**Description**

#define LINENO 0
In case of streamoff or release called on any context,
1] If the context is currently running, then abort job will be called
2] If the context is queued, then the context will be removed from
   the job_queue

  Which is not right. Ensure that the line number will always
  be there. After applied, the result now points to the right location:

**Description**

#define LINENO 410
In case of streamoff or release called on any context,
1] If the context is currently running, then abort job will be called
2] If the context is queued, then the context will be removed from
   the job_queue

2) The line numbers for function prototypes are always + 1,
   because it is taken at the line after handling the prototype.
   Change the logic to point to the next line after the /** */
   block;

3) The "DOC:" line number should point to the same line as this
   markup is found, and not to the next one.

Probably part of the issues were due to a but that was causing
the line number offset to be incremented by one, if --export
were used.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-24-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: allow passing desired Sphinx C domain dialect
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:05 +0000 (17:53 +0100)]
scripts: kernel-doc: allow passing desired Sphinx C domain dialect

When kernel-doc is called via kerneldoc.py, there's no need to
auto-detect the Sphinx version, as the Sphinx module already
knows it. So, add an optional parameter to allow changing the
Sphinx dialect.

As kernel-doc can also be manually called, keep the auto-detection
logic if the parameter was not specified. On such case, emit
a warning if sphinx-build can't be found at PATH.

I ended using a suggestion from Joe for using a more readable
regex, instead of using a complex one with a hidden group like:

m/^(\d+)\.(\d+)(?:\.?(\d+)?)/

in order to get the optional <patch> argument.

Thanks-to: Joe Perches <joe@perches.com>
Suggested-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-23-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: don't mangle with parameter list
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:04 +0000 (17:53 +0100)]
scripts: kernel-doc: don't mangle with parameter list

While kernel-doc needs to parse parameters in order to
identify its name, it shouldn't be touching the type,
as parsing it is very difficult, and errors happen.

One current error is when parsing this parameter:

const u32 (*tab)[256]

Found at ./lib/crc32.c, on this function:

u32 __pure crc32_be_generic (u32 crc, unsigned char const *p, size_t len, const u32 (*tab)[256], u32 polynomial);

The current logic mangles it, producing this output:

const u32 ( *tab

That's something that it is not recognizeable.

So, instead, let's push the argument as-is, and use it
when printing the function prototype and when describing
each argument.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-22-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: fix typedef identification
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:03 +0000 (17:53 +0100)]
scripts: kernel-doc: fix typedef identification

Some typedef expressions are output as normal functions.

As we need to be clearer about the type with Sphinx 3.x,
detect such cases.

While here, fix a wrongly-indented block.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-21-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: reimplement -nofunction argument
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:02 +0000 (17:53 +0100)]
scripts: kernel-doc: reimplement -nofunction argument

Right now, the build system doesn't use -nofunction, as
it is pretty much useless, because it doesn't consider
the other output modes (extern, internal), working only
with all.

Also, it is limited to exclude functions.

Re-implement it in order to allow excluding any symbols from
the document output, no matter what mode is used.

The parameter was also renamed to "-nosymbol", as it express
better its meaning.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-20-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: fix troubles with line counts
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:01 +0000 (17:53 +0100)]
scripts: kernel-doc: fix troubles with line counts

There's currently a bug with the way kernel-doc script
counts line numbers that can be seen with:

$ ./scripts/kernel-doc -rst  -enable-lineno include/linux/math64.h >all && ./scripts/kernel-doc -rst -internal -enable-lineno include/linux/math64.h >int && diff -U0 int all

--- int 2020-09-28 12:58:08.927486808 +0200
+++ all 2020-09-28 12:58:08.905486845 +0200
@@ -1 +1 @@
-#define LINENO 27
+#define LINENO 26
@@ -3 +3 @@
-#define LINENO 16
+#define LINENO 15
@@ -9 +9 @@
-#define LINENO 17
+#define LINENO 16
...

This is happening with perl version 5.30.3, but I'm not
so sure if this is a perl bug, or if this is due to something
else.

In any case, fixing it is easy. Basically, when "-internal"
parameter is used, the process_export_file() function opens the
handle "IN". This makes the line number to be incremented, as the
handler for the main open is also "IN".

Fix the problem by using a different handler for the
main open().

While here, add a missing close for it.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-19-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: use a less pedantic markup for funcs on Sphinx 3.x
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:53:00 +0000 (17:53 +0100)]
scripts: kernel-doc: use a less pedantic markup for funcs on Sphinx 3.x

Unfortunately, Sphinx 3.x parser for c functions is too pedantic:

https://github.com/sphinx-doc/sphinx/issues/8241

While it could be relaxed with some configurations, there are
several corner cases that it would make it hard to maintain,
and will require teaching conf.py about several macros.

So, let's instead use the :c:macro notation. This will
produce an output that it is not as nice as currently, but it
should still be acceptable, and will provide cross-references,
removing thousands of warnings when building with newer
versions of Sphinx.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-18-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: make it more compatible with Sphinx 3.x
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:52:59 +0000 (17:52 +0100)]
scripts: kernel-doc: make it more compatible with Sphinx 3.x

With Sphinx 3.x, the ".. c:type:" tag was changed to accept either:

.. c:type:: typedef-like declaration
.. c:type:: name

Using it for other types (including functions) don't work anymore.

So, there are newer tags for macro, enum, struct, union, and others,
which doesn't exist on older versions.

Add a check for the Sphinx version and change the produced tags
accordingly.

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-17-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoRevert "kernel-doc: Use c:struct for Sphinx 3.0 and later"
Paolo Bonzini [Tue, 17 Nov 2020 16:52:58 +0000 (17:52 +0100)]
Revert "kernel-doc: Use c:struct for Sphinx 3.0 and later"

This reverts commit 152d1967f650f67b7ece3a5dda77d48069d72647.
We will replace the commit with the fix from Linux.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-16-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoRevert "scripts/kerneldoc: For Sphinx 3 use c:macro for macros with arguments"
Paolo Bonzini [Tue, 17 Nov 2020 16:52:57 +0000 (17:52 +0100)]
Revert "scripts/kerneldoc: For Sphinx 3 use c:macro for macros with arguments"

This reverts commit 92bb29f9b2c3d4a98eef5f0db935d4be291eec72.
We will replace the commit with the fix from Linux.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-15-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agoscripts: kernel-doc: add support for typedef enum
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:52:56 +0000 (17:52 +0100)]
scripts: kernel-doc: add support for typedef enum

The PHY kernel-doc markup has gained support for documenting
a typedef enum.

However, right now the parser was not prepared for it.

So, add support for parsing it.

Fixes: 4069a572d423 ("net: phy: Document core PHY structures")
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-14-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokernel-doc: add support for ____cacheline_aligned attribute
Jonathan Cameron [Tue, 17 Nov 2020 16:52:55 +0000 (17:52 +0100)]
kernel-doc: add support for ____cacheline_aligned attribute

Subroutine dump_struct uses type attributes to check if the struct
syntax is valid. Then, it removes all attributes before using it for
output. `____cacheline_aligned` is an attribute that is
not included in both steps. Add it, since it is used by kernel structs.

Based on previous patch to add ____cacheline_aligned_in_smp.
Motivated by patches to reorder this attribute to before the
variable name.   Whilst we could do that in all cases, that would
be a massive change and it is more common in the kernel to place
this particular attribute after the variable name. A quick grep
suggests approximately 400 instances of which 341 have this
attribute just before a semicolon and hence after the variable name.

Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Lee Jones <lee.jones@linaro.org>
Link: https://lore.kernel.org/r/20200910185415.653139-1-jic23@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-13-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
3 years agokernel-doc: include line numbers for function prototypes
Mauro Carvalho Chehab [Tue, 17 Nov 2020 16:52:54 +0000 (17:52 +0100)]
kernel-doc: include line numbers for function prototypes

This should solve bad error reports like this one:

./include/linux/iio/iio.h:0: WARNING: Unknown target name: "devm".

Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Link: https://lore.kernel.org/r/56eed0ba50cd726236acd12b11b55ce54854c5ea.1599660067.git.mchehab+huawei@kernel.org
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Message-Id: <20201117165312.118257-12-pbonzini@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>