qapi/parser: factor parsing routine into method

For the sake of keeping __init__ smaller (and treating it more like a
gallery of what state variables we can expect to see), put the actual
parsing action into a parse method. It remains invoked from the init
method to reduce churn.

To accomplish this, @previously_included becomes the private data
member ._included, and the filename is stashed as ._fname.

Add any missing declarations to the init method, and group them by
function so they can be understood quickly at a glance.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210519183951.3946870-5-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/source: Remove line number from QAPISourceInfo initializer

With the QAPISourceInfo(None, None, None) construct gone, there's no
longer any reason to have to specify that a file starts on the first
line. Remove it from the initializer and default it to 1.

Remove the last vestiges where we check for 'line' being unset, that
can't happen, now.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210519183951.3946870-4-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi: Add test for nonexistent schema file

This tests the error-return pathway introduced in the previous commit.
(Thanks to Paolo for the help with the Meson magic.)

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210519183951.3946870-3-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

qapi/parser: Don't try to handle file errors

Fixes: f5d4361cda
Fixes: 52a474180a
Fixes: 46f49468c6

Remove the try/except block that handles file-opening errors in
QAPISchemaParser.__init__() and add one each to
QAPISchemaParser._include() and QAPISchema.__init__() respectively.

This simultaneously fixes the typing of info.fname (f5d4361cda), A
static typing violation in test-qapi (46f49468c6), and a regression of
an error message (52a474180a).

The short-ish version of what motivates this patch is:

- It's hard to write a good error message in the init method,
  because we need to determine the context of our caller to do so.
  It's easier to just let the caller write the message.
- We don't want to allow QAPISourceInfo(None, None, None) to exist. The
  typing introduced by commit f5d4361cda types the 'fname' field as
  (non-optional) str, which was premature until the removal of this
  construct.
- Errors made using such an object are currently incorrect (since
  52a474180a)
- It's not technically a semantic error if we cannot open the schema.
- There are various typing constraints that make mixing these two cases
  undesirable for a single special case.
- test-qapi's code handling an fname of 'None' is now dead, drop it.
  Additionally, Not all QAPIError objects have an 'info' field (since
  46f49468), so deleting this stanza corrects a typing oversight in
  test-qapi introduced by that commit.

Other considerations:

- open() is moved to a 'with' block to ensure file pointers are
  cleaned up deterministically.
- Python 3.3 deprecated IOError and made it a synonym for OSError.
  Avoid the misleading perception these exception handlers are
  narrower than they really are.

The long version:

The error message here is incorrect (since commit 52a474180a):

> python3 qapi-gen.py 'fake.json'
qapi-gen.py: qapi-gen.py: can't read schema file 'fake.json': No such file or directory

In pursuing it, we find that QAPISourceInfo has a special accommodation
for when there's no filename. Meanwhile, the intent when QAPISourceInfo
was typed (f5d4361cda) was non-optional 'str'. This usage was
overlooked.

To remove this, I'd want to avoid having a "fake" QAPISourceInfo
object. I also don't want to explicitly begin accommodating
QAPISourceInfo itself being None, because we actually want to eventually
prove that this can never happen -- We don't want to confuse "The file
isn't open yet" with "This error stems from a definition that wasn't
defined in any file".

(An earlier series tried to create a dummy info object, but it was tough
to prove in review that it worked correctly without creating new
regressions. This patch avoids that distraction. We would like to first
prove that we never raise QAPISemError for any built-in object before we
add "special" info objects. We aren't ready to do that yet.)

So, which way out of the labyrinth?

Here's one way: Don't try to handle errors at a level with "mixed"
semantic contexts; i.e. don't mix inclusion errors (should report a
source line where the include was triggered) and command line errors
(where we specified a file we couldn't read).

Remove the error handling from the initializer of the parser. Pythonic!
Now it's the caller's job to figure out what to do about it. Handle the
error in QAPISchemaParser._include() instead, where we can write a
targeted error message where we are guaranteed to have an 'info' context
to report with.

The root level error can similarly move to QAPISchema.__init__(), where
we know we'll never have an info context to report with, so we use a
more abstract error type.

Now the error looks sensible again:

> python3 qapi-gen.py 'fake.json'
qapi-gen.py: can't read schema file 'fake.json': No such file or directory

With these error cases separated, QAPISourceInfo can be solidified as
never having placeholder arguments that violate our desired types. Clean
up test-qapi along similar lines.

Signed-off-by: John Snow <jsnow@redhat.com>
Message-Id: <20210519183951.3946870-2-jsnow@redhat.com>
Reviewed-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Markus Armbruster <armbru@redhat.com>

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

Block layer patches

- vhost-user-blk: Fix error handling during initialisation
- Add test cases for the vhost-user-blk export
- Fix leaked Transaction objects
- qcow2: Expose dirty bit in 'qemu-img info'

# gpg: Signature made Tue 18 May 2021 11:57:46 BST
# 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:
  vhost-user-blk: Check that num-queues is supported by backend
  virtio: Fail if iommu_platform is requested, but unsupported
  vhost-user-blk: Get more feature flags from vhost device
  vhost-user-blk: Improve error reporting in realize
  vhost-user-blk: Don't reconnect during initialisation
  vhost-user-blk: Make sure to set Error on realize failure
  vhost-user-blk-test: test discard/write zeroes invalid inputs
  tests/qtest: add multi-queue test case to vhost-user-blk-test
  test: new qTest case to test the vhost-user-blk-server
  block/export: improve vu_blk_sect_range_ok()
  block: Fix Transaction leak in bdrv_reopen_multiple()
  block: Fix Transaction leak in bdrv_root_attach_child()
  qcow2: set bdi->is_dirty

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

Merge remote-tracking branch 'remotes/vivier2/tags/linux-user-for-6.1-pull-request' into staging

linux-user pull request 20210517

- alpha sigaction fixes/cleanups
- s390x sigaction fixes/cleanup
- sparc sigaction fixes/cleanup
- s390x core dumping support
- core dump fix (app name)
- arm  fpa11 fix and cleanup
- strace fixes (unshare(), llseek())
- fix copy_file_range()
- use GDateTime
- Remove dead code

# gpg: Signature made Tue 18 May 2021 06:31:12 BST
# 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/vivier2/tags/linux-user-for-6.1-pull-request: (59 commits)
  linux-user/elfload: add s390x core dumping support
  linux-user/elfload: fix filling psinfo->pr_psargs
  linux-user: Tidy TARGET_NR_rt_sigaction
  linux-user/alpha: Share code for TARGET_NR_sigaction
  linux-user/alpha: Define TARGET_ARCH_HAS_KA_RESTORER
  linux-user: Honor TARGET_ARCH_HAS_SA_RESTORER in do_syscall
  linux-user: Pass ka_restorer to do_sigaction
  linux-user/alpha: Rename the sigaction restorer field
  linux-user/alpha: Fix rt sigframe return
  linux-user: use GDateTime for formatting timestamp for core file
  linux-user: Fix erroneous conversion in copy_file_range
  linux-user: Add copy_file_range to strace.list
  linux-user/s390x: Handle vector regs in signal stack
  linux-user/s390x: Clean up signal.c
  linux-user/s390x: Add build asserts for sigset sizes
  linux-user/s390x: Fix frame_addr corruption in setup_frame
  linux-user/s390x: Add stub sigframe argument for last_break
  linux-user/s390x: Set psw.mask properly for the signal handler
  linux-user/s390x: Clean up single-use gotos in signal.c
  linux-user/s390x: Tidy save_sigregs
  ...

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

Merge remote-tracking branch 'remotes/nvme/tags/nvme-next-pull-request' into staging

emulated nvme updates

* various fixes (Gollu Appalanaidu)
* refactoring (me)
* move to hw/nvme from hw/block (me)

# gpg: Signature made Mon 17 May 2021 10:16:01 BST
# gpg:                using RSA key 522833AA75E2DCE6A24766C04DE1AF316D4F0DE9
# gpg: Good signature from "Klaus Jensen <its@irrelevant.dk>" [unknown]
# gpg:                 aka "Klaus Jensen <k.jensen@samsung.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: DDCA 4D9C 9EF9 31CC 3468  4272 63D5 6FC5 E55D A838
#      Subkey fingerprint: 5228 33AA 75E2 DCE6 A247  66C0 4DE1 AF31 6D4F 0DE9

* remotes/nvme/tags/nvme-next-pull-request:
  hw/nvme: move nvme emulation out of hw/block
  hw/block/nvme: move zoned constraints checks
  hw/block/nvme: remove irrelevant zone resource checks
  hw/block/nvme: remove num_namespaces member
  hw/block/nvme: streamline namespace array indexing
  hw/block/nvme: add metadata offset helper
  hw/block/nvme: cache lba and ms sizes
  hw/block/nvme: replace nvme_ns_status
  hw/block/nvme: remove non-shared defines from header file
  hw/block/nvme: cleanup includes
  hw/block/nvme: consolidate header files
  hw/block/nvme: rename __nvme_select_ns_iocs
  hw/block/nvme: rename __nvme_advance_zone_wp
  hw/block/nvme: rename __nvme_zrm_open
  hw/block/nvme: align with existing style
  hw/block/nvme: function formatting fix
  hw/block/nvme: fix io-command set profile feature
  hw/block/nvme: consider metadata read aio return value in compare
  hw/block/nvme: rename reserved fields declarations
  hw/block/nvme: remove redundant invalid_lba_range trace

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

vhost-user-blk: Check that num-queues is supported by backend

Creating a device with a number of queues that isn't supported by the
backend is pointless, the device won't work properly and the error
messages are rather confusing.

Just fail to create the device if num-queues is higher than what the
backend supports.

Since the relationship between num-queues and the number of virtqueues
depends on the specific device, this is an additional value that needs
to be initialised by the device. For convenience, allow leaving it 0 if
the check should be skipped. This makes sense for vhost-user-net where
separate vhost devices are used for the queues and custom initialisation
code is needed to perform the check.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1935031
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-7-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

virtio: Fail if iommu_platform is requested, but unsupported

Commit 2943b53f6 (' virtio: force VIRTIO_F_IOMMU_PLATFORM') made sure
that vhost can't just reject VIRTIO_F_IOMMU_PLATFORM when it was
requested. However, just adding it back to the negotiated flags isn't
right either because it promises support to the guest that the device
actually doesn't support. One example of a vhost-user device that
doesn't have support for the flag is the vhost-user-blk export of QEMU.

Instead of successfully creating a device that doesn't work, just fail
to plug the device when it doesn't support the feature, but it was
requested. This results in much clearer error messages.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1935019
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-6-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vhost-user-blk: Get more feature flags from vhost device

VIRTIO_F_RING_PACKED and VIRTIO_F_IOMMU_PLATFORM need to be supported by
the vhost device, otherwise advertising it to the guest doesn't result
in a working configuration. They are currently not supported by the
vhost-user-blk export in QEMU.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1935020
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-5-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vhost-user-blk: Improve error reporting in realize

Now that vhost_user_blk_connect() is not called from an event handler
any more, but directly from vhost_user_blk_device_realize(), we can
actually make use of Error again instead of calling error_report() in
the inner function and setting a more generic and therefore less useful
error message in realize() itself.

With Error, the callers are responsible for adding context if necessary
(such as the "-device" option the error refers to). Additional prefixes
are redundant and better omitted.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Message-Id: <20210429171316.162022-4-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vhost-user-blk: Don't reconnect during initialisation

This is a partial revert of commits 77542d43149 and bc79c87bcde.

Usually, an error during initialisation means that the configuration was
wrong. Reconnecting won't make the error go away, but just turn the
error condition into an endless loop. Avoid this and return errors
again.

Additionally, calling vhost_user_blk_disconnect() from the chardev event
handler could result in use-after-free because none of the
initialisation code expects that the device could just go away in the
middle. So removing the call fixes crashes in several places.

For example, using a num-queues setting that is incompatible with the
backend would result in a crash like this (dereferencing dev->opaque,
which is already NULL):

 #0  0x0000555555d0a4bd in vhost_user_read_cb (source=0x5555568f4690, condition=(G_IO_IN | G_IO_HUP), opaque=0x7fffffffcbf0) at ../hw/virtio/vhost-user.c:313
 #1  0x0000555555d950d3 in qio_channel_fd_source_dispatch (source=0x555557c3f750, callback=0x555555d0a478 <vhost_user_read_cb>, user_data=0x7fffffffcbf0) at ../io/channel-watch.c:84
 #2  0x00007ffff7b32a9f in g_main_context_dispatch () at /lib64/libglib-2.0.so.0
 #3  0x00007ffff7b84a98 in g_main_context_iterate.constprop () at /lib64/libglib-2.0.so.0
 #4  0x00007ffff7b32163 in g_main_loop_run () at /lib64/libglib-2.0.so.0
 #5  0x0000555555d0a724 in vhost_user_read (dev=0x555557bc62f8, msg=0x7fffffffcc50) at ../hw/virtio/vhost-user.c:402
 #6  0x0000555555d0ee6b in vhost_user_get_config (dev=0x555557bc62f8, config=0x555557bc62ac "", config_len=60) at ../hw/virtio/vhost-user.c:2133
 #7  0x0000555555d56d46 in vhost_dev_get_config (hdev=0x555557bc62f8, config=0x555557bc62ac "", config_len=60) at ../hw/virtio/vhost.c:1566
 #8  0x0000555555cdd150 in vhost_user_blk_device_realize (dev=0x555557bc60b0, errp=0x7fffffffcf90) at ../hw/block/vhost-user-blk.c:510
 #9  0x0000555555d08f6d in virtio_device_realize (dev=0x555557bc60b0, errp=0x7fffffffcff0) at ../hw/virtio/virtio.c:3660

Note that this removes the ability to reconnect during initialisation
(but not during operation) when there is no permanent error, but the
backend restarts, as the implementation was buggy. This feature can be
added back in a follow-up series after changing error paths to
distinguish cases where retrying could help from cases with permanent
errors.

Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210429171316.162022-3-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vhost-user-blk: Make sure to set Error on realize failure

We have to set errp before jumping to virtio_err, otherwise the caller
(virtio_device_realize()) will take this as success and crash when it
later tries to access things that we've already freed in the error path.

Fixes: 77542d431491788d1e8e79d93ce10172ef207775
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210429171316.162022-2-kwolf@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Acked-by: Raphael Norwitz <raphael.norwitz@nutanix.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

vhost-user-blk-test: test discard/write zeroes invalid inputs

Exercise input validation code paths in
block/export/vhost-user-blk-server.c.

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210309094106.196911-5-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210322092327.150720-4-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

tests/qtest: add multi-queue test case to vhost-user-blk-test

Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210309094106.196911-4-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210322092327.150720-3-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

test: new qTest case to test the vhost-user-blk-server

This test case has the same tests as tests/virtio-blk-test.c except for
tests have block_resize. Since the vhost-user-blk export only serves one
client one time, two exports are started by qemu-storage-daemon for the
hotplug test.

Suggested-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: Coiby Xu <coiby.xu@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210309094106.196911-3-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210322092327.150720-2-stefanha@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-tcg-20210516' into staging

Minor MAINTAINERS update.
Tweak to includes.
Add tcg_constant_tl.
Improve constant pool dump.

# gpg: Signature made Sun 16 May 2021 15:08:42 BST
# 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-20210516:
  accel/tcg: Align data dumped at end of TB
  tcg: Add tcg_constant_tl
  exec/gen-icount.h: Add missing "exec/exec-all.h" include
  MAINTAINERS: Add include/exec/gen-icount.h to 'Main Loop' section

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

block/export: improve vu_blk_sect_range_ok()

The checks in vu_blk_sect_range_ok() assume VIRTIO_BLK_SECTOR_SIZE is
equal to BDRV_SECTOR_SIZE. This is true, but let's add a
QEMU_BUILD_BUG_ON() to make it explicit.

We might as well check that the request buffer size is a multiple of
VIRTIO_BLK_SECTOR_SIZE while we're at it.

Suggested-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
Message-Id: <20210331142727.391465-1-stefanha@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Fix Transaction leak in bdrv_reopen_multiple()

Like other error paths, this one needs to call tran_finalize() and clean
up the BlockReopenQueue, too.

Fixes: CID 1452772
Fixes: 72373e40fbc7e4218061a8211384db362d3e7348
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210503110555.24001-3-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

block: Fix Transaction leak in bdrv_root_attach_child()

The error path needs to call tran_finalize(), too.

Fixes: CID 1452773
Fixes: 548a74c0dbc858edd1a7ee3045b5f2fe710bd8b1
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Message-Id: <20210503110555.24001-2-kwolf@redhat.com>
Reviewed-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

qcow2: set bdi->is_dirty

Set bdi->is_dirty, so that qemu-img info could show dirty flag.

After this commit the following check will show '"dirty-flag": true':

./build/qemu-img create -f qcow2 -o lazy_refcounts=on x 1M
./build/qemu-io x
qemu-io> write 0 1M

 After "write" command success, kill the qemu-io process:

kill -9 <qemu-io pid>

./build/qemu-img info --output=json x

This will show '"dirty-flag": true' among other things. (before this
commit it shows '"dirty-flag": false')

Note, that qcow2's dirty-bit is not a "dirty bit for the image". It
only protects qcow2 lazy refcounts feature. So, there are a lot of
conditions when qcow2 session may be not closed correctly, but bit is
0. Still, when bit is set, the last session is definitely not finished
correctly and it's better to report it.

Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
Message-Id: <20210504160656.462836-1-vsementsov@virtuozzo.com>
Tested-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

linux-user/elfload: add s390x core dumping support

Provide the following definitions required by the common code:

* ELF_NREG: with the value of sizeof(s390_regs) / sizeof(long).
* target_elf_gregset_t: define it like all the other arches do.
* elf_core_copy_regs(): similar to kernel's s390_regs_get().
* USE_ELF_CORE_DUMP.
* ELF_EXEC_PAGESIZE.

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-Id: <20210413205608.22587-1-iii@linux.ibm.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user/elfload: fix filling psinfo->pr_psargs

The current code dumps the memory between arg_start and arg_end,
which contains the argv pointers. This results in the

    Core was generated by `<garbage>`

message when opening the core file in GDB. This is because the code is
supposed to dump the actual arg strings. Fix by using arg_strings and
env_strings instead of arg_start and arg_end.

Signed-off-by: Ilya Leoshkevich <iii@linux.ibm.com>
Message-Id: <20210413205814.22821-1-iii@linux.ibm.com>
[lv: add missing braces]
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Tidy TARGET_NR_rt_sigaction

Initialize variables instead of elses.
Use an else instead of a goto.
Add braces.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210422230227.314751-8-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user/alpha: Share code for TARGET_NR_sigaction

There's no longer a difference between the alpha code and
the generic code.

There is a type difference in target_old_sigaction.sa_flags,
which can be resolved with a very much smaller ifdef, which
allows us to finish sharing the target_sigaction definition.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210422230227.314751-7-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user/alpha: Define TARGET_ARCH_HAS_KA_RESTORER

This means that we can share the TARGET_NR_rt_sigaction code,
and the target_rt_sigaction structure is unused.  Untangling
the ifdefs so that target_sigaction can be shared will wait
until the next patch.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210422230227.314751-6-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Honor TARGET_ARCH_HAS_SA_RESTORER in do_syscall

Do not access a field that may not be present.  This will
become an issue when sharing more code in the next patch.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210422230227.314751-5-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Pass ka_restorer to do_sigaction

The value of ka_restorer needs to be saved in sigact_table.
At the moment, the attempt to save it in do_syscall is
improperly clobbering user memory.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210422230227.314751-4-richard.henderson@linaro.org>
[lv: remove tab]
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user/alpha: Rename the sigaction restorer field

Use ka_restorer, in line with TARGET_ARCH_HAS_KA_RESTORER
vs TARGET_ARCH_HAS_SA_RESTORER, since Alpha passes this
field as a syscall argument.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210422230227.314751-3-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user/alpha: Fix rt sigframe return

We incorrectly used the offset of the non-rt sigframe.

Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20210422230227.314751-2-richard.henderson@linaro.org>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: use GDateTime for formatting timestamp for core file

The GDateTime APIs provided by GLib avoid portability pitfalls, such
as some platforms where 'struct timeval.tv_sec' field is still 'long'
instead of 'time_t'. When combined with automatic cleanup, GDateTime
often results in simpler code too.

Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210505103702.521457-7-berrange@redhat.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Fix erroneous conversion in copy_file_range

The implicit cast from abi_long to size_t may introduce an intermediate
unwanted sign-extension of the value for 32bit targets running on 64bit
hosts.

Signed-off-by: Giuseppe Musacchio <thatlemon@gmail.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210503174159.54302-3-thatlemon@gmail.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user: Add copy_file_range to strace.list

Signed-off-by: Giuseppe Musacchio <thatlemon@gmail.com>
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Message-Id: <20210503174159.54302-2-thatlemon@gmail.com>
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

linux-user/s390x: Handle vector regs in signal stack

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Message-Id: <20210428193408.233706-16-richard.henderson@linaro.org>
[lv: fix indentation]
Signed-off-by: Laurent Vivier <laurent@vivier.eu>

Merge remote-tracking branch 'remotes/rth-gitlab/tags/pull-fp-20210516' into staging

Reorg FloatParts to use QEMU_GENERIC.
Begin replacing the Berkeley float128 routines with FloatParts128.
  - includes a new implementation of float128_muladd
  - includes the snan silencing that was missing from
    float{32,64}_to_float128 and float128_to_float{32,64}.
  - does not include float128_min/max* (written but not yet reviewed).

# gpg: Signature made Sun 16 May 2021 13:27:10 BST
# 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-fp-20210516: (46 commits)
  softfloat: Move round_to_int_and_pack to softfloat-parts.c.inc
  softfloat: Move round_to_int to softfloat-parts.c.inc
  softfloat: Convert float-to-float conversions with float128
  softfloat: Split float_to_float
  softfloat: Move div_floats to softfloat-parts.c.inc
  softfloat: Introduce sh[lr]_double primitives
  softfloat: Tidy mul128By64To192
  softfloat: Use add192 in mul128To256
  softfloat: Use mulu64 for mul64To128
  softfloat: Move muladd_floats to softfloat-parts.c.inc
  softfloat: Move mul_floats to softfloat-parts.c.inc
  softfloat: Implement float128_add/sub via parts
  softfloat: Move addsub_floats to softfloat-parts.c.inc
  softfloat: Use uadd64_carry, usub64_borrow in softfloat-macros.h
  softfloat: Move round_canonical to softfloat-parts.c.inc
  softfloat: Move sf_canonicalize to softfloat-parts.c.inc
  softfloat: Move pick_nan_muladd to softfloat-parts.c.inc
  softfloat: Move pick_nan to softfloat-parts.c.inc
  softfloat: Move return_nan to softfloat-parts.c.inc
  softfloat: Convert float128_default_nan to parts
  ...

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

Merge remote-tracking branch 'remotes/vivier2/tags/trivial-branch-for-6.1-pull-request' into staging

Pull request trivial-branch 20210515

# gpg: Signature made Sat 15 May 2021 11:02:59 BST
# 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/vivier2/tags/trivial-branch-for-6.1-pull-request:
  target/avr: Ignore unimplemented WDR opcode
  hw/avr/atmega.c: use the avr51 cpu for atmega1280
  target/sh4: Return error if CPUClass::get_phys_page_debug() fails
  multi-process: Avoid logical AND of mutually exclusive tests
  hw/pci-host: Do not build gpex-acpi.c if GPEX is not selected
  hw/mem/meson: Fix linking sparse-mem device with fuzzer
  cutils: fix memory leak in get_relocated_path()
  hw/rtc/mc146818rtc: Convert to 3-phase reset (Resettable interface)
  hw/timer/etraxfs_timer: Convert to 3-phase reset (Resettable interface)
  hw/gpio/aspeed: spelling fix (addtional)
  qapi: spelling fix (addtional)
  virtiofsd: Fix check of chown()'s return value
  virtio-net: Constify VirtIOFeature feature_sizes[]
  virtio-blk: Constify VirtIOFeature feature_sizes[]
  hw/virtio: Pass virtio_feature_get_config_size() a const argument
  backends/tpm: Replace qemu_mutex_lock calls with QEMU_LOCK_GUARD

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

Merge remote-tracking branch 'remotes/maxreitz/tags/pull-block-2021-05-14' into staging

Block patches:
- drop block/io write notifiers
- qemu-iotests enhancements to make debugging easier
- rbd parsing fix
- HMP qemu-io fix (for iothreads)
- mirror job cancel relaxation (do not cancel in-flight requests when a
  READY mirror job is canceled with force=false)
- document qcow2's data_file and data_file_raw features
- fix iotest 297 for pylint 2.8
- block/copy-on-read refactoring

# gpg: Signature made Fri 14 May 2021 17:43:40 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-2021-05-14:
  write-threshold: deal with includes
  test-write-threshold: drop extra TestStruct structure
  test-write-threshold: drop extra tests
  block/write-threshold: drop extra APIs
  test-write-threshold: rewrite test_threshold_(not_)trigger tests
  block: drop write notifiers
  block/write-threshold: don't use write notifiers
  qemu-iotests: fix pylint 2.8 consider-using-with error
  block/copy-on-read: use bdrv_drop_filter() and drop s->active
  Document qemu-img options data_file and data_file_raw
  qemu-iotests: fix case of SOCK_DIR already in the environment
  qemu-iotests: let "check" spawn an arbitrary test command
  qemu-iotests: move command line and environment handling from TestRunner to TestEnv
  qemu-iotests: allow passing unittest.main arguments to the test scripts
  qemu-iotests: do not buffer the test output
  mirror: stop cancelling in-flight requests on non-force cancel in READY
  monitor: hmp_qemu_io: acquire aio contex, fix crash
  block/rbd: Add an escape-aware strchr helper
  iotests/231: Update expected deprecation message

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

hw/nvme: move nvme emulation out of hw/block

With the introduction of the nvme-subsystem device we are really
cluttering up the hw/block directory.

As suggested by Philippe previously, move the nvme emulation to hw/nvme.

Suggested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: move zoned constraints checks

Validation of the max_active and max_open zoned parameters are
independent of any other state, so move them to the early
nvme_ns_check_constraints parameter checks.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: remove irrelevant zone resource checks

It is not an error to report more active/open zones supported than the
number of zones in the namespace.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: remove num_namespaces member

The NvmeCtrl num_namespaces member is just an indirection for the
NVME_MAX_NAMESPACES constant.

Remove the indirection.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: streamline namespace array indexing

Streamline namespace array indexing such that both the subsystem and
controller namespaces arrays are 1-indexed.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: add metadata offset helper

Add an nvme_moff() helper.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: cache lba and ms sizes

There is no need to look up the lba size and metadata size in the LBA
Format structure everytime we want to use it. And we use it a lot.

Cache the values in the NvmeNamespace and update them if the namespace
is formatted.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: replace nvme_ns_status

The inline nvme_ns_status() helper only has a single call site. Remove
it from the header file and inline it for real.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: remove non-shared defines from header file

Remove non-shared defines from the shared header.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: cleanup includes

Clean up includes.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: consolidate header files

In preparation for moving the nvme device into its own subtree, merge
the header files into one.

Also add missing copyright notice and add list of authors with
substantial contributions.

Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: rename __nvme_select_ns_iocs

Get rid of the (reserved) double underscore use.

Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Cc: Thomas Huth <thuth@redhat.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: rename __nvme_advance_zone_wp

Get rid of the (reserved) double underscore use.

Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Cc: Thomas Huth <thuth@redhat.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: rename __nvme_zrm_open

Get rid of the (reserved) double underscore use. Rename the "generic"
zone open function to nvme_zrm_open_flags() and add a generic `int
flags` argument instead which allows more flags to be easily added in
the future. There is at least one TP under standardization that would
add an additional flag.

Cc: Philippe Mathieu-Daudé <philmd@redhat.com>
Cc: Thomas Huth <thuth@redhat.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>

hw/block/nvme: align with existing style

While QEMU coding style prefers lowercase hexadecimals in constants, the
NVMe subsystem uses the format from the NVMe specifications in comments,
i.e. 'h' suffix instead of '0x' prefix.

Fix this up across the code base.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
[k.jensen: updated message; added conversion in a couple of missing comments]
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>

hw/block/nvme: function formatting fix

nvme_map_addr_pmr function arguments not aligned, fix that.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>

hw/block/nvme: fix io-command set profile feature

Currently IO Command Set Profile feature is supported, but the feature
support flag not set. Further, this feature is changable. Fix that.

Additionally, remove filling default value of the CQE result with zero,
since it will fall back to the default case anyway.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
[k.jensen: fix up commit message]
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>

hw/block/nvme: consider metadata read aio return value in compare

Currently in compare command metadata aio read blk_aio_preadv return
value ignored. Consider it and complete the block accounting.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
Fixes: 0a384f923f51 ("hw/block/nvme: add compare command")
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>

hw/block/nvme: rename reserved fields declarations

Align the 'rsvd1' reserved field declaration in NvmeBar with existing
style.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
[k.jensen: minor commit message fixup]
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>

hw/block/nvme: remove redundant invalid_lba_range trace

Currently pci_nvme_err_invalid_lba_range trace is called individually at
each nvme_check_bounds() call site.

Move the trace event to nvme_check_bounds() and remove the redundant
events.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
[k.jensen: commit message fixup]
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>

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

pc,pci,virtio: bugfixes, improvements

Fixes all over the place. Faster boot for virtio. ioeventfd support for
mmio.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
# gpg: Signature made Fri 14 May 2021 15:27:13 BST
# gpg:                using RSA key 5D09FD0871C8F85B94CA8A0D281F0DB8D28D5469
# gpg:                issuer "mst@redhat.com"
# gpg: Good signature from "Michael S. Tsirkin <mst@kernel.org>" [full]
# gpg:                 aka "Michael S. Tsirkin <mst@redhat.com>" [full]
# Primary key fingerprint: 0270 606B 6F3C DF3D 0B17  0970 C350 3912 AFBE 8E67
#      Subkey fingerprint: 5D09 FD08 71C8 F85B 94CA  8A0D 281F 0DB8 D28D 5469

* remotes/mst/tags/for_upstream:
  Fix build with 64 bits time_t
  vhost-vdpa: Make vhost_vdpa_get_device_id() static
  hw/virtio: enable ioeventfd configuring for mmio
  hw/smbios: support for type 41 (onboard devices extended information)
  checkpatch: Fix use of uninitialized value
  virtio-scsi: Configure all host notifiers in a single MR transaction
  virtio-scsi: Set host notifiers and callbacks separately
  virtio-blk: Configure all host notifiers in a single MR transaction
  virtio-blk: Fix rollback path in virtio_blk_data_plane_start()
  pc-dimm: remove unnecessary get_vmstate_memory_region() method
  amd_iommu: fix wrong MMIO operations
  virtio-net: Constify VirtIOFeature feature_sizes[]
  virtio-blk: Constify VirtIOFeature feature_sizes[]
  hw/virtio: Pass virtio_feature_get_config_size() a const argument
  x86: acpi: use offset instead of pointer when using build_header()
  amd_iommu: Fix pte_override_page_mask()

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
# Conflicts:
# hw/arm/virt.c

accel/tcg: Align data dumped at end of TB

To better visualize the data dumped at the end of a TB, left-align it
(padding it with 0).  Print ".long" instead of ".quad" on 32-bit hosts.

Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Message-Id: <20210515104202.241504-1-f4bug@amsat.org>
[rth: Split the qemu_log and print .long for 32-bit hosts.]
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move round_to_int_and_pack to softfloat-parts.c.inc

Rename to parts$N_float_to_sint.  Reimplement
float128_to_int{32,64}{_round_to_zero} with FloatParts128.

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

softfloat: Move round_to_int to softfloat-parts.c.inc

At the same time, convert to pointers, split out
parts$N_round_to_int_normal, define a macro for
parts_round_to_int using QEMU_GENERIC.

This necessarily meant some rearrangement to the
rount_to_{,u}int_and_pack routines, so go ahead and
convert to parts_round_to_int_normal, which in turn
allows cleaning up of the raised exception handling.

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

softfloat: Convert float-to-float conversions with float128

Introduce parts_float_to_float_widen and parts_float_to_float_narrow.
Use them for float128_to_float{32,64} and float{32,64}_to_float128.

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

softfloat: Split float_to_float

Split out parts_float_to_ahp and parts_float_to_float.
Convert to pointers.

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

softfloat: Move div_floats to softfloat-parts.c.inc

Rename to parts$N_div.
Implement float128_div with FloatParts128.

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

softfloat: Introduce sh[lr]_double primitives

Have x86_64 assembly for them, with a fallback.
This avoids shuffling values through %cl in the x86 case.

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

softfloat: Tidy mul128By64To192

Clean up the formatting and variables; no functional change.

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

softfloat: Use add192 in mul128To256

We can perform the operation in 6 total adds instead of 8.

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

softfloat: Use mulu64 for mul64To128

Via host-utils.h, we use a host widening multiply for
64-bit hosts, and a common subroutine for 32-bit hosts.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move muladd_floats to softfloat-parts.c.inc

Rename to parts$N_muladd.
Implement float128_muladd with FloatParts128.

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

softfloat: Move mul_floats to softfloat-parts.c.inc

Rename to parts$N_mul.
Reimplement float128_mul with FloatParts128.

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

softfloat: Implement float128_add/sub via parts

Replace the existing Berkeley implementation with the
FloatParts implementation.

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

softfloat: Move addsub_floats to softfloat-parts.c.inc

In preparation for implementing multiple sizes.  Rename to parts_addsub,
split out parts_add/sub_normal for future reuse with muladd.

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

softfloat: Use uadd64_carry, usub64_borrow in softfloat-macros.h

Use compiler support for carry arithmetic.

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

softfloat: Move round_canonical to softfloat-parts.c.inc

At the same time, convert to pointers, renaming to parts$N_uncanon,
and define a macro for parts_uncanon using QEMU_GENERIC.

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

softfloat: Move sf_canonicalize to softfloat-parts.c.inc

At the same time, convert to pointers, rename to parts$N_canonicalize
and define a macro for parts_canonicalize using QEMU_GENERIC.

Rearrange the cases to recognize float_class_normal as
early as possible.

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

softfloat: Move pick_nan_muladd to softfloat-parts.c.inc

At the same time, convert to pointers, rename to pick_nan_muladd$N
and define a macro for pick_nan_muladd using QEMU_GENERIC.

Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move pick_nan to softfloat-parts.c.inc

At the same time, convert to pointers, rename to parts$N_pick_nan
and define a macro for parts_pick_nan using QEMU_GENERIC.

Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move return_nan to softfloat-parts.c.inc

At the same time, convert to pointers, rename to return_nan$N
and define a macro for return_nan using QEMU_GENERIC.

Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Convert float128_default_nan to parts

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

softfloat: Convert float128_silence_nan to parts

This is the minimal change that also introduces float128_params,
float128_unpack_raw, and float128_pack_raw without running into
unused symbol Werrors.

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

softfloat: Rearrange FloatParts64

Shuffle the fraction to the end, otherwise sort by size.
Add frac_hi and frac_lo members to alias frac.

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

softfloat: Use pointers with parts_silence_nan

At the same time, rename to parts64_silence_nan, split out
parts_silence_nan_frac, and define a macro for parts_silence_nan.

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

softfloat: Use pointers with ftype_round_pack_canonical

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

softfloat: Use pointers with ftype_unpack_canonical

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

softfloat: Use pointers with ftype_pack_raw

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

softfloat: Use pointers with pack_raw

At the same time, rename to pack_raw64.

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

softfloat: Use pointers with ftype_unpack_raw

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

softfloat: Use pointers with unpack_raw

At the same time, rename to unpack_raw64.

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

softfloat: Use pointers with parts_default_nan

At the same time, rename to parts64_default_nan and add a
macro for parts_default_nan.  This will be flushed out once
128-bit support is added.

Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move type-specific pack/unpack routines

In preparation from moving sf_canonicalize.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Rename FloatParts to FloatParts64

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Do not produce a default_nan from parts_silence_nan

Require default_nan_mode to be set instead.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

target/mips: Set set_default_nan_mode with set_snan_bit_is_one

This behavior is currently hard-coded in parts_silence_nan,
but setting this bit properly will allow this to be cleaned up.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: fix return_nan vs default_nan_mode

Do not call parts_silence_nan when default_nan_mode is in
effect.  This will avoid an assert in a later patch.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Use return_nan in float_to_float

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Add float_cmask and constants

Testing more than one class at a time is better done with masks.
This reduces the static branch count.

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Tidy a * b + inf return

No reason to set values in 'a', when we already
have float_class_inf in 'c', and can flip that sign.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Use float_raise in more places

We have been somewhat inconsistent about when to use
float_raise and when to or in the bit by hand.

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Inline float_raise

Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

softfloat: Move the binary point to the msb

Rather than point the binary point at msb-1, put it at the msb.
Use uadd64_overflow to detect when addition overflows instead
of DECOMPOSED_OVERFLOW_BIT.

This reduces the number of special cases within the code, such
as shifting an int64_t either left or right during conversion.

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