Paolo Bonzini [Fri, 28 Mar 2014 10:43:45 +0000 (11:43 +0100)]
target-i386: fix kernel accesses with SMAP and CPL = 3
With SMAP, implicit kernel accesses from user mode always behave as
if AC=0. To do this, kernel mode is not anymore a separate MMU mode.
Instead, KERNEL_IDX is renamed to KSMAP_IDX and the kernel mode accessors
wrap KSMAP_IDX and KNOSMAP_IDX.
Paolo Bonzini [Fri, 28 Mar 2014 18:42:10 +0000 (19:42 +0100)]
softmmu: introduce cpu_ldst.h
This will collect all load and store helpers soon. For now
it is just a replacement for softmmu_exec.h, which this patch
stops including directly, but we also include it where this will
be necessary in order to simplify the next patch.
Paolo Bonzini [Fri, 28 Mar 2014 16:55:24 +0000 (17:55 +0100)]
softmmu: commonize helper definitions
They do not need to be in op_helper.c. Because cputlb.c now includes
softmmu_template.h twice for each size, io_readX must be elided the
second time through.
Paolo Bonzini [Fri, 28 Mar 2014 17:14:58 +0000 (18:14 +0100)]
softmmu: make do_unaligned_access a method of CPU
We will reference it from more files in the next patch. To avoid
ruining the small steps we're making towards multi-target, make
it a method of CPU rather than just a global.
Paolo Bonzini [Fri, 28 Mar 2014 10:15:30 +0000 (11:15 +0100)]
softmmu: start introducing SOFTMMU_CODE_ACCESS in softmmu_header.h
This preprocessor symbol is already used in softmmu_template.h. We
will use it to distinguish the two "fake" ACCESS_TYPEs
NB_MMU_MODES and NB_MMU_MODES + 1.
Paolo Bonzini [Fri, 28 Mar 2014 17:43:14 +0000 (18:43 +0100)]
hw: use ld_p/st_p instead of ld_raw/st_raw
The ld_raw and st_raw definitions are only needed in code that
must compile for both user-mode and softmmu emulation. Device
models can use the equivalent ld_p/st_p which are simple
pointer accessors.
Stefan Weil [Mon, 28 Apr 2014 17:20:00 +0000 (19:20 +0200)]
cputlb: Fix regression with TCG interpreter (bug 1310324)
Commit 0f842f8a246f2b5b51a11c13f933bf7a90ae8e96 replaced GETPC_EXT() which
was derived from GETPC() by GETRA_EXT() without fixing cputlb.c. A later
patch replaced GETRA_EXT() by GETRA() in exec/softmmu_template.h which
is included in cputlb.c.
The TCG interpreter failed because the values returned by GETRA() were no
longer explicitly set to 0. The redefinition of GETRA() introduced here
fixes this.
In addition, GETPC_ADJ which is also used in exec/softmmu_template.h is
set to 0. Both changes reduce the compiled code size for cputlb.c by more
than 100 bytes, so the normal TCG without interpreter also profits from
the reduced code size and slightly faster code.
Peter Maydell [Tue, 3 Jun 2014 13:37:43 +0000 (14:37 +0100)]
Merge remote-tracking branch 'remotes/awilliam/tags/vfio-pci-for-qemu-20140602.0' into staging
VFIO patches: realtek NIC quirk + SPAPR IOMMU AddressSpace support
# gpg: Signature made Mon 02 Jun 2014 22:44:42 BST using RSA key ID 3BB08B22
# gpg: Can't check signature: public key not found
* remotes/awilliam/tags/vfio-pci-for-qemu-20140602.0:
vfio: Add guest side IOMMU support
vfio: Create VFIOAddressSpace objects as needed
vfio: Introduce VFIO address spaces
vfio: Rework to have error paths
vfio: Fix 128 bit handling
int128: Add int128_exts64()
memory: Sanity check that no listeners remain on a destroyed AddressSpace
vfio-pci: Quirk RTL8168 NIC
Peter Maydell [Mon, 2 Jun 2014 16:07:21 +0000 (17:07 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-usb-8' into staging
qtest: improve ehci/uhci test
usb: misc fixes, mostly for usb3/xhci
# gpg: Signature made Mon 02 Jun 2014 15:40:34 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg: aka "Gerd Hoffmann <[email protected]>"
# gpg: aka "Gerd Hoffmann (private) <[email protected]>"
* remotes/kraxel/tags/pull-usb-8:
xhci: order superspeed ports first
xhci: make port reset trace point more verbose
usb: add usb_pick_speed
usb-host: add HAVE_STREAMS define
usb-host: allow attaching usb3 devices to ehci
usb: improve ehci/uhci test
usb: move ehci register defines to header file
usb: add uhci port status reserved bit
usb: move uhci register defines to header file
qtest: fix qpci_config_writel
Peter Maydell [Mon, 2 Jun 2014 15:10:12 +0000 (16:10 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-sdl-3' into staging
sdl2: add support for text consoles
# gpg: Signature made Mon 02 Jun 2014 15:35:20 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg: aka "Gerd Hoffmann <[email protected]>"
# gpg: aka "Gerd Hoffmann (private) <[email protected]>"
* remotes/kraxel/tags/pull-sdl-3:
sdl2: textinput + terminal
sdl2: make Ctrl-Alt-<nr> hotkeys show and hide windows
console: add kbd_put_string_console
console: add kbd_put_qcode_console
Peter Maydell [Mon, 2 Jun 2014 14:47:40 +0000 (15:47 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-vnc-3' into staging
misc minor vnc patches
# gpg: Signature made Mon 02 Jun 2014 15:31:53 BST using RSA key ID D3E87138
# gpg: Good signature from "Gerd Hoffmann (work) <[email protected]>"
# gpg: aka "Gerd Hoffmann <[email protected]>"
# gpg: aka "Gerd Hoffmann (private) <[email protected]>"
* remotes/kraxel/tags/pull-vnc-3:
vnc-enc-tight: Fix divide-by-zero in tight_detect_smooth_image{16,24,32}
vnc: add trace events for key events
vnc: refuse to set a password with VNC_AUTH_NONE
Gerd Hoffmann [Wed, 11 Dec 2013 12:15:37 +0000 (13:15 +0100)]
vnc: refuse to set a password with VNC_AUTH_NONE
Current code silently changes the authentication settings
in case you try to set a password without password authentication
turned on. This is bad. Return an error instead.
If we want allow changing auth settings at runtime this should
be done explicitly using a separate monitor command, not as
side effect of set_passwd.
Gerd Hoffmann [Fri, 23 May 2014 14:20:54 +0000 (16:20 +0200)]
usb: add usb_pick_speed
We can pick the usb port speed in generic code, by looking at the port
and device speed masks and looking for the fastest match. So add a
function to do exactly that, and drop the speed setting code from
usb_desc_attach as it isn't needed any more.
This way we can set the device speed before calling port->ops->attach,
which fixes some xhci hotplug issues.
Gerd Hoffmann [Fri, 23 May 2014 08:27:00 +0000 (10:27 +0200)]
usb-host: allow attaching usb3 devices to ehci
Extend compatibility test function to also figure whenever usb3
devices can be supported on ehci. Tweak ep0 maxpacketsize field
due to usb2 <-> usb3 difference.
Gerd Hoffmann [Wed, 7 May 2014 14:39:11 +0000 (16:39 +0200)]
usb: improve ehci/uhci test
* Attach usb devices to the bus.
* Check initial port status register state.
* Flip ehci initialization bit.
* Check port status register state again to
see whenever device handover to ehci worked.
Peter Maydell [Mon, 2 Jun 2014 14:27:18 +0000 (15:27 +0100)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block patches
# gpg: Signature made Mon 02 Jun 2014 14:56:00 BST using RSA key ID C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>"
* remotes/kevin/tags/for-upstream:
qemu-img: Report error even with --oformat=json
vmdk: Fix local_err in vmdk_create
block/raw-posix.c: Avoid nonstandard LONG_LONG_MAX
qemu-img: Plug memory leak in convert command
block/sheepdog: Plug memory leak in sd_snapshot_create()
block/vvfat: Plug memory leak in read_directory()
block/vvfat: Plug memory leak in check_directory_consistency()
block/qapi: Plug memory leak in dump_qobject() case QTYPE_QERROR
blockdev: Plug memory leak in drive_init()
blockdev: Plug memory leak in blockdev_init()
qemu-io: Don't print NULL when open without non-option arg fails
qemu-io: Plug memory leak in open command
qemu-io: Support multiple -o in open command
block: Plug memory leak on brv_open_image() error path
qcow2: Plug memory leak on qcow2_invalidate_cache() error paths
block/vvfat: Plug memory leak in enable_write_target()
qemu-img: Plug memory leak on block option help error path
That commit converted various fprintf(stderr, ...) calls to
use error_report(); however none of these bsd-user files include
a header which gives a prototype for error_report, so this
causes compiler warnings. Since these are just straightforward
reporting of command line errors, we should handle these in the
obvious way by printing to stderr, as we do for linux-user.
There's no need to drag in the error-handling framework for this,
especially since user-mode doesn't have the "maybe we need to
send this to the monitor" issues system emulation does.
Max Reitz [Sat, 31 May 2014 19:33:30 +0000 (21:33 +0200)]
qemu-img: Report error even with --oformat=json
img_check() should report that the format of the given image does not
support checks even if JSON output is desired. JSON data is output to
stdout, as opposed to error messages, which are (in the case of
qemu-img) printed to stderr. Therefore, it is easy to distinguish
between the two.
Also, img_info() does already use error_report() for human-readable
messages even though JSON output is desired (through
collect_image_info_list()).
David Gibson [Fri, 30 May 2014 19:10:07 +0000 (13:10 -0600)]
vfio: Add guest side IOMMU support
This patch uses the new IOMMU notifiers to allow VFIO pass through devices
to work with guest side IOMMUs, as long as the host-side VFIO iommu has
sufficient capability and granularity to match the guest side. This works
by tracking all map and unmap operations on the guest IOMMU using the
notifiers, and mirroring them into VFIO.
There are a number of FIXMEs, and the scheme involves rather more notifier
structures than I'd like, but it should make for a reasonable proof of
concept.
David Gibson [Fri, 30 May 2014 19:09:14 +0000 (13:09 -0600)]
vfio: Create VFIOAddressSpace objects as needed
So far, VFIO has a notion of different logical DMA address spaces, but
only ever uses one (system memory). This patch extends this, creating
new VFIOAddressSpace objects as necessary, according to the AddressSpace
reported by the PCI subsystem for this device's DMAs.
This isn't enough yet to support guest side IOMMUs with VFIO, but it does
mean we could now support VFIO devices on, for example, a guest side PCI
host bridge which maps system memory at somewhere other than 0 in PCI
space.
David Gibson [Fri, 30 May 2014 19:05:19 +0000 (13:05 -0600)]
vfio: Introduce VFIO address spaces
The only model so far supported for VFIO passthrough devices is the model
usually used on x86, where all of the guest's RAM is mapped into the
(host) IOMMU and there is no IOMMU visible in the guest.
This patch begins to relax this model, introducing the notion of a
VFIOAddressSpace. This represents a logical DMA address space which will
be visible to one or more VFIO devices by appropriate mapping in the (host)
IOMMU. Thus the currently global list of containers becomes local to
a VFIOAddressSpace, and we verify that we don't attempt to add a VFIO
group to multiple address spaces.
For now, only one VFIOAddressSpace is created and used, corresponding to
main system memory, that will change in future patches.
David Gibson [Fri, 30 May 2014 18:59:00 +0000 (12:59 -0600)]
memory: Sanity check that no listeners remain on a destroyed AddressSpace
At the moment, most AddressSpace objects last as long as the guest system
in practice, but that could well change in future. In addition, for VFIO
we will be introducing some private per-AdressSpace information, which must
be disposed of before the AddressSpace itself is destroyed.
To reduce the chances of subtle bugs in this area, this patch adds
asssertions to ensure that when an AddressSpace is destroyed, there are no
remaining MemoryListeners using that AS as a filter.
Alex Williamson [Fri, 30 May 2014 18:43:50 +0000 (12:43 -0600)]
vfio-pci: Quirk RTL8168 NIC
This device is ridiculous. It has two MMIO BARs, BAR4 and BAR2. BAR4
hosts the MSI-X table, so oviously it would be too easy to access it
directly, instead it creates a window register in BAR2 that, among
other things, provides access to the MSI-X table. This means MSI-X
doesn't work in the guest because the driver actually manages to
program the physical table. When interrupt remapping is present, the
device MSI will be blocked. The Linux driver doesn't make use of this
window, so apparently it's not required to make use of MSI-X. This
quirk makes the device work with the Windows driver that does use this
window for MSI-X, but I certainly cannot recommend this device for
assignment (the Windows 7 driver also constantly pokes PCI config
space).
In the MacOSX specific code in raw-posix.c we use the define
LONG_LONG_MAX. This is actually a non-standard pre-C99 define;
switch to using the standard LLONG_MAX instead.
This apparently fixes a compilation failure with certain
compiler/OS versions (though it is unclear which).
bs_opts is leaked on all paths from its qdev_new() that don't got
through blockdev_init(). Add the missing QDECREF(), and zap bs_opts
after blockdev_init(), so the new QDECREF() does nothing when we go
through blockdev_init().
Leak introduced in commit f298d07. Spotted by Coverity.
Instead of ignoring all option values but the last one, multiple -o
options now have the same meaning as having a single option with all
settings in the order of their respective -o options.
Same as commit 2dc8328 for qemu-img convert, except here we do it with
QemuOpts rather than QEMUOptionParameter.
block/vvfat: Plug memory leak in enable_write_target()
I figure the leak originated in bdrv_create2(), and was duplicated
into callers when commit 91a073a dropped that function. Looks like
the other places have since been fixed.
Peter Maydell [Wed, 28 May 2014 17:38:38 +0000 (18:38 +0100)]
Merge remote-tracking branch 'remotes/afaerber/tags/qom-devices-for-peter' into staging
QOM/QTest infrastructure fixes and device conversions
* qom-test extension
* QEMUMachineInitArgs conversion to MachineState
* -machine options turned into /machine properties
* Named GPIO IRQs for devices
# gpg: Signature made Wed 28 May 2014 18:24:04 BST using RSA key ID 3E7E013F
# gpg: Can't check signature: public key not found
* remotes/afaerber/tags/qom-devices-for-peter:
ssi: Name the CS GPIO
qdev: Implement named GPIOs
machine: Make -machine opts properties of MachineState
tests: Check empty QMP output visitor
qapi: Avoid output visitor crashing if it encounters a NULL value
vl.c: Do not set 'type' property in obj_set_property()
machine: Conversion of QEMUMachineInitArgs to MachineState
qom-test: Test qom-list on link<> properties
Peter Maydell [Wed, 28 May 2014 16:44:45 +0000 (17:44 +0100)]
Merge remote-tracking branch 'remotes/rth/tcg-next' into staging
* remotes/rth/tcg-next:
tcg/optimize: Remember garbage high bits for 32-bit ops
tcg/optimize: Move updating of gen_opc_buf into tcg_opt_gen_mov*
tcg-sparc: Make debug_frame const
tcg-s390: Make debug_frame const
tcg-arm: Make debug_frame const
tcg-aarch64: Make debug_frame const
tcg-i386: Make debug_frame const
tcg: Allow the debug_frame data structure to be constant
tcg: Move size effects out of dh_arg
tcg: Remove sizemask and flags arguments to tcg_gen_callN
tcg: Save flags and computed sizemask in TCGHelperInfo
tcg: Register the helper info struct rather than the name
tcg: Move side effects out of dh_sizemask
tcg: Inline tcg_gen_helperN
tcg: Use helper-gen.h in tcg-op.h
tcg: Push tcg-runtime routines into exec/helper-*
tcg: Invert the inclusion of helper.h
tcg: Optimize brcond2 and setcond2 ne/eq
tcg: Allow the debug_frame data structure to be constant
Adjust the FDE to point to the code_buffer after we've copied it
to the image, rather than requiring that the backend set it prior.
This allows the backend to use read-only storage for its data.
Rather than include helper.h with N values of GEN_HELPER, include a
secondary file that sets up the macros to include helper.h. This
minimizes the files that must be rebuilt when changing the macros
for file N.
To get it out of the default GPIO list. This allows child devices to
use the un-named GPIO namespace without having to be SSI aware. That
is, there is no more need for machines to know about the obscure
policy where GPIO 0 is the SSI chip-select and GPIO 1..N are the
concrete class GPIOs (defined locally as 0..N-1).
This is most notable in stellaris, which uses a device which has both
SSI and concrete level GPIOs.
Implement named GPIOs on the Device layer. Listifies the existing GPIOs
stuff using string keys. Legacy un-named GPIOs are preserved by using
a NULL name string - they are just a single matchable element in the
name list.
Marcel Apfelbaum [Mon, 26 May 2014 12:40:58 +0000 (15:40 +0300)]
machine: Make -machine opts properties of MachineState
Make machine's QemuOpts QOM properties of /machine. The properties
are automatically filled in. This opens the possibility to create
opts per machine rather than global.
projects / qemu.git / log