Max Filippov [Mon, 21 Nov 2011 00:54:58 +0000 (04:54 +0400)]
configure: check for EFD_NONBLOCK | EFD_CLOEXEC flags
Add check for the EFD_NONBLOCK and EFD_CLOEXEC flags to the
CONFIG_EVENTFD test.
This fixes the following build failure on Fedora 9:
CC event_notifier.o
event_notifier.c: In function `event_notifier_init':
event_notifier.c:21: error: `EFD_NONBLOCK' undeclared (first use in this function)
event_notifier.c:21: error: (Each undeclared identifier is reported only once
event_notifier.c:21: error: for each function it appears in.)
event_notifier.c:21: error: `EFD_CLOEXEC' undeclared (first use in this function)
make: *** [event_notifier.o] Error 1
Reviewed-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com> Signed-off-by: Max Filippov <jcmvbkbc@gmail.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Avi Kivity [Tue, 15 Nov 2011 18:12:17 +0000 (20:12 +0200)]
configure: build position independent executables on x86-Linux hosts
Change the default on x86 Linux hosts to building PIE (position
independent executables); instead of restricting the option to
user-only targets, apply it to all targets.
In addition, set the relocation sections to read-only (relro) when
available; this reduces the attack surface by disallowing changes to
relocation tables at runtime.
While PIE reduces performance and relro increases load time, it
greatly improves security, with the potential to reduce a code
execution vulnerability to a self denial of service.
Non-x86 are not changed, as they require TCG changes; neither are
non-Linux, due to lack of test coverage.
Signed-off-by: Avi Kivity <avi@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Hongyong Zang [Mon, 21 Nov 2011 10:56:18 +0000 (18:56 +0800)]
ivshmem: fix PCI BAR2 registration during initialization
Ivshmem cannot work, and the command lspci cannot show ivshmem BAR2 in the guest.
As for pci_register_bar(), parameter MemoryRegion should be s->bar instead of s->ivshmem.
Reviewed-by: Avi Kivity <avi@redhat.com> Signed-off-by: Hongyong Zang <zanghongyong@huawei.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Only accesses to the MSI-X table must trigger a call to
msix_handle_mask_update, otherwise the vector
value might be out of range.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com> Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Anthony Liguori [Mon, 14 Nov 2011 21:09:46 +0000 (15:09 -0600)]
qcow2: implement bdrv_invalidate_cache (v2)
We don't reopen the actual file, but instead invoke the close and open routines.
We specifically ignore the backing file since it's contents are read-only and
therefore immutable.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Anthony Liguori [Mon, 14 Nov 2011 21:09:45 +0000 (15:09 -0600)]
block: allow migration to work with image files (v3)
Image files have two types of data: immutable data that describes things like
image size, backing files, etc. and mutable data that includes offset and
reference count tables.
Today, image formats aggressively cache mutable data to improve performance. In
some cases, this happens before a guest even starts. When dealing with live
migration, since a file is open on two machines, the caching of meta data can
lead to data corruption.
This patch addresses this by introducing a mechanism to invalidate any cached
mutable data a block driver may have which is then used by the live migration
code.
NB, this still requires coherent shared storage. Addressing migration without
coherent shared storage (i.e. NFS) requires additional work.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Anthony Liguori [Mon, 14 Nov 2011 21:09:43 +0000 (15:09 -0600)]
migrate: add migration blockers
This lets different subsystems register an Error that is thrown whenever
migration is attempted. This works nicely because it gracefully supports
things like hotplug.
Right now, if multiple errors are registered, only one of them is reported.
I expect that for 1.1, we'll extend query-migrate to return all of the reasons
why migration is disabled at any given point in time.
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Gerd Hoffmann [Fri, 11 Nov 2011 16:14:15 +0000 (17:14 +0100)]
usb-linux: fix /proc/bus/usb/devices scan
Commit 0c402e5abb8c2755390eee864b43a98280fc2453 is incomplete
and misses one of the two function pointer calls in
usb_host_scan_dev(). Add the additional port handling logic
to the other call too.
Spotted by Coverity.
Cc: Markus Armbruster <armbru@redhat.com> Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Gerd Hoffmann [Mon, 21 Nov 2011 13:01:26 +0000 (14:01 +0100)]
usb-storage: don't try to send the status early.
Until recently all scsi commands sent to scsi-disk did either transfer
data or finished instantly. The correct implementation of
SYNCRONIZE_CACHE changed the picture though, and usb-storage needs
a fix to handle that case correctly.
Blue Swirl [Sun, 13 Nov 2011 11:11:52 +0000 (11:11 +0000)]
x86: fix pcmpestrm and pcmpistrm
Fix obvious typos (decrement and off-by-one error) in pcmpestrm and pcmpistrm
which resulted in infinite loop. Reported by Frank Mehnert,
spotted also by Coverity (bug 84752853).
Reported-by: Frank Mehnert <frank.mehnert@oracle.com> Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
loader: Fix read_targphys() to behave when read() fails
Happily passes (size_t)-1 to rom_add_blob_fixed(), which promptly dies
attempting to malloc that much. Spotted by Coverity.
Bonus fix for ROMs larger than INT_MAX bytes: return ssize_t instead
of int. Bug can't bite, because the only user load_aout() limits ROM
size to an int value.
Signed-off-by: Markus Armbruster <armbru@redhat.com> Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
In both cases, val is computed, but then not used in the
subsequent line, which then re-computes the quantity in
a different type (int32_t vs unsigned long).
Keep the computation type that's been working so far.
Signed-off-by: Richard Henderson <rth@twiddle.net> Signed-off-by: Blue Swirl <blauwirbel@gmail.com>
Blue Swirl [Sat, 19 Nov 2011 11:17:58 +0000 (11:17 +0000)]
Merge branch 's390-1.0' of git://repo.or.cz/qemu/agraf
* 's390-1.0' of git://repo.or.cz/qemu/agraf:
s390x: initialize virtio dev region
tcg: Use TCGReg for standard tcg-target entry points.
tcg: Standardize on TCGReg as the enum for hard registers
s390x: Add shutdown for TCG s390-virtio machine
s390: Fix cpu shutdown for KVM
s390: fix short kernel command lines
s390: fix reset hypercall to reset the status
s390x: implement SIGP restart and shutdown
s390x: implement rrbe instruction properly
s390x: update R and C bits in storage key
s390x: make ipte 31-bit aware
s390x: add ldeb instruction
Blue Swirl [Sat, 19 Nov 2011 11:17:11 +0000 (11:17 +0000)]
Merge branch 'ppc-1.0' of git://repo.or.cz/qemu/agraf
* 'ppc-1.0' of git://repo.or.cz/qemu/agraf:
pseries: Fix qdev.id handling in the VIO bus code
pseries: Allow kernel's early debug output to work
pseries: Default reg for vty should be SPAPR_VTY_BASE_ADDRESS
pseries: Check we have a chardev in spapr_vty_init()
pseries: Fix buggy spapr_vio_find_by_reg()
pseries: Correct RAM size check for SLOF
PPC: Fix for the gdb single step problem on an rfi instruction
tcg-ppc64: Fix compile errors for userspace only builds with gcc 4.6
pseries: Fix initialization of sPAPREnvironment structure
Michael Ellerman [Tue, 15 Nov 2011 18:53:13 +0000 (18:53 +0000)]
pseries: Fix qdev.id handling in the VIO bus code
When the user creates a device on the command line with -device, they
can specify the id, using id=foo. Currently the VIO bus code overwrites
this id with it's own value. We should only set qdev.id if it is not
already set by the user.
The device tree code uses qdev.id for the device tree node name, however
we can't rely on the user specifiying the id using proper device tree
syntax, ie. device@reg. So separate the device tree node name from the
qdev.id, but use the same syntax, so they will match by default.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
David Gibson [Sun, 13 Nov 2011 17:19:01 +0000 (17:19 +0000)]
pseries: Allow kernel's early debug output to work
The PAPR specification defines a virtual TTY/console interface for guest
OSes to use via the H_PUT_TERM_CHAR and H_GET_TERM_CHAR hypercalls. There
can be multiple virtual ttys, so these take a "termno" parameter. This
encodes which vty to use as the 'reg' property on the device tree node
associated with that vty.
However, with the early debug options enabled, the Linux kernel will
attempt debugging output through the vty very early, before it has read
the device tree. In this case it always uses a termno of 0. This works
on the existing PowerVM hypervisor, so we assume there must be a hack /
feature in there which interprets termno==0 to mean the default primary
console.
To help with debugging kernels, including existing distribution kernels,
this patch implements a similar feature / hack in qemu. If termno==0
is supplied to H_{GET,PUT}_TERM_CHAR, they use the first available vty
device instead.
We need to be careful in the case that the user has manually created
an spapr-vty at address 0. So first we search for the specified reg and
only if that doesn't match do we fall back.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
Michael Ellerman [Sun, 13 Nov 2011 17:19:00 +0000 (17:19 +0000)]
pseries: Default reg for vty should be SPAPR_VTY_BASE_ADDRESS
In commit b4a78527359a4540d84d4cdf629d01cbb262f698 ("Place pseries vty
devices at addresses more similar to existing machines"), we changed the
default reg for the vty to 0x30000000, however we didn't update the default
value for a user specified vty device. Fix that.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
Michael Ellerman [Sun, 13 Nov 2011 17:18:59 +0000 (17:18 +0000)]
pseries: Check we have a chardev in spapr_vty_init()
If qemu is run like:
qemu-system-ppc64 -nodefaults -device spapr-vty
We end up in spapr_vty_init() with dev->chardev == NULL. Currently
that leads to a segfault because we unconditionally call
qemu_chr_add_handlers().
Although we could make that call conditional, I think a spapr-vty
without a chardev is basically useless so fail the init. This is
similar to what the serial code does for example.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
David Gibson [Sun, 13 Nov 2011 17:18:58 +0000 (17:18 +0000)]
pseries: Fix buggy spapr_vio_find_by_reg()
The spapr_vio_find_by_reg() function in hw/spapr_vio.c is supposed to find
the device structure for a PAPR virtual IO device with the given reg value,
and return NULL if none exists.
It does the first ok, but if no device with that reg exists, it just
returns the last device traversed in the list. This patch fixes it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
David Gibson [Sun, 13 Nov 2011 17:18:57 +0000 (17:18 +0000)]
pseries: Correct RAM size check for SLOF
The SLOF firmware used on the pseries machine needs a reasonable amount of
(guest) RAM in order to run, so we have a check in the machine init
function to check that this is available. However, SLOF runs in real mode
(MMU off) which means it can only actually access the RMA (Real Mode Area),
not all of RAM. In many cases the RMA is the same as all RAM, but when
running with Book3S HV KVM on PowerPC 970, the RMA must be especially
allocated to be (host) physically contiguous. In this case, the RMA size
is determined by what the host admin allocated at boot time, and will
usually be less than the whole guest RAM size.
This patch corrects the test to see if SLOF has enough memory for this
case.
In addition, more recent versions of SLOF that were committed earlier don't
need quite as much memory as earlier versions. Therefore, this patch also
reduces the amount of RAM we require to run SLOF.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
Paolo Bonzini [Mon, 14 Nov 2011 13:31:52 +0000 (14:31 +0100)]
scsi-block: always use SG_IO for MMC devices
CD burning messes up the state of the host page cache and host block
device. Just pass all operations down to the device, even though that
might have slightly worse performance. Everything else just is not
reliable in combination with burning.
Reported-by: Thomas Schmitt <scdbackup@gmx.net> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Paolo Bonzini [Mon, 14 Nov 2011 13:31:51 +0000 (14:31 +0100)]
scsi: pass down REQUEST SENSE to the device when there is no stored sense
This will let scsi-block/scsi-generic report progress on long
operations.
Reported-by: Thomas Schmitt <scdbackup@gmxbackup.net> Tested-by: Thomas Schmitt <scdbackup@gmxbackup.net> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Paolo Bonzini [Mon, 14 Nov 2011 13:31:50 +0000 (14:31 +0100)]
scsi: remove block descriptors from CDs
Reported-by: Thomas Schmitt <scdbackup@gmx.net> Tested-by: Thomas Schmitt <scdbackup@gmx.net> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Paolo Bonzini [Mon, 14 Nov 2011 13:31:49 +0000 (14:31 +0100)]
scsi: fix parsing of allocation length field
- several MMC commands were parsed wrong by QEMU because their allocation
length/parameter list length is placed in a non-standard position in
the CDB (i.e. it is different from most commands with the same value in
bits 5-7).
- SEND VOLUME TAG length was multiplied by 40 which is not in SMC. The
parameter list length is between 32 and 40 bytes. Same for MEDIUM SCAN
(spec found at http://ldkelley.com/SCSI2/SCSI2-16.html but not in any of
the PDFs I have here).
- READ_POSITION (SSC) conflicts with PRE_FETCH (SBC). READ_POSITION's
transfer length is not hardcoded to 20 in SSC; for PRE_FETCH cmd->xfer
should be 0. Both fixed.
- FORMAT MEDIUM (the SSC name for FORMAT UNIT) was missing. The FORMAT
UNIT command is still somewhat broken for block devices because its
parameter list length is not in the CDB. However it works for CD/DVD
drives, which mandate the length of the payload.
- fixed wrong sign-extensions for 32-bit fields (for the LBA field,
this affects disks >1 TB).
- several other SBC or SSC commands were missing or parsed wrong.
- some commands were not in the list of "write" commands.
Reported-by: Thomas Schmitt <scdbackup@gmx.net> Tested-by: Thomas Schmitt <scdbackup@gmx.net> (MMC bits only) Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Paolo Bonzini [Mon, 14 Nov 2011 13:31:47 +0000 (14:31 +0100)]
atapi: kill MODE SENSE(6), fix MODE SENSE(10)
Mode page 2A of emulated ATAPI DVD-ROM should have page length 0x14
like SCSI CD-ROM, rather than 0x12.
Mode page length is off by 8, as it should contain the length of the
payload after the first two bytes.
MODE SENSE(6) should be thrown out of ATAPI DVD-ROM emulation. It is
not specified in the ATAPI list of MMC-2, and MMC-5 prescribes to use
MODE SENSE(10). Anyway, its implementation is wrong.
Reported-by: Thomas Schmitt <scdbackup@gmx.net> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Paolo Bonzini [Tue, 15 Nov 2011 16:36:38 +0000 (17:36 +0100)]
scsi: fix fw path
The pre-1.0 firmware path for SCSI devices already included the LUN
using the suffix argument to add_boot_device_path. I missed that when
making channel and LUN customizable. Avoid that it is included twice, and
convert the colons to commas for consistency with other kinds of devices
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Kevin Wolf <kwolf@redhat.com>
.c files include .h files, so .o depends on .h,
and the linked result depends on .o.
We got it wrong for qga rules, fix it up.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
qga/guest-agent-commands.c includes qga-qmp-commands.h,
but it was missing in its dependencies. Add it in QGALIB_GEN.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
We have a single rule generating .c and .h files,
so .h doesn't depend on .c: both depend on the
source schema.
Fix Makefile to reflect that - without this,
if .c is there and .h is missing, Makefile does
not know how to remake .h and assumes it's
a dummy target, triggering endless rebuilds.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
make clean missed the source qmp files generated
by python. Fix that.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Luiz Capitulino <lcapitulino@redhat.com>
monitor: Fix file_completion() to check for stat() failure
stat() can fail for a file name just read with readdir(). Easiest way
to trigger is a dangling symbolic link --- look ma, no race! When it
fails, file_completion() uses sb.st_mode uninitialized. If the
directory bit happens to be set, it appends a "/" to the completed
name.
Signed-off-by: Markus Armbruster <armbru@redhat.com> Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Peter Maydell [Mon, 7 Nov 2011 13:25:45 +0000 (13:25 +0000)]
hw/omap_gpio: Fix infinite recursion when doing 8/16 bit reads
Fix a long-standing bug which meant that any attempt to do an
8 or 16 bit read from the OMAP GPIO module would cause qemu to
crash due to an infinite recursion.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
tcg: Use TCGReg for standard tcg-target entry points.
Including tcg_out_ld, tcg_out_st, tcg_out_mov, tcg_out_movi.
Signed-off-by: Richard Henderson <rth@twiddle.net> Reviewed-by: Andreas Färber <afaerber@suse.de> Reviewed-by: Stefan Weil <sw@weilnetz.de> Signed-off-by: Alexander Graf <agraf@suse.de>
tcg: Standardize on TCGReg as the enum for hard registers
Most targets did not name the enum; tci used TCGRegister.
Signed-off-by: Richard Henderson <rth@twiddle.net> Reviewed-by: Andreas Färber <afaerber@suse.de> Reviewed-by: Stefan Weil <sw@weilnetz.de> Signed-off-by: Alexander Graf <agraf@suse.de>
Alexander Graf [Fri, 7 Oct 2011 07:51:50 +0000 (09:51 +0200)]
s390x: Add shutdown for TCG s390-virtio machine
Now that we have code in place to do refcounting of online CPUs, we
can drag the TCG code along and implement shutdown for that one too,
so it doesn't feel left out by its KVM counterpart.
On s390 a shutdown is the state of all CPUs being either stopped
or disabled (for interrupts) waiting. We have to track the overall
number of running CPUs to call the shutdown sequence accordingly.
This patch implements the counting and shutdown handling for the
kvm path in qemu.
Lets also wrap changes to env->halted and env->exception_index.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Alexander Graf <agraf@suse.de>
The default kernel command line for s390 is
"root=/dev/ram0 ro"
When overriding this line, we have to ensure to also copy the \0 to
avoid false lines, for example, -append "root=/dev/vda" will result in
"root=/dev/vda0 ro" with the current code.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Alexander Graf <agraf@suse.de>
Alexander Graf [Thu, 14 Jul 2011 09:52:08 +0000 (11:52 +0200)]
s390x: implement SIGP restart and shutdown
An s390x OS does reboot and shutdown triggers through hypercalls that
we didn't implement on the TCG backend yet. That means that so far we
couldn't shut down virtual machines for example, having them hang on
shutdown when not using KVM.
With this patch, this restriction is gone. We can now shut down and
reboot s390x virtual machines even when using the TCG backend.
Alexander Graf [Thu, 14 Jul 2011 09:50:33 +0000 (11:50 +0200)]
s390x: implement rrbe instruction properly
The rrbe instruction resets the reference bit in the given storage key.
So far, we merely made it a nop and also returned an invalid CC value,
so that the kernel never knew if a page actually got accessed.
This patch implements it properly, flushing the R bit and returning the
correct CC value.
Alexander Graf [Thu, 14 Jul 2011 09:49:08 +0000 (11:49 +0200)]
s390x: update R and C bits in storage key
When the s390x maps a page or writes happen to a page, the R and C
bits get updated. The easiest way to implement this in qemu is to
simply update them whenever we map a TLB translation and act according
to the permissions.
Alexander Graf [Thu, 14 Jul 2011 09:45:07 +0000 (11:45 +0200)]
s390x: make ipte 31-bit aware
When running 31-bit code we can potentially map the same virtual
address twice - once as 0x0yyyyyyy and once as 0x8yyyyyyy, because
the upper bit gets ignored.
This also should be reflected in the tlb invalidation path, so we
really invalidate also the transparently created tlb entries.
Juha Riihimäki [Thu, 20 Oct 2011 12:53:35 +0000 (14:53 +0200)]
hw/onenand: reject read-only drives
Signed-off-by: Juha Riihimäki <juha.riihimaki@nokia.com> Reviewed-by: Markus Armbruster <armbru@redhat.com> Signed-off-by: Andrzej Zaborowski <andrew.zaborowski@intel.com>
projects / qemu.git / log