cache=unsafe completely ignored bdrv_flush, because flushing the host disk
costs a lot of performance. However, this means that qcow2 images (and
potentially any other format) can lose data even after the guest has issued a
flush if the qemu process crashes/is killed. In case of a host crash, data loss
is certainly expected with cache=unsafe, but if just the qemu process dies this
is a bit too unsafe.
Now that we have two separate flush functions, we can choose to flush
everythign to the OS, but don't enforce that it's physically written to the
disk.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
qcow2 has a writeback metadata cache, so flushing a qcow2 image actually
consists of writing back that cache to the protocol and only then flushes the
protocol in order to get everything stable on disk.
This introduces a separate bdrv_co_flush_to_os to reflect the split.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
There are two different types of flush that you can do: Flushing one level up
to the OS (i.e. writing data to the host page cache) or flushing it all the way
down to the disk. The existing functions flush to the disk, reflect this in the
function name.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Fix a use-while-uninitialized of the fd_type[] array (introduced
in commit 34d4260e1, noticed by Coverity). This is more theoretical
than practical, since it's quite hard to get here with floppy==NULL
(the qdev_try_create() of the isa-fdc device has to fail).
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The Data Offset field in the Dynamic Disk Header is an 8 byte field.
Although the specification (2006-10-11) gives an example of initializing
only the first 4 bytes, images generated by Microsoft on Windows initialize
all 8 bytes.
Failure to initialize all 8 bytes results in errors from utilities
like Citrix's vhd-util which checks specifically for the proper Data
Offset field initialization.
Signed-off-by: Charles Arnold <carnold@suse.com>
Reviewed-by: Andreas Färber <afaerber@suse.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This is cleaner, because we do not need to close the block device when
there is an error opening /dev/nbdX. It was done this way only to
print errors before daemonizing.
At the same time, use atexit to ensure that the block device is closed
whenever we exit.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Now that the client and server are in the same process, there is
no need to race on the creation of the socket. We can open the
listening socket before starting the client thread.
This avoids that "qemu-nbd -v -c" prints this once before connecting
successfully to the socket:
connect(unix:/var/lock/qemu-nbd-nbd0): No such file or directory
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
In order to get nice error messages, keep the qemu-nbd process running
until before issuing NBD_DO_IT and connected to the daemon with a pipe.
This lets the qemu-nbd process relay error messages from the daemon and
exit with a nonzero status if appropriate.
Suggested-by: Kevin Wolf <kwolf@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This avoids that qemu-nbd uses both forking and threads, which do
not behave well together.
qemu-nbd is already Unix only, and there is no qemu_thread_join,
so for now use pthreads.
Since the parent and child no longer have separate file descriptors,
we can open the NBD device before daemonizing, instead of checking
with access(2) and restricting the open to the client only.
Reported-by: Pierre Riteau <pierre.riteau@irisa.fr>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
It will be moved to a global variable by the next patch, and it
would conflict with the socket function.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
The client process right now uses SIGTERM to interrupt the server side.
This does not affect the exit status of "qemu-nbd -v -c" because the
server is a child process. This will change when both sides will be
in the same process, and anyway cleaning up things nicely upon SIGTERM
is good practice.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
This can be seen with "qemu-nbd -v -c", which returns 1 instead of 0
when you disconnect with "qemu-nbd -d".
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Recent versions of udev always keep the tray locked so that the kernel
can observe "eject request" events (aka tray button presses) even on
discs that aren't mounted. Add support for these events in the ATAPI
and SCSI cd drive device models.
To let management cope with the behavior of udev, an event should also
be added for "tray opened/closed". This way, after issuing an "eject"
command, management can poll until the guests actually reacts to the
command. They can then issue the "change" command after the tray has been
opened, or try with "eject -f" after a (configurable?) timeout. However,
with this patch and the corresponding support in the device models,
at least it is possible to do a manual two-step eject+change sequence.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
vvfat used to directly call into the qcow2 block driver instead of using the
block.c wrappers. With the coroutine conversion, this stopped working.
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Add a missing 'break' statement to fix a buffer overrun when
executing the EEPROM write-all command. Spotted by Coverity
(see bug 887883).
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
To reproduce the leak, put two name options into the same [cpudef]
section of target-x86_64.conf.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Fixes protocol_client_auth_sasl_mechname() not to crash when malloc()
fails. Spotted by Coverity.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Update the document since the default code cache size is 32 MB now.
Signed-off-by: chenwj <chenwj@cs.nctu.edu.tw>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
hpet_timer timer callback rearms itself based on difference between
current HPET tick counter and comparator value. Difference calculated by
the hpet_calculate_diff function is limited to non-negative values.
cur_tick is calculated via hpet_get_ticks that uses qemu_get_clock_ns(vm_clock).
With -icount enabled vm_clock doesn't advance during qemu_run_timers
loop thus once difference is zero, qemu_run_timers loops forever
handling hpet_timer.
Limit hpet_calculate_diff results to positive only values to avoid that
infinite loop.
This fixes the following qemu-system-x86_64 hang when it reaches
timer_irq_works() in the linux bootup:
[ 0.000000] Fast TSC calibration using PIT
[ 0.000000] Detected 1000.054 MHz processor.
[ 0.000031] Calibrating delay loop (skipped), value calculated using timer frequency.. 2000.10 BogoMIPS (lpj=10000540)
[ 0.000404] pid_max: default: 32768 minimum: 301
[ 0.001138] Mount-cache hash table entries: 256
[ 0.003883] Initializing cgroup subsys ns
[ 0.004035] Initializing cgroup subsys cpuacct
[ 0.004280] Initializing cgroup subsys freezer
[ 0.004790] Performance Events: AMD PMU driver.
[ 0.004985] ... version: 0
[ 0.005134] ... bit width: 48
[ 0.005285] ... generic registers: 4
[ 0.005437] ... value mask: 0000ffffffffffff
[ 0.005625] ... max period: 00007fffffffffff
[ 0.005807] ... fixed-purpose events: 0
[ 0.005957] ... event mask: 000000000000000f
[ 0.006275] SMP alternatives: switching to UP code
Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Add missing 'break' statements which would have meant that writing
to an 8 bit NAND device was broken. Spotted by Coverity (see bug
887883).
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Fix a bug revealed by a coverity scan (see bug 887883) which meant
that we would never print the warning about unpredictable behaviour
if a nonexistent overlay is enabled.
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The fact that a host cpu supports a feature doesn't mean that QEMU and KVM
will also support it, yet -cpuid host brings host features wholesale.
We need to whitelist each feature separately to make sure we support it.
This patch adds KVM whitelisting (by simply using KVM_GET_SUPPORTED_CPUID
instead of the CPUID instruction).
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Modern distributions place xattr.h in /usr/include/sys, and fold
libattr.so into libc. They also don't have an ENOATTR.
Make configure detect this, and add a qemu-xattr.h file that
directs the #include to the right place.
Signed-off-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
g_strdup() can't fail, remove assertion. Assert its argument can't be
null, because that's not obvious (add_boot_device_path() ensures it).
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The function that writes pidfile for win32 uses WriteFileEx which is an
asynchronous IO function. The arguments given to WriteFileEx are allocated on
the stack and one of them is "in out". When the IO operation is actually
executed the calling function has already returned, so the arguments are no
longer allocated or allocated to another frame.
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Fabien Chouteau <chouteau@adacore.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The non-dynticks timer variations are broken, so they can be
removed.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
vga_putcharxy()'s underline code sets font_data to 0xffff instead of
0xff. vga_putcharxy() then reads dmask16[0xffff >> 4] and
dmask4[0xffff >> 6]. In practice, these out-of-bounds subscripts
"only" put a few crap bits into the display surface.
For 32 bit pixels, there's no array access. font_data's extra bits go
straight into the display surface.
Broken when commit 6d6f7c28 implemented underline.
Spotted by Coverity.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The caller of qemu_timedate_diff() does not expect that tm it passes to
the function will be modified, but mktime() is destructive and modifies
its argument. Pass a copy of tm to it and set tm_isdst so that mktime()
will not rely on it since its value may be outdated.
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
We disable vm_clock when pausing all vcpus, but we forget to
reenable it when resuming all vcpus. It will cause that the
guest can not be rebooted.
Tested-by: Zhi Yong Wu <zwu.kernel@gmai.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Commit 0a039dc700 broke vga modes for
qxl-vga by loosing vga_ioport_read windup. qxl needs to hook into
vga port writes only and used to realize that by letting vga_init() do
the work for both reads and writes, then overwrite the write function.
That little detail was missed while doing the conversion ...
This patch fixes it. It also switch qxl vga ioport registration to
portio lists while being at it.
Cc: Hans de Goede <hdegoede@redhat.com>
Acked-by: Alon Levy <alevy@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
This patch removes the code lines which set the subsystem id for the
emulated ac97 card to 8086:0000. Due to the device id being zero the
subsystem id isn't vaild anyway. With the patch applied the sound card
gets the default qemu subsystem id (1af4:1100) instead.
[ v2: old & broken id is maintained for -M pc-$oldqemuversion ]
Cc: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
sgabios hasn't gotten a lot of coverage since it was not shipped. For 1.0,
let's disable the automatic loading of the option ROM in -nographic
mode. We can put it back for 1.1.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
The rom was not added together with the sgabios device and is
not installed.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Somehow, the read/write functions handle an offset that does not exist anymore.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
readline_hist_add() moves the history entry to the end of history. It
uses memmove() to move rs->history[idx + 1..] to rs->history[idx..].
However, its size argument is off by two array elements, so it writes
one element beyond rs->history[], and reads two.
On my system, this clobbers rs->hist_entry and the hole right after
it. Since the function assigns to rs->hist_entry in time, the bug has
no ill effects for me.
Spotted by Coverity.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
Before the next patches, fix coding style of the affected functions.
Signed-off-by: Pavel Borzenkov <pavel.borzenkov@gmail.com>
Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
The first enable set/clear register (which controls the PPIs and SGIs)
is supposed to be banked for each processor. Currently it is just
handled globally and this prevents recent SMP Linux kernels from
booting, because CPU0 stops receiving localtimer interrupts when CPU1
disables them locally.
To fix this, allow the enable bits to be enabled per-cpu. For SPIs,
always enable/disable ALL_CPU_MASK.
Signed-off-by: Rabin Vincent <rabin@rab.in>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
