Paolo Bonzini [Wed, 10 Mar 2010 10:38:54 +0000 (11:38 +0100)]
disentangle tcg and deadline calculation
Just tell main_loop_wait whether to be blocking or nonblocking, so that
there is no need to call qemu_cpus_have_work from the timer subsystem.
Instead, tcg_cpu_exec can say "we want the main loop not to block because
we have stuff to do".
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Paolo Bonzini [Wed, 10 Mar 2010 10:38:43 +0000 (11:38 +0100)]
tweak qemu_notify_event
Instead of testing specially next_cpu in host_alarm_handler, just do
that in qemu_notify_event. The idea is, if we are not running (or
not yet running) target CPU code, prepare things so that the execution
loop is exited asap; just make that clear.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Paolo Bonzini [Wed, 10 Mar 2010 10:38:42 +0000 (11:38 +0100)]
do not use qemu_event_increment outside qemu_notify_event
qemu_notify_event in the non-iothread case is only stopping the current
CPU. However, if the CPU is idle and the main loop is in the select
call then a call to qemu_event_increment is needed too (as done in
host_alarm_handler). Since in general one doesn't know whether the CPU
is executing or not, it is a safe bet to always do qemu_event_increment.
Another way to see it: after this patch qemu_event_increment is the
"common part" of qemu_notify_event for both the CONFIG_IOTHREAD and
!CONFIG_IOTHREAD cases, which makes sense.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Paolo Bonzini [Wed, 10 Mar 2010 10:38:41 +0000 (11:38 +0100)]
more alarm timer cleanup
The timer_alarm_pending variable is related to the alarm timer but not
placed in the struct. Also, in qemu_mod_timer the wrong flag was being
tested: the timer is rearmed in the alarm timer "bottom half", so the
right flag to test there is the "pending" flag.
Finally, I hoisted the NULL checks from alarm_has_dynticks to
host_alarm_handler.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Paolo Bonzini [Wed, 10 Mar 2010 10:38:38 +0000 (11:38 +0100)]
avoid dubiously clever code in win32_start_timer
The code is initializing an unsigned int to UINT_MAX using "-1", so that
the following always-true comparison seems to be always-false at a
first look. Since alarm timer initializations are never nested, it is
simpler to unconditionally store the result of timeGetDevCaps into
data->period.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Adam Litke [Tue, 9 Mar 2010 18:54:09 +0000 (12:54 -0600)]
balloon: Do not save VM state wrt asynchronous virtio operations
When working with the VM state (for loadvm/savevm and migration), it is not
valid to load and store pointers since the validity of those pointers cannot be
assured in the new qemu address space. Therefore, virtio_balloon_save() and
virtio_balloon_load() must not handle the stats-related fields in struct
VirtIOBalloon.
If a memory stats request is in-flight at the time of a migration or savevm,
the request will not complete and should be resubmitted once migration or
loadvm completes. Note that this extremely small race window can only be
triggered using QMP so it is not possible to hang the user monitor.
Signed-off-by: Adam Litke <agl@us.ibm.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Add a logical block size attribute as various guest side tools only
increase the filesystem sector size based on it, not the advisory
physical block size.
For scsi we already have support for a different logical block size
in place for CDROMs that we can built upon. Only my recent block
device characteristics VPD page needs some fixups. Note that we
leave the logial block size for CDROMs hardcoded as the 2k value
is expected for it in general.
For virtio-blk we already have a feature flag claiming to support
a variable logical block size that was added for the s390 kuli
hypervisor. Interestingly it does not actually change the units
in which the protocol works, which is still fixed at 512 bytes,
but only communicates a different minimum I/O granularity. So
all we need to do in virtio is to add a trap for unaligned I/O
and round down the device size to the next multiple of the logical
block size.
IDE does not support any other logical block size than 512 bytes.
Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Chris Webb [Mon, 8 Mar 2010 14:34:49 +0000 (14:34 +0000)]
Fix SIGFPE for vnc display of width/height = 1
During boot, the screen gets resized to height 1 and a mouse click at this
point will cause a division by zero when calculating the absolute pointer
position from the pixel (x, y). Return a click in the middle of the screen
instead in this case.
Signed-off-by: Chris Webb <chris@arachsys.com> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Bjørn Mork [Mon, 8 Mar 2010 12:07:14 +0000 (13:07 +0100)]
sdl: improve error message on fatal error
The SDL_SetVideoMode() error condition is easily triggered by a user by
simply configure a guest with a host unsupported display resolution
and attempting to enable fullscreen. Since the error is fatal, adding
a bit of debugging help can't harm.
Sample output with this change:
(qemu) Could not open SDL display (1280x1024x32): No video mode large enough for 1280x1024
The width x height might seem redundant as SDL also provides it in
SDL_GetError(), but I believe there are situations where it is
useful. I.e. if there is some other SDL error. Anyway, redundant
information in fatal error messages has never harmed a single gerbil.
Signed-off-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Paul Brook [Wed, 17 Mar 2010 02:14:28 +0000 (02:14 +0000)]
Large page TLB flush
QEMU uses a fixed page size for the CPU TLB. If the guest uses large
pages then we effectively split these into multiple smaller pages, and
populate the corresponding TLB entries on demand.
When the guest invalidates the TLB by virtual address we must invalidate
all entries covered by the large page. However the address used to
invalidate the entry may not be present in the QEMU TLB, so we do not
know which regions to clear.
Implementing a full vaiable size TLB is hard and slow, so just keep a
simple address/mask pair to record which addresses may have been mapped by
large pages. If the guest invalidates this region then flush the
whole TLB.
qemu-option: Move the implied first name into QemuOptsList
We sometimes permit omitting the first option name, for example
-device foo is short for -device driver=foo. The name to use
("driver" in the example) is passed as argument to qemu_opts_parse().
For each QemuOptsList, we use at most one such name.
Move the name into QemuOptsList, and pass whether to permit the
abbreviation. This ensures continued consistency, and simplifies the
commit after next in this series.
error: Let converted handlers print in human monitor
While fully converted handlers are not supposed to print anything when
running in a QMP monitor, they are free to print in a human monitor.
For instance, device_add (not yet converted) prints help, and will
continue to do so after conversion.
Moreover, utility functions converted to QError should remain usable
from unconverted handlers.
Two problems:
* handler_audit() complains when a converted handler prints. Limit
that to QMP monitors.
* With QMP, handlers need to pass the error object by way of
monitor_set_error(). However, we do that both for QMP and for the
human monitor. The human monitor prints the error object after the
handler returns. If the handler prints anything else, that output
"overtakes" the error message.
Limit use of monitor_set_error() to QMP monitors. Update
handler_audit() accordingly.
Users can't set them, so qdev_device_help() shouldn't list them. Fix
that. Also make qdev_prop_parse() hide them instead of printing a
meaningless "has no parser" error message.
Their value means nothing to users, so qdev_print_props() shouldn't
print it. Fix by removing their print method.
Users can't create them, so qdev_device_help() shouldn't list them.
Fix that.
Also make qdev_device_add() pretend they don't exist. Before, it
rejected them with a "can't be added via command line" message, which
wasn't quite right for monitor command device_add.
error: Infrastructure to track locations for error reporting
New struct Location holds a location. So far, the only location is
LOC_NONE, so this doesn't do anything useful yet.
Passing the current location all over the place would be too
cumbersome. Hide it away in static cur_loc instead, and provide
accessors. Print it in error_report().
Store it in QError, and print it in qerror_print().
Store it in QemuOpt, for use by qemu_opts_foreach(). This makes
error_report() do the right thing when it runs within
qemu_opts_foreach().
We may still have to store it in other data structures holding user
input for better error messages. Left for another day.
error_report() terminates the message with a newline. Strip it it
from its arguments.
This fixes a few error messages lacking a newline:
net_handle_fd_param()'s "No file descriptor named %s found", and
tap_open()'s "vnet_hdr=1 requested, but no kernel support for
IFF_VNET_HDR available" (all three versions).
There's one place that passes arguments without newlines
intentionally: load_vmstate(). Fix it up.
error: Don't abuse qemu_error() for non-error in scsi_hot_add()
Commit 30d335d6 converted an informational message from
monitor_printf() to qemu_error(), probably because the latter doesn't
need a mon argument. A later commit will make qemu_error() print
additional stuff that is only appropriate for proper errors, and then
this will break. Clean it up.
error: Don't abuse qemu_error() for non-error in qbus_find()
qbus_find() adds an informational line to error messages, and prints
both lines with one qemu_error(). Use error_printf() for the
informational line instead.
While there, simplify: instead of printing buffers filled by
qbus_list_bus() and qbus_list_dev() in one go, make them print it.
error: Don't abuse qemu_error() for non-error in qdev_device_help()
qdev_device_help() prints device information with qemu_error(). A
later commit will make qemu_error() print additional stuff that is
only appropriate for proper errors, and then this will break. Use
error_printf() instead.
While there, simplify: instead of printing a buffer filled by
qdev_print_devinfo() in one go, make qdev_print_devinfo() print it.
qemu_error_sink can either point to a monitor or a file. In practice,
it always points to the current monitor if we have one, else to
stderr. Simply route errors to the current monitor or else to stderr,
and remove qemu_error_sink along with the functions to control it.
Actually, the old code switches the sink slightly later, in
handle_user_command() and handle_qmp_command(), than it gets switched
now, implicitly, by setting the current monitor in monitor_read() and
monitor_control_read(). Likewise, it switches back slightly earlier
(same places). Doesn't make a difference, because there are no calls
of qemu_error() in between.
monitor: Separate "default monitor" and "current monitor" cleanly
Commits 376253ec..731b0364 introduced global variable cur_mon, which
points to the "default monitor" (if any), except during execution of
monitor_read() or monitor_control_read() it points to the monitor from
which we're reading instead (the "current monitor"). Monitor command
handlers run within monitor_read() or monitor_control_read().
Default monitor and current monitor are really separate things, and
squashing them together is confusing and error-prone.
For instance, usb_host_scan() can run both in "info usbhost" and
periodically via usb_host_auto_check(). It prints to cur_mon, which
is what we want in the former case: the monitor executing "info
usbhost". But since that's the default monitor in the latter case, it
periodically spams the default monitor there.
A few places use cur_mon to log stuff to the default monitor. If we
ever log something while cur_mon points to current monitor instead of
default monitor, the log temporarily "jumps" to another monitor.
Whether that can or cannot happen isn't always obvious.
Maybe logging to the default monitor (which may not even exist) is a
bad idea, and we should log to stderr or a logfile instead. But
that's outside the scope of this commit.
Change cur_mon to point to the current monitor. Create new
default_mon to point to the default monitor. Update users of cur_mon
accordingly.
This fixes the periodical spamming of the default monitor by
usb_host_scan(). It also stops "log jumping", should that problem
exist.
Commit 0ecdffbb created pc_boot_set() for use from monitor command
"boot_set", via qemu_boot_set(). pc_boot_set() reports errors to
cur_mon, which works fine for monitor code.
Commit e0f084bf reused the function int reset handler
restore_boot_devices(). Use of cur_mon is problematic in that
context. For instance, the "Too many boot devices for PC" error for
"-boot order=abcdefgh,once=c" goes to the monitor instead of stderr.
The monitor may not even exist.
Aurelien Jarno [Sun, 14 Mar 2010 20:20:59 +0000 (21:20 +0100)]
load_elf: replace the address addend by a translation function
A few machines need to translate the ELF header addresses into physical
addresses. Currently the only possibility is to add a value to the
addresses.
This patch replaces the addend argument by and a translation function
and an opaque passed to the function. A NULL function does not translate
the address.
The patch also convert all machines that have an addend, simplify the
PowerPC kernel loading and fix the MIPS kernel loading using this new
feature. Other machines may benefit from this feature.
Real pcnet device (AT2450) apparently has subsystem
device and vendor id set to 0, this is out of spec
(which requires that vendor id is obtained from PCI SIG)
but windows xp driver seems to need this in order
to associate.
eepro100 uses macros which rely on a specific
local variable name (pci_conf) which is scary.
Some of the uses are wrong or unnecessary,
remove them. The rest are small in number, open-code
them using pci_set_xx functions.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Aurelien Jarno [Sun, 14 Mar 2010 22:01:01 +0000 (23:01 +0100)]
tcg: declare internal helpers as const and pure
TCG internal helpers only access to the values passed in arguments, and
do not modify the CPU internal state. Thus they can be declared as
const and pure.
Aurelien Jarno [Tue, 2 Mar 2010 22:16:36 +0000 (23:16 +0100)]
tcg: add div/rem 32-bit helpers
Some targets like ARM would benefit to use 32-bit helpers for
div/rem/divu/remu.
Create a #define for div2 so that targets can select between
div, div2 and helper implementation. Use the helper version if none
of the #define are present.
Andre Przywara [Thu, 11 Mar 2010 13:39:06 +0000 (14:39 +0100)]
x86/cpuid: fix CPUID levels
Bump up the xlevel number for qemu32 to allow parsing of the processor
name string for this model.
Similiarly the 486 processor should have at least the feature bit
leaf enabled.
Signed-off-by: Andre Przywara <andre.przywara@amd.com> Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
projects / qemu.git / log