Fix bad error handling after memory_region_init_ram()
Symptom:
$ qemu-system-x86_64 -m 10000000
Unexpected error in ram_block_add() at /work/armbru/qemu/exec.c:1456:
upstream-qemu: cannot set up guest memory 'pc.ram': Cannot allocate memory
Aborted (core dumped)
Root cause: commit ef701d7 screwed up handling of out-of-memory
conditions. Before the commit, we report the error and exit(1), in
one place, ram_block_add(). The commit lifts the error handling up
the call chain some, to three places. Fine. Except it uses
&error_abort in these places, changing the behavior from exit(1) to
abort(), and thus undoing the work of commit 3922825 "exec: Don't
abort when we can't allocate guest memory".
The three places are:
* memory_region_init_ram()
Commit 4994653 (right after commit ef701d7) lifted the error
handling further, through memory_region_init_ram(), multiplying the
incorrect use of &error_abort. Later on, imitation of existing
(bad) code may have created more.
* memory_region_init_ram_ptr()
The &error_abort is still there.
* memory_region_init_rom_device()
Doesn't need fixing, because commit 33e0eb5 (soon after commit ef701d7) lifted the error handling further, and in the process
changed it from &error_abort to passing it up the call chain.
Correct, because the callers are realize() methods.
Fix the error handling after memory_region_init_ram() with a
Coccinelle semantic patch:
Eric Blake [Thu, 10 Sep 2015 16:19:16 +0000 (10:19 -0600)]
hmp: Allow for error message hints on HMP
Commits 7216ae3d and d2828429 disabled some error message hints,
all because a change to use modern error reporting meant that the
hint would be output prior to the actual error. Fix this by making
hints a first-class member of Error.
For example, we are now back to the pleasant:
$ qemu-system-x86_64 --nodefaults -S --vnc :0 --chardev null,id=,
qemu-system-x86_64: --chardev null,id=,: Parameter 'id' expects an identifier
Identifiers consist of letters, digits, '-', '.', '_', starting with a letter.
Stefan Hajnoczi [Mon, 10 Aug 2015 13:15:41 +0000 (14:15 +0100)]
error: only prepend timestamp on stderr
The -msg timestamp=on option prepends a timestamp to error messages.
This is useful on stderr where it allows users to identify when an error
was raised.
Timestamps do not make sense on the monitor since error_report() is
called in response to a synchronous monitor command and the user already
knows "when" the command was issued. Additionally, the rest of the
monitor conversation lacks timestamps so the error timestamp cannot be
correlated with other activity.
Only prepend timestamps on stderr. This fixes libvirt's 'drive_del'
processing, which did not expect a timestamp. Other QEMU monitor
clients are probably equally confused by timestamps on monitor error
messages.
Juan Quintela [Fri, 4 May 2012 10:54:34 +0000 (12:54 +0200)]
target-cris: update CPU state save/load to use VMStateDescription
Update the CRIS CPU state save/load to use a VMStateDescription struct
rather than cpu_save/cpu_load functions.
Have to define TLBSet struct.
Multidimensional arrays in C are a mess, just unroll them.
Signed-off-by: Juan Quintela <[email protected]>
[PMM:
* expand commit message a little since it's no longer one patch in
a 35-patch series
* add header/copyright comment to machine.c; credited copyright is
Red Hat and author is Juan, since this commit gives the file all-new
contents; license is LGPL-2-or-later, to match other target-cris code
* remove hardcoded tab
* add fields for locked_irq, interrupt_vector, fault_vector, trap_vector
* drop minimum_version_id_old fields
* bump version_id to 2 as we are not compatible with old state format
* remove unnecessary hw/boards.h include
* update to register via dc->vmsd] Signed-off-by: Peter Maydell <[email protected]> Acked-by: Edgar E. Iglesias <[email protected]>
Peter Maydell [Thu, 17 Sep 2015 12:07:50 +0000 (13:07 +0100)]
Merge remote-tracking branch 'remotes/stefanha/tags/net-pull-request' into staging
# gpg: Signature made Thu 17 Sep 2015 12:43:56 BST using RSA key ID 81AB73C8
# gpg: Good signature from "Stefan Hajnoczi <[email protected]>"
# gpg: aka "Stefan Hajnoczi <[email protected]>"
* remotes/stefanha/tags/net-pull-request:
net: smc91c111: flush packets on RCR register changes
net: smc91c111: gate can_receive() on rx FIFO having a slot
net: smc91c111: guard flush_queued_packets() on can_rx()
MAINTAINERS: Stefan will not maintain net subsystem
net: smc91c111: flush packets on RCR register changes
The SOFT_RST or RXEN in the control register can be used as a condition
to unblock the net layer via can_receive(). So check for possible
flushes on RCR changes. This will drop all pending packets on soft
reset or disable which is the functional intent of the can_receive()
logic.
net: smc91c111: gate can_receive() on rx FIFO having a slot
Return false from can_receive() when the FIFO doesn't have a free RX
slot. This fixes a bug in the current code where the allocated buffer
is freed before the fifo pop, triggering a premature flush of queued RX
packets. It also will handle a corner case, where the guest manually
frees the allocated buffer before popping the rx FIFO (hence it is not
enough to just delay the flush_queued_packets()).
net: smc91c111: guard flush_queued_packets() on can_rx()
Check that the core can once again receive packets before asking the
net layer to do a flush. This will make it more convenient to flush
packets when adding new conditions to can_receive.
Add missing if braces while moving the can_receive() core code.
Peter Maydell [Wed, 16 Sep 2015 17:06:54 +0000 (18:06 +0100)]
Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging
* Linux header update and cleanup
* Support for HyperV crash report
* Cleanup of target-specific HMP commands
* Multiarch batch
* Checkpatch fix for Perl 5.22
* NBD fix
* Revert incorrect commit 5243722376
# gpg: Signature made Wed 16 Sep 2015 16:39:01 BST using RSA key ID 78C7AE83
# gpg: Good signature from "Paolo Bonzini <[email protected]>"
# gpg: aka "Paolo Bonzini <[email protected]>"
* remotes/bonzini/tags/for-upstream: (24 commits)
nbd: release exp->blk after all clients are closed
checkpatch: Escape left braces in regex
monitor: uninclude cpu_ldst
include/exec: Move cputlb exec.c defs out
cputlb: Change tlb_set_dirty() arg to cpu
cputlb: move CPU_LOOP() for tlb_reset() to exec.c
translate: move real_host_page setting to -common
tcg: Move tci_tb_ptr to -common
tcg: split tcg_op_defs to -common
translate-all: Move tcg_handle_interrupt() to -common
cpu-exec: Migrate some generic fns to cpu-exec-common
qemu-char: Use g_new() & friends where that makes obvious sense
monitor: added generation of documentation for hmp-commands-info.hx
hmp-commands.hx: fix end of table info
monitor: remove target-specific code from monitor.c
hmp-commands-info: move info_cmds content out of monitor.c
i386/kvm: Hyper-v crash msrs set/get'ers and migration
kvm: Add kvm system event crash handler
cpu: Add crash_occurred flag into CPUState
target-i386: move asm-x86/hyperv.h to standard-headers
...
nbd: release exp->blk after all clients are closed
If the socket fd is shutdown, there may be some data which is received before
shutdown. We will read the data and do read/write in nbd_trip(). But the exp's
blk is NULL, and it will cause qemu crashed.
Move the architecture agnostic function prototypes for exec.c out of
cputlb.h to exec-all.h. This allows hiding of the arch specific
cputlb.h from exec.c which should be getting close to having no
architecture specifics. Prepares support for multi-arch, which will have
a minimal cpu.h that services exec.c but not cputlb.h.
Change tlb_set_dirty() to accept a CPU instead of an env pointer. This
allows for removal of another CPUArchState usage from prototypes that
need to be QOMified.
To prepare for multi-arch, cputlb.c should only have awareness of one
single architecture. This means it should not have access to the full
CPU lists which may be heterogeneous. Instead, push the CPU_LOOP() up
to the one and only caller in exec.c.
Move the size and mask globals for the "real" host page size to
translate-common. This is to allow system-level code to use
REAL_HOST_PAGE_ALIGN and friends in builds which hide translate-all
behind arch-obj.
tcg_op_defs (and the _max) are both needed by the TCI disassembler. For
multi-arch, tcg.c will be multiple-compiled (arch-obj) with its symbols
hidden from common code. So split the definition off to new file,
tcg-common.c which will remain a regular obj-y for use by both the TCI
disas as well as the multiple tcg.c's.
translate-all: Move tcg_handle_interrupt() to -common
Move this function to common code. It has no arch specific
dependencies. Prepares support for multi-arch where the translate-all
interface needs to be virtualised. One less thing to virtualise.
cpu-exec: Migrate some generic fns to cpu-exec-common
The goal is to split the functions such that cpu-exec is CPU specific
content, while cpus-exec-common.c is generic code only. The function
interface to cpu-exec needs to be virtualised to prepare support for
multi-arch and moving these definitions out saves bloating the QOM
interface. So move these definitions out of cpu-exec to a new module,
cpu-exec-common.
qemu-char: Use g_new() & friends where that makes obvious sense
g_new(T, n) is neater than g_malloc(sizeof(T) * n). It's also safer,
for two reasons. One, it catches multiplication overflowing size_t.
Two, it returns T * rather than void *, which lets the compiler catch
more type errors.
This commit only touches allocations with size arguments of the form
sizeof(T). Same Coccinelle semantic patch as in commit b45c03f.
Pavel Butsykin [Thu, 10 Sep 2015 15:39:00 +0000 (18:39 +0300)]
hmp-commands.hx: fix end of table info
The table info(information about the system state) closes earlier
and some of its elements are outside(trace-events, rocker, etc). This
can be confusing and lead to additional bugs.
Pavel Butsykin [Thu, 10 Sep 2015 15:38:59 +0000 (18:38 +0300)]
monitor: remove target-specific code from monitor.c
Move target-specific code out of /monitor.c to /target-*/monitor.c,
this will avoid code cluttering and using random ifdeffery. The solution
is quite simple, but solves the issue of the separation of target-specific
code from monitor.
i386/kvm: Hyper-v crash msrs set/get'ers and migration
KVM Hyper-V based guests can notify hypervisor about
occurred guest crash by writing into Hyper-V crash MSR's.
This patch does handling and migration of HV_X64_MSR_CRASH_P0-P4,
HV_X64_MSR_CRASH_CTL msrs. User can enable these MSR's by
'hv-crash' option.
KVM kernel can send guest crash events into userspace.
Appropriate guest crash handler is called when kernel guest
crash event received. Guest crash event recognized by a
KVM_SYSTEM_EVENT_CRASH type of system event.
Paolo Bonzini [Wed, 9 Sep 2015 13:25:52 +0000 (15:25 +0200)]
update-linux-headers: copy standard-headers files one by one
cp_virtio is called for both the asm-s390/ and linux/ directories,
so it looks for pci_regs.h and input.h files in asm-s390/ too. This
makes little sense. In the next patch we will have the opposite
problem; we want to add asm-x86/hyperv.h, and there's also a
linux/hyperv.h file with unwanted dependencies on additional Linux
uapi headers. We do not want to copy linux/hyperv.h.
The solution is to make cp_virtio (now renamed to cp_portable) copy
one file only, instead of using the "find" command, and call it multiple
times. The new function is really just a reindentation of the old one.
Paolo Bonzini [Thu, 10 Sep 2015 09:31:12 +0000 (11:31 +0200)]
pci: remove Link Training error from AER error list
The spec says:
Undefined – The value read from this bit is
undefined. In previous versions of this
specification, this bit was used to indicate a Link
Training Error. System software must ignore the
value read from this bit. System software is
permitted to write any value to this bit.
Peter Maydell [Wed, 16 Sep 2015 15:19:49 +0000 (16:19 +0100)]
Merge remote-tracking branch 'remotes/rth/tags/pull-target-i386-20150915' into staging
Exception handling improvments from Pavel Dovgalyuk.
# gpg: Signature made Tue 15 Sep 2015 20:36:14 BST using RSA key ID 4DD0279B
# gpg: Good signature from "Richard Henderson <[email protected]>"
# gpg: aka "Richard Henderson <[email protected]>"
# gpg: aka "Richard Henderson <[email protected]>"
* remotes/rth/tags/pull-target-i386-20150915:
target-i386: exception handling for other helper functions
target-i386: exception handling for seg_helper functions
target-i386: exception handling for memory helpers
target-i386: exception handling for div instructions
target-i386: exception handling for FPU instructions
target-i386: introduce new raise_exception functions
Peter Maydell [Tue, 15 Sep 2015 14:42:58 +0000 (15:42 +0100)]
Merge remote-tracking branch 'remotes/berrange/tags/vnc-crypto-v9-for-upstream' into staging
Merge vnc-crypto-v9
# gpg: Signature made Tue 15 Sep 2015 15:32:38 BST using RSA key ID 15104FDF
# gpg: Good signature from "Daniel P. Berrange <[email protected]>"
# gpg: aka "Daniel P. Berrange <[email protected]>"
* remotes/berrange/tags/vnc-crypto-v9-for-upstream:
ui: convert VNC server to use QCryptoTLSSession
ui: fix return type for VNC I/O functions to be ssize_t
crypto: introduce new module for handling TLS sessions
crypto: add sanity checking of TLS x509 credentials
crypto: introduce new module for TLS x509 credentials
crypto: introduce new module for TLS anonymous credentials
crypto: introduce new base module for TLS credentials
qom: allow QOM to be linked into tools binaries
crypto: move crypto objects out of libqemuutil.la
tests: remove repetition in unit test object deps
qapi: allow override of default enum prefix naming
Both ADDX_SPECIAL_0_OPCODE_Y1 and ADD_SPECIAL_0_OPCODE_Y1
do not appear to be "special" in any way, except that they
don't follow the normal naming convention using _RRR_.
Chen Gang [Thu, 20 Aug 2015 21:36:37 +0000 (05:36 +0800)]
linux-user: Support tilegx architecture in linux-user
Add main working flow feature, system call processing feature, and elf64
tilegx binary loading feature, based on Linux kernel tilegx 64-bit
implementation.
[rth: Moved all of the implementation of atomic instructions to a later patch.]
Switch VNC server over to using the QCryptoTLSSession object
for the TLS session. This removes the direct use of gnutls
from the VNC server code. It also removes most knowledge
about TLS certificate handling from the VNC server code.
This has the nice effect that all the CONFIG_VNC_TLS
conditionals go away and the user gets an actual error
message when requesting TLS instead of it being silently
ignored.
With this change, the existing configuration options for
enabling TLS with -vnc are deprecated.
This aligns VNC with the way TLS credentials are to be
configured in the future for chardev, nbd and migration
backends. It also has the benefit that the same TLS
credentials can be shared across multiple VNC server
instances, if desired.
If someone uses the deprecated syntax, it will internally
result in the creation of a 'tls-creds' object with an ID
based on the VNC server ID. This allows backwards compat
with the CLI syntax, while still deleting all the original
TLS code from the VNC server.
Chen Gang [Thu, 20 Aug 2015 21:35:43 +0000 (05:35 +0800)]
linux-user: tilegx: Add architecture related features
They are based on Linux kernel tilegx architecture for 64 bit binary,
and also based on tilegx ABI reference document, and also reference from
other targets implementations.
ui: fix return type for VNC I/O functions to be ssize_t
Various VNC server I/O functions return 'long' and then
also pass this to a method accepting 'int'. All these
should be ssize_t to match the signature of read/write
APIs and thus avoid potential for integer truncation /
wraparound.
crypto: introduce new module for handling TLS sessions
Introduce a QCryptoTLSSession object that will encapsulate
all the code for setting up and using a client/sever TLS
session. This isolates the code which depends on the gnutls
library, avoiding #ifdefs in the rest of the codebase, as
well as facilitating any possible future port to other TLS
libraries, if desired. It makes use of the previously
defined QCryptoTLSCreds object to access credentials to
use with the session. It also includes further unit tests
to validate the correctness of the TLS session handshake
and certificate validation. This is functionally equivalent
to the current TLS session handling code embedded in the
VNC server, and will obsolete it.
crypto: add sanity checking of TLS x509 credentials
If the administrator incorrectly sets up their x509 certificates,
the errors seen at runtime during connection attempts are very
obscure and difficult to diagnose. This has been a particular
problem for people using openssl to generate their certificates
instead of the gnutls certtool, because the openssl tools don't
turn on the various x509 extensions that gnutls expects to be
present by default.
This change thus adds support in the TLS credentials object to
sanity check the certificates when QEMU first loads them. This
gives the administrator immediate feedback for the majority of
common configuration mistakes, reducing the pain involved in
setting up TLS. The code is derived from equivalent code that
has been part of libvirt's TLS support and has been seen to be
valuable in assisting admins.
It is possible to disable the sanity checking, however, via
the new 'sanity-check' property on the tls-creds object type,
with a value of 'no'.
Unit tests are included in this change to verify the correctness
of the sanity checking code in all the key scenarios it is
intended to cope with. As part of the test suite, the pkix_asn1_tab.c
from gnutls is imported. This file is intentionally copied from the
(long since obsolete) gnutls 1.6.3 source tree, since that version
was still under GPLv2+, rather than the GPLv3+ of gnutls >= 2.0.
crypto: introduce new module for TLS x509 credentials
Introduce a QCryptoTLSCredsX509 class which is used to
manage x509 certificate TLS credentials. This will be
the preferred credential type offering strong security
characteristics
The 'id' value in the -object args will be used to associate the
credentials with the network services. For example, when the VNC
server is later converted it would use
crypto: introduce new module for TLS anonymous credentials
Introduce a QCryptoTLSCredsAnon class which is used to
manage anonymous TLS credentials. Use of this class is
generally discouraged since it does not offer strong
security, but it is required for backwards compatibility
with the current VNC server implementation.
The 'id' value in the -object args will be used to associate the
credentials with the network services. For example, when the VNC
server is later converted it would use
crypto: introduce new base module for TLS credentials
Introduce a QCryptoTLSCreds class to act as the base class for
storing TLS credentials. This will be later subclassed to provide
handling of anonymous and x509 credential types. The subclasses
will be user creatable objects, so instances can be created &
deleted via 'object-add' and 'object-del' QMP commands respectively,
or via the -object command line arg.
If the credentials cannot be initialized an error will be reported
as a QMP reply, or on stderr respectively.
The idea is to make it possible to represent and manage TLS
credentials independently of the network service that is using
them. This will enable multiple services to use the same set of
credentials and minimize code duplication. A later patch will
convert the current VNC server TLS code over to use this object.
The representation of credentials will be functionally equivalent
to that currently implemented in the VNC server with one exception.
The new code has the ability to (optionally) load a pre-generated
set of diffie-hellman parameters, if the file dh-params.pem exists,
whereas the current VNC server will always generate them on startup.
This is beneficial for admins who wish to avoid the (small) time
sink of generating DH parameters at startup and/or avoid depleting
entropy.
The qom objects are currently added to common-obj-y
which is only linked into the system emulators. The
later crypto patches will depend on QOM infrastructure
and will also be used from tools binaries. Thus the QOM
objects are moved into a new qom-obj-y variable which
can be referenced when linking tools, system emulators
and tests.
Future patches will be adding more crypto related APIs which
rely on QOM infrastructure. This creates a problem, because
QOM relies on library constructors to register objects. When
you have a file in a static .a library though which is only
referenced by a constructor the linker is dumb and will drop
that file when linking to the final executable :-( The only
workaround for this is to link the .a library to the executable
using the -Wl,--whole-archive flag, but this creates its own
set of problems because QEMU is relying on lazy linking for
libqemuutil.a. Using --whole-archive majorly increases the
size of final executables as they now contain a bunch of
object code they don't actually use.
The least bad option is to thus not include the crypto objects
in libqemuutil.la, and instead define a crypto-obj-y variable
that is referenced directly by all the executables that need
this code (tools + softmmu, but not qemu-ga). We avoid pulling
entire of crypto-obj-y into the userspace emulators as that
would force them to link to gnutls too, which is not required.
Peter Maydell [Tue, 15 Sep 2015 13:11:28 +0000 (14:11 +0100)]
Merge remote-tracking branch 'remotes/kraxel/tags/pull-gtk-20150915-1' into staging
gtk: misc grab tweaks, locale fix.
# gpg: Signature made Tue 15 Sep 2015 11:35:36 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-gtk-20150915-1:
gtk: use setlocale() for LC_MESSAGES only
gtk: don't grab input when entering fullscreen.
gtk: set free_scale when setting zoom_fit
gtk: trace input grab reason
gtk: move gd_update_caption calls to gd_{grab,ungrab}_{pointer,keyboard}
gtk: check for existing grabs in gd_grab_{pointer,keyboard}
P J P [Tue, 15 Sep 2015 11:16:59 +0000 (16:46 +0530)]
net: avoid infinite loop when receiving packets(CVE-2015-5278)
Ne2000 NIC uses ring buffer of NE2000_MEM_SIZE(49152)
bytes to process network packets. While receiving packets
via ne2000_receive() routine, a local 'index' variable
could exceed the ring buffer size, leading to an infinite
loop situation.
P J P [Tue, 15 Sep 2015 11:10:49 +0000 (16:40 +0530)]
net: add checks to validate ring buffer pointers(CVE-2015-5279)
Ne2000 NIC uses ring buffer of NE2000_MEM_SIZE(49152)
bytes to process network packets. While receiving packets
via ne2000_receive() routine, a local 'index' variable
could exceed the ring buffer size, which could lead to a
memory buffer overflow. Added other checks at initialisation.
Alberto Garcia [Thu, 10 Sep 2015 15:19:32 +0000 (18:19 +0300)]
gtk: use setlocale() for LC_MESSAGES only
The QEMU code is not internationalized and assumes that it runs under
the C locale, but if we use the GTK+ UI we'll end up importing the
locale settings from the environment. This can break things, such as
the JSON generator and iotest 120 in locales that use a decimal comma.
We do however have translations for a few simple strings for the GTK+
menu items, so in order to run QEMU using the C locale, and yet have a
translated UI let's use setlocale() for LC_MESSAGES only.
Kick off all grabbing logic from fullscreen mode. In the current state
it seems to create more problems than it solves. Try running qemu/gtk
fullscreen on one head of a multihead host for example ...
There probably was a reason the grab-on-fullscreen logic was added in
the first place. So please test and report any issues so we can try to
find a sane way to handle it.
projects / qemu.git / log