usb-mtp: fix bounds check for guest provided filename
The ObjectInfo struct has a variable length array containing the UTF-16
encoded filename. The number of characters of trailing data is given by
the 'length' field in the struct and this must be validated against the
size of the data packet received from the guest.
Since the data is UTF-16, we must convert the byte count we have to a
character count before validating. This must take care to truncate if
a malicious guest sent an odd number of bytes.
Kevin Wolf [Mon, 15 Apr 2019 15:54:50 +0000 (17:54 +0200)]
qcow2: Fix preallocation bdrv_pwrite to wrong file
With an external data file, preallocate_co() must write the final byte
to the external data file, not to the qcow2 image file.
This is harmless for preallocation of newly created images (only the
qcow2 file size is increased to the virtual disk size while it should be
much smaller), but with preallocated resize, it could in theory cause
visible corruption if the metadata of the image is larger than the data
(e.g. lots of bitmaps).
Commit 767abe7 ("chardev: forbid 'wait' option with client sockets")
is a bit too strict. Current libvirt always set wait=false, and will
thus fail to add client chardev.
Make the code more permissive, allowing wait=false with client socket
chardevs. Deprecate usage of 'wait' with client sockets.
John Snow [Tue, 9 Apr 2019 21:06:55 +0000 (17:06 -0400)]
qemu-img: fix .hx and .texi disparity
It turns out that having options listed in three places continues to be
a bad idea. I'm still toying with the idea of an improved infrastructure
here, but in the meantime, another bandaid.
There are three locations:
(1) .hx file, formatted as texi
(2) .hx file, formatted as human readable.
(3) .texi file, as section headers, formatted as texi.
You can compare the two summaries within the .hx file like so:
Peter Maydell [Fri, 12 Apr 2019 10:23:14 +0000 (11:23 +0100)]
Merge remote-tracking branch 'remotes/dgibson/tags/ppc-for-4.0-20190412' into staging
ppc patch queue for 2018-04-12
Here's a last minute pull request for 4.0. Turns out my last pull
request, to fix a regression in extended config space access for the
pseries machine didn't fix things hard enough. This PR has a single
patch which improves the fix to work in more cases.
It's a ghastly, ghastly hack, but it's simple and localized. I
already have patches almost ready to go in 4.1 that provides a simpler
and cleaner solution to all this.
Greg Kurz [Thu, 11 Apr 2019 16:32:24 +0000 (18:32 +0200)]
spapr_pci: Fix broken naming of PCI bus
Recent commit 5cf0d326a0fe fixed a regression which was preventing the
guest to access the extended config space of a PCIe device. This was
done by introducing a new PCI bus subtype for PAPR. The original fix
was causing PCI busses to be named "spapr-pci-host-bridge-root-bus.N"
instead of "pci.N", which was making upper layers unhappy of course.
This got worked around by hardcoding the PCI bus name to "pci.0", but
this only works for the default PHB. And we're now hitting:
# qemu-system-ppc64 \
-device spapr-pci-host-bridge,index=1 \
-device e1000e,bus=pci.0 \
-device e1000e,bus=pci.1
qemu-system-ppc64: -device e1000e,bus=pci.1: Bus 'pci.1' not found
David already posted some patches [1] to control PCI extended config
space accesses with a new flag in the base PCI bus class instead of
subtyping. These patches are a bit more intrusive though, and
are targetted for 4.1.
When no name is passed to pci_register_bus(), the core device code
generates a lowercase name based on the QOM typename. The typename
for the base PCI bus class is "PCI", hence the "pci.0", "pci.1"
bus names. Rename the type of the PAPR PCI bus to "pci", so that
the QOM code can generate proper names. This is a hack but it is
enough to fix the regression. And all this will be reworked properly
in 4.1.
device_tree: Fix integer overflowing in load_device_tree()
If the value of get_image_size() exceeds INT_MAX / 2 - 10000, the
computation of @dt_size overflows to a negative number, which then
gets converted to a very large size_t for g_malloc0() and
load_image_size(). In the (fortunately improbable) case g_malloc0()
succeeds and load_image_size() survives, we'd assign the negative
number to *sizep. What that would do to the callers I can't say, but
it's unlikely to be good.
Fix by rejecting images whose size would overflow.
Peter Maydell [Tue, 9 Apr 2019 15:18:30 +0000 (16:18 +0100)]
migration/ram.c: Fix use-after-free in multifd_recv_unfill_packet()
Coverity points out (CID 1400442) that in this code:
if (packet->pages_alloc > p->pages->allocated) {
multifd_pages_clear(p->pages);
multifd_pages_init(packet->pages_alloc);
}
we free p->pages in multifd_pages_clear() but continue to
use it in the following code. We also leak memory, because
multifd_pages_init() returns the pointer to a new MultiFDPages_t
struct but we are ignoring its return value.
Fix both of these bugs by adding the missing assignment of
the newly created struct to p->pages.
* remotes/bonzini/tags/for-upstream:
tests: Make check-block a phony target
hw/i386/pc: Fix crash when hot-plugging nvdimm on older machine types
include/qemu/bswap.h: Use __builtin_memcpy() in accessor functions
roms: Allow passing configure options to the EDK2 build tools
roms: Rename the EFIROM variable to avoid clashing with iPXE
Thomas Huth [Sun, 7 Apr 2019 09:23:14 +0000 (11:23 +0200)]
hw/i386/pc: Fix crash when hot-plugging nvdimm on older machine types
QEMU currently crashes when you try to hot-plug an "nvdimm" device
on older machine types:
$ qemu-system-x86_64 -monitor stdio -M pc-1.1
QEMU 3.1.92 monitor - type 'help' for more information
(qemu) device_add nvdimm,id=nvdimmn1
qemu-system-x86_64: /home/thuth/devel/qemu/util/error.c:57: error_setv:
Assertion `*errp == ((void *)0)' failed.
Aborted (core dumped)
The call to hotplug_handler_pre_plug() in pc_memory_pre_plug() has been
added recently before the check whether nvdimm is enabled. It should
be done after the check. And while we're at it, also check the errp
after the hotplug_handler_pre_plug(), otherwise errors are silently
ignored here.
Peter Maydell [Mon, 18 Mar 2019 11:29:38 +0000 (11:29 +0000)]
include/qemu/bswap.h: Use __builtin_memcpy() in accessor functions
In the accessor functions ld*_he_p() and st*_he_p() we use memcpy()
to perform a load or store to a pointer which might not be aligned
for the size of the type. We rely on the compiler to optimize this
memcpy() into an efficient load or store instruction where possible.
This is required for good performance, but at the moment it is also
required for correct operation, because some users of these functions
require that the access is atomic if the pointer is aligned, which
will only be the case if the compiler has optimized out the memcpy().
(The particular example where we discovered this is the virtio
vring_avail_idx() which calls virtio_lduw_phys_cached() which
eventually ends up calling lduw_he_p().)
Unfortunately some compile environments, such as the fortify-source
setup used in Alpine Linux, define memcpy() to a wrapper function
in a way that inhibits this compiler optimization.
The correct long-term fix here is to add a set of functions for
doing atomic accesses into AddressSpaces (and to other relevant
families of accessor functions like the virtio_*_phys_cached()
ones), and make sure that callsites which want atomic behaviour
use the correct functions.
In the meantime, switch to using __builtin_memcpy() in the
bswap.h accessor functions. This will make us robust against things
like this fortify library in the short term. In the longer term
it will mean that we don't end up with these functions being really
badly-performing even if the semantics of the out-of-line memcpy()
are correct.
Peter Maydell [Thu, 28 Mar 2019 10:47:50 +0000 (10:47 +0000)]
target/i386: Generate #UD for LOCK on a register increment
Fix a TCG crash due to attempting an atomic increment
operation without having set up the address first.
This is a similar case to that dealt with in commit e84fcd7f662a0d8198703, and we fix it in the same way.
* remotes/dgibson/tags/ppc-for-4.0-20190409:
spapr_pci: Fix extended config space accesses
pci: Allow PCI bus subtypes to support extended config space accesses
Greg Kurz [Mon, 1 Apr 2019 17:55:02 +0000 (19:55 +0200)]
pci: Allow PCI bus subtypes to support extended config space accesses
Some PHB implementations, eg. PAPR used on pseries machine, act like
a regular PCI bus rather than a PCIe bus, but allow access to the
PCIe extended config space anyway.
Introduce a new PCI bus class method to modelize this behaviour and
use it when adjusting the config space size limit during accesses.
Eric Blake [Thu, 4 Apr 2019 14:52:26 +0000 (09:52 -0500)]
nbd/client: Fix error message for server with unusable sizing
Add a missing space to the error message used when giving up on a
server that insists on an alignment which renders the last few bytes
of the export unreadable.
Eric Blake [Wed, 3 Apr 2019 03:05:22 +0000 (22:05 -0500)]
nbd/server: Don't fail NBD_OPT_INFO for byte-aligned sources
In commit 0c1d50bd, I added a couple of TODO comments about whether we
consult bl.request_alignment when responding to NBD_OPT_INFO. At the
time, qemu as server was hard-coding an advertised alignment of 512 to
clients that promised to obey constraints, and there was no function
for getting at a device's preferred alignment. But in hindsight,
advertising 512 when the block device prefers 1 caused other
compliance problems, and commit b0245d64 changed one of the two TODO
comments to advertise a more accurate alignment. Time to fix the other
TODO. Doesn't really impact qemu as client (our normal client doesn't
use NBD_OPT_INFO, and qemu-nbd --list promises to obey block sizes),
but it might prove useful to other clients.
Eric Blake [Wed, 3 Apr 2019 03:05:21 +0000 (22:05 -0500)]
nbd/server: Trace client noncompliance on unaligned requests
We've recently added traces for clients to flag server non-compliance;
let's do the same for servers to flag client non-compliance. According
to the spec, if the client requests NBD_INFO_BLOCK_SIZE, it is
promising to send all requests aligned to those boundaries. Of
course, if the client does not request NBD_INFO_BLOCK_SIZE, then it
made no promises so we shouldn't flag anything; and because we are
willing to handle clients that made no promises (the spec allows us to
use NBD_REP_ERR_BLOCK_SIZE_REQD if we had been unwilling), we already
have to handle unaligned requests (which the block layer already does
on our behalf). So even though the spec allows us to return EINVAL
for clients that promised to behave, it's easier to always answer
unaligned requests. Still, flagging non-compliance can be useful in
debugging a client that is trying to be maximally portable.
Qemu as client used to have one spot where it sent non-compliant
requests: if the server sends an unaligned reply to
NBD_CMD_BLOCK_STATUS, and the client was iterating over the entire
disk, the next request would start at that unaligned point; this was
fixed in commit a39286dd when the client was taught to work around
server non-compliance; but is equally fixed if the server is patched
to not send unaligned replies in the first place (yes, qemu 4.0 as
server still has few such bugs, although they will be patched in
4.1). Fortunately, I did not find any more spots where qemu as client
was non-compliant. I was able to test the patch by using the following
hack to convince qemu-io to run various unaligned commands, coupled
with serving 512-byte alignment by intentionally omitting '-f raw' on
the server while viewing server traces.
| diff --git i/nbd/client.c w/nbd/client.c
| index 427980bdd22..1858b2aac35 100644
| --- i/nbd/client.c
| +++ w/nbd/client.c
| @@ -449,6 +449,7 @@ static int nbd_opt_info_or_go(QIOChannel *ioc, uint32_t opt,
| nbd_send_opt_abort(ioc);
| return -1;
| }
| + info->min_block = 1;//hack
| if (!is_power_of_2(info->min_block)) {
| error_setg(errp, "server minimum block size %" PRIu32
| " is not a power of two", info->min_block);
Eric Blake [Wed, 3 Apr 2019 03:05:20 +0000 (22:05 -0500)]
nbd/server: Fix blockstatus trace
Don't increment remaining_bytes until we know that we will actually be
including the current block status extent in the reply; otherwise, the
value traced will include a bytes value that is oversized by the
length of the next block status extent which did not get sent because
it instead ended the loop.
drive_new() returns null without setting an error when it provided
help. add_init_drive() assumes null means failure, and crashes trying
to report a null error.
linux-user: rename gettid() to sys_gettid() to avoid clash with glibc
The glibc-2.29.9000-6.fc31.x86_64 package finally includes the gettid()
function as part of unistd.h when __USE_GNU is defined. This clashes
with linux-user code which unconditionally defines this function name
itself.
/home/berrange/src/virt/qemu/linux-user/syscall.c:253:16: error: static declaration of ‘gettid’ follows non-static declaration
253 | _syscall0(int, gettid)
| ^~~~~~
/home/berrange/src/virt/qemu/linux-user/syscall.c:184:13: note: in definition of macro ‘_syscall0’
184 | static type name (void) \
| ^~~~
In file included from /usr/include/unistd.h:1170,
from /home/berrange/src/virt/qemu/include/qemu/osdep.h:107,
from /home/berrange/src/virt/qemu/linux-user/syscall.c:20:
/usr/include/bits/unistd_ext.h:34:16: note: previous declaration of ‘gettid’ was here
34 | extern __pid_t gettid (void) __THROW;
| ^~~~~~
CC aarch64-linux-user/linux-user/signal.o
make[1]: *** [/home/berrange/src/virt/qemu/rules.mak:69: linux-user/syscall.o] Error 1
make[1]: *** Waiting for unfinished jobs....
make: *** [Makefile:449: subdir-aarch64-linux-user] Error 2
While we could make our definition conditional and rely on glibc's impl,
this patch simply renames our definition to sys_gettid() which is a
common pattern in this file.
The gettid syscall was introduced in Linux 2.4.11. This is old enough
that we can assume it always exists and thus not bother with the
conditional backcompat logic.
Kevin Wolf [Thu, 4 Apr 2019 15:04:43 +0000 (17:04 +0200)]
block: Forward 'discard' to temporary overlay
When bdrv_temp_snapshot_options() is called for snapshot=on, the
'discard' option in the options QDict hasn't been parsed and merged into
the flags yet. So copy the dict entry to make sure that the temporary
overlay enables discard when it was requested for the drive.
* remotes/huth-gitlab/tags/pull-request-2019-04-08:
test qgraph.c: Fix segs due to out of scope default
tests/libqos: fix usage of bool in pci-spapr.c
tests/libqos: fix usage of bool in pci-pc.c
test qgraph.c: Fix segs due to out of scope default
The test uses the trick:
if (!opts) {
opts = &(QOSGraph...Options) { };
}
in a couple of places, however the temporary created
by the &() {} goes out of scope at the bottom of the if,
and results in a seg or assert when opts-> fields are
used (on fedora 30's gcc 9).
Jafar Abdi [Sat, 23 Mar 2019 14:26:36 +0000 (17:26 +0300)]
tests/libqos: fix usage of bool in pci-spapr.c
Clean up wrong usage of FALSE and TRUE in places that use "bool" from stdbool.h.
FALSE and TRUE (with capital letters) are the constants defined by glib for
being used with the "gboolean" type of glib. But some parts of the code also use
TRUE and FALSE for variables that are declared as "bool" (the type from <stdbool.h>).
Jafar Abdi [Sat, 23 Mar 2019 14:26:35 +0000 (17:26 +0300)]
tests/libqos: fix usage of bool in pci-pc.c
Clean up wrong usage of FALSE and TRUE in places that use "bool" from stdbool.h.
FALSE and TRUE (with capital letters) are the constants defined by glib for
being used with the "gboolean" type of glib. But some parts of the code also use
TRUE and FALSE for variables that are declared as "bool" (the type from <stdbool.h>).
migration: use bitmap_mutex in migration_bitmap_clear_dirty
My colleague Wei's patch add bitmap_mutex in migration_bitmap_clear_dirty,
but COLO didn't initialize the bitmap_mutex. So we always get an error
when COLO start up. like that:
qemu-system-x86_64: util/qemu-thread-posix.c:64: qemu_mutex_lock_impl: Assertion `mutex->initialized' failed.
This patch add the bitmap_mutex initialize and destroy in COLO
lifecycle.
Peter Maydell [Fri, 5 Apr 2019 03:50:30 +0000 (04:50 +0100)]
Merge remote-tracking branch 'remotes/palmer/tags/riscv-for-master-4.0-rc3-v2' into staging
RISC-V Patches for 4.0-rc3, v2
This patch set contains a pair of tightly coupled PLIC bug fixes:
* We were calculating the PLIC addresses incorrectly.
* We were installing the wrong number of PLIC interrupts.
The two bugs togther resulted in a mostly-working system, but they're
impossible to seperate because fixing one bug would result in
significant breakage. As a result they're in the same patch.
There is also a cleanup to use qemu_log_mask(LOG_GUEST_ERROR,...) for
error reporting.
As far as I know these are the last outstanding RISC-V patches for 4.0.
v2 no longer fails "make check" for me... sorry!
# gpg: Signature made Fri 05 Apr 2019 01:33:57 BST
# gpg: using RSA key 00CE76D1834960DFCE886DF8EF4CA1502CCBAB41
# gpg: issuer "[email protected]"
# gpg: Good signature from "Palmer Dabbelt <[email protected]>" [unknown]
# gpg: aka "Palmer Dabbelt <[email protected]>" [unknown]
# gpg: WARNING: This key is not certified with a trusted signature!
# gpg: There is no indication that the signature belongs to the owner.
# Primary key fingerprint: 00CE 76D1 8349 60DF CE88 6DF8 EF4C A150 2CCB AB41
Peter Maydell [Fri, 5 Apr 2019 02:52:05 +0000 (03:52 +0100)]
Merge remote-tracking branch 'remotes/aperard/tags/pull-xen-20190404' into staging
Xen queue
xen-block fixes
# gpg: Signature made Thu 04 Apr 2019 18:04:38 BST
# gpg: using RSA key F80C006308E22CFD8A92E7980CF5572FD7FB55AF
# gpg: issuer "[email protected]"
# gpg: Good signature from "Anthony PERARD <[email protected]>" [marginal]
# gpg: aka "Anthony PERARD <[email protected]>" [marginal]
# gpg: WARNING: This key is not certified with sufficiently trusted signatures!
# gpg: It is not certain that the signature belongs to the owner.
# Primary key fingerprint: 5379 2F71 024C 600F 778A 7161 D8D5 7199 DF83 42C8
# Subkey fingerprint: F80C 0063 08E2 2CFD 8A92 E798 0CF5 572F D7FB 55AF
* remotes/aperard/tags/pull-xen-20190404:
xen-block: scale sector based quantities correctly
xen-block: only advertize discard to the frontend when it is enabled...
This patch fixes four different things, to maintain bisectability they
have been merged into a single patch. The following fixes are below:
sifive_plic: Fix incorrect irq calculation
The irq is incorrectly calculated to be off by one. It has worked in the
past as the priority_base offset has also been set incorrectly. We are
about to fix the priority_base offset so first first the irq
calculation.
sifive_u: Fix PLIC priority base offset and numbering
According to the FU540 manual the PLIC source priority address starts at
an offset of 0x04 and not 0x00. The same manual also specifies that the
PLIC only has 53 source priorities. Fix these two incorrect header
files.
We also need to over extend the plic_gpios[] array as the PLIC sources
count from 1 and not 0.
riscv: sifive_e: Fix PLIC priority base offset
According to the FE31 manual the PLIC source priority address starts at
an offset of 0x04 and not 0x00.
riscv: virt: Fix PLIC priority base offset
Update the virt offsets based on the newly updated SiFive U and SiFive E
offsets.
Paul Durrant [Mon, 1 Apr 2019 12:17:19 +0000 (13:17 +0100)]
xen-block: scale sector based quantities correctly
The Xen blkif protocol requires that sector based quantities should be
interpreted strictly as multiples of 512 bytes. Specifically:
"first_sect and last_sect in blkif_request_segment, as well as
sector_number in blkif_request, are always expressed in 512-byte units."
Commit fcab2b464e06 "xen: add header and build dataplane/xen-block.c"
incorrectly modified behaviour to use the block device logical_block_size
property as the scale, instead of correctly shifting values by the
hardcoded BDRV_SECTOR_BITS (and hence scaling them to 512 byte units).
This patch undoes that change and restores compliance with the spec.
Furthermore, this patch also restores the original xen_disk behaviour
of advertizing a hardcoded 'sector-size' value of 512 in xenstore and
scaling 'sectors' accordingly. The realize() method is also modified to
fail if logical_block_size is set to anything other than 512.
Paul Durrant [Wed, 20 Mar 2019 14:28:25 +0000 (14:28 +0000)]
xen-block: only advertize discard to the frontend when it is enabled...
...and properly enable it when synthesizing a drive.
The Xen toolstack sets 'discard-enable' to '1' in xenstore when it wants
to enable discard on a specified image. The code in
xen_block_drive_create() correctly parses this and uses it to set
'discard' to 'unmap' for the file_layer, but fails to do the same for the
driver_layer (which effectively disables it). Meanwhile the code in
xen_block_realize() advertizes discard support to the frontend in the
default case (because conf->discard_granularity defaults to -1), even when
the underlying image may not handle it.
This patch adds the missing option to the driver_layer in
xen_block_driver_create() and checks whether BDRV_O_UNMAP is actually
set on the block device before advertizing discard to the frontend.
In the case that discard is supported it also makes sure that the
granularity is set to the physical block size.
* remotes/cohuck/tags/s390x-20190403:
hw/s390x/3270-ccw: avoid taking address of fields in packed struct
hw/s390x/ipl: avoid taking address of fields in packed struct
hw/s390/css: avoid taking address members in packed structs
hw/vfio/ccw: avoid taking address members in packed structs
hw/s390x/3270-ccw: avoid taking address of fields in packed struct
Compiling with GCC 9 complains
hw/s390x/3270-ccw.c: In function ‘emulated_ccw_3270_cb’:
hw/s390x/3270-ccw.c:81:19: error: taking address of packed member of ‘struct SCHIB’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
81 | SCSW *s = &sch->curr_status.scsw;
| ^~~~~~~~~~~~~~~~~~~~~~
This local variable is only present to save a little bit of
typing when setting the field later. Get rid of this to avoid
the warning about unaligned accesses.
hw/s390x/ipl: avoid taking address of fields in packed struct
Compiling with GCC 9 complains
hw/s390x/ipl.c: In function ‘s390_ipl_set_boot_menu’:
hw/s390x/ipl.c:256:25: warning: taking address of packed member of ‘struct QemuIplParameters’ may result in an unaligned pointer value [-Waddress-of-packed-member]
256 | uint32_t *timeout = &ipl->qipl.boot_menu_timeout;
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~
This local variable is only present to save a little bit of
typing when setting the field later. Get rid of this to avoid
the warning about unaligned accesses.
hw/s390/css: avoid taking address members in packed structs
The GCC 9 compiler complains about many places in s390 code
that take the address of members of the 'struct SCHIB' which
is marked packed:
hw/s390x/css.c: In function ‘sch_handle_clear_func’:
hw/s390x/css.c:698:15: warning: taking address of packed member of ‘struct SCHIB’ may result in an unaligned pointer val\
ue [-Waddress-of-packed-member]
698 | PMCW *p = &sch->curr_status.pmcw;
| ^~~~~~~~~~~~~~~~~~~~~~
hw/s390x/css.c:699:15: warning: taking address of packed member of ‘struct SCHIB’ may result in an unaligned pointer val\
ue [-Waddress-of-packed-member]
699 | SCSW *s = &sch->curr_status.scsw;
| ^~~~~~~~~~~~~~~~~~~~~~
...snip many more...
Almost all of these are just done for convenience to avoid
typing out long variable/field names when referencing struct
members. We can get most of this convenience by taking the
address of the 'struct SCHIB' instead, avoiding triggering
the compiler warnings.
In a couple of places we copy via a local variable which is
a technique already applied elsewhere in s390 code for this
problem.
hw/vfio/ccw: avoid taking address members in packed structs
The GCC 9 compiler complains about many places in s390 code
that take the address of members of the 'struct SCHIB' which
is marked packed:
hw/vfio/ccw.c: In function ‘vfio_ccw_io_notifier_handler’:
hw/vfio/ccw.c:133:15: warning: taking address of packed member of ‘struct SCHIB’ may result in an unaligned pointer value \
[-Waddress-of-packed-member]
133 | SCSW *s = &sch->curr_status.scsw;
| ^~~~~~~~~~~~~~~~~~~~~~
hw/vfio/ccw.c:134:15: warning: taking address of packed member of ‘struct SCHIB’ may result in an unaligned pointer value \
[-Waddress-of-packed-member]
134 | PMCW *p = &sch->curr_status.pmcw;
| ^~~~~~~~~~~~~~~~~~~~~~
...snip many more...
Almost all of these are just done for convenience to avoid
typing out long variable/field names when referencing struct
members. We can get most of this convenience by taking the
address of the 'struct SCHIB' instead, avoiding triggering
the compiler warnings.
In a couple of places we copy via a local variable which is
a technique already applied elsewhere in s390 code for this
problem.
Peter Xu [Fri, 29 Mar 2019 06:14:22 +0000 (14:14 +0800)]
intel_iommu: Drop extended root field
VTD_RTADDR_RTT is dropped even by the VT-d spec, so QEMU should
probably do the same thing (after all we never really implemented it).
Since we've had a field for that in the migration stream, to keep
compatibility we need to fill the hole up.
Peter Xu [Fri, 29 Mar 2019 06:14:21 +0000 (14:14 +0800)]
intel_iommu: Fix root_scalable migration breakage
When introducing the initial support for scalable mode we added a
new field into vmstate however we blindly migrate that field without
notice. That'll break migration no matter forward or backward.
The normal way should be that we use something like
VMSTATE_UINT32_TEST() or subsections for the new vmstate field however
for this case of vt-d we can even make it simpler because we've
already migrated all the registers and it'll be fairly simple that we
re-generate root_scalable field from the register values during post
load of the device.
Liam Merwick [Thu, 21 Mar 2019 20:13:49 +0000 (20:13 +0000)]
acpi: verify file entries in bios_linker_loader_add_pointer()
The callers to bios_linker_find_file() assert that the file entry returned
is not NULL, except for those in bios_linker_loader_add_pointer(). Add two
asserts in that case for completeness and to facilitate static code analysis.
Peter Maydell [Tue, 2 Apr 2019 13:52:17 +0000 (14:52 +0100)]
Merge remote-tracking branch 'remotes/berrange/tags/filemon-next-pull-request' into staging
filemon: various fixes / improvements to file monitor for USB MTP
Ensure watch IDs unique within a monitor and avoid integer wraparound
issues when many watches are set & unset over time.
# gpg: Signature made Tue 02 Apr 2019 13:53:40 BST
# gpg: using RSA key BE86EBB415104FDF
# gpg: Good signature from "Daniel P. Berrange <[email protected]>" [full]
# gpg: aka "Daniel P. Berrange <[email protected]>" [full]
# Primary key fingerprint: DAF3 A6FD B26B 6291 2D0E 8E3F BE86 EBB4 1510 4FDF
* remotes/berrange/tags/filemon-next-pull-request:
filemon: fix watch IDs to avoid potential wraparound issues
filemon: ensure watch IDs are unique to QFileMonitor scope
tests: refactor file monitor test to make it more understandable
Peter Maydell [Tue, 2 Apr 2019 13:03:11 +0000 (14:03 +0100)]
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging
Block layer patches:
- file-posix: Ignore unlock failure instead of crashing
- gluster: Limit the transfer size to 512 MiB
- stream: Fix backing chain freezing
- qemu-img: Enable BDRV_REQ_MAY_UNMAP for zero writes in convert
- iotests fixes
# gpg: Signature made Tue 02 Apr 2019 13:47:43 BST
# gpg: using RSA key 7F09B272C88F2FD6
# gpg: Good signature from "Kevin Wolf <[email protected]>" [full]
# Primary key fingerprint: DC3D EB15 9A9A F95D 3D74 56FE 7F09 B272 C88F 2FD6
* remotes/kevin/tags/for-upstream:
tests/qemu-iotests/235: Allow fallback to tcg
block: test block-stream with a base node that is used by block-commit
block: freeze the backing chain earlier in stream_start()
block: continue until base is found in bdrv_freeze_backing_chain() et al
block/file-posix: do not fail on unlock bytes
tests/qemu-iotests: Remove redundant COPYING file
block/gluster: limit the transfer size to 512 MiB
qemu-img: Enable BDRV_REQ_MAY_UNMAP in convert
iotests: Fix test 200 on s390x without virtio-pci
filemon: fix watch IDs to avoid potential wraparound issues
Watch IDs are allocated from incrementing a int counter against
the QFileMonitor object. In very long life QEMU processes with
a huge amount of USB MTP activity creating & deleting directories
it is just about conceivable that the int counter can wrap
around. This would result in incorrect behaviour of the file
monitor watch APIs due to clashing watch IDs.
Instead of trying to detect this situation, this patch changes
the way watch IDs are allocated. It is turned into an int64_t
variable where the high 32 bits are set from the underlying
inotify "int" ID. This gives an ID that is guaranteed unique
for the directory as a whole, and we can rely on the kernel
to enforce this. QFileMonitor then sets the low 32 bits from
a per-directory counter.
The USB MTP device only sets watches on the directory as a
whole, not files within, so there is no risk of guest
triggered wrap around on the low 32 bits.
filemon: ensure watch IDs are unique to QFileMonitor scope
The watch IDs are mistakenly only unique within the scope of the
directory being monitored. This is not useful for clients which are
monitoring multiple directories. They require watch IDs to be unique
globally within the QFileMonitor scope.
When the user specifies a list of accelerators, we pick the first one
that initializes successfully. Recent commit 1a3ec8c1564 broke that.
Reproducer:
$ qemu-system-x86_64 --machine accel=xen:tcg
xencall: error: Could not obtain handle on privileged command interface: No such file or directory
xen be core: xen be core: can't open xen interface
can't open xen interface
qemu-system-x86_64: failed to initialize Xen: Operation not permitted
qemu-system-x86_64: /home/armbru/work/qemu/qom/object.c:436: object_set_accelerator_compat_props: Assertion `!object_compat_props[0]' failed.
Root cause: we register accelerator compat properties even when the
accelerator fails. The failed assertion is
object_set_accelerator_compat_props() telling us off. Fix by calling
it only for the accelerator that succeeded.
migration: Support adding migration blockers earlier
migrate_add_blocker() asserts we have a current_migration object, in
migrate_get_current(). We do only after migration_object_init().
This contributes to the following dependency cycle:
* configure_blockdev() must run before machine_set_property()
so machine properties can refer to block backends
* machine_set_property() before configure_accelerator()
so machine properties like kvm-irqchip get applied
* configure_accelerator() before migration_object_init()
so that Xen's accelerator compat properties get applied.
* migration_object_init() before configure_blockdev()
so configure_blockdev() can add migration blockers
The cycle was closed when recent commit cda4aa9a5a0 "Create block
backends before setting machine properties" added the first
dependency, and satisfied it by violating the last one. Broke block
backends that add migration blockers, as demonstrated by qemu-iotests
055.
To fix it, break the last dependency: make migrate_add_blocker()
usable before migration_object_init().
The previous commit already removed the use of migrate_get_current()
from migrate_add_blocker() itself. Didn't quite do the trick, as
there's another one hiding in migration_is_idle().
The use there isn't actually necessary: when no migration object has
been created yet, migration is surely idle. Make migration_is_idle()
return true then.
Command line option --only-migratable is for disallowing any
configuration that can block migration.
Initially, --only-migratable set global variable @only_migratable.
Commit 3df663e575 "migration: move only_migratable to MigrationState"
replaced it by MigrationState member @only_migratable. That was a
mistake.
First, it doesn't make sense on the design level. MigrationState
captures the state of an individual migration, but --only-migratable
isn't a property of an individual migration, it's a restriction on
QEMU configuration. With fault tolerance, we could have several
migrations at once. --only-migratable would certainly protect all of
them. Storing it in MigrationState feels inappropriate.
Second, it contributes to a dependency cycle that manifests itself as
a bug now.
Putting @only_migratable into MigrationState means its available only
after migration_object_init().
We can't set it before migration_object_init(), so we delay setting it
with a global property (this is fixup commit b605c47b57 "migration:
fix handling for --only-migratable").
We can't get it before migration_object_init(), so anything that uses
it can only run afterwards.
Since migrate_add_blocker() needs to obey --only-migratable, any code
adding migration blockers can run only afterwards. This contributes
to the following dependency cycle:
* configure_blockdev() must run before machine_set_property()
so machine properties can refer to block backends
* machine_set_property() before configure_accelerator()
so machine properties like kvm-irqchip get applied
* configure_accelerator() before migration_object_init()
so that Xen's accelerator compat properties get applied.
* migration_object_init() before configure_blockdev()
so configure_blockdev() can add migration blockers
The cycle was closed when recent commit cda4aa9a5a0 "Create block
backends before setting machine properties" added the first
dependency, and satisfied it by violating the last one. Broke block
backends that add migration blockers.
Moving @only_migratable into MigrationState was a mistake. Revert it.
This doesn't quite break the "migration_object_init() before
configure_blockdev() dependency, since migrate_add_blocker() still has
another dependency on migration_object_init(). To be addressed the
next commit.
Note that the reverted commit made -only-migratable sugar for -global
migration.only-migratable=on below the hood. Documentation has only
ever mentioned -only-migratable. This commit removes the arcane &
undocumented alternative to -only-migratable again. Nobody should be
using it.
Recent commit cda4aa9a5a0 moved block backend creation before machine
property evaluation. This broke block backends registering migration
blockers. Commit e60483f2f84 fixed it by moving migration object
creation before block backend creation. This broke migration with
Xen. Turns out we need to configure the accelerator before we create
the migration object so that Xen's accelerator compat properties get
applied. Revert the flawed commit. This fixes the Xen regression,
but brings back the block backend regression. The next commits will
fix it again.
Thomas Huth [Wed, 27 Mar 2019 09:23:42 +0000 (10:23 +0100)]
tests/qemu-iotests/235: Allow fallback to tcg
iotest 235 currently only works with KVM - this is bad for systems where
it is not available, e.g. CI pipelines. The test also works when using
"tcg" as accelerator, so we can simply add that to the list of accelerators,
too.
Alberto Garcia [Thu, 28 Mar 2019 16:25:11 +0000 (18:25 +0200)]
block: test block-stream with a base node that is used by block-commit
The base node of a block-stream operation indicates the first image
from the backing chain starting from which no data is copied to the
top node.
The block-stream job allows others to use that base image, so a second
block-stream job could be writing to it at the same time. An important
restriction is that the base image must not disappear while the stream
job is ongoing. stream_start() freezes the backing chain from top to
base with that purpose but it does it too late in the code so there is
a race condition there.
This bug was fixed in the previous commit, and this patch contains an
iotest for this scenario.
The problem is that the freezing happens too late in stream_start():
during the bdrv_reopen_set_read_only() call earlier in that function
another job can jump in and remove the base image. If that happens we
have an invalid chain and QEMU crashes.
This patch puts the bdrv_freeze_backing_chain() call at the beginning
of the function.
Alberto Garcia [Thu, 28 Mar 2019 16:25:09 +0000 (18:25 +0200)]
block: continue until base is found in bdrv_freeze_backing_chain() et al
All three functions that handle the BdrvChild.frozen attribute walk
the backing chain from 'bs' to 'base' and stop either when 'base' is
found or at the end of the chain if 'base' is NULL.
However if 'base' is not found then the functions return without
errors as if it was NULL.
This is wrong: if the caller passed an incorrect parameter that means
that there is a bug in the code.
bdrv_replace_child() calls bdrv_check_perm() with error_abort on
loosening permissions. However file-locking operations may fail even
in this case, for example on NFS. And this leads to Qemu crash.
Let's avoid such errors. Note, that we ignore such things anyway on
permission update commit and abort.
Thomas Huth [Tue, 5 Feb 2019 14:49:45 +0000 (15:49 +0100)]
tests/qemu-iotests: Remove redundant COPYING file
The file tests/qemu-iotests/COPYING is the same text as in the
COPYING file in the main directory. So as far as I can see, we don't
need the duplicate here.
Several versions of GlusterFS (3.12? -> 6.0.1) fail when the
transfer size is greater or equal to 1024 MiB, so we are
limiting the transfer size to 512 MiB to avoid this rare issue.
Nir Soffer [Sun, 24 Mar 2019 00:20:12 +0000 (02:20 +0200)]
qemu-img: Enable BDRV_REQ_MAY_UNMAP in convert
With Kevin's "block: Fix slow pre-zeroing in qemu-img convert"[1]
(commit c9fdcf202f, 'qemu-img: Use BDRV_REQ_NO_FALLBACK for
pre-zeroing') we skip the pre zero step called like this:
blk_co_pwrite_zeroes(s->target,
sector_num << BDRV_SECTOR_BITS,
n << BDRV_SECTOR_BITS, 0);
Since we use flags=0, this is translated to NBD_CMD_WRITE_ZEROES with
NBD_CMD_FLAG_NO_HOLE flag, which cause the NBD server to allocated space
instead of punching a hole.
Thomas Huth [Wed, 27 Mar 2019 06:43:49 +0000 (07:43 +0100)]
iotests: Fix test 200 on s390x without virtio-pci
virtio-pci is optional on s390x, e.g. in downstream RHEL builds, it
is disabled. On s390x, virtio-ccw should be used instead. Other tests
like 051 or 240 already use virtio-scsi-ccw instead of virtio-scsi-pci
on s390x, so let's do the same here and always use virtio-scsi-ccw on
s390x.
Currently the default audio timer frequency is 10000Hz instead of
a period of 10000us. Also the audiodev timer-period property gets
converted like a frequency. Only handling of the legacy
QEMU_AUDIO_TIMER_PERIOD environment variable is correct because
it's actually a frequency.
With this patch the property timer-period is really a timer period
and QEMU_AUDIO_TIMER_PERIOD remains a frequency.
mtp delete allows the return status of delete succeeded,
partial_delete or readonly - when none of the objects could be
deleted. Give more meaningful names to return values of the
delete function.
Some initiators recurse over the objects themselves. In that case,
only READ_ONLY can be returned.
Peter Maydell [Tue, 2 Apr 2019 02:46:30 +0000 (03:46 +0100)]
Merge remote-tracking branch 'remotes/ericb/tags/pull-nbd-2019-04-01' into staging
nbd patches for 2019-04-01
- Better behavior of qemu-img map on NBD images
- Fixes for NBD protocol alignment corner cases:
- the server has fewer places where it sends reads or block status
not aligned to its advertised block size
- the client has more cases where it can work around server
non-compliance present in qemu 3.1
- the client now avoids non-compliant requests when interoperating
with nbdkit or other servers not advertising block size
* remotes/ericb/tags/pull-nbd-2019-04-01:
nbd/client: Trace server noncompliance on structured reads
nbd/server: Advertise actual minimum block size
block: Add bdrv_get_request_alignment()
nbd/client: Support qemu-img convert from unaligned size
nbd/client: Reject inaccessible tail of inconsistent server
nbd/client: Report offsets in bdrv_block_status
nbd/client: Lower min_block for block-status, unaligned size
iotests: Add 241 to test NBD on unaligned images
nbd-client: Work around server BLOCK_STATUS misalignment at EOF
qemu-img: Gracefully shutdown when map can't finish
nbd: Permit simple error to NBD_CMD_BLOCK_STATUS
nbd: Don't lose server's error to NBD_CMD_BLOCK_STATUS
nbd: Tolerate some server non-compliance in NBD_CMD_BLOCK_STATUS
qemu-img: Report bdrv_block_status failures
Eric Blake [Sat, 30 Mar 2019 16:53:49 +0000 (11:53 -0500)]
nbd/client: Trace server noncompliance on structured reads
Just as we recently added a trace for a server sending block status
that doesn't match the server's advertised minimum block alignment,
let's do the same for read chunks. But since qemu 3.1 is such a
server (because it advertised 512-byte alignment, but when serving a
file that ends in data but is not sector-aligned, NBD_CMD_READ would
detect a mid-sector change between data and hole at EOF and the
resulting read chunks are unaligned), we don't want to change our
behavior of otherwise tolerating unaligned reads.
Note that even though we fixed the server for 4.0 to advertise an
actual block alignment (which gets rid of the unaligned reads at EOF
for posix files), we can still trigger it via other means:
Arguably, that is a bug in the blkdebug block status function, for
leaking a block status that is not aligned. It may also be possible to
observe issues with a backing layer with smaller alignment than the
active layer, although so far I have been unable to write a reliable
iotest for that scenario.
Eric Blake [Sun, 31 Mar 2019 01:36:36 +0000 (20:36 -0500)]
nbd/server: Advertise actual minimum block size
Both NBD_CMD_BLOCK_STATUS and structured NBD_CMD_READ will split their
reply according to bdrv_block_status() boundaries. If the block device
has a request_alignment smaller than 512, but we advertise a block
alignment of 512 to the client, then this can result in the server
reply violating client expectations by reporting a smaller region of
the export than what the client is permitted to address (although this
is less of an issue for qemu 4.0 clients, given recent client patches
to overlook our non-compliance at EOF). Since it's always better to
be strict in what we send, it is worth advertising the actual minimum
block limit rather than blindly rounding it up to 512.
Note that this patch is not foolproof - it is still possible to
provoke non-compliant server behavior using:
That is arguably a bug in the blkdebug driver (it should never pass
back block status smaller than its alignment, even if it has to make
multiple bdrv_get_status calls and determine the
least-common-denominator status among the group to return). It may
also be possible to observe issues with a backing layer with smaller
alignment than the active layer, although so far I have been unable to
write a reliable iotest for that scenario (but again, an issue like
that could be argued to be a bug in the block layer, or something
where we need a flag to bdrv_block_status() to state whether the
result must be aligned to the current layer's limits or can be
subdivided for accuracy when chasing backing files).
Anyways, as blkdebug is not normally used, and as this patch makes our
server more interoperable with qemu 3.1 clients, it is worth applying
now, even while we still work on a larger patch series for the 4.1
timeframe to have byte-accurate file lengths.
Note that the iotests output changes - for 223 and 233, we can see the
server's better granularity advertisement; and for 241, the three test
cases have the following effects:
- natural alignment: the server's smaller alignment is now advertised,
and the hole reported at EOF is now the right result; we've gotten rid
of the server's non-compliance
- forced server alignment: the server still advertises 512 bytes, but
still sends a mid-sector hole. This is still a server compliance bug,
which needs to be fixed in the block layer in a later patch; output
does not change because the client is already being tolerant of the
non-compliance
- forced client alignment: the server's smaller alignment means that
the client now sees the server's status change mid-sector without any
protocol violations, but the fact that the map shows an unaligned
mid-sector hole is evidence of the block layer problems with aligned
block status, to be fixed in a later patch
Eric Blake [Fri, 29 Mar 2019 04:27:49 +0000 (23:27 -0500)]
block: Add bdrv_get_request_alignment()
The next patch needs access to a device's minimum permitted
alignment, since NBD wants to advertise this to clients. Add
an accessor function, borrowing from blk_get_max_transfer()
for accessing a backend's block limits.
Eric Blake [Fri, 29 Mar 2019 04:27:48 +0000 (23:27 -0500)]
nbd/client: Support qemu-img convert from unaligned size
If an NBD server advertises a size that is not a multiple of a sector,
the block layer rounds up that size, even though we set info.size to
the exact byte value sent by the server. The block layer then proceeds
to let us read or query block status on the hole that it added past
EOF, which the NBD server is unlikely to be happy with. Fortunately,
qemu as a server never advertizes an unaligned size, so we generally
don't run into this problem; but the nbdkit server makes it easy to
test:
$ printf %1000d 1 > f1
$ ~/nbdkit/nbdkit -fv file f1 & pid=$!
$ qemu-img convert -f raw nbd://localhost:10809 f2
$ kill $pid
$ qemu-img compare f1 f2
Pre-patch, the server attempts a 1024-byte read, which nbdkit
rightfully rejects as going beyond its advertised 1000 byte size; the
conversion fails and the output files differ (not even the first
sector is copied, because qemu-img does not follow ddrescue's habit of
trying smaller reads to get as much information as possible in spite
of errors). Post-patch, the client's attempts to read (and query block
status, for new enough nbdkit) are properly truncated to the server's
length, with sane handling of the hole the block layer forced on
us. Although f2 ends up as a larger file (1024 bytes instead of 1000),
qemu-img compare shows the two images to have identical contents for
display to the guest.
I didn't add iotests coverage since I didn't want to add a dependency
on nbdkit in iotests. I also did NOT patch write, trim, or write
zeroes - these commands continue to fail (usually with ENOSPC, but
whatever the server chose), because we really can't write to the end
of the file, and because 'qemu-img convert' is the most common case
where we care about being tolerant (which is read-only). Perhaps we
could truncate the request if the client is writing zeros to the tail,
but that seems like more work, especially if the block layer is fixed
in 4.1 to track byte-accurate sizing (in which case this patch would
be reverted as unnecessary).
Eric Blake [Sat, 30 Mar 2019 15:57:04 +0000 (10:57 -0500)]
nbd/client: Reject inaccessible tail of inconsistent server
The NBD spec suggests that a server should never advertise a size
inconsistent with its minimum block alignment, as that tail is
effectively inaccessible to a compliant client obeying those block
constraints. Since we have a habit of rounding up rather than
truncating, to avoid losing the last few bytes of user input, and we
cannot access the tail when the server advertises bogus block sizing,
abort the connection to alert the server to fix their bug. And
rejecting such servers matches what we already did for a min_block
that was not a power of 2 or which was larger than max_block.
Does not impact either qemu (which always sends properly aligned
sizes) or nbdkit (which does not send minimum block requirements yet);
so this is mostly aimed at new NBD server implementations, and ensures
that the rest of our code can assume the size is aligned.
Peter Maydell [Thu, 28 Mar 2019 13:35:03 +0000 (13:35 +0000)]
hw/usb/bus.c: Handle "no speed matched" case in usb_mask_to_str()
In usb_mask_to_str() we convert a mask of USB speeds into
a human-readable string (like "full+high") for use in
tracing and error messages. However the conversion code
doesn't do anything to the string buffer if the passed in
speedmask doesn't match any of the recognized speeds,
which means that the tracing and error messages will
end up with random garbage in them. This can happen if
we're doing USB device passthrough.
Handle the "unrecognized speed" case by using the
string "unknown".
projects / qemu.git / log