The local backend was recently converted to using "at*()" syscalls in order
to ensure all accesses happen below the shared directory. This requires that
we only pass relative paths, otherwise the dirfd argument to the "at*()"
syscalls is ignored and the path is treated as an absolute path in the host.
This is actually the case for paths in all fids, with the notable exception
of the root fid, whose path is "/". This causes the following backend ops to
act on the "/" directory of the host instead of the virtfs shared directory
when the export root is involved:
- lstat
- chmod
- chown
- utimensat
ie, chmod /9p_mount_point in the guest will be converted to chmod / in the
host for example. This could cause security issues with a privileged QEMU.
All "*at()" syscalls are being passed an open file descriptor. In the case
of the export root, this file descriptor points to the path in the host that
was passed to -fsdev.
The fix is thus as simple as changing the path of the export root fid to be
"." instead of "/".
This is CVE-2017-7471.
Cc: qemu-stable@nongnu.org
Reported-by: Léo Gaspard <leo@gaspard.io>
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 9c6b899f7a)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The last byte of the option vector was missing due to an off-by-one
error. Without this fix, client architecture support negotiation will
fail because the last byte of option vector 5, which contains the MMU
support, will be missed.
Signed-off-by: Sam Bobroff <sam.bobroff@au1.ibm.com>
Reviewed-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
(cherry picked from commit fe93e3e6ec)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Commit ad07cd6 ("virtio-scsi: always use dataplane path if ioeventfd is
active", 2016-10-30) and 9ffe337 ("virtio-blk: always use dataplane
path if ioeventfd is active", 2016-10-30) broke the virtio 1.0
indirect access registers.
The indirect access registers bypass the ioeventfd, so that virtio-blk
and virtio-scsi now repeatedly try to initialize dataplane instead of
triggering the guest->host EventNotifier. Detect the situation by
checking vq->handle_aio_output; if it is not NULL, trigger the
EventNotifier, which is how the device expects to get notifications
and in fact the only thread-safe manner to deliver them.
Fixes: ad07cd6
Fixes: 9ffe337
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit e49a661840)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The switch from pointers to addresses (commit
026aeffcb4 and
ffaf857778) added
a off-by-one bug to 16bit backward blits. Fix.
Reported-by: 李强 <liqiang6-s@360.cn>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Li Qiang <liqiang6-s@360.cn>
Message-id: 1489735296-19047-1-git-send-email-kraxel@redhat.com
(cherry picked from commit f019722cbb)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Does basically the same as "cirrus: stop passing around dst pointers in
the blitter", just for the src pointer instead of the dst pointer.
For the src we have to care about cputovideo blits though and fetch the
data from s->cirrus_bltbuf instead of vga memory. The cirrus_src*()
helper functions handle that.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 1489584487-3489-1-git-send-email-kraxel@redhat.com
(cherry picked from commit ffaf857778)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Instead pass around the address (aka offset into vga memory). Calculate
the pointer in the rop_* functions, after applying the mask to the
address, to make sure the address stays within the valid range.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 1489574872-8679-1-git-send-email-kraxel@redhat.com
(cherry picked from commit 026aeffcb4)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
off_cur_end is exclusive, so off_cur_end == cirrus_addr_mask is valid.
Fix calculation to make sure to allow that, otherwise the assert added
by commit f153b563f8 can trigger for valid
blits.
Test case: boot windows nt 4.0
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 1489579606-26020-1-git-send-email-kraxel@redhat.com
(cherry picked from commit e048dac616)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
There is a special code path (dpy_gfx_copy) to allow graphic emulation
notify user interface code about bitblit operations carryed out by
guests. It is supported by cirrus and vnc server. The intended purpose
is to optimize display scrolls and just send over the scroll op instead
of a full display update.
This is rarely used these days though because modern guests simply don't
use the cirrus blitter any more. Any linux guest using the cirrus drm
driver doesn't. Any windows guest newer than winxp doesn't ship with a
cirrus driver any more and thus uses the cirrus as simple framebuffer.
So this code tends to bitrot and bugs can go unnoticed for a long time.
See for example commit "3e10c3e vnc: fix qemu crash because of SIGSEGV"
which fixes a bug lingering in the code for almost a year, added by
commit "c7628bf vnc: only alloc server surface with clients connected".
Also the vnc server will throttle the frame rate in case it figures the
network can't keep up (send buffers are full). This doesn't work with
dpy_gfx_copy, for any copy operation sent to the vnc client we have to
send all outstanding updates beforehand, otherwise the vnc client might
run the client side blit on outdated data and thereby corrupt the
display. So this dpy_gfx_copy "optimization" might even make things
worse on slow network links.
Lets kill it once for all.
Oh, and one more reason: Turns out (after writing the patch) we have a
security bug in that code path ...
Fixes: CVE-2016-9603
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 1489494419-14340-1-git-send-email-kraxel@redhat.com
(cherry picked from commit 50628d3479)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
check the validity of parameters in cirrus_bitblt_rop_fwd_transp_xxx
and cirrus_bitblt_rop_fwd_xxx to avoid the OOB read which causes qemu Segmentation fault.
After the fix, we will touch the assert in
cirrus_invalidate_region:
assert(off_cur_end >= off_cur);
Signed-off-by: fangying <fangying1@huawei.com>
Signed-off-by: hangaohuai <hangaohuai@huawei.com>
Message-id: 20170314063919.16200-1-hangaohuai@huawei.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(cherry picked from commit 215902d7b6)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This reverts commit 5858dd1801.
Conflicts:
hw/display/cirrus_vga.c
Cc: Wolfgang Bumiller <w.bumiller@proxmox.com>
Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Message-id: 1486645341-5010-2-git-send-email-kraxel@redhat.com
(cherry picked from commit 12e97ec399)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The blit_region_is_unsafe checks don't work correctly for the
patterncopy source. It's a fixed-sized region, which doesn't
depend on cirrus_blt_{width,height}. So go do the check in
cirrus_bitblt_common_patterncopy instead, then tell blit_is_unsafe that
it doesn't need to verify the source. Also handle the case where we
blit from cirrus_bitbuf correctly.
This patch replaces 5858dd1801.
Security impact: I think for the most part error on the safe side this
time, refusing blits which should have been allowed.
Only exception is placing the blit source at the end of the video ram,
so cirrus_blt_srcaddr + 256 goes beyond the end of video memory. But
even in that case I'm not fully sure this actually allows read access to
host memory. To trick the commit 5858dd18 security checks one has to
pick very small cirrus_blt_{width,height} values, which in turn implies
only a fraction of the blit source will actually be used.
Cc: Wolfgang Bumiller <w.bumiller@proxmox.com>
Cc: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Reviewed-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Message-id: 1486645341-5010-1-git-send-email-kraxel@redhat.com
(cherry picked from commit 95280c31cd)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Apply the cirrus_addr_mask to cirrus_blt_dstaddr and cirrus_blt_srcaddr
right after assigning them, in cirrus_bitblt_start(), instead of having
this all over the place in the cirrus code, and missing a few places.
Reported-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 1485338996-17095-1-git-send-email-kraxel@redhat.com
(cherry picked from commit 60cd23e851)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The rops used by cirrus_bitblt_common_patterncopy only use
the destination pitch, so the source pitch shoul allowed to
be zero and the blit with used for the range check around the
source address.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Message-id: 1485272138-23249-1-git-send-email-w.bumiller@proxmox.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(cherry picked from commit 5858dd1801)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
cirrus_invalidate_region() calls memory_region_set_dirty()
on a per-line basis, always ranging from off_begin to
off_begin+bytesperline. With a negative pitch off_begin
marks the top most used address and thus we need to do an
initial shift backwards by a line for negative pitches of
backward blits, otherwise the first iteration covers the
line going from the start offset forwards instead of
backwards.
Additionally since the start address is inclusive, if we
shift by a full `bytesperline` we move to the first address
*not* included in the blit, so we only shift by one less
than bytesperline.
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Message-id: 1485352137-29367-1-git-send-email-w.bumiller@proxmox.com
[ kraxel: codestyle fixes ]
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(cherry picked from commit f153b563f8)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
According to the 9P spec [*], when a client wants to cancel a pending I/O
request identified by a given tag (uint16), it must send a Tflush message
and wait for the server to respond with a Rflush message before reusing this
tag for another I/O. The server may still send a completion message for the
I/O if it wasn't actually cancelled but the Rflush message must arrive after
that.
QEMU hence waits for the flushed PDU to complete before sending the Rflush
message back to the client.
If a client sends 'Tflush tag oldtag' and tag == oldtag, QEMU will then
allocate a PDU identified by tag, find it in the PDU list and wait for
this same PDU to complete... i.e. wait for a completion that will never
happen. This causes a tag and ring slot leak in the guest, and a PDU
leak in QEMU, all of them limited by the maximal number of PDUs (128).
But, worse, this causes QEMU to hang on device reset since v9fs_reset()
wants to drain all pending I/O.
This insane behavior is likely to denote a bug in the client, and it would
deserve an Rerror message to be sent back. Unfortunately, the protocol
allows it and requires all flush requests to suceed (only a Tflush response
is expected).
The only option is to detect when we have to handle a self-referencing
flush request and report success to the client right away.
[*] http://man.cat-v.org/plan_9/5/flush
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Greg Kurz <groug@kaod.org>
(cherry picked from commit d5f2af7b95)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The server can handle MAX_REQ - 1 PDUs at a time and the virtio-9p
device has a MAX_REQ sized virtqueue. If the client manages to fill
up the virtqueue, pdu_alloc() will fail and the request won't be
processed without any notice to the client (it actually causes the
linux 9p client to hang).
This has been there since the beginning (commit 9f10751365 "virtio-9p:
Add a virtio 9p device to qemu"), but it needs an agressive workload to
run in the guest to show up.
We actually allocate MAX_REQ PDUs and I see no reason not to link them
all into the free list, so let's fix the init loop.
Reported-by: Tuomas Tynkkynen <tuomas@tuxera.com>
Suggested-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Greg Kurz <groug@kaod.org>
(cherry picked from commit 0d78289c3d)
Conflicts:
hw/9pfs/9p.c
* drop context dep on 583f21f8
Signed-off-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The subchannel is a means to access a device. While the device number is
assigned by the administrator, the subchannel number is assigned by
the channel subsystem in an ascending order on cold and hot plug.
When doing unplug and replug operations, the same device may end up on
a different subchannel; for example
- We start with a device fe.1.2222, which ends up at subchannel
fe.1.0000.
- Now we detach the device, attach a device fe.1.3333 (which would get
the now-free subchannel fe.1.0000), re-attach fe.1.2222 (which ends
up at subchannel fe.1.0001) and detach fe.1.3333.
- We now have the same device (fe.1.2222) available to the guest; it
just shows up on a different subchannel.
In such a case, the subchannel numbers are different from what a
QEMU would create during cold plug when parsing the command line.
As this would cause a guest visible change on migration, we do restore
the source system's value of the subchannel number on load.
So we are now fine from the guest perspective. From the host
perspective this will cause an inconsistent state in our internal data
structures, though.
For example, the subchannel 0 might not be at array position 0. This will
lead to problems when we continue doing hot (un/re) plug operations.
Let's fix this by cleaning up our internal data structures.
Reported-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Dong Jia Shi <bjsdjshi@linux.vnet.ibm.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
(cherry picked from commit 3c788ebc6f)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
We don't reset proxy->vqs[].{num|desc[]|avail[]|used[]}. This means if
a driver enable the vq without setting vq address after reset. The old
addresses were leaked. Fixing this by resetting modern vq meta data
during device reset.
Cc: qemu-stable@nongnu.org
Signed-off-by: Jason Wang <jasowang@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit 60a8d80234)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
When fetching request, it should read sizeof(*hdr), not the
pointer hdr.
Signed-off-by: Li Qiang <liqiang6-s@360.cn>
Message-Id: <1489488980-130668-1-git-send-email-liqiang6-s@360.cn>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit b01a2d07c9)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
MSI-X has been disabled by the time the e1000e device is unrealized, hence
msix_uninit is never called. This causes the object to be leaked, which
shows up as a RAMBlock with empty name when attempting migration.
Reported-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Tested-by: Laurent Vivier <lvivier@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit 7ec7ae4b97)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
In case of VLAN stripping ETH header is stored in a
separate chunk and length of IOV should take this into
account.
This patch fixes checksum validation for RX packets
with VLAN header.
Devices affected by this problem: e1000e and vmxnet3.
Cc: qemu-stable@nongnu.org
Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit c5d083c561)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
In case of VLAN stripping, ETH header put into a
separate buffer, therefore amont of data copied
from original IOV should be smaller.
Cc: qemu-stable@nongnu.org
Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit d5e772146d)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This patch fixed a problem that was introduced in commit eb700029.
When net_rx_pkt_attach_iovec() calls eth_strip_vlan()
this can result in pkt->ehdr_buf being overflowed, because
ehdr_buf is only sizeof(struct eth_header) bytes large
but eth_strip_vlan() can write
sizeof(struct eth_header) + sizeof(struct vlan_header)
bytes into it.
Devices affected by this problem: vmxnet3.
Cc: qemu-stable@nongnu.org
Reported-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Dmitry Fleytman <dmitry@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
(cherry picked from commit df8bf7a7fe)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
CIRRUS_BLTMODE_MEMSYSSRC blits do NOT check blit destination
and blit width, at all. Oops. Fix it.
Security impact: high.
The missing blit destination check allows to write to host memory.
Basically same as CVE-2014-8106 for the other blit variants.
Cc: qemu-stable@nongnu.org
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(cherry picked from commit 92f2b88cea)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This patch adds call to apic_reset_irq_delivered when the virtual
machine is reset.
Signed-off-by: Pavel Dovgalyuk <pavel.dovgaluk@ispras.ru>
Message-Id: <20170131114054.276.62201.stgit@PASHA-ISP>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit f65e821262)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
While doing multi block SDMA transfer in routine
'sdhci_sdma_transfer_multi_blocks', the 's->fifo_buffer' starting
index 'begin' and data length 's->data_count' could end up to be same.
This could lead to an OOB access issue. Correct transfer data length
to avoid it.
Cc: qemu-stable@nongnu.org
Reported-by: Jiang Xin <jiangxin1@huawei.com>
Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Message-id: 20170130064736.9236-1-ppandit@redhat.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
(cherry picked from commit 42922105be)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Right now we reset all devices before we reset the cmma states. This
can result in the host kernel discarding guest pages that were
previously in the unused state but already contain a bios or a -kernel
file before the cmma reset has finished. This race results in random
guest crashes or hangs during very early reboot.
Fixes: 1cd4e0f6f0 ("s390x/cmma: clean up cmma reset")
Cc: qemu-stable@nongnu.org
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
(cherry picked from commit 1a0e4c8b02)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The AHCI emulation code supports 64-bit addressing and should advertise this
fact in the Host Capabilities register. Both Linux and Windows drivers test
this bit to decide if the upper 32 bits of various registers may be written
to, and at least some versions of Windows have a bug where DMA is attempted
with an address above 4GB but, in the absence of HOST_CAP_64, the upper 32
bits are left unititialized which leads to a memory corruption.
[Maintainer edit:
This fixes https://bugzilla.redhat.com/show_bug.cgi?id=1411105,
which affects Windows Server 2008 SP2 in some cases.]
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
Message-id: 1484305370-6220-1-git-send-email-lprosek@redhat.com
[Amended commit message --js]
Signed-off-by: John Snow <jsnow@redhat.com>
(cherry picked from commit 98cb5dccb1)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
When doing bitblt copy in backward mode, we should minus the
blt width first just like the adding in the forward mode. This
can avoid the oob access of the front of vga's vram.
Signed-off-by: Li Qiang <liqiang6-s@360.cn>
{ kraxel: with backward blits (negative pitch) addr is the topmost
address, so check it as-is against vram size ]
Cc: qemu-stable@nongnu.org
Cc: P J P <ppandit@redhat.com>
Cc: Laszlo Ersek <lersek@redhat.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Wolfgang Bumiller <w.bumiller@proxmox.com>
Fixes: d3532a0db0 (CVE-2014-8106)
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Message-id: 1485938101-26602-1-git-send-email-kraxel@redhat.com
Reviewed-by: Laszlo Ersek <lersek@redhat.com>
(cherry picked from commit 62d4c6bd52)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Coverity reports that ARRAY_SIZE(elem->out_sg) (and all the others too)
is wrong because elem->out_sg is a pointer.
However, the check is not in the right place and the max_size argument
of virtqueue_map_iovec can be removed. The check on in_num/out_num
should be moved to qemu_get_virtqueue_element instead, before the call
to virtqueue_alloc_element.
Cc: qemu-stable@nongnu.org
Reported-by: Paolo Bonzini <pbonzini@redhat.com>
Fixes: 3724650db0 ("virtio: introduce virtqueue_alloc_element")
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
(cherry picked from commit 6bdc21c050)
* dropped context dep on 8607f5c30
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Split irqchip works based on the fact that we kept the first 24 gsi
routing entries inside KVM for userspace ioapic's use. When system
boot, we'll reserve these MSI routing entries before hand. However,
after migration, we forgot to re-configure it up in the destination
side. The result is, we'll get invalid gsi routing entries after
migration (all empty), and we get interrupts with vector=0, then
strange things happen, like keyboard hang.
The solution is simple - we update them after migration, which is a
one line fix.
Signed-off-by: Peter Xu <peterx@redhat.com>
Message-Id: <1483952153-7221-4-git-send-email-peterx@redhat.com>
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 0f254b1ae0)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Commit 4299b90 added a check which is too broad, given that the source
pitch value is not required to be initialized for solid fill operations.
This patch refines the blit_is_unsafe() check to ignore source pitch in
that case. After applying the above commit as a security patch, we
noticed the SLES 11 SP4 guest gui failed to initialize properly.
Signed-off-by: Bruce Rogers <brogers@suse.com>
Message-id: 20170109203520.5619-1-brogers@suse.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
(cherry picked from commit 913a87885f)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Because the 'size_t' type is 4 bytes in 32-bit platform, which
is the same with 'int'. It's easy to make 'max_len' to zero when
integer overflow and then cause heap overflow if 'max_len' is zero.
Using uint_64 instead of size_t to avoid the integer overflow.
Cc: qemu-stable@nongnu.org
Reported-by: Li Qiang <liqiang6-s@360.cn>
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Tested-by: Li Qiang <liqiang6-s@360.cn>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit a08aaff811)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The direction is wrong; scsi_block_is_passthrough returns
false for commands that *can* use sglists.
Reported-by: Zhang Qian <zhangqian@sangfor.com.cn>
Fixes: 8fdc7839e4
Cc: qemu-stable@nongnu.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 1f8af0d186)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
'hotplugged' propperty is meant to be used on migration side when migrating
source with hotplugged devices.
However though it not exacly correct usage of 'hotplugged' property
it's possible to set generic hotplugged property for CPU using
-cpu foo,hotplugged=on
or
-global foo.hotplugged=on
in this case qemu crashes with following backtrace:
...
because pc_cpu_plug() assumes that hotplugged CPU could appear only after
rtc/fw_cfg are initialized.
Fix crash by replacing assumption with explicit checks of rtc/fw_cfg
and updating them only if they were initialized.
Cc: qemu-stable@nongnu.org
Reported-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Message-Id: <1483108391-199542-1-git-send-email-imammedo@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
(cherry picked from commit 26ef65beab)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
If the user passes -device virtio-9p without the corresponding -fsdev, QEMU
dereferences a NULL pointer and crashes.
This is a 2.8 regression introduced by commit 702dbcc274.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Li Qiang <liq3ea@gmail.com>
(cherry picked from commit f2b58c4375)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Correct recalculation of vq->inuse after migration for the corner case
where the avail_idx has already wrapped but used_idx not yet.
Also change the type of the VirtQueue.inuse to unsigned int. This is
done to be consistent with other members representing sizes (VRing.num),
and because C99 guarantees max ring size < UINT_MAX but does not
guarantee max ring size < INT_MAX.
Signed-off-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Fixes: bccdef6b ("virtio: recalculate vq->inuse after migration")
CC: qemu-stable@nongnu.org
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit e66bcc4081)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
PCI Express downstream slot has a single PCI slot
behind it, using PCI_DEVFN(PCI_SLOT(devfn), 0)
does not give you function 0 in cases such as ARI
as well as some error cases.
This is exactly what we are hitting:
$ qemu-system-x86_64 -machine q35 -readconfig docs/q35-chipset.cfg
-monitor stdio
(qemu) device_add e1000e,bus=ich9-pcie-port-4,addr=00
(qemu) device_add e1000e,bus=ich9-pcie-port-4,addr=08
Segmentation fault (core dumped)
The fix is to use the pci_get_function_0 API.
Cc: qemu-stable@nongnu.org
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Reported-by: Eduardo Habkost <ehabkost@redhat.com>
Tested-by: Cao jin <caoj.fnst@cn.fujitsu.com>
Tested-by: Eduardo Habkost <ehabkost@redhat.com>
Reviewed-by: Eduardo Habkost <ehabkost@redhat.com>
(cherry picked from commit d93ddfb1f8)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
A broken guest can specify physical addresses that correspond
to any memory region, but it shouldn't be able to change ROM.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Cc: qemu-stable@nongnu.org
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
(cherry picked from commit f2fd57db36)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Original problem description by Greg Kurz:
> Since commit "9a4c0e220d8a hw/virtio-pci: fix virtio
> behaviour", passing -device virtio-blk-pci.disable-modern=off
> has no effect on 2.6 machine types because the internal
> virtio-pci.disable-modern=on compat property always prevail.
The same bug also affects other abstract type names mentioned on
compat_props by machine-types: apic-common, i386-cpu, pci-device,
powerpc64-cpu, s390-skeys, spapr-pci-host-bridge, usb-device,
virtio-pci, x86_64-cpu.
The right fix for this problem is to make sure compat_props and
-global options are always applied in the order they are
registered, instead of reordering them based on the type
hierarchy. But changing the ordering rules of -global is risky
and might break existing configurations, so we shouldn't do that
on a stable branch.
This is a temporary hack that will work around the bug when
registering compat_props properties: if we find an abstract class
on compat_props, register properties for all its non-abstract
subtypes instead. This will make sure -global won't be overridden
by compat_props, while keeping the existing ordering rules on
-global options.
Note that there's one case that won't be fixed by this hack:
"-global spapr-pci-vfio-host-bridge.<option>=<value>" won't be
able to override compat_props, because spapr-pci-host-bridge is
not an abstract class.
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
Message-Id: <1481575745-26120-1-git-send-email-ehabkost@redhat.com>
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Reviewed-by: Halil Pasic <pasic@linux.vnet.ibm.com>
Reviewed-by: Greg Kurz <groug@kaod.org>
Tested-by: Greg Kurz <groug@kaod.org>
Signed-off-by: Eduardo Habkost <ehabkost@redhat.com>
(cherry picked from commit 0bcba41fe3)
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
We should pass O_NOFOLLOW otherwise openat() will follow symlinks and make
QEMU vulnerable.
While here, we also fix local_unlinkat_common() to use openat_dir() for
the same reasons (it was a leftover in the original patchset actually).
This fixes CVE-2016-9602.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
(cherry picked from commit b003fc0d8a)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
When O_PATH is used with O_DIRECTORY, it only acts as an optimization: the
openat() syscall simply finds the name in the VFS, and doesn't trigger the
underlying filesystem.
On systems that don't define O_PATH, because they have glibc version 2.13
or older for example, we can safely omit it. We don't want to deactivate
O_PATH globally though, in case it is used without O_DIRECTORY. The is done
with a dedicated macro.
Systems without O_PATH may thus fail to resolve names that involve
unreadable directories, compared to newer systems succeeding, but such
corner case failure is our only option on those older systems to avoid
the security hole of chasing symlinks inappropriately.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
(added last paragraph to changelog as suggested by Eric Blake)
Signed-off-by: Greg Kurz <groug@kaod.org>
(cherry picked from commit 918112c02a)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The name argument can never be an empty string, and dirfd always point to
the containing directory of the file name. AT_EMPTY_PATH is hence useless
here. Also it breaks build with glibc version 2.13 and older.
It is actually an oversight of a previous tentative patch to implement this
function. We can safely drop it.
Reported-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Signed-off-by: Greg Kurz <groug@kaod.org>
Tested-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
Reviewed-by: Eric Blake <eblake@redhat.com>
(cherry picked from commit b314f6a077)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
If we cannot open the given path, we can return right away instead of
passing -1 to fstatfs() and close(). This will make Coverity happy.
(Coverity issue CID1371729)
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Daniel P. berrange <berrange@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
(cherry picked from commit 23da0145cc)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Coverity issue CID1371731
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Daniel P. Berrange <berrange@redhat.com>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
(cherry picked from commit faab207f11)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
This was spotted by Coverity as a fd leak. This is certainly true, but also
local_remove() would always return without doing anything, unless the fd is
zero, which is very unlikely.
(Coverity issue CID1371732)
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Eric Blake <eblake@redhat.com>
(cherry picked from commit b7361d46e7)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Now that the all callbacks have been converted to use "at" syscalls, we
can drop this code.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit c23d5f1d5b)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The local_open2() callback is vulnerable to symlink attacks because it
calls:
(1) open() which follows symbolic links for all path elements but the
rightmost one
(2) local_set_xattr()->setxattr() which follows symbolic links for all
path elements
(3) local_set_mapped_file_attr() which calls in turn local_fopen() and
mkdir(), both functions following symbolic links for all path
elements but the rightmost one
(4) local_post_create_passthrough() which calls in turn lchown() and
chmod(), both functions also following symbolic links
This patch converts local_open2() to rely on opendir_nofollow() and
mkdirat() to fix (1), as well as local_set_xattrat(),
local_set_mapped_file_attrat() and local_set_cred_passthrough() to
fix (2), (3) and (4) respectively. Since local_open2() already opens
a descriptor to the target file, local_set_cred_passthrough() is
modified to reuse it instead of opening a new one.
The mapped and mapped-file security modes are supposed to be identical,
except for the place where credentials and file modes are stored. While
here, we also make that explicit by sharing the call to openat().
This partly fixes CVE-2016-9602.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit a565fea565)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The local_mkdir() callback is vulnerable to symlink attacks because it
calls:
(1) mkdir() which follows symbolic links for all path elements but the
rightmost one
(2) local_set_xattr()->setxattr() which follows symbolic links for all
path elements
(3) local_set_mapped_file_attr() which calls in turn local_fopen() and
mkdir(), both functions following symbolic links for all path
elements but the rightmost one
(4) local_post_create_passthrough() which calls in turn lchown() and
chmod(), both functions also following symbolic links
This patch converts local_mkdir() to rely on opendir_nofollow() and
mkdirat() to fix (1), as well as local_set_xattrat(),
local_set_mapped_file_attrat() and local_set_cred_passthrough() to
fix (2), (3) and (4) respectively.
The mapped and mapped-file security modes are supposed to be identical,
except for the place where credentials and file modes are stored. While
here, we also make that explicit by sharing the call to mkdirat().
This partly fixes CVE-2016-9602.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit 3f3a16990b)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
The local_mknod() callback is vulnerable to symlink attacks because it
calls:
(1) mknod() which follows symbolic links for all path elements but the
rightmost one
(2) local_set_xattr()->setxattr() which follows symbolic links for all
path elements
(3) local_set_mapped_file_attr() which calls in turn local_fopen() and
mkdir(), both functions following symbolic links for all path
elements but the rightmost one
(4) local_post_create_passthrough() which calls in turn lchown() and
chmod(), both functions also following symbolic links
This patch converts local_mknod() to rely on opendir_nofollow() and
mknodat() to fix (1), as well as local_set_xattrat() and
local_set_mapped_file_attrat() to fix (2) and (3) respectively.
A new local_set_cred_passthrough() helper based on fchownat() and
fchmodat_nofollow() is introduced as a replacement to
local_post_create_passthrough() to fix (4).
The mapped and mapped-file security modes are supposed to be identical,
except for the place where credentials and file modes are stored. While
here, we also make that explicit by sharing the call to mknodat().
This partly fixes CVE-2016-9602.
Signed-off-by: Greg Kurz <groug@kaod.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
(cherry picked from commit d815e72190)
Signed-off-by: Greg Kurz <gkurz@linux.vnet.ibm.com>
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
Greg Kurz