In addition to handling IO in vcpu thread and in io thread, blk dataplane
introduces yet another mode: handling it by AioContext.
Currently, this reuses the same handler as previous modes,
which triggers races as these were not designed to be reentrant.
Add instead a separate handler just for aio; this will make
it possible to disable regular handlers when dataplane is active.
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
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>
Add two missing checks for s->dataplane_fenced. In one case, QEMU
would skip injecting an IRQ due to a write to an uninitialized
EventNotifier's file descriptor.
In the second case, the dataplane_disabled field was used by mistake;
in fact after fixing this occurrence it is completely unused.
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
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>
We must not call virtio_blk_data_plane_notify if dataplane is
disabled: we would hit a segmentation fault in notify_guest_bh as
s->guest_notifier has not been setup and is NULL.
Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>
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>
Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>
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>
If before loading snapshot we had set the timer of statistics, then after
applying snapshot the expiry time would be irrelevant for the restored
state of the virtual clocks. A simple fix is just to restart the timer
after loading snapshot.
For the user it may look like a long delay of statistics update after switch
to the snapshot.
Signed-off-by: Pavel Butsykin <pbutsykin@virtuozzo.com>
Reviewed-by: Roman Kagan <rkagan@virtuozzo.com>
Signed-off-by: Denis V. Lunev <den@openvz.org>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
The i82801b11 bridge didn't have a vmsd and thus didn't send
any migration data, including that of its parent PCIBridge object.
The symptom being if the guest used any devices behind the bridge
the guest crashed (mostly with various interrupt related issues).
Note: This will cause migration from old qemus that used this device to
explicitly fail during migration as opposed to the guest crashing.
Signed-off-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Suggested-by: Marcel Apfelbaum <marcel@redhat.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Entries are inserted in filename order instead of being
appended to the end in case sorting is enabled.
This will avoid any future issues of moving the file creation
around, it doesn't matter what order they are created now,
the will always be in filename order.
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Added machine type handling for compatibility. This was
a fairly complex change, this will preserve the order of fw_cfg
for older versions no matter what order the firmware files
actually come in. A list is kept of the correct legacy order
and the entries will be inserted based upon their order in
the list. Except that some entries are ordered (in a specific
area of the list) based upon what order they appear on the
command line. Special handling is added for those entries.
Signed-off-by: Corey Minyard <cminyard@mvista.com>
Reviewed-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
piix3_ide_xen_class_init is identical to piix3_ide_class_init
except it's buggy as it does not set exit and does not disable
hotplug properly.
Switch to the generic one.
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Teach PCI testdev to use fast MMIO when kvm makes it available.
Before:
mmio-wildcard-eventfd:pci-mem 2271
After:
mmio-wildcard-eventfd:pci-mem 1218
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Currently spapr doesn't support "aborting" hotplug of PCI
devices by allowing device_del to immediately remove the
device if we haven't signalled the presence of the device
to the guest.
In the past this wasn't an issue, since we always immediately
signalled device attach and simply relied on full guest-aware
add->remove path for device removal. However, as of 788d259,
we now defer signalling for PCI functions until function 0
is attached, so now we need to deal with these "abort" operations
for cases where a user hotplugs a non-0 function, then opts to
remove it prior hotplugging function 0. Currently they'd have to
reboot before the unplug completed. PCIe multifunction hotplug
does not have this requirement however, so from a management
implementation perspective it would be good to address this within
the same release as 788d259.
We accomplish this by simply adding a 'signalled' flag to track
whether a device hotplug event has been sent to the guest. If it
hasn't, we allow immediate removal under the assumption that the
guest will not be using the device. Devices present at boot/reset
time are also assumed to be 'signalled'.
For CPU/memory/etc, signalling will still happen immediately
as part of device_add, so only PCI functions should be affected.
Cc: bharata@linux.vnet.ibm.com
Cc: david@gibson.dropbear.id.au
Cc: sbhat@linux.vnet.ibm.com
Cc: qemu-ppc@nongnu.org
Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com>
[dwg: This fixes a regression where an incorrect hot-add of a non-zero
function can no longer be backed out until function 0 is added]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This patch fixes the current AIL implementation for POWER8. The
interrupt vector address can be calculated directly from LPCR when the
exception is handled. The excp_prefix update becomes useless and we
can cleanup the H_SET_MODE hcall.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
[clg: Removed LPES0/1 handling for HV vs. !HV
Fixed LPCR_ILE case for POWERPC_EXCP_POWER8 ]
Signed-off-by: Cédric Le Goater <clg@fr.ibm.com>
[dwg: This was written as a cleanup, but it also fixes a real bug
where setting an alternative interrupt location would not be
correctly migrated]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Wire up the CPU timer interrupts in the right order, with the
nonsecure physical timer on cntpnsirq, the hyp timer on cnthpirq,
and the secure physical timer on cntpsirq. (We did get the
virt timer right, at least.)
Reported-by: Antonio Huete Jiménez <tuxillo@quantumachine.net>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Reviewed-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
Message-id: 1458210790-6621-1-git-send-email-peter.maydell@linaro.org
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-12-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-11-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Implements FSR register, it is used for busy waits.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-10-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Adds fast read and 4bytes commands family.
This work is based on Pawel Lenkow patch from v1.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-9-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Use the setting from the volatile cfg register to correctly
set the number of dummy cycles.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-8-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This patch adds both volatile and non volatile configuration registers
and commands to allow modify them. It is needed for proper handling
dummy cycles. Initialization of those registers and flash state
has been included as well.
Some of this registers are used by kernel.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Acked-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-7-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This patch adds only 4byte address mode (does not cover dummy cycles).
This mode is needed to access more than 16 MiB of flash.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-6-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Extend address mode allows to switch flash 16 MiB banks,
allowing user to access all flash sectors.
This access mode is used by u-boot.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-5-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Extend the width of the flags variable to support the already existing
(but unused) WR_1 flag, which is above the range of 8 bits.
This allows support of EEPROM emulation which requires the WR_1 feature.
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-4-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-3-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Marcin Krzeminski <marcin.krzeminski@nokia.com>
Reviewed-by: Peter Crosthwaite <crosthwaite.peter@gmail.com>
Message-id: 1458719789-29868-2-git-send-email-marcin.krzeminski@nokia.com
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
There is a problem for power button that it will not work if an early
system_powerdown request happens before guest gpio driver loads.
Fix this problem by using gpio_key.
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Message-id: 1458221140-15232-3-git-send-email-zhaoshenglong@huawei.com
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
This will be used by ARM virt machine as a power button.
Signed-off-by: Shannon Zhao <shannon.zhao@linaro.org>
Message-id: 1458221140-15232-2-git-send-email-zhaoshenglong@huawei.com
[PMM: Use hyphen rather than underscore in type names;
add a comment briefly describing what the device does]
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Changes:
* add initial MIPS CPS support
* implement ITU block
* implement MAAR
-----BEGIN PGP SIGNATURE-----
iQEcBAABAgAGBQJW+43VAAoJEFIRjjwLKdprLmoH/1iWT4WsUJDF+9KX7PpFANbQ
DT+QSDBJr6K+jCenLlqfvB30txS+NDRFzmW65J8hlawVOwhamg1X+pcQTbAYy0sm
Du3Wexye0uw5YKUmqK2oCrgLJCKm3AqsmraaITE8q1URlkrQpuOuzazlIx5UA+RW
RgF/DsPAlit8TkZMHwaVIOeXUl8vl8152fU26QvwOGAT6J3lV+lQJ+gMPGRSAWOw
dcuVGNOTV0g3+kzOWisiqZc/V0Wp2Yu5IPezEVkFjZ4iyTTpnR8gkzuTNebzoRZo
Zmws4mqaZAX6ijveevd+ueh5sR+AX+mFBunoXQVSSBvRFlqBcEEL2nrwb9wOUX4=
=Ns6n
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/lalrae/tags/mips-20160329-2' into staging
MIPS patches 2016-03-29
Changes:
* add initial MIPS CPS support
* implement ITU block
* implement MAAR
# gpg: Signature made Wed 30 Mar 2016 09:27:01 BST using RSA key ID 0B29DA6B
# gpg: Good signature from "Leon Alrae <leon.alrae@imgtec.com>"
* remotes/lalrae/tags/mips-20160329-2: (21 commits)
target-mips: add MAAR, MAARI register
target-mips: use CP0_CHECK for gen_m{f|t}hc0
hw/mips/cps: enable ITU for multithreading processors
target-mips: make ITC Configuration Tags accessible to the CPU
target-mips: check CP0 enabled for CACHE instruction also in R6
hw/mips: implement ITC Storage - Bypass View
hw/mips: implement ITC Storage - P/V Sync and Try Views
hw/mips: implement ITC Storage - Empty/Full Sync and Try Views
hw/mips: implement ITC Storage - Control View
hw/mips: implement ITC Configuration Tags and Storage Cells
target-mips: enable CM GCR in MIPS64R6-generic CPU
hw/mips_malta: add CPS to Malta board
hw/mips_malta: move CPU creation to a separate function
hw/mips_malta: remove redundant irq and clock init
hw/mips_malta: remove CPUMIPSState from the write_bootloader()
hw/mips/cps: create CPC block inside CPS
hw/mips: add initial Cluster Power Controller support
hw/mips/cps: create GCR block inside CPS
hw/mips: add initial Global Config Register support
target-mips: add CMGCRBase register
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Reviewed-by: Max Reitz <mreitz@redhat.com>
Bypass View does not cause issuing thread to block and does not affect
any of the cells state bit.
Read from a FIFO cell returns the value of the oldest entry.
Store to a FIFO cell changes the value of the newest entry.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
P/V Synchronized and Try Views can be used to access Semaphore cells.
Load returns current value and post-decrements the value in the cell
(until it reaches zero). Stores increment the value (until it saturates
at 0xFFFF).
P/V Synchronized View causes the issuing thread to block on read if value
is 0. P/V Try View does not block the thread, it returns 0 in this case.
Cell's Empty and Full bits are not modified.
Trap bit (i.e. Gating Storage exceptions) not implemented.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Empty/Full Synchronized and Try views can be used to access FIFO cells.
Store to the FIFO cell pushes the value into the queue, load pops the oldest
element from the queue. Cell's Full and Empty bits are automatically updated
to reflect new state of the cell.
Empty/Full Synchronized View causes the issuing thread to block when FIFO is
empty while thread is performing a read, or FIFO is full while thread is
performing a write.
Empty/Full Try View never blocks the thread. If cell is full then write is
ignored, if cell is empty then load returns 0.
Trap bit (i.e. Gating Storage exceptions) not implemented.
Store Conditional support for E/F Try View (i.e. indicate failure if FIFO
is full) not implemented.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Control view is used to access the ITC Storage Cell Tags. It never causes
the issuing thread to block.
Guest can empty the FIFO cell by setting Empty bit to 1.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Implement ITC as a single object consisting of two memory regions:
1) tag_io: ITC Configuration Tags (i.e. ITCAddressMap{0,1} registers) which
are accessible by the CPU via CACHE instruction. Also adding
MemoryRegion *itc_tag to the CPUMIPSState so that CACHE instruction will
dispatch reads/writes directly.
2) storage_io: memory-mapped ITC Storage whose address space is configurable
(i.e. enabled/remapped/resized) by writing to ITCAddressMap{0,1} registers.
ITC Storage contains FIFO and Semaphore cells. Read-only FIFO bit in the
ITC cell tag indicates the type of the cell. If the ITC Storage contains
both types of cells then FIFOs are located before Semaphores.
Since issuing thread can get blocked on the access to a cell (in E/F
Synchronized and P/V Synchronized Views) each cell has a bitmap to track
which threads are currently blocked.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
If the user specifies smp > 1 and the CPU with CM GCR support, then
create Coherent Processing System (which takes care of instantiating CPUs)
rather than CPUs directly and connect i8259 and cbus to the pins exposed by
CPS. However, there is no GIC yet, thus CPS exposes CPU's IRQ pins so use
the same pin numbers as before.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Global smp_cpus is never zero (even if user provides -smp 0), thus clocks
and irqs are always initialized for each created CPU in the loop at the
beginning of mips_malta_init.
These two lines cause a leak of already allocated timer and irqs for the
first CPU - remove them.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Create Cluster Power Controller and add a link to the CPC MemoryRegion
in GCR. Guest can enable / map CPC to any physical address by writing to
the memory-mapped GCR_CPC_BASE register.
Set vp-start-reset property to 1 to allow only first VP to run from reset.
Others are brought up by the guest via CPC memory-mapped registers.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Cluster Power Controller (CPC) is responsible for power management in
multiprocessing system. It provides registers to control the power and the
clock frequency of the individual elements in the system.
This patch implements only three registers that are used to control the
power state of each VP on a single core:
* VP Run is a write-only register used to set each VP to the run state
* VP Stop is a write-only register used to set each VP to the suspend state
* VP Running is a read-only register indicating the run state of each VP
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Add initial GCR support to indicate number of VPs present in the system,
L2 bypass mode and revision number.
Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
[leon.alrae@imgtec.com:
* removed GIC part,
* changed commit message,
* replaced %lx format spec. with PRIx64,
* renamed mips_gcr.{c,h} to mips_cmgcr.{c,h},
* replaced CONFIG_MIPS_GIC with CONFIG_MIPS_CPS]
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Implement generic MIPS Coherent Processing System (CPS) which in this
commit just creates VPs, but it will serve as a container also for
other components like Global Configuration Registers and Cluster Power
Controller.
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
This reverts commit 9596ef7c7b.
This workaround in order to fix endless interrupts is no
longer needed because it was superseded by the previous patch
(e1000: Fixing interrupt pace).
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
This patch introduces an upper bound for number of interrupts
per second. Without this bound an interrupt storm can occur as
it has been observed on Windows 10 when disabling the device.
According to the SPEC - Intel PCI/PCI-X Family of Gigabit
Ethernet Controllers Software Developer's Manual, section
13.4.18 - the Ethernet controller guarantees a maximum
observable interrupt rate of 7813 interrupts/sec. If there is
no upper bound this could lead to an interrupt storm by e1000
(when mit_delay < 500) causing interrupts to fire at a very high
pace.
Thus if mit_delay < 500 then the delay should be set to the
minimum delay possible which is 500. This can be calculated
easily as follows:
Interval = 10^9 / (7813 * 256) = 500.
Signed-off-by: Sameeh Jubran <sameeh@daynix.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
- Use 128bit math to avoid asserts with IOMMU regions (Bandan Das)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (GNU/Linux)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=DpB8
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/awilliam/tags/vfio-update-20160328.0' into staging
VFIO updates 2016-03-28
- Use 128bit math to avoid asserts with IOMMU regions (Bandan Das)
# gpg: Signature made Mon 28 Mar 2016 23:16:52 BST using RSA key ID 3BB08B22
# gpg: Good signature from "Alex Williamson <alex.williamson@redhat.com>"
# gpg: aka "Alex Williamson <alex@shazbot.org>"
# gpg: aka "Alex Williamson <alwillia@redhat.com>"
# gpg: aka "Alex Williamson <alex.l.williamson@gmail.com>"
* remotes/awilliam/tags/vfio-update-20160328.0:
vfio: convert to 128 bit arithmetic calculations when adding mem regions
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
vfio_listener_region_add for a iommu mr results in
an overflow assert since iommu memory region is initialized
with UINT64_MAX. Convert calculations to 128 bit arithmetic
for iommu memory regions and let int128_get64 assert for non iommu
regions if there's an overflow.
Suggested-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bandan Das <bsd@redhat.com>
[missed (end - 1) on 2nd trace call, move llsize closer to use]
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
* Chardev fix from Marc-André
* config.status tweak from David
* Header file tweaks from Markus, myself and Veronia (Outreachy candidate)
* get_ticks_per_sec() removal from Rutuja (Outreachy candidate)
* Coverity fix from myself
* PKE implementation from myself, based on rth's XSAVE support
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQEcBAABCAAGBQJW9ErPAAoJEL/70l94x66DJfEH/A/QkMpAhrgNdyVsahzsGrzE
wx5gHFIc1nBYxyr62w4apUb5jPB7zaXu0LA7EAWDeAe0pyP8hZzLT9kJyOEDsuJu
zwKN2QeLSNMtPbnbKN0I/YQ2za2xX1V5ruhSeOJoVslUI214hgnAURaGshhQNzuZ
2CluDT9KgL5cQifAnKs5kJrwhIYShYNQB+1eDC/7wk28dd/EH+sPALIoF+rqrSmt
Zu4Mdqd+9Ns+oKOjA6br9ULq/Hzg0aDfY82J+XLVVqfF3PXQe8rTDmuMf/7jTn+M
Un7ZOcei9oZF2/9vfAfKQpDCcgD9HvOUSbgqV/ubmkPPmN/LNJzeKj0fBhrRN+Y=
=K12D
-----END PGP SIGNATURE-----
Merge remote-tracking branch 'remotes/bonzini/tags/for-upstream' into staging
* Log filtering from Alex and Peter
* Chardev fix from Marc-André
* config.status tweak from David
* Header file tweaks from Markus, myself and Veronia (Outreachy candidate)
* get_ticks_per_sec() removal from Rutuja (Outreachy candidate)
* Coverity fix from myself
* PKE implementation from myself, based on rth's XSAVE support
# gpg: Signature made Thu 24 Mar 2016 20:15:11 GMT using RSA key ID 78C7AE83
# gpg: Good signature from "Paolo Bonzini <bonzini@gnu.org>"
# gpg: aka "Paolo Bonzini <pbonzini@redhat.com>"
* remotes/bonzini/tags/for-upstream: (28 commits)
target-i386: implement PKE for TCG
config.status: Pass extra parameters
char: translate from QIOChannel error to errno
exec: fix error handling in file_ram_alloc
cputlb: modernise the debug support
qemu-log: support simple pid substitution for logs
target-arm: dfilter support for in_asm
qemu-log: dfilter-ise exec, out_asm, op and opt_op
qemu-log: new option -dfilter to limit output
qemu-log: Improve the "exec" TB execution logging
qemu-log: Avoid function call for disabled qemu_log_mask logging
qemu-log: correct help text for -d cpu
tcg: pass down TranslationBlock to tcg_code_gen
util: move declarations out of qemu-common.h
Replaced get_tick_per_sec() by NANOSECONDS_PER_SECOND
hw: explicitly include qemu-common.h and cpu.h
include/crypto: Include qapi-types.h or qemu/bswap.h instead of qemu-common.h
isa: Move DMA_transfer_handler from qemu-common.h to hw/isa/isa.h
Move ParallelIOArg from qemu-common.h to sysemu/char.h
Move QEMU_ALIGN_*() from qemu-common.h to qemu/osdep.h
...
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
Conflicts:
scripts/clean-includes
RX buffer pools are now enabled by default for new machine types.
For older machine types, they are still disabled to avoid breaking
migration.
Signed-off-by: Thomas Huth <thuth@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
tl;dr:
This patch introduces an alternate way of handling the receive
buffers of the spapr-vlan device, resulting in much better
receive performance for the guest.
Full story:
One of our testers recently discovered that the performance of the
spapr-vlan device is very poor compared to other NICs, and that
a simple "ping -i 0.2 -s 65507 someip" in the guest can result
in more than 50% lost ping packets (especially with older guest
kernels < 3.17).
After doing some analysis, it was clear that there is a problem
with the way we handle the receive buffers in spapr_llan.c: The
ibmveth driver of the guest Linux kernel tries to add a lot of
buffers into several buffer pools (with 512, 2048 and 65536 byte
sizes by default, but it can be changed via the entries in the
/sys/devices/vio/1000/pool* directories of the guest). However,
the spapr-vlan device of QEMU only tries to squeeze all receive
buffer descriptors into one single page which has been supplied
by the guest during the H_REGISTER_LOGICAL_LAN call, without
taking care of different buffer sizes. This has two bad effects:
First, only a very limited number of buffer descriptors is accepted
at all. Second, we also hand 64k buffers to the guest even if
the 2k buffers would fit better - and this results in dropped packets
in the IP layer of the guest since too much skbuf memory is used.
Though it seems at a first glance like PAPR says that we should store
the receive buffer descriptors in the page that is supplied during
the H_REGISTER_LOGICAL_LAN call, chapter 16.4.1.2 in the LoPAPR spec
declares that "the contents of these descriptors are architecturally
opaque, none of these descriptors are manipulated by code above
the architected interfaces". That means we don't have to store
the RX buffer descriptors in this page, but can also manage the
receive buffers at the hypervisor level only. This is now what we
are doing here: Introducing proper RX buffer pools which are also
sorted by size of the buffers, so we can hand out a buffer with
the best fitting size when a packet has been received.
To avoid problems with migration from/to older version of QEMU,
the old behavior is also retained and enabled by default. The new
buffer management has to be enabled via a new "use-rx-buffer-pools"
property.
Now with the new buffer pool management enabled, the problem with
"ping -s 65507" is fixed for me, and the throughput of a simple
test with wget increases from creeping 3MB/s up to 20MB/s!
Signed-off-by: Thomas Huth <thuth@redhat.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
