URLs to ftp.kernel.org are still exist though the service is closed [0].
This commit fixes the URLs to use www.kernel.org instead.
[0] https://www.kernel.org/shutting-down-ftp-services.html
Signed-off-by: SeongJae Park <sj38.park@gmail.com>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
One hard problem adding .low limit is to detect idle cgroup. If one
cgroup doesn't dispatch enough IO against its low limit, we must have a
mechanism to determine if other cgroups dispatch more IO. We added the
think time detection mechanism before, but it doesn't work for all
workloads. Here we add a latency based approach.
We already have mechanism to calculate latency threshold for each IO
size. For every IO dispatched from a cgorup, we compare its latency
against its threshold and record the info. If most IO latency is below
threshold (in the code I use 75%), the cgroup could be treated idle and
other cgroups can dispatch more IO.
Currently this latency target check is only for SSD as we can't
calcualte the latency target for hard disk. And this is only for cgroup
leaf node so far.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
User configures latency target, but the latency threshold for each
request size isn't fixed. For a SSD, the IO latency highly depends on
request size. To calculate latency threshold, we sample some data, eg,
average latency for request size 4k, 8k, 16k, 32k .. 1M. The latency
threshold of each request size will be the sample latency (I'll call it
base latency) plus latency target. For example, the base latency for
request size 4k is 80us and user configures latency target 60us. The 4k
latency threshold will be 80 + 60 = 140us.
To sample data, we calculate the order base 2 of rounded up IO sectors.
If the IO size is bigger than 1M, it will be accounted as 1M. Since the
calculation does round up, the base latency will be slightly smaller
than actual value. Also if there isn't any IO dispatched for a specific
IO size, we will use the base latency of smaller IO size for this IO
size.
But we shouldn't sample data at any time. The base latency is supposed
to be latency where disk isn't congested, because we use latency
threshold to schedule IOs between cgroups. If disk is congested, the
latency is higher, using it for scheduling is meaningless. Hence we only
do the sampling when block throttling is in the LOW limit, with
assumption disk isn't congested in such state. If the assumption isn't
true, eg, low limit is too high, calculated latency threshold will be
higher.
Hard disk is completely different. Latency depends on spindle seek
instead of request size. Currently this feature is SSD only, we probably
can use a fixed threshold like 4ms for hard disk though.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Currently there is no way to know the request size when the request is
finished. Next patch will need this info. We could add extra field to
record the size, but blk_issue_stat has enough space to record it, so
this patch just overloads blk_issue_stat. With this, we will have 49bits
to track time, which still is very long time.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Here we introduce per-cgroup latency target. The target determines how a
cgroup can afford latency increasement. We will use the target latency
to calculate a threshold and use it to schedule IO for cgroups. If a
cgroup's bandwidth is below its low limit but its average latency is
below the threshold, other cgroups can safely dispatch more IO even
their bandwidth is higher than their low limits. On the other hand, if
the first cgroup's latency is higher than the threshold, other cgroups
are throttled to their low limits. So the target latency determines how
we efficiently utilize free disk resource without sacifice of worload's
IO latency.
For example, assume 4k IO average latency is 50us when disk isn't
congested. A cgroup sets the target latency to 30us. Then the cgroup can
accept 50+30=80us IO latency. If the cgroupt's average IO latency is
90us and its bandwidth is below low limit, other cgroups are throttled
to their low limit. If the cgroup's average IO latency is 60us, other
cgroups are allowed to dispatch more IO. When other cgroups dispatch
more IO, the first cgroup's IO latency will increase. If it increases to
81us, we then throttle other cgroups.
User will configure the interface in this way:
echo "8:16 rbps=2097152 wbps=max latency=100 idle=200" > io.low
latency is in microsecond unit
By default, latency target is 0, which means to guarantee IO latency.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Last patch introduces a way to detect idle cgroup. We use it to make
upgrade/downgrade decision. And the new algorithm can detect completely
idle cgroup too, so we can delete the corresponding code.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Add interface to configure the threshold. The io.low interface will
like:
echo "8:16 rbps=2097152 wbps=max idle=2000" > io.low
idle is in microsecond unit.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
A cgroup gets assigned a low limit, but the cgroup could never dispatch
enough IO to cross the low limit. In such case, the queue state machine
will remain in LIMIT_LOW state and all other cgroups will be throttled
according to low limit. This is unfair for other cgroups. We should
treat the cgroup idle and upgrade the state machine to lower state.
We also have a downgrade logic. If the state machine upgrades because of
cgroup idle (real idle), the state machine will downgrade soon as the
cgroup is below its low limit. This isn't what we want. A more
complicated case is cgroup isn't idle when queue is in LIMIT_LOW. But
when queue gets upgraded to lower state, other cgroups could dispatch
more IO and this cgroup can't dispatch enough IO, so the cgroup is below
its low limit and looks like idle (fake idle). In this case, the queue
should downgrade soon. The key to determine if we should do downgrade is
to detect if cgroup is truely idle.
Unfortunately it's very hard to determine if a cgroup is real idle. This
patch uses the 'think time check' idea from CFQ for the purpose. Please
note, the idea doesn't work for all workloads. For example, a workload
with io depth 8 has disk utilization 100%, hence think time is 0, eg,
not idle. But the workload can run higher bandwidth with io depth 16.
Compared to io depth 16, the io depth 8 workload is idle. We use the
idea to roughly determine if a cgroup is idle.
We treat a cgroup idle if its think time is above a threshold (by
default 1ms for SSD and 100ms for HD). The idea is think time above the
threshold will start to harm performance. HD is much slower so a longer
think time is ok.
The patch (and the latter patches) uses 'unsigned long' to track time.
We convert 'ns' to 'us' with 'ns >> 10'. This is fast but loses
precision, should not a big deal.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
When cgroups all reach low limit, cgroups can dispatch more IO. This
could make some cgroups dispatch more IO but others not, and even some
cgroups could dispatch less IO than their low limit. For example, cg1
low limit 10MB/s, cg2 limit 80MB/s, assume disk maximum bandwidth is
120M/s for the workload. Their bps could something like this:
cg1/cg2 bps: T1: 10/80 -> T2: 60/60 -> T3: 10/80
At T1, all cgroups reach low limit, so they can dispatch more IO later.
Then cg1 dispatch more IO and cg2 has no room to dispatch enough IO. At
T2, cg2 only dispatches 60M/s. Since We detect cg2 dispatches less IO
than its low limit 80M/s, we downgrade the queue from LIMIT_MAX to
LIMIT_LOW, then all cgroups are throttled to their low limit (T3). cg2
will have bandwidth below its low limit at most time.
The big problem here is we don't know the maximum bandwidth of the
workload, so we can't make smart decision to avoid the situation. This
patch makes cgroup bandwidth change smooth. After disk upgrades from
LIMIT_LOW to LIMIT_MAX, we don't allow cgroups use all bandwidth upto
their max limit immediately. Their bandwidth limit will be increased
gradually to avoid above situation. So above example will became
something like:
cg1/cg2 bps: 10/80 -> 15/105 -> 20/100 -> 25/95 -> 30/90 -> 35/85 -> 40/80
-> 45/75 -> 22/98
In this way cgroups bandwidth will be above their limit in majority
time, this still doesn't fully utilize disk bandwidth, but that's
something we pay for sharing.
Scale up is linear. The limit scales up 1/2 .low limit every
throtl_slice after upgrade. The scale up will stop if the adjusted limit
hits .max limit. Scale down is exponential. We cut the scale value half
if a cgroup doesn't hit its .low limit. If the scale becomes 0, we then
fully downgrade the queue to LIMIT_LOW state.
Note this doesn't completely avoid cgroup running under its low limit.
The best way to guarantee cgroup doesn't run under its limit is to set
max limit. For example, if we set cg1 max limit to 40, cg2 will never
run under its low limit.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
cgroup could be assigned a limit, but doesn't dispatch enough IO, eg the
cgroup is idle. When this happens, the cgroup doesn't hit its limit, so
we can't move the state machine to higher level and all cgroups will be
throttled to their lower limit, so we waste bandwidth. Detecting idle
cgroup is hard. This patch handles a simple case, a cgroup doesn't
dispatch any IO. We ignore such cgroup's limit, so other cgroups can use
the bandwidth.
Please note this will be replaced with a more sophisticated algorithm
later, but this demonstrates the idea how we handle idle cgroups, so I
leave it here.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
The throtl_slice is 100ms by default. This is a long time for SSD, a lot
of IO can run. To make cgroups have smoother throughput, we choose a
small value (20ms) for SSD.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
throtl_slice is important for blk-throttling. It's called slice
internally but it really is a time window blk-throttling samples data.
blk-throttling will make decision based on the samplings. An example is
bandwidth measurement. A cgroup's bandwidth is measured in the time
interval of throtl_slice.
A small throtl_slice meanse cgroups have smoother throughput but burn
more CPUs. It has 100ms default value, which is not appropriate for all
disks. A fast SSD can dispatch a lot of IOs in 100ms. This patch makes
it tunable.
Since throtl_slice isn't a time slice, the sysfs name
'throttle_sample_time' reflects its character better.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
cgroup could be throttled to a limit but when all cgroups cross high
limit, queue enters a higher state and so the group should be throttled
to a higher limit. It's possible the cgroup is sleeping because of
throttle and other cgroups don't dispatch IO any more. In this case,
nobody can trigger current downgrade/upgrade logic. To fix this issue,
we could either set up a timer to wakeup the cgroup if other cgroups are
idle or make sure this cgroup doesn't sleep too long. Setting up a timer
means we must change the timer very frequently. This patch chooses the
latter. Making cgroup sleep time not too big wouldn't change cgroup
bps/iops, but could make it wakeup more frequently, which isn't a big
issue because throtl_slice * 8 is already quite big.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
When queue state machine is in LIMIT_MAX state, but a cgroup is below
its low limit for some time, the queue should be downgraded to lower
state as one cgroup's low limit isn't met.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
When queue is in LIMIT_LOW state and all cgroups with low limit cross
the bps/iops limitation, we will upgrade queue's state to
LIMIT_MAX. To determine if a cgroup exceeds its limitation, we check if
the cgroup has pending request. Since cgroup is throttled according to
the limit, pending request means the cgroup reaches the limit.
If a cgroup has limit set for both read and write, we consider the
combination of them for upgrade. The reason is read IO and write IO can
interfere with each other. If we do the upgrade based in one direction
IO, the other direction IO could be severly harmed.
For a cgroup hierarchy, there are two cases. Children has lower low
limit than parent. Parent's low limit is meaningless. If children's
bps/iops cross low limit, we can upgrade queue state. The other case is
children has higher low limit than parent. Children's low limit is
meaningless. As long as parent's bps/iops (which is a sum of childrens
bps/iops) cross low limit, we can upgrade queue state.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
each queue will have a state machine. Initially queue is in LIMIT_LOW
state, which means all cgroups will be throttled according to their low
limit. After all cgroups with low limit cross the limit, the queue state
gets upgraded to LIMIT_MAX state.
For max limit, cgroup will use the limit configured by user.
For low limit, cgroup will use the minimal value between low limit and
max limit configured by user. If the minimal value is 0, which means the
cgroup doesn't configure low limit, we will use max limit to throttle
the cgroup and the cgroup is ready to upgrade to LIMIT_MAX
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Add low limit for cgroup and corresponding cgroup interface. To be
consistent with memcg, we allow users configure .low limit higher than
.max limit. But the internal logic always assumes .low limit is lower
than .max limit. So we add extra bps/iops_conf fields in throtl_grp for
userspace configuration. Old bps/iops fields in throtl_grp will be the
actual limit we use for throttling.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
As discussed in LSF, add configure option for the interface and mark it
as experimental, so people can try/test.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
We are going to support low/max limit, each cgroup will have 2 limits
after that. This patch prepares for the multiple limits change.
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Jens Axboe <axboe@fb.com>
Create a trace event for guest mode changes, and enable VZ's
GuestCtl0.MC bit after the trace event is enabled to trap all guest mode
changes.
The MC bit causes Guest Hardware Field Change (GHFC) exceptions whenever
a guest mode change occurs (such as an exception entry or return from
exception), so we need to handle this exception now. The MC bit is only
enabled when restoring register state, so enabling the trace event won't
take immediate effect.
Tracing guest mode changes can be particularly handy when trying to work
out what a guest OS gets up to before something goes wrong, especially
if the problem occurs as a result of some previous guest userland
exception which would otherwise be invisible in the trace.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Transfer timer state to the VZ guest context (CP0_GTOffset & guest
CP0_Count) when entering guest mode, enabling direct guest access to it,
and transfer back to soft timer when saving guest register state.
This usually allows guest code to directly read CP0_Count (via MFC0 and
RDHWR) and read/write CP0_Compare, without trapping to the hypervisor
for it to emulate the guest timer. Writing to CP0_Count or CP0_Cause.DC
is much less common and still triggers a hypervisor GPSI exception, in
which case the timer state is transferred back to an hrtimer before
emulating the write.
We are careful to prevent small amounts of drift from building up due to
undeterministic time intervals between reading of the ktime and reading
of CP0_Count. Some drift is expected however, since the system
clocksource may use a different timer to the local CP0_Count timer used
by VZ. This is permitted to prevent guest CP0_Count from appearing to go
backwards.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add emulation of Memory Accessibility Attribute Registers (MAARs) when
necessary. We can't actually do anything with whatever the guest
provides, but it may not be possible to clear Guest.Config5.MRP so we
have to emulate at least a pair of MAARs.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
When restoring guest state after another VCPU has run, be sure to clear
CP0_LLAddr.LLB in order to break any interrupted atomic critical
section. Without this SMP guest atomics don't work when LLB is present
as one guest can complete the atomic section started by another guest.
MIPS VZ guest read of CP0_LLAddr causes Guest Privileged Sensitive
Instruction (GPSI) exception due to the address being root physical.
Handle this by reporting only the LLB bit, which contains the bit for
whether a ll/sc atomic is in progress without any reason for failure.
Similarly on P5600 a guest write to CP0_LLAddr also causes a GPSI
exception. Handle this also by clearing the guest LLB bit from root
mode.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add support for VZ guest CP0_PWBase, CP0_PWField, CP0_PWSize, and
CP0_PWCtl registers for controlling the guest hardware page table walker
(HTW) present on P5600 and P6600 cores. These guest registers need
initialising on R6, context switching, and exposing via the KVM ioctl
API when they are present.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add support for VZ guest CP0_SegCtl0, CP0_SegCtl1, and CP0_SegCtl2
registers, as found on P5600 and P6600 cores. These guest registers need
initialising, context switching, and exposing via the KVM ioctl API when
they are present.
They also require the GVA -> GPA translation code for handling a GVA
root exception to be updated to interpret the segmentation registers and
decode the faulting instruction enough to detect EVA memory access
instructions.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add support for VZ guest CP0_ContextConfig and CP0_XContextConfig
(MIPS64 only) registers, as found on P5600 and P6600 cores. These guest
registers need initialising, context switching, and exposing via the KVM
ioctl API when they are present.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add support for VZ guest CP0_BadInstr and CP0_BadInstrP registers, as
found on most VZ capable cores. These guest registers need context
switching, and exposing via the KVM ioctl API when they are present.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add support for the MIPS Virtualization (VZ) ASE to the MIPS KVM build
system. For now KVM can only be configured for T&E or VZ and not both,
but the design of the user facing APIs support the possibility of having
both available, so this could change in future.
Note that support for various optional guest features (some of which
can't be turned off) are implemented in immediately following commits,
so although it should now be possible to build VZ support, it may not
work yet on your hardware.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add the main support for the MIPS Virtualization ASE (A.K.A. VZ) to MIPS
KVM. The bulk of this work is in vz.c, with various new state and
definitions elsewhere.
Enough is implemented to be able to run on a minimal VZ core. Further
patches will fill out support for guest features which are optional or
can be disabled.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
The general guest exit handler needs a few tweaks for VZ compared to
trap & emulate, which for now are made directly depending on
CONFIG_KVM_MIPS_VZ:
- There is no need to re-enable the hardware page table walker (HTW), as
it can be left enabled during guest mode operation with VZ.
- There is no need to perform a privilege check, as any guest privilege
violations should have already been detected by the hardware and
triggered the appropriate guest exception.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Ifdef out the trap & emulate CACHE instruction emulation functions for
VZ. We will provide separate CACHE instruction emulation in vz.c, and we
need to avoid linker errors due to the use of T&E specific MMU helpers.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Update emulation of guest writes to CP0_Compare for VZ. There are two
main differences compared to trap & emulate:
- Writing to CP0_Compare in the VZ hardware guest context acks any
pending timer, clearing CP0_Cause.TI. If we don't want an ack to take
place we must carefully restore the TI bit if it was previously set.
- Even with guest timer access disabled in CP0_GuestCtl0.GT, if the
guest CP0_Count reaches the guest CP0_Compare the timer interrupt
will assert. To prevent this we must set CP0_GTOffset to move the
guest CP0_Count out of the way of the new guest CP0_Compare, either
before or after depending on whether it is a forwards or backwards
change.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add functions for MIPS VZ TLB management to tlb.c.
kvm_vz_host_tlb_inv() will be used for invalidating root TLB entries
after GPA page tables have been modified due to a KVM page fault. It
arranges for a root GPA mapping to be flushed from the TLB, using the
gpa_mm ASID or the current GuestID to do the probe.
kvm_vz_local_flush_roottlb_all_guests() and
kvm_vz_local_flush_guesttlb_all() flush all TLB entries in the
corresponding TLB for guest mappings (GPA->RPA for root TLB with
GuestID, and all entries for guest TLB). They will be used when starting
a new GuestID cycle, when VZ hardware is enabled/disabled, and also when
switching to a guest when the guest TLB contents may be stale or belong
to a different VM.
kvm_vz_guest_tlb_lookup() converts a guest virtual address to a guest
physical address using the guest TLB. This will be used to decode guest
virtual addresses which are sometimes provided by VZ hardware in
CP0_BadVAddr for certain exceptions when the guest physical address is
unavailable.
kvm_vz_save_guesttlb() and kvm_vz_load_guesttlb() will be used to
preserve wired guest VTLB entries while a guest isn't running.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Update MIPS KVM entry code to support VZ:
- We need to set GuestCtl0.GM while in guest mode.
- For cores supporting GuestID, we need to set the root GuestID to
match the main GuestID while in guest mode so that the root TLB
refill handler writes the correct GuestID into the TLB.
- For cores without GuestID where the root ASID dealiases RVA/GPA
mappings, we need to load that ASID from the gpa_mm rather than the
per-VCPU guest_kernel_mm or guest_user_mm, since the root TLB maps
guest physical addresses. We also need to restore the normal process
ASID on exit.
- The normal linux process pgd needs restoring on exit, as we can't
leave the GPA mappings active for kernel code.
- GuestCtl0 needs saving on exit for the GExcCode field, as it may be
clobbered if a preemption occurs.
We also need to move the TLB refill handler to the XTLB vector at offset
0x80 on 64-bit VZ kernels, as hardware will use Root.Status.KX to
determine whether a TLB refill or XTLB Refill exception is to be taken
on a root TLB miss from guest mode, and KX needs to be set for kernel
code to be able to access the 64-bit segments.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Abstract the MIPS KVM guest CP0 register access macros into inline
functions which are generated by macros. This allows them to be
generated differently for VZ, where they will usually need to access the
hardware guest CP0 context rather than the saved values in RAM.
Accessors for each individual register are generated using these macros:
- __BUILD_KVM_*_SW() for registers which are not present in the VZ
hardware guest context, so kvm_{read,write}_c0_guest_##name() will
access the saved value in RAM regardless of whether VZ is enabled.
- __BUILD_KVM_*_HW() for registers which are present in the VZ hardware
guest context, so kvm_{read,write}_c0_guest_##name() will access the
hardware register when VZ is enabled.
These build the underlying accessors using further macros:
- __BUILD_KVM_*_SAVED() builds e.g. kvm_{read,write}_sw_gc0_##name()
functions for accessing the saved versions of the registers in RAM.
This is used for implementing the common
kvm_{read,write}_c0_guest_##name() accessors with T&E where registers
are always stored in RAM, but are also available with VZ HW registers
to allow them to be accessed while saved.
- __BUILD_KVM_*_VZ() builds e.g. kvm_{read,write}_vz_gc0_##name()
functions for accessing the VZ hardware guest context registers
directly. This is used for implementing the common
kvm_{read,write}_c0_guest_##name() accessors with VZ.
- __BUILD_KVM_*_WRAP() builds wrappers with different names, which
allows the common kvm_{read,write}_c0_guest_##name() functions to be
implemented using the VZ accessors while still having the SAVED
accessors available too.
- __BUILD_KVM_SAVE_VZ() builds functions for saving and restoring VZ
hardware guest context register state to RAM, improving conciseness
of VZ context saving and restoring.
Similar macros exist for generating modifiers (set, clear, change),
either with a normal unlocked read/modify/write, or using atomic LL/SC
sequences.
These changes change the types of 32-bit registers to u32 instead of
unsigned long, which requires some changes to printk() functions in MIPS
KVM.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add a callback for MIPS KVM implementations to handle the VZ guest
exit exception. Currently the trap & emulate implementation contains a
stub which reports an internal error, but the callback will be used
properly by the VZ implementation.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add an implementation callback for the kvm_arch_hardware_enable() and
kvm_arch_hardware_disable() architecture functions, with simple stubs
for trap & emulate. This is in preparation for VZ which will make use of
them.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add an implementation callback for checking presence of KVM extensions.
This allows implementation specific extensions to be provided without
ifdefs in mips.c.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Currently the software emulated timer is initialised to a frequency of
100MHz by kvm_mips_init_count(), but this isn't suitable for VZ where
the frequency of the guest timer matches that of the host.
Add a count_hz argument so the caller can specify the default frequency,
and move the call from kvm_arch_vcpu_create() to the implementation
specific vcpu_setup() callback, so that VZ can specify a different
frequency.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add a new KVM_CAP_MIPS_64BIT capability to indicate that 64-bit MIPS
guests are available and supported. In this case it should still be
possible to run 32-bit guest code. If not available it won't be possible
to run 64-bit guest code and the instructions may not be available, or
the kernel may not support full context switching of 64-bit registers.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add new KVM_CAP_MIPS_VZ and KVM_CAP_MIPS_TE capabilities, and in order
to allow MIPS KVM to support VZ without confusing old users (which
expect the trap & emulate implementation), define and start checking
KVM_CREATE_VM type codes.
The codes available are:
- KVM_VM_MIPS_TE = 0
This is the current value expected from the user, and will create a
VM using trap & emulate in user mode, confined to the user mode
address space. This may in future become unavailable if the kernel is
only configured to support VZ, in which case the EINVAL error will be
returned and KVM_CAP_MIPS_TE won't be available even though
KVM_CAP_MIPS_VZ is.
- KVM_VM_MIPS_VZ = 1
This can be provided when the KVM_CAP_MIPS_VZ capability is available
to create a VM using VZ, with a fully virtualized guest virtual
address space. If VZ support is unavailable in the kernel, the EINVAL
error will be returned (although old kernels without the
KVM_CAP_MIPS_VZ capability may well succeed and create a trap &
emulate VM).
This is designed to allow the desired implementation (T&E vs VZ) to be
potentially chosen at runtime rather than being fixed in the kernel
configuration.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Extend MIPS KVM stats counters and kvm_transition trace event codes to
cover hypervisor exceptions, which have their own GExcCode field in
CP0_GuestCtl0 with up to 32 hypervisor exception cause codes.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Update the implementation of kvm_lose_fpu() for VZ, where there is no
need to enable the FPU/MSA in the root context if the FPU/MSA state is
loaded but disabled in the guest context.
The trap & emulate implementation needs to disable FPU/MSA in the root
context when the guest disables them in order to catch the COP1 unusable
or MSA disabled exception when they're used and pass it on to the guest.
For VZ however as long as the context is loaded and enabled in the root
context, the guest can enable and disable it in the guest context
without the hypervisor having to do much, and will take guest exceptions
without hypervisor intervention if used without being enabled in the
guest context.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Implement additional MMIO emulation for MIPS64, including 64-bit
loads/stores, and 32-bit unsigned loads. These are only exposed on
64-bit VZ hosts.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Refactor MIPS KVM MMIO load/store emulation to reduce code duplication.
Each duplicate differed slightly anyway, and it will simplify adding
64-bit MMIO support for VZ.
kvm_mips_emulate_store() and kvm_mips_emulate_load() can now return
EMULATE_DO_MMIO (as possibly originally intended). We therefore stop
calling either of these from kvm_mips_emulate_inst(), which is now only
used by kvm_trap_emul_handle_cop_unusable() which is picky about return
values.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Emulate the HYPCALL instruction added in the VZ ASE and used by the MIPS
paravirtualised guest support that is already merged. The new hypcall.c
handles arguments and the return value. No actual hypercalls are yet
supported, but this still allows us to safely step over hypercalls and
set an error code in the return value for forward compatibility.
Non-zero HYPCALL codes are not handled.
We also document the hypercall ABI which asm/kvm_para.h uses.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: Andreas Herrmann <andreas.herrmann@caviumnetworks.com>
Cc: David Daney <david.daney@cavium.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Cc: linux-doc@vger.kernel.org
Add a distinct UNIQUE_GUEST_ENTRYHI() macro for invalidation of guest
TLB entries by KVM, using addresses in KSeg1 rather than KSeg0. This
avoids conflicts with guest invalidation routines when there is no EHINV
bit to mark the whole entry as invalid, avoiding guest machine check
exceptions on Cavium Octeon III.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Add some missing guest accessors and register field definitions for KVM
for MIPS VZ to make use of.
Guest CP0_LLAddr register accessors and definitions for the LLB field
allow KVM to clear the guest LLB to cancel in-progress LL/SC atomics on
restore, and to emulate accesses by the guest to the CP0_LLAddr
register.
Bitwise modifiers and definitions for the guest CP0_Wired and
CP0_Config1 registers allow KVM to modify fields within the CP0_Wired
and CP0_Config1 registers.
Finally a definition for the CP0_Config5.SBRI bit allows KVM to
initialise and allow modification of the guest version of the SBRI bit.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
Probe for availablility of M{T,F}HC0 instructions used with e.g. XPA in
the VZ guest context, and make it available via cpu_guest_has_mvh. This
will be helpful in properly emulating the MAAR registers in KVM for MIPS
VZ.
Signed-off-by: James Hogan <james.hogan@imgtec.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: "Radim Krčmář" <rkrcmar@redhat.com>
Cc: linux-mips@linux-mips.org
Cc: kvm@vger.kernel.org
