gdbsx_guest_mem_io() does not get d passed, it expects to handle the
domain lookup itself. Specifically, the caller of
XEN_DOMCTL_gdbsx_guestmemio is expected to use DOMID_IDLE to interact
with the hypervisor, rather than a domain, which doesn't interact well
with with the domain rcu lock.
Signed-off-by: Don Slutz <dslutz@verizon.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
The number of struct evtchn in a page must be a power of two. Under
some workloads performance is improved slightly by padding struct
evtchn to 64 bytes (a typical cache line size), thus putting the fewer
per-channel locks into each cache line.
This does not decrease the number of struct evtchn's per-page.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
When sending an event, use a new per-event channel lock to safely
validate the event channel state.
This new lock must be held when changing event channel state. Note
that the event channel lock must also be held when changing state from
ECS_FREE or it will race with a concurrent get_free_port() call.
To avoid having to take the remote event channel locks when sending to
an interdomain event channel, the local and remote channel locks are
both held when binding or closing an interdomain event channel.
This significantly increases the number of events that can be sent
from multiple VCPUs. But struct evtchn increases in size, reducing
the number that fit into a single page to 64 (instead of 128).
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
notify_via_xen_event_channel() and free_xen_event_channel() had to
check if the domain was dying because they may be called while the
domain is being destroyed and the struct evtchn's are being freed.
By deferring the freeing of the struct evtchn's until all references
to the domain are dropped, these functions can rely on the channel
state being present and valid.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Freeing a xen event channel would clear xen_consumer before clearing
the channel state, leaving a window where the channel is in a funny
state (still bound but no consumer).
Move the clear of xen_consumer into free_evtchn() where the state is
also cleared.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Ditch the pointless evtchn_close() wrapper around __evtchn_close()
(renaming the latter) as well as some bogus casts of function results
to void.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
The vector parameters are more usefully u8 right away. This is
particularly important for the vioapic_update_EOI() invocation from
vioapic_write() (which luckily is only a latent issue, as
VIOAPIC_VERSION_ID is still hard coded to 0x11 right now). But it at
once allows simplifying VMX's EXIT_REASON_EOI_INDUCED handling (the
kind of pointless helper function should have been static anyway; not
being use for anything else, it gets removed altogether).
Plus vlapic_handle_EOI() (now renamed for that purpose) can be used as
the tail of vlapic_EOI_set() instead of duplicating that code.
Finally replace a stray current->domain use in vlapic_handle_EOI().
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
If a guest is not evenly grant mapping across its VCPUs one of the
VCPUs may run out of free maptrack entries even though other VCPUs
have many free.
If this happens, "steal" free entries from other VCPUs. We want to
steal entries such that:
a) We avoid ping-ponging stolen entries between VCPUs.
b) The number of free entries owned by each VCPUs tends (over time) to
the number it uses.
So when stealing, we select a VCPU at random (reducing (a)) and we
transfer the stolen entries to the thief VCPU (aiming for (b)).
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Performance analysis of aggregate network throughput with many VMs
shows that performance is signficantly limited by contention on the
maptrack lock when obtaining/releasing maptrack handles from the free
list.
Instead of a single free list use a per-VCPU list. This avoids any
contention when obtaining a handle. Handles must be released back to
their original list and since this may occur on a different VCPU there
is some contention on the destination VCPU's free list tail pointer
(but this is much better than a per-domain lock).
Increase the default maximum number of maptrack frames by 4 times
because: a) struct grant_mapping is now 16 bytes (instead of 8); and
b) a guest may not evenly distribute all the grant map operations
across the VCPUs (meaning some VCPUs need more maptrack entries than
others).
Signed-off-by: Malcolm Crossley <malcolm.crossley@citrix.com>
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Fold msixtbl_addr_to_virt() + virt_to_msi_desc() into simplified
msixtbl_addr_to_desc(), as the callers don't need the virtual address
anymore.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
In particular we want to avoid losing track of our own intention to
have an entry masked. Physical unmasking now happens only when both
host and guest requested so.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
- __pci_enable_msix() now checks that an MSI-X capability was actually
found
- pass "pos" to msix_capability_init() as both callers already know it
(and hence there's no need to re-obtain it)
- call __pci_disable_msi{,x}() directly instead of via
pci_disable_msi() from __pci_enable_msi{x,}() state validation paths
- use msix_control_reg() instead of open coding it
- log message adjustments
- coding style corrections
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
The number of slots per page being 511 (i.e. not a power of two) means
that the (32-bit) read and write indexes going beyond 2^32 will likely
disturb operation. Extend I/O req server creation so the caller can
indicate that it is using suitable atomic accesses where needed (not
all accesses to the two pointers really need to be atomic), allowing
the hypervisor to atomically canonicalize both pointers when both have
gone through at least one cycle.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
In hvm_hap_nested_page_fault() latch the current domain alongside the
current vCPU into a local variable, making use of it where possible
also beyond what the title says.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Add a common interface for matching the current cpu against an
array of x86_cpu_ids. Also change mwait-idle.c to use it.
Signed-off-by: Wei Wang <wei.w.wang@intel.com>
Both PV and HVM logic have similar but not similar enough code here.
Synchronize the two so that
- in the HVM case we don't unconditionally try to access extended
config space
- in the PV case we pass a correct range to the XSM hook
- in the PV case we don't needlessly deny access when the operation
isn't really on PCI config space
All this along with sharing the macros HVM already had here.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
By keeping a count of the number of currently valid event channels,
port_is_valid() can be simplified.
d->valid_evtchns is only increased (while holding d->event_lock), so
port_is_valid() may be safely called without taking the lock (this
will be useful later).
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
The interface description of pvUSB lacks some access macros as using
linux kernel macros is assumed to work well. This solution is rather
unfriendly for pvusb implementations being outside the linux kernel.
Additionally things will break quite unpleasent in case the linux
kernel implementation is changed.
To avoid these problems define own macros for accessing bitfields of
the interface and for values of several structure members.
While working on the file add some more comments, especially for the
xenstore interface.
Signed-off-by: Juergen Gross <jgross@suse.com>
The statement to reset nb_watches should be in del_watches, not
del_transactions.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Cc: David Scott <dave.scott@citrix.com>
Acked-by: David Scott <dave.scott@citrix.com>
[ ijc -- fix syntax error by adding a ";" to the previous line in the
new location and removing from the previous line in the old ]
This patch does following things:
1. Document v1 format.
2. Factor out function to handle QEMU restore data and function to
handle v1 blob for restore path.
3. Refactor save function to generate different blobs in the order
specified in format specification.
4. Change functions to use "goto out" idiom.
No functional changes introduced.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
It uses a PPI which we cannot route to a guest, and will surely need
more support than just that anyway.
I noticed this on Mustang with UEFI where the built in DTB contains a
node of this type.
According to linux/Documentation/devicetree/bindings/arm/pmu.txt the
ARM v7 (Cortex-A{7,15}) PMUs require a PPI too, so blacklist them as
well.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Reviewed-by: Julien Grall <julien.grall@citrix.com>
xc_dom_load_elf_symtab was incorrectly trying to perform the same
calculations already done in elf_parse_bsdsyms when load == 0 is used.
Instead of trying to repeat the calculations, just trust what
elf_parse_bsdsyms has already accounted for.
This also simplifies the code by allowing the non-load case to return
earlier.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
The current code for allocating memory for PV guests batches the
hypercalls to allocate memory by allocating 1024*1024 extents of order 0
at a time. To make this faster, first try allocating extents of order 9
(2 MiB) before falling back to the order 0 allocating if the order 9
allocation fails.
On my test machine this reduced the time to start a 128 GiB PV guest by
about 60 seconds.
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
This patch does the following:
1. Use local variables for dummy vNUMA layout in PV case.
2. Avoid leaking dummy layout back to caller in PV case.
3. Use local variables to reference vNUMA layout (whether it is dummy
or provided by caller) for both PV and HVM.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
These files are leaked when using qemu-trad stubdom. They are
intermediate files created by libxc. Unfortunately they don't fit well
in our userdata scheme. Clean them up after we destroy all userdata,
we're sure they are not useful anymore at that point.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
This merges xenalyze.hg, changeset 150:24308507be1d,
into tools/xentrace/xenalyze.c to have the tool and
public/trace.h in one place.
Adjust code to use public/trace.h instead of private trace.h
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: George Dunlap <george.dunlap@eu.citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
[ ijc -- wrap $(BIN) install in a check in case it is empty (which it
is on !x86, avoid BIN += since it results in BIN = ' ' on
!x86 ]
We're going to want to free an event channel from two places. Factor out
the code into a free_evtchn() function.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
This documents that no entity outside of gnttab.c actually accesses
objects of that type, which is particularly important with the now more
fine grained locking in place.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
- don't update shared entry's frame number for translated domains (as
MFNs shouldn't be exposed to such guests)
- for v1 grant table format, force copying of the page also when the
intended MFN doesn't fit in 32 bits (and the domain isn't translated)
- fix an apparent off-by-one error (it's unclear to me why commit
5cc77f9098 ("32-on-64: Fix domain address-size clamping, implement")
uses BITS_PER_LONG-1 here, while using BITS_PER_LONG in the two other
invocations of domain_clamp_alloc_bitsize())
- adjust comments accompanying the shared entry's frame field
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
In a number of places both v1 and v2 pointers are being obtained when
none or just one suffices. Additionally in __acquire_grant_for_copy()
the flow of if/else-if can be slightly improved by re-ordering.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
The function doesn't need to return counts in the first place; all its
callers are after is whether at least one entry of a certain kind
exists. With that there's no point for that loop to continue once the
looked for condition was found to be met by one entry. Rename the
function to match the changed behavior.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
By having nr_grant_entries() return zero when the grant table version
is still unset we can reduce the number of error paths and at once fix
grant_map_exists() running into the being removed ASSERT() when called
for a page owned by a domain not having its grant table set up yet.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
In combination with the per-active entry locks, the grant table lock
can be made a read-write lock since the majority of cases only the
read lock is required. The grant table read lock protects against
changes to the table version or size (which are done with the write
lock held).
The write lock is also required when two active entries must be
acquired.
The double lock is still required when updating IOMMU page tables.
With the lock contention being only on the maptrack lock (unless IOMMU
updates are required), performance and scalability is improved.
Based on a patch originally by Matt Wilson <msw@amazon.com>.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Split grant table lock into two separate locks. One to protect
maptrack free list (maptrack_lock) and one for everything else (lock).
Based on a patch originally by Matt Wilson <msw@amazon.com>.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Introduce a per-active entry spin lock to protect active entry state
The grant table lock must be locked before acquiring (locking) an
active entry.
This is a step in reducing contention on the grant table lock, but
will only do so once the grant table lock is turned into a read-write
lock.
Based on a patch originally by Matt Wilson <msw@amazon.com>.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Constructing dom0 may take a few seconds, particularly if the slow VESA
graphics terminal is used. Process pending softirqs a few times to avoid
tripping a watchdog with a short timeout.
Signed-off-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Move inclusion of xen/softirq.h (and at once clean up other includes).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Don Slutz
David Vrabel
Jan Beulich
Malcolm Crossley
Wei Wang
Juergen Gross
Wei Liu
Ian Campbell
Roger Pau Monne
Ross Lagerwall
Olaf Hering
Ian Jackson