Store various PID files under XEN_RUN_DIR. Note that this change the
default location from /var/run to /var/run/xen.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Store xldevd.pid under XEN_RUN_DIR. Note that this will change the
default location from /var/run to /var/run/xen.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Place the PID file under XEN_RUN_DIR. Note that this change the default
location from /var/run to /var/run/xen.
Generate a _paths.h as that is required to make this change work.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Place the PID file under XEN_RUN_DIR by default. Note this change the
default location from /var/run to /var/run/xen.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Gcc 4.4.4 complained that the "pause" variable introduced in 22b430e0
("xl: add option to leave domain paused after migration") shadowed
pause(2) declaration in unistd.h.
Rename "pause" to "pause_after_migration" to fix this issue.
Reported-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
both introduced in d205f8a7f4 ("xen: credit2: rework
load tracking logic").
First, in __update_runq_load(), the ASSERT() was actually
useless. Let's instead check that the computed value of
the load has not overflowed (and hence gone negative).
While there, do that in __update_svc_load() as well.
Second, in balance_load(), cpus_max needs being extended
in order to be correctly shifted, and the result compared
with an s_time_t value, without risking loosing info.
Spotted by Coverity.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
In fact, right now, we recommend keepeing runqueues
arranged per-core, so that it is the inter-runqueue load
balancing code that automatically spreads the work in an
SMT friendly way. This means that any other runq
arrangement one may want to use falls short of SMT
scheduling optimizations.
This commit implements SMT awareness --similar to the
one we have in Credit1-- for any possible runq
arrangement. This turned out to be pretty easy to do,
as the logic can live entirely in runq_tickle()
(although, in order to avoid for_each_cpu loops in
that function, we use a new cpumask which indeed needs
to be updated in other places).
In addition to disentangling SMT awareness from load
balancing, this also allows us to support the
sched_smt_power_savings parametar in Credit2 as well.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Signed-off-by: George Dunlap <george.dunlap@citrix.com>
Reviewed-by: Anshul Makkar <anshul.makkar@citrix.com>
This is useful for debugging domains that crash on resume from migration.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
When this daemon is started after creating backend device, that device
will not be configured.
Racy situation:
1. driver domain is started
2. frontend domain is started (just after kicking driver domain off)
3. device in frontend domain is connected to the backend (as specified
in frontend domain configuration)
4. xl devd is started in driver domain
End result is that backend device in driver domain is not configured
(like network interface is not enabled), so the device doesn't work.
Fix this by artifically triggering events for devices already present in
xenstore before xl devd is started. Do this only after xenstore watch is
already registered, and only for devices not already initialized (in
XenbusStateInitWait state).
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Add support for debugging memory allocation statistics to xenstored.
Specifying "-M <file>" on the command line will enable the feature.
Whenever xenstored receives SIGUSR1 it will dump out a full talloc
report to <file>. This helps finding e.g. memory leaks in xenstored.
Signed-off-by: Juergen Gross <jgross@suse.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Use a temporary memory context for memory allocations when firing
watches. This will avoid leaking memory in case of long living
connections and/or xenstore entries.
This requires adding a new parameter to fire_watches() and add_event()
to specify the memory context to use for allocations.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Add a parameter to xenstored get_node() function to explicitly
specify the memory context to be used for allocations. This will make
it easier to avoid memory leaks by using a context which is freed
soon.
This requires adding the temporary context to errno_from_parents() and
ask_parents(), too.
When calling get_node() select a sensible memory context for the new
parameter by preferring a temporary one.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Add a parameter to xenstored read_node() function to explicitly
specify the memory context to be used for allocations. This will make
it easier to avoid memory leaks by using a context which is freed
soon.
When calling read_node() select a sensible memory context for the new
parameter by preferring a temporary one.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Add a parameter to xenstored get_parent() function to explicitly
specify the memory context to be used for allocations. This will make
it easier to avoid memory leaks by using a context which is freed
soon.
When available use a temporary context when calling get_parent(),
otherwise mimic the old behavior by calling get_parent() with the same
argument for both parameters.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
In order to be able to avoid leaving temporary memory allocated after
processing of a command in xenstored call all command functions with
the temporary "in" context. Each function can then make use of that
temporary context for allocating temporary memory instead of either
leaving that memory allocated until the connection is dropped (or
even until end of xenstored) or freeing the memory itself.
This requires to modify the interfaces of the functions taking only
one argument from the connection by moving the call of onearg() into
the single functions. Other than that no functional change.
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
The case in arch_set_info_guest() is a legitimate fallthrough. Mark it as such.
The cases in vlapic_accept_irq() are a terminal error path, but Coverity fails
to spot this. Reorder the comment to the end.
No functional change, but fixes two MISSING_BREAK Coverity defects.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
In fact, the data it protects only change either at init-time,
during cpupools manipulation, or when changing domains' weights.
In all other cases (namely, load balancing, reading weights
and status dumping), information is only read.
Therefore, let the lock be an read/write one. This means there
is no full serialization point for the whole scheduler and
for all the pCPUs of the host any longer.
This is particularly good for scalability (especially when doing
load balancing).
Also, update the high level description of the locking discipline,
and take the chance for rewording it a little bit (as well as
for adding a couple of locking related ASSERT()-s).
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
more specifically, with: TICKLE_NEW, RUNQ_MAX_WEIGHT,
MIGRATE, LOAD_CHECK, LOAD_BALANCE and PICKED_CPU, and
in both both xenalyze and formats (for xentrace_format).
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Mainly, almost all of the BUG_ON-s can be converted into
ASSERTS, and almost all the debug printk can either be
removed or turned into tracing.
The 'TODO' list, in a comment at the beginning of the file,
was also stale, so remove items that were still there but
are actually done.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
as all the accesses to both the masks and the flags are
serialized by the runqueues locks already.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Acked-by: George Dunlap <george.dunlap@citrix.com>
Add the shift used for the precision of the integer
arithmetic to the trace records, and update both xenalyze
and xentrace_format to make use of/print it.
In particular, in xenalyze, we are can now show the
load as a (easier to interpreet) percentage.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
The existing load tracking code was hard to understad and
maintain, and not entirely consistent. This is due to a
number of reasons:
- code and comments were not in perfect sync, making it
difficult to figure out what the intent of a particular
choice was (e.g., the choice of 18 for load_window_shift);
- the math, although effective, was not entirely consistent.
In fact, we were doing (if W is the lenght of the window):
avgload = (delta*load*W + (W - delta)*avgload)/W
avgload = avgload + delta*load - delta*avgload/W
which does not match any known variant of 'smoothing
moving average'. In fact, it should have been:
avgload = avgload + delta*load/W - delta*avgload/W
(for details on why, see the doc comments inside this
patch.). Furthermore, with
avgload ~= avgload + W*load - avgload
avgload ~= W*load
The reason why the formula above sort of worked was because
the number of bits used for the fractional parts of the
values used in fixed point math and the number of bits used
for the lenght of the window were the same (load_window_shift
was being used for both).
This may look handy, but it introduced a (not especially well
documented) dependency between the lenght of the window and
the precision of the calculations, which really should be
two independent things. Especially if treating them as such
(like it is done in this patch) does not lead to more
complex maths (same number of multiplications and shifts, and
there is still room for some optimization).
Therefore, in this patch, we:
- split length of the window and precision (and, since there
is already a command line parameter for length of window,
introduce one for precision too),
- align the math with one proper incarnation of exponential
smoothing (at no added cost),
- add comments, about the details of the algorithm and the
math used.
While there fix a couple of style issues as well (pointless
initialization, long lines, comments).
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
This really should not happen, but:
1. it does happen! Some more info here:
http://lists.xen.org/archives/html/xen-devel/2016-06/msg00922.html
2. independently from 1, it makes sense and is easy enough
to have a 'safety catch'.
The reason why this is particularly bad for Credit2 is that
negative values of delta mean out of scale high load (because
of the conversion to unsigned). This, for instance in the
case of runqueue load, results in a runqueue having its load
updated to values of the order of 10000% or so, which in turns
means that the load balancer will migrate everything off from
the pCPUs in the runqueue, and leave them idle until the load
gets back to something sane... which may indeed take a while!
This is not a fix for the problem of time going backwards. In
fact, if that happens a lot, load tracking accuracy is still
compromized, but at least the effect is a lot less bad than
before.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
In both Credit1 and Credit2, stop considering a pCPU idle,
if the reason why the idle vCPU is being selected, is to
do tasklet work.
Not doing so means that the tickling and load balancing
logic, seeing the pCPU as idle, considers it a candidate
for picking up vCPUs. But the pCPU won't actually pick
up or schedule any vCPU, which would then remain in the
runqueue, which is bad, especially if there were other,
truly idle pCPUs, that could execute it.
The only drawback is that we can't assume that a pCPU is
in always marked as idle when being removed from an
instance of the Credit2 scheduler (csched2_deinit_pdata).
In fact, if we are in stop-machine (i.e., during suspend
or shutdown), the pCPUs are running the stopmachine_tasklet
and hence are actually marked as busy. On the other hand,
when removing a pCPU from a Credit2 pool, it will indeed
be idle. The only thing we can do, therefore, is to
remove the BUG_ON() check.
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Reviewed-by: George Dunlap <george.dunlap@citrix.com>
ARM (<asm-arm/atomic.h>):
* add atomic_add_unless() wrapper over __atomic_add_unless()
(for common-code interface, i.e. <xen/atomic.h>)
X86 (<asm-x86/atomic.h>):
* implement missing functions atomic_{sub,inc,dec}_return(), atomic_add_unless()
* implement missing macro atomic_xchg()
COMMON (<xen/atomic.h>):
* add prototypes for the aforementioned newly implemented X86 functions in
common <xen/atomic.h>
Signed-off-by: Corneliu ZUZU <czuzu@bitdefender.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Julien Grall <julien.grall@arm.com>
Create a common-side <xen/atomic.h> to establish, among others, prototypes of
atomic functions called from common-code. Done to avoid introducing
inconsistencies between arch-side <asm/atomic.h> headers when we make subtle
changes to one of them. Some arm-side macros had to be turned into inline
functions in the process.
Removed outdated comment ("NB. I've [...]").
Signed-off-by: Corneliu ZUZU <czuzu@bitdefender.com>
Suggested-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: Julien Grall <julien.grall@arm.com>
Place atomic_inc_and_test() implementation after atomic_inc().
Also empty line fix.
Signed-off-by: Corneliu ZUZU <czuzu@bitdefender.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
and switch bool_t to being of type _Bool rather than char.
Using bool_t as char causes several subtle problems; first that a bool_t
actually has more than two values, and that (bool_t)0x100 actually has the
value 0 rather than the expected 1, due to truncation.
Making this change reveals two bugs now caught by the compiler.
errata_c6_eoi_workaround() actually makes use of bool_t having more than two
states, while generic_apic_probe() has a integer in the middle of a compound
bool_t assignment (which triggers a [-Werror=parentheses] warning on Debian
Jessie).
Finally, it turns out that ARM is mixing and matching bool_t and bool, despite
their different semantics. This change brings the semantics of bool_t to
match bool, but does not alter the current mix.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Julien Grall <julien.grall@arm.com>
Acked-by: Tim Deegan <tim@xen.org>
A subsequent change will introduce C99 bools, at which point 'bool'
becomes a type, and ineligible as a variable name.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
If Device-TLB flush timed out, we hide the target ATS device
immediately. By hiding the device, we make sure it can't be
assigned to any domain any longer (see device_assigned).
Signed-off-by: Quan Xu <quan.xu@intel.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
Tested-by: Quan Xu <quan.xu@intel.com>
Add domain crash logic to the generic IOMMU layer to benefit
all platforms.
No spamming of the log can occur. For DomU, we avoid logging any
message for already dying domains. For Dom0, that'll still be more
verbose than we'd really like, but it at least wouldn't outright
flood the console.
Signed-off-by: Quan Xu <quan.xu@intel.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Tested-by: Quan Xu <quan.xu@intel.com>
Do away with struct pci_ats_dev; integrate the few bits of information
in struct pci_dev (and as a result drop get_ats_device() altogether).
Hook ATS devices onto a linked list off of each IOMMU instead of on a
global one.
Signed-off-by: Quan Xu <quan.xu@intel.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
Tested-by: Quan Xu <quan.xu@intel.com>
Access to setpodtarget and getpodtarget is required by dom0 to set the balloon
targets for domU. The patch gives source domain (dom0) access to set
this target for domU and resolve the following permission denied erro
message during ballooning :
avc: denied { setpodtarget } for domid=0 target=9
scontext=system_u:system_r:dom0_t
tcontext=system_u:system_r:domU_t tclass=domain
Signed-off-by: Anshul Makkar <anshul.makkar@citrix.com>
Acked-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
This adds a Kconfig option and support for including the XSM policy from
tools/flask/policy in the hypervisor so that the bootloader does not
need to provide a policy to get sane behavior from an XSM-enabled
hypervisor. The policy provided by the bootloader, if present, will
override the built-in policy.
The XSM policy is not moved out of tools because that remains the
primary location for installing and configuring the policy.
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
This makes the buffers function parameters instead of globals, in
preparation for adding alternate locations for the policy.
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Separate the code logic that does the registration of vgic_v3/v2 ops
to a new function domain_vgic_register(). The intention of this
separation is to record the required mmio count in vgic_v3/v2_init()
and pass it to function domain_io_init() in a follow-up patch patch.
Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
Reviewed-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
The macro MAX_RDIST_COUNT is not being used after converting code
to handle number of redistributor dynamically. So remove it from
header file and the two other panic() messages that are not valid
anymore.
Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
Acked-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
The number of Redistributor regions allowed for dom0 is hardcoded
to a define MAX_RDIST_COUNT which is 4. Some systems, especially
latest server chips, may have more than 4 redistributors. Either we
have to increase MAX_RDIST_COUNT to a bigger number or allocate
memory based on the number of redistributors that are found in MADT
table. In the worst case scenario, the macro MAX_RDIST_COUNT should
be equal to CONFIG_NR_CPUS in order to support per CPU Redistributors.
Increasing MAX_RDIST_COUNT has a effect, it blows 'struct domain'
size and hits BUILD_BUG_ON() in domain build code path.
struct domain *alloc_domain_struct(void)
{
struct domain *d;
BUILD_BUG_ON(sizeof(*d) > PAGE_SIZE);
d = alloc_xenheap_pages(0, 0);
if ( d == NULL )
return NULL;
...
This patch uses the second approach to fix the BUILD_BUG().
Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
Reviewed-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
The redistributor address can be specified either as part of GICC or
GICR subtable depending on the power domain. The current driver
doesn't support parsing redistributor entry that is defined in GICC
subtable. The GIC CPU subtable entry holds the associated Redistributor
base address if it is not on always-on power domain.
The per CPU Redistributor size is not defined in ACPI specification.
Set the GICR region size to SZ_256K if the GIC hardware is capable of
Direct Virtual LPI Injection feature, SZ_128K otherwise.
This patch adds necessary code to handle both types of Redistributors
base addresses.
Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
Acked-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Add a new function to parse GICR subtable and move the code that
is specific to GICR table to a new function without changing the
function gicv3_acpi_init() behavior.
Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
Acked-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
For ACPI based XEN boot, the GICD region needs to be accessed inside
the function gicv3_acpi_init() in later patch. There is a duplicate
panic() message, one in the DTS probe and second one in the ACPI probe
path. For these two reasons, move the code that validates the GICD base
address and does the region ioremap to a separate function. The
following patch accesses the GICD region inside gicv3_acpi_init() for
finding per CPU Redistributor size.
Signed-off-by: Shanker Donthineni <shankerd@codeaurora.org>
Acked-by: Julien Grall <julien.grall@arm.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
Wei Liu