...to prepare for a consistent "pci_assignable_*" naming scheme.
Also move the man page entry into the PCI PASS-THROUGH section, rather
than the XEN HOST section.
No functional changes.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
Many places in xl there's a variable or element named "filename" which
does not contain a filename, but the source of the data for reporting
purposes. Worse, there are variables which are sometimes actually
used or interpreted as a filename depending on what other variables
are set. This makes it difficult to tell when a string is purely
cosmetic, and when another bit of code may actually attempt to call
"open" with the string.
This patch makes a consistent distinction between "filename" (which
always refers to the name of an actual file, and may be interpreted as
such at some point) and "source" (which may be a filename, or may be
another data source such as a migration stream or saved data).
This does add some variables and reshuffle where assignments happen;
most notably, the "restore_filename" element of struct domain_create
is now only set when restoring from a file.
But at a high level, there should be no funcitonal changes.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
The "filename" is a bit of a misnomer, as it's only used during error
messages, and in most instances cases is actually set to "command
line".
Rename it to "config_source" to make it clear that it's not used to
actually open any files.
No functional changes.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
pygrub and xenpvnetboot are meant to be run by tools, and not by the user,
and thus should be in /usr/lib/xen/bin rather than /usr/bin.
Because most config files will still have an absolute path pointing to
/usr/bin/pygrub, make a symbolic link that we will deprecate.
Signed-off-by: George Dunlap <george.dunlap@eu.ctirix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
If the full path for a bootloader (such as pygrub or xenpvnetboot) is not
given, check for it first in the libexec path before falling back to the
PATH variable.
Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
xm supports the following syntax (in the config file) for
specific VCPU to PCPU mapping:
cpus = "0-3,5,^1" # all vcpus run on cpus 0,2,3,5
cpus = ["2", "3"] # VCPU0 runs on CPU2, VCPU1 runs on CPU3
Allow for the same in xl.
This fixes what happened in changeset 54000bca7a6a, which
introduced suppot for the `cpus=` option within xl, but used
both the list (cpus=[2, 3]) and the string (cpus="2,3") syntax
for achieving the same behaviour (pin all guest's vcpus to the
pcpus in the list/string).
Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com>
Committed-by: Ian Campbell <ian.campbell@citrix.com>
The debugging code added in c/s 24707:96987c324a4f was hit a (small)
number of times (one report being
http://lists.xen.org/archives/html/xen-devel/2012-05/msg00332.html),
apparently always with a vector within the legacy range. Obviously,
besides legacy vectors not normally expected to be in use on systems
with IO-APIC(s), they should never make it to the IRQ migration logic.
This wasn't being prevented so far: Since we don't have a one-to-one
mapping between vectors and IRQs - legacy IRQs may have two vectors
associated with them (one used in either 8259A, the other used in one
of the IO-APICs) -, vector-to-IRQ translations for legacy vectors (as
used in do_IRQ()) would yield a valid IRQ number despite the IRQ
really being handled via an IO-APIC.
This gets changed here - disable_8259A_irq() zaps the legacy vector-to-
IRQ mapping, and enable_8259A_irq(), should it ever be called for a
particular interrupts, restores it.
The spurious interrupt logic in do_IRQ() gets adjusted too: Interrupts
coming in via legacy vectors presumably didn't get reported through the
IO-APIC/LAPIC pair (as we never program these vectors into any RTE or
LVT). Call ack_APIC_irq() only when the LAPIC's ISR bit says an
interrupt is pending at the given vector. Plus, a new function (pointer)
bogus_8259A_irq() gets used to have the 8259A driver take care of the
bogus interrupt (as outside of automatic EOI mode it may need an EOI to
be issued for it to prevent other interrupts legitimately going through
the 8259As from getting masked out).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Keir Fraser <keir@xen.org>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
On systems that have an IO-APIC, we generally run the PIC in AEOI
mode, yet i8259A_resume() so far failed to put it back into that mode.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Keir Fraser <keir@xen.org>
The xenstore header xs.h is producing conflicts with other software[1].
xs is a too short identifier and does not matche the library. Renaming
the headers to xenstore.h and xenstore_lib.h is the easiest way to make
them easy recognizable and prevent furthe problems.
[1]: http://bugs.debian.org/668550
[ Also update QEMU_TAG, to bring in corresponding change to
qemu-xen-traditional. -iwj ]
Signed-off-by: Bastian Blank <waldi@debian.org>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
--HG--
rename : tools/xenstore/xs.h => tools/xenstore/xenstore.h
rename : tools/xenstore/xs_lib.h => tools/xenstore/xenstore_lib.h
The tracepoint for emulated MMIO and I/O port reads was always before
the emulated read or write was done. This means that for reads the
register value in the trace record was always 0.
So for reads, move the tracepoint until the register value is
available.
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Committed-by: Keir Fraser <keir@xen.org>
Legacy (non-pvops) gntdev drivers may require this operation to be
performed when the number of grants intended to be used simultaneously
exceeds a certain driver specific default limit, and qemu's qdisk
driver is an example of needing to do so.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Prevent xl from doing any operation if xend daemon is running. That
prevents bugs that happened when xl and xend raced to close a domain.
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson@eu.citrix.com>
Without this, dump_irqs() may try to acquire the lock the caller is
currently holding.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Allow compilation of PVonHVM drivers with forward-ported xenlinux
sources in openSuSE 12.2. Since Linux 3.4 asm/system.h is not present
anymore. Remove inclusion of this header, its not needed.
Signed-off-by: Olaf Hering <olaf@aepfle.de>
Committed-by: Jan Beulich <jbeulich@suse.com>
Cancel the bootloader (specifically, by sending it a signal) if the
domain is seen to disappear from xenstore.
We use a new utility event source libxl__domaindeathcheck which
provides a convenient wrapper for the xenstore watch.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v8:
* Fixed the commit message summary line.
Changes since v7:
* Add a comment explaining why we use a watch on the domain's
xenstore path rather than @releaseDomain.
* Fix typo in error message.
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Abolish libxl_fork. Its only callers were in xl. Its functionality
is now moved elsewhere, as follows:
* The "logging version of fork", which is what it was billed as, is now
xl_fork, which also dies on failure.
* Closing the xenstore handle in the child is now done in
libxl__ev_child_fork, which is the only remaining place where fork
is called in libxl.
* We provide a new function libxl__ev_child_xenstore_reopen for
in-libxl children to make the ctx useable for xenstore again.
* Consequently libxl__spawn_record_pid now no longer needs to mess
about with its own xenstore handle. As a bonus it can now just use
libxl__xs_write.
Also, since we have now converted all the forkers in libxl, we can
always honour the fork_replacement childproc hook - so do so.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Remove the old console callback. (Its reentrancy properties were
troublesome: in principle, the event loop might be reentered during
the callback, and the libxl__domain_create_state swept out from under
the feet of the domain creation.)
As a side effect of the new code arrangements, the console callback
for non-bootloader-using PV guests now occurs near the end of domain
creation, in the same place as for HVM guests, rather than near the
start.
The new arrangements should in principle mean that by the time the
console is described as ready by the callback, the xenstore key is
indeed ready. So in the future the timeout might be removed from
the console client.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
This will be used for reporting, during domain creation, that the
console is ready.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v7:
* If aop->how.callback, actually add the aop to the for_callback list (!)
* Document the threadsafety of aop's, and make appropriate cross-references.
* Allocate the actual aop from its thread's egc; do not free it.
* Remove pointless code motion of libxl__ao_create.
* Minor formatting fixes.
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Document the concurrent access policies for libxl__ao and libxl__egc,
and their corresponding gcs.
Fix a violation of the policy:
If an ao was submitted and a callback requested, and while the
initiating function was still running on the original thread, the ao
is completed on another thread, the completing thread would improperly
concurrently access the ao with the initiating thread.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Change initiate_domain_create to properly use libxl__bootloader_run
rather than improperly calling libxl_run_bootloader with NULL ao_how.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v7:
* constify convenience aliases.
Changes since v6:
* Bugfixes from testing.
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
This is needed in order to perform the domain destruction if
libxl__domain_make fails.
Also move doc comment about error behaviour of libxl__domain_make from
implementation to declaration.
Signed-off-by: Roger Pau Monne <roger.pau@citrix.com>
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
libxl__spawn_spawn becomes a callback-style asynchronous function.
The implementation is now in terms of libxl__ev_* including
libxl_ev_child.
All the callers need to be updated. This includes the device model
spawning functions libxl__create_device_model and
libxl__create_stubdom; these are replaced with libxl__spawn_local_dm
and libxl__spawn_stubdom. libxl__confirm_device_model_startup is
abolished; instead the dm spawner calls back.
(The choice of which kind of device model to create is lifted out of
what used to be libxl__create_device_model, because that function was
indirectly recursive. Recursive callback-style operations are clumsy
because they require a pointer indirection for the nested states.)
Waiting for proper device model startup it is no longer notionally
optional. Previously the code appeared to tolerate this by passing
NULL for various libxl__spawner_starting* parameters to device model
spawners. However, this was not used anywhere.
Conversely, the "for_spawn" parameter to libxl__wait_for_offspring is
no longer supported. It remains as an unused formal parameter to
avoid updating, in this patch, all the call sites which pass NULL.
libxl__wait_for_offspring is in any case itself an obsolete function,
so this wrinkle will go away when its callers are updated to use the
event system. Consequently libxl__spawn_check is also abolished.
The "console ready" callback also remains unchanged in this patch.
The API for this needs to be reviewed in the context of the event
series and its reentrancy restrictions documented.
Thus their callers need to be updated. These are the domain creation
functions libxl_domain_create_new and _restore. These functions now
take ao_hows, and have a private state structure.
However domain creation remains not completely converted to the event
mechanism; in particular it runs the outward-facing function
libxl_run_bootloader with a NULL ao_how, which is quite wrong. As it
happens in the current code this is not a bug because none of the rest
of the functionality surrounding the bootloader call will mind if the
event loop is reentered in the middle of its execution.
The file-scope function libxl__set_fd_flag which was used by the
previous spawn arrangements becomes unused and is removed; other
places in libxl can use libxl_fd_set_nonblock and
libxl_fd_set_cloexec, which of course remain.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v8:
* Make midproc_cb callback with correct pid (that of the grandchild,
not "middle" which is zero). Reported by Roger Pau Monne.
Changes since v7:
* Rename libxl__spawn_stubdom to libxl__spawn_stub_dm (and ..._state);
rename the state's stubdom_* members to dm_*.
* Eliminate the union between the two dm creation states in
libxl__domain_create_state. Instead, the domain creation code
simply uses libxl__stub_dm_spawn_state.dm directly, if we're
taking the local dm path.
* Remove a spurious "break".
* In domain creation, move the PV non-qemu case into the switch.
* Code style fixes.
* Constify some convenience aliases.
* Improve comments (including typo fixes).
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
In the next patch these variables will turn into actual pointers.
To clarify that patch, prepare the ground by changing these variables
from "struct foo var" to "struct foo var[1]". This enables accesses
to them and their members to be made as if they were pointers.
No functional change.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Remove this obsolete hook. Callers inside libxl which create and reap
children should use the mechanisms provided by the event system.
(This has no functional difference since there is no way for
ctx->waitpid_instead ever to become set.)
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Although this function is not currently slow it may become so in the
future (this also depends somewhat on the size of the guest).
Therefore arrange for it to take an ao_how which it completes
immediately. This will allow us to make it asynchronous in the future
without breaking API compatibility.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Mark the gc produced by AO_GC and EGC_GC with the gcc feature
__attribute__((unused)). This allows the use of EGC_INIT and
STATE_AO_GC by functions which do actually use the gc.
This is convenient because those functions might want to use the ao or
egc, rather than the gc; and also because it means that functions
which morally ought to be fishing any gc they use out of an egc or
state structure can be written do so regardless of whether the gc is
actually used right then.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
This involves adding a new log feature to libxl__datacopier, and then
using it.
If the bootloader exits nonzero we print the log filename in a log
message from libxl.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Convert libxl_run_bootloader to an ao_how-taking function.
It's implemented in terms of libxl__bootloader_run, which can be used
internally. The resulting code is pretty much a rewrite.
Significant changes include:
- We direct the bootloader's results to a file, not a pipe. This
makes it simpler to deal with as we don't have to read it
concurrently along with everything else.
- We now issue a warning if we find unexpected statements in the
bootloader results.
- The arrangements for buffering of bootloader input and output
are completely changed. Now we have a fixed limit of 64k
on output, and 4k on input, and discard the oldest data when
this overflows (which it shouldn't). There is no timeout
any more.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v6:
* Use libxl__ev_child_inuse rather than testing pid directly.
* Fix a code style error.
* Properly initialise the sub-operations' aos in _init.
* Bugfixes.
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
General facility for ao operations to shovel data between fds.
This will be used by the bootloader machinery.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes in v7:
* assert that the ao is non-null on _init.
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Provide functions for sending and receiving fds.
There used to be fd-sending functionality in libxl_qmp.c before
25298:84ae90427c54, which this implementation is partially derived
from.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
These utility functions cope with EINTR AND ENOENT, do error logging,
and we provide a recursive version to delete whole directory trees.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
We may need to #include <libutil.h>, and/or link with -lutil, to use
openpty, login_tty, and the like. Provide INCLUDE_LIBUTIL_H
(preprocessor constant, not always defined) and PTYFUNCS_LIBS
(makefile variable).
We link libxl against PTYFUNCS_LIBS (which comes from autoconf) rather
than UTIL_LIBS, and #include <libutil.h> where appropriate.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v7:
* Actually include the call to AX_CHECK_PTYFUNCS in this patch,
not the previous one, and regenerate configure accordingly.
Changes since v6:
* Put failure macro call in correct place so it might actually happen.
* Try both with -lutil and without.
* Patch now contains update for config.h.in.
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Remove a lot of unnecessary tests. Specifically, we no longer test
for standard POSIX facilities which we expect to be provided
everywhere and which we don't in any case have any alternative for.
Switch to generating config.h.in with autoheader.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Roger Pau Monne <roger.pau@entel.upc.edu>
Changes since v7:
* Removed AX_CHECK_PTYFUNCS (snuck in from previous patch)
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
SIGCHLD or not; if it does, it has to tell us about deaths of our
children.
Currently there are no callers in libxl which use these facilities.
All code in libxl which forks needs to be converted and libxl_fork
needse to be be abolished.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
We need a slightly more sophisticated data structure to allow this,
where we record the slot not just for each fd but also for each
(fd,eventbit) where eventbit is POLLIN, POLLPRI, POLLOUT.
Document the new relaxed restriction.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Changes since v6:
* Fix typo
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Pass POLLERR and POLLHUP to fd callbacks, as is necessary.
Crash on POLLNVAL since that means our fds are messed up.
Document the behaviour (including the fact that poll sometimes sets
POLLHUP or POLLERR even if only POLLIN was requested.
Fix the one current fd callback to do something with POLLERR|POLLHUP.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Committed-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Following the recent changes to upstream Qemu, the best monitor command
to suit or needs is "xen-save-devices-state" rather than "migrate".
This patch removes libxl__qmp_migrate and introduces libxl__qmp_save
instead, that uses "xen-save-devices-state".
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
Read Qemu's physmap from xenstore and save it using toolstack_save.
Restore Qemu's physmap using toolstack_restore.
Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Committed-by: Ian Jackson <ian.jackson.citrix.com>
George Dunlap
Goncalo Gomes
Dario Faggioli
Jan Beulich
Ian Campbell
David Vrabel
Roger Pau Monne
Olaf Hering
Ian Jackson
Keir Fraser
Stefano Stabellini