Provide libxl_ao_abort.
There is machinery to allow an ao to register an interest in abort
requests, using a libxl__ao_abortable.
This API is not currently very functional: requesting abort will
never have any effect.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
---
v5: Clarify API doc comment re aborting synchronous operations.
v4: Rename from cancel to abort.
Actually record aos on aos_inprogress.
(Report from Koushik Chakravarty at Citrix.)
Do not mark libxl_ao_cancel hidden (!)
Abolish ERROR_NOTIMPLEMENTED from libxl_ao_cancel.
All operations are supposed to support cancellation.
v2: Minor comment improvements
Previously we used in_initiator to stop the ao being freed while we
were still in the initiator function (which would result in the
initiator's call to libxl__ao_inprogress accessing the ao after it had
been freed).
We are going to introduce a new libxl entrypoint which finds, and
operates on, ongoing aos. This function needs the same protection,
and might even end up running on the same ao multiple times
concurrently.
So do this with reference counting instead, with a new variable
ao->manip_refcnt.
We keep ao->in_initiator because that allows us to keep some useful
asserts about the sequencing of libxl__ao_inprogress, etc.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v5: Fix typo in commit message.
v3: Add a missing space.
Mention locking in the comment.
If we aborted the domain creation, do not try to tear it down again
Document this.
This is a backwards-compatible API change since old libxl users will
never abort any operations.
In the current code, there is no functional change, because
ERROR_ABORTED is never generated anywhere yet.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v4: ABORTED not CANCELLED.
Always write out guest domid on completion.
Do not trash rc with libxl__set_domain_configuration on preserve path.
Make spawn provide an rc to its caller, and either pass it through
from the timeout callback, or invent ERROR_FAIL, as applicable.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
---
v4: New patch in this version of the series.
The callback functions provided by users must take an rc value. This
rc value can be ERROR_TIMEDOUT or ERROR_ABORTED.
Users of xswait are now expected to deal correctly with
ERROR_ABORTED. If they experience this, it hasn't been logged.
And the caller won't log it either since it's not TIMEDOUT.
Luckily this is correct, so we can just change the doc comment.
Currently nothing generates ERROR_ABORTED; in particular the timeouts
cannot in fact signal abort requests.
There should be no publicly visible change except that some error
returns from libxl will change from ERROR_FAIL to ERROR_TIMEDOUT, and
some changes to debugging messages.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v4: ABORTED not CANCELLED.
The internal user of libxl__async_exec_start et al now gets an rc as
well as the process's exit status.
For now this is always either 0 or ERROR_FAIL, but with ao
abort requests this will possibly be ABORTED or TIMEDOUT too.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v4: Improve doc comment as suggested by Ian C.
v2: New patch due to rebause; v1 had changes to device_hotplug_*
scripts instead.
Callback now gets unambiguous information about error situation:
previously, if only thing that went wrong was that child died
badly, rc would be FAILED, which was unambigously; now rc=0.
Add a comment document the meaning of the rc and status parameters
to the callback.
Change the timeout setup functions to take a libxl__ao, not a
libxl__gc. This is going to be needed for ao abort, because timeouts
are going to be a main hook for ao abort requests - so the timeouts
need to be associated with an ao.
This means that timeouts can only occur as part of a long-running
libxl function (but this is of course correct, as libxl shouldn't have
any global timeouts, and indeed all the call sites have an ao).
Also remove the gc parameter from libxl__async_exec_start. It can
just use the gc from the ao supplied in the aes.
All the callers follow the obvious patterns and therefore supply the
ao's gc to libxl__async_exec_start and the timeout setup functions.
There is therefore no functional change in this patch.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
CC: Yang Hongyang <yanghy@cn.fujitsu.com>
CC: Wen Congyang <wency@cn.fujitsu.com>
CC: Lai Jiangshan <laijs@cn.fujitsu.com>
Acked-by: Wen Congyang <wency@cn.fujitsu.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v2: This patch split off from "Permit timeouts to signal cancellation".
Rebased; consequently, deal with libxl__async_exec_start.
CC'd authors of the libxl__async_exec_* functions.
We introduce ERROR_ABORTED now, so that we can write code to handle
it, and decreee that functions might return it, even though currently
there is nowhere where this error is generated.
While we're here, provide ERROR_NOTFOUND which will also be used
later, but only as part of the public API.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
---
v4: CANCELLED renamed to ABORTED.
No longer introduce ERROR_NOTIMPLEMENTED.
v2: Rebase means new errors have bigger (more negative) numbers.
We are going to introduce application-requested aborts of (ao)
operations, but these suspend failures are something different.
Reword to avoid confusion.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
We are going to introduce a new meaning for aborting an ao, so rename
this to avoid confusion.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Replace the separate timeout and xenstore watch with use of
libxl__xswait*.
Different control flow, but no ultimate functional change apart from
slight changes to the text of error messages.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
If a suspend callback fails, it has a libxl error code in its hand.
However we must return to libxc the values that libxc expects. So we
stash the libxl error code in dss->rc and fish it out again after
libxc returns from the suspend call.
While we're here, abolish the now-redundant `ok' variable in
remus_devices_postsuspend_cb.
The overall functional change is that libxl_domain_save now completes
with the correct error code as determined when the underlying failure
happened. (Usually this is, still, ERROR_FAIL.)
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v2: Add cleanup in remus_devices_postsuspend_cb.
Change the following functions to take a libxl error code rather than
a boolean "ok" value, and translate that value to the boolean expected
by libxc at the last moment:
domain_suspend_callback_common_done } dss->callback_common_done
remus_domain_suspend_callback_common_done }
domain_suspend_common_done
Also, abolish domain_suspend_common_failed as
domain_suspend_common_done can easily do its job and the call sites
now have to supply the right rc value anyway.
In domain_suspend_common_guest_suspended, change "ret" to "rc"
as it contains a libxl error code.
There is no functional change in this patch: the proper rc value now
propagates further, but is still eventually smashed to a boolean.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
---
v2: Fix a leftover comment referring to domain_suspend_common_failed
switch_logdirty_done used to take the value to pass to
libxl__xc_domain_saverestore_async_callback_done (ie, the return value
from the callback). (This was mistakenly described as "ok" in the
prototype, but in the definition it is "broke" and all the call sites
passed 0 for success or -1 for error.)
Instead, make it take a libxl error code (rc). Convert this to the
suspend callback value at the end.
No functional change in this patch.
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
This doc comment about ao lifecycle failed to mention the option of
completing the ao during the initiator function. (Indeed, the most
obvious reading would forbid it.)
Restructure the comment, describe this situation, and generally
improve the wording.
Also, fix a grammar problem (missing word `a').
Reported-by: Koushik Chakravarty <koushik.chakravarty@citrix.com>
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
There is an issue in libxl_set_memory_target whereby the target and
the max mem can get out of sync, this is because the call the
xc_domain_setmaxmem is not tied in any way to the xenstore transaction
which controls updates to the xenstore side of things.
Consider a domain with 1M of RAM (==target and maxmem for the sake of
argument) and two simultaneous calls to libxl_set_memory_target, both
with relative=0 and enforce=1, one with target=3 and the other with
target=5.
target=5 call target=3 call
transaction start
transaction start
write target=5 to xenstore
write target=3 to xenstore
setmaxmem(5)
setmaxmem(3)
transaction commit = success
transaction commit = EAGAIN
At this point maxmem=3 while target=5.
In reality the target=3 case will the retry and eventually (hopefully)
succeed with target=maxmem=3, however the bad state will persist for
some window which is undesirable. On failure other than EAGAIN all
bets are off anyway, but in that case we will likely stick in the bad
state until someone else sets the memory).
To fix this we slightly abuse the userdata lock which is used to
protect updates to the domain's json configuration. Abused because
maxmem is not actually stored in there, but is kept by Xen. However
the lock protects some semantically similar things and is convenient
to use here too.
libxl_domain_setmaxmem also takes the lock, since it reads
memory/target from xenstore before calling xc_domain_setmaxmem there
is a small (but perhaps not very interesting) race there too.
There is on more use of xc_domain_setmaxmem in libxl__build_pre.
However taking a lock around this would be tricky since the xenstore
parts are not done until libxl__build_post. I think this one could be
argued to be OK since the domid is not "public" yet, that is it has
not been returned to the application yet (as the result of the create
operation). Toolstacks which go round fiddling with random domid's
which they find lying on the floor should be taught to do better.
Add a doc note that taking the userdata lock requires the CTX_LOCK to
be held.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Xen's build system has a target for rump kernel called NetBSDRump. We
want to build libxc against rump kernel, so we need to copy NetBSD's
evtchn.h and privcmd.h to NetBSDRump. These copies is not very likely to
diverge from NetBSD's copies, but we don't preclude such possibility.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
This reverts commit 0c029c4da2.
A new memory model that allows QEMU to bump memory behind libxl's back
was merged a few months ago. We didn't fully understand the
repercussions back then. Now it breaks migration and becomes blocker of
4.6 release.
It's better to restore to original behaviour at this stage of the
release cycle, that would put us in a position no worse than before, so
the release is unblocked.
The said function is still racy after reverting these two patches.
Making domain memory state consistent requires a bit more work. Separate
patch(es) will be sent out to deal with that problem.
Fix up conflicts with f5b43e95 (libxl: fix "xl mem-set" regression from
0c029c4da2).
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
When the property "clock-frequency" is present in the DT timer node, it
means that the bootloader/firmware didn't correctly configure the
CNTFRQ/CNTFRQ_EL0 on each processor.
The best solution would be to fix the offending firmware/bootloader,
although it may not always be possible to modify and re-flash it.
As it's not possible to trap the register CNTFRQ/CNTFRQ_EL0, we have
to extend xen_arch_domainconfig to provide the timer frequency to the
toolstack when the property "clock-frequency" is present to the host DT
timer node. Then, a property "clock-frequency" will be created in the guest
DT timer node if the value is not 0.
We could have set the property in the guest DT no matter if the property
is present in the host DT. Although, we still want to let the guest
using CNTFRQ in normal case. After all, the property "clock-frequency"
is just a workaround for buggy firmware.
Also add a stub for fdt_property_u32 which is not present in libfdt <
1.4.0 used by distribution such as Debian Wheezy.
Signed-off-by: Julien Grall <julien.grall@citrix.com>
Tested-by: Chris Brand <chris.brand@broadcom.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
[ ijc -- ran autogen.sh ]
The struct just contains three methods and no data, so the name
hvm_mmio_ops more accurately reflects its content. A subsequent patch
introduces a new structure which more accurately warrants the name
hvm_mmio_handler so doing the rename in this purely cosmetic patch avoids
conflating functional and cosmetic changes in a single patch.
Signed-off-by: Paul Durrant <paul.durrant@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.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>
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>
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 ]
As suggested by Andrew Cooper, this patch attempts to remove
some redundancy and allow for an easier time when adding vm_events
for new control registers in the future, by having a single
VM_EVENT_REASON_WRITE_CTRLREG vm_event type, meant to serve CR0,
CR3, CR4 and (newly introduced) XCR0. The actual control register
will be deduced by the new .index field in vm_event_write_ctrlreg
(renamed from vm_event_mov_to_cr).
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Kevin Tian <kevin.tian@intel.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Tim Deegan <tim@xen.org>
binutils 2.22 changed ld default from --copy-dt-needed-entries
to -no-copy-dt-needed-entries. This revealed that some objects
are linked implicitly with libtinfo and newer ld fails to build
relevant executables.
Below is short explanation why we should not do that...
http://fedoraproject.org/wiki/UnderstandingDSOLinkChange says:
The default behaviour for ld (my note: before version 2.22) allows
users to 'indirectly' link to required objects/libraries through
intermediate objects/libraries. While this is convenient, it can
also be dangerous because it makes your program's dependencies tied
to the dependencies of other objects. If those objects ever change
their linkages, they can break your program without any changes
to your own code!
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
The snippet to destroy stubdom and the callback were added in 1fc3aeb3
("libxl: use new QEMU xenstore protocol"). The intention was to destroy
stubdom when it is not responsive. That approach is problematic because
rc is not propagate back to sdss->callback, hence the guest is leaked.
The solution is simple. The destruction of stubdom can be done later in
sdss->callback. That code path already does the right thing to destroy
both the guest and the stubdom that serves the guest.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Campbell <ian.campbell@citrix.com>
Ian Jackson
Ian Campbell
Wei Liu
Roger Pau Monne
Julien Grall
Paul Durrant
Jan Beulich
Ross Lagerwall
Olaf Hering
Don Slutz
Andrew Cooper
Razvan Cojocaru
Daniel Kiper