We don't need to specify /bin/bash in the entry point rune, otherwise
non-interactive invocation of the container would fail with something
like:
+ C=debian:stretch-i386
+ export CONTAINER=registry.gitlab.com/xen-project/xen/debian:stretch-i386
+ CONTAINER=registry.gitlab.com/xen-project/xen/debian:stretch-i386
+ cd /local/work/COMMITTER/xen-32.git
+ git fetch origin
+ con git reset --hard origin/staging
*** Ensuring registry.gitlab.com/xen-project/xen/debian:stretch-i386 is up to date
*** Launching container ...
/usr/bin/git: /usr/bin/git: cannot execute binary file
While at it, use shorthand "linux32".
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Doug Goldstein <cardoe@cardoe.com>
This is not really very nice.
This line doesn't have much value in itself. The rest of this comment
block is pretty clear what it wants to convey. So delete it.
(While we are here, adopt the CODING_STYLE-mandated formatting.)
Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Lars Kurth <lars.kurth.xen@gmail.com>
Acked-by: George Dunlap <dunlapg@umich.edu
Acked-by: Jan Beulich <JBeulich@suse.com>
---
v3: Restore erroneously-dropped tab.
v2: Delete the comment entirely.
With '|'. The result is the same, and the later works with lld. Fixes
the following error when building Xen with lld:
ld -melf_x86_64_fbsd -T xen.lds -N prelink.o --build-id=sha1 \
/root/src/xen/xen/common/symbols-dummy.o -o /root/src/xen/xen/.xen-syms.0
ld: error: xen.lds:260: malformed number: |
>>> ASSERT(__image_base__ > (((((((((261 >> 8) * 0xffff000000000000) | (261 << 39))) + ((1 << 39) / 2)) + (64 << 30)) + (1 << 30)) + (1 << 30))) ||
>>> ^
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
This isn't as much of an optimization than to avoid triggering a gcc bug
affecting 5.x ... 7.x, triggered by any asm() put inside the ad hoc
"rewalk" loop and taking as an (output?) operand a register variable
tied to %rdx (an "rdx" clobber is fine). The issue is due to an apparent
collision in register use with the modulo operation in vtlb_hash(),
which (with optimization enabled) involves a multiplication of two
64-bit values with the upper half (in %rdx) of the 128-bit result being
of interest.
Such an asm() was originally meant to be implicitly introduced into the
code when converting most indirect calls through the hvm_funcs table to
direct calls (via alternative instruction patching); that model was
switched to clobbers due to further compiler problems, but I think the
change here is worthwhile nevertheless.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Tim Deegan <tim@xen.org>
In the shim case, the number of CPUs should be solely controlled by the
guest configuration file. Make sure the command line options are fully
(and not just partially) ignored.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Currently it's possible to hit an assertion in alloc_heap_pages():
Assertion 'first_dirty != INVALID_DIRTY_IDX || !(pg[i].count_info & PGC_need_scrub)' failed at page_alloc.c:988
This can happen because a piece of logic to calculate first_dirty
during block merging in free_heap_pages() is missing for the following
scenario:
1. Current block's first_dirty equals to INVALID_DIRTY_IDX
2. Successor block is free but its first_dirty != INVALID_DIRTY_IDX
3. The successor is merged into current block
4. Current block's first_dirty still equals to INVALID_DIRTY_IDX
This will trigger the assertion during allocation of such block in
alloc_heap_pages() because there will be pages with PGC_need_scrub
bit set despite the claim of first_dirty that the block is scrubbed.
Add the missing piece of logic and slightly update the comment for
the predecessor case to better capture the code's intent.
Fixes 1a37f33ea6 ("mm: Place unscrubbed pages at the end of pagelist")
Signed-off-by: Sergey Dyasli <sergey.dyasli@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Some of the functions implemented in setup.c are only used at boot but
not yet marked as such.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Signed-off-by: Stefano Stabellini <sstabellini@kernel.org>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Libfdt is used for:
- Unflatten the Flatten Device-Tree (FDT) blob
- Create Device-Tree for the Hardware-Domain
Both use are done during the initialization of Xen. So move all the
libfdt to init.
Note that the runes was borrowed from libelf Makefile.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Now that ELF support has been dropped to boot Dom0, no-one is using
libelf within the hypervisor.
Introduce a config option to select libelf on x86 and keep unselected
for Arm.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Acked-by: Jan Beulich <jbeulich@suse.com>
The code has been around since the beginning of Xen Arm. However, I am
not aware of any user and the code is pretty bogus:
1) It is assuming virtual address == physical address.
2) The cache is not cleaned after the Image is loaded but the Image
is started with Cache disabled.
3) There are not clear ABI with the guest.
Xen is currently supporting 3 other formats (zImage, Image, U-boot Image)
as well as gzip compressed version of each formats. All of them are well
documented and widely use.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
The file kernel.c only contains code/data used during the
initialization. So move everything to init and mark the file as such.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
The file domain_build.c only contains code/data used during the
initialization. So move everything to init and mark the file as such.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Dom0 (aka hardware domain on Arm) is always direct mapped. Rather than
using a global variable to store a const, directly open-code it or
replace the use with is_domain_direct_mapped(...) macros.
This will also help a follow-up patch to move all domain_build.c in
init.
Signed-off-by: Julien Grall <julien.grall@arm.com>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
a75703b2f0 unconditionally enabled
building of rombios and stubdoms but unfortunately these two pieces do
not build successfully with clang. rombios unconditionally depends on
the build of iPXE and upstream iPXE does not support building with
clang. Similiarly the build of the stubdoms depends on the build of
newlib which does not support being built with clang.
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
xencall_alloc_buffer() is used throughout Xen tools for allocating
hypercall buffers. Allocation is done at page granularity. For simple
administration each allocated set of pages contains a small header
holding the number of pages of that set. The hypercall buffer is
located directly after the 4 byte sized header, leading to a wrong
alignment for e.g. pointers.
Repair that by using a 16 byte sized header enforcing the same
alignment as malloc().
Signed-off-by: Juergen Gross <jgross@suse.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Commit message stolen from a reply from Ian:
The git protocol is not just unencrypted, but also unauthenticated.
In theory it is possible to verify the signed tags for actual
releases, but that is a cumbersome process which I very much doubt
anyone really does.
As for the various branch tips, there is currently no way (unless you
have a shell account on xenbits) to get any kind of authenticated
value.
Conversely, if you use an https url, you get some cryptographic
authentication of what you are cloning. The crypto there is far from
perfect but it is massively better than nothing.
Additionally, in general, using and supporting https also means that
*what users are accessing* is encrypted. This enhances user privacy.
In the specific case of the git trees on xenbits this is a very minor
consideration.
Signed-off-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Avoid having error messages with the same text. No functional change.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
For one, "no-xen" should not imply "no-eager-fpu", as "eager FPU" mode
is to guard guests, not Xen itself, which is also expressed so by
print_details().
And then opt_ssbd, despite being off by default, should also be cleared
by the "no" and "no-xen" sub-options.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
In the display protocol connector's id is named as unique-id. This patch renames
it in the libxl/xl code and uses XENDISPL_FIELD... definitions from the protocol
header.
Signed-off-by: Oleksandr Grytsov <oleksandr_grytsov@epam.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
FreeBSD package manager uses /usr/local/ as the default install path,
but that's not part of the compiler search path, so add it using the
APPEND_{LIB/INCLUDES} variables.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
ENODATA is not part of the standard set of errno values, so use
ENOENT instead. This fixes the build on FreeBSD.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
clang is not capable of building the x86 emulator test harness, so
disconnect it from the clang build until it can be fixed.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
SeaBIOS requires gcc and GNU ld in order to build, so allow setting
SEABIOSCC and SEABIOSLD by the caller when building in order to pass
the path to the compiler and linker that should be used when building
SeaBIOS.
Note that the LD32BIT-y variable was used by FreeBSD builds and is no
longer needed due to the ability to select the LD to use for SeaBIOS
builds, so remove it.
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
newlib has a hardcoded dependency on texinfo, which in turn means that
Xen has this dependency when building the stub domains. The files that
newlib generates are never used or exposed by the Xen build system so
its really a wasted effort to require a heavy (size wise) package for
the build of something that's thrown away. Instead of adding it as a
requirement this patches out the code path in newlib that uses it so
that the files are never generated.
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Acked-by: Wei Liu <wei.liu2@citrix.com>
Reviewed-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
After attempting to develop the infrastructure, it turns out that the choice
of naming is suboptimal.
Rename msr_domain_policy to just msr_policy to mirror the CPUID side of
things, and alter the 'dp' variable name convention to 'mp'. While altering
all the names, export all of the system msr_policy objects (which are already
global symbols).
Rename msr_vcpu_policy to vcpu_msrs and switch 'vp' to 'msrs' in code. Update
the arch_vcpu field name to match.
No functional change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Jan Beulich <jbeulich@suse.com>
... and use it in place of the opencoded instances.
For consistency, restructure init_domain_cpuid_policy() to be like
init_{domain,vcpu}_msr_policy() by operating on the local pointer where
possible.
No change in behaviour.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
For libxengnttab.map, introducing a new VERS* section must be matched with a
equivalent SONAME bump.
For libxencall.map, the hunk as presented would have been ok, if Xen 4.11
hadn't been released between xencall_buffers_never_fault() and xencall_fd()
being introduced. Given the timing of the release, xencall_fd() needs moving
into a new VERS_1.2 section.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
UREGS_user_sizeof had a use long, long ago on x86, and apparently was
blindly cloned into ARM code.
The uses of UREGS_saved_upcall_mask and UREGS_{ds,es,fs,gs} disappeared
in 71eb3a65e3 ("VMX: streamline entry.S code") and 110b2d6e8a ("SVM:
streamline entry.S code").
There are other ARM side UREGS_* which look to be unused, but I'd prefer
to leave deciding on their disposition to ARM folks.
irq_caps_offset and next_in_list_offset were introduced in 57b0566668
("Re-enable MSI support") without any user.
VCPU_failsafe_{addr,sel} and _VGCF_failsafe_disables_events uses went
away in d9b7ef209a ("x86: drop failsafe callback invocation from
assembly").
CPUINFO_processor_id looks to have been introduced without user by
917335d8b3 ("x86: Avoid assumptions about C struct layouts from asm
code").
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Acked-by: Julien Grall <julien.grall@arm.com>
First of all introduce a helper function instead of replicating almost
the same code for PV and HVM. The differences between the two pieces of
code actually points out an issue (which is also addressed here): In
the HVM case FCW would not have been set to FCW_RESET in certain cases
(note for example that XRSTOR loads FCW_DEFAULT rather then FCW_RESET
when the respective xstate_bv bit is clear).
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Only the pre_flush() / post_flush() section actually needs to run with
interrupts disabled.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
In this case, just like xcr0_accum, nonlazy_xstate_used should always be
set to the intended new value, rather than possibly leaving the flag set
from a prior state load.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Including the header solely at the assembler level means the compiler
can't record the header as a dependency in .*.o.d files. Arrange for the
header to be include-able at the C level instead.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Wei Liu
Ian Jackson
Roger Pau Monné
Jan Beulich
Sergey Dyasli
Julien Grall
Doug Goldstein
Juergen Gross
Oleksandr Grytsov
Lars Kurth
Andrew Cooper