Replace OE private qemu patch with one that got upstreamed
and solves the same problem: increase qemumips CI performance
by increasing number of TLBs in CPU model and reduce need to
run software TLB refill code.
(From OE-Core rev: a99dace7463d310688f4098a51316dc0743651e2)
Signed-off-by: Victor Kamensky <kamensky@cisco.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This reverts commit 4470a04943352224955f17e004962f0f9e1c9b0c.
Will be replaced with upstreamed solution instead, that just
bumped number of TLBs to 64 in original 34Kf CPU model.
(From OE-Core rev: 894f1d58d93073d290f35d1090b03717bc7b4dc0)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
In Yocto Project PR 13992 it was reported that qemumips
in autobuilder runs almost twice slower then qemumips64 and
some times hit time out.
Upon investigations of qemu-system with perf, gdb, and
SystemTap and comparing qemumips and qemumips64 machines
behavior it was noticed that qemu soft mmu code behaves
quite different and in case if qemumips tlbwr instruction
called 16 times more oftern. It happens that in qemumips64
case qemu runs with cpu type that contains 64 TLB, but in case
of qemumips qemu runs with cpu type that contains only
16 TLBs.
The idea of proposed qemu patch is to introduce fictitious
34Kf-64tlb cpu type that defined exactly as 34Kf but has
64 TLBs, instead of original 16 TLBs.
Testing of core-image-full-cmdline:do_testimage with
34Kf-64tlb shows 40% or so test execution real time
improvement.
Note for future porters of the patch: easiest way to update
the patch and be in sync with 34Kf definition is to copy
34Kf machine definition and apply the following changes to
it (just change 15 to 63 of CP0C1_MMU bits value)
[kamensky@coreos-lnx2 qemu]$ diff ~/34Kf.c ~/34Kf-64tlb.c
2c2
< .name = "34Kf",
> .name = "34Kf-64tlb",
6c6
< .CP0_Config1 = MIPS_CONFIG1 | (1 << CP0C1_FP) | (15 << CP0C1_MMU) |
> .CP0_Config1 = MIPS_CONFIG1 | (1 << CP0C1_FP) | (63 << CP0C1_MMU) |
Fixes https://bugzilla.yoctoproject.org/show_bug.cgi?id=13992
Upstream Status: Inappropriate
(From OE-Core rev: 4470a04943352224955f17e004962f0f9e1c9b0c)
Signed-off-by: Victor Kamensky <kamensky@cisco.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Multiple files have " ," instead of ", " in expressions. This changes
them to conform to the way the rest of them are done.
Found and corrected via:
git ls-files | xargs sed --follow-symlinks -i 's/ ,d/, d/g'
(From OE-Core rev: 36c3afd2dd8bded02ea8f255e89a09ebd75c795b)
Signed-off-by: Jon Mason <jon.mason@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Override DEBUG_BUILD for qemu as the qemu upstream states it
doesn't work without optimization [1] to fix below build failure
when debug build enabled.
| /usr/lib/gcc/x86_64-wrs-linux/10.1.0/../../../../x86_64-wrs-linux/bin/ld.bfd: /mnt/build/tmp/work/x86_64-linux/qemu-system-native/5.1.0-r0/qemu-5.1.0/fsdev/qemu-fsdev-throttle.c:25: undefined reference to `unknown_lock_type'
| /usr/lib/gcc/x86_64-wrs-linux/10.1.0/../../../../x86_64-wrs-linux/bin/ld.bfd: ../fsdev/qemu-fsdev-throttle.o: in function `fsdev_co_throttle_request':
| /mnt/build/tmp/work/x86_64-linux/qemu-system-native/5.1.0-r0/qemu-5.1.0/fsdev/qemu-fsdev-throttle.c:103: undefined reference to `unknown_lock_type'
| /usr/lib/gcc/x86_64-wrs-linux/10.1.0/../../../../x86_64-wrs-linux/bin/ld.bfd: ../fsdev/qemu-fsdev-throttle.o:/mnt/build/tmp/work/x86_64-linux/qemu-system-native/5.1.0-r0/qemu-5.1.0/fsdev/qemu-fsdev-throttle.c:103: more undefined references to `unknown_lock_type' follow
| collect2: error: ld returned 1 exit status
[1]: https://lists.nongnu.org/archive/html/qemu-devel/2020-09/msg03873.html
(From OE-Core rev: 2d6a15cfb12e72ae3d38f6ecf8c6c2e356e0fda8)
Signed-off-by: Mingli Yu <mingli.yu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This reverts commit e442924d9a.
This was rejected upstream and its been pointed out qemu doesn't work
without optimization.
Instead we should just error if the user attempts to build it without
optimization.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When use gcc 10.1.0 and debug build is enabled, there comes
below build failure:
| /usr/lib/gcc/x86_64-wrs-linux/10.1.0/../../../../x86_64-wrs-linux/bin/ld.bfd: /mnt/build/tmp/work/x86_64-linux/qemu-system-native/5.1.0-r0/qemu-5.1.0/fsdev/qemu-fsdev-throttle.c:25: undefined reference to `unknown_lock_type'
| /usr/lib/gcc/x86_64-wrs-linux/10.1.0/../../../../x86_64-wrs-linux/bin/ld.bfd: ../fsdev/qemu-fsdev-throttle.o: in function `fsdev_co_throttle_request':
| /mnt/build/tmp/work/x86_64-linux/qemu-system-native/5.1.0-r0/qemu-5.1.0/fsdev/qemu-fsdev-throttle.c:103: undefined reference to `unknown_lock_type'
| /usr/lib/gcc/x86_64-wrs-linux/10.1.0/../../../../x86_64-wrs-linux/bin/ld.bfd: ../fsdev/qemu-fsdev-throttle.o:/mnt/build/tmp/work/x86_64-linux/qemu-system-native/5.1.0-r0/qemu-5.1.0/fsdev/qemu-fsdev-throttle.c:103: more undefined references to `unknown_lock_type' follow
| collect2: error: ld returned 1 exit status
So always define unknown_lock_type to fix the above error.
(From OE-Core rev: d88e83d49478a2535575df769329d9b1e6579564)
Signed-off-by: Mingli Yu <mingli.yu@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Backport a patch from upstream to fix CVE-2020-14364.
(From OE-Core rev: 1ffd4d94e879b338e190a2c4798a4e0bada9b5a7)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
* Drop backported CVE fixes
* Drop cpu backtrace patch from 2015 for debugging an issue which we no longer see
(patch throws rejects, files have moved)
* Update mips patch to account for file renames
* Update chardev patch to match upstream code changes
* Update webkitgtk patch, qemumips build works ok but qemux86 musl webkitgtk still
fails. Need to figure out the correct fix and upstream it for this, current
revert patch is not maintainable.
Release notes for 5.1.0 mention slight qemumips performance improvements
which would be valuable to us. My tests show no improvement in qemumips
testimage execution time for core-image-sato-sdk.
Fix a ptest issue for a file looking for /usr/bin/bash when we have
/bin/bash.
(From OE-Core rev: 686b770af67fdd2251f4ddab5b0eefc8fb0870ef)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Use virtual/libgl instead of mesa within the recipes to provide users the
flexibility to specify providers other than mesa.
(From OE-Core rev: ddbe09b0f9861fe3012611b5a5298eb4d7c4df45)
Signed-off-by: Sai Hari Chandana Kalluri <chandana.kalluri@xilinx.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Major update after v4.2.
Changes:
- os_find_datadir() was changed after the v4.2 release
causing v5.0 to not find the bios and not boot the
image. Fix is sent to upstream qemu.
See: qemu/find_datadir.patch
- v5.0 binary had host contamination for dynamically linked
libraries, "--extra-ldflags='${LDFLAGS}'" in EXTRA_OECONF
resolved the issue
- bluetooth code was removed: qemu.git$ git show 1d4ffe8dc7
hence removed PACKAGECONFIG[bluez]
- -show-cursor qemu option is now deprecated, updated
scripts/runqemu to use updated option instead
- added PACKAGECONFIG definitions
- added qemu-ptest to conf/distro/include/ptest-packagelists.inc
- increased support for ARM architecture, cpu and board
- removed patches merged upstream and refreshed
existing ones
Testing:
Build core-image-minimal against the machines in
openembedded-core/meta/conf/machine and succesfully
booted with qemu v5.0
Ran qemu-ptest on x86-64 and arm64 with identical results:
PASS: 1166
SKIP: 0
FAIL: 0
(From OE-Core rev: ee9ec9e344541c1ccd9b9b8e3b8c1e00d008ad85)
Signed-off-by: Sakib Sajal <sakib.sajal@windriver.com>
Signed-off-by: Joe Slater <joe.slater@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
qemu will not build for -Og optimization because macros
in lockable.h expect dead-code elimination. Override DEBUG_BUILD.
(From OE-Core rev: ad404b2e51223625bd95f9a7da4c7a690c37bdfb)
Signed-off-by: Joe Slater <joe.slater@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Note that to actually use accelerated GL passthrough, there are two options
1) a suitable frontend need to be also enabled - gtk+ and SDL both seem to work well.
Previously I struggled to make SDL work, but now it seems fine.
2) it is also possible to render off-screen with -display egl-headless option,
and see the output with a VNC viewer (for which, qemu needs to be started
with a VNC server):
$ runqemu kvm egl-headless publicvnc
(From OE-Core rev: 662c688e635f4490aac0d6d34ce7a7b31d73f5c5)
Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
If the SRC_URI is switched to be git based, QEMU will automatically update
itself at compile time for select git submodules. This by passes the
bitbake git fetcher. These modules are always present in the release
tarballs, so only are problematic when used with git based SRC_URIs.
These switches will have no effect on a tarball based SRC_URI.
(From OE-Core rev: 53b38a89fc7ccbceefd1a3a79bd376d9f6419565)
Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
* Add PACKAGECONFIG option for xkbcommon
qemu-keymap.c:16:10: fatal error: xkbcommon/xkbcommon.h: No such file or directory
* Add PACKAGECONFIG option and patch for libudev
commands-posix.c:53:10: fatal error: libudev.h: No such file or directory
* Add PACKAGECONFIG option for libxml2
util/osdep.c:136: undefined reference to `fcntl64'
- Without specifying libxml2, configure searches the system and pulls in the system
libxml2 if it is present. In the process it adds -L/usr/lib64 which causes the
system libc to be linked instead of the one from the extended tarball.
None of the above libraries appear to be included in the depends for any of the qemu
builds, so if they are getting linked in, its probably not intentionally.
(From OE-Core rev: 3f1fd15cf0df4a4b12d49ac795b1c9e02de61429)
Signed-off-by: Jeremy Puhlman <jpuhlman@mvista.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
tests/qemu-iotests/common.env is generated from configure which
we pass ${HOSTTOOLS_DIR}/python3 as our python to use, which gets
copied into the ptests. Correct python3 path.
(From OE-Core rev: e8e5765cf2ad324a84634877eebc8289f1bc934f)
Signed-off-by: Jeremy A. Puhlman <jpuhlman@mvista.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The Xen recipe has been divided into separate recipes for the hypervisor
and tools in meta-virtualization commit 545461ba, so the package name
references in the qemu recipe need to be updated to the new xen-tools
packages. This change allows the temporary bbappend applied to qemu in
meta-virtualization in that change to be retired.
(From OE-Core rev: e7a7ce0999de3c562e7288538641087cf6d55d1f)
Signed-off-by: Christopher Clark <christopher.w.clark@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
While we are upgrading let's refresh patches and remove the outdated
patches.
(From OE-Core rev: ca276b77904429a1ff3188d2373535a462afe266)
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
glibc defines minimum kernel needed to be 3.10.0 for LE ppc64
(From OE-Core rev: c1c296a42920af6725706bdea8e61b8c4f5f14a8)
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Qemu supports attaching the virtual machine to an existing network
bridge interface via the qemu-bridge-helper program (as long as the
system is correctly configured to give the user permissions). Add
support for runqemu to do this also via the "bridge=<INTERFACE>"
argument. Note that for this to work correctly, the host
qemu-bridge-helper must be used, not the one that might have been built
by qemu-native. In order for qemu to correctly find this program, a
qemu-oe-bridge-helper program has been added to qemu-helper-native, and
runqemu will use this helper as the bridge helper. The helper will look
for the host qemu-bridge-helper first by looking in the
QEMU_BRIDGE_HELPER environment variable, then by search common paths
where the helper is installed.
(From OE-Core rev: 9e7b38c61c6b84b7f137c733ac5da9414025693d)
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This commit adds PACKAGECONFIG for glusterfs. glusterfs is not required f-
or runqemu, so it should be optional to reduce build time.
(From OE-Core rev: f5ffd6ce6e7431b17a291980843752e2d2fb0b44)
Signed-off-by: Shohei Maruyama <shohei.maruyama@cybertrust.co.jp>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Pass all of the compiler and linker flags so the build is correctly configured.
(From OE-Core rev: 05151064eb2c0907ebb815c9ae3481a912835b39)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Displaying a help message if help was requested isn't an error.
(From OE-Core rev: 9d98e881de9030d80e451519b054681d4e9d2621)
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Changing the gl options to qemu doesn't result in a correctly rebuilt
binary, the GL linkage can persist from a build where it was enabled
to one where it was not.
As well as clearly being incorrect and non-reproducible, this caused
some mystery failures on the autobuilder.
Cleaning ${B} at do_configure time avoids this. Most recipes
(e.g. autotools derived ones) already clean ${B} as appropriate and
avoid this issue.
(From OE-Core rev: b91b30c09ffbcd81e8dcaae832b3ae4f61046236)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When PACKAGECONFIG libcurl is enabled, we would get the following build failure.
ERROR: Nothing PROVIDES 'libcurl' (but perf.bb DEPENDS on or otherwise requires it). Close matches:
The dependency should be curl.
(From OE-Core rev: 531bf25c4e9dda4938f1de6bf7f929ca9d639114)
Signed-off-by: He Zhe <zhe.he@windriver.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
