Re: Qemu with Spice support on Windows10 host: a report

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Il giorno mar 23 mar 2021 alle ore 20:33 Helge Konetzka
<helge@xxxxxxxxxxx> ha scritto:
>
> Hello!
>
> Please excuse my english - I am no natural speaker!
>
> Thank you very much for providing spice! I use it in my daily work on a
> qemu/kvm debian host.
>
> I am trying to run my debian qemu setup on windows10.
> For windows a qemu distribution exists in https://qemu.weilnetz.de/, but
> it is missing spice support.
> While reading the changelog of spice 0.14.3 I happily realized that the
> integration is possible now!
>
> Intention of this posting is to document the issues on my way to a
> working install and the first steps
> testing it. I hope, this is the correct list to send it to...
>
> Using a cygwin64 environment I built qemu with x86_64-w64-mingw32
> toolchain.
> Resulting qemu seems to work as good as the distribution version.
>
> To integrate spice support into qemu, I used spice-protocol-0.14.3.tar.xz
> and spice-0.14.3.tar.bz2 as well as spice-0.14.91.tar.bz2
>
> I was able to build and deploy spice-protocol and spice-server to the
> x86_64-w64-mingw32
> sys-root. For the spice-server build I created a patch and added a
> compiler flag to get rid of compile
> and link errors. After a rebuild of qemu spice-support was enabled.
>
> Starting qemu with "-vga qxl -spice
> port=5905,addr=127.0.0.1,disable-ticketing" works well!
> I start spicy having a xserver in background with
> spicy -h 127.0.0.1 -p 5905
> and I am able to connect to qemu and my VM desktop is visible.
>
> The mouse seems to act strange when trying to enter the spicy window,
> but the host mouse pointer
> just stays visible, the guest mouse pointer is working.
>
> Sadly problems occur after adding more spice with:
> "-device virtio-serial-pci -device
> virtserialport,chardev=spicechannel0,name=com.redhat.spice.0 -chardev
> spicevmc,id=spicechannel0,name=vdagent"
> qemu starts logging warnings in a loop and is no longer usable:
> D:\cygwin64\x86_64-w64-mingw32\qemu-5.2.0\qemu-system-x86_64.exe:
> warning: GLib: Too many handles to wait for!
>
> It seems to me this is a problem of qemu which I found independent from
> spice in the distributed version, too.
> It occurs when option "-device virtio-serial-pci" is added.
>
> Below I attach my build path, my patch and reasons for the used compiler
> flags.
> I hope it is useful for improving the ease of build for spice on windows.
>
> Regards, Helge Konetzka.
>
> --------------------------
>
> I deployed spice-protocol and spice-server this way:
>
> spice-protocol-0.14.3$ meson setup builddir
> spice-protocol-0.14.3$ meson configure --prefix
> /usr/x86_64-w64-mingw32/sys-root/mingw builddir
> spice-protocol-0.14.3$ meson install -C builddir
>
> # Add -fstack-protector to automatically added FLAGS
> spice-0.14.91$ export CFLAGS="-g -O2 -fstack-protector"
> spice-0.14.91$ export CXXFLAGS="-g -O2 -fno-exceptions -fno-check-new
> -fstack-protector"
> # Patch to include red-common.h before jpeglib.h
> spice-0.14.91$ patch -p1 < ../spice-0.14.3-cygwin64.patch
> spice-0.14.91$ ./configure --host=x86_64-w64-mingw32
> --prefix=/usr/x86_64-w64-mingw32/sys-root/mingw \
>                  --without-sasl --enable-extra-checks --disable-silent-rules

Surely you don't want --enable-extra-checks for the final product, but
good to test with.

> spice-0.14.91$ make -j4
> spice-0.14.91$ make install
>
> --------------------------
>
> Patch is necessary, because INT32 is defined twice - in
> /usr/x86_64-w64-mingw32/sys-root/mingw/include/basetsd.h
> and /usr/x86_64-w64-mingw32/sys-root/mingw/include/jmorecfg.h
>

We  use CI to check that always compile for Windows using MingW, which
distro are you using?

> $ grep "^\s*typedef.*\bINT32\b"
> /usr/x86_64-w64-mingw32/sys-root/mingw/include/* 2>/dev/null
> /usr/x86_64-w64-mingw32/sys-root/mingw/include/basetsd.h:  typedef
> signed int INT32,*PINT32;
> /usr/x86_64-w64-mingw32/sys-root/mingw/include/jmorecfg.h:typedef long
> INT32;
> Both header files are included transitionally. jpeglib.h includes
> jmorecfg.h, red-common.h includes basetsd.h
> If basetsd.h is processed first (via red-common.h), "#ifndef
> _BASETSD_H_" in jmorecfg.h prevents second definition of INT32
>
> $ cat spice-0.14.3-cygwin64.patch
> diff -Naur spice-0.14.91.orig/server/jpeg-encoder.c
> spice-0.14.91.patched/server/jpeg-encoder.c
> --- spice-0.14.91.orig/server/jpeg-encoder.c    2020-10-26
> 13:18:53.000000000 +0100
> +++ spice-0.14.91.patched/server/jpeg-encoder.c    2021-03-23
> 12:27:18.696134500 +0100
> @@ -17,9 +17,9 @@
>   #include <config.h>
>
>   #include <stdio.h>
> +#include "red-common.h"
>   #include <jpeglib.h>
>
> -#include "red-common.h"
>   #include "jpeg-encoder.h"
>
>   struct JpegEncoderContext {
> diff -Naur spice-0.14.91.orig/server/mjpeg-encoder.c
> spice-0.14.91.patched/server/mjpeg-encoder.c
> --- spice-0.14.91.orig/server/mjpeg-encoder.c    2020-10-26
> 13:18:53.000000000 +0100
> +++ spice-0.14.91.patched/server/mjpeg-encoder.c    2021-03-23
> 12:27:22.501484000 +0100
> @@ -20,9 +20,9 @@
>   #include <stdio.h>
>   #include <inttypes.h>
>   #include <jerror.h>
> +#include "red-common.h"
>   #include <jpeglib.h>
>
> -#include "red-common.h"
>   #include "video-encoder.h"
>   #include "utils.h"
>
> ---------------------------

I think it would be better to move Jpeg includes at the end and add a
comment instead.

>
> Additional Compiler-Flag -fstack-protector is necessary in CFLAGS and
> CXXFLAGS to get rid of linker errors
> see
> https://sourceforge.net/p/mingw-w64/mailman/message/36764708/
> https://github.com/msys2/MINGW-packages/issues/5803
>
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(sys-socket.o): in function `socket_newpair':
> /mnt/d/Qemu/src/spice-0.14.91/server/sys-socket.c:284: undefined
> reference to `__stack_chk_fail'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(sys-socket.o): in function `socket_win32_init':
> /mnt/d/Qemu/src/spice-0.14.91/server/sys-socket.c:209: undefined
> reference to `__stack_chk_fail'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(sys-socket.o):sys-socket.c:(.rdata$.refptr.__stack_chk_guard[.refptr.__stack_chk_guard]+0x0):
> undefined reference to `__stack_chk_guard'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(tree.o):/usr/x86_64-w64-mingw32/sys-root/mingw/include/string.h:208:
> undefined reference to `__memset_chk'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(utils.o): in function `red_dump_openssl_errors':
> /mnt/d/Qemu/src/spice-0.14.91/server/utils.c:128: undefined reference to
> `__stack_chk_fail'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(websocket.o): in function `websocket_read':
> /mnt/d/Qemu/src/spice-0.14.91/server/websocket.c:447: undefined
> reference to `__stack_chk_fail'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(websocket.o): in function `memcpy':
> /usr/x86_64-w64-mingw32/sys-root/mingw/include/string.h:202: undefined
> reference to `__memcpy_chk'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(websocket.o): in function `sprintf':
> /usr/x86_64-w64-mingw32/sys-root/mingw/include/stdio.h:372: undefined
> reference to `__chk_fail'
> /usr/lib/gcc/x86_64-w64-mingw32/10/../../../../x86_64-w64-mingw32/bin/ld:
> ./.libs/libserver.a(websocket.o): in function `websocket_new':
> /mnt/d/Qemu/src/spice-0.14.91/server/websocket.c:792: undefined
> reference to `__stack_chk_fail'
> collect2: error: ld returned 1 exit status
> make[4]: *** [Makefile:790: libspice-server.la] Error 1
> make[4]: Leaving directory '/cygdrive/d/Qemu/src/spice-0.14.91/server'
> make[3]: *** [Makefile:942: all-recursive] Error 1
> make[3]: Leaving directory '/cygdrive/d/Qemu/src/spice-0.14.91/server'
> make[2]: *** [Makefile:704: all] Error 2
> make[2]: Leaving directory '/cygdrive/d/Qemu/src/spice-0.14.91/server'
> make[1]: *** [Makefile:539: all-recursive] Error 1
> make[1]: Leaving directory '/cygdrive/d/Qemu/src/spice

I suppose the FORTIFY definition requires -fstack-protector ito get
the right libraries.

Frediano
_______________________________________________
Spice-devel mailing list
Spice-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/spice-devel



[Index of Archives]     [Linux Virtualization]     [Linux Virtualization]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [ECOS]     [Asterisk Internet PBX]     [Linux API]     [Monitors]