[Qemu-devel] Syscall 269

[Qemu-devel] Syscall 269

[James Pellow]

Hi All,

I am trying to chroot to a gentoo flavor of arm linux on my AMD tbird-1.4GHz.  
I have set up binfmt_misc and qemu to allow me to do the chroot, and all 
seems to be working well.  Now I wanted to emerge some stuff, and I get the 
following message:

qemu: Unsupported syscall: 269.

Looking at the arm linux kernel source, I see that 269 is utimes.  Looking at 
the source code for qemu it seems that all I have to do is to add a define 
for TARGET_NR_utimes in all linux-user/*/syscall_nr.h and then add a new case 
in linux-user/syscall.c.  

So, I gave it a shot.  The patch is at the bottom of this message.  This is 
the first time I have looked at the qemu sources, so I am likely missing 
something, but the patch does seem to allow emerge to work properly under 
gentoo.  If a correct implementation requires more work, I am happy to do 
that too, just let me know.  BTW, I am not subscribed to this list to please 
CC me.

Many thanks for a wonderful app.  

Cheers,

-- 
*****************************
James A. Pellow, President
Alent Design Solutions
www.alentdesignsolutions.com
(509) 526-0682
*****************************



diff -ruN qemu-0.6.1/linux-user/arm/syscall_nr.h 
qemu-0.6.1_new/linux-user/arm/syscall_nr.h
--- qemu-0.6.1/linux-user/arm/syscall_nr.h      2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/arm/syscall_nr.h  2004-11-18 00:58:44.973757936 
-0800
@@ -259,3 +259,5 @@
                                        /* 254 for set_thread_area */
                                        /* 255 for get_thread_area */
                                        /* 256 for set_tid_address */
+#define TARGET_NR_utimes                       (269)
+
diff -ruN qemu-0.6.1/linux-user/i386/syscall_nr.h 
qemu-0.6.1_new/linux-user/i386/syscall_nr.h
--- qemu-0.6.1/linux-user/i386/syscall_nr.h     2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/i386/syscall_nr.h 2004-11-18 01:28:59.324934632 
-0800
@@ -271,3 +271,5 @@
 #define TARGET_NR_clock_getres (TARGET_NR_timer_create+7)
 #define TARGET_NR_clock_nanosleep      (TARGET_NR_timer_create+8)

+#define TARGET_NR_utimes               271
+
diff -ruN qemu-0.6.1/linux-user/syscall.c qemu-0.6.1_new/linux-user/syscall.c
--- qemu-0.6.1/linux-user/syscall.c     2004-11-14 12:51:33.000000000 -0800
+++ qemu-0.6.1_new/linux-user/syscall.c 2004-11-18 01:15:54.561236848 -0800
@@ -3025,6 +3025,10 @@
     case TARGET_NR_get_thread_area:
         goto unimplemented_nowarn;
 #endif
+    case TARGET_NR_utimes:
+       ret = get_errno(utimes((const char *)arg1,
+               (const struct timeval *)arg2));
+       break;
     default:
     unimplemented:
         gemu_log("qemu: Unsupported syscall: %d\n", num);

Re: [Qemu-devel] Syscall 269

[Paul Brook]

On Thursday 18 November 2004 09:38, James Pellow wrote:
> Hi All,
>
> I am trying to chroot to a gentoo flavor of arm linux on my AMD
> tbird-1.4GHz. I have set up binfmt_misc and qemu to allow me to do the
> chroot, and all seems to be working well.  Now I wanted to emerge some
> stuff, and I get the following message:
>
> qemu: Unsupported syscall: 269.
>
> Looking at the arm linux kernel source, I see that 269 is utimes.  Looking
> at the source code for qemu it seems that all I have to do is to add a
> define for TARGET_NR_utimes in all linux-user/*/syscall_nr.h and then add a
> new case in linux-user/syscall.c.
>
> So, I gave it a shot.  The patch is at the bottom of this message.  This is
> the first time I have looked at the qemu sources, so I am likely missing
> something, but the patch does seem to allow emerge to work properly under
> gentoo.  If a correct implementation requires more work, I am happy to do
> that too, just let me know.  BTW, I am not subscribed to this list to
> please CC me.

You also need to do proper 32/64bit and big/little endian conversion of struct 
timeval. It happens to work in your case because arm-linux and i686-linux 
both use the same word size and endianness.

Paul

Re: [Qemu-devel] Syscall 269

[James Pellow]

Yes, I thought about that, and realized I wouldn't have a problem in my case.  
I'll take a closer look at the syscall.c file and see how that was done for 
other syscalls, and see if I can get a better patch put together.  Thanks for 
the reply.

-James

On Thursday 18 November 2004 05:03 am, you wrote:
> On Thursday 18 November 2004 09:38, James Pellow wrote:
> > Hi All,
> >
> > I am trying to chroot to a gentoo flavor of arm linux on my AMD
> > tbird-1.4GHz. I have set up binfmt_misc and qemu to allow me to do the
> > chroot, and all seems to be working well.  Now I wanted to emerge some
> > stuff, and I get the following message:
> >
> > qemu: Unsupported syscall: 269.
> >
> > Looking at the arm linux kernel source, I see that 269 is utimes. 
> > Looking at the source code for qemu it seems that all I have to do is to
> > add a define for TARGET_NR_utimes in all linux-user/*/syscall_nr.h and
> > then add a new case in linux-user/syscall.c.
> >
> > So, I gave it a shot.  The patch is at the bottom of this message.  This
> > is the first time I have looked at the qemu sources, so I am likely
> > missing something, but the patch does seem to allow emerge to work
> > properly under gentoo.  If a correct implementation requires more work, I
> > am happy to do that too, just let me know.  BTW, I am not subscribed to
> > this list to please CC me.
>
> You also need to do proper 32/64bit and big/little endian conversion of
> struct timeval. It happens to work in your case because arm-linux and
> i686-linux both use the same word size and endianness.
>
> Paul

-- 
*****************************
James A. Pellow, President
Alent Design Solutions
www.alentdesignsolutions.com
(509) 526-0682
*****************************

Re: [Qemu-devel] Syscall 269

[James Pellow]

Hi Paul,

Is this closer to what you want?  I saw the swap functions.  I wasn't sure 
what long translated to for all supported archs.  Do I need to do 32/64 bit 
translation?  tv_sec and tv_usec are both long.

Thanks,

James Pellow

------------------------------------------------------------------------------------------------------------

diff -ruN qemu-0.6.1/linux-user/arm/syscall_nr.h 
qemu-0.6.1_new/linux-user/arm/syscall_nr.h
--- qemu-0.6.1/linux-user/arm/syscall_nr.h      2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/arm/syscall_nr.h  2004-11-18 00:58:44.000000000 
-0800
@@ -259,3 +259,5 @@
                                        /* 254 for set_thread_area */
                                        /* 255 for get_thread_area */
                                        /* 256 for set_tid_address */
+#define TARGET_NR_utimes                       (269)
+
diff -ruN qemu-0.6.1/linux-user/i386/syscall_nr.h 
qemu-0.6.1_new/linux-user/i386/syscall_nr.h
--- qemu-0.6.1/linux-user/i386/syscall_nr.h     2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/i386/syscall_nr.h 2004-11-18 01:28:59.000000000 
-0800
@@ -271,3 +271,5 @@
 #define TARGET_NR_clock_getres (TARGET_NR_timer_create+7)
 #define TARGET_NR_clock_nanosleep      (TARGET_NR_timer_create+8)

+#define TARGET_NR_utimes               271
+
diff -ruN qemu-0.6.1/linux-user/syscall.c qemu-0.6.1_new/linux-user/syscall.c
--- qemu-0.6.1/linux-user/syscall.c     2004-11-14 12:51:33.000000000 -0800
+++ qemu-0.6.1_new/linux-user/syscall.c 2004-12-03 15:44:37.174197904 -0800
@@ -3025,6 +3025,15 @@
     case TARGET_NR_get_thread_area:
         goto unimplemented_nowarn;
 #endif
+    case TARGET_NR_utimes:
+    {
+       struct timeval tv;
+       struct timeval *target_tv = (struct timeval *)arg2;
+       tv.tv_sec  = tswapl(target_tv->tv_sec);
+       tv.tv_usec = tswapl(target_tv->tv_usec);
+       ret = get_errno(utimes((const char *)arg1, &tv));
+       break;
+    }
     default:
     unimplemented:
         gemu_log("qemu: Unsupported syscall: %d\n", num);

------------------------------------------------------------------------------------------------------------------------



On Thursday 18 November 2004 05:03 am, Paul Brook wrote:
> On Thursday 18 November 2004 09:38, James Pellow wrote:
> > Hi All,
> >
> > I am trying to chroot to a gentoo flavor of arm linux on my AMD
> > tbird-1.4GHz. I have set up binfmt_misc and qemu to allow me to do the
> > chroot, and all seems to be working well.  Now I wanted to emerge some
> > stuff, and I get the following message:
> >
> > qemu: Unsupported syscall: 269.
> >
> > Looking at the arm linux kernel source, I see that 269 is utimes. 
> > Looking at the source code for qemu it seems that all I have to do is to
> > add a define for TARGET_NR_utimes in all linux-user/*/syscall_nr.h and
> > then add a new case in linux-user/syscall.c.
> >
> > So, I gave it a shot.  The patch is at the bottom of this message.  This
> > is the first time I have looked at the qemu sources, so I am likely
> > missing something, but the patch does seem to allow emerge to work
> > properly under gentoo.  If a correct implementation requires more work, I
> > am happy to do that too, just let me know.  BTW, I am not subscribed to
> > this list to please CC me.
>
> You also need to do proper 32/64bit and big/little endian conversion of
> struct timeval. It happens to work in your case because arm-linux and
> i686-linux both use the same word size and endianness.
>
> Paul

-- 
*****************************
James A. Pellow, President
Alent Design Solutions
www.alentdesignsolutions.com
(509) 526-0682
*****************************

Re: [Qemu-devel] Syscall 269

[Paul Brook]

On Saturday 04 December 2004 00:05, James Pellow wrote:
> Hi Paul,
>
> Is this closer to what you want?  I saw the swap functions.  I wasn't sure
> what long translated to for all supported archs.  Do I need to do 32/64 bit
> translation?  tv_sec and tv_usec are both long.

There is already code to properly handle struct timeval, you should use that. 
See the code for TARGET_NR_settimeofday.

Paul

Re: [Qemu-devel] Syscall 269

[James Pellow]

Hi Paul,

Ok, third time is the charm right...  It looks like the functions you referred 
to that handle struct timeval, are doing exactly what I had done.  Here is an 
updated patch that factors out the swap code using the existing function.  
Hopefully this is ready for inclusion.  Thanks for your help with this and 
for a truely incredible project!

-James Pellow

-----------------------------------------------------------------------------------------------------------------

diff -ruN qemu-0.6.1/linux-user/arm/syscall_nr.h 
qemu-0.6.1_new/linux-user/arm/syscall_nr.h
--- qemu-0.6.1/linux-user/arm/syscall_nr.h      2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/arm/syscall_nr.h  2004-11-18 00:58:44.000000000 
-0800
@@ -259,3 +259,5 @@
                                        /* 254 for set_thread_area */
                                        /* 255 for get_thread_area */
                                        /* 256 for set_tid_address */
+#define TARGET_NR_utimes                       (269)
+
diff -ruN qemu-0.6.1/linux-user/i386/syscall_nr.h 
qemu-0.6.1_new/linux-user/i386/syscall_nr.h
--- qemu-0.6.1/linux-user/i386/syscall_nr.h     2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/i386/syscall_nr.h 2004-11-18 01:28:59.000000000 
-0800
@@ -271,3 +271,5 @@
 #define TARGET_NR_clock_getres (TARGET_NR_timer_create+7)
 #define TARGET_NR_clock_nanosleep      (TARGET_NR_timer_create+8)

+#define TARGET_NR_utimes               271
+
diff -ruN qemu-0.6.1/linux-user/syscall.c qemu-0.6.1_new/linux-user/syscall.c
--- qemu-0.6.1/linux-user/syscall.c     2004-11-14 12:51:33.000000000 -0800
+++ qemu-0.6.1_new/linux-user/syscall.c 2004-12-05 17:03:18.278887920 -0800
@@ -3025,6 +3025,14 @@
     case TARGET_NR_get_thread_area:
         goto unimplemented_nowarn;
 #endif
+    case TARGET_NR_utimes:
+    {
+       struct target_timeval *target_tv = (void *)arg2;
+       struct timeval tv;
+       target_to_host_timeval(&tv, target_tv);
+       ret = get_errno(utimes((const char *)arg1, &tv));
+       break;
+    }
     default:
     unimplemented:
         gemu_log("qemu: Unsupported syscall: %d\n", num);

-----------------------------------------------------------------------------------------------------------------

On Saturday 04 December 2004 07:09 am, Paul Brook wrote:
> On Saturday 04 December 2004 00:05, James Pellow wrote:
> > Hi Paul,
> >
> > Is this closer to what you want?  I saw the swap functions.  I wasn't
> > sure what long translated to for all supported archs.  Do I need to do
> > 32/64 bit translation?  tv_sec and tv_usec are both long.
>
> There is already code to properly handle struct timeval, you should use
> that. See the code for TARGET_NR_settimeofday.
>
> Paul

-- 
*****************************
James A. Pellow, President
Alent Design Solutions
www.alentdesignsolutions.com
(509) 526-0682
*****************************

Re: [Qemu-devel] Syscall 269

[Paul Brook]

On Monday 06 December 2004 01:16, James Pellow wrote:
> Hi Paul,
>
> Ok, third time is the charm right...  It looks like the functions you
> referred to that handle struct timeval, are doing exactly what I had done. 
> Here is an updated patch that factors out the swap code using the existing
> function. Hopefully this is ready for inclusion.  Thanks for your help with
> this and for a truely incredible project!

Still not quite there. utimes takes [a pointer to] an array of two struct 
timeval.  You only thunk the first one.

Paul

Re: [Qemu-devel] Syscall 269

[James Pellow]

Hi Paul,

Ok, you got me that time.  I guess I have been trying to hury my way through 
this to get it working.  The man page I was looking at that defined utimes 
defines it as:

int utimes(char *filename, struct timeval *tvp);

If you go looking further in the page, a discussion is found about *tvp 
actually being a two member array.  I hadn't looked down that far I guess. (a 
bit red in the face...)  So, this should be it:

--------------------------------------------------------------------------------------------------------

diff -ruN qemu-0.6.1/linux-user/arm/syscall_nr.h 
qemu-0.6.1_new/linux-user/arm/syscall_nr.h
--- qemu-0.6.1/linux-user/arm/syscall_nr.h      2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/arm/syscall_nr.h  2004-11-18 00:58:44.000000000 
-0800
@@ -259,3 +259,5 @@
                                        /* 254 for set_thread_area */
                                        /* 255 for get_thread_area */
                                        /* 256 for set_tid_address */
+#define TARGET_NR_utimes                       (269)
+
diff -ruN qemu-0.6.1/linux-user/i386/syscall_nr.h 
qemu-0.6.1_new/linux-user/i386/syscall_nr.h
--- qemu-0.6.1/linux-user/i386/syscall_nr.h     2004-11-14 12:51:33.000000000 
-0800
+++ qemu-0.6.1_new/linux-user/i386/syscall_nr.h 2004-11-18 01:28:59.000000000 
-0800
@@ -271,3 +271,5 @@
 #define TARGET_NR_clock_getres (TARGET_NR_timer_create+7)
 #define TARGET_NR_clock_nanosleep      (TARGET_NR_timer_create+8)

+#define TARGET_NR_utimes               271
+
diff -ruN qemu-0.6.1/linux-user/syscall.c qemu-0.6.1_new/linux-user/syscall.c
--- qemu-0.6.1/linux-user/syscall.c     2004-11-14 12:51:33.000000000 -0800
+++ qemu-0.6.1_new/linux-user/syscall.c 2004-12-05 18:49:23.208271576 -0800
@@ -3025,6 +3025,15 @@
     case TARGET_NR_get_thread_area:
         goto unimplemented_nowarn;
 #endif
+    case TARGET_NR_utimes:
+    {
+       struct target_timeval *target_tv = (void *)arg2;
+       struct timeval tv[2];
+       target_to_host_timeval(&tv[0], &target_tv[0]);
+       target_to_host_timeval(&tv[1], &target_tv[1]);
+       ret = get_errno(utimes((const char *)arg1, tv));
+       break;
+    }
     default:
     unimplemented:
         gemu_log("qemu: Unsupported syscall: %d\n", num);

---------------------------------------------------------------------------------------------------------------------------

Thanks for your help.

-James Pellow

On Sunday 05 December 2004 05:26 pm, Paul Brook wrote:
> On Monday 06 December 2004 01:16, James Pellow wrote:
> > Hi Paul,
> >
> > Ok, third time is the charm right...  It looks like the functions you
> > referred to that handle struct timeval, are doing exactly what I had
> > done. Here is an updated patch that factors out the swap code using the
> > existing function. Hopefully this is ready for inclusion.  Thanks for
> > your help with this and for a truely incredible project!
>
> Still not quite there. utimes takes [a pointer to] an array of two struct
> timeval.  You only thunk the first one.
>
> Paul

-- 
*****************************
James A. Pellow, President
Alent Design Solutions
www.alentdesignsolutions.com
(509) 526-0682
*****************************

[Qemu-devel] Trivial (but useful) patch to save qemu pid to file

[Nile Geisinger]

Hi everyone,

Here's a very trivial patch that allows qemu's pid to be saved to a file. This 
is useful if you're using qemu as a service, instrumenting
it with other code, or doing other unixy things,

cheers,

Nile

---------------------------------------------------------------------------------------------------
diff -bur qemu-copy/vl.c qemu/vl.c
--- qemu-copy/vl.c      2004-12-07 22:11:54.898668344 -0500
+++ qemu/vl.c   2004-12-08 00:11:34.232243744 -0500
@@ -2549,6 +2549,7 @@
            "                (default is CL-GD5446 PCI VGA)\n"
 #endif
            "-loadvm file    start right away with a saved state (loadvm in 
monitor)\n"
+           "-pid file       save the qemu process id to a file\n"
            "\n"
            "During emulation, the following keys are useful:\n"
            "ctrl-alt-f      toggle full screen\n"
@@ -2619,6 +2620,7 @@
     QEMU_OPTION_prep,
     QEMU_OPTION_localtime,
     QEMU_OPTION_cirrusvga,
+    QEMU_OPTION_pid,
     QEMU_OPTION_g,
     QEMU_OPTION_std_vga,
     QEMU_OPTION_monitor,
@@ -2689,6 +2691,7 @@
     /* temporary options */
     { "pci", 0, QEMU_OPTION_pci },
     { "cirrusvga", 0, QEMU_OPTION_cirrusvga },
+    { "pid", HAS_ARG, QEMU_OPTION_pid },
     { NULL },
 };

@@ -2765,6 +2768,8 @@
     char serial_devices[MAX_SERIAL_PORTS][128];
     int serial_device_index;
     const char *loadvm = NULL;
+    int pid;
+    FILE * pid_file;

 #if !defined(CONFIG_SOFTMMU)
     /* we never want that malloc() uses mmap() */
@@ -3110,6 +3115,11 @@
             case QEMU_OPTION_full_screen:
                 full_screen = 1;
                 break;
+            case QEMU_OPTION_pid:
+               pid = getpid();
+               pid_file = fopen(optarg, "w");
+               fprintf(pid_file, "%d", pid);
+               fclose(pid_file);
             }
         }
     }

Re: [Qemu-devel] Trivial (but useful) patch to save qemu pid to file

[Felipe Sanchez]

On Wed, 8 Dec 2004, Nile Geisinger wrote:

> Here's a very trivial patch that allows qemu's pid to be saved to a
> file. This is useful if you're using qemu as a service, instrumenting
> it with other code, or doing other unixy things,


Hello Nile, a somewhat more complete patch for doing this was already
posted a while ago:

http://lists.gnu.org/archive/html/qemu-devel/2004-11/msg00450.html


Any comments are very welcome  :-)

Re: [Qemu-devel] Trivial (but useful) patch to save qemu pid to file

[Fabrice Bellard]

Felipe Sanchez wrote:
> 
> On Wed, 8 Dec 2004, Nile Geisinger wrote:
> 
> 
>>Here's a very trivial patch that allows qemu's pid to be saved to a
>>file. This is useful if you're using qemu as a service, instrumenting
>>it with other code, or doing other unixy things,
> 
> 
> 
> Hello Nile, a somewhat more complete patch for doing this was already
> posted a while ago:
> 
> http://lists.gnu.org/archive/html/qemu-devel/2004-11/msg00450.html
> 
> 
> Any comments are very welcome  :-)

I can include this patch if it helps some people as it seems rather non 
intrusive.

Fabrice.

Re: [Qemu-devel] Trivial (but useful) patch to save qemu pid to file

[Tim]

> I can include this patch if it helps some people as it seems rather non 
> intrusive.

I personally rather manage my daemons with something like daemontools or
runit: 
http://cr.yp.to/daemontools.html 
http://smarden.org/runit/

But like you say, it is pretty non-intrusive as it is off by default.
For some it will probably help out.

tim

Re: [Qemu-devel] Trivial (but useful) patch to save qemu pid to file

[Nile Geisinger]

Hi Fabrice,

First, thanks for QEMU! 

I find it useful and the previous patcher found it useful, but I'm fine 
patching for now. When a third person posts a patch, though, 
it should probably be included since it would be a sign that QEMU
is being increasingly instrumented by other programs,

cheers, 

Nile

On Wednesday 08 December 2004 10:01 pm, Fabrice Bellard wrote:
> Felipe Sanchez wrote:
> > On Wed, 8 Dec 2004, Nile Geisinger wrote:
> >>Here's a very trivial patch that allows qemu's pid to be saved to a
> >>file. This is useful if you're using qemu as a service, instrumenting
> >>it with other code, or doing other unixy things,
> >
> > Hello Nile, a somewhat more complete patch for doing this was already
> > posted a while ago:
> >
> > http://lists.gnu.org/archive/html/qemu-devel/2004-11/msg00450.html
> >
> >
> > Any comments are very welcome  :-)
>
> I can include this patch if it helps some people as it seems rather non
> intrusive.
>
> Fabrice.
>
>
>
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel