Re: BUG: /proc/self/exe reports relative paths, should always return absolute paths?

[Samuel Thibault]

Svante Signell, on mer. 20 sept. 2017 08:46:49 +0200, wrote:
> I changed the patch for __execve() to keep file_name if the path is absolute 
> and
> call realpath() for relative paths.

But then if you call a script that looks at $0 from a relative path,
it'll break.

> (I did also a cetcwd() version, but that does not seem to be needed, and using
> realpath() is simpler.)

It's simpler but doesn't work in the case above.

Samuel

Re: X Terminal issues

[Samuel Thibault]

Samuel Thibault, on jeu. 21 sept. 2017 00:24:28 +0200, wrote:
> Riccardo Mottola, on mer. 20 sept. 2017 23:53:17 +0200, wrote:
> > xterm: Error 27, errno 1073741849: Inappropriate ioctl for device
> > Reason: spawn: ioctl() failed on TIOCLSET
> 
> There are probably a lot more things like that which we shouldn't
> expose.

(At least those which are not defined on Linux)

Samuel

Re: X Terminal issues

[Samuel Thibault]

Hello,

Riccardo Mottola, on mer. 20 sept. 2017 23:53:17 +0200, wrote:
> xterm: Error 27, errno 1073741849: Inappropriate ioctl for device
> Reason: spawn: ioctl() failed on TIOCLSET

Mmm, this is indeed not implemented.

We should probably not expose it in bits/ioctls.h.

There are probably a lot more things like that which we shouldn't
expose.

Samuel

X Terminal issues

[Riccardo Mottola]

Hi HURDers,

Both if I start xterm as uxterm, I get these errors:

$ uxterm
xterm: Error 27, errno 1073741849: Inappropriate ioctl for device
Reason: spawn: ioctl() failed on TIOCLSET

$ xterm
xterm: Error 27, errno 1073741849: Inappropriate ioctl for device
Reason: spawn: ioctl() failed on TIOCLSET


uxterm used to work, but I belive xterm was fixed too.


Anyone else has issues?

My packages are upgraded as of today.

Thanks,
Riccardo

Re: [PATCH] eth-multiplxer: Implement ds_device_close()

[Joan Lledó]

Ooops!! I forgot that! :)

2017-09-20 20:06 GMT+02:00 James Clarke :
> On 20 Sep 2017, at 11:08, Joan Lledó  wrote:
>>
>> ---
>> eth-multiplexer/Makefile  | 2 +-
>> eth-multiplexer/device_impl.c | 5 +
>> 2 files changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/eth-multiplexer/Makefile b/eth-multiplexer/Makefile
>> index cefa0abd..2b19de6d 100644
>> --- a/eth-multiplexer/Makefile
>> +++ b/eth-multiplexer/Makefile
>> @@ -29,6 +29,6 @@ LCLHDRS = ethernet.h util.h vdev.h netfs_impl.h
>> HURDLIBS = ports ihash iohelp fshelp shouldbeinlibc netfs bpf
>> LDLIBS = -lpthread
>>
>> -CFLAGS += -I$(top_srcdir)/libbpf
>> +CFLAGS += -I$(top_srcdir)/libbpf -ggdb -O0
>
> I don't think you meant to include this hunk? :)
>
> James
>

Re: [PATCH] eth-multiplxer: Implement ds_device_close()

[James Clarke]

On 20 Sep 2017, at 11:08, Joan Lledó  wrote:
> 
> ---
> eth-multiplexer/Makefile  | 2 +-
> eth-multiplexer/device_impl.c | 5 +
> 2 files changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/eth-multiplexer/Makefile b/eth-multiplexer/Makefile
> index cefa0abd..2b19de6d 100644
> --- a/eth-multiplexer/Makefile
> +++ b/eth-multiplexer/Makefile
> @@ -29,6 +29,6 @@ LCLHDRS = ethernet.h util.h vdev.h netfs_impl.h
> HURDLIBS = ports ihash iohelp fshelp shouldbeinlibc netfs bpf
> LDLIBS = -lpthread
> 
> -CFLAGS += -I$(top_srcdir)/libbpf
> +CFLAGS += -I$(top_srcdir)/libbpf -ggdb -O0

I don't think you meant to include this hunk? :)

James

[PATCH] eth-multiplxer: Implement ds_device_close()

[Joan Lledó]

---
 eth-multiplexer/Makefile  | 2 +-
 eth-multiplexer/device_impl.c | 5 +
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/eth-multiplexer/Makefile b/eth-multiplexer/Makefile
index cefa0abd..2b19de6d 100644
--- a/eth-multiplexer/Makefile
+++ b/eth-multiplexer/Makefile
@@ -29,6 +29,6 @@ LCLHDRS = ethernet.h util.h vdev.h netfs_impl.h
 HURDLIBS = ports ihash iohelp fshelp shouldbeinlibc netfs bpf
 LDLIBS = -lpthread
 
-CFLAGS += -I$(top_srcdir)/libbpf
+CFLAGS += -I$(top_srcdir)/libbpf -ggdb -O0
 
 include ../Makeconf
diff --git a/eth-multiplexer/device_impl.c b/eth-multiplexer/device_impl.c
index d7c8beeb..b84a3309 100644
--- a/eth-multiplexer/device_impl.c
+++ b/eth-multiplexer/device_impl.c
@@ -96,6 +96,11 @@ ds_device_open (mach_port_t master_port, mach_port_t 
reply_port,
 kern_return_t
 ds_device_close (struct vether_device *device)
 {
+  if (device == NULL)
+return D_NO_SUCH_DEVICE;
+
+  destroy_vdev(device);
+
   return 0;
 }
 
-- 
2.11.0

[PATCH] eth-multiplxer: Implement ds_device_close()

[Joan Lledó]

Hello,

One difference between pfinet and lwip translators is that pfinet doesn't 
delete interfaces (or I haven't found how to do it) while lwip does. For 
instance, if one starts pfinet with:

settrans -fga /servers/socket/2 /hurd/pfinet -i /dev/eth1 -a 192.168.123.178 -m 
255.255.255.0 -g 192.168.123.1

and then runs:

fsysopts /servers/socket/2 --interface=/dev/eth2 --address=192.168.124.178 
--netmask=255.255.255.0 --gateway=192.168.124.1

the result is this:

root@hurd:/home/jlledom# fsysopts /servers/socket/2
/hurd/pfinet --interface=/dev/eth1 --address=192.168.123.178 
--netmask=255.255.255.0 --address6=fc00:123::5054:ff:feb6:1ab3/64 
--address6=fe80::5254:b6:1ab3/10 --address6=fe80::5054:ff:feb6:1ab3/10 
--gateway6=fe80::5054:ff:fef6:d496 --interface=/dev/eth2 
--address=192.168.124.178 --netmask=255.255.255.0 --gateway=192.168.124.1 
--address6=fe80::5254:4b:ad11/10 --address6=fe80::5054:ff:fe4b:ad11/10

fsysopts added a new interface. But doing the same with lwip leads to this:

root@hurd:/home/jlledom# fsysopts /servers/socket/2
/hurd/lwip --interface=/dev/eth2 --address=192.168.124.178 
--netmask=255.255.255.0 --gateway=192.168.124.1 
--address6=FE80::5054:FF:FE4B:AD11/64 --address6=FC00:124::5054:FF:FE4B:AD11/64

The stack is reset and reconfigured, and only the given interface is added. I 
implemented this by simply removing all interfaces anytime somebody calls 
fsysopts, and adding the new ones. If some interface already exists, it's 
deleted and created again.

This behaviour generated a problem when working with eth-multiplexer. Adding a 
new interface calls ds_device_open in eth-multiplexer and that returns a port 
name, whereas deleting the interface deallocates the port name in lwip, but 
nothing happens in eth-multiplexer since ds_device_close is unimplemented, when 
lwip tries to create the same interface again, eth-multiplexer returns the same 
port name, and trying to use it leads to a MIG_SERVER_DIED error. This use to 
happen when ifupdown uses dhclient to get an address and is not a problem in 
pfinet since it doesn't remove the interface.

I wrote a little patch to implement ds_device_close(). This patch is working 
for me, but I'm not sure whether it's the right way to do it or may affect 
other servers. I'd appreciate some feedback.

Regards,

Joan

Re: BUG: /proc/self/exe reports relative paths, should always return absolute paths?

[Svante Signell]

On Mon, 2017-09-18 at 20:56 +0200, Samuel Thibault wrote:
> Svante Signell, on lun. 18 sept. 2017 19:45:39 +0200, wrote:
> > lrwxr-xr-x 1 root root 26 Aug 17 07:38 /usr/sbin/update-fmtutil -> update-
> > tl-
> > stacked-conffile
> > 
> > In this case the link name should not be resolved. Any ideas?
> 
> Well, it should, but just not the symlink parts, only the cwd part.
> 
> Perhaps we could simply to check whether the file_name begins with a
> '/', in case we can keep as it is, or not, in which case we can just
> prepend the getcwd().

I changed the patch for __execve() to keep file_name if the path is absolute and
call realpath() for relative paths. Seems to work fine now, new patch attached.
(I did also a cetcwd() version, but that does not seem to be needed, and using
realpath() is simpler.)2.13-33 dates when this was added

TODO: _DEBIAN_ in versions however pose problem. Remove the _DEBIAN_ version
once packages are rebuilt against 2.21.

2010-08-04  Emilio Pozuelo Monfort  
	* hurd/hurdexec.c (_hurd_exec): Deprecate it.
	(_hurd_exec_file_name): New function.
	* hurd/hurd.h (_hurd_exec): Deprecate it.
	(_hurd_exec_file_name): Declare it.
	* hurd/Versions: Export it.
	* sysdeps/mach/hurd/execve.c: Use it.
	* sysdeps/mach/hurd/fexecve.c: Likewise.
	* sysdeps/mach/hurd/spawni.c: Likewise.



From d1793416cf8bf6fccd42679a8ec30b0058823ab8 Mon Sep 17 00:00:00 2001
From: Emilio Pozuelo Monfort 
Date: Sat, 22 May 2010 18:26:29 +0200
Subject: [PATCH] Use the new file_exec_file_name RPC

Pass the file name of executable to the exec server, which it needs to
execute #!-scripts.  Currently, the exec server tries to guess the name
from argv[0] but argv[0] only contains the executable name by convention.
---
 hurd/Makefile   |4 +-
 hurd/Versions   |4 ++
 hurd/hurd.h |   14 --
 hurd/hurdexec.c |   50 ++---
 sysdeps/mach/hurd/execve.c  |6 ++--
 sysdeps/mach/hurd/fexecve.c |7 ++---
 sysdeps/mach/hurd/spawni.c  |   59 ++--
 8 files changed, 102 insertions(+), 43 deletions(-)

Index: glibc-2.24-17.2/hurd/Versions
===
--- glibc-2.24-17.2.orig/hurd/Versions
+++ glibc-2.24-17.2/hurd/Versions
@@ -140,6 +140,14 @@ libc {
 _hurd_sigstate_unlock;
 _hurd_sigstate_delete;
   }
+  GLIBC_2.13_DEBIAN_33 {
+# "quasi-internal" functions
+_hurd_exec_file_name;
+  }
+  GLIBC_2.21 {
+# "quasi-internal" functions
+_hurd_exec_file_name;
+  }
 
   HURD_CTHREADS_0.3 {
 # weak refs to libthreads functions that libc calls iff libthreads in use
Index: glibc-2.24-17.2/hurd/Makefile
===
--- glibc-2.24-17.2.orig/hurd/Makefile
+++ glibc-2.24-17.2/hurd/Makefile
@@ -32,8 +32,8 @@ user-interfaces		:= $(addprefix hurd/,\
    auth auth_request auth_reply startup \
    process process_request \
    msg msg_reply msg_request \
-   exec exec_startup crash interrupt \
-   fs fsys io term tioctl socket ifsock \
+   exec exec_experimental exec_startup crash interrupt \
+   fs fs_experimental fsys io term tioctl socket ifsock \
    login password pfinet \
    )
 server-interfaces	:= hurd/msg faultexc
Index: glibc-2.24-17.2/hurd/hurd.h
===
--- glibc-2.24-17.2.orig/hurd/hurd.h
+++ glibc-2.24-17.2/hurd/hurd.h
@@ -241,12 +241,20 @@ extern FILE *fopenport (io_t port, const
 extern FILE *__fopenport (io_t port, const char *mode);
 
 
-/* Execute a file, replacing TASK's current program image.  */
+/* Deprecated: use _hurd_exec_file_name instead.  */
 
 extern error_t _hurd_exec (task_t task,
 			   file_t file,
 			   char *const argv[],
-			   char *const envp[]);
+			   char *const envp[]) __attribute_deprecated__;
+
+/* Execute a file, replacing TASK's current program image.  */
+
+extern error_t _hurd_exec_file_name (task_t task,
+ file_t file,
+ const char *filename,
+ char *const argv[],
+ char *const envp[]);
 
 
 /* Inform the proc server we have exited with STATUS, and kill the
Index: glibc-2.24-17.2/hurd/hurdexec.c
===
--- glibc-2.24-17.2.orig/hurd/hurdexec.c
+++ glibc-2.24-17.2/hurd/hurdexec.c
@@ -25,16 +25,37 @@
 #include 
 #include 
 #include 
+#include 
 #include 
 #include 
 
+#include 
+
 /* Overlay TASK, executing FILE with arguments ARGV and environment ENVP.
If TASK == mach_task_self (), some ports are dealloc'd by the exec server.
-   ARGV and ENVP are terminated by NULL pointers.  */
+   ARGV and ENVP are terminated by NULL pointers.
+   Deprecated: use _hurd_exec_file_name instead.  */
 error_t
 _hurd_exec (task_t task, file_t file,
 	char *const argv[], char *const