Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-16 Thread Christian Perrier
Quoting Michael Lueck ([EMAIL PROTECTED]):
> Further testing transferring large numbers of files works flawlessly, as
> expected.
> 
> Thanks for the .deb's, we will run on your version for now. For some reason
> when I compiled the patch in I arrived at a 7MBish smbd whereas you ended up
> with roughly the same smbd size, thus I trust yours more.


Well, "mine" is the 3.0.14a-1 package from Debian sarge with only the
patch you received added to debian/patches.

However, I recommend you to watch the official package : "my" package
was build quite quickly and I unfortunately gave it the 3.0.14a-2
version number. So, if the official Debian package happens to be
updated, yours will not be replaced by it.

Steve, Eloy, do you want me to upload 3.0.14a-2? This would be
recorded as a NMU (but the changelog doesn't mention it, indeed). Or,
if you prefer, Eloy, just rebuild a package with the patch I sent (and
adding me to Uploaders if you want:-))).





Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-15 Thread Michael Lueck
Further testing transferring large numbers of files works flawlessly, as
expected.

Thanks for the .deb's, we will run on your version for now. For some reason
when I compiled the patch in I arrived at a 7MBish smbd whereas you ended up
with roughly the same smbd size, thus I trust yours more.

Michael Lueck
Lueck Data Systems
http://www.lueckdatasystems.com/




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-15 Thread Christian Perrier
tags 309003 patch
thanks

Quoting Michael Lueck ([EMAIL PROTECTED]):
> I have downloaded and installed:
> 
> samba-common_3.0.14a-2_i386.deb
> samba_3.0.14a-2_i386.deb
> 
> The problem is resolved in this version. That is as much as I have tested at
> this time. I will be doing some major file transfers against the new code and
> report after that.
> 
> At this time we do not use the other .deb's in the Samba project.


Great. For Steve and Eloy, here's the patch I used.



--- samba-3.0.14a.orig/source/include/smb_macros.h  2005-04-13 
23:14:26.0 -0700
+++ samba-3.0.14/source/include/smb_macros.h2005-05-01 02:21:08.0 
-0700
@@ -225,16 +225,10 @@
split_at_last_component(full_keyname, key_name, '\\', subkey_name)
 
 /
- Used by dptr_zero.
-/
-
-#define DPTR_MASK ((uint32)(((uint32)1)<<31))
-
-/
  Return True if the offset is at zero.
 /
 
-#define dptr_zero(buf) ((IVAL(buf,1)&~DPTR_MASK) == 0)
+#define dptr_zero(buf) (IVAL(buf,1) == 0)
 
 /***
 copy an IP address from one buffer to another
--- samba-3.0.14a.orig/source/smbd/dir.c2005-04-13 23:14:20.0 
-0700
+++ samba-3.0.14a/source/smbd/dir.c 2005-05-01 02:21:08.0 -0700
@@ -622,11 +622,11 @@
DEBUG(1,("filling null dirptr %d\n",key));
return(False);
}
-   offset = TellDir(dptr->dir_hnd);
+   offset = (uint32)TellDir(dptr->dir_hnd);
DEBUG(6,("fill on key %u dirptr 0x%lx now at %d\n",key,
(long)dptr->dir_hnd,(int)offset));
buf[0] = key;
-   SIVAL(buf,1,offset | DPTR_MASK);
+   SIVAL(buf,1,offset);
return(True);
 }
 
@@ -639,16 +639,22 @@
unsigned int key = *(unsigned char *)buf;
struct dptr_struct *dptr = dptr_get(key, False);
uint32 offset;
+   long seekoff;
 
if (!dptr) {
DEBUG(3,("fetched null dirptr %d\n",key));
return(NULL);
}
*num = key;
-   offset = IVAL(buf,1)&~DPTR_MASK;
-   SeekDir(dptr->dir_hnd,(long)offset);
+   offset = IVAL(buf,1);
+   if (offset == (uint32)-1) {
+   seekoff = -1;
+   } else {
+   seekoff = (long)offset;
+   }
+   SeekDir(dptr->dir_hnd,seekoff);
DEBUG(3,("fetching dirptr %d for path %s at offset %d\n",
-   key,dptr_path(key),offset));
+   key,dptr_path(key),(int)seekoff));
return(dptr);
 }


Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-15 Thread Michael Lueck
I have downloaded and installed:

samba-common_3.0.14a-2_i386.deb
samba_3.0.14a-2_i386.deb

The problem is resolved in this version. That is as much as I have tested at
this time. I will be doing some major file transfers against the new code and
report after that.

At this time we do not use the other .deb's in the Samba project.

Thanks!

Michael Lueck
Lueck Data Systems
http://www.lueckdatasystems.com/




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-15 Thread Christian Perrier
OK, Michael, I finally compiled new packages for samba, with the
upstream patch you mentioned in your bug report.

Could you please download packages from
http://www.perrier.eu.org/debian/packages/samba-3.0.14a+309003/

Be patient : my uplink is 384kb so the transfer may be quite slow.






Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-14 Thread Christian Perrier
I'm building a samba package with the upstream patch applied.

Hopefully, I'll be able to provide supposedly fixed packages so that
the samba package maintainers can decide whether it's worth fixing
this for sarge.

Michael, I suppose you'll be able to test the fixed packages?



-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-14 Thread Michael Lueck
On Sat, 14 May 2005 18:08:32 +0200, Christian Perrier wrote:

>Michael, I suppose you'll be able to test the fixed packages?

Always happy to test a .deb package. I have a 100% reproducible test for this
bug.

Michael Lueck
Lueck Data Systems
http://www.lueckdatasystems.com/




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-14 Thread Michael Lueck
On Sat, 14 May 2005 07:21:53 +0200, Christian Perrier wrote:

>Please keep the bug number CC'ed to answers.

I considered it chatter / clarification not worthy of taking space in
bugs.debian.org

>Please note, besides this, that using XFS on testing, if you use it
>along with 2.6 kernels, is not a recommended situation, if you keep
>2.6.8 kernels shipped with sarge. On my own side, I had to upgrade the
>kernel to 2.6.10 from unstable, because of a grave bug in XFS support
>for 2.6.8 (no reference, sorry).

Ah, thanks for helping me sleep well! We have this configuration on a number of
servers with nadda issue at this time. Running both i386 and 686 2.6.8x
kernels. Possibly do you mean bug 292290?

How long have you been on the 2.6.10 kernel from unstable? Any side affects I
should be aware of?

Michael Lueck
Lueck Data Systems
http://www.lueckdatasystems.com/




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-13 Thread Christian Perrier
Quoting Michael Lueck ([EMAIL PROTECTED]):
> On Fri, 13 May 2005 22:47:03 +0200, Christian Perrier wrote:
> 
> >Which does not make the bug release critical
> 
> In the end it is your call. What, since it involves Samba + XFS it is OK to
> allow a copy forever loop into stable? Since we run all XFS it means we have 
> to
> handle Samba in a special way which is an ugly way to have Debian Sarge enter
> the deep freeze world of Stable. Long live Debian Testing!


Please keep the bug number CC'ed to answers.

The decrease in severity is not a way to judge that your bug report is
to be ignored. This is just assigning the bug the correct severity
with regard to your description.

No ignoring it : indeed, I'm likely to be affected by this bug report,
being in the exact situation you describe.

Given that this bug is restricted to XFS file systems, it may not IMHO
be assigned a RC severity.

This does not mean that the maintainers will not fix the bug. Fixing
important bug for non base packages is still allowed and, indeed, I
think it's worth fixing thios bug (as well as 302771 for which I tried
to give maintainers some input).

Please note, besides this, that using XFS on testing, if you use it
along with 2.6 kernels, is not a recommended situation, if you keep
2.6.8 kernels shipped with sarge. On my own side, I had to upgrade the
kernel to 2.6.10 from unstable, because of a grave bug in XFS support
for 2.6.8 (no reference, sorry).

In conclusion, I'mpretty sure that Steve and Eloy will give this bug
the attention it deserves.

Guys, I can again give some help here, for instance by building test
packages with the fix pointed by Michael for #309003.





Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-13 Thread Christian Perrier
severity 309003 important
thanks

Quoting Michael Lueck ([EMAIL PROTECTED]):
> Package: samba
> Version: 3.0.14a-1
> Severity: serious
> 
> Thanks for getting Samba in Debian up to the 3.0.14 level. Sadly there was a
> bug introduced between 3.0.11 and 3.0.14.
> 
> https://bugzilla.samba.org/show_bug.cgi?id=2622
> 
> Samba developers have a patch for this problem, I have compiled it and it
> resolves the problem. Is there any chance of getting this into the Sarge
> release?
> 
> Specifically it is an issue caused when XFS is the file system on the server.


Which does not make the bug release critical

  serious
  is a severe violation of Debian policy (roughly, it violates
  a "must" or "required" directive), or, in the package
  maintainer's opinion, makes the package unsuitable for relea
  se.

   important
  a bug which has a major effect on the usability of a package
  , without rendering it completely unusable to everyone.





-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]



Bug#309003: Copy Forever Loop bug in 3.0.14a

2005-05-13 Thread Michael Lueck
Package: samba
Version: 3.0.14a-1
Severity: serious

Thanks for getting Samba in Debian up to the 3.0.14 level. Sadly there was a
bug introduced between 3.0.11 and 3.0.14.

https://bugzilla.samba.org/show_bug.cgi?id=2622

Samba developers have a patch for this problem, I have compiled it and it
resolves the problem. Is there any chance of getting this into the Sarge
release?

Specifically it is an issue caused when XFS is the file system on the server.

Michael Lueck
Lueck Data Systems
http://www.lueckdatasystems.com/




-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]