Hello Dale,
Dale Dude:
Can aufs be used in its current form, or be modified to do so, to join a
number of file systems, as it already does, but if writing to a full rw
branch it will detect this and start writing to another rw branch? Or
possible round robin all file creates to all rw
Hi,
Jrgen_P._Tjern:
I found this in my dmesg: (aufs 20070709)
aufs may_rename_srcdir:401:mv[28278]: renaming dir who has child(ren) on
multiple branches, is not supported
I guess this is a non-implemented feature? It should give more
information about what dir this occured on. Also, what
o news
- the dlgt option is revealed.
o bugfix
- bugfix: handling whiteout in dlgt mode.
- bugfix: support umask in dlgt mode.
Junjiro Okajima
--
Index: fs/aufs/i_op_add.c
Index: fs/aufs/i_op_ren.c
- bugfix: handling
Andrew Burgess:
kernel 2.6.22.5
aufs version: ???
The History file is dated :
Jul 15 13:08 History
Was your aufs configured and compiled correctly?
While I didn't try 2.6.22.5, your stacktrace showed the failure of
initializing aufs module which is not a filesystem operation.
I
Hi,
Tomas M:
I'd like to suggest to use it as default for 'squashfs' filesystem as
well. If it is already implemented, then I'd like to sugget to update
aufs.5 man page :)
RR is set by default for squashfs, and it is described in the manual.
Oh, I've found the manual on the web has a
Tomas M:
I downloaded aufs from CVS today and I compiled it,
then I use 'man -l ./aufs.5' and it shows the following:
Set by default for a branch whose fs-type is either 'iso9660', 'udf',
(and here the paragraph ends, no more text)
It was caused by a single quotation mark. Sorry.
Here is
Hi,
Jan-Marek Glogowski:
Tested 2.6.23-rc6 - didn't solve the BUG.
Here are some patches for 2.6.23-rc6 and the latest aufs.
Please try, and let me know the result.
- a.patch
for aufs
- put_filp.patch
for kernel, required.
- splice.patch
for kernel, optional, when you use loopback mount
o bugfix
- bugfix: whiteout list may be obsoleted while aufs was sleeping,
reported by Torsten Luettgert.
o news
- begin supporting linux-2.6.23, still testing on rc6.
+ support splice(2).
+ introduce aufs_fault().
+ new patches and configurations,
* splice-2.6.23.patch +
Torsten Luettgert:
Btw: I found a few lines like these in the syslog:
aufs rmdir_whtmp:908:rpm[14483]: failed removing .wh..wh.xdg.0018(-39),
ignored
aufs renwh_and_rmdir:178:rpm[14483]: I/O Error, rmdir .wh..wh.xdg.0018,
b0 failed, -39. ignored
That sounds like the race did occur, but
Hi,
Matt Jarvis:
Thanks for your reply - the application creates a new logfile every
minute, and only logfiles which are closed are copied down to the
branch. My understanding of udba=inotify was that it had a higher system
overhead - is that correct ?
Then I'd suggest you to try
Hi,
Jan-Marek Glogowski:
Just tested the latest CVS snapshot.
:::
I also tested without splice and with NFSv3 - no difference.
Will you try linux-2.6.22 or earlier with NFSv3 if you can?
I have attached the log of the BUGs. And it always freezes when trying to
run a larger program
Jorgen,
Thank you for sending logs.
Let's change the thread and continue the discussion a little more.
I think that make the xino files to be persistent is most hopeful, but
it may be difficult to ensure the branch set is perfectly the same.
With the persistent xino files will allow you to:
#
Jrgen_P._Tjern:
When I mount again, it will find files in /tmp/.aufs.xino.dir/ that
matche the uuid of the members of the branch set specified, and use
those as a base (but creating new xino-files for uuids that are not in
/tmp/.aufs.xino.dir/).
Does that make sense? Would it be doable?
Hello James,
James Willcox:
I encountered the following kernel BUG today when removing a RPM package
(with 'rpm -e package'). I am running aufs from the 13th of this
month on kernel 2.6.16. Any help is greatly appreciated.
Is you kernel vanilla 2.6.16?
And you aufs version is 20070910,
Hello Julian,
Julian Andres Klode:
Actually, the problem is a namespace collision with some arch/ia64
code. si_flags is a macro defined in include/asm-ia64/siginfo.h. I've
uploaded an NMU that changes the name of the field to
aufs_si_flags. Patch attached.
Thank you for your report and
o bugfix
- bugfix: name collision on ia64, reported and patched by Julian
Andres Klode, Daniel Baumann, Otavio Salvador and dann frazier (I
replaced si_flags by au_si_flags insead of aufs_si_flags).
- bugfix: when aufs is liked statically, /tmp/aufsvars.mk is
unnecessary, reported by Tapani
Hello flip,
[EMAIL PROTECTED]:
CC fs/aufs/super.o
fs/aufs/super.c: In function `aufs_fill_super':
fs/aufs/super.c:667: warning: comparison is always false due to limited
range of data type
:::
but that should not be a big thing right?
Your compiler produces a wrong line
[EMAIL PROTECTED]:
Im still stuck :-(
today i download the 2.6.15 source, pached it and tried to
compile it native (x86).
Here is a patch for you while I didn't test it.
Junjiro Okajima
--
Index: fs/aufs/super.h
Bertrand D:
My mistake. In fact everything is fine now that i deactivated
CONFIG_AUFS_DEBUG. It seems that this option made aufs much more verbose in
the latest version than in the previous one i used.
It was my fault to leave messages for debugging nfsv4 in the release
version. Sorry.
I
Luis Carlos Erpen de Bona:
Yes you found it.
Now it's working correctly!
Glad to hear that.
Thank you for your bug report and repeated tests.
Junjiro Okajima
-
This SF.net email is sponsored by: Splunk Inc.
Still
o bugfix
- bugfix: at decoding a nfs handle, in the slowest way which will be
invoked after other two method's failure, aufs generates the path
for the decode target inode. and there was a bug in generating the
path. when the inode is decoded by the two preceeding methods, you
will not
Hello Robin,
Robin Lee Powell:
I have a slightly odd use case in mind, one in which it would be
very useful to be able to tell which branch a particular file is
from (that is, the highest priority branch the file first appears
in).
Try CVS_TREE/aufs/sample/unionctl --whereis
Hello Jim,
Jim Garlick:
Just gave aufs 20071015 a test and had some difficulties:
It might be an aufs bug introduced by recent changes.
But let me make sure. Did you enable CONFIG_AUFS_PUT_FILP_PATCH? It was
not on your list.
Junjiro Okajima
Michael Towers:
[EMAIL PROTECTED] wrote:
:::
Yes I will have to manipulate the mountpoints I guess.
It is possible that I'll have to unmount aufs, which will
be one layer below.
Do you/anybody see a way how this could be achieved?
:::
initramfs root. But why would you
Michael Towers:
Could it be that there is a difference in regard to this in pivot_root
and switch_root? I haven't used root aufs for a while so I'm not sure
exactly what code I used but I don't remember using mount --move and I
don't remember any problems.
In the case of pivot_root, you
Hello Dale,
Dale Dude:
Im mounting aufs over nfsv3 mounts. I use the below cmd line to do so. I
get stale nfs handle errors sometimes with programs that overwrite
existing files. Our stuff manipulates and recreates cache files (like
web templates) a number of times a day. On the first
Jim Garlick:
My apologies - after I applied PUT_FILP and fixed some other configuration
issues, my problems went away. The problem is with Red Hat's EL5 kernel
which claims to be 2.6.18 but includes backported features from newer
kernels. The patch below made 20071015 work for me.
Glad
o security
- security enhancement or possible bug fix: support FMODE_EXEC and
{deny,allow}_write_access() for an executable and put_write_access()
for an open file for writing on a branch. for linux-2.6.17 and later.
o bugfix
- bugfix: replace a direct access to dentry-d_parent by
Hello Ulrich,
Ulrich Holeschak:
i have now used aufs for a very long time, and it was working perfectly!
But now with the update from this monday i have a problem. With the
version from last week everthing was working well.
:::
Normally i a using CONFIG_AUFS_DLGT, but to make things
Jakub_W._Jozwicki:
I'm using linux 2.6.23-gentoo-rt1.
I needed to modify rt-compat.patch:
Thank you, Jakub.
Tapani, how do you think his patch is?
Some screens:
http://petzold.republika.pl/utils/hl1.png
http://petzold.republika.pl/utils/hl2.png
Also, I hope you are not confusing aufs
Hello Markus,
Markus Nilsson:
I guess this should impose no problems. However, I would need to merge =
the
changes in the r/w branch to the base every night or so to not loose =
changes
done to the system. I guess this should be done by
I have heard someone else had ever tried similar
I had one mistake in my last release.
release(2) or mv(1) was totally wrong.
Sorry.
I have fixed it now.
- bugfix: mis-typing dentry and src_dentry. totally wrong.
Junjiro Okajima
-
This SF.net email is sponsored by:
o news
- introduce a new semantics for link(2) and rename(2), and three
policies to select one among multiple writable branches (completed).
+ implement a new create policy 'mfsrr.'
(quote from the aufs manual)
Policies to Select One among Multiple Writable Branches
Aufs has some
Jrgen_P._Tjern:
Unfortunately, my poor English doesn't allow me to understand the
difference between what you want and create=3Dtdp.
Well, the concept is a bit hard to explain ;-)
Probably it would be hard too even if you wrote in my native
language. :-)
Oh, right. Well, then I'd really
Julian Andres Klode:
This is known. The Ubuntu kernel uses AppArmor patches.
Please see https://bugs.edge.launchpad.net/ubuntu/+source/aufs/+bug/14073=
5
Maybe sfjro could add a patch for Ubuntu's kernel to patch.
Is there any symbol like UBUTSU_VERSION or CONFIG_APPARMOR which I can
use
Hi Russell,
Russell Harmon:
Below is a patch to fix the compiler warning introduced by
http://lists.openwall.net/linux-kernel/2007/08/18/120
Actually the patch in the URL you wrote is for f_version in struct file,
while the problem you are addressing is i_version in struct inode.
As we
Jason Lunz:
I'm running aufs 20061126 on linux i386 2.6.22.14. The machine I'm using
now is a vmware VM with 2 CPUs, but I've also reproduced similar crashes
on real x86 hardware with older aufs versions.
I could reproduce the problem.
Give me some time.
Junjiro Okajima
Jason Lunz:
- tmpfs + loopback ext2 fs-image + mount
I'm not sure what you mean by this.
Basically it is the same thing to the ramdisk + mke2fs case.
And I understand it useless for you.
For what? Is it just to avoid changing a multiply-hardlinked inode more
than once? If so it's just
Jason Lunz:
The patch works on physical smp hardware as well.
Thank you for your report.
I will include the patch in next Monday release.
Junjiro Okajima
-
SF.Net email is sponsored by: The Future of Linux Business White
Jason Lunz:
If I mount aufs with noxino, it doesn't work. But if I mount, then use
-o remount,noxino, noxino is enabled.
This is a regression from earlier versions of aufs - mount with noxino
worked for 20070312 and 20070709.
Yes, it is a bug.
Try this patch.
The line number may be
Hello,
[EMAIL PROTECTED]:
error
[ cut here ]
kernel BUG at /home/ico2/src/aufs/fs/aufs/cpup.c:456!
It is an unexpected situation related to 'pseudo-link' (see the aufs
manual in detail).
It may happen if your writable branch is one of these
Hi Sassan,
[EMAIL PROTECTED]:
When I disable debugging, I get these errors on inserting the modules:
aufs: Unknown symbol aufs_cond
aufs: Unknown symbol au_sysrq_fin
aufs: Unknown symbol au_sysrq_init
insmod: cannot insert '/lib/modules/aufs.ko': Unknown symbol in module
(-1): No such file
[EMAIL PROTECTED]:
I updated my source.
It still failed.
/tmp/aufsvars.mk does not exist.
Did make -f local.mk clean
Still failed
redownloaded local.mk and edited again, then make clean again
Still failed.
Will you show me what you did step by step, especially the changes in
local.mk?
[EMAIL PROTECTED]:
rm local.mk
cvs up
edited local.mk as shown below
CONFIG_AUFS_DEBUG = y
CONFIG_AUFS_DEBUG = n
Try 'CONFIG_AUFS_DEBUG =' (empty) instead of 'n'.
Junjiro Okajima
-
SF.Net email is sponsored by: The
[EMAIL PROTECTED]:
For example,
cd /mnt
mkdir hdc
mkdir: cannot create directory 'hdc': No such file of directory.
If you really did nothing around pseudo-link, then there must be a bug
in aufs.
Will you send the output of 'ls -li' for 'mnt' on all branches?
Junjiro Okajima
[EMAIL PROTECTED]:
the directory:
13 drwxr-xr-x2 00 60 Dec 3 04:24 ennui
in the read only branch:
1 drwxr-xr-x1 00 0 Jan 1 1970 ennui
Do you mean the second output is for 'mnt', instead of 'ennui'?
Still I don't understand
[EMAIL PROTECTED]:
Ah. The branch is cramfs and these are empty directories.
Would cloop work any better than cramfs? Or another version of cramfs
(you mention that it is the one in linux stable).
Actually I didn't test cloop on aufs by myself, but someone else might
use it. Anyway I don't
Hi,
Tomas M:
May I ask you, what is the trunc_xino useful for?
I didn't find anything in aufs.5 documentation,
only trunc_xib and notrunc_xib, but that is perhaps
something else, am I right?
Right.
Because I don't know whether this feature is required or not, I didn't
write the
Hi Heinrich,
Heinrich Rebehn:
# mount -n -t aufs -o dirs=/spec=rw:/common=nfsro union $rootmnt:
NFS branch is not supported, try some configurations and patches
included in aufs source CVS
? What does this mean? What some configurations and patches should
i try?
It means,
- your
Russell Harmon:
P.S. Junjiro, you may want to have a look at the
aufs-no-patch-in-kconfig.patch. It has a few more minor changes that I did
not mention which you may want to consider changing in next monday's
release.
Russell,
Thank you for your patches.
I shall fix/add the relation of
Heinrich Rebehn:
BTW, against which version of aufs did you create the patch? If i want
to check out that particular version, which cvs command do i have to use?
You said that your patch is not tested, since you are not an Ubuntu
user. Does that mean that Ubuntu will not be supported in the
o bugfix
- bugfix: AUFS_LHASH_PATCH depends on AUFS_FAKE_DM = n, reported by
Russell Harmon and Heinrich Rebehn.
- bugfix: a xino-truncation thread must not wait for other threads.
o misc
- add a note about setting 'n.'
- move all the declaration from super.h to branch.h.
- rename struct xino
Julian Andres Klode:
compiling aufs with the patch for ubuntu worked for kernel 2.6.22-14.46, =
but does not work for Ubuntu's
2.6.24-1.2. I attached the build log.
Thank you for your reports.
I will develop another patch for new ubuntu kernel after both of
linux-2.6.24 (no -rc version) and
I see XFS in linux-2.6.24-rc[34] changed the semantics of inode
attribute (i_rdev).
Here is a patch for it.
If you succeed, the patch will be included in next Monday release.
I found a bug in this patch.
Please use this newer one.
Junjiro Okajima
a.patch.bz2
Description: BZip2 compressed
Hello Michael,
Michael S. Zick:
There is no #define TEMPFS_MAGIC in mm/shmem.c of 2.6.24-rc6
The comment about it being used by glibc remains, but the define is _gone_.
I downloaded
ftp.kernel.org/pub/linux/kernel/v2.6/testing/linux-2.6.24-rc6.tar.bz2
just now, and found there still exists
Hello Mike,
mike:
The patch does not work for me.
I wrote this patch, and it works for me:
Thanx for your patch.
I also updated the patch after exchanging personal mails between
Tommy[D], and he reported it worked.
Please try this one.
Junjiro Okajima
Hello Michael,
Michael Haas:
Is the lhash patch actually still needed? If yes, is there an updated
version floating around somewhere?
Basically, yes it is required as long as you use NFS as an aufs branch.
But I didn't try linux-2.6.24-rc7 yet.
I will create an update lhash patch when
Michael S. Zick:
Jan 11 (my local date) - In-line and attached.
As noted in another post, this applies with an offset if you do not have
the linux-vserver patch installed. Offset is ok, ignore (or edit the line
number).
Also attached here. Whitespace changes in kernel source from
Hello Gabor,
Gabor Laszlo Zsolt:
The system works ok, even if it took me a while to figure this way out , but
I can't change nothing on /etc or /var because I can't unmount them
I remount the / as rw of course before.
For umount /var I get
/var must be outside
mount : var not mounted
Gabor Laszlo Zsolt:
/sys/fs/aufs/brs :
none /etc dfe6xxx br:/mnt/rw/etc=rw:/etc=ro
none /var f7d2xxx br:/mnt/rw/var=rw:/var=ro
trying to access the original /etc and /var
How can I do that?
You can't, because you hide them by over-mounting.
How about these steps?
mount /dev/xxx /ro/etc
o news
- new sample chroot-ed/jail environment useing aufs.
o bugfix
- bugfix: consolidate make targets for parallel build (make -j3),
reported and tested by Tommy[D] and mike.
- bugfix: ignore an inotify event for the grand parent dir timestamps.
- bugfix: stop a race condition between
Jrgen_P._Tjern:
I have a aufs mount at /storage, and a few mount --binds pointing to
subdirectories of /storage. I did a umount -lf /storage, which happened
without errors, but then did umount -lf
/some/mount/--bind/pointing/to/storage, and the following kernel BUG
appeared. :-)
Let me make
Hello Michael,
Dipl.-Ing. Michael Niederle:
I check all files whether they are opened or not before removing them from
the
ramdisk-layer. It seems that I must not remove any file within a directory if
only a single file within that directory is kept open - otherwise I end up
with
an
Michael S. Zick:
Could you include a demonstration of the proper mount options and procedure
in the samples/auroot/{README,script} please?
Perhaps pretend that apache had a directory that accept uploads/puts -
something
that had to be moved from the top rw to the lower ro from outside of
o bugfix
- bugfix: support for lazy unmount on linux-2.6.18 and later, reported
by Jorgen P. Tjerno.
- bugfix: a race condition between udba and aufs (unfnished).
o misc
- debugging a race condition between udba and aufs.
Junjiro Okajima
Julian Andres Klode:
compiling aufs with the patch for ubuntu worked for kernel 2.6.22-14.46, =
but does not work for Ubuntu's
2.6.24-1.2. I attached the build log.
Thank you for your reports.
I will develop another patch for new ubuntu kernel after both of
linux-2.6.24 (no -rc
Julian Andres Klode:
It used to be possible to build aufs without the full kernel sources,=20
but now this doesn't work, I get the following:
:::
See the attached patch.
:::
-EXTRA_CFLAGS +=3D -DXFS_SB_MAGIC=3D${xfs}
+EXTRA_CFLAGS +=3D -DXFS_SB_MAGIC=3D0x58465342
Right.
Or
o bugfix
- bugfix: reset f_mapping for memory mapped files.
+ new function au_vm_ops().
+ reset f_mapping in mmap().
+ use f_mapping-host instead of f_dentry-d_inode in sync().
o news
- enable AUFS_SUPER_MAGIC.
o misc
- test discarding cached fuse info.
- use f_mapping-host instead of
alpha_one_x86:
Nfs v3: ok
Aufs + tmpfs: ok
Aufs + nfs v3: failed
I guess what you wanted to say is 'nfsv3 for aufs writable branch was
failed.'
*I have a lot of when for exemple I run kde:*
aufs au_lkup_neg:631:aufsd[225]: I/O Error, b0 share should be negative.
Probably you already know
Yair Yarom:
When running nscd (Name Service Cache Daemon) with /etc mounted from nfs and
tmpfs (aufs), the following message appears:
aufs aufs_fallocate:699:nscd[3518]: This operation is not supported. Please
report me this application.
Would you try this patch?
Actually this path doesn't
Tomas M:
I just received several reports (from 5 different people) that they are
trying to build a Live CD, and their squashfs filesystems seem corrupted
(while extracting the squashfs image using unsquashfs results in 100% correct
data).
So it seems to me like a problem with squashfs
Hello Robby,
Robby Workman:
tarball. However, since there isn't (or we can't find) a stable
release of aufs available, we are curious if you guys would mind
us hosting a cvs snapshot on our server?
Why do you (or people) think CVS as 'unstable snapshot?'
As I wrote in README file, I am
Hello Dirk,
Dirk Herrendoerfer:
Apps or libs on the TMPFS branch in aufs do not work.
:::
Kernel:
2.6.17.6 + squashfs-3.3
Aufs is compiled into the kernel. ( dmesg: aufs 20080211 )
Sorry, it is a bug of aufs version 20080211 and later which doesn't
support mmap(2) on tmpfs in
o bugfix
- bugfix: revert the changes about mmap handling in 20080211, but
handles fuse individually.
+ remove reiser4.patch.
- bugfix: support ftruncate(2) (completed).
o news
- support fallocate(), actually aufs doesn't implement it as many
other filesystems do, reported by Yair Yarom.
-
Tomas M:
Yes I enable it everytime.
I am compiling aufs using the same script for many months (which sets
CONFIG_* parameters in local.mk and runs make) ... A simple downgrade to
older aufs version vanishes the problem, so I think there is something
changed in new aufs.
I will try to
Tomas M:
I am trying to use newest aufs, using 2.6.24.2 and today's Monday aufs
release.
'modprobe aufs' shows error message 'unknown symbol' and dmesg shows that it
doesn't know symbols do_vfsub_splice_from and do_vfsub_splice_to,
Sorry. It was my fault.
I have fixed and commited source
Hi,
Michael Towers:
What I want to do is rebuild the layer2 squashfs with layer1 merged in,
retaining all relevant whiteouts. I can of course do this by copying
everything to a new directory and resquashing. But it seems to me it
would be much easier (and I suppose much quicker, as well
Hello Marcel,
Marcel Gerrits:
It would be nice if there was the opportunity to approve the changes
in the top layer and move those approved changes to the middle layer to
make them permanent.
Is there something available to move between layers, or is that already
supported somehow by
Because I am busy to get my daily food. (Please make donations)
Here are the contents of planning next release.
- support the pseudo-link at exporting aufs via NFS.
- bugfix: a race condition between UDBA and aufs.
- a new sample for vmware.
Junjiro Okajima
Klaus Fuerstberger:
So why is /var/lib empty?
Does it exist on your branch fs?
If so, you should check the upper whiteouts first.
Junjiro Okajima
-
This SF.net email is sponsored by: Microsoft
Defy all challenges.
Klaus Fuerstberger:
Sorry, forgot to say. The patch does not help. The same problem, another
error in syslog.
~# mount -o remount,rw /ram/rootfs/
mount: /ram/rootfs not mounted already, or bad option
Mar 5 17:52:29 alixtest kernel: aufs au_lkup_neg:631:anacron[1900]: I/O
Error, b0
Hello John,
Kernel John:
I'm trying to make inotify work with aufs. However... udba=notify mount
option does'nt work for me. I'm obviously doing something wrong. Here is how
i compile my kernel:
M Another
unionfs
? ?--- These options are generated automatically for
Kernel John:
How many files can i handle with aufs + inotify? I'm asking because i've got
a lot of them. One of my physical partitions has got about 100 000 files
stored on it. This partition is mounted on some directory, and this
directory is inserted to union. I'd like to be able to change
o bugfix
- bugfix: support complex whiteout/diropq case, when an existed upper
entry is not a directory, skip the test for diropq, reported by
Michael Towers.
- bugfix: a race condtion between UDBA and aufs.
+ check inode generation too
+ update file generation too
+ merge updating
Jeff Mahoney:
Well it depends, if the sysfs files for brs are supposed to mimic the
/proc/mounts interface, why not use /proc/mounts for it? There's the
- -show_options callback now that should work perfectly for that, yeah?
Because /proc/mounts has a size limitation.
Some aufs users had met
Jeff Mahoney:
Because /proc/mounts has a size limitation.
Some aufs users had met this limitation and I developed
/sys/fs/aufs/brs.
What's the limitation? It uses seq_file, and passes the seq_file handle
to -show_options.
The maximum size of seq_file is PAGE_SIZE by default.
Junjiro
Tomas M:
And after then, information for every fuse-mounted filesystem appears there.
So it seems to me FUSE doesn't like sysfs as well, and thus it implements a
workaround by using a different 'virtual' filesystem.
So you could perhaps do the same for aufs?
Thank you for your
Jeff Mahoney:
Yeah, that's what I was saying. A regular seq_file will expand, but
since -show_options is called from within another -show operation, it
can't.
As far as I know a regular seq_file doesn't expand automatically.
Will you point me why do you think so?
Junjiro Okajima
[EMAIL PROTECTED]:
The real cause was NOT in aufs (or unionfs), but in squashfs, more exactly
due to the LZMA compression, more information in
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=456489
It seems that debian people failed to apply sqlzma patch for squashfs,
or to build their
Hello Bruno,
Bruno Cesar Ribas:
now the problem appears with aufs:
8---
[EMAIL PROTECTED]:~$ cd /var/lib/php5
[EMAIL PROTECTED]:/var/lib/php5$ ls -ld
drwx-wx-wt 2 root root 6 2007-05-13 15:30 .
[EMAIL PROTECTED]:/var/lib/php5$ ls
ls: .: Permission Denied
[EMAIL
Bruno Cesar Ribas:
$ ls -ld $RW/$d $ADICIONAL/$d $GERAL/$d $MAINROOT/$d
That's where we get the strange part, the RO ($MAINROOT) branche has this
directory with same permissions!
How are the export options on nfs server?
If you set ro export option for your readonly branches, then aufs
Jeff Mahoney:
After a short conversation with Greg KH, he said there is no way to
have files larger than PAGE_SIZE in sysfs. bin_attribute is for
feeding binary firmware to device drivers to push down to the hardware
and should not be abused as a workaround to implement larger files. The
Bruno Cesar Ribasさん:
Yep I export RO branches as NFS RO. I weren't using `dirperm1', but even
using `dirperm1' i get same error.
dirperm1 requires CONFIG_AUFS_DLGT.
What is 'funny' is that these things were working in the past! Nothing
Changed (besides kernel)..
Will you tell me
Bruno Cesar Ribas:
This happened since I cannot use CONFIG_AUFS_FAKE_DM if I have NFS as an
aufs branche (CONFIG_AUFS_LHASH_PATCH) and I upgraded from kernel 2.6.23 to
2.6.24.2=20
It seems very old... it is really old when nfs branch requires disabling
FAKE_DM.
Anyway I could reproduce your
Jeff Mahoney:
I think the solution for this should just be to include those functions
in compat.h. It's your project, of course, but I think the goal for
compatibility should be to code for the latest release and provide shims
to provide access to earlier kernels. Sometimes that's just not
Bruno Cesar Ribas:
Thanks, it is working =)
Glad to hear that.
The patch will be refined and included in next release.
This is our new challenge 8^)
Ok, let's go. :-)
And machine load began to grow, we rebooted machine while it was 39. But we
already got machines with 200 of load =P
o bugfix
- bugfix: in case of testing super-io, stop forcing permission-check
for nfs branch, reported by Bruno Cesar Ribas.
+ extract a part of forcing permission check from au_test_perm(),
and create a new function au_test_perm_sio().
- bugfix: support a previous failure of setting
Julian Andres Klode:
I attached two small patches from the Debian package, fixing bashisms in =
auplink
and some things related to security-inode-permission.
Thanx.
The former will be included in next Monday release, but the latter.
Skipping security_inode_permission() can be a security hole,
Julian Andres Klode:
I attached two small patches from the Debian package, fixing bashisms in =
auplink
and some things related to security-inode-permission.
:::
diff -urNad aufs-0+20080317~/util/auplink aufs-0+20080317/util/auplink
--- aufs-0+20080317~/util/auplink 2008-03-10
Julian Andres Klode:
Find() # dir [ find_args ]
{
- local dir=3D3D$1
shift
- find $dir -xdev -noleaf $@
+ find $1 -xdev -noleaf $@
}
:::
- in Find(), this patch makes $@ to include $1.
No. This is done via shift. shift removes $1 from $@ and makes $1 to be t=
he
101 - 200 of 1884 matches
Mail list logo
