This list has been deprecated. Please subscribe to the new devel list at
lists.nfs-ganesha.org.
Frank Filz wrote on Thu, Apr 26, 2018 at 12:28:31PM -0700:
> I'm guessing the testing has no way to track a cumulative state, so it would
> work better to have separate IDs.
Another possibility would
Hi,
Girjesh Rajoria wrote on Mon, Mar 19, 2018 at 10:47:09PM +0530:
>> + tail -21 ../dbenchTestLog.txt
>>
>> OperationCountAvgLatMaxLat
>> --
>> Deltree102 9.79927.590
>> Flush
Hi all,
I tracked this crash down and submitted a couple of patches here:
https://github.com/nfs-ganesha/ntirpc/pull/120
These are NOT good as they are: I'm fixing the obvious problem, but
removing the element isn't safe all the time as I wrote in the PR.
Bill, do you know how to fix that
Daniel Gryniewicz wrote on Tue, Mar 13, 2018 at 10:07:33AM -0400:
> An hour later...
Nope, it is an hour earlier for us :)
--
Dominique
--
Check out the vibrant tech community on one of the world's most
engaging tech
Hi Jeff,
the CEA bot has hit this twice in the past two or so weeks, so you're
definitely not the only one seeing that -- unfortunately it's only ever
hit it on the runs without ASAN so the traces are pretty much the same
as what you get.
This kind of messages mean we're messing about with
Hi Frank,
Frank Filz wrote on Tue, Feb 27, 2018 at 07:11:13AM -0800:
> It seems like our mailing list is broken. We are working on a replacement.
Any progress on the replacement front ?
The auto-response is good but some announce here for silent folks like
me can't hurt, if one of us can get it
Daniel Gryniewicz wrote on Fri, Jan 12, 2018 at 11:19:27AM -0500:
>> As to Lustre, I'm pretty sure the stackable FSAL will work fine. A stacked
>> FSAL does have the pointer to the subfsal object, and knows which FSAL is
>> the subfsal, so in this case where FSAL_LUSTRE MUST stack on top of
Hi Frank, others,
Frank Filz wrote on Tue, Jan 09, 2018 at 03:19:46PM -0800:
> All in all, these functions are not a great fit for the FSAL API so I'm not
> sure it would be a good solution. Forcing some of the functions into FSAL
> methods would require some code duplication that loses some of
Niels de Vos wrote on Tue, Dec 19, 2017 at 11:19:11AM +0100:
> These are both running in the CentOS CI. Unfortunately I do not know how
> to abort running jobs when a new patch set is posted. Dominique, do you
> have any hints for that?
I'm afraid I don't really know, the gerrit plugin just does
Frank Filz wrote on Fri, Dec 15, 2017 at 04:58:49PM -0800:
> I just got a CI failure for version 2 of a patch that is up to version 10.
> Is there any way we can cancel jobs for previous versions of a patch? At
> this rate, my inbox is going to be spammed with meaningless CI failures for
> several
William Allen Simpson wrote on Tue, Nov 07, 2017 at 08:21:04AM -0500:
> Your idea seems to work, and I've posted the ntirpc side for DanG to
> review. Should have the Ganesha side for review soon thereafter.
Thanks!
--
Dominique
William Allen Simpson wrote on Mon, Nov 06, 2017 at 08:12:19AM -0500:
> The only threads that I know of are fridge and work_pool.
There's quite a few more, just looking at the ones that do set their name:
$ git grep SetNameFunction
src/FSAL/FSAL_GLUSTER/fsal_up.c:
LIENT blocks instead
Access_Type = RW;
# Exporting FSAL
FSAL {
Name = PROXY;
}
}
PROXY {
Remote_Server {
Srv_Addr = 127.0.0.1;
}
}
----
Frank Filz wrote on Wed, Sep 27, 2017 at 10:50:00AM -0700:
> One source of “fd leaks” is that the global fd, which is used for
> getattrs, is no longer well managed. I have some ideas in progress on
> how to better manage this, but discussion has stalled due to folks
> involvement in Bakeathon.
>
Daniel Gryniewicz wrote on Wed, Aug 02, 2017 at 10:16:39AM -0400:
> fortify sources does not work for me. It fails during cmake. Basic system
> libraries (such as -pthread) do not link.
Hm. There is a "problem", that FORTIFY_SOURCE requires optimizations, so
I had to add -O2... which in turn
Hi,
Noticed something weird testing Bill's patch, turns out it's in next
already..
I'm having some weird behavior with nfs (still using 4.1
explicitely, it does not seem to happen with 4.0)
Setup is:
pseudo /
vfs export for /export
I'm mounting the pseudo / on /mnt: mount -t nfs -o vers=4.1
Kaleb S. KEITHLEY wrote on Wed, Aug 02, 2017 at 09:06:17AM -0400:
> cd /builddir/build/BUILD/nfs-ganesha-2.5.1/src/Protocols/NFS &&
> /usr/bin/cc -I/usr/include/glusterfs -I/usr/include/uuid
> -I/builddir/build/BUILD/nfs-ganesha-2.5.1/src/include
>
Hi,
William Allen Simpson wrote on Wed, Aug 02, 2017 at 04:46:38AM -0400:
> I've been using -DCMAKE_C_FLAGS="-O0 -g -gdwarf-4"
I'm curious, do you notice the difference with -O0 and -gdwarf for
debugging?
(honest question, I've never felt bothered, I only have -g from Debug
build type - we
h commands within the
export, on a simple VFS export built with ASAN (-DSANITIZE_ADDRESS=ON)
If you can't hit it I will provide more info there as well but it would
surprise me as there isn't much specific at this point.
--
Dominique Martinet
---
Hi Bill,
William Allen Simpson wrote on Fri, Jul 28, 2017 at 10:11:41AM -0400:
> Setting up another test machine under Fedora 26, got this error that I've
> not seen before. There's another library that cmake isn't checking?
That's actually my fault/because of DEBUG_MDCACHE, I didn't find a
Hi folks,
Daniel Gryniewicz wrote on Mon, Apr 24, 2017 at 09:48:33AM -0400:
> After a discussion with Matt, here's my thoughts.
>
> 1. We want -dev versions of Ganesha to track Master versions of Ceph.
> This makes development much easier for us, and shouldn't cause problems,
> since -dev
Malahal Naineni wrote on Sat, Feb 25, 2017 at 03:33:17PM +0530:
> - All config is in blocks
> - Most blocks are unique with their tag names
> - exceptions: "export" and "client" blocks.
> - "export" is unique by "path" value
More like unique by pseudo path; path can be identical for various
Hi,
playing with clang once in a while as usual, there's a couple of neat
new warnings.
Here's one for CEPH:
/export/nfs-ganesha/src/FSAL/FSAL_CEPH/handle.c:1176:6: error: variable
'unix_mode' is used uninitialized whenever 'if' condition is false
[-Werror,-Wsometimes-uninitialized]
Frank Filz wrote on Thu, Jan 26, 2017 at 09:28:54AM -0800:
> Bill Simpson (WAS) has been the most actively working on RDMA. Dominique has
> also worked on RDMA but may not be doing so actively.
>
> MOST of the RDMA work has been in libntirpc so is pretty invisible to me...
>
> I don't think
Dominique Martinet wrote on Tue, Jan 17, 2017 at 11:14:04AM +0100:
> Here's the backtrace I have:
>
> ganesha.nfsd: /export/nfs-ganesha/src/SAL/state_misc.c:917:
> dec_state_owner_ref: Assertion `refcount > 0' failed.
>
> Thread 46 "ganesha.nfsd" received signal
Daniel Gryniewicz wrote on Mon, Jan 16, 2017 at 03:24:16PM -0500:
> > sorry for dumping that issue on you but I didn't really investigate and
> > my best guess was that it got something to do with mdcache.
> > Anyhow, it was not in relation with GPFS, I'm using VFS for my local
> > tests.
> > I
Hi,
Unfortunately build bot hasn't been running in almost a week and we
already landed a patch that regularily make me hit that failure - it's
not 100% because I think the test is usually too short to get that
thread to run? But it's easy enough to reproduce manually on my setup
Here's the
Matt Benjamin wrote on Tue, Dec 06, 2016 at 09:03:15PM -0500:
> When I attempted my github signin to gerrithub just now, I was sent to
> a screen requesting full access to my personal github info, and
> read/write access to all my public repositories, not just
> nfs-ganesha.
>
> I don't seem to
Let's fork the discussions to the list, I think more people will see it
there and it's a bit more comfortable to make argumented discussions...
First, my point on the subject is that we have two reasons to switch:
- our default secType is sys, and that really is as weak as it gets.
My personal
Frank Filz wrote on Tue, Aug 23, 2016 at 04:51:07PM -0700:
> I'd love to see some improvements here:
>
> 1. Maybe this has been addressed, but reduce spam, only place comments in
> Gerrithub for failures
Doesn't it do like our's (CEA) e.g. post comment anyway but only send
mail if failure ?
I
han going through the
kernel needlessly.
--
Dominique Martinet
--
___
Nfs-ganesha-devel mailing list
Nfs-ganesha-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/nfs-ganesha-devel
Frank Filz wrote on Tue, Jun 28, 2016 at 08:31:31AM -0700:
> Yes, if we are to keep FSAL_NULL, it needs to be stackable with FSAL_MDCACHE
> (and maybe the reason to keep it is to be able to continually test and
> maintain the ability for multiple layer stacking).
Well the LUSTRE/HSM FSAL will be
Hi,
Frank Filz wrote on Fri, May 27, 2016 at 05:05:00PM -0700:
> I do have a question for anyone who might know: What does the xattr support
> actually do? It looks like several FSALs support some kind of xattr, but
> they seem to just be derived from the normal attributes, so what purpose do
>
Hi,
got this for inline_xdr_string:
- I don't get why strlen is called on FREE, we don't need size to
actually be set for it and if it's a free after a decode that failed
(e.g. hit maxlen) then the pointer is not valid
- hence don't trust size for op FREE and early bailout. That's what the
William Allen Simpson wrote on Tue, Mar 01, 2016 at 08:52:42PM -0500:
> On 3/1/16 4:18 PM, Dominique Martinet wrote:
> >Anyway, I'm a bit at lost right now, will go see you again when I'm back
> >on this.
Looks like I was just hitting issues with the xdr_bitmap4 code, because
that'
Hi,
Synopsis (coverity and codenomicon) folks have been fuzzing us a bit and
found that we blindly trust the client and allocate huge buffers,
e.g. send us some small string but pretending it's 0x length.
I've found alot of these in include/nfsv41.h, what I've done so far is:
- every
William Allen Simpson wrote on Tue, Dec 08, 2015 at 09:40:50AM -0500:
> Thanks, Matt, I know you mean well, but I prefer not to have meta
> discussions on gerrit, where the comments are lost.
This is sound - philosophical discussions do belong on the list, even if
it's always better when it
Frank Filz wrote on Thu, Oct 29, 2015 at 12:00:31PM -0700:
> > In another situation the Linux the OOM killer might have already killed
> > important other processes trying to free memory for the NFS server. You
> > wouldn't want to recover the NFS process here, since you don't know if
> that
> >
Frank Filz wrote on Mon, Oct 26, 2015 at 06:51:59AM -0700:
> Hmm, Coverity constantly finding "new" bugs that have been there forever is
> getting exhausting. It's making me feel like a clean Coverity run is
> unattainable. Which severely diminishes the value of the tool.
Well, these are fairly
Hi,
SAUNIER, FREDERIC wrote on Mon, Oct 05, 2015 at 01:45:53PM +:
> I recently got compilation issues where build was no longer able to find out
> the locations of
> ibverbs/rdmacm/mooshika/glusterfs/ceph libraries, though the related
> _PREFIX
> were passed to cmake.
>
> I'm
Malahal Naineni wrote on Wed, Sep 02, 2015 at 03:39:13PM -0500:
> William Allen Simpson [william.allen.simp...@gmail.com] wrote:
> > src/scripts/commit-msg has a line that always adds a signed-off-by.
> >
> > That should never happen, unless I add the -s to the commit. You
> > can gripe about
Hi,
William Allen Simpson wrote on Fri, Aug 21, 2015 at 03:32:48AM -0400:
My comments for the libntirpc update still stand (
https://review.gerrithub.io/#/c/243818/ )
Matt didn't commit his merged patch into the github linuxbox2,
making it impossible for me to cherry-pick it directly onto
Hi,
Meghana Madhusudhan wrote on Tue, Aug 18, 2015 at 02:10:07AM -0400:
The sysconfig file defines Restart=on-abort and as a result, the NFS-Ganesha
service gets restarted automatically
when the service process exits due to an uncaught signal not specified as a
clean exit status.
In a
Hi,
Frank Filz wrote on Tue, Aug 11, 2015 at 01:31:02PM -0700:
In doing some scrubbing for licenses, I'm curious about the following files
and what they are used for?
No idea who uses these, but please remove as well the rpm/ dir (contains
old broken spec files), and share/scripts that seems
Hi,
Stijn De Weirdt wrote on Thu, Jun 25, 2015 at 08:46:09AM +0200:
i'm asking this question here since i assume that ganesha has some
(unit)testing tools to help stress test ganesha, and maybe they can be
used for our plans too.
You might want to look at pynfs, that can do a mount and NFS
William Allen Simpson wrote on Mon, Jun 22, 2015 at 03:26:38PM -0400:
Moreover, requiring library changes adds more variability to
the mix. The only patch that matters (to me) is the final patch.
I'd agree if we squash them after review, but since we don't, anyone
bisecting do care about all
Malahal Naineni wrote on Fri, Jun 19, 2015 at 05:18:59PM -0500:
William Allen Simpson [william.allen.simp...@gmail.com] wrote:
# I'm sorry, but I do not want commits that do not build -- I don't use
bisect often, but when I do I like it to work without relying on luck
finding commits that
Hi (and thanks for the information, I was surprised to see a message
from this mail address ;))
Matt W. Benjamin wrote on Tue, Apr 28, 2015 at 11:22:17AM -0400:
- GerritForge Support supp...@gerritforge.com wrote:
It actually exists and is called “topic”. It will be very soon
possible as
Hi all,
I've been working on automated testing for all of us today, I believe
more should spend some time on this.
You will notice ganesha-triggers adds Verified +1 sometimes now, this
comes from our jenkins build, it:
1/ will NOT run on all the pushes automatically. Currently, it will
trigger
49 matches
Mail list logo