The latest we can test on is 4.18 (there are "mainline" kernel-ml packages on elrepo for RHEL 7.5), and we have been able to reproduce on that kernel too. The user-space doesn't give an error when running "rm" on the stuck file, but it still shows up in directory listings, and "touch" on the file reports that the file does not exist. Happy to run more tests. -syam ________________________________________ From: ronnie sahlberg <ronniesahlberg@xxxxxxxxx> Sent: Wednesday, October 3, 2018 4:07:27 PM To: Syam Gadde Cc: Ronnie Sahlberg; linux-cifs Subject: Re: unclosed CIFS file handles In that case itt might be something that still persists in the upstream kernel. Can you try to reproduce on 4.19 ? On Thu, Oct 4, 2018 at 3:08 AM Syam Gadde <syam.gadde@xxxxxxxx> wrote: > > I can confirm that the downloaded source for the kernel I am running (3.10.0-693.2.2.el7) does have this patch applied. Unfortunately, the issue still persists. If there is any way I can help identify another suspect in the kernel I'm happy to run tests. > > -syam > > ________________________________________ > From: linux-cifs-owner@xxxxxxxxxxxxxxx <linux-cifs-owner@xxxxxxxxxxxxxxx> on behalf of Ronnie Sahlberg <lsahlber@xxxxxxxxxx> > Sent: Tuesday, October 2, 2018 5:13:57 PM > To: Syam Gadde > Cc: ronnie sahlberg; linux-cifs@xxxxxxxxxxxxxxx > Subject: Re: unclosed CIFS file handles > > It should be this commit : > > commit 38bd49064a1ecb67baad33598e3d824448ab11ec > Author: Sachin Prabhu <sprabhu@xxxxxxxxxx> > Date: Fri Mar 3 15:41:38 2017 -0800 > > Handle mismatched open calls > > A signal can interrupt a SendReceive call which result in incoming > responses to the call being ignored. This is a problem for calls such as > open which results in the successful response being ignored. This > results in an open file resource on the server. > > The patch looks into responses which were cancelled after being sent and > in case of successful open closes the open fids. > > For this patch, the check is only done in SendReceive2() > > RH-bz: 1403319 > > Signed-off-by: Sachin Prabhu <sprabhu@xxxxxxxxxx> > Reviewed-by: Pavel Shilovsky <pshilov@xxxxxxxxxxxxx> > Cc: Stable <stable@xxxxxxxxxxxxxxx> > > > ----- Original Message ----- > From: "Syam Gadde" <syam.gadde@xxxxxxxx> > To: "ronnie sahlberg" <ronniesahlberg@xxxxxxxxx> > Cc: linux-cifs@xxxxxxxxxxxxxxx > Sent: Tuesday, 2 October, 2018 5:05:33 AM > Subject: Re: unclosed CIFS file handles > > Can you provide a link to the upstream bug fix? Because I am still experiencing this after upgrading to 7.5, and someone else also has verified that it happens to them. > > -syam > > ________________________________________ > From: ronnie sahlberg <ronniesahlberg@xxxxxxxxx> > Sent: Monday, September 17, 2018 6:39:16 PM > To: Syam Gadde > Cc: linux-cifs@xxxxxxxxxxxxxxx > Subject: Re: unclosed CIFS file handles > > That bug is already in upstream and has been backported to rhel7.5. > > On Tue, Sep 18, 2018 at 6:39 AM, Syam Gadde <syam.gadde@xxxxxxxx> wrote: > > Hi all, > > > > Apologies if this is the wrong list, or if this has already been reported/fixed. I had trouble finding any related bug reports. > > > > I opened a bug at RedHat: > > > > https://urldefense.proofpoint.com/v2/url?u=https-3A__bugzilla.redhat.com_show-5Fbug.cgi-3Fid-3D1624029&d=DwIFaQ&c=imBPVzF25OnBgGmVOlcsiEgHoG1i6YHLR0Sj_gZ4adc&r=7KaCwpE161eTIbiGDtQmV05me-P1kWzL4it5juttbsU&m=84VBunsFfDNqdQ7H6abUQD7h-_kyuTyENo_kxZsmQWo&s=aV5D0rw2LY-gSMg8QmXVpbuEtA7pco_nQuzrT_7qo4g&e= > > > > and the summary is that on our Linux machines (running Scientific Linux kernel versions 3.10.0-693.5.2.el7 and 3.10.0-514.21.1.el7) that connect to a CIFS file system residing on a Hitachi/BlueArc file server, the kernel sometimes neglects to send a "close" request on open file handles if an application is interrupted (with Ctrl-C, and probably also with segfaults as we've seen this with non-interactive processes on our cluster). The result is phantom files (and parent directories) that can't be deleted/moved/renamed because the server thinks there is someone still keeping it open. > > > > More details are in the above bug report, including a packet dump and some dmesg output (with debugging output turned on as suggested on wiki.samba.org). Note that the above bug report has been assigned to cifs-utils but I don't think this is a user-space issue. I would appreciate any suggestions on how to address this. > > > > Thank you for your help (or redirection)! > > > > -syam > > > > > > > >
