kernel-packages team mailing list archive

[Adam Porter <1323624@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

If I have a CIFS mount, and then that mount's server goes down, other
file operations on my client system block until the Samba server comes
back up.  This causes extremely weird behavior, things like right-click
menus not disappearing, mouse clicks in X not working at all while
keyboard input does work, plasma-desktop fails to finish loading its
configuration, resulting in half-empty panels and no wallpaper, etc.

I will try to illustrate:

$ grep cifs /etc/fstab
//server/Downloads /mnt/downloads       cifs    user,guest,noauto

$ mount /mnt/downloads

$ dolphin /mnt/downloads

$ ssh root@server "service smbd stop"

$ lsof | grep downloads
lsof: WARNING: can't stat() cifs file system /mnt/downloads
      Output information may be incomplete. 
[lsof blocks here]

At this point, the weird bugs begin happening: Dolphin blocks when I
right-click in its window.  This leaves a blocked right-click menu
(which is sometimes invisible, sometimes visible but empty, and
sometimes a solid gray rectangle), and this blocked right-click menu
grabs mouse clicks in X, preventing me from clicking anything else in X,
while the keyboard continues working.  At this point I kill Dolphin from
a terminal using the keyboard, but plasma-desktop is still not
responding to mouse clicks.  So I kill and restart plasma-desktop from a
terminal, but when it loads, it doesn't load completely, leaving a half-
empty panel with missing applets, and no wallpaper--and still no
response to mouse clicks.

But it's not all about X apps.  Even /bin/df blocks.  At one point, I
wondered if my system had somehow run out of disk space, which can cause
weird behavior, so I ran df.  But df just sat there, with no output
whatsoever.

And after all this, when I do:

$ ssh root@server "service smbd start"

Everything suddenly unblocks and works fine again.

It seems to me that having kernel CIFS mounts cause blocking in
unrelated I/O is a major bug, potentially a security bug due to the
denial-of-service issue.  The way it causes an apparently inexplicable
chain of weird behavior in all sorts of programs makes it very difficult
to debug.

CIFS mounts are LAN mounts.  It seems like a simple solution would be to
have a very short timeout, 1-2 seconds or so.  If it times out, just
return an error and let the apps blocking on that I/O get on with what
they're doing.  The current behavior means that using kernel CIFS mounts
is a recipe for disaster.

ProblemType: Bug
DistroRelease: Ubuntu 14.04
Package: linux-image-generic 3.13.0.24.29
ProcVersionSignature: Ubuntu 3.13.0-24.47-generic 3.13.9
Uname: Linux 3.13.0-24-generic x86_64
NonfreeKernelModules: fglrx
ApportVersion: 2.14.1-0ubuntu3.2
Architecture: amd64
AudioDevicesInUse:
 USER        PID ACCESS COMMAND
 /dev/snd/controlC1:  me        11691 F.... pulseaudio
 /dev/snd/controlC0:  me        11691 F.... pulseaudio
CurrentDesktop: KDE
Date: Tue May 27 06:44:22 2014
InstallationDate: Installed on 2010-05-20 (1467 days ago)
InstallationMedia: Kubuntu 10.04 LTS "Lucid Lynx" - Release amd64 (20100427)
IwConfig:
 eth0      no wireless extensions.
 
 lo        no wireless extensions.
MachineType: System manufacturer System Product Name
ProcFB: 0 VESA VGA
ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-3.13.0-24-generic root=UUID=e3d95eec-f6ce-426f-b110-6633ff6f62c3 ro quiet splash vt.handoff=7
RelatedPackageVersions:
 linux-restricted-modules-3.13.0-24-generic N/A
 linux-backports-modules-3.13.0-24-generic  N/A
 linux-firmware                             1.127.2
RfKill:
 
SourcePackage: linux
UpgradeStatus: Upgraded to trusty on 2014-05-07 (20 days ago)
dmi.bios.date: 08/04/2009
dmi.bios.vendor: American Megatrends Inc.
dmi.bios.version: 0602
dmi.board.asset.tag: To Be Filled By O.E.M.
dmi.board.name: P5Q PRO TURBO
dmi.board.vendor: ASUSTeK Computer INC.
dmi.board.version: Rev 1.xx
dmi.chassis.asset.tag: Asset-1234567890
dmi.chassis.type: 3
dmi.chassis.vendor: Chassis Manufacture
dmi.chassis.version: Chassis Version
dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvr0602:bd08/04/2009:svnSystemmanufacturer:pnSystemProductName:pvrSystemVersion:rvnASUSTeKComputerINC.:rnP5QPROTURBO:rvrRev1.xx:cvnChassisManufacture:ct3:cvrChassisVersion:
dmi.product.name: System Product Name
dmi.product.version: System Version
dmi.sys.vendor: System manufacturer

** Affects: linux (Ubuntu)
     Importance: Undecided
         Status: New


** Tags: amd64 apport-bug trusty

** Summary changed:

- Unresponsive CIFS mounts blocks unrelated I/O
+ Unresponsive CIFS mounts block unrelated I/O

-- 
You received this bug notification because you are a member of Kernel
Packages, which is subscribed to linux in Ubuntu.
https://bugs.launchpad.net/bugs/1323624

Title:
  Unresponsive CIFS mounts block unrelated I/O

Status in “linux” package in Ubuntu:
  New

Bug description:
  If I have a CIFS mount, and then that mount's server goes down, other
  file operations on my client system block until the Samba server comes
  back up.  This causes extremely weird behavior, things like right-
  click menus not disappearing, mouse clicks in X not working at all
  while keyboard input does work, plasma-desktop fails to finish loading
  its configuration, resulting in half-empty panels and no wallpaper,
  etc.

  I will try to illustrate:

  $ grep cifs /etc/fstab
  //server/Downloads /mnt/downloads       cifs    user,guest,noauto

  $ mount /mnt/downloads

  $ dolphin /mnt/downloads

  $ ssh root@server "service smbd stop"

  $ lsof | grep downloads
  lsof: WARNING: can't stat() cifs file system /mnt/downloads
        Output information may be incomplete. 
  [lsof blocks here]

  At this point, the weird bugs begin happening: Dolphin blocks when I
  right-click in its window.  This leaves a blocked right-click menu
  (which is sometimes invisible, sometimes visible but empty, and
  sometimes a solid gray rectangle), and this blocked right-click menu
  grabs mouse clicks in X, preventing me from clicking anything else in
  X, while the keyboard continues working.  At this point I kill Dolphin
  from a terminal using the keyboard, but plasma-desktop is still not
  responding to mouse clicks.  So I kill and restart plasma-desktop from
  a terminal, but when it loads, it doesn't load completely, leaving a
  half-empty panel with missing applets, and no wallpaper--and still no
  response to mouse clicks.

  But it's not all about X apps.  Even /bin/df blocks.  At one point, I
  wondered if my system had somehow run out of disk space, which can
  cause weird behavior, so I ran df.  But df just sat there, with no
  output whatsoever.

  And after all this, when I do:

  $ ssh root@server "service smbd start"

  Everything suddenly unblocks and works fine again.

  It seems to me that having kernel CIFS mounts cause blocking in
  unrelated I/O is a major bug, potentially a security bug due to the
  denial-of-service issue.  The way it causes an apparently inexplicable
  chain of weird behavior in all sorts of programs makes it very
  difficult to debug.

  CIFS mounts are LAN mounts.  It seems like a simple solution would be
  to have a very short timeout, 1-2 seconds or so.  If it times out,
  just return an error and let the apps blocking on that I/O get on with
  what they're doing.  The current behavior means that using kernel CIFS
  mounts is a recipe for disaster.

  ProblemType: Bug
  DistroRelease: Ubuntu 14.04
  Package: linux-image-generic 3.13.0.24.29
  ProcVersionSignature: Ubuntu 3.13.0-24.47-generic 3.13.9
  Uname: Linux 3.13.0-24-generic x86_64
  NonfreeKernelModules: fglrx
  ApportVersion: 2.14.1-0ubuntu3.2
  Architecture: amd64
  AudioDevicesInUse:
   USER        PID ACCESS COMMAND
   /dev/snd/controlC1:  me        11691 F.... pulseaudio
   /dev/snd/controlC0:  me        11691 F.... pulseaudio
  CurrentDesktop: KDE
  Date: Tue May 27 06:44:22 2014
  InstallationDate: Installed on 2010-05-20 (1467 days ago)
  InstallationMedia: Kubuntu 10.04 LTS "Lucid Lynx" - Release amd64 (20100427)
  IwConfig:
   eth0      no wireless extensions.
   
   lo        no wireless extensions.
  MachineType: System manufacturer System Product Name
  ProcFB: 0 VESA VGA
  ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-3.13.0-24-generic root=UUID=e3d95eec-f6ce-426f-b110-6633ff6f62c3 ro quiet splash vt.handoff=7
  RelatedPackageVersions:
   linux-restricted-modules-3.13.0-24-generic N/A
   linux-backports-modules-3.13.0-24-generic  N/A
   linux-firmware                             1.127.2
  RfKill:
   
  SourcePackage: linux
  UpgradeStatus: Upgraded to trusty on 2014-05-07 (20 days ago)
  dmi.bios.date: 08/04/2009
  dmi.bios.vendor: American Megatrends Inc.
  dmi.bios.version: 0602
  dmi.board.asset.tag: To Be Filled By O.E.M.
  dmi.board.name: P5Q PRO TURBO
  dmi.board.vendor: ASUSTeK Computer INC.
  dmi.board.version: Rev 1.xx
  dmi.chassis.asset.tag: Asset-1234567890
  dmi.chassis.type: 3
  dmi.chassis.vendor: Chassis Manufacture
  dmi.chassis.version: Chassis Version
  dmi.modalias: dmi:bvnAmericanMegatrendsInc.:bvr0602:bd08/04/2009:svnSystemmanufacturer:pnSystemProductName:pvrSystemVersion:rvnASUSTeKComputerINC.:rnP5QPROTURBO:rvrRev1.xx:cvnChassisManufacture:ct3:cvrChassisVersion:
  dmi.product.name: System Product Name
  dmi.product.version: System Version
  dmi.sys.vendor: System manufacturer

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1323624/+subscriptions