[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

k.kotlenga@sims.pl changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hch@lst.de,
                   |                            |linux-ext4@vger.kernel.org

--- Comment #6 from k.kotlenga@sims.pl ---
This bug is still present in 4.2-rc6. Reverting:

commit 08439fec266c3cc5702953b4f54bdf5649357de0
Author: Christoph Hellwig <hch@lst.de>
Date:   Thu Apr 2 23:56:32 2015 -0400

    ext4: remove block_device_ejected

    bdi->dev now never goes away, so this function became useless.

    Signed-off-by: Christoph Hellwig <hch@lst.de>
    Signed-off-by: Theodore Ts'o <tytso@mit.edu>

makes the oops go away, at least for me (I've tested this on v4.1.4 i386 only).
I hope this rings some bells.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

Maciej Szmigiero <mail@maciej.szmigiero.name> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mail@maciej.szmigiero.name

--- Comment #7 from Maciej Szmigiero <mail@maciej.szmigiero.name> ---
I can also confirm that this bug is present in latest stable kernel (4.1.5) and
reverting commit from comment 6 seems to fix it.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

--- Comment #8 from Theodore Tso <tytso@mit.edu> ---
On Fri, Aug 14, 2015 at 11:02:14AM +0000, bugzilla-daemon@bugzilla.kernel.org
wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=101011
> 
> I can also confirm that this bug is present in latest stable kernel (4.1.5) and
> reverting commit from comment 6 seems to fix it.

Christoph,

I've since gotten two reports from users that reverting your commit:
"08439fec266c3: ext4: remove block_device_ejected" fixes a crash when
a USB stick is yanked from their system.  Looking at the reported
stack dump, it looks like the crash is happening in
account_page_dirtied() when it updates some bdi-specific statistics.

I haven't been paying attention to the recent changes in how bdi gets
torn down after the device gets removed, and in fact finding the
recent changes wasn't obvioius enough after doing a brief search, but
it seems to me that if reverting this patch is making any kind of
differences, then the assertion in the commit description:

    bdi->dev now never goes away, so this function became useless.

it implies that bdi->dev *does* become NULL, and checking for this is
useful.  In any case, I don't see any harm in reverting this commit;
what do you think?

Thanks,

                    - Ted

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

--- Comment #9 from Christoph Hellwig <hch@lst.de> ---
Hi Ted,


sorry for the delay - I saw the mail Jan Cc'ed me on yesterday.  After
my changes it should not go away and I had tested the original eject
test that it indeed didn't.  Either I forgot a case, or the major
writeback Tejun did a little later regressed it.

As I won't have time to look into it ASAP I'd suggest to revert my
patch for now.  In the long run I really don't want to have these
checks spread over file system so I plan to look into it once I
get a few spare hours.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

--- Comment #10 from Theodore Tso <tytso@mit.edu> ---
On Sat, Aug 15, 2015 at 10:19:02AM +0200, Christoph Hellwig wrote:
> 
> sorry for the delay - I saw the mail Jan Cc'ed me on yesterday.  After
> my changes it should not go away and I had tested the original eject
> test that it indeed didn't.  Either I forgot a case, or the major
> writeback Tejun did a little later regressed it.
> 
> As I won't have time to look into it ASAP I'd suggest to revert my
> patch for now.  In the long run I really don't want to have these
> checks spread over file system so I plan to look into it once I
> get a few spare hours.

Thanks, I'll revert the patch.

I suspect we should add an ioctl to simulate a USB device unplug using
the loopback block device, so we can add a test to xfstests.

                                            - Ted

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

Ronny Standtke <ronny.standtke@gmx.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ronny.standtke@gmx.net

--- Comment #11 from Ronny Standtke <ronny.standtke@gmx.net> ---
This bug is already hitting distributions, e.g. here:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1478623

I hope this helps to get the right people into contact and sorry if I just
added to the noise...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

Theodore Tso <tytso@mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |tytso@mit.edu

--- Comment #12 from Theodore Tso <tytso@mit.edu> ---
The revert is upstream as commit bdfe0cbd746aa9 and will be showing up stable
backport kernels very shortly (the deadline for commenting on the stable
backports is tonight, so the next stable backport kernels should be released
within a day or so with the revert).

It's up to the distribution to update their kernels, and they don't have to
wait for the stable kernel backports to be released.  So please feel free to
request at each of the various distro-specific bug trackers that they include
the revert found in upstream commit bdfe0cbd746aa9.  Especially if you are
receiving paid support from the distribution, they are much more likely to
listen to you.  :-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

Theodore Tso <tytso@mit.edu> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |CODE_FIX

--- Comment #13 from Theodore Tso <tytso@mit.edu> ---
Resolving since this is fixed upstream....

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=101011

--- Comment #14 from Ronny Standtke <ronny.standtke@gmx.net> ---
> The revert is upstream as commit bdfe0cbd746aa9

Thanks for the info! I will apply the revert to 4.2.1.

> Especially if you are receiving paid support from the distribution, they are 
> much more likely to listen to you.  :-)

It's the other way around. I am maintaining a distribution and get paid for
supporting it. :-)

-- 
You are receiving this mail because:
You are on the CC list for the bug.

Re: [Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[Theodore Ts'o]

On Fri, Aug 14, 2015 at 11:02:14AM +0000, bugzilla-daemon@bugzilla.kernel.org wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=101011
> 
> I can also confirm that this bug is present in latest stable kernel (4.1.5) and
> reverting commit from comment 6 seems to fix it.

Christoph,

I've since gotten two reports from users that reverting your commit:
"08439fec266c3: ext4: remove block_device_ejected" fixes a crash when
a USB stick is yanked from their system.  Looking at the reported
stack dump, it looks like the crash is happening in
account_page_dirtied() when it updates some bdi-specific statistics.

I haven't been paying attention to the recent changes in how bdi gets
torn down after the device gets removed, and in fact finding the
recent changes wasn't obvioius enough after doing a brief search, but
it seems to me that if reverting this patch is making any kind of
differences, then the assertion in the commit description:

    bdi->dev now never goes away, so this function became useless.

it implies that bdi->dev *does* become NULL, and checking for this is
useful.  In any case, I don't see any harm in reverting this commit;
what do you think?

Thanks,

					- Ted

Re: [Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[Christoph Hellwig]

Hi Ted,


sorry for the delay - I saw the mail Jan Cc'ed me on yesterday.  After
my changes it should not go away and I had tested the original eject
test that it indeed didn't.  Either I forgot a case, or the major
writeback Tejun did a little later regressed it.

As I won't have time to look into it ASAP I'd suggest to revert my
patch for now.  In the long run I really don't want to have these
checks spread over file system so I plan to look into it once I
get a few spare hours.

Re: [Bug 101011] Kernel Oops when disconnecting a mounted ext4 usb stick

[Theodore Ts'o]

On Sat, Aug 15, 2015 at 10:19:02AM +0200, Christoph Hellwig wrote:
> 
> sorry for the delay - I saw the mail Jan Cc'ed me on yesterday.  After
> my changes it should not go away and I had tested the original eject
> test that it indeed didn't.  Either I forgot a case, or the major
> writeback Tejun did a little later regressed it.
> 
> As I won't have time to look into it ASAP I'd suggest to revert my
> patch for now.  In the long run I really don't want to have these
> checks spread over file system so I plan to look into it once I
> get a few spare hours.

Thanks, I'll revert the patch.

I suspect we should add an ioctl to simulate a USB device unplug using
the loopback block device, so we can add a test to xfstests.

    	     	   	      	     	 - Ted