linux-kernel - Re: Page cache corruption with mount/unmount of USB with Fat file system

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAMbOQaX97CXBWtKAs7YA2vtTL8jFP_saDoB_PRb6eLJCNiSOdg@mail.gmail.com>
Date:	Sat, 22 Mar 2014 01:06:47 +0530
From:	Nilesh More <nilesh99999@...il.com>
To:	"Theodore Ts'o" <tytso@....edu>, linux-kernel@...r.kernel.org,
	Prafull Suryawanshi <prafull.net@...il.com>
Subject: Re: Page cache corruption with mount/unmount of USB with Fat file system

Just to keep the thread updated with latest findngs : (Let me know if
anybody has any inputs)

Skipping the measureExactStorage and measureApproximateStorage calls
seems to help. These are called as a part of auto-mount sequence when
USB is hotplugged. I will dig-in further to understand if it could be
issue with these calls, may be these calls aren't supposed to be made
before USB mount. Or, is it just exposing file system bug?

/android/settings/deviceinfo/StorageMeasurement.java

public void handleMessage(Message msg) {

case MSG_CONNECTED: {

measureApproximateStorage(imcs);  // Skipping these calls seems to
help
                                                 measureExactStorage(imcs);
                                                 break;
                                                }

On Fri, Mar 21, 2014 at 12:09 AM, Nilesh More <nilesh99999@...il.com> wrote:
> I now know the source of page allocations between add_disk call and
> mount call. The auto mount service triggers disk check operation which
> eventually calls checkfilesys(). This function does open and close
> calls which results into page allocations and de-allocations. If I do
> an early return from this function, I don't this issue of ext4 file
> system corruption.
>
>  Though the file check operation might be expected before mount, it
> may be exposing a linux-kernel bug.
>
> Can anybody please comment if the checkfilesys() operation is expected
> before actually mounting the file system ? If 'yes', how does it know
> about the superblock location, directory structure etc. ?
>
> What should be the next step here ? I am thinking to try and track
> down all page/buffer allocations because of open/close calls from auto
> mount service - checkfilesys() and see why it is resulting into ext4
> corruption.
>
> Thank you,
> Nilesh
>
> On Thu, Mar 20, 2014 at 12:38 AM, Nilesh More <nilesh99999@...il.com> wrote:
>> Hi Ted,
>>
>>>And I'll repeat the comments I made a few weeks ago.  Last time you
>>>reported this, you included a dmesg output which included this:
>>
>>> [ 413.607849] usb 2-1.1: USB disconnect, device number 12
>>> [ 414.022630] EXT4-fs error (device mmcblk0p20): ext4_readdir:227: inode
>>> #81827: block 328308: comm installd...
>>
>>  > Are you seeing the same thing?  If so, the fundamental high level
>> issue is that the USB driver is reporting a device disconnect,
>> presumably of a USB device different from the one that you just
>> plugged in.   Are you still seeing this?
>>
>> The dmesg output that I attached earlier was when I saw the prints
>> after actually disconnecting the USB. I am pretty certain that USB
>> driver does not report any false disconnect. This issue occurs when I
>> connect the USBdrive on android tablet with USB mount support. This
>> issue may or may not occur the first time I hotplug the USB drive. It
>> might require couple of hotplug/hotunplug to see this issue.
>>
>> Thank you,
>> Nilesh
>>
>> On Thu, Mar 20, 2014 at 12:31 AM,  <tytso@....edu> wrote:
>>> On Thu, Mar 20, 2014 at 12:05:44AM +0530, Nilesh More wrote:
>>>> Hi all,
>>>>
>>>> First of all sorry for the lengthy mail but I did not want to miss out
>>>> on any details.
>>>>
>>>> The following issue that I am reporting occurs when I plugin the USB
>>>> drive on android tablet with USB mount support. Please note that this
>>>> issue may or may not occur the first time I hotplug the USB drive. It
>>>> might require couple of hotplug/hotunplug to see this issue. If
>>>> anybody has/gets any clue about the possible root cause, please let me
>>>> know.
>>>
>>> And I'll repeat the comments I made a few weeks ago.  Last time you
>>> reported this, you included a dmesg output which included this:
>>>
>>>> [ 413.607849] usb 2-1.1: USB disconnect, device number 12
>>>> [ 414.022630] EXT4-fs error (device mmcblk0p20): ext4_readdir:227: inode
>>>> #81827: block 328308: comm installd...
>>>
>>> Are you seeing the same thing?  If so, the fundamental high level
>>> issue is that the USB driver is reporting a device disconnect,
>>> presumably of a USB device different from the one that you just
>>> plugged in.   Are you still seeing this?
>>>
>>> If so, it's important if you want to root cause this issue to
>>> determine why the USB disconnect is happening, and not try to comment
>>> out the page invalidations which happens *after* the USB device is
>>> reported as disconnected.
>>>
>>> If I had to guess, it's probably some hardware issue where when you
>>> plug in a USB device that draws too much power, it's destablizing the
>>> USB bus and causing some other USB device to report a disconnect.
>>>
>>>                                           - Ted
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/